[ { "path": ".gitignore", "content": ".*.swp\n*.o\n*.log\ndump.rdb\nredis-benchmark\nredis-check-aof\nredis-check-dump\nredis-cli\nredis-sentinel\nredis-server\ndoc-tools\nrelease\nmisc/*\nsrc/release.h\nappendonly.aof\nSHORT_TERM_TODO\nrelease.h\nsrc/transfer.sh\nsrc/configs\nredis.ds\nsrc/redis.conf\ndeps/lua/src/lua\ndeps/lua/src/luac\ndeps/lua/src/liblua.a\n.make-*\n.prerequisites\n*.dSYM\n" }, { "path": "00-RELEASENOTES", "content": "Hello! This file is just a placeholder, since this is the \"unstable\" branch\nof Redis, the place where all the development happens.\n\nThere is no release notes for this branch, it gets forked into another branch\nevery time there is a partial feature freeze in order to eventually create\na new stable release.\n\nUsually \"unstable\" is stable enough for you to use it in development enviromnets\nhowever you should never use it in production environments. It is possible\nto download the latest stable release here:\n\n http://download.redis.io/releases/redis-stable.tar.gz\n\nMore information is available at http://redis.io\n\nHappy hacking!\n" }, { "path": "BUGS", "content": "Plese check https://github.com/antirez/redis/issues\n" }, { "path": "CONTRIBUTING", "content": "Note: by contributing code to the Redis project in any form, including sending\na pull request via Github, a code fragment or patch via private email or\npublic discussion groups, you agree to release your code under the terms\nof the BSD license that you can find in the COPYING file included in the Redis\nsource distribution. You will include BSD license in the COPYING file within\neach source file that you contribute.\n\n# IMPORTANT: HOW TO USE REDIS GITHUB ISSUES\n\n* Github issues SHOULD ONLY BE USED to report bugs, and for DETAILED feature\n requests. Everything else belongs to the Redis Google Group.\n\n PLEASE DO NOT POST GENERAL QUESTIONS that are not about bugs or suspected\n bugs in the Github issues system. We'll be very happy to help you and provide\n all the support in the Redis Google Group.\n\n Redis Google Group address:\n \n https://groups.google.com/forum/?fromgroups#!forum/redis-db\n\n# How to provide a patch for a new feature\n\n1. Drop a message to the Redis Google Group with a proposal of semantics/API.\n\n2. If in step 1 you get an acknowledge from the project leaders, use the\n following procedure to submit a patch:\n\n a. Fork Redis on github ( http://help.github.com/fork-a-repo/ )\n b. Create a topic branch (git checkout -b my_branch)\n c. Push to your branch (git push origin my_branch)\n d. Initiate a pull request on github ( http://help.github.com/send-pull-requests/ )\n e. Done :)\n\nThanks!\n" }, { "path": "COPYING", "content": "Copyright (c) 2006-2014, Salvatore Sanfilippo\nAll rights reserved.\n\nRedistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:\n\n * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.\n * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.\n * Neither the name of Redis nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.\n\nTHIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n" }, { "path": "INSTALL", "content": "See README\n" }, { "path": "MANIFESTO", "content": "[Note: this is the Redis manifesto, for general information about\n installing and running Redis read the README file instead.]\n\nRedis Manifesto\n===============\n\n1 - A DSL for Abstract Data Types. Redis is a DSL (Domain Specific Language)\n that manipulates abstract data types and implemented as a TCP daemon.\n Commands manipulate a key space where keys are binary-safe strings and\n values are different kinds of abstract data types. Every data type\n represents an abstract version of a fundamental data structure. For instance\n Redis Lists are an abstract representation of linked lists. In Redis, the\n essence of a data type isn't just the kind of operations that the data types\n support, but also the space and time complexity of the data type and the\n operations performed upon it.\n\n2 - Memory storage is #1. The Redis data set, composed of defined key-value\n pairs, is primarily stored in the computer's memory. The amount of memory in\n all kinds of computers, including entry-level servers, is increasing\n significantly each year. Memory is fast, and allows Redis to have very\n predictable performance. Datasets composed of 10k or 40 millions keys will\n perform similarly. Complex data types like Redis Sorted Sets are easy to\n implement and manipulate in memory with good performance, making Redis very\n simple. Redis will continue to explore alternative options (where data can\n be optionally stored on disk, say) but the main goal of the project remains\n the development of an in-memory database.\n\n3 - Fundamental data structures for a fundamental API. The Redis API is a direct\n consequence of fundamental data structures. APIs can often be arbitrary but\n not an API that resembles the nature of fundamental data structures. If we\n ever meet intelligent life forms from another part of the universe, they'll\n likely know, understand and recognize the same basic data structures we have\n in our computer science books. Redis will avoid intermediate layers in API,\n so that the complexity is obvious and more complex operations can be\n performed as the sum of the basic operations.\n\n4 - Code is like a poem; it's not just something we write to reach some\n practical result. Sometimes people that are far from the Redis philosophy\n suggest using other code written by other authors (frequently in other\n languages) in order to implement something Redis currently lacks. But to us\n this is like if Shakespeare decided to end Enrico IV using the Paradiso from\n the Divina Commedia. Is using any external code a bad idea? Not at all. Like\n in \"One Thousand and One Nights\" smaller self contained stories are embedded\n in a bigger story, we'll be happy to use beautiful self contained libraries\n when needed. At the same time, when writing the Redis story we're trying to\n write smaller stories that will fit in to other code.\n\n5 - We're against complexity. We believe designing systems is a fight against\n complexity. We'll accept to fight the complexity when it's worthwhile but\n we'll try hard to recognize when a small feature is not worth 1000s of lines\n of code. Most of the time the best way to fight complexity is by not\n creating it at all.\n\n6 - Two levels of API. The Redis API has two levels: 1) a subset of the API fits\n naturally into a distributed version of Redis and 2) a more complex API that\n supports multi-key operations. Both are useful if used judiciously but\n there's no way to make the more complex multi-keys API distributed in an\n opaque way without violating our other principles. We don't want to provide\n the illusion of something that will work magically when actually it can't in\n all cases. Instead we'll provide commands to quickly migrate keys from one\n instance to another to perform multi-key operations and expose the tradeoffs\n to the user.\n\n7 - We optimize for joy. We believe writing code is a lot of hard work, and the\n only way it can be worth is by enjoying it. When there is no longer joy in\n writing code, the best thing to do is stop. To prevent this, we'll avoid\n taking paths that will make Redis less of a joy to develop.\n" }, { "path": "Makefile", "content": "# Top level makefile, the real shit is at src/Makefile\n\ndefault: all\n\n.DEFAULT:\n\tcd src && $(MAKE) $@\n\ninstall:\n\tcd src && $(MAKE) $@\n\n.PHONY: install\n" }, { "path": "README", "content": "Where to find complete Redis documentation?\n-------------------------------------------\n\nThis README is just a fast \"quick start\" document. You can find more detailed\ndocumentation at http://redis.io\n\nBuilding Redis\n--------------\n\nRedis can be compiled and used on Linux, OSX, OpenBSD, NetBSD, FreeBSD.\nWe support big endian and little endian architectures.\n\nIt may compile on Solaris derived systems (for instance SmartOS) but our\nsupport for this platform is \"best effort\" and Redis is not guaranteed to\nwork as well as in Linux, OSX, and *BSD there.\n\nIt is as simple as:\n\n % make\n\nYou can run a 32 bit Redis binary using:\n\n % make 32bit\n\nAfter building Redis is a good idea to test it, using:\n\n % make test\n\nFixing problems building 32 bit binaries\n---------\n\nIf after building Redis with a 32 bit target you need to rebuild it\nwith a 64 bit target, or the other way around, you need to perform a\n\"make distclean\" in the root directory of the Redis distribution.\n\nIn case of build errors when trying to build a 32 bit binary of Redis, try\nthe following steps:\n\n* Install the packages libc6-dev-i386 (also try g++-multilib).\n* Try using the following command line instead of \"make 32bit\":\n\n make CFLAGS=\"-m32 -march=native\" LDFLAGS=\"-m32\"\n\nAllocator\n---------\n\nSelecting a non-default memory allocator when building Redis is done by setting\nthe `MALLOC` environment variable. Redis is compiled and linked against libc\nmalloc by default, with the exception of jemalloc being the default on Linux\nsystems. This default was picked because jemalloc has proven to have fewer\nfragmentation problems than libc malloc.\n\nTo force compiling against libc malloc, use:\n\n % make MALLOC=libc\n\nTo compile against jemalloc on Mac OS X systems, use:\n\n % make MALLOC=jemalloc\n\nVerbose build\n-------------\n\nRedis will build with a user friendly colorized output by default.\nIf you want to see a more verbose output use the following:\n\n % make V=1\n\nRunning Redis\n-------------\n\nTo run Redis with the default configuration just type:\n\n % cd src\n % ./redis-server\n \nIf you want to provide your redis.conf, you have to run it using an additional\nparameter (the path of the configuration file):\n\n % cd src\n % ./redis-server /path/to/redis.conf\n\nIt is possible to alter the Redis configuration passing parameters directly\nas options using the command line. Examples:\n\n % ./redis-server --port 9999 --slaveof 127.0.0.1 6379\n % ./redis-server /etc/redis/6379.conf --loglevel debug\n\nAll the options in redis.conf are also supported as options using the command\nline, with exactly the same name.\n\nPlaying with Redis\n------------------\n\nYou can use redis-cli to play with Redis. Start a redis-server instance,\nthen in another terminal try the following:\n\n % cd src\n % ./redis-cli\n redis> ping\n PONG\n redis> set foo bar\n OK\n redis> get foo\n \"bar\"\n redis> incr mycounter\n (integer) 1\n redis> incr mycounter\n (integer) 2\n redis> \n\nYou can find the list of all the available commands here:\n\n http://redis.io/commands\n\nInstalling Redis\n-----------------\n\nIn order to install Redis binaries into /usr/local/bin just use:\n\n % make install\n\nYou can use \"make PREFIX=/some/other/directory install\" if you wish to use a\ndifferent destination.\n\nMake install will just install binaries in your system, but will not configure\ninit scripts and configuration files in the appropriate place. This is not\nneeded if you want just to play a bit with Redis, but if you are installing\nit the proper way for a production system, we have a script doing this\nfor Ubuntu and Debian systems:\n\n % cd utils\n % ./install_server.sh\n\nThe script will ask you a few questions and will setup everything you need\nto run Redis properly as a background daemon that will start again on\nsystem reboots.\n\nYou'll be able to stop and start Redis using the script named\n/etc/init.d/redis_, for instance /etc/init.d/redis_6379.\n\nCode contributions\n---\n\nNote: by contributing code to the Redis project in any form, including sending\na pull request via Github, a code fragment or patch via private email or\npublic discussion groups, you agree to release your code under the terms\nof the BSD license that you can find in the COPYING file included in the Redis\nsource distribution.\n\nPlease see the CONTRIBUTING file in this source distribution for more\ninformation.\n\nEnjoy!\n" }, { "path": "README.rst", "content": "Redis 3.0 源码注释\n============================\n\n本项目是注释版的 Redis 3.0 源码，\n原始代码来自： https://github.com/antirez/redis 。\n\n这份注释是我在创作新版《Redis 设计与实现》期间，\n为了了解 Redis 的内部实现而制作的，\n所有在书中有介绍的内容，\n在源码中都进行了相应的注释。\n\n在注释的过程中，\n除了少量空格和空行方面的调整外，\n没有对原始代码进行任何其他改动，\n最大程度地保证了代码的“原汁原味”。\n\n希望这份注释源码能给大家学习和了解 Redis 带来一点帮助。\n\n另外，\n`新版《Redis 设计与实现》 `_\\ 正在各大网店发售中，\n希望大家可以多多支持这本书。\n\nHave fun!\n\n| 黄健宏（huangz）\n| 2014 年 6 月 28 日\n\n\n附录：各个源码文件的作用简介\n------------------------------\n\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| 文件 | 作用 |\n+===================================================================+===================================================================+\n| ``adlist.c`` 、 ``adlist.h`` | 双端链表数据结构的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``ae.c`` 、 ``ae.h`` 、 ``ae_epoll.c`` 、 ``ae_evport.c`` 、 | 事件处理器，以及各个具体实现。 |\n| ``ae_kqueue.c`` 、 ``ae_select.c`` | |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``anet.c`` 、 ``anet.h`` | Redis 的异步网络框架，内容主要为对 socket 库的包装。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``aof.c`` | AOF 功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``asciilogo.h`` | 保存了 Redis 的 ASCII LOGO 。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``bio.c`` 、 ``bio.h`` | Redis 的后台 I/O 程序，用于将 I/O 操作放到子线程里面执行， |\n| | 减少 I/O 操作对主线程的阻塞。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``bitops.c`` | 二进制位操作命令的实现文件。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``blocked.c`` | 用于实现 BLPOP 命令和 WAIT 命令的阻塞效果。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``cluster.c`` 、 ``cluster.h`` | Redis 的集群实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``config.c`` 、 ``config.h`` | Redis 的配置管理实现，负责读取并分析配置文件， |\n| | 然后根据这些配置修改 Redis 服务器的各个选项。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``crc16.c`` 、 ``crc64.c`` 、 ``crc64.h`` | 计算 CRC 校验和。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``db.c`` | 数据库实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``debug.c`` | 调试实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``dict.c`` 、 ``dict.h`` | 字典数据结构的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``endianconv.c`` 、 ``endianconv.h`` | 二进制的大端、小端转换函数。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``fmacros.h`` | 一些移植性方面的宏。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``help.h`` | ``utils/generate-command-help.rb`` 程序自动生成的命令帮助信息。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``hyperloglog.c`` | HyperLogLog 数据结构的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``intset.c`` 、 ``intset.h`` | 整数集合数据结构的实现，用于优化 SET 类型。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``lzf_c.c`` 、 ``lzf_d.c`` 、 ``lzf.h`` 、 ``lzfP.h`` | Redis 对字符串和 RDB 文件进行压缩时使用的 LZF 压缩算法的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``Makefile`` 、 ``Makefile.dep`` | 构建文件。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``memtest.c`` | 内存测试。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``mkreleasehdr.sh`` | 用于生成释出信息的脚本。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``multi.c`` | Redis 的事务实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``networking.c`` | Redis 的客户端网络操作库， |\n| | 用于实现命令请求接收、发送命令回复等工作， |\n| | 文件中的函数大多为 write 、 read 、 close 等函数的包装， |\n| | 以及各种协议的分析和构建函数。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``notify.c`` | Redis 的数据库通知实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``object.c`` | Redis 的对象系统实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``pqsort.c`` 、 ``pqsort.h`` | 快速排序（QuickSort）算法的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``pubsub.c`` | 发布与订阅功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``rand.c`` 、 ``rand.h`` | 伪随机数生成器。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``rdb.c`` 、 ``rdb.h`` | RDB 持久化功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redisassert.h`` | Redis 自建的断言系统。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis-benchmark.c`` | Redis 的性能测试程序。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis.c`` | 负责服务器的启动、维护和关闭等事项。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis-check-aof.c`` 、 ``redis-check-dump.c`` | RDB 文件和 AOF 文件的合法性检查程序。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis-cli.c`` | Redis 客户端的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis.h`` | Redis 的主要头文件，记录了 Redis 中的大部分数据结构， |\n| | 包括服务器状态和客户端状态。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``redis-trib.rb`` | Redis 集群的管理程序。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``release.c`` 、 ``release.h`` | 记录和生成 Redis 的释出版本信息。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``replication.c`` | 复制功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``rio.c`` 、 ``rio.h`` | Redis 对文件 I/O 函数的包装， |\n| | 在普通 I/O 函数的基础上增加了显式缓存、以及计算校验和等功能。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``scripting.c`` | 脚本功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``sds.c`` 、 ``sds.h`` | SDS 数据结构的实现，SDS 为 Redis 的默认字符串表示。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``sentinel.c`` | Redis Sentinel 的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``setproctitle.c`` | 进程环境设置函数。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``sha1.c`` 、 ``sha1.h`` | SHA1 校验和计算函数。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``slowlog.c`` 、 ``slowlog.h`` | 慢查询功能的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``solarisfixes.h`` | 针对 Solaris 系统的补丁。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``sort.c`` | SORT 命令的实现。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``syncio.c`` | 同步 I/O 操作。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``testhelp.h`` | 测试辅助宏。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``t_hash.c`` 、 ``t_list.c`` 、 ``t_set.c`` 、 ``t_string.c`` 、 | 定义了 Redis 的各种数据类型，以及这些数据类型的命令。 |\n| ``t_zset.c`` | |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``util.c`` 、 ``util.h`` | 各种辅助函数。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``valgrind.sup`` | valgrind 的suppression文件。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``version.h`` | 记录了 Redis 的版本号。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``ziplist.c`` 、 ``ziplist.h`` | ZIPLIST 数据结构的实现，用于优化 LIST 类型。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``zipmap.c`` 、 ``zipmap.h`` | ZIPMAP 数据结构的实现，在 Redis 2.6 以前用与优化 HASH 类型， |\n| | Redis 2.6 开始已经废弃。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n| ``zmalloc.c`` 、 ``zmalloc.h`` | 内存管理程序。 |\n+-------------------------------------------------------------------+-------------------------------------------------------------------+\n" }, { "path": "deps/Makefile", "content": "# Redis dependency Makefile\n\nuname_S:= $(shell sh -c 'uname -s 2>/dev/null || echo not')\n\nCCCOLOR=\"\\033[34m\"\nLINKCOLOR=\"\\033[34;1m\"\nSRCCOLOR=\"\\033[33m\"\nBINCOLOR=\"\\033[37;1m\"\nMAKECOLOR=\"\\033[32;1m\"\nENDCOLOR=\"\\033[0m\"\n\ndefault:\n\t@echo \"Explicit target required\"\n\n.PHONY: default\n\n# Prerequisites target\n.make-prerequisites:\n\t@touch $@\n\n# Clean everything when CFLAGS is different\nifneq ($(shell sh -c '[ -f .make-cflags ] && cat .make-cflags || echo none'), $(CFLAGS))\n.make-cflags: distclean\n\t-(echo \"$(CFLAGS)\" > .make-cflags)\n.make-prerequisites: .make-cflags\nendif\n\n# Clean everything when LDFLAGS is different\nifneq ($(shell sh -c '[ -f .make-ldflags ] && cat .make-ldflags || echo none'), $(LDFLAGS))\n.make-ldflags: distclean\n\t-(echo \"$(LDFLAGS)\" > .make-ldflags)\n.make-prerequisites: .make-ldflags\nendif\n\ndistclean:\n\t-(cd hiredis && $(MAKE) clean) > /dev/null || true\n\t-(cd linenoise && $(MAKE) clean) > /dev/null || true\n\t-(cd lua && $(MAKE) clean) > /dev/null || true\n\t-(cd jemalloc && [ -f Makefile ] && $(MAKE) distclean) > /dev/null || true\n\t-(rm -f .make-*)\n\n.PHONY: distclean\n\nhiredis: .make-prerequisites\n\t@printf '%b %b\\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)\n\tcd hiredis && $(MAKE) static\n\n.PHONY: hiredis\n\nlinenoise: .make-prerequisites\n\t@printf '%b %b\\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)\n\tcd linenoise && $(MAKE)\n\n.PHONY: linenoise\n\nifeq ($(uname_S),SunOS)\n\t# Make isinf() available\n\tLUA_CFLAGS= -D__C99FEATURES__=1\nendif\n\nLUA_CFLAGS+= -O2 -Wall -DLUA_ANSI $(CFLAGS)\nLUA_LDFLAGS+= $(LDFLAGS)\n\nlua: .make-prerequisites\n\t@printf '%b %b\\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)\n\tcd lua/src && $(MAKE) all CFLAGS=\"$(LUA_CFLAGS)\" MYLDFLAGS=\"$(LUA_LDFLAGS)\"\n\n.PHONY: lua\n\nJEMALLOC_CFLAGS= -std=gnu99 -Wall -pipe -g3 -O3 -funroll-loops $(CFLAGS)\nJEMALLOC_LDFLAGS= $(LDFLAGS)\n\njemalloc: .make-prerequisites\n\t@printf '%b %b\\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)\n\tcd jemalloc && ./configure --with-jemalloc-prefix=je_ --enable-cc-silence CFLAGS=\"$(JEMALLOC_CFLAGS)\" LDFLAGS=\"$(JEMALLOC_LDFLAGS)\"\n\tcd jemalloc && $(MAKE) CFLAGS=\"$(JEMALLOC_CFLAGS)\" LDFLAGS=\"$(JEMALLOC_LDFLAGS)\" lib/libjemalloc.a\n\n.PHONY: jemalloc\n" }, { "path": "deps/hiredis/.gitignore", "content": "/hiredis-test\n/hiredis-example*\n/*.o\n/*.so\n/*.dylib\n/*.a\n" }, { "path": "deps/hiredis/CHANGELOG.md", "content": "### 0.10.1\n\n* Makefile overhaul. Important to check out if you override one or more\n variables using environment variables or via arguments to the \"make\" tool.\n\n* Issue #45: Fix potential memory leak for a multi bulk reply with 0 elements\n being created by the default reply object functions.\n\n* Issue #43: Don't crash in an asynchronous context when Redis returns an error\n reply after the connection has been made (this happens when the maximum\n number of connections is reached).\n\n### 0.10.0\n\n* See commit log.\n\n" }, { "path": "deps/hiredis/COPYING", "content": "Copyright (c) 2009-2011, Salvatore Sanfilippo \nCopyright (c) 2010-2011, Pieter Noordhuis \n\nAll rights reserved.\n\nRedistribution and use in source and binary forms, with or without\nmodification, are permitted provided that the following conditions are met:\n\n* Redistributions of source code must retain the above copyright notice,\n this list of conditions and the following disclaimer.\n\n* Redistributions in binary form must reproduce the above copyright notice,\n this list of conditions and the following disclaimer in the documentation\n and/or other materials provided with the distribution.\n\n* Neither the name of Redis nor the names of its contributors may be used\n to endorse or promote products derived from this software without specific\n prior written permission.\n\nTHIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND\nANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\nWARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\nDISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR\nANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES\n(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;\nLOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON\nANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS\nSOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n" }, { "path": "deps/hiredis/Makefile", "content": "# Hiredis Makefile\n# Copyright (C) 2010-2011 Salvatore Sanfilippo \n# Copyright (C) 2010-2011 Pieter Noordhuis \n# This file is released under the BSD license, see the COPYING file\n\nOBJ=net.o hiredis.o sds.o async.o\nBINS=hiredis-example hiredis-test\nLIBNAME=libhiredis\n\nHIREDIS_MAJOR=0\nHIREDIS_MINOR=10\n\n# Fallback to gcc when $CC is not in $PATH.\nCC:=$(shell sh -c 'type $(CC) >/dev/null 2>/dev/null && echo $(CC) || echo gcc')\nOPTIMIZATION?=-O3\nWARNINGS=-Wall -W -Wstrict-prototypes -Wwrite-strings\nDEBUG?= -g -ggdb\nREAL_CFLAGS=$(OPTIMIZATION) -fPIC $(CFLAGS) $(WARNINGS) $(DEBUG) $(ARCH)\nREAL_LDFLAGS=$(LDFLAGS) $(ARCH)\n\nDYLIBSUFFIX=so\nSTLIBSUFFIX=a\nDYLIB_MINOR_NAME=$(LIBNAME).$(DYLIBSUFFIX).$(HIREDIS_MAJOR).$(HIREDIS_MINOR)\nDYLIB_MAJOR_NAME=$(LIBNAME).$(DYLIBSUFFIX).$(HIREDIS_MAJOR)\nDYLIBNAME=$(LIBNAME).$(DYLIBSUFFIX)\nDYLIB_MAKE_CMD=$(CC) -shared -Wl,-soname,$(DYLIB_MINOR_NAME) -o $(DYLIBNAME) $(LDFLAGS)\nSTLIBNAME=$(LIBNAME).$(STLIBSUFFIX)\nSTLIB_MAKE_CMD=ar rcs $(STLIBNAME)\n\n# Platform-specific overrides\nuname_S := $(shell sh -c 'uname -s 2>/dev/null || echo not')\nifeq ($(uname_S),SunOS)\n REAL_LDFLAGS+= -ldl -lnsl -lsocket\n DYLIB_MAKE_CMD=$(CC) -G -o $(DYLIBNAME) -h $(DYLIB_MINOR_NAME) $(LDFLAGS)\n INSTALL= cp -r\nendif\nifeq ($(uname_S),Darwin)\n DYLIBSUFFIX=dylib\n DYLIB_MINOR_NAME=$(LIBNAME).$(HIREDIS_MAJOR).$(HIREDIS_MINOR).$(DYLIBSUFFIX)\n DYLIB_MAJOR_NAME=$(LIBNAME).$(HIREDIS_MAJOR).$(DYLIBSUFFIX)\n DYLIB_MAKE_CMD=$(CC) -shared -Wl,-install_name,$(DYLIB_MINOR_NAME) -o $(DYLIBNAME) $(LDFLAGS)\nendif\n\nall: $(DYLIBNAME) $(BINS)\n\n# Deps (use make dep to generate this)\nnet.o: net.c fmacros.h net.h hiredis.h\nasync.o: async.c async.h hiredis.h sds.h dict.c dict.h\nexample.o: example.c hiredis.h\nhiredis.o: hiredis.c fmacros.h hiredis.h net.h sds.h\nsds.o: sds.c sds.h\ntest.o: test.c hiredis.h\n\n$(DYLIBNAME): $(OBJ)\n\t$(DYLIB_MAKE_CMD) $(OBJ)\n\n$(STLIBNAME): $(OBJ)\n\t$(STLIB_MAKE_CMD) $(OBJ)\n\ndynamic: $(DYLIBNAME)\nstatic: $(STLIBNAME)\n\n# Binaries:\nhiredis-example-libevent: example-libevent.c adapters/libevent.h $(STLIBNAME)\n\t$(CC) -o $@ $(REAL_CFLAGS) $(REAL_LDFLAGS) -levent example-libevent.c $(STLIBNAME)\n\nhiredis-example-libev: example-libev.c adapters/libev.h $(STLIBNAME)\n\t$(CC) -o $@ $(REAL_CFLAGS) $(REAL_LDFLAGS) -lev example-libev.c $(STLIBNAME)\n\nifndef AE_DIR\nhiredis-example-ae:\n\t@echo \"Please specify AE_DIR (e.g. /src)\"\n\t@false\nelse\nhiredis-example-ae: example-ae.c adapters/ae.h $(STLIBNAME)\n\t$(CC) -o $@ $(REAL_CFLAGS) $(REAL_LDFLAGS) -I$(AE_DIR) $(AE_DIR)/ae.o $(AE_DIR)/zmalloc.o example-ae.c $(STLIBNAME)\nendif\n\nhiredis-%: %.o $(STLIBNAME)\n\t$(CC) -o $@ $(REAL_LDFLAGS) $< $(STLIBNAME)\n\ntest: hiredis-test\n\t./hiredis-test\n\ncheck: hiredis-test\n\techo \\\n\t\t\"daemonize yes\\n\" \\\n\t\t\"pidfile /tmp/hiredis-test-redis.pid\\n\" \\\n\t\t\"port 56379\\n\" \\\n\t\t\"bind 127.0.0.1\\n\" \\\n\t\t\"unixsocket /tmp/hiredis-test-redis.sock\" \\\n\t\t\t| redis-server -\n\t./hiredis-test -h 127.0.0.1 -p 56379 -s /tmp/hiredis-test-redis.sock || \\\n\t\t\t( kill `cat /tmp/hiredis-test-redis.pid` && false )\n\tkill `cat /tmp/hiredis-test-redis.pid`\n\n.c.o:\n\t$(CC) -std=c99 -pedantic -c $(REAL_CFLAGS) $<\n\nclean:\n\trm -rf $(DYLIBNAME) $(STLIBNAME) $(BINS) hiredis-example* *.o *.gcda *.gcno *.gcov\n\ndep:\n\t$(CC) -MM *.c\n\n# Installation related variables and target\nPREFIX?=/usr/local\nINCLUDE_PATH?=include/hiredis\nLIBRARY_PATH?=lib\nINSTALL_INCLUDE_PATH= $(PREFIX)/$(INCLUDE_PATH)\nINSTALL_LIBRARY_PATH= $(PREFIX)/$(LIBRARY_PATH)\n\nifeq ($(uname_S),SunOS)\n INSTALL?= cp -r\nendif\n\nINSTALL?= cp -a\n\ninstall: $(DYLIBNAME) $(STLIBNAME)\n\tmkdir -p $(INSTALL_INCLUDE_PATH) $(INSTALL_LIBRARY_PATH)\n\t$(INSTALL) hiredis.h async.h adapters $(INSTALL_INCLUDE_PATH)\n\t$(INSTALL) $(DYLIBNAME) $(INSTALL_LIBRARY_PATH)/$(DYLIB_MINOR_NAME)\n\tcd $(INSTALL_LIBRARY_PATH) && ln -sf $(DYLIB_MINOR_NAME) $(DYLIB_MAJOR_NAME)\n\tcd $(INSTALL_LIBRARY_PATH) && ln -sf $(DYLIB_MAJOR_NAME) $(DYLIBNAME)\n\t$(INSTALL) $(STLIBNAME) $(INSTALL_LIBRARY_PATH)\n\n32bit:\n\t@echo \"\"\n\t@echo \"WARNING: if this fails under Linux you probably need to install libc6-dev-i386\"\n\t@echo \"\"\n\t$(MAKE) CFLAGS=\"-m32\" LDFLAGS=\"-m32\"\n\ngprof:\n\t$(MAKE) CFLAGS=\"-pg\" LDFLAGS=\"-pg\"\n\ngcov:\n\t$(MAKE) CFLAGS=\"-fprofile-arcs -ftest-coverage\" LDFLAGS=\"-fprofile-arcs\"\n\ncoverage: gcov\n\tmake check\n\tmkdir -p tmp/lcov\n\tlcov -d . -c -o tmp/lcov/hiredis.info\n\tgenhtml --legend -o tmp/lcov/report tmp/lcov/hiredis.info\n\nnoopt:\n\t$(MAKE) OPTIMIZATION=\"\"\n\n.PHONY: all test check clean dep install 32bit gprof gcov noopt\n" }, { "path": "deps/hiredis/README.md", "content": "# HIREDIS\n\nHiredis is a minimalistic C client library for the [Redis](http://redis.io/) database.\n\nIt is minimalistic because it just adds minimal support for the protocol, but\nat the same time it uses an high level printf-alike API in order to make it\nmuch higher level than otherwise suggested by its minimal code base and the\nlack of explicit bindings for every Redis command.\n\nApart from supporting sending commands and receiving replies, it comes with\na reply parser that is decoupled from the I/O layer. It\nis a stream parser designed for easy reusability, which can for instance be used\nin higher level language bindings for efficient reply parsing.\n\nHiredis only supports the binary-safe Redis protocol, so you can use it with any\nRedis version >= 1.2.0.\n\nThe library comes with multiple APIs. There is the\n*synchronous API*, the *asynchronous API* and the *reply parsing API*.\n\n## UPGRADING\n\nVersion 0.9.0 is a major overhaul of hiredis in every aspect. However, upgrading existing\ncode using hiredis should not be a big pain. The key thing to keep in mind when\nupgrading is that hiredis >= 0.9.0 uses a `redisContext*` to keep state, in contrast to\nthe stateless 0.0.1 that only has a file descriptor to work with.\n\n## Synchronous API\n\nTo consume the synchronous API, there are only a few function calls that need to be introduced:\n\n redisContext *redisConnect(const char *ip, int port);\n void *redisCommand(redisContext *c, const char *format, ...);\n void freeReplyObject(void *reply);\n\n### Connecting\n\nThe function `redisConnect` is used to create a so-called `redisContext`. The\ncontext is where Hiredis holds state for a connection. The `redisContext`\nstruct has an integer `err` field that is non-zero when an the connection is in\nan error state. The field `errstr` will contain a string with a description of\nthe error. More information on errors can be found in the **Errors** section.\nAfter trying to connect to Redis using `redisConnect` you should\ncheck the `err` field to see if establishing the connection was successful:\n\n redisContext *c = redisConnect(\"127.0.0.1\", 6379);\n if (c->err) {\n printf(\"Error: %s\\n\", c->errstr);\n // handle error\n }\n\n### Sending commands\n\nThere are several ways to issue commands to Redis. The first that will be introduced is\n`redisCommand`. This function takes a format similar to printf. In the simplest form,\nit is used like this:\n\n reply = redisCommand(context, \"SET foo bar\");\n\nThe specifier `%s` interpolates a string in the command, and uses `strlen` to\ndetermine the length of the string:\n\n reply = redisCommand(context, \"SET foo %s\", value);\n\nWhen you need to pass binary safe strings in a command, the `%b` specifier can be\nused. Together with a pointer to the string, it requires a `size_t` length argument\nof the string:\n\n reply = redisCommand(context, \"SET foo %b\", value, valuelen);\n\nInternally, Hiredis splits the command in different arguments and will\nconvert it to the protocol used to communicate with Redis.\nOne or more spaces separates arguments, so you can use the specifiers\nanywhere in an argument:\n\n reply = redisCommand(context, \"SET key:%s %s\", myid, value);\n\n### Using replies\n\nThe return value of `redisCommand` holds a reply when the command was\nsuccessfully executed. When an error occurs, the return value is `NULL` and\nthe `err` field in the context will be set (see section on **Errors**).\nOnce an error is returned the context cannot be reused and you should set up\na new connection.\n\nThe standard replies that `redisCommand` are of the type `redisReply`. The\n`type` field in the `redisReply` should be used to test what kind of reply\nwas received:\n\n* **`REDIS_REPLY_STATUS`**:\n * The command replied with a status reply. The status string can be accessed using `reply->str`.\n The length of this string can be accessed using `reply->len`.\n\n* **`REDIS_REPLY_ERROR`**:\n * The command replied with an error. The error string can be accessed identical to `REDIS_REPLY_STATUS`.\n\n* **`REDIS_REPLY_INTEGER`**:\n * The command replied with an integer. The integer value can be accessed using the\n `reply->integer` field of type `long long`.\n\n* **`REDIS_REPLY_NIL`**:\n * The command replied with a **nil** object. There is no data to access.\n\n* **`REDIS_REPLY_STRING`**:\n * A bulk (string) reply. The value of the reply can be accessed using `reply->str`.\n The length of this string can be accessed using `reply->len`.\n\n* **`REDIS_REPLY_ARRAY`**:\n * A multi bulk reply. The number of elements in the multi bulk reply is stored in\n `reply->elements`. Every element in the multi bulk reply is a `redisReply` object as well\n and can be accessed via `reply->element[..index..]`.\n Redis may reply with nested arrays but this is fully supported.\n\nReplies should be freed using the `freeReplyObject()` function.\nNote that this function will take care of freeing sub-replies objects\ncontained in arrays and nested arrays, so there is no need for the user to\nfree the sub replies (it is actually harmful and will corrupt the memory).\n\n**Important:** the current version of hiredis (0.10.0) free's replies when the\nasynchronous API is used. This means you should not call `freeReplyObject` when\nyou use this API. The reply is cleaned up by hiredis _after_ the callback\nreturns. This behavior will probably change in future releases, so make sure to\nkeep an eye on the changelog when upgrading (see issue #39).\n\n### Cleaning up\n\nTo disconnect and free the context the following function can be used:\n\n void redisFree(redisContext *c);\n\nThis function immediately closes the socket and then free's the allocations done in\ncreating the context.\n\n### Sending commands (cont'd)\n\nTogether with `redisCommand`, the function `redisCommandArgv` can be used to issue commands.\nIt has the following prototype:\n\n void *redisCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen);\n\nIt takes the number of arguments `argc`, an array of strings `argv` and the lengths of the\narguments `argvlen`. For convenience, `argvlen` may be set to `NULL` and the function will\nuse `strlen(3)` on every argument to determine its length. Obviously, when any of the arguments\nneed to be binary safe, the entire array of lengths `argvlen` should be provided.\n\nThe return value has the same semantic as `redisCommand`.\n\n### Pipelining\n\nTo explain how Hiredis supports pipelining in a blocking connection, there needs to be\nunderstanding of the internal execution flow.\n\nWhen any of the functions in the `redisCommand` family is called, Hiredis first formats the\ncommand according to the Redis protocol. The formatted command is then put in the output buffer\nof the context. This output buffer is dynamic, so it can hold any number of commands.\nAfter the command is put in the output buffer, `redisGetReply` is called. This function has the\nfollowing two execution paths:\n\n1. The input buffer is non-empty:\n * Try to parse a single reply from the input buffer and return it\n * If no reply could be parsed, continue at *2*\n2. The input buffer is empty:\n * Write the **entire** output buffer to the socket\n * Read from the socket until a single reply could be parsed\n\nThe function `redisGetReply` is exported as part of the Hiredis API and can be used when a reply\nis expected on the socket. To pipeline commands, the only things that needs to be done is\nfilling up the output buffer. For this cause, two commands can be used that are identical\nto the `redisCommand` family, apart from not returning a reply:\n\n void redisAppendCommand(redisContext *c, const char *format, ...);\n void redisAppendCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen);\n\nAfter calling either function one or more times, `redisGetReply` can be used to receive the\nsubsequent replies. The return value for this function is either `REDIS_OK` or `REDIS_ERR`, where\nthe latter means an error occurred while reading a reply. Just as with the other commands,\nthe `err` field in the context can be used to find out what the cause of this error is.\n\nThe following examples shows a simple pipeline (resulting in only a single call to `write(2)` and\na single call to `read(2)`):\n\n redisReply *reply;\n redisAppendCommand(context,\"SET foo bar\");\n redisAppendCommand(context,\"GET foo\");\n redisGetReply(context,&reply); // reply for SET\n freeReplyObject(reply);\n redisGetReply(context,&reply); // reply for GET\n freeReplyObject(reply);\n\nThis API can also be used to implement a blocking subscriber:\n\n reply = redisCommand(context,\"SUBSCRIBE foo\");\n freeReplyObject(reply);\n while(redisGetReply(context,&reply) == REDIS_OK) {\n // consume message\n freeReplyObject(reply);\n }\n\n### Errors\n\nWhen a function call is not successful, depending on the function either `NULL` or `REDIS_ERR` is\nreturned. The `err` field inside the context will be non-zero and set to one of the\nfollowing constants:\n\n* **`REDIS_ERR_IO`**:\n There was an I/O error while creating the connection, trying to write\n to the socket or read from the socket. If you included `errno.h` in your\n application, you can use the global `errno` variable to find out what is\n wrong.\n\n* **`REDIS_ERR_EOF`**:\n The server closed the connection which resulted in an empty read.\n\n* **`REDIS_ERR_PROTOCOL`**:\n There was an error while parsing the protocol.\n\n* **`REDIS_ERR_OTHER`**:\n Any other error. Currently, it is only used when a specified hostname to connect\n to cannot be resolved.\n\nIn every case, the `errstr` field in the context will be set to hold a string representation\nof the error.\n\n## Asynchronous API\n\nHiredis comes with an asynchronous API that works easily with any event library.\nExamples are bundled that show using Hiredis with [libev](http://software.schmorp.de/pkg/libev.html)\nand [libevent](http://monkey.org/~provos/libevent/).\n\n### Connecting\n\nThe function `redisAsyncConnect` can be used to establish a non-blocking connection to\nRedis. It returns a pointer to the newly created `redisAsyncContext` struct. The `err` field\nshould be checked after creation to see if there were errors creating the connection.\nBecause the connection that will be created is non-blocking, the kernel is not able to\ninstantly return if the specified host and port is able to accept a connection.\n\n redisAsyncContext *c = redisAsyncConnect(\"127.0.0.1\", 6379);\n if (c->err) {\n printf(\"Error: %s\\n\", c->errstr);\n // handle error\n }\n\nThe asynchronous context can hold a disconnect callback function that is called when the\nconnection is disconnected (either because of an error or per user request). This function should\nhave the following prototype:\n\n void(const redisAsyncContext *c, int status);\n\nOn a disconnect, the `status` argument is set to `REDIS_OK` when disconnection was initiated by the\nuser, or `REDIS_ERR` when the disconnection was caused by an error. When it is `REDIS_ERR`, the `err`\nfield in the context can be accessed to find out the cause of the error.\n\nThe context object is always free'd after the disconnect callback fired. When a reconnect is needed,\nthe disconnect callback is a good point to do so.\n\nSetting the disconnect callback can only be done once per context. For subsequent calls it will\nreturn `REDIS_ERR`. The function to set the disconnect callback has the following prototype:\n\n int redisAsyncSetDisconnectCallback(redisAsyncContext *ac, redisDisconnectCallback *fn);\n\n### Sending commands and their callbacks\n\nIn an asynchronous context, commands are automatically pipelined due to the nature of an event loop.\nTherefore, unlike the synchronous API, there is only a single way to send commands.\nBecause commands are sent to Redis asynchronously, issuing a command requires a callback function\nthat is called when the reply is received. Reply callbacks should have the following prototype:\n\n void(redisAsyncContext *c, void *reply, void *privdata);\n\nThe `privdata` argument can be used to curry arbitrary data to the callback from the point where\nthe command is initially queued for execution.\n\nThe functions that can be used to issue commands in an asynchronous context are:\n\n int redisAsyncCommand(\n redisAsyncContext *ac, redisCallbackFn *fn, void *privdata,\n const char *format, ...);\n int redisAsyncCommandArgv(\n redisAsyncContext *ac, redisCallbackFn *fn, void *privdata,\n int argc, const char **argv, const size_t *argvlen);\n\nBoth functions work like their blocking counterparts. The return value is `REDIS_OK` when the command\nwas successfully added to the output buffer and `REDIS_ERR` otherwise. Example: when the connection\nis being disconnected per user-request, no new commands may be added to the output buffer and `REDIS_ERR` is\nreturned on calls to the `redisAsyncCommand` family.\n\nIf the reply for a command with a `NULL` callback is read, it is immediately free'd. When the callback\nfor a command is non-`NULL`, the memory is free'd immediately following the callback: the reply is only\nvalid for the duration of the callback.\n\nAll pending callbacks are called with a `NULL` reply when the context encountered an error.\n\n### Disconnecting\n\nAn asynchronous connection can be terminated using:\n\n void redisAsyncDisconnect(redisAsyncContext *ac);\n\nWhen this function is called, the connection is **not** immediately terminated. Instead, new\ncommands are no longer accepted and the connection is only terminated when all pending commands\nhave been written to the socket, their respective replies have been read and their respective\ncallbacks have been executed. After this, the disconnection callback is executed with the\n`REDIS_OK` status and the context object is free'd.\n\n### Hooking it up to event library *X*\n\nThere are a few hooks that need to be set on the context object after it is created.\nSee the `adapters/` directory for bindings to *libev* and *libevent*.\n\n## Reply parsing API\n\nHiredis comes with a reply parsing API that makes it easy for writing higher\nlevel language bindings.\n\nThe reply parsing API consists of the following functions:\n\n redisReader *redisReaderCreate(void);\n void redisReaderFree(redisReader *reader);\n int redisReaderFeed(redisReader *reader, const char *buf, size_t len);\n int redisReaderGetReply(redisReader *reader, void **reply);\n\nThe same set of functions are used internally by hiredis when creating a\nnormal Redis context, the above API just exposes it to the user for a direct\nusage.\n\n### Usage\n\nThe function `redisReaderCreate` creates a `redisReader` structure that holds a\nbuffer with unparsed data and state for the protocol parser.\n\nIncoming data -- most likely from a socket -- can be placed in the internal\nbuffer of the `redisReader` using `redisReaderFeed`. This function will make a\ncopy of the buffer pointed to by `buf` for `len` bytes. This data is parsed\nwhen `redisReaderGetReply` is called. This function returns an integer status\nand a reply object (as described above) via `void **reply`. The returned status\ncan be either `REDIS_OK` or `REDIS_ERR`, where the latter means something went\nwrong (either a protocol error, or an out of memory error).\n\n### Customizing replies\n\nThe function `redisReaderGetReply` creates `redisReply` and makes the function\nargument `reply` point to the created `redisReply` variable. For instance, if\nthe response of type `REDIS_REPLY_STATUS` then the `str` field of `redisReply`\nwill hold the status as a vanilla C string. However, the functions that are\nresponsible for creating instances of the `redisReply` can be customized by\nsetting the `fn` field on the `redisReader` struct. This should be done\nimmediately after creating the `redisReader`.\n\nFor example, [hiredis-rb](https://github.com/pietern/hiredis-rb/blob/master/ext/hiredis_ext/reader.c)\nuses customized reply object functions to create Ruby objects.\n\n### Reader max buffer\n\nBoth when using the Reader API directly or when using it indirectly via a\nnormal Redis context, the redisReader structure uses a buffer in order to\naccumulate data from the server.\nUsually this buffer is destroyed when it is empty and is larger than 16\nkb in order to avoid wasting memory in unused buffers\n\nHowever when working with very big payloads destroying the buffer may slow\ndown performances considerably, so it is possible to modify the max size of\nan idle buffer changing the value of the `maxbuf` field of the reader structure\nto the desired value. The special value of 0 means that there is no maximum\nvalue for an idle buffer, so the buffer will never get freed.\n\nFor instance if you have a normal Redis context you can set the maximum idle\nbuffer to zero (unlimited) just with:\n\n context->reader->maxbuf = 0;\n\nThis should be done only in order to maximize performances when working with\nlarge payloads. The context should be set back to `REDIS_READER_MAX_BUF` again\nas soon as possible in order to prevent allocation of useless memory.\n\n## AUTHORS\n\nHiredis was written by Salvatore Sanfilippo (antirez at gmail) and\nPieter Noordhuis (pcnoordhuis at gmail) and is released under the BSD license.\n" }, { "path": "deps/hiredis/adapters/ae.h", "content": "/*\n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __HIREDIS_AE_H__\n#define __HIREDIS_AE_H__\n#include \n#include \n#include \"../hiredis.h\"\n#include \"../async.h\"\n\ntypedef struct redisAeEvents {\n redisAsyncContext *context;\n aeEventLoop *loop;\n int fd;\n int reading, writing;\n} redisAeEvents;\n\nstatic void redisAeReadEvent(aeEventLoop *el, int fd, void *privdata, int mask) {\n ((void)el); ((void)fd); ((void)mask);\n\n redisAeEvents *e = (redisAeEvents*)privdata;\n redisAsyncHandleRead(e->context);\n}\n\nstatic void redisAeWriteEvent(aeEventLoop *el, int fd, void *privdata, int mask) {\n ((void)el); ((void)fd); ((void)mask);\n\n redisAeEvents *e = (redisAeEvents*)privdata;\n redisAsyncHandleWrite(e->context);\n}\n\nstatic void redisAeAddRead(void *privdata) {\n redisAeEvents *e = (redisAeEvents*)privdata;\n aeEventLoop *loop = e->loop;\n if (!e->reading) {\n e->reading = 1;\n aeCreateFileEvent(loop,e->fd,AE_READABLE,redisAeReadEvent,e);\n }\n}\n\nstatic void redisAeDelRead(void *privdata) {\n redisAeEvents *e = (redisAeEvents*)privdata;\n aeEventLoop *loop = e->loop;\n if (e->reading) {\n e->reading = 0;\n aeDeleteFileEvent(loop,e->fd,AE_READABLE);\n }\n}\n\nstatic void redisAeAddWrite(void *privdata) {\n redisAeEvents *e = (redisAeEvents*)privdata;\n aeEventLoop *loop = e->loop;\n if (!e->writing) {\n e->writing = 1;\n aeCreateFileEvent(loop,e->fd,AE_WRITABLE,redisAeWriteEvent,e);\n }\n}\n\nstatic void redisAeDelWrite(void *privdata) {\n redisAeEvents *e = (redisAeEvents*)privdata;\n aeEventLoop *loop = e->loop;\n if (e->writing) {\n e->writing = 0;\n aeDeleteFileEvent(loop,e->fd,AE_WRITABLE);\n }\n}\n\nstatic void redisAeCleanup(void *privdata) {\n redisAeEvents *e = (redisAeEvents*)privdata;\n redisAeDelRead(privdata);\n redisAeDelWrite(privdata);\n free(e);\n}\n\nstatic int redisAeAttach(aeEventLoop *loop, redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n redisAeEvents *e;\n\n /* Nothing should be attached when something is already attached */\n if (ac->ev.data != NULL)\n return REDIS_ERR;\n\n /* Create container for context and r/w events */\n e = (redisAeEvents*)malloc(sizeof(*e));\n e->context = ac;\n e->loop = loop;\n e->fd = c->fd;\n e->reading = e->writing = 0;\n\n /* Register functions to start/stop listening for events */\n ac->ev.addRead = redisAeAddRead;\n ac->ev.delRead = redisAeDelRead;\n ac->ev.addWrite = redisAeAddWrite;\n ac->ev.delWrite = redisAeDelWrite;\n ac->ev.cleanup = redisAeCleanup;\n ac->ev.data = e;\n\n return REDIS_OK;\n}\n#endif\n" }, { "path": "deps/hiredis/adapters/libev.h", "content": "/*\n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __HIREDIS_LIBEV_H__\n#define __HIREDIS_LIBEV_H__\n#include \n#include \n#include \n#include \"../hiredis.h\"\n#include \"../async.h\"\n\ntypedef struct redisLibevEvents {\n redisAsyncContext *context;\n struct ev_loop *loop;\n int reading, writing;\n ev_io rev, wev;\n} redisLibevEvents;\n\nstatic void redisLibevReadEvent(EV_P_ ev_io *watcher, int revents) {\n#if EV_MULTIPLICITY\n ((void)loop);\n#endif\n ((void)revents);\n\n redisLibevEvents *e = (redisLibevEvents*)watcher->data;\n redisAsyncHandleRead(e->context);\n}\n\nstatic void redisLibevWriteEvent(EV_P_ ev_io *watcher, int revents) {\n#if EV_MULTIPLICITY\n ((void)loop);\n#endif\n ((void)revents);\n\n redisLibevEvents *e = (redisLibevEvents*)watcher->data;\n redisAsyncHandleWrite(e->context);\n}\n\nstatic void redisLibevAddRead(void *privdata) {\n redisLibevEvents *e = (redisLibevEvents*)privdata;\n struct ev_loop *loop = e->loop;\n ((void)loop);\n if (!e->reading) {\n e->reading = 1;\n ev_io_start(EV_A_ &e->rev);\n }\n}\n\nstatic void redisLibevDelRead(void *privdata) {\n redisLibevEvents *e = (redisLibevEvents*)privdata;\n struct ev_loop *loop = e->loop;\n ((void)loop);\n if (e->reading) {\n e->reading = 0;\n ev_io_stop(EV_A_ &e->rev);\n }\n}\n\nstatic void redisLibevAddWrite(void *privdata) {\n redisLibevEvents *e = (redisLibevEvents*)privdata;\n struct ev_loop *loop = e->loop;\n ((void)loop);\n if (!e->writing) {\n e->writing = 1;\n ev_io_start(EV_A_ &e->wev);\n }\n}\n\nstatic void redisLibevDelWrite(void *privdata) {\n redisLibevEvents *e = (redisLibevEvents*)privdata;\n struct ev_loop *loop = e->loop;\n ((void)loop);\n if (e->writing) {\n e->writing = 0;\n ev_io_stop(EV_A_ &e->wev);\n }\n}\n\nstatic void redisLibevCleanup(void *privdata) {\n redisLibevEvents *e = (redisLibevEvents*)privdata;\n redisLibevDelRead(privdata);\n redisLibevDelWrite(privdata);\n free(e);\n}\n\nstatic int redisLibevAttach(EV_P_ redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n redisLibevEvents *e;\n\n /* Nothing should be attached when something is already attached */\n if (ac->ev.data != NULL)\n return REDIS_ERR;\n\n /* Create container for context and r/w events */\n e = (redisLibevEvents*)malloc(sizeof(*e));\n e->context = ac;\n#if EV_MULTIPLICITY\n e->loop = loop;\n#else\n e->loop = NULL;\n#endif\n e->reading = e->writing = 0;\n e->rev.data = e;\n e->wev.data = e;\n\n /* Register functions to start/stop listening for events */\n ac->ev.addRead = redisLibevAddRead;\n ac->ev.delRead = redisLibevDelRead;\n ac->ev.addWrite = redisLibevAddWrite;\n ac->ev.delWrite = redisLibevDelWrite;\n ac->ev.cleanup = redisLibevCleanup;\n ac->ev.data = e;\n\n /* Initialize read/write events */\n ev_io_init(&e->rev,redisLibevReadEvent,c->fd,EV_READ);\n ev_io_init(&e->wev,redisLibevWriteEvent,c->fd,EV_WRITE);\n return REDIS_OK;\n}\n\n#endif\n" }, { "path": "deps/hiredis/adapters/libevent.h", "content": "/*\n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __HIREDIS_LIBEVENT_H__\n#define __HIREDIS_LIBEVENT_H__\n#include \n#include \"../hiredis.h\"\n#include \"../async.h\"\n\ntypedef struct redisLibeventEvents {\n redisAsyncContext *context;\n struct event rev, wev;\n} redisLibeventEvents;\n\nstatic void redisLibeventReadEvent(int fd, short event, void *arg) {\n ((void)fd); ((void)event);\n redisLibeventEvents *e = (redisLibeventEvents*)arg;\n redisAsyncHandleRead(e->context);\n}\n\nstatic void redisLibeventWriteEvent(int fd, short event, void *arg) {\n ((void)fd); ((void)event);\n redisLibeventEvents *e = (redisLibeventEvents*)arg;\n redisAsyncHandleWrite(e->context);\n}\n\nstatic void redisLibeventAddRead(void *privdata) {\n redisLibeventEvents *e = (redisLibeventEvents*)privdata;\n event_add(&e->rev,NULL);\n}\n\nstatic void redisLibeventDelRead(void *privdata) {\n redisLibeventEvents *e = (redisLibeventEvents*)privdata;\n event_del(&e->rev);\n}\n\nstatic void redisLibeventAddWrite(void *privdata) {\n redisLibeventEvents *e = (redisLibeventEvents*)privdata;\n event_add(&e->wev,NULL);\n}\n\nstatic void redisLibeventDelWrite(void *privdata) {\n redisLibeventEvents *e = (redisLibeventEvents*)privdata;\n event_del(&e->wev);\n}\n\nstatic void redisLibeventCleanup(void *privdata) {\n redisLibeventEvents *e = (redisLibeventEvents*)privdata;\n event_del(&e->rev);\n event_del(&e->wev);\n free(e);\n}\n\nstatic int redisLibeventAttach(redisAsyncContext *ac, struct event_base *base) {\n redisContext *c = &(ac->c);\n redisLibeventEvents *e;\n\n /* Nothing should be attached when something is already attached */\n if (ac->ev.data != NULL)\n return REDIS_ERR;\n\n /* Create container for context and r/w events */\n e = (redisLibeventEvents*)malloc(sizeof(*e));\n e->context = ac;\n\n /* Register functions to start/stop listening for events */\n ac->ev.addRead = redisLibeventAddRead;\n ac->ev.delRead = redisLibeventDelRead;\n ac->ev.addWrite = redisLibeventAddWrite;\n ac->ev.delWrite = redisLibeventDelWrite;\n ac->ev.cleanup = redisLibeventCleanup;\n ac->ev.data = e;\n\n /* Initialize and install read/write events */\n event_set(&e->rev,c->fd,EV_READ,redisLibeventReadEvent,e);\n event_set(&e->wev,c->fd,EV_WRITE,redisLibeventWriteEvent,e);\n event_base_set(base,&e->rev);\n event_base_set(base,&e->wev);\n return REDIS_OK;\n}\n#endif\n" }, { "path": "deps/hiredis/async.c", "content": "/*\n * Copyright (c) 2009-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"fmacros.h\"\n#include \n#include \n#include \n#include \n#include \n#include \n#include \"async.h\"\n#include \"net.h\"\n#include \"dict.c\"\n#include \"sds.h\"\n\n#define _EL_ADD_READ(ctx) do { \\\n if ((ctx)->ev.addRead) (ctx)->ev.addRead((ctx)->ev.data); \\\n } while(0)\n#define _EL_DEL_READ(ctx) do { \\\n if ((ctx)->ev.delRead) (ctx)->ev.delRead((ctx)->ev.data); \\\n } while(0)\n#define _EL_ADD_WRITE(ctx) do { \\\n if ((ctx)->ev.addWrite) (ctx)->ev.addWrite((ctx)->ev.data); \\\n } while(0)\n#define _EL_DEL_WRITE(ctx) do { \\\n if ((ctx)->ev.delWrite) (ctx)->ev.delWrite((ctx)->ev.data); \\\n } while(0)\n#define _EL_CLEANUP(ctx) do { \\\n if ((ctx)->ev.cleanup) (ctx)->ev.cleanup((ctx)->ev.data); \\\n } while(0);\n\n/* Forward declaration of function in hiredis.c */\nvoid __redisAppendCommand(redisContext *c, char *cmd, size_t len);\n\n/* Functions managing dictionary of callbacks for pub/sub. */\nstatic unsigned int callbackHash(const void *key) {\n return dictGenHashFunction((unsigned char*)key,sdslen((char*)key));\n}\n\nstatic void *callbackValDup(void *privdata, const void *src) {\n ((void) privdata);\n redisCallback *dup = malloc(sizeof(*dup));\n memcpy(dup,src,sizeof(*dup));\n return dup;\n}\n\nstatic int callbackKeyCompare(void *privdata, const void *key1, const void *key2) {\n int l1, l2;\n ((void) privdata);\n\n l1 = sdslen((sds)key1);\n l2 = sdslen((sds)key2);\n if (l1 != l2) return 0;\n return memcmp(key1,key2,l1) == 0;\n}\n\nstatic void callbackKeyDestructor(void *privdata, void *key) {\n ((void) privdata);\n sdsfree((sds)key);\n}\n\nstatic void callbackValDestructor(void *privdata, void *val) {\n ((void) privdata);\n free(val);\n}\n\nstatic dictType callbackDict = {\n callbackHash,\n NULL,\n callbackValDup,\n callbackKeyCompare,\n callbackKeyDestructor,\n callbackValDestructor\n};\n\nstatic redisAsyncContext *redisAsyncInitialize(redisContext *c) {\n redisAsyncContext *ac = realloc(c,sizeof(redisAsyncContext));\n c = &(ac->c);\n\n /* The regular connect functions will always set the flag REDIS_CONNECTED.\n * For the async API, we want to wait until the first write event is\n * received up before setting this flag, so reset it here. */\n c->flags &= ~REDIS_CONNECTED;\n\n ac->err = 0;\n ac->errstr = NULL;\n ac->data = NULL;\n\n ac->ev.data = NULL;\n ac->ev.addRead = NULL;\n ac->ev.delRead = NULL;\n ac->ev.addWrite = NULL;\n ac->ev.delWrite = NULL;\n ac->ev.cleanup = NULL;\n\n ac->onConnect = NULL;\n ac->onDisconnect = NULL;\n\n ac->replies.head = NULL;\n ac->replies.tail = NULL;\n ac->sub.invalid.head = NULL;\n ac->sub.invalid.tail = NULL;\n ac->sub.channels = dictCreate(&callbackDict,NULL);\n ac->sub.patterns = dictCreate(&callbackDict,NULL);\n return ac;\n}\n\n/* We want the error field to be accessible directly instead of requiring\n * an indirection to the redisContext struct. */\nstatic void __redisAsyncCopyError(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n ac->err = c->err;\n ac->errstr = c->errstr;\n}\n\nredisAsyncContext *redisAsyncConnect(const char *ip, int port) {\n redisContext *c = redisConnectNonBlock(ip,port);\n redisAsyncContext *ac = redisAsyncInitialize(c);\n __redisAsyncCopyError(ac);\n return ac;\n}\n\nredisAsyncContext *redisAsyncConnectUnix(const char *path) {\n redisContext *c = redisConnectUnixNonBlock(path);\n redisAsyncContext *ac = redisAsyncInitialize(c);\n __redisAsyncCopyError(ac);\n return ac;\n}\n\nint redisAsyncSetConnectCallback(redisAsyncContext *ac, redisConnectCallback *fn) {\n if (ac->onConnect == NULL) {\n ac->onConnect = fn;\n\n /* The common way to detect an established connection is to wait for\n * the first write event to be fired. This assumes the related event\n * library functions are already set. */\n _EL_ADD_WRITE(ac);\n return REDIS_OK;\n }\n return REDIS_ERR;\n}\n\nint redisAsyncSetDisconnectCallback(redisAsyncContext *ac, redisDisconnectCallback *fn) {\n if (ac->onDisconnect == NULL) {\n ac->onDisconnect = fn;\n return REDIS_OK;\n }\n return REDIS_ERR;\n}\n\n/* Helper functions to push/shift callbacks */\nstatic int __redisPushCallback(redisCallbackList *list, redisCallback *source) {\n redisCallback *cb;\n\n /* Copy callback from stack to heap */\n cb = malloc(sizeof(*cb));\n if (source != NULL) {\n memcpy(cb,source,sizeof(*cb));\n cb->next = NULL;\n }\n\n /* Store callback in list */\n if (list->head == NULL)\n list->head = cb;\n if (list->tail != NULL)\n list->tail->next = cb;\n list->tail = cb;\n return REDIS_OK;\n}\n\nstatic int __redisShiftCallback(redisCallbackList *list, redisCallback *target) {\n redisCallback *cb = list->head;\n if (cb != NULL) {\n list->head = cb->next;\n if (cb == list->tail)\n list->tail = NULL;\n\n /* Copy callback from heap to stack */\n if (target != NULL)\n memcpy(target,cb,sizeof(*cb));\n free(cb);\n return REDIS_OK;\n }\n return REDIS_ERR;\n}\n\nstatic void __redisRunCallback(redisAsyncContext *ac, redisCallback *cb, redisReply *reply) {\n redisContext *c = &(ac->c);\n if (cb->fn != NULL) {\n c->flags |= REDIS_IN_CALLBACK;\n cb->fn(ac,reply,cb->privdata);\n c->flags &= ~REDIS_IN_CALLBACK;\n }\n}\n\n/* Helper function to free the context. */\nstatic void __redisAsyncFree(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n redisCallback cb;\n dictIterator *it;\n dictEntry *de;\n\n /* Execute pending callbacks with NULL reply. */\n while (__redisShiftCallback(&ac->replies,&cb) == REDIS_OK)\n __redisRunCallback(ac,&cb,NULL);\n\n /* Execute callbacks for invalid commands */\n while (__redisShiftCallback(&ac->sub.invalid,&cb) == REDIS_OK)\n __redisRunCallback(ac,&cb,NULL);\n\n /* Run subscription callbacks callbacks with NULL reply */\n it = dictGetIterator(ac->sub.channels);\n while ((de = dictNext(it)) != NULL)\n __redisRunCallback(ac,dictGetEntryVal(de),NULL);\n dictReleaseIterator(it);\n dictRelease(ac->sub.channels);\n\n it = dictGetIterator(ac->sub.patterns);\n while ((de = dictNext(it)) != NULL)\n __redisRunCallback(ac,dictGetEntryVal(de),NULL);\n dictReleaseIterator(it);\n dictRelease(ac->sub.patterns);\n\n /* Signal event lib to clean up */\n _EL_CLEANUP(ac);\n\n /* Execute disconnect callback. When redisAsyncFree() initiated destroying\n * this context, the status will always be REDIS_OK. */\n if (ac->onDisconnect && (c->flags & REDIS_CONNECTED)) {\n if (c->flags & REDIS_FREEING) {\n ac->onDisconnect(ac,REDIS_OK);\n } else {\n ac->onDisconnect(ac,(ac->err == 0) ? REDIS_OK : REDIS_ERR);\n }\n }\n\n /* Cleanup self */\n redisFree(c);\n}\n\n/* Free the async context. When this function is called from a callback,\n * control needs to be returned to redisProcessCallbacks() before actual\n * free'ing. To do so, a flag is set on the context which is picked up by\n * redisProcessCallbacks(). Otherwise, the context is immediately free'd. */\nvoid redisAsyncFree(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n c->flags |= REDIS_FREEING;\n if (!(c->flags & REDIS_IN_CALLBACK))\n __redisAsyncFree(ac);\n}\n\n/* Helper function to make the disconnect happen and clean up. */\nstatic void __redisAsyncDisconnect(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n\n /* Make sure error is accessible if there is any */\n __redisAsyncCopyError(ac);\n\n if (ac->err == 0) {\n /* For clean disconnects, there should be no pending callbacks. */\n assert(__redisShiftCallback(&ac->replies,NULL) == REDIS_ERR);\n } else {\n /* Disconnection is caused by an error, make sure that pending\n * callbacks cannot call new commands. */\n c->flags |= REDIS_DISCONNECTING;\n }\n\n /* For non-clean disconnects, __redisAsyncFree() will execute pending\n * callbacks with a NULL-reply. */\n __redisAsyncFree(ac);\n}\n\n/* Tries to do a clean disconnect from Redis, meaning it stops new commands\n * from being issued, but tries to flush the output buffer and execute\n * callbacks for all remaining replies. When this function is called from a\n * callback, there might be more replies and we can safely defer disconnecting\n * to redisProcessCallbacks(). Otherwise, we can only disconnect immediately\n * when there are no pending callbacks. */\nvoid redisAsyncDisconnect(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n c->flags |= REDIS_DISCONNECTING;\n if (!(c->flags & REDIS_IN_CALLBACK) && ac->replies.head == NULL)\n __redisAsyncDisconnect(ac);\n}\n\nstatic int __redisGetSubscribeCallback(redisAsyncContext *ac, redisReply *reply, redisCallback *dstcb) {\n redisContext *c = &(ac->c);\n dict *callbacks;\n dictEntry *de;\n int pvariant;\n char *stype;\n sds sname;\n\n /* Custom reply functions are not supported for pub/sub. This will fail\n * very hard when they are used... */\n if (reply->type == REDIS_REPLY_ARRAY) {\n assert(reply->elements >= 2);\n assert(reply->element[0]->type == REDIS_REPLY_STRING);\n stype = reply->element[0]->str;\n pvariant = (tolower(stype[0]) == 'p') ? 1 : 0;\n\n if (pvariant)\n callbacks = ac->sub.patterns;\n else\n callbacks = ac->sub.channels;\n\n /* Locate the right callback */\n assert(reply->element[1]->type == REDIS_REPLY_STRING);\n sname = sdsnewlen(reply->element[1]->str,reply->element[1]->len);\n de = dictFind(callbacks,sname);\n if (de != NULL) {\n memcpy(dstcb,dictGetEntryVal(de),sizeof(*dstcb));\n\n /* If this is an unsubscribe message, remove it. */\n if (strcasecmp(stype+pvariant,\"unsubscribe\") == 0) {\n dictDelete(callbacks,sname);\n\n /* If this was the last unsubscribe message, revert to\n * non-subscribe mode. */\n assert(reply->element[2]->type == REDIS_REPLY_INTEGER);\n if (reply->element[2]->integer == 0)\n c->flags &= ~REDIS_SUBSCRIBED;\n }\n }\n sdsfree(sname);\n } else {\n /* Shift callback for invalid commands. */\n __redisShiftCallback(&ac->sub.invalid,dstcb);\n }\n return REDIS_OK;\n}\n\nvoid redisProcessCallbacks(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n redisCallback cb;\n void *reply = NULL;\n int status;\n\n while((status = redisGetReply(c,&reply)) == REDIS_OK) {\n if (reply == NULL) {\n /* When the connection is being disconnected and there are\n * no more replies, this is the cue to really disconnect. */\n if (c->flags & REDIS_DISCONNECTING && sdslen(c->obuf) == 0) {\n __redisAsyncDisconnect(ac);\n return;\n }\n \n /* If monitor mode, repush callback */\n if(c->flags & REDIS_MONITORING) {\n __redisPushCallback(&ac->replies,&cb);\n }\n\n /* When the connection is not being disconnected, simply stop\n * trying to get replies and wait for the next loop tick. */\n break;\n }\n\n /* Even if the context is subscribed, pending regular callbacks will\n * get a reply before pub/sub messages arrive. */\n if (__redisShiftCallback(&ac->replies,&cb) != REDIS_OK) {\n /*\n * A spontaneous reply in a not-subscribed context can be the error\n * reply that is sent when a new connection exceeds the maximum\n * number of allowed connections on the server side.\n *\n * This is seen as an error instead of a regular reply because the\n * server closes the connection after sending it.\n *\n * To prevent the error from being overwritten by an EOF error the\n * connection is closed here. See issue #43.\n *\n * Another possibility is that the server is loading its dataset.\n * In this case we also want to close the connection, and have the\n * user wait until the server is ready to take our request.\n */\n if (((redisReply*)reply)->type == REDIS_REPLY_ERROR) {\n c->err = REDIS_ERR_OTHER;\n snprintf(c->errstr,sizeof(c->errstr),\"%s\",((redisReply*)reply)->str);\n __redisAsyncDisconnect(ac);\n return;\n }\n /* No more regular callbacks and no errors, the context *must* be subscribed or monitoring. */\n assert((c->flags & REDIS_SUBSCRIBED || c->flags & REDIS_MONITORING));\n if(c->flags & REDIS_SUBSCRIBED)\n __redisGetSubscribeCallback(ac,reply,&cb);\n }\n\n if (cb.fn != NULL) {\n __redisRunCallback(ac,&cb,reply);\n c->reader->fn->freeObject(reply);\n\n /* Proceed with free'ing when redisAsyncFree() was called. */\n if (c->flags & REDIS_FREEING) {\n __redisAsyncFree(ac);\n return;\n }\n } else {\n /* No callback for this reply. This can either be a NULL callback,\n * or there were no callbacks to begin with. Either way, don't\n * abort with an error, but simply ignore it because the client\n * doesn't know what the server will spit out over the wire. */\n c->reader->fn->freeObject(reply);\n }\n }\n\n /* Disconnect when there was an error reading the reply */\n if (status != REDIS_OK)\n __redisAsyncDisconnect(ac);\n}\n\n/* Internal helper function to detect socket status the first time a read or\n * write event fires. When connecting was not succesful, the connect callback\n * is called with a REDIS_ERR status and the context is free'd. */\nstatic int __redisAsyncHandleConnect(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n\n if (redisCheckSocketError(c,c->fd) == REDIS_ERR) {\n /* Try again later when connect(2) is still in progress. */\n if (errno == EINPROGRESS)\n return REDIS_OK;\n\n if (ac->onConnect) ac->onConnect(ac,REDIS_ERR);\n __redisAsyncDisconnect(ac);\n return REDIS_ERR;\n }\n\n /* Mark context as connected. */\n c->flags |= REDIS_CONNECTED;\n if (ac->onConnect) ac->onConnect(ac,REDIS_OK);\n return REDIS_OK;\n}\n\n/* This function should be called when the socket is readable.\n * It processes all replies that can be read and executes their callbacks.\n */\nvoid redisAsyncHandleRead(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n\n if (!(c->flags & REDIS_CONNECTED)) {\n /* Abort connect was not successful. */\n if (__redisAsyncHandleConnect(ac) != REDIS_OK)\n return;\n /* Try again later when the context is still not connected. */\n if (!(c->flags & REDIS_CONNECTED))\n return;\n }\n\n if (redisBufferRead(c) == REDIS_ERR) {\n __redisAsyncDisconnect(ac);\n } else {\n /* Always re-schedule reads */\n _EL_ADD_READ(ac);\n redisProcessCallbacks(ac);\n }\n}\n\nvoid redisAsyncHandleWrite(redisAsyncContext *ac) {\n redisContext *c = &(ac->c);\n int done = 0;\n\n if (!(c->flags & REDIS_CONNECTED)) {\n /* Abort connect was not successful. */\n if (__redisAsyncHandleConnect(ac) != REDIS_OK)\n return;\n /* Try again later when the context is still not connected. */\n if (!(c->flags & REDIS_CONNECTED))\n return;\n }\n\n if (redisBufferWrite(c,&done) == REDIS_ERR) {\n __redisAsyncDisconnect(ac);\n } else {\n /* Continue writing when not done, stop writing otherwise */\n if (!done)\n _EL_ADD_WRITE(ac);\n else\n _EL_DEL_WRITE(ac);\n\n /* Always schedule reads after writes */\n _EL_ADD_READ(ac);\n }\n}\n\n/* Sets a pointer to the first argument and its length starting at p. Returns\n * the number of bytes to skip to get to the following argument. */\nstatic char *nextArgument(char *start, char **str, size_t *len) {\n char *p = start;\n if (p[0] != '$') {\n p = strchr(p,'$');\n if (p == NULL) return NULL;\n }\n\n *len = (int)strtol(p+1,NULL,10);\n p = strchr(p,'\\r');\n assert(p);\n *str = p+2;\n return p+2+(*len)+2;\n}\n\n/* Helper function for the redisAsyncCommand* family of functions. Writes a\n * formatted command to the output buffer and registers the provided callback\n * function with the context. */\nstatic int __redisAsyncCommand(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, char *cmd, size_t len) {\n redisContext *c = &(ac->c);\n redisCallback cb;\n int pvariant, hasnext;\n char *cstr, *astr;\n size_t clen, alen;\n char *p;\n sds sname;\n\n /* Don't accept new commands when the connection is about to be closed. */\n if (c->flags & (REDIS_DISCONNECTING | REDIS_FREEING)) return REDIS_ERR;\n\n /* Setup callback */\n cb.fn = fn;\n cb.privdata = privdata;\n\n /* Find out which command will be appended. */\n p = nextArgument(cmd,&cstr,&clen);\n assert(p != NULL);\n hasnext = (p[0] == '$');\n pvariant = (tolower(cstr[0]) == 'p') ? 1 : 0;\n cstr += pvariant;\n clen -= pvariant;\n\n if (hasnext && strncasecmp(cstr,\"subscribe\\r\\n\",11) == 0) {\n c->flags |= REDIS_SUBSCRIBED;\n\n /* Add every channel/pattern to the list of subscription callbacks. */\n while ((p = nextArgument(p,&astr,&alen)) != NULL) {\n sname = sdsnewlen(astr,alen);\n if (pvariant)\n dictReplace(ac->sub.patterns,sname,&cb);\n else\n dictReplace(ac->sub.channels,sname,&cb);\n }\n } else if (strncasecmp(cstr,\"unsubscribe\\r\\n\",13) == 0) {\n /* It is only useful to call (P)UNSUBSCRIBE when the context is\n * subscribed to one or more channels or patterns. */\n if (!(c->flags & REDIS_SUBSCRIBED)) return REDIS_ERR;\n\n /* (P)UNSUBSCRIBE does not have its own response: every channel or\n * pattern that is unsubscribed will receive a message. This means we\n * should not append a callback function for this command. */\n } else if(strncasecmp(cstr,\"monitor\\r\\n\",9) == 0) {\n /* Set monitor flag and push callback */\n c->flags |= REDIS_MONITORING;\n __redisPushCallback(&ac->replies,&cb);\n } else {\n if (c->flags & REDIS_SUBSCRIBED)\n /* This will likely result in an error reply, but it needs to be\n * received and passed to the callback. */\n __redisPushCallback(&ac->sub.invalid,&cb);\n else\n __redisPushCallback(&ac->replies,&cb);\n }\n\n __redisAppendCommand(c,cmd,len);\n\n /* Always schedule a write when the write buffer is non-empty */\n _EL_ADD_WRITE(ac);\n\n return REDIS_OK;\n}\n\nint redisvAsyncCommand(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, const char *format, va_list ap) {\n char *cmd;\n int len;\n int status;\n len = redisvFormatCommand(&cmd,format,ap);\n status = __redisAsyncCommand(ac,fn,privdata,cmd,len);\n free(cmd);\n return status;\n}\n\nint redisAsyncCommand(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, const char *format, ...) {\n va_list ap;\n int status;\n va_start(ap,format);\n status = redisvAsyncCommand(ac,fn,privdata,format,ap);\n va_end(ap);\n return status;\n}\n\nint redisAsyncCommandArgv(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, int argc, const char **argv, const size_t *argvlen) {\n char *cmd;\n int len;\n int status;\n len = redisFormatCommandArgv(&cmd,argc,argv,argvlen);\n status = __redisAsyncCommand(ac,fn,privdata,cmd,len);\n free(cmd);\n return status;\n}\n" }, { "path": "deps/hiredis/async.h", "content": "/*\n * Copyright (c) 2009-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __HIREDIS_ASYNC_H\n#define __HIREDIS_ASYNC_H\n#include \"hiredis.h\"\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\nstruct redisAsyncContext; /* need forward declaration of redisAsyncContext */\nstruct dict; /* dictionary header is included in async.c */\n\n/* Reply callback prototype and container */\ntypedef void (redisCallbackFn)(struct redisAsyncContext*, void*, void*);\ntypedef struct redisCallback {\n struct redisCallback *next; /* simple singly linked list */\n redisCallbackFn *fn;\n void *privdata;\n} redisCallback;\n\n/* List of callbacks for either regular replies or pub/sub */\ntypedef struct redisCallbackList {\n redisCallback *head, *tail;\n} redisCallbackList;\n\n/* Connection callback prototypes */\ntypedef void (redisDisconnectCallback)(const struct redisAsyncContext*, int status);\ntypedef void (redisConnectCallback)(const struct redisAsyncContext*, int status);\n\n/* Context for an async connection to Redis */\ntypedef struct redisAsyncContext {\n /* Hold the regular context, so it can be realloc'ed. */\n redisContext c;\n\n /* Setup error flags so they can be used directly. */\n int err;\n char *errstr;\n\n /* Not used by hiredis */\n void *data;\n\n /* Event library data and hooks */\n struct {\n void *data;\n\n /* Hooks that are called when the library expects to start\n * reading/writing. These functions should be idempotent. */\n void (*addRead)(void *privdata);\n void (*delRead)(void *privdata);\n void (*addWrite)(void *privdata);\n void (*delWrite)(void *privdata);\n void (*cleanup)(void *privdata);\n } ev;\n\n /* Called when either the connection is terminated due to an error or per\n * user request. The status is set accordingly (REDIS_OK, REDIS_ERR). */\n redisDisconnectCallback *onDisconnect;\n\n /* Called when the first write event was received. */\n redisConnectCallback *onConnect;\n\n /* Regular command callbacks */\n redisCallbackList replies;\n\n /* Subscription callbacks */\n struct {\n redisCallbackList invalid;\n struct dict *channels;\n struct dict *patterns;\n } sub;\n} redisAsyncContext;\n\n/* Functions that proxy to hiredis */\nredisAsyncContext *redisAsyncConnect(const char *ip, int port);\nredisAsyncContext *redisAsyncConnectUnix(const char *path);\nint redisAsyncSetConnectCallback(redisAsyncContext *ac, redisConnectCallback *fn);\nint redisAsyncSetDisconnectCallback(redisAsyncContext *ac, redisDisconnectCallback *fn);\nvoid redisAsyncDisconnect(redisAsyncContext *ac);\nvoid redisAsyncFree(redisAsyncContext *ac);\n\n/* Handle read/write events */\nvoid redisAsyncHandleRead(redisAsyncContext *ac);\nvoid redisAsyncHandleWrite(redisAsyncContext *ac);\n\n/* Command functions for an async context. Write the command to the\n * output buffer and register the provided callback. */\nint redisvAsyncCommand(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, const char *format, va_list ap);\nint redisAsyncCommand(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, const char *format, ...);\nint redisAsyncCommandArgv(redisAsyncContext *ac, redisCallbackFn *fn, void *privdata, int argc, const char **argv, const size_t *argvlen);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif\n" }, { "path": "deps/hiredis/dict.c", "content": "/* Hash table implementation.\n *\n * This file implements in memory hash tables with insert/del/replace/find/\n * get-random-element operations. Hash tables will auto resize if needed\n * tables of power of two in size are used, collisions are handled by\n * chaining. See the source code for more information... :)\n *\n * Copyright (c) 2006-2010, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"fmacros.h\"\n#include \n#include \n#include \n#include \"dict.h\"\n\n/* -------------------------- private prototypes ---------------------------- */\n\nstatic int _dictExpandIfNeeded(dict *ht);\nstatic unsigned long _dictNextPower(unsigned long size);\nstatic int _dictKeyIndex(dict *ht, const void *key);\nstatic int _dictInit(dict *ht, dictType *type, void *privDataPtr);\n\n/* -------------------------- hash functions -------------------------------- */\n\n/* Generic hash function (a popular one from Bernstein).\n * I tested a few and this was the best. */\nstatic unsigned int dictGenHashFunction(const unsigned char *buf, int len) {\n unsigned int hash = 5381;\n\n while (len--)\n hash = ((hash << 5) + hash) + (*buf++); /* hash * 33 + c */\n return hash;\n}\n\n/* ----------------------------- API implementation ------------------------- */\n\n/* Reset an hashtable already initialized with ht_init().\n * NOTE: This function should only called by ht_destroy(). */\nstatic void _dictReset(dict *ht) {\n ht->table = NULL;\n ht->size = 0;\n ht->sizemask = 0;\n ht->used = 0;\n}\n\n/* Create a new hash table */\nstatic dict *dictCreate(dictType *type, void *privDataPtr) {\n dict *ht = malloc(sizeof(*ht));\n _dictInit(ht,type,privDataPtr);\n return ht;\n}\n\n/* Initialize the hash table */\nstatic int _dictInit(dict *ht, dictType *type, void *privDataPtr) {\n _dictReset(ht);\n ht->type = type;\n ht->privdata = privDataPtr;\n return DICT_OK;\n}\n\n/* Expand or create the hashtable */\nstatic int dictExpand(dict *ht, unsigned long size) {\n dict n; /* the new hashtable */\n unsigned long realsize = _dictNextPower(size), i;\n\n /* the size is invalid if it is smaller than the number of\n * elements already inside the hashtable */\n if (ht->used > size)\n return DICT_ERR;\n\n _dictInit(&n, ht->type, ht->privdata);\n n.size = realsize;\n n.sizemask = realsize-1;\n n.table = calloc(realsize,sizeof(dictEntry*));\n\n /* Copy all the elements from the old to the new table:\n * note that if the old hash table is empty ht->size is zero,\n * so dictExpand just creates an hash table. */\n n.used = ht->used;\n for (i = 0; i < ht->size && ht->used > 0; i++) {\n dictEntry *he, *nextHe;\n\n if (ht->table[i] == NULL) continue;\n\n /* For each hash entry on this slot... */\n he = ht->table[i];\n while(he) {\n unsigned int h;\n\n nextHe = he->next;\n /* Get the new element index */\n h = dictHashKey(ht, he->key) & n.sizemask;\n he->next = n.table[h];\n n.table[h] = he;\n ht->used--;\n /* Pass to the next element */\n he = nextHe;\n }\n }\n assert(ht->used == 0);\n free(ht->table);\n\n /* Remap the new hashtable in the old */\n *ht = n;\n return DICT_OK;\n}\n\n/* Add an element to the target hash table */\nstatic int dictAdd(dict *ht, void *key, void *val) {\n int index;\n dictEntry *entry;\n\n /* Get the index of the new element, or -1 if\n * the element already exists. */\n if ((index = _dictKeyIndex(ht, key)) == -1)\n return DICT_ERR;\n\n /* Allocates the memory and stores key */\n entry = malloc(sizeof(*entry));\n entry->next = ht->table[index];\n ht->table[index] = entry;\n\n /* Set the hash entry fields. */\n dictSetHashKey(ht, entry, key);\n dictSetHashVal(ht, entry, val);\n ht->used++;\n return DICT_OK;\n}\n\n/* Add an element, discarding the old if the key already exists.\n * Return 1 if the key was added from scratch, 0 if there was already an\n * element with such key and dictReplace() just performed a value update\n * operation. */\nstatic int dictReplace(dict *ht, void *key, void *val) {\n dictEntry *entry, auxentry;\n\n /* Try to add the element. If the key\n * does not exists dictAdd will suceed. */\n if (dictAdd(ht, key, val) == DICT_OK)\n return 1;\n /* It already exists, get the entry */\n entry = dictFind(ht, key);\n /* Free the old value and set the new one */\n /* Set the new value and free the old one. Note that it is important\n * to do that in this order, as the value may just be exactly the same\n * as the previous one. In this context, think to reference counting,\n * you want to increment (set), and then decrement (free), and not the\n * reverse. */\n auxentry = *entry;\n dictSetHashVal(ht, entry, val);\n dictFreeEntryVal(ht, &auxentry);\n return 0;\n}\n\n/* Search and remove an element */\nstatic int dictDelete(dict *ht, const void *key) {\n unsigned int h;\n dictEntry *de, *prevde;\n\n if (ht->size == 0)\n return DICT_ERR;\n h = dictHashKey(ht, key) & ht->sizemask;\n de = ht->table[h];\n\n prevde = NULL;\n while(de) {\n if (dictCompareHashKeys(ht,key,de->key)) {\n /* Unlink the element from the list */\n if (prevde)\n prevde->next = de->next;\n else\n ht->table[h] = de->next;\n\n dictFreeEntryKey(ht,de);\n dictFreeEntryVal(ht,de);\n free(de);\n ht->used--;\n return DICT_OK;\n }\n prevde = de;\n de = de->next;\n }\n return DICT_ERR; /* not found */\n}\n\n/* Destroy an entire hash table */\nstatic int _dictClear(dict *ht) {\n unsigned long i;\n\n /* Free all the elements */\n for (i = 0; i < ht->size && ht->used > 0; i++) {\n dictEntry *he, *nextHe;\n\n if ((he = ht->table[i]) == NULL) continue;\n while(he) {\n nextHe = he->next;\n dictFreeEntryKey(ht, he);\n dictFreeEntryVal(ht, he);\n free(he);\n ht->used--;\n he = nextHe;\n }\n }\n /* Free the table and the allocated cache structure */\n free(ht->table);\n /* Re-initialize the table */\n _dictReset(ht);\n return DICT_OK; /* never fails */\n}\n\n/* Clear & Release the hash table */\nstatic void dictRelease(dict *ht) {\n _dictClear(ht);\n free(ht);\n}\n\nstatic dictEntry *dictFind(dict *ht, const void *key) {\n dictEntry *he;\n unsigned int h;\n\n if (ht->size == 0) return NULL;\n h = dictHashKey(ht, key) & ht->sizemask;\n he = ht->table[h];\n while(he) {\n if (dictCompareHashKeys(ht, key, he->key))\n return he;\n he = he->next;\n }\n return NULL;\n}\n\nstatic dictIterator *dictGetIterator(dict *ht) {\n dictIterator *iter = malloc(sizeof(*iter));\n\n iter->ht = ht;\n iter->index = -1;\n iter->entry = NULL;\n iter->nextEntry = NULL;\n return iter;\n}\n\nstatic dictEntry *dictNext(dictIterator *iter) {\n while (1) {\n if (iter->entry == NULL) {\n iter->index++;\n if (iter->index >=\n (signed)iter->ht->size) break;\n iter->entry = iter->ht->table[iter->index];\n } else {\n iter->entry = iter->nextEntry;\n }\n if (iter->entry) {\n /* We need to save the 'next' here, the iterator user\n * may delete the entry we are returning. */\n iter->nextEntry = iter->entry->next;\n return iter->entry;\n }\n }\n return NULL;\n}\n\nstatic void dictReleaseIterator(dictIterator *iter) {\n free(iter);\n}\n\n/* ------------------------- private functions ------------------------------ */\n\n/* Expand the hash table if needed */\nstatic int _dictExpandIfNeeded(dict *ht) {\n /* If the hash table is empty expand it to the intial size,\n * if the table is \"full\" dobule its size. */\n if (ht->size == 0)\n return dictExpand(ht, DICT_HT_INITIAL_SIZE);\n if (ht->used == ht->size)\n return dictExpand(ht, ht->size*2);\n return DICT_OK;\n}\n\n/* Our hash table capability is a power of two */\nstatic unsigned long _dictNextPower(unsigned long size) {\n unsigned long i = DICT_HT_INITIAL_SIZE;\n\n if (size >= LONG_MAX) return LONG_MAX;\n while(1) {\n if (i >= size)\n return i;\n i *= 2;\n }\n}\n\n/* Returns the index of a free slot that can be populated with\n * an hash entry for the given 'key'.\n * If the key already exists, -1 is returned. */\nstatic int _dictKeyIndex(dict *ht, const void *key) {\n unsigned int h;\n dictEntry *he;\n\n /* Expand the hashtable if needed */\n if (_dictExpandIfNeeded(ht) == DICT_ERR)\n return -1;\n /* Compute the key hash value */\n h = dictHashKey(ht, key) & ht->sizemask;\n /* Search if this slot does not already contain the given key */\n he = ht->table[h];\n while(he) {\n if (dictCompareHashKeys(ht, key, he->key))\n return -1;\n he = he->next;\n }\n return h;\n}\n\n" }, { "path": "deps/hiredis/dict.h", "content": "/* Hash table implementation.\n *\n * This file implements in memory hash tables with insert/del/replace/find/\n * get-random-element operations. Hash tables will auto resize if needed\n * tables of power of two in size are used, collisions are handled by\n * chaining. See the source code for more information... :)\n *\n * Copyright (c) 2006-2010, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __DICT_H\n#define __DICT_H\n\n#define DICT_OK 0\n#define DICT_ERR 1\n\n/* Unused arguments generate annoying warnings... */\n#define DICT_NOTUSED(V) ((void) V)\n\ntypedef struct dictEntry {\n void *key;\n void *val;\n struct dictEntry *next;\n} dictEntry;\n\ntypedef struct dictType {\n unsigned int (*hashFunction)(const void *key);\n void *(*keyDup)(void *privdata, const void *key);\n void *(*valDup)(void *privdata, const void *obj);\n int (*keyCompare)(void *privdata, const void *key1, const void *key2);\n void (*keyDestructor)(void *privdata, void *key);\n void (*valDestructor)(void *privdata, void *obj);\n} dictType;\n\ntypedef struct dict {\n dictEntry **table;\n dictType *type;\n unsigned long size;\n unsigned long sizemask;\n unsigned long used;\n void *privdata;\n} dict;\n\ntypedef struct dictIterator {\n dict *ht;\n int index;\n dictEntry *entry, *nextEntry;\n} dictIterator;\n\n/* This is the initial size of every hash table */\n#define DICT_HT_INITIAL_SIZE 4\n\n/* ------------------------------- Macros ------------------------------------*/\n#define dictFreeEntryVal(ht, entry) \\\n if ((ht)->type->valDestructor) \\\n (ht)->type->valDestructor((ht)->privdata, (entry)->val)\n\n#define dictSetHashVal(ht, entry, _val_) do { \\\n if ((ht)->type->valDup) \\\n entry->val = (ht)->type->valDup((ht)->privdata, _val_); \\\n else \\\n entry->val = (_val_); \\\n} while(0)\n\n#define dictFreeEntryKey(ht, entry) \\\n if ((ht)->type->keyDestructor) \\\n (ht)->type->keyDestructor((ht)->privdata, (entry)->key)\n\n#define dictSetHashKey(ht, entry, _key_) do { \\\n if ((ht)->type->keyDup) \\\n entry->key = (ht)->type->keyDup((ht)->privdata, _key_); \\\n else \\\n entry->key = (_key_); \\\n} while(0)\n\n#define dictCompareHashKeys(ht, key1, key2) \\\n (((ht)->type->keyCompare) ? \\\n (ht)->type->keyCompare((ht)->privdata, key1, key2) : \\\n (key1) == (key2))\n\n#define dictHashKey(ht, key) (ht)->type->hashFunction(key)\n\n#define dictGetEntryKey(he) ((he)->key)\n#define dictGetEntryVal(he) ((he)->val)\n#define dictSlots(ht) ((ht)->size)\n#define dictSize(ht) ((ht)->used)\n\n/* API */\nstatic unsigned int dictGenHashFunction(const unsigned char *buf, int len);\nstatic dict *dictCreate(dictType *type, void *privDataPtr);\nstatic int dictExpand(dict *ht, unsigned long size);\nstatic int dictAdd(dict *ht, void *key, void *val);\nstatic int dictReplace(dict *ht, void *key, void *val);\nstatic int dictDelete(dict *ht, const void *key);\nstatic void dictRelease(dict *ht);\nstatic dictEntry * dictFind(dict *ht, const void *key);\nstatic dictIterator *dictGetIterator(dict *ht);\nstatic dictEntry *dictNext(dictIterator *iter);\nstatic void dictReleaseIterator(dictIterator *iter);\n\n#endif /* __DICT_H */\n" }, { "path": "deps/hiredis/example-ae.c", "content": "#include \n#include \n#include \n#include \n#include \"hiredis.h\"\n#include \"async.h\"\n#include \"adapters/ae.h\"\n\n/* Put event loop in the global scope, so it can be explicitly stopped */\nstatic aeEventLoop *loop;\n\nvoid getCallback(redisAsyncContext *c, void *r, void *privdata) {\n redisReply *reply = r;\n if (reply == NULL) return;\n printf(\"argv[%s]: %s\\n\", (char*)privdata, reply->str);\n\n /* Disconnect after receiving the reply to GET */\n redisAsyncDisconnect(c);\n}\n\nvoid connectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Connected...\\n\");\n}\n\nvoid disconnectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Disconnected...\\n\");\n}\n\nint main (int argc, char **argv) {\n signal(SIGPIPE, SIG_IGN);\n\n redisAsyncContext *c = redisAsyncConnect(\"127.0.0.1\", 6379);\n if (c->err) {\n /* Let *c leak for now... */\n printf(\"Error: %s\\n\", c->errstr);\n return 1;\n }\n\n loop = aeCreateEventLoop();\n redisAeAttach(loop, c);\n redisAsyncSetConnectCallback(c,connectCallback);\n redisAsyncSetDisconnectCallback(c,disconnectCallback);\n redisAsyncCommand(c, NULL, NULL, \"SET key %b\", argv[argc-1], strlen(argv[argc-1]));\n redisAsyncCommand(c, getCallback, (char*)\"end-1\", \"GET key\");\n aeMain(loop);\n return 0;\n}\n\n" }, { "path": "deps/hiredis/example-libev.c", "content": "#include \n#include \n#include \n#include \n#include \"hiredis.h\"\n#include \"async.h\"\n#include \"adapters/libev.h\"\n\nvoid getCallback(redisAsyncContext *c, void *r, void *privdata) {\n redisReply *reply = r;\n if (reply == NULL) return;\n printf(\"argv[%s]: %s\\n\", (char*)privdata, reply->str);\n\n /* Disconnect after receiving the reply to GET */\n redisAsyncDisconnect(c);\n}\n\nvoid connectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Connected...\\n\");\n}\n\nvoid disconnectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Disconnected...\\n\");\n}\n\nint main (int argc, char **argv) {\n signal(SIGPIPE, SIG_IGN);\n\n redisAsyncContext *c = redisAsyncConnect(\"127.0.0.1\", 6379);\n if (c->err) {\n /* Let *c leak for now... */\n printf(\"Error: %s\\n\", c->errstr);\n return 1;\n }\n\n redisLibevAttach(EV_DEFAULT_ c);\n redisAsyncSetConnectCallback(c,connectCallback);\n redisAsyncSetDisconnectCallback(c,disconnectCallback);\n redisAsyncCommand(c, NULL, NULL, \"SET key %b\", argv[argc-1], strlen(argv[argc-1]));\n redisAsyncCommand(c, getCallback, (char*)\"end-1\", \"GET key\");\n ev_loop(EV_DEFAULT_ 0);\n return 0;\n}\n" }, { "path": "deps/hiredis/example-libevent.c", "content": "#include \n#include \n#include \n#include \n#include \"hiredis.h\"\n#include \"async.h\"\n#include \"adapters/libevent.h\"\n\nvoid getCallback(redisAsyncContext *c, void *r, void *privdata) {\n redisReply *reply = r;\n if (reply == NULL) return;\n printf(\"argv[%s]: %s\\n\", (char*)privdata, reply->str);\n\n /* Disconnect after receiving the reply to GET */\n redisAsyncDisconnect(c);\n}\n\nvoid connectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Connected...\\n\");\n}\n\nvoid disconnectCallback(const redisAsyncContext *c, int status) {\n if (status != REDIS_OK) {\n printf(\"Error: %s\\n\", c->errstr);\n return;\n }\n printf(\"Disconnected...\\n\");\n}\n\nint main (int argc, char **argv) {\n signal(SIGPIPE, SIG_IGN);\n struct event_base *base = event_base_new();\n\n redisAsyncContext *c = redisAsyncConnect(\"127.0.0.1\", 6379);\n if (c->err) {\n /* Let *c leak for now... */\n printf(\"Error: %s\\n\", c->errstr);\n return 1;\n }\n\n redisLibeventAttach(c,base);\n redisAsyncSetConnectCallback(c,connectCallback);\n redisAsyncSetDisconnectCallback(c,disconnectCallback);\n redisAsyncCommand(c, NULL, NULL, \"SET key %b\", argv[argc-1], strlen(argv[argc-1]));\n redisAsyncCommand(c, getCallback, (char*)\"end-1\", \"GET key\");\n event_base_dispatch(base);\n return 0;\n}\n" }, { "path": "deps/hiredis/example.c", "content": "#include \n#include \n#include \n\n#include \"hiredis.h\"\n\nint main(void) {\n unsigned int j;\n redisContext *c;\n redisReply *reply;\n\n struct timeval timeout = { 1, 500000 }; // 1.5 seconds\n c = redisConnectWithTimeout((char*)\"127.0.0.2\", 6379, timeout);\n if (c->err) {\n printf(\"Connection error: %s\\n\", c->errstr);\n exit(1);\n }\n\n /* PING server */\n reply = redisCommand(c,\"PING\");\n printf(\"PING: %s\\n\", reply->str);\n freeReplyObject(reply);\n\n /* Set a key */\n reply = redisCommand(c,\"SET %s %s\", \"foo\", \"hello world\");\n printf(\"SET: %s\\n\", reply->str);\n freeReplyObject(reply);\n\n /* Set a key using binary safe API */\n reply = redisCommand(c,\"SET %b %b\", \"bar\", 3, \"hello\", 5);\n printf(\"SET (binary API): %s\\n\", reply->str);\n freeReplyObject(reply);\n\n /* Try a GET and two INCR */\n reply = redisCommand(c,\"GET foo\");\n printf(\"GET foo: %s\\n\", reply->str);\n freeReplyObject(reply);\n\n reply = redisCommand(c,\"INCR counter\");\n printf(\"INCR counter: %lld\\n\", reply->integer);\n freeReplyObject(reply);\n /* again ... */\n reply = redisCommand(c,\"INCR counter\");\n printf(\"INCR counter: %lld\\n\", reply->integer);\n freeReplyObject(reply);\n\n /* Create a list of numbers, from 0 to 9 */\n reply = redisCommand(c,\"DEL mylist\");\n freeReplyObject(reply);\n for (j = 0; j < 10; j++) {\n char buf[64];\n\n snprintf(buf,64,\"%d\",j);\n reply = redisCommand(c,\"LPUSH mylist element-%s\", buf);\n freeReplyObject(reply);\n }\n\n /* Let's check what we have inside the list */\n reply = redisCommand(c,\"LRANGE mylist 0 -1\");\n if (reply->type == REDIS_REPLY_ARRAY) {\n for (j = 0; j < reply->elements; j++) {\n printf(\"%u) %s\\n\", j, reply->element[j]->str);\n }\n }\n freeReplyObject(reply);\n\n return 0;\n}\n" }, { "path": "deps/hiredis/fmacros.h", "content": "#ifndef __HIREDIS_FMACRO_H\n#define __HIREDIS_FMACRO_H\n\n#if !defined(_BSD_SOURCE)\n#define _BSD_SOURCE\n#endif\n\n#if defined(__sun__)\n#define _POSIX_C_SOURCE 200112L\n#elif defined(__linux__) || defined(__OpenBSD__) || defined(__NetBSD__)\n#define _XOPEN_SOURCE 600\n#else\n#define _XOPEN_SOURCE\n#endif\n\n#endif\n" }, { "path": "deps/hiredis/hiredis.c", "content": "/*\n * Copyright (c) 2009-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"fmacros.h\"\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"hiredis.h\"\n#include \"net.h\"\n#include \"sds.h\"\n\nstatic redisReply *createReplyObject(int type);\nstatic void *createStringObject(const redisReadTask *task, char *str, size_t len);\nstatic void *createArrayObject(const redisReadTask *task, int elements);\nstatic void *createIntegerObject(const redisReadTask *task, long long value);\nstatic void *createNilObject(const redisReadTask *task);\n\n/* Default set of functions to build the reply. Keep in mind that such a\n * function returning NULL is interpreted as OOM. */\nstatic redisReplyObjectFunctions defaultFunctions = {\n createStringObject,\n createArrayObject,\n createIntegerObject,\n createNilObject,\n freeReplyObject\n};\n\n/* Create a reply object */\nstatic redisReply *createReplyObject(int type) {\n redisReply *r = calloc(1,sizeof(*r));\n\n if (r == NULL)\n return NULL;\n\n r->type = type;\n return r;\n}\n\n/* Free a reply object */\nvoid freeReplyObject(void *reply) {\n redisReply *r = reply;\n size_t j;\n\n switch(r->type) {\n case REDIS_REPLY_INTEGER:\n break; /* Nothing to free */\n case REDIS_REPLY_ARRAY:\n if (r->element != NULL) {\n for (j = 0; j < r->elements; j++)\n if (r->element[j] != NULL)\n freeReplyObject(r->element[j]);\n free(r->element);\n }\n break;\n case REDIS_REPLY_ERROR:\n case REDIS_REPLY_STATUS:\n case REDIS_REPLY_STRING:\n if (r->str != NULL)\n free(r->str);\n break;\n }\n free(r);\n}\n\nstatic void *createStringObject(const redisReadTask *task, char *str, size_t len) {\n redisReply *r, *parent;\n char *buf;\n\n r = createReplyObject(task->type);\n if (r == NULL)\n return NULL;\n\n buf = malloc(len+1);\n if (buf == NULL) {\n freeReplyObject(r);\n return NULL;\n }\n\n assert(task->type == REDIS_REPLY_ERROR ||\n task->type == REDIS_REPLY_STATUS ||\n task->type == REDIS_REPLY_STRING);\n\n /* Copy string value */\n memcpy(buf,str,len);\n buf[len] = '\\0';\n r->str = buf;\n r->len = len;\n\n if (task->parent) {\n parent = task->parent->obj;\n assert(parent->type == REDIS_REPLY_ARRAY);\n parent->element[task->idx] = r;\n }\n return r;\n}\n\nstatic void *createArrayObject(const redisReadTask *task, int elements) {\n redisReply *r, *parent;\n\n r = createReplyObject(REDIS_REPLY_ARRAY);\n if (r == NULL)\n return NULL;\n\n if (elements > 0) {\n r->element = calloc(elements,sizeof(redisReply*));\n if (r->element == NULL) {\n freeReplyObject(r);\n return NULL;\n }\n }\n\n r->elements = elements;\n\n if (task->parent) {\n parent = task->parent->obj;\n assert(parent->type == REDIS_REPLY_ARRAY);\n parent->element[task->idx] = r;\n }\n return r;\n}\n\nstatic void *createIntegerObject(const redisReadTask *task, long long value) {\n redisReply *r, *parent;\n\n r = createReplyObject(REDIS_REPLY_INTEGER);\n if (r == NULL)\n return NULL;\n\n r->integer = value;\n\n if (task->parent) {\n parent = task->parent->obj;\n assert(parent->type == REDIS_REPLY_ARRAY);\n parent->element[task->idx] = r;\n }\n return r;\n}\n\nstatic void *createNilObject(const redisReadTask *task) {\n redisReply *r, *parent;\n\n r = createReplyObject(REDIS_REPLY_NIL);\n if (r == NULL)\n return NULL;\n\n if (task->parent) {\n parent = task->parent->obj;\n assert(parent->type == REDIS_REPLY_ARRAY);\n parent->element[task->idx] = r;\n }\n return r;\n}\n\nstatic void __redisReaderSetError(redisReader *r, int type, const char *str) {\n size_t len;\n\n if (r->reply != NULL && r->fn && r->fn->freeObject) {\n r->fn->freeObject(r->reply);\n r->reply = NULL;\n }\n\n /* Clear input buffer on errors. */\n if (r->buf != NULL) {\n sdsfree(r->buf);\n r->buf = NULL;\n r->pos = r->len = 0;\n }\n\n /* Reset task stack. */\n r->ridx = -1;\n\n /* Set error. */\n r->err = type;\n len = strlen(str);\n len = len < (sizeof(r->errstr)-1) ? len : (sizeof(r->errstr)-1);\n memcpy(r->errstr,str,len);\n r->errstr[len] = '\\0';\n}\n\nstatic size_t chrtos(char *buf, size_t size, char byte) {\n size_t len = 0;\n\n switch(byte) {\n case '\\\\':\n case '\"':\n len = snprintf(buf,size,\"\\\"\\\\%c\\\"\",byte);\n break;\n case '\\n': len = snprintf(buf,size,\"\\\"\\\\n\\\"\"); break;\n case '\\r': len = snprintf(buf,size,\"\\\"\\\\r\\\"\"); break;\n case '\\t': len = snprintf(buf,size,\"\\\"\\\\t\\\"\"); break;\n case '\\a': len = snprintf(buf,size,\"\\\"\\\\a\\\"\"); break;\n case '\\b': len = snprintf(buf,size,\"\\\"\\\\b\\\"\"); break;\n default:\n if (isprint(byte))\n len = snprintf(buf,size,\"\\\"%c\\\"\",byte);\n else\n len = snprintf(buf,size,\"\\\"\\\\x%02x\\\"\",(unsigned char)byte);\n break;\n }\n\n return len;\n}\n\nstatic void __redisReaderSetErrorProtocolByte(redisReader *r, char byte) {\n char cbuf[8], sbuf[128];\n\n chrtos(cbuf,sizeof(cbuf),byte);\n snprintf(sbuf,sizeof(sbuf),\n \"Protocol error, got %s as reply type byte\", cbuf);\n __redisReaderSetError(r,REDIS_ERR_PROTOCOL,sbuf);\n}\n\nstatic void __redisReaderSetErrorOOM(redisReader *r) {\n __redisReaderSetError(r,REDIS_ERR_OOM,\"Out of memory\");\n}\n\nstatic char *readBytes(redisReader *r, unsigned int bytes) {\n char *p;\n if (r->len-r->pos >= bytes) {\n p = r->buf+r->pos;\n r->pos += bytes;\n return p;\n }\n return NULL;\n}\n\n/* Find pointer to \\r\\n. */\nstatic char *seekNewline(char *s, size_t len) {\n int pos = 0;\n int _len = len-1;\n\n /* Position should be < len-1 because the character at \"pos\" should be\n * followed by a \\n. Note that strchr cannot be used because it doesn't\n * allow to search a limited length and the buffer that is being searched\n * might not have a trailing NULL character. */\n while (pos < _len) {\n while(pos < _len && s[pos] != '\\r') pos++;\n if (s[pos] != '\\r') {\n /* Not found. */\n return NULL;\n } else {\n if (s[pos+1] == '\\n') {\n /* Found. */\n return s+pos;\n } else {\n /* Continue searching. */\n pos++;\n }\n }\n }\n return NULL;\n}\n\n/* Read a long long value starting at *s, under the assumption that it will be\n * terminated by \\r\\n. Ambiguously returns -1 for unexpected input. */\nstatic long long readLongLong(char *s) {\n long long v = 0;\n int dec, mult = 1;\n char c;\n\n if (*s == '-') {\n mult = -1;\n s++;\n } else if (*s == '+') {\n mult = 1;\n s++;\n }\n\n while ((c = *(s++)) != '\\r') {\n dec = c - '0';\n if (dec >= 0 && dec < 10) {\n v *= 10;\n v += dec;\n } else {\n /* Should not happen... */\n return -1;\n }\n }\n\n return mult*v;\n}\n\nstatic char *readLine(redisReader *r, int *_len) {\n char *p, *s;\n int len;\n\n p = r->buf+r->pos;\n s = seekNewline(p,(r->len-r->pos));\n if (s != NULL) {\n len = s-(r->buf+r->pos);\n r->pos += len+2; /* skip \\r\\n */\n if (_len) *_len = len;\n return p;\n }\n return NULL;\n}\n\nstatic void moveToNextTask(redisReader *r) {\n redisReadTask *cur, *prv;\n while (r->ridx >= 0) {\n /* Return a.s.a.p. when the stack is now empty. */\n if (r->ridx == 0) {\n r->ridx--;\n return;\n }\n\n cur = &(r->rstack[r->ridx]);\n prv = &(r->rstack[r->ridx-1]);\n assert(prv->type == REDIS_REPLY_ARRAY);\n if (cur->idx == prv->elements-1) {\n r->ridx--;\n } else {\n /* Reset the type because the next item can be anything */\n assert(cur->idx < prv->elements);\n cur->type = -1;\n cur->elements = -1;\n cur->idx++;\n return;\n }\n }\n}\n\nstatic int processLineItem(redisReader *r) {\n redisReadTask *cur = &(r->rstack[r->ridx]);\n void *obj;\n char *p;\n int len;\n\n if ((p = readLine(r,&len)) != NULL) {\n if (cur->type == REDIS_REPLY_INTEGER) {\n if (r->fn && r->fn->createInteger)\n obj = r->fn->createInteger(cur,readLongLong(p));\n else\n obj = (void*)REDIS_REPLY_INTEGER;\n } else {\n /* Type will be error or status. */\n if (r->fn && r->fn->createString)\n obj = r->fn->createString(cur,p,len);\n else\n obj = (void*)(size_t)(cur->type);\n }\n\n if (obj == NULL) {\n __redisReaderSetErrorOOM(r);\n return REDIS_ERR;\n }\n\n /* Set reply if this is the root object. */\n if (r->ridx == 0) r->reply = obj;\n moveToNextTask(r);\n return REDIS_OK;\n }\n\n return REDIS_ERR;\n}\n\nstatic int processBulkItem(redisReader *r) {\n redisReadTask *cur = &(r->rstack[r->ridx]);\n void *obj = NULL;\n char *p, *s;\n long len;\n unsigned long bytelen;\n int success = 0;\n\n p = r->buf+r->pos;\n s = seekNewline(p,r->len-r->pos);\n if (s != NULL) {\n p = r->buf+r->pos;\n bytelen = s-(r->buf+r->pos)+2; /* include \\r\\n */\n len = readLongLong(p);\n\n if (len < 0) {\n /* The nil object can always be created. */\n if (r->fn && r->fn->createNil)\n obj = r->fn->createNil(cur);\n else\n obj = (void*)REDIS_REPLY_NIL;\n success = 1;\n } else {\n /* Only continue when the buffer contains the entire bulk item. */\n bytelen += len+2; /* include \\r\\n */\n if (r->pos+bytelen <= r->len) {\n if (r->fn && r->fn->createString)\n obj = r->fn->createString(cur,s+2,len);\n else\n obj = (void*)REDIS_REPLY_STRING;\n success = 1;\n }\n }\n\n /* Proceed when obj was created. */\n if (success) {\n if (obj == NULL) {\n __redisReaderSetErrorOOM(r);\n return REDIS_ERR;\n }\n\n r->pos += bytelen;\n\n /* Set reply if this is the root object. */\n if (r->ridx == 0) r->reply = obj;\n moveToNextTask(r);\n return REDIS_OK;\n }\n }\n\n return REDIS_ERR;\n}\n\nstatic int processMultiBulkItem(redisReader *r) {\n redisReadTask *cur = &(r->rstack[r->ridx]);\n void *obj;\n char *p;\n long elements;\n int root = 0;\n\n /* Set error for nested multi bulks with depth > 7 */\n if (r->ridx == 8) {\n __redisReaderSetError(r,REDIS_ERR_PROTOCOL,\n \"No support for nested multi bulk replies with depth > 7\");\n return REDIS_ERR;\n }\n\n if ((p = readLine(r,NULL)) != NULL) {\n elements = readLongLong(p);\n root = (r->ridx == 0);\n\n if (elements == -1) {\n if (r->fn && r->fn->createNil)\n obj = r->fn->createNil(cur);\n else\n obj = (void*)REDIS_REPLY_NIL;\n\n if (obj == NULL) {\n __redisReaderSetErrorOOM(r);\n return REDIS_ERR;\n }\n\n moveToNextTask(r);\n } else {\n if (r->fn && r->fn->createArray)\n obj = r->fn->createArray(cur,elements);\n else\n obj = (void*)REDIS_REPLY_ARRAY;\n\n if (obj == NULL) {\n __redisReaderSetErrorOOM(r);\n return REDIS_ERR;\n }\n\n /* Modify task stack when there are more than 0 elements. */\n if (elements > 0) {\n cur->elements = elements;\n cur->obj = obj;\n r->ridx++;\n r->rstack[r->ridx].type = -1;\n r->rstack[r->ridx].elements = -1;\n r->rstack[r->ridx].idx = 0;\n r->rstack[r->ridx].obj = NULL;\n r->rstack[r->ridx].parent = cur;\n r->rstack[r->ridx].privdata = r->privdata;\n } else {\n moveToNextTask(r);\n }\n }\n\n /* Set reply if this is the root object. */\n if (root) r->reply = obj;\n return REDIS_OK;\n }\n\n return REDIS_ERR;\n}\n\nstatic int processItem(redisReader *r) {\n redisReadTask *cur = &(r->rstack[r->ridx]);\n char *p;\n\n /* check if we need to read type */\n if (cur->type < 0) {\n if ((p = readBytes(r,1)) != NULL) {\n switch (p[0]) {\n case '-':\n cur->type = REDIS_REPLY_ERROR;\n break;\n case '+':\n cur->type = REDIS_REPLY_STATUS;\n break;\n case ':':\n cur->type = REDIS_REPLY_INTEGER;\n break;\n case '$':\n cur->type = REDIS_REPLY_STRING;\n break;\n case '*':\n cur->type = REDIS_REPLY_ARRAY;\n break;\n default:\n __redisReaderSetErrorProtocolByte(r,*p);\n return REDIS_ERR;\n }\n } else {\n /* could not consume 1 byte */\n return REDIS_ERR;\n }\n }\n\n /* process typed item */\n switch(cur->type) {\n case REDIS_REPLY_ERROR:\n case REDIS_REPLY_STATUS:\n case REDIS_REPLY_INTEGER:\n return processLineItem(r);\n case REDIS_REPLY_STRING:\n return processBulkItem(r);\n case REDIS_REPLY_ARRAY:\n return processMultiBulkItem(r);\n default:\n assert(NULL);\n return REDIS_ERR; /* Avoid warning. */\n }\n}\n\nredisReader *redisReaderCreate(void) {\n redisReader *r;\n\n r = calloc(sizeof(redisReader),1);\n if (r == NULL)\n return NULL;\n\n r->err = 0;\n r->errstr[0] = '\\0';\n r->fn = &defaultFunctions;\n r->buf = sdsempty();\n r->maxbuf = REDIS_READER_MAX_BUF;\n if (r->buf == NULL) {\n free(r);\n return NULL;\n }\n\n r->ridx = -1;\n return r;\n}\n\nvoid redisReaderFree(redisReader *r) {\n if (r->reply != NULL && r->fn && r->fn->freeObject)\n r->fn->freeObject(r->reply);\n if (r->buf != NULL)\n sdsfree(r->buf);\n free(r);\n}\n\nint redisReaderFeed(redisReader *r, const char *buf, size_t len) {\n sds newbuf;\n\n /* Return early when this reader is in an erroneous state. */\n if (r->err)\n return REDIS_ERR;\n\n /* Copy the provided buffer. */\n if (buf != NULL && len >= 1) {\n /* Destroy internal buffer when it is empty and is quite large. */\n if (r->len == 0 && r->maxbuf != 0 && sdsavail(r->buf) > r->maxbuf) {\n sdsfree(r->buf);\n r->buf = sdsempty();\n r->pos = 0;\n\n /* r->buf should not be NULL since we just free'd a larger one. */\n assert(r->buf != NULL);\n }\n\n newbuf = sdscatlen(r->buf,buf,len);\n if (newbuf == NULL) {\n __redisReaderSetErrorOOM(r);\n return REDIS_ERR;\n }\n\n r->buf = newbuf;\n r->len = sdslen(r->buf);\n }\n\n return REDIS_OK;\n}\n\nint redisReaderGetReply(redisReader *r, void **reply) {\n /* Default target pointer to NULL. */\n if (reply != NULL)\n *reply = NULL;\n\n /* Return early when this reader is in an erroneous state. */\n if (r->err)\n return REDIS_ERR;\n\n /* When the buffer is empty, there will never be a reply. */\n if (r->len == 0)\n return REDIS_OK;\n\n /* Set first item to process when the stack is empty. */\n if (r->ridx == -1) {\n r->rstack[0].type = -1;\n r->rstack[0].elements = -1;\n r->rstack[0].idx = -1;\n r->rstack[0].obj = NULL;\n r->rstack[0].parent = NULL;\n r->rstack[0].privdata = r->privdata;\n r->ridx = 0;\n }\n\n /* Process items in reply. */\n while (r->ridx >= 0)\n if (processItem(r) != REDIS_OK)\n break;\n\n /* Return ASAP when an error occurred. */\n if (r->err)\n return REDIS_ERR;\n\n /* Discard part of the buffer when we've consumed at least 1k, to avoid\n * doing unnecessary calls to memmove() in sds.c. */\n if (r->pos >= 1024) {\n sdsrange(r->buf,r->pos,-1);\n r->pos = 0;\n r->len = sdslen(r->buf);\n }\n\n /* Emit a reply when there is one. */\n if (r->ridx == -1) {\n if (reply != NULL)\n *reply = r->reply;\n r->reply = NULL;\n }\n return REDIS_OK;\n}\n\n/* Calculate the number of bytes needed to represent an integer as string. */\nstatic int intlen(int i) {\n int len = 0;\n if (i < 0) {\n len++;\n i = -i;\n }\n do {\n len++;\n i /= 10;\n } while(i);\n return len;\n}\n\n/* Helper that calculates the bulk length given a certain string length. */\nstatic size_t bulklen(size_t len) {\n return 1+intlen(len)+2+len+2;\n}\n\nint redisvFormatCommand(char **target, const char *format, va_list ap) {\n const char *c = format;\n char *cmd = NULL; /* final command */\n int pos; /* position in final command */\n sds curarg, newarg; /* current argument */\n int touched = 0; /* was the current argument touched? */\n char **curargv = NULL, **newargv = NULL;\n int argc = 0;\n int totlen = 0;\n int j;\n\n /* Abort if there is not target to set */\n if (target == NULL)\n return -1;\n\n /* Build the command string accordingly to protocol */\n curarg = sdsempty();\n if (curarg == NULL)\n return -1;\n\n while(*c != '\\0') {\n if (*c != '%' || c[1] == '\\0') {\n if (*c == ' ') {\n if (touched) {\n newargv = realloc(curargv,sizeof(char*)*(argc+1));\n if (newargv == NULL) goto err;\n curargv = newargv;\n curargv[argc++] = curarg;\n totlen += bulklen(sdslen(curarg));\n\n /* curarg is put in argv so it can be overwritten. */\n curarg = sdsempty();\n if (curarg == NULL) goto err;\n touched = 0;\n }\n } else {\n newarg = sdscatlen(curarg,c,1);\n if (newarg == NULL) goto err;\n curarg = newarg;\n touched = 1;\n }\n } else {\n char *arg;\n size_t size;\n\n /* Set newarg so it can be checked even if it is not touched. */\n newarg = curarg;\n\n switch(c[1]) {\n case 's':\n arg = va_arg(ap,char*);\n size = strlen(arg);\n if (size > 0)\n newarg = sdscatlen(curarg,arg,size);\n break;\n case 'b':\n arg = va_arg(ap,char*);\n size = va_arg(ap,size_t);\n if (size > 0)\n newarg = sdscatlen(curarg,arg,size);\n break;\n case '%':\n newarg = sdscat(curarg,\"%\");\n break;\n default:\n /* Try to detect printf format */\n {\n static const char intfmts[] = \"diouxX\";\n char _format[16];\n const char *_p = c+1;\n size_t _l = 0;\n va_list _cpy;\n\n /* Flags */\n if (*_p != '\\0' && *_p == '#') _p++;\n if (*_p != '\\0' && *_p == '0') _p++;\n if (*_p != '\\0' && *_p == '-') _p++;\n if (*_p != '\\0' && *_p == ' ') _p++;\n if (*_p != '\\0' && *_p == '+') _p++;\n\n /* Field width */\n while (*_p != '\\0' && isdigit(*_p)) _p++;\n\n /* Precision */\n if (*_p == '.') {\n _p++;\n while (*_p != '\\0' && isdigit(*_p)) _p++;\n }\n\n /* Copy va_list before consuming with va_arg */\n va_copy(_cpy,ap);\n\n /* Integer conversion (without modifiers) */\n if (strchr(intfmts,*_p) != NULL) {\n va_arg(ap,int);\n goto fmt_valid;\n }\n\n /* Double conversion (without modifiers) */\n if (strchr(\"eEfFgGaA\",*_p) != NULL) {\n va_arg(ap,double);\n goto fmt_valid;\n }\n\n /* Size: char */\n if (_p[0] == 'h' && _p[1] == 'h') {\n _p += 2;\n if (*_p != '\\0' && strchr(intfmts,*_p) != NULL) {\n va_arg(ap,int); /* char gets promoted to int */\n goto fmt_valid;\n }\n goto fmt_invalid;\n }\n\n /* Size: short */\n if (_p[0] == 'h') {\n _p += 1;\n if (*_p != '\\0' && strchr(intfmts,*_p) != NULL) {\n va_arg(ap,int); /* short gets promoted to int */\n goto fmt_valid;\n }\n goto fmt_invalid;\n }\n\n /* Size: long long */\n if (_p[0] == 'l' && _p[1] == 'l') {\n _p += 2;\n if (*_p != '\\0' && strchr(intfmts,*_p) != NULL) {\n va_arg(ap,long long);\n goto fmt_valid;\n }\n goto fmt_invalid;\n }\n\n /* Size: long */\n if (_p[0] == 'l') {\n _p += 1;\n if (*_p != '\\0' && strchr(intfmts,*_p) != NULL) {\n va_arg(ap,long);\n goto fmt_valid;\n }\n goto fmt_invalid;\n }\n\n fmt_invalid:\n va_end(_cpy);\n goto err;\n\n fmt_valid:\n _l = (_p+1)-c;\n if (_l < sizeof(_format)-2) {\n memcpy(_format,c,_l);\n _format[_l] = '\\0';\n newarg = sdscatvprintf(curarg,_format,_cpy);\n\n /* Update current position (note: outer blocks\n * increment c twice so compensate here) */\n c = _p-1;\n }\n\n va_end(_cpy);\n break;\n }\n }\n\n if (newarg == NULL) goto err;\n curarg = newarg;\n\n touched = 1;\n c++;\n }\n c++;\n }\n\n /* Add the last argument if needed */\n if (touched) {\n newargv = realloc(curargv,sizeof(char*)*(argc+1));\n if (newargv == NULL) goto err;\n curargv = newargv;\n curargv[argc++] = curarg;\n totlen += bulklen(sdslen(curarg));\n } else {\n sdsfree(curarg);\n }\n\n /* Clear curarg because it was put in curargv or was free'd. */\n curarg = NULL;\n\n /* Add bytes needed to hold multi bulk count */\n totlen += 1+intlen(argc)+2;\n\n /* Build the command at protocol level */\n cmd = malloc(totlen+1);\n if (cmd == NULL) goto err;\n\n pos = sprintf(cmd,\"*%d\\r\\n\",argc);\n for (j = 0; j < argc; j++) {\n pos += sprintf(cmd+pos,\"$%zu\\r\\n\",sdslen(curargv[j]));\n memcpy(cmd+pos,curargv[j],sdslen(curargv[j]));\n pos += sdslen(curargv[j]);\n sdsfree(curargv[j]);\n cmd[pos++] = '\\r';\n cmd[pos++] = '\\n';\n }\n assert(pos == totlen);\n cmd[pos] = '\\0';\n\n free(curargv);\n *target = cmd;\n return totlen;\n\nerr:\n while(argc--)\n sdsfree(curargv[argc]);\n free(curargv);\n\n if (curarg != NULL)\n sdsfree(curarg);\n\n /* No need to check cmd since it is the last statement that can fail,\n * but do it anyway to be as defensive as possible. */\n if (cmd != NULL)\n free(cmd);\n\n return -1;\n}\n\n/* Format a command according to the Redis protocol. This function\n * takes a format similar to printf:\n *\n * %s represents a C null terminated string you want to interpolate\n * %b represents a binary safe string\n *\n * When using %b you need to provide both the pointer to the string\n * and the length in bytes. Examples:\n *\n * len = redisFormatCommand(target, \"GET %s\", mykey);\n * len = redisFormatCommand(target, \"SET %s %b\", mykey, myval, myvallen);\n */\nint redisFormatCommand(char **target, const char *format, ...) {\n va_list ap;\n int len;\n va_start(ap,format);\n len = redisvFormatCommand(target,format,ap);\n va_end(ap);\n return len;\n}\n\n/* Format a command according to the Redis protocol. This function takes the\n * number of arguments, an array with arguments and an array with their\n * lengths. If the latter is set to NULL, strlen will be used to compute the\n * argument lengths.\n */\nint redisFormatCommandArgv(char **target, int argc, const char **argv, const size_t *argvlen) {\n char *cmd = NULL; /* final command */\n int pos; /* position in final command */\n size_t len;\n int totlen, j;\n\n /* Calculate number of bytes needed for the command */\n totlen = 1+intlen(argc)+2;\n for (j = 0; j < argc; j++) {\n len = argvlen ? argvlen[j] : strlen(argv[j]);\n totlen += bulklen(len);\n }\n\n /* Build the command at protocol level */\n cmd = malloc(totlen+1);\n if (cmd == NULL)\n return -1;\n\n pos = sprintf(cmd,\"*%d\\r\\n\",argc);\n for (j = 0; j < argc; j++) {\n len = argvlen ? argvlen[j] : strlen(argv[j]);\n pos += sprintf(cmd+pos,\"$%zu\\r\\n\",len);\n memcpy(cmd+pos,argv[j],len);\n pos += len;\n cmd[pos++] = '\\r';\n cmd[pos++] = '\\n';\n }\n assert(pos == totlen);\n cmd[pos] = '\\0';\n\n *target = cmd;\n return totlen;\n}\n\nvoid __redisSetError(redisContext *c, int type, const char *str) {\n size_t len;\n\n c->err = type;\n if (str != NULL) {\n len = strlen(str);\n len = len < (sizeof(c->errstr)-1) ? len : (sizeof(c->errstr)-1);\n memcpy(c->errstr,str,len);\n c->errstr[len] = '\\0';\n } else {\n /* Only REDIS_ERR_IO may lack a description! */\n assert(type == REDIS_ERR_IO);\n strerror_r(errno,c->errstr,sizeof(c->errstr));\n }\n}\n\nstatic redisContext *redisContextInit(void) {\n redisContext *c;\n\n c = calloc(1,sizeof(redisContext));\n if (c == NULL)\n return NULL;\n\n c->err = 0;\n c->errstr[0] = '\\0';\n c->obuf = sdsempty();\n c->reader = redisReaderCreate();\n return c;\n}\n\nvoid redisFree(redisContext *c) {\n if (c->fd > 0)\n close(c->fd);\n if (c->obuf != NULL)\n sdsfree(c->obuf);\n if (c->reader != NULL)\n redisReaderFree(c->reader);\n free(c);\n}\n\n/* Connect to a Redis instance. On error the field error in the returned\n * context will be set to the return value of the error function.\n * When no set of reply functions is given, the default set will be used. */\nredisContext *redisConnect(const char *ip, int port) {\n redisContext *c = redisContextInit();\n c->flags |= REDIS_BLOCK;\n redisContextConnectTcp(c,ip,port,NULL);\n return c;\n}\n\nredisContext *redisConnectWithTimeout(const char *ip, int port, struct timeval tv) {\n redisContext *c = redisContextInit();\n c->flags |= REDIS_BLOCK;\n redisContextConnectTcp(c,ip,port,&tv);\n return c;\n}\n\nredisContext *redisConnectNonBlock(const char *ip, int port) {\n redisContext *c = redisContextInit();\n c->flags &= ~REDIS_BLOCK;\n redisContextConnectTcp(c,ip,port,NULL);\n return c;\n}\n\nredisContext *redisConnectUnix(const char *path) {\n redisContext *c = redisContextInit();\n c->flags |= REDIS_BLOCK;\n redisContextConnectUnix(c,path,NULL);\n return c;\n}\n\nredisContext *redisConnectUnixWithTimeout(const char *path, struct timeval tv) {\n redisContext *c = redisContextInit();\n c->flags |= REDIS_BLOCK;\n redisContextConnectUnix(c,path,&tv);\n return c;\n}\n\nredisContext *redisConnectUnixNonBlock(const char *path) {\n redisContext *c = redisContextInit();\n c->flags &= ~REDIS_BLOCK;\n redisContextConnectUnix(c,path,NULL);\n return c;\n}\n\n/* Set read/write timeout on a blocking socket. */\nint redisSetTimeout(redisContext *c, struct timeval tv) {\n if (c->flags & REDIS_BLOCK)\n return redisContextSetTimeout(c,tv);\n return REDIS_ERR;\n}\n\n/* Use this function to handle a read event on the descriptor. It will try\n * and read some bytes from the socket and feed them to the reply parser.\n *\n * After this function is called, you may use redisContextReadReply to\n * see if there is a reply available. */\nint redisBufferRead(redisContext *c) {\n char buf[1024*16];\n int nread;\n\n /* Return early when the context has seen an error. */\n if (c->err)\n return REDIS_ERR;\n\n nread = read(c->fd,buf,sizeof(buf));\n if (nread == -1) {\n if (errno == EAGAIN && !(c->flags & REDIS_BLOCK)) {\n /* Try again later */\n } else {\n __redisSetError(c,REDIS_ERR_IO,NULL);\n return REDIS_ERR;\n }\n } else if (nread == 0) {\n __redisSetError(c,REDIS_ERR_EOF,\"Server closed the connection\");\n return REDIS_ERR;\n } else {\n if (redisReaderFeed(c->reader,buf,nread) != REDIS_OK) {\n __redisSetError(c,c->reader->err,c->reader->errstr);\n return REDIS_ERR;\n }\n }\n return REDIS_OK;\n}\n\n/* Write the output buffer to the socket.\n *\n * Returns REDIS_OK when the buffer is empty, or (a part of) the buffer was\n * succesfully written to the socket. When the buffer is empty after the\n * write operation, \"done\" is set to 1 (if given).\n *\n * Returns REDIS_ERR if an error occured trying to write and sets\n * c->errstr to hold the appropriate error string.\n */\nint redisBufferWrite(redisContext *c, int *done) {\n int nwritten;\n\n /* Return early when the context has seen an error. */\n if (c->err)\n return REDIS_ERR;\n\n if (sdslen(c->obuf) > 0) {\n nwritten = write(c->fd,c->obuf,sdslen(c->obuf));\n if (nwritten == -1) {\n if (errno == EAGAIN && !(c->flags & REDIS_BLOCK)) {\n /* Try again later */\n } else {\n __redisSetError(c,REDIS_ERR_IO,NULL);\n return REDIS_ERR;\n }\n } else if (nwritten > 0) {\n if (nwritten == (signed)sdslen(c->obuf)) {\n sdsfree(c->obuf);\n c->obuf = sdsempty();\n } else {\n sdsrange(c->obuf,nwritten,-1);\n }\n }\n }\n if (done != NULL) *done = (sdslen(c->obuf) == 0);\n return REDIS_OK;\n}\n\n/* Internal helper function to try and get a reply from the reader,\n * or set an error in the context otherwise. */\nint redisGetReplyFromReader(redisContext *c, void **reply) {\n if (redisReaderGetReply(c->reader,reply) == REDIS_ERR) {\n __redisSetError(c,c->reader->err,c->reader->errstr);\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\nint redisGetReply(redisContext *c, void **reply) {\n int wdone = 0;\n void *aux = NULL;\n\n /* Try to read pending replies */\n if (redisGetReplyFromReader(c,&aux) == REDIS_ERR)\n return REDIS_ERR;\n\n /* For the blocking context, flush output buffer and read reply */\n if (aux == NULL && c->flags & REDIS_BLOCK) {\n /* Write until done */\n do {\n if (redisBufferWrite(c,&wdone) == REDIS_ERR)\n return REDIS_ERR;\n } while (!wdone);\n\n /* Read until there is a reply */\n do {\n if (redisBufferRead(c) == REDIS_ERR)\n return REDIS_ERR;\n if (redisGetReplyFromReader(c,&aux) == REDIS_ERR)\n return REDIS_ERR;\n } while (aux == NULL);\n }\n\n /* Set reply object */\n if (reply != NULL) *reply = aux;\n return REDIS_OK;\n}\n\n\n/* Helper function for the redisAppendCommand* family of functions.\n *\n * Write a formatted command to the output buffer. When this family\n * is used, you need to call redisGetReply yourself to retrieve\n * the reply (or replies in pub/sub).\n */\nint __redisAppendCommand(redisContext *c, char *cmd, size_t len) {\n sds newbuf;\n\n newbuf = sdscatlen(c->obuf,cmd,len);\n if (newbuf == NULL) {\n __redisSetError(c,REDIS_ERR_OOM,\"Out of memory\");\n return REDIS_ERR;\n }\n\n c->obuf = newbuf;\n return REDIS_OK;\n}\n\nint redisvAppendCommand(redisContext *c, const char *format, va_list ap) {\n char *cmd;\n int len;\n\n len = redisvFormatCommand(&cmd,format,ap);\n if (len == -1) {\n __redisSetError(c,REDIS_ERR_OOM,\"Out of memory\");\n return REDIS_ERR;\n }\n\n if (__redisAppendCommand(c,cmd,len) != REDIS_OK) {\n free(cmd);\n return REDIS_ERR;\n }\n\n free(cmd);\n return REDIS_OK;\n}\n\nint redisAppendCommand(redisContext *c, const char *format, ...) {\n va_list ap;\n int ret;\n\n va_start(ap,format);\n ret = redisvAppendCommand(c,format,ap);\n va_end(ap);\n return ret;\n}\n\nint redisAppendCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen) {\n char *cmd;\n int len;\n\n len = redisFormatCommandArgv(&cmd,argc,argv,argvlen);\n if (len == -1) {\n __redisSetError(c,REDIS_ERR_OOM,\"Out of memory\");\n return REDIS_ERR;\n }\n\n if (__redisAppendCommand(c,cmd,len) != REDIS_OK) {\n free(cmd);\n return REDIS_ERR;\n }\n\n free(cmd);\n return REDIS_OK;\n}\n\n/* Helper function for the redisCommand* family of functions.\n *\n * Write a formatted command to the output buffer. If the given context is\n * blocking, immediately read the reply into the \"reply\" pointer. When the\n * context is non-blocking, the \"reply\" pointer will not be used and the\n * command is simply appended to the write buffer.\n *\n * Returns the reply when a reply was succesfully retrieved. Returns NULL\n * otherwise. When NULL is returned in a blocking context, the error field\n * in the context will be set.\n */\nstatic void *__redisBlockForReply(redisContext *c) {\n void *reply;\n\n if (c->flags & REDIS_BLOCK) {\n if (redisGetReply(c,&reply) != REDIS_OK)\n return NULL;\n return reply;\n }\n return NULL;\n}\n\nvoid *redisvCommand(redisContext *c, const char *format, va_list ap) {\n if (redisvAppendCommand(c,format,ap) != REDIS_OK)\n return NULL;\n return __redisBlockForReply(c);\n}\n\nvoid *redisCommand(redisContext *c, const char *format, ...) {\n va_list ap;\n void *reply = NULL;\n va_start(ap,format);\n reply = redisvCommand(c,format,ap);\n va_end(ap);\n return reply;\n}\n\nvoid *redisCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen) {\n if (redisAppendCommandArgv(c,argc,argv,argvlen) != REDIS_OK)\n return NULL;\n return __redisBlockForReply(c);\n}\n" }, { "path": "deps/hiredis/hiredis.h", "content": "/*\n * Copyright (c) 2009-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __HIREDIS_H\n#define __HIREDIS_H\n#include /* for size_t */\n#include /* for va_list */\n#include /* for struct timeval */\n\n#define HIREDIS_MAJOR 0\n#define HIREDIS_MINOR 10\n#define HIREDIS_PATCH 1\n\n#define REDIS_ERR -1\n#define REDIS_OK 0\n\n/* When an error occurs, the err flag in a context is set to hold the type of\n * error that occured. REDIS_ERR_IO means there was an I/O error and you\n * should use the \"errno\" variable to find out what is wrong.\n * For other values, the \"errstr\" field will hold a description. */\n#define REDIS_ERR_IO 1 /* Error in read or write */\n#define REDIS_ERR_EOF 3 /* End of file */\n#define REDIS_ERR_PROTOCOL 4 /* Protocol error */\n#define REDIS_ERR_OOM 5 /* Out of memory */\n#define REDIS_ERR_OTHER 2 /* Everything else... */\n\n/* Connection type can be blocking or non-blocking and is set in the\n * least significant bit of the flags field in redisContext. */\n#define REDIS_BLOCK 0x1\n\n/* Connection may be disconnected before being free'd. The second bit\n * in the flags field is set when the context is connected. */\n#define REDIS_CONNECTED 0x2\n\n/* The async API might try to disconnect cleanly and flush the output\n * buffer and read all subsequent replies before disconnecting.\n * This flag means no new commands can come in and the connection\n * should be terminated once all replies have been read. */\n#define REDIS_DISCONNECTING 0x4\n\n/* Flag specific to the async API which means that the context should be clean\n * up as soon as possible. */\n#define REDIS_FREEING 0x8\n\n/* Flag that is set when an async callback is executed. */\n#define REDIS_IN_CALLBACK 0x10\n\n/* Flag that is set when the async context has one or more subscriptions. */\n#define REDIS_SUBSCRIBED 0x20\n\n/* Flag that is set when monitor mode is active */\n#define REDIS_MONITORING 0x40\n\n#define REDIS_REPLY_STRING 1\n#define REDIS_REPLY_ARRAY 2\n#define REDIS_REPLY_INTEGER 3\n#define REDIS_REPLY_NIL 4\n#define REDIS_REPLY_STATUS 5\n#define REDIS_REPLY_ERROR 6\n\n#define REDIS_READER_MAX_BUF (1024*16) /* Default max unused reader buffer. */\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n/* This is the reply object returned by redisCommand() */\ntypedef struct redisReply {\n int type; /* REDIS_REPLY_* */\n long long integer; /* The integer when type is REDIS_REPLY_INTEGER */\n int len; /* Length of string */\n char *str; /* Used for both REDIS_REPLY_ERROR and REDIS_REPLY_STRING */\n size_t elements; /* number of elements, for REDIS_REPLY_ARRAY */\n struct redisReply **element; /* elements vector for REDIS_REPLY_ARRAY */\n} redisReply;\n\ntypedef struct redisReadTask {\n int type;\n int elements; /* number of elements in multibulk container */\n int idx; /* index in parent (array) object */\n void *obj; /* holds user-generated value for a read task */\n struct redisReadTask *parent; /* parent task */\n void *privdata; /* user-settable arbitrary field */\n} redisReadTask;\n\ntypedef struct redisReplyObjectFunctions {\n void *(*createString)(const redisReadTask*, char*, size_t);\n void *(*createArray)(const redisReadTask*, int);\n void *(*createInteger)(const redisReadTask*, long long);\n void *(*createNil)(const redisReadTask*);\n void (*freeObject)(void*);\n} redisReplyObjectFunctions;\n\n/* State for the protocol parser */\ntypedef struct redisReader {\n int err; /* Error flags, 0 when there is no error */\n char errstr[128]; /* String representation of error when applicable */\n\n char *buf; /* Read buffer */\n size_t pos; /* Buffer cursor */\n size_t len; /* Buffer length */\n size_t maxbuf; /* Max length of unused buffer */\n\n redisReadTask rstack[9];\n int ridx; /* Index of current read task */\n void *reply; /* Temporary reply pointer */\n\n redisReplyObjectFunctions *fn;\n void *privdata;\n} redisReader;\n\n/* Public API for the protocol parser. */\nredisReader *redisReaderCreate(void);\nvoid redisReaderFree(redisReader *r);\nint redisReaderFeed(redisReader *r, const char *buf, size_t len);\nint redisReaderGetReply(redisReader *r, void **reply);\n\n/* Backwards compatibility, can be removed on big version bump. */\n#define redisReplyReaderCreate redisReaderCreate\n#define redisReplyReaderFree redisReaderFree\n#define redisReplyReaderFeed redisReaderFeed\n#define redisReplyReaderGetReply redisReaderGetReply\n#define redisReplyReaderSetPrivdata(_r, _p) (int)(((redisReader*)(_r))->privdata = (_p))\n#define redisReplyReaderGetObject(_r) (((redisReader*)(_r))->reply)\n#define redisReplyReaderGetError(_r) (((redisReader*)(_r))->errstr)\n\n/* Function to free the reply objects hiredis returns by default. */\nvoid freeReplyObject(void *reply);\n\n/* Functions to format a command according to the protocol. */\nint redisvFormatCommand(char **target, const char *format, va_list ap);\nint redisFormatCommand(char **target, const char *format, ...);\nint redisFormatCommandArgv(char **target, int argc, const char **argv, const size_t *argvlen);\n\n/* Context for a connection to Redis */\ntypedef struct redisContext {\n int err; /* Error flags, 0 when there is no error */\n char errstr[128]; /* String representation of error when applicable */\n int fd;\n int flags;\n char *obuf; /* Write buffer */\n redisReader *reader; /* Protocol reader */\n} redisContext;\n\nredisContext *redisConnect(const char *ip, int port);\nredisContext *redisConnectWithTimeout(const char *ip, int port, struct timeval tv);\nredisContext *redisConnectNonBlock(const char *ip, int port);\nredisContext *redisConnectUnix(const char *path);\nredisContext *redisConnectUnixWithTimeout(const char *path, struct timeval tv);\nredisContext *redisConnectUnixNonBlock(const char *path);\nint redisSetTimeout(redisContext *c, struct timeval tv);\nvoid redisFree(redisContext *c);\nint redisBufferRead(redisContext *c);\nint redisBufferWrite(redisContext *c, int *done);\n\n/* In a blocking context, this function first checks if there are unconsumed\n * replies to return and returns one if so. Otherwise, it flushes the output\n * buffer to the socket and reads until it has a reply. In a non-blocking\n * context, it will return unconsumed replies until there are no more. */\nint redisGetReply(redisContext *c, void **reply);\nint redisGetReplyFromReader(redisContext *c, void **reply);\n\n/* Write a command to the output buffer. Use these functions in blocking mode\n * to get a pipeline of commands. */\nint redisvAppendCommand(redisContext *c, const char *format, va_list ap);\nint redisAppendCommand(redisContext *c, const char *format, ...);\nint redisAppendCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen);\n\n/* Issue a command to Redis. In a blocking context, it is identical to calling\n * redisAppendCommand, followed by redisGetReply. The function will return\n * NULL if there was an error in performing the request, otherwise it will\n * return the reply. In a non-blocking context, it is identical to calling\n * only redisAppendCommand and will always return NULL. */\nvoid *redisvCommand(redisContext *c, const char *format, va_list ap);\nvoid *redisCommand(redisContext *c, const char *format, ...);\nvoid *redisCommandArgv(redisContext *c, int argc, const char **argv, const size_t *argvlen);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif\n" }, { "path": "deps/hiredis/net.c", "content": "/* Extracted from anet.c to work properly with Hiredis error reporting.\n *\n * Copyright (c) 2006-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"fmacros.h\"\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"net.h\"\n#include \"sds.h\"\n\n/* Defined in hiredis.c */\nvoid __redisSetError(redisContext *c, int type, const char *str);\n\nstatic void __redisSetErrorFromErrno(redisContext *c, int type, const char *prefix) {\n char buf[128];\n size_t len = 0;\n\n if (prefix != NULL)\n len = snprintf(buf,sizeof(buf),\"%s: \",prefix);\n strerror_r(errno,buf+len,sizeof(buf)-len);\n __redisSetError(c,type,buf);\n}\n\nstatic int redisSetReuseAddr(redisContext *c, int fd) {\n int on = 1;\n if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,NULL);\n close(fd);\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\nstatic int redisCreateSocket(redisContext *c, int type) {\n int s;\n if ((s = socket(type, SOCK_STREAM, 0)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,NULL);\n return REDIS_ERR;\n }\n if (type == AF_INET) {\n if (redisSetReuseAddr(c,s) == REDIS_ERR) {\n return REDIS_ERR;\n }\n }\n return s;\n}\n\nstatic int redisSetBlocking(redisContext *c, int fd, int blocking) {\n int flags;\n\n /* Set the socket nonblocking.\n * Note that fcntl(2) for F_GETFL and F_SETFL can't be\n * interrupted by a signal. */\n if ((flags = fcntl(fd, F_GETFL)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"fcntl(F_GETFL)\");\n close(fd);\n return REDIS_ERR;\n }\n\n if (blocking)\n flags &= ~O_NONBLOCK;\n else\n flags |= O_NONBLOCK;\n\n if (fcntl(fd, F_SETFL, flags) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"fcntl(F_SETFL)\");\n close(fd);\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\nstatic int redisSetTcpNoDelay(redisContext *c, int fd) {\n int yes = 1;\n if (setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &yes, sizeof(yes)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"setsockopt(TCP_NODELAY)\");\n close(fd);\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\n#define __MAX_MSEC (((LONG_MAX) - 999) / 1000)\n\nstatic int redisContextWaitReady(redisContext *c, int fd, const struct timeval *timeout) {\n struct pollfd wfd[1];\n long msec;\n\n msec = -1;\n wfd[0].fd = fd;\n wfd[0].events = POLLOUT;\n\n /* Only use timeout when not NULL. */\n if (timeout != NULL) {\n if (timeout->tv_usec > 1000000 || timeout->tv_sec > __MAX_MSEC) {\n close(fd);\n return REDIS_ERR;\n }\n\n msec = (timeout->tv_sec * 1000) + ((timeout->tv_usec + 999) / 1000);\n\n if (msec < 0 || msec > INT_MAX) {\n msec = INT_MAX;\n }\n }\n\n if (errno == EINPROGRESS) {\n int res;\n\n if ((res = poll(wfd, 1, msec)) == -1) {\n __redisSetErrorFromErrno(c, REDIS_ERR_IO, \"poll(2)\");\n close(fd);\n return REDIS_ERR;\n } else if (res == 0) {\n errno = ETIMEDOUT;\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,NULL);\n close(fd);\n return REDIS_ERR;\n }\n\n if (redisCheckSocketError(c, fd) != REDIS_OK)\n return REDIS_ERR;\n\n return REDIS_OK;\n }\n\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,NULL);\n close(fd);\n return REDIS_ERR;\n}\n\nint redisCheckSocketError(redisContext *c, int fd) {\n int err = 0;\n socklen_t errlen = sizeof(err);\n\n if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &err, &errlen) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"getsockopt(SO_ERROR)\");\n close(fd);\n return REDIS_ERR;\n }\n\n if (err) {\n errno = err;\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,NULL);\n close(fd);\n return REDIS_ERR;\n }\n\n return REDIS_OK;\n}\n\nint redisContextSetTimeout(redisContext *c, struct timeval tv) {\n if (setsockopt(c->fd,SOL_SOCKET,SO_RCVTIMEO,&tv,sizeof(tv)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"setsockopt(SO_RCVTIMEO)\");\n return REDIS_ERR;\n }\n if (setsockopt(c->fd,SOL_SOCKET,SO_SNDTIMEO,&tv,sizeof(tv)) == -1) {\n __redisSetErrorFromErrno(c,REDIS_ERR_IO,\"setsockopt(SO_SNDTIMEO)\");\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\nint redisContextConnectTcp(redisContext *c, const char *addr, int port, struct timeval *timeout) {\n int s, rv;\n char _port[6]; /* strlen(\"65535\"); */\n struct addrinfo hints, *servinfo, *p;\n int blocking = (c->flags & REDIS_BLOCK);\n\n snprintf(_port, 6, \"%d\", port);\n memset(&hints,0,sizeof(hints));\n hints.ai_family = AF_INET;\n hints.ai_socktype = SOCK_STREAM;\n\n /* Try with IPv6 if no IPv4 address was found. We do it in this order since\n * in a Redis client you can't afford to test if you have IPv6 connectivity\n * as this would add latency to every connect. Otherwise a more sensible\n * route could be: Use IPv6 if both addresses are available and there is IPv6\n * connectivity. */\n if ((rv = getaddrinfo(addr,_port,&hints,&servinfo)) != 0) {\n hints.ai_family = AF_INET6;\n if ((rv = getaddrinfo(addr,_port,&hints,&servinfo)) != 0) {\n __redisSetError(c,REDIS_ERR_OTHER,gai_strerror(rv));\n return REDIS_ERR;\n }\n }\n for (p = servinfo; p != NULL; p = p->ai_next) {\n if ((s = socket(p->ai_family,p->ai_socktype,p->ai_protocol)) == -1)\n continue;\n\n if (redisSetBlocking(c,s,0) != REDIS_OK)\n goto error;\n if (connect(s,p->ai_addr,p->ai_addrlen) == -1) {\n if (errno == EHOSTUNREACH) {\n close(s);\n continue;\n } else if (errno == EINPROGRESS && !blocking) {\n /* This is ok. */\n } else {\n if (redisContextWaitReady(c,s,timeout) != REDIS_OK)\n goto error;\n }\n }\n if (blocking && redisSetBlocking(c,s,1) != REDIS_OK)\n goto error;\n if (redisSetTcpNoDelay(c,s) != REDIS_OK)\n goto error;\n\n c->fd = s;\n c->flags |= REDIS_CONNECTED;\n rv = REDIS_OK;\n goto end;\n }\n if (p == NULL) {\n char buf[128];\n snprintf(buf,sizeof(buf),\"Can't create socket: %s\",strerror(errno));\n __redisSetError(c,REDIS_ERR_OTHER,buf);\n goto error;\n }\n\nerror:\n rv = REDIS_ERR;\nend:\n freeaddrinfo(servinfo);\n return rv; // Need to return REDIS_OK if alright\n}\n\nint redisContextConnectUnix(redisContext *c, const char *path, struct timeval *timeout) {\n int s;\n int blocking = (c->flags & REDIS_BLOCK);\n struct sockaddr_un sa;\n\n if ((s = redisCreateSocket(c,AF_LOCAL)) < 0)\n return REDIS_ERR;\n if (redisSetBlocking(c,s,0) != REDIS_OK)\n return REDIS_ERR;\n\n sa.sun_family = AF_LOCAL;\n strncpy(sa.sun_path,path,sizeof(sa.sun_path)-1);\n if (connect(s, (struct sockaddr*)&sa, sizeof(sa)) == -1) {\n if (errno == EINPROGRESS && !blocking) {\n /* This is ok. */\n } else {\n if (redisContextWaitReady(c,s,timeout) != REDIS_OK)\n return REDIS_ERR;\n }\n }\n\n /* Reset socket to be blocking after connect(2). */\n if (blocking && redisSetBlocking(c,s,1) != REDIS_OK)\n return REDIS_ERR;\n\n c->fd = s;\n c->flags |= REDIS_CONNECTED;\n return REDIS_OK;\n}\n" }, { "path": "deps/hiredis/net.h", "content": "/* Extracted from anet.c to work properly with Hiredis error reporting.\n *\n * Copyright (c) 2006-2011, Salvatore Sanfilippo \n * Copyright (c) 2010-2011, Pieter Noordhuis \n *\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __NET_H\n#define __NET_H\n\n#include \"hiredis.h\"\n\n#if defined(__sun)\n#define AF_LOCAL AF_UNIX\n#endif\n\nint redisCheckSocketError(redisContext *c, int fd);\nint redisContextSetTimeout(redisContext *c, struct timeval tv);\nint redisContextConnectTcp(redisContext *c, const char *addr, int port, struct timeval *timeout);\nint redisContextConnectUnix(redisContext *c, const char *path, struct timeval *timeout);\n\n#endif\n" }, { "path": "deps/hiredis/sds.c", "content": "/* SDSLib, A C dynamic strings library\n *\n * Copyright (c) 2006-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \n#include \n#include \n#include \n#include \n#include \"sds.h\"\n#include \"zmalloc.h\"\n\n/* Create a new sds string with the content specified by the 'init' pointer\n * and 'initlen'.\n * If NULL is used for 'init' the string is initialized with zero bytes.\n *\n * The string is always null-termined (all the sds strings are, always) so\n * even if you create an sds string with:\n *\n * mystring = sdsnewlen(\"abc\",3\");\n *\n * You can print the string with printf() as there is an implicit \\0 at the\n * end of the string. However the string is binary safe and can contain\n * \\0 characters in the middle, as the length is stored in the sds header. */\nsds sdsnewlen(const void *init, size_t initlen) {\n struct sdshdr *sh;\n\n if (init) {\n sh = zmalloc(sizeof(struct sdshdr)+initlen+1);\n } else {\n sh = zcalloc(sizeof(struct sdshdr)+initlen+1);\n }\n if (sh == NULL) return NULL;\n sh->len = initlen;\n sh->free = 0;\n if (initlen && init)\n memcpy(sh->buf, init, initlen);\n sh->buf[initlen] = '\\0';\n return (char*)sh->buf;\n}\n\n/* Create an empty (zero length) sds string. Even in this case the string\n * always has an implicit null term. */\nsds sdsempty(void) {\n return sdsnewlen(\"\",0);\n}\n\n/* Create a new sds string starting from a null termined C string. */\nsds sdsnew(const char *init) {\n size_t initlen = (init == NULL) ? 0 : strlen(init);\n return sdsnewlen(init, initlen);\n}\n\n/* Duplicate an sds string. */\nsds sdsdup(const sds s) {\n return sdsnewlen(s, sdslen(s));\n}\n\n/* Free an sds string. No operation is performed if 's' is NULL. */\nvoid sdsfree(sds s) {\n if (s == NULL) return;\n zfree(s-sizeof(struct sdshdr));\n}\n\n/* Set the sds string length to the length as obtained with strlen(), so\n * considering as content only up to the first null term character.\n *\n * This function is useful when the sds string is hacked manually in some\n * way, like in the following example:\n *\n * s = sdsnew(\"foobar\");\n * s[2] = '\\0';\n * sdsupdatelen(s);\n * printf(\"%d\\n\", sdslen(s));\n *\n * The output will be \"2\", but if we comment out the call to sdsupdatelen()\n * the output will be \"6\" as the string was modified but the logical length\n * remains 6 bytes. */\nvoid sdsupdatelen(sds s) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n int reallen = strlen(s);\n sh->free += (sh->len-reallen);\n sh->len = reallen;\n}\n\n/* Modify an sds string on-place to make it empty (zero length).\n * However all the existing buffer is not discarded but set as free space\n * so that next append operations will not require allocations up to the\n * number of bytes previously available. */\nvoid sdsclear(sds s) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n sh->free += sh->len;\n sh->len = 0;\n sh->buf[0] = '\\0';\n}\n\n/* Enlarge the free space at the end of the sds string so that the caller\n * is sure that after calling this function can overwrite up to addlen\n * bytes after the end of the string, plus one more byte for nul term.\n * \n * Note: this does not change the *length* of the sds string as returned\n * by sdslen(), but only the free buffer space we have. */\nsds sdsMakeRoomFor(sds s, size_t addlen) {\n struct sdshdr *sh, *newsh;\n size_t free = sdsavail(s);\n size_t len, newlen;\n\n if (free >= addlen) return s;\n len = sdslen(s);\n sh = (void*) (s-(sizeof(struct sdshdr)));\n newlen = (len+addlen);\n if (newlen < SDS_MAX_PREALLOC)\n newlen *= 2;\n else\n newlen += SDS_MAX_PREALLOC;\n newsh = zrealloc(sh, sizeof(struct sdshdr)+newlen+1);\n if (newsh == NULL) return NULL;\n\n newsh->free = newlen - len;\n return newsh->buf;\n}\n\n/* Reallocate the sds string so that it has no free space at the end. The\n * contained string remains not altered, but next concatenation operations\n * will require a reallocation.\n *\n * After the call, the passed sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call. */\nsds sdsRemoveFreeSpace(sds s) {\n struct sdshdr *sh;\n\n sh = (void*) (s-(sizeof(struct sdshdr)));\n sh = zrealloc(sh, sizeof(struct sdshdr)+sh->len+1);\n sh->free = 0;\n return sh->buf;\n}\n\n/* Return the total size of the allocation of the specifed sds string,\n * including:\n * 1) The sds header before the pointer.\n * 2) The string.\n * 3) The free buffer at the end if any.\n * 4) The implicit null term.\n */\nsize_t sdsAllocSize(sds s) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n\n return sizeof(*sh)+sh->len+sh->free+1;\n}\n\n/* Increment the sds length and decrements the left free space at the\n * end of the string according to 'incr'. Also set the null term\n * in the new end of the string.\n *\n * This function is used in order to fix the string length after the\n * user calls sdsMakeRoomFor(), writes something after the end of\n * the current string, and finally needs to set the new length.\n *\n * Note: it is possible to use a negative increment in order to\n * right-trim the string.\n *\n * Usage example:\n *\n * Using sdsIncrLen() and sdsMakeRoomFor() it is possible to mount the\n * following schema, to cat bytes coming from the kernel to the end of an\n * sds string without copying into an intermediate buffer:\n *\n * oldlen = sdslen(s);\n * s = sdsMakeRoomFor(s, BUFFER_SIZE);\n * nread = read(fd, s+oldlen, BUFFER_SIZE);\n * ... check for nread <= 0 and handle it ...\n * sdsIncrLen(s, nread);\n */\nvoid sdsIncrLen(sds s, int incr) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n\n assert(sh->free >= incr);\n sh->len += incr;\n sh->free -= incr;\n assert(sh->free >= 0);\n s[sh->len] = '\\0';\n}\n\n/* Grow the sds to have the specified length. Bytes that were not part of\n * the original length of the sds will be set to zero.\n *\n * if the specified length is smaller than the current length, no operation\n * is performed. */\nsds sdsgrowzero(sds s, size_t len) {\n struct sdshdr *sh = (void*)(s-(sizeof(struct sdshdr)));\n size_t totlen, curlen = sh->len;\n\n if (len <= curlen) return s;\n s = sdsMakeRoomFor(s,len-curlen);\n if (s == NULL) return NULL;\n\n /* Make sure added region doesn't contain garbage */\n sh = (void*)(s-(sizeof(struct sdshdr)));\n memset(s+curlen,0,(len-curlen+1)); /* also set trailing \\0 byte */\n totlen = sh->len+sh->free;\n sh->len = len;\n sh->free = totlen-sh->len;\n return s;\n}\n\n/* Append the specified binary-safe string pointed by 't' of 'len' bytes to the\n * end of the specified sds string 's'.\n *\n * After the call, the passed sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call. */\nsds sdscatlen(sds s, const void *t, size_t len) {\n struct sdshdr *sh;\n size_t curlen = sdslen(s);\n\n s = sdsMakeRoomFor(s,len);\n if (s == NULL) return NULL;\n sh = (void*) (s-(sizeof(struct sdshdr)));\n memcpy(s+curlen, t, len);\n sh->len = curlen+len;\n sh->free = sh->free-len;\n s[curlen+len] = '\\0';\n return s;\n}\n\n/* Append the specified null termianted C string to the sds string 's'.\n *\n * After the call, the passed sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call. */\nsds sdscat(sds s, const char *t) {\n return sdscatlen(s, t, strlen(t));\n}\n\n/* Append the specified sds 't' to the existing sds 's'.\n *\n * After the call, the modified sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call. */\nsds sdscatsds(sds s, const sds t) {\n return sdscatlen(s, t, sdslen(t));\n}\n\n/* Destructively modify the sds string 's' to hold the specified binary\n * safe string pointed by 't' of length 'len' bytes. */\nsds sdscpylen(sds s, const char *t, size_t len) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n size_t totlen = sh->free+sh->len;\n\n if (totlen < len) {\n s = sdsMakeRoomFor(s,len-sh->len);\n if (s == NULL) return NULL;\n sh = (void*) (s-(sizeof(struct sdshdr)));\n totlen = sh->free+sh->len;\n }\n memcpy(s, t, len);\n s[len] = '\\0';\n sh->len = len;\n sh->free = totlen-len;\n return s;\n}\n\n/* Like sdscpylen() but 't' must be a null-termined string so that the length\n * of the string is obtained with strlen(). */\nsds sdscpy(sds s, const char *t) {\n return sdscpylen(s, t, strlen(t));\n}\n\n/* Like sdscatpritf() but gets va_list instead of being variadic. */\nsds sdscatvprintf(sds s, const char *fmt, va_list ap) {\n va_list cpy;\n char *buf, *t;\n size_t buflen = 16;\n\n while(1) {\n buf = zmalloc(buflen);\n if (buf == NULL) return NULL;\n buf[buflen-2] = '\\0';\n va_copy(cpy,ap);\n vsnprintf(buf, buflen, fmt, cpy);\n if (buf[buflen-2] != '\\0') {\n zfree(buf);\n buflen *= 2;\n continue;\n }\n break;\n }\n t = sdscat(s, buf);\n zfree(buf);\n return t;\n}\n\n/* Append to the sds string 's' a string obtained using printf-alike format\n * specifier.\n *\n * After the call, the modified sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call.\n *\n * Example:\n *\n * s = sdsempty(\"Sum is: \");\n * s = sdscatprintf(s,\"%d+%d = %d\",a,b,a+b).\n *\n * Often you need to create a string from scratch with the printf-alike\n * format. When this is the need, just use sdsempty() as the target string:\n *\n * s = sdscatprintf(sdsempty(), \"... your format ...\", args);\n */\nsds sdscatprintf(sds s, const char *fmt, ...) {\n va_list ap;\n char *t;\n va_start(ap, fmt);\n t = sdscatvprintf(s,fmt,ap);\n va_end(ap);\n return t;\n}\n\n/* Remove the part of the string from left and from right composed just of\n * contiguous characters found in 'cset', that is a null terminted C string.\n *\n * After the call, the modified sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call.\n *\n * Example:\n *\n * s = sdsnew(\"AA...AA.a.aa.aHelloWorld :::\");\n * s = sdstrim(s,\"A. :\");\n * printf(\"%s\\n\", s);\n *\n * Output will be just \"Hello World\".\n */\nsds sdstrim(sds s, const char *cset) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n char *start, *end, *sp, *ep;\n size_t len;\n\n sp = start = s;\n ep = end = s+sdslen(s)-1;\n while(sp <= end && strchr(cset, *sp)) sp++;\n while(ep > start && strchr(cset, *ep)) ep--;\n len = (sp > ep) ? 0 : ((ep-sp)+1);\n if (sh->buf != sp) memmove(sh->buf, sp, len);\n sh->buf[len] = '\\0';\n sh->free = sh->free+(sh->len-len);\n sh->len = len;\n return s;\n}\n\n/* Turn the string into a smaller (or equal) string containing only the\n * substring specified by the 'start' and 'end' indexes.\n *\n * start and end can be negative, where -1 means the last character of the\n * string, -2 the penultimate character, and so forth.\n *\n * The interval is inclusive, so the start and end characters will be part\n * of the resulting string.\n *\n * The string is modified in-place.\n *\n * Example:\n *\n * s = sdsnew(\"Hello World\");\n * sdstrim(s,1,-1); => \"ello Worl\"\n */\nvoid sdsrange(sds s, int start, int end) {\n struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));\n size_t newlen, len = sdslen(s);\n\n if (len == 0) return;\n if (start < 0) {\n start = len+start;\n if (start < 0) start = 0;\n }\n if (end < 0) {\n end = len+end;\n if (end < 0) end = 0;\n }\n newlen = (start > end) ? 0 : (end-start)+1;\n if (newlen != 0) {\n if (start >= (signed)len) {\n newlen = 0;\n } else if (end >= (signed)len) {\n end = len-1;\n newlen = (start > end) ? 0 : (end-start)+1;\n }\n } else {\n start = 0;\n }\n if (start && newlen) memmove(sh->buf, sh->buf+start, newlen);\n sh->buf[newlen] = 0;\n sh->free = sh->free+(sh->len-newlen);\n sh->len = newlen;\n}\n\n/* Apply tolower() to every character of the sds string 's'. */\nvoid sdstolower(sds s) {\n int len = sdslen(s), j;\n\n for (j = 0; j < len; j++) s[j] = tolower(s[j]);\n}\n\n/* Apply toupper() to every character of the sds string 's'. */\nvoid sdstoupper(sds s) {\n int len = sdslen(s), j;\n\n for (j = 0; j < len; j++) s[j] = toupper(s[j]);\n}\n\n/* Compare two sds strings s1 and s2 with memcmp().\n *\n * Return value:\n *\n * 1 if s1 > s2.\n * -1 if s1 < s2.\n * 0 if s1 and s2 are exactly the same binary string.\n *\n * If two strings share exactly the same prefix, but one of the two has\n * additional characters, the longer string is considered to be greater than\n * the smaller one. */\nint sdscmp(const sds s1, const sds s2) {\n size_t l1, l2, minlen;\n int cmp;\n\n l1 = sdslen(s1);\n l2 = sdslen(s2);\n minlen = (l1 < l2) ? l1 : l2;\n cmp = memcmp(s1,s2,minlen);\n if (cmp == 0) return l1-l2;\n return cmp;\n}\n\n/* Split 's' with separator in 'sep'. An array\n * of sds strings is returned. *count will be set\n * by reference to the number of tokens returned.\n *\n * On out of memory, zero length string, zero length\n * separator, NULL is returned.\n *\n * Note that 'sep' is able to split a string using\n * a multi-character separator. For example\n * sdssplit(\"foo_-_bar\",\"_-_\"); will return two\n * elements \"foo\" and \"bar\".\n *\n * This version of the function is binary-safe but\n * requires length arguments. sdssplit() is just the\n * same function but for zero-terminated strings.\n */\nsds *sdssplitlen(const char *s, int len, const char *sep, int seplen, int *count) {\n int elements = 0, slots = 5, start = 0, j;\n sds *tokens;\n\n if (seplen < 1 || len < 0) return NULL;\n\n tokens = zmalloc(sizeof(sds)*slots);\n if (tokens == NULL) return NULL;\n\n if (len == 0) {\n *count = 0;\n return tokens;\n }\n for (j = 0; j < (len-(seplen-1)); j++) {\n /* make sure there is room for the next element and the final one */\n if (slots < elements+2) {\n sds *newtokens;\n\n slots *= 2;\n newtokens = zrealloc(tokens,sizeof(sds)*slots);\n if (newtokens == NULL) goto cleanup;\n tokens = newtokens;\n }\n /* search the separator */\n if ((seplen == 1 && *(s+j) == sep[0]) || (memcmp(s+j,sep,seplen) == 0)) {\n tokens[elements] = sdsnewlen(s+start,j-start);\n if (tokens[elements] == NULL) goto cleanup;\n elements++;\n start = j+seplen;\n j = j+seplen-1; /* skip the separator */\n }\n }\n /* Add the final element. We are sure there is room in the tokens array. */\n tokens[elements] = sdsnewlen(s+start,len-start);\n if (tokens[elements] == NULL) goto cleanup;\n elements++;\n *count = elements;\n return tokens;\n\ncleanup:\n {\n int i;\n for (i = 0; i < elements; i++) sdsfree(tokens[i]);\n zfree(tokens);\n *count = 0;\n return NULL;\n }\n}\n\n/* Free the result returned by sdssplitlen(), or do nothing if 'tokens' is NULL. */\nvoid sdsfreesplitres(sds *tokens, int count) {\n if (!tokens) return;\n while(count--)\n sdsfree(tokens[count]);\n zfree(tokens);\n}\n\n/* Create an sds string from a long long value. It is much faster than:\n *\n * sdscatprintf(sdsempty(),\"%lld\\n\", value);\n */\nsds sdsfromlonglong(long long value) {\n char buf[32], *p;\n unsigned long long v;\n\n v = (value < 0) ? -value : value;\n p = buf+31; /* point to the last character */\n do {\n *p-- = '0'+(v%10);\n v /= 10;\n } while(v);\n if (value < 0) *p-- = '-';\n p++;\n return sdsnewlen(p,32-(p-buf));\n}\n\n/* Append to the sds string \"s\" an escaped string representation where\n * all the non-printable characters (tested with isprint()) are turned into\n * escapes in the form \"\\n\\r\\a....\" or \"\\x\".\n *\n * After the call, the modified sds string is no longer valid and all the\n * references must be substituted with the new pointer returned by the call. */\nsds sdscatrepr(sds s, const char *p, size_t len) {\n s = sdscatlen(s,\"\\\"\",1);\n while(len--) {\n switch(*p) {\n case '\\\\':\n case '\"':\n s = sdscatprintf(s,\"\\\\%c\",*p);\n break;\n case '\\n': s = sdscatlen(s,\"\\\\n\",2); break;\n case '\\r': s = sdscatlen(s,\"\\\\r\",2); break;\n case '\\t': s = sdscatlen(s,\"\\\\t\",2); break;\n case '\\a': s = sdscatlen(s,\"\\\\a\",2); break;\n case '\\b': s = sdscatlen(s,\"\\\\b\",2); break;\n default:\n if (isprint(*p))\n s = sdscatprintf(s,\"%c\",*p);\n else\n s = sdscatprintf(s,\"\\\\x%02x\",(unsigned char)*p);\n break;\n }\n p++;\n }\n return sdscatlen(s,\"\\\"\",1);\n}\n\n/* Helper function for sdssplitargs() that returns non zero if 'c'\n * is a valid hex digit. */\nint is_hex_digit(char c) {\n return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') ||\n (c >= 'A' && c <= 'F');\n}\n\n/* Helper function for sdssplitargs() that converts an hex digit into an\n * integer from 0 to 15 */\nint hex_digit_to_int(char c) {\n switch(c) {\n case '0': return 0;\n case '1': return 1;\n case '2': return 2;\n case '3': return 3;\n case '4': return 4;\n case '5': return 5;\n case '6': return 6;\n case '7': return 7;\n case '8': return 8;\n case '9': return 9;\n case 'a': case 'A': return 10;\n case 'b': case 'B': return 11;\n case 'c': case 'C': return 12;\n case 'd': case 'D': return 13;\n case 'e': case 'E': return 14;\n case 'f': case 'F': return 15;\n default: return 0;\n }\n}\n\n/* Split a line into arguments, where every argument can be in the\n * following programming-language REPL-alike form:\n *\n * foo bar \"newline are supported\\n\" and \"\\xff\\x00otherstuff\"\n *\n * The number of arguments is stored into *argc, and an array\n * of sds is returned.\n *\n * The caller should free the resulting array of sds strings with\n * sdsfreesplitres().\n *\n * Note that sdscatrepr() is able to convert back a string into\n * a quoted string in the same format sdssplitargs() is able to parse.\n *\n * The function returns the allocated tokens on success, even when the\n * input string is empty, or NULL if the input contains unbalanced\n * quotes or closed quotes followed by non space characters\n * as in: \"foo\"bar or \"foo'\n */\nsds *sdssplitargs(const char *line, int *argc) {\n const char *p = line;\n char *current = NULL;\n char **vector = NULL;\n\n *argc = 0;\n while(1) {\n /* skip blanks */\n while(*p && isspace(*p)) p++;\n if (*p) {\n /* get a token */\n int inq=0; /* set to 1 if we are in \"quotes\" */\n int insq=0; /* set to 1 if we are in 'single quotes' */\n int done=0;\n\n if (current == NULL) current = sdsempty();\n while(!done) {\n if (inq) {\n if (*p == '\\\\' && *(p+1) == 'x' &&\n is_hex_digit(*(p+2)) &&\n is_hex_digit(*(p+3)))\n {\n unsigned char byte;\n\n byte = (hex_digit_to_int(*(p+2))*16)+\n hex_digit_to_int(*(p+3));\n current = sdscatlen(current,(char*)&byte,1);\n p += 3;\n } else if (*p == '\\\\' && *(p+1)) {\n char c;\n\n p++;\n switch(*p) {\n case 'n': c = '\\n'; break;\n case 'r': c = '\\r'; break;\n case 't': c = '\\t'; break;\n case 'b': c = '\\b'; break;\n case 'a': c = '\\a'; break;\n default: c = *p; break;\n }\n current = sdscatlen(current,&c,1);\n } else if (*p == '\"') {\n /* closing quote must be followed by a space or\n * nothing at all. */\n if (*(p+1) && !isspace(*(p+1))) goto err;\n done=1;\n } else if (!*p) {\n /* unterminated quotes */\n goto err;\n } else {\n current = sdscatlen(current,p,1);\n }\n } else if (insq) {\n if (*p == '\\\\' && *(p+1) == '\\'') {\n p++;\n current = sdscatlen(current,\"'\",1);\n } else if (*p == '\\'') {\n /* closing quote must be followed by a space or\n * nothing at all. */\n if (*(p+1) && !isspace(*(p+1))) goto err;\n done=1;\n } else if (!*p) {\n /* unterminated quotes */\n goto err;\n } else {\n current = sdscatlen(current,p,1);\n }\n } else {\n switch(*p) {\n case ' ':\n case '\\n':\n case '\\r':\n case '\\t':\n case '\\0':\n done=1;\n break;\n case '\"':\n inq=1;\n break;\n case '\\'':\n insq=1;\n break;\n default:\n current = sdscatlen(current,p,1);\n break;\n }\n }\n if (*p) p++;\n }\n /* add the token to the vector */\n vector = zrealloc(vector,((*argc)+1)*sizeof(char*));\n vector[*argc] = current;\n (*argc)++;\n current = NULL;\n } else {\n /* Even on empty input string return something not NULL. */\n if (vector == NULL) vector = zmalloc(sizeof(void*));\n return vector;\n }\n }\n\nerr:\n while((*argc)--)\n sdsfree(vector[*argc]);\n zfree(vector);\n if (current) sdsfree(current);\n *argc = 0;\n return NULL;\n}\n\n/* Modify the string substituting all the occurrences of the set of\n * characters specified in the 'from' string to the corresponding character\n * in the 'to' array.\n *\n * For instance: sdsmapchars(mystring, \"ho\", \"01\", 2)\n * will have the effect of turning the string \"hello\" into \"0ell1\".\n *\n * The function returns the sds string pointer, that is always the same\n * as the input pointer since no resize is needed. */\nsds sdsmapchars(sds s, const char *from, const char *to, size_t setlen) {\n size_t j, i, l = sdslen(s);\n\n for (j = 0; j < l; j++) {\n for (i = 0; i < setlen; i++) {\n if (s[j] == from[i]) {\n s[j] = to[i];\n break;\n }\n }\n }\n return s;\n}\n\n/* Join an array of C strings using the specified separator (also a C string).\n * Returns the result as an sds string. */\nsds sdsjoin(char **argv, int argc, char *sep) {\n sds join = sdsempty();\n int j;\n\n for (j = 0; j < argc; j++) {\n join = sdscat(join, argv[j]);\n if (j != argc-1) join = sdscat(join,sep);\n }\n return join;\n}\n\n#ifdef SDS_TEST_MAIN\n#include \n#include \"testhelp.h\"\n\nint main(void) {\n {\n struct sdshdr *sh;\n sds x = sdsnew(\"foo\"), y;\n\n test_cond(\"Create a string and obtain the length\",\n sdslen(x) == 3 && memcmp(x,\"foo\\0\",4) == 0)\n\n sdsfree(x);\n x = sdsnewlen(\"foo\",2);\n test_cond(\"Create a string with specified length\",\n sdslen(x) == 2 && memcmp(x,\"fo\\0\",3) == 0)\n\n x = sdscat(x,\"bar\");\n test_cond(\"Strings concatenation\",\n sdslen(x) == 5 && memcmp(x,\"fobar\\0\",6) == 0);\n\n x = sdscpy(x,\"a\");\n test_cond(\"sdscpy() against an originally longer string\",\n sdslen(x) == 1 && memcmp(x,\"a\\0\",2) == 0)\n\n x = sdscpy(x,\"xyzxxxxxxxxxxyyyyyyyyyykkkkkkkkkk\");\n test_cond(\"sdscpy() against an originally shorter string\",\n sdslen(x) == 33 &&\n memcmp(x,\"xyzxxxxxxxxxxyyyyyyyyyykkkkkkkkkk\\0\",33) == 0)\n\n sdsfree(x);\n x = sdscatprintf(sdsempty(),\"%d\",123);\n test_cond(\"sdscatprintf() seems working in the base case\",\n sdslen(x) == 3 && memcmp(x,\"123\\0\",4) ==0)\n\n sdsfree(x);\n x = sdstrim(sdsnew(\"xxciaoyyy\"),\"xy\");\n test_cond(\"sdstrim() correctly trims characters\",\n sdslen(x) == 4 && memcmp(x,\"ciao\\0\",5) == 0)\n\n y = sdsrange(sdsdup(x),1,1);\n test_cond(\"sdsrange(...,1,1)\",\n sdslen(y) == 1 && memcmp(y,\"i\\0\",2) == 0)\n\n sdsfree(y);\n y = sdsrange(sdsdup(x),1,-1);\n test_cond(\"sdsrange(...,1,-1)\",\n sdslen(y) == 3 && memcmp(y,\"iao\\0\",4) == 0)\n\n sdsfree(y);\n y = sdsrange(sdsdup(x),-2,-1);\n test_cond(\"sdsrange(...,-2,-1)\",\n sdslen(y) == 2 && memcmp(y,\"ao\\0\",3) == 0)\n\n sdsfree(y);\n y = sdsrange(sdsdup(x),2,1);\n test_cond(\"sdsrange(...,2,1)\",\n sdslen(y) == 0 && memcmp(y,\"\\0\",1) == 0)\n\n sdsfree(y);\n y = sdsrange(sdsdup(x),1,100);\n test_cond(\"sdsrange(...,1,100)\",\n sdslen(y) == 3 && memcmp(y,\"iao\\0\",4) == 0)\n\n sdsfree(y);\n y = sdsrange(sdsdup(x),100,100);\n test_cond(\"sdsrange(...,100,100)\",\n sdslen(y) == 0 && memcmp(y,\"\\0\",1) == 0)\n\n sdsfree(y);\n sdsfree(x);\n x = sdsnew(\"foo\");\n y = sdsnew(\"foa\");\n test_cond(\"sdscmp(foo,foa)\", sdscmp(x,y) > 0)\n\n sdsfree(y);\n sdsfree(x);\n x = sdsnew(\"bar\");\n y = sdsnew(\"bar\");\n test_cond(\"sdscmp(bar,bar)\", sdscmp(x,y) == 0)\n\n sdsfree(y);\n sdsfree(x);\n x = sdsnew(\"aar\");\n y = sdsnew(\"bar\");\n test_cond(\"sdscmp(bar,bar)\", sdscmp(x,y) < 0)\n\n {\n int oldfree;\n\n sdsfree(x);\n x = sdsnew(\"0\");\n sh = (void*) (x-(sizeof(struct sdshdr)));\n test_cond(\"sdsnew() free/len buffers\", sh->len == 1 && sh->free == 0);\n x = sdsMakeRoomFor(x,1);\n sh = (void*) (x-(sizeof(struct sdshdr)));\n test_cond(\"sdsMakeRoomFor()\", sh->len == 1 && sh->free > 0);\n oldfree = sh->free;\n x[1] = '1';\n sdsIncrLen(x,1);\n test_cond(\"sdsIncrLen() -- content\", x[0] == '0' && x[1] == '1');\n test_cond(\"sdsIncrLen() -- len\", sh->len == 2);\n test_cond(\"sdsIncrLen() -- free\", sh->free == oldfree-1);\n }\n }\n test_report()\n return 0;\n}\n#endif\n" }, { "path": "deps/hiredis/sds.h", "content": "/* SDSLib, A C dynamic strings library\n *\n * Copyright (c) 2006-2010, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __SDS_H\n#define __SDS_H\n\n#define SDS_MAX_PREALLOC (1024*1024)\n\n#include \n#include \n\ntypedef char *sds;\n\nstruct sdshdr {\n int len;\n int free;\n char buf[];\n};\n\nstatic inline size_t sdslen(const sds s) {\n struct sdshdr *sh = (void*)(s-(sizeof(struct sdshdr)));\n return sh->len;\n}\n\nstatic inline size_t sdsavail(const sds s) {\n struct sdshdr *sh = (void*)(s-(sizeof(struct sdshdr)));\n return sh->free;\n}\n\nsds sdsnewlen(const void *init, size_t initlen);\nsds sdsnew(const char *init);\nsds sdsempty(void);\nsize_t sdslen(const sds s);\nsds sdsdup(const sds s);\nvoid sdsfree(sds s);\nsize_t sdsavail(const sds s);\nsds sdsgrowzero(sds s, size_t len);\nsds sdscatlen(sds s, const void *t, size_t len);\nsds sdscat(sds s, const char *t);\nsds sdscatsds(sds s, const sds t);\nsds sdscpylen(sds s, const char *t, size_t len);\nsds sdscpy(sds s, const char *t);\n\nsds sdscatvprintf(sds s, const char *fmt, va_list ap);\n#ifdef __GNUC__\nsds sdscatprintf(sds s, const char *fmt, ...)\n __attribute__((format(printf, 2, 3)));\n#else\nsds sdscatprintf(sds s, const char *fmt, ...);\n#endif\n\nsds sdstrim(sds s, const char *cset);\nvoid sdsrange(sds s, int start, int end);\nvoid sdsupdatelen(sds s);\nvoid sdsclear(sds s);\nint sdscmp(const sds s1, const sds s2);\nsds *sdssplitlen(const char *s, int len, const char *sep, int seplen, int *count);\nvoid sdsfreesplitres(sds *tokens, int count);\nvoid sdstolower(sds s);\nvoid sdstoupper(sds s);\nsds sdsfromlonglong(long long value);\nsds sdscatrepr(sds s, const char *p, size_t len);\nsds *sdssplitargs(const char *line, int *argc);\nsds sdsmapchars(sds s, const char *from, const char *to, size_t setlen);\nsds sdsjoin(char **argv, int argc, char *sep);\n\n/* Low level functions exposed to the user API */\nsds sdsMakeRoomFor(sds s, size_t addlen);\nvoid sdsIncrLen(sds s, int incr);\nsds sdsRemoveFreeSpace(sds s);\nsize_t sdsAllocSize(sds s);\n\n#endif\n" }, { "path": "deps/hiredis/test.c", "content": "#include \"fmacros.h\"\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"hiredis.h\"\n\nenum connection_type {\n CONN_TCP,\n CONN_UNIX\n};\n\nstruct config {\n enum connection_type type;\n\n struct {\n const char *host;\n int port;\n } tcp;\n\n struct {\n const char *path;\n } unix;\n};\n\n/* The following lines make up our testing \"framework\" :) */\nstatic int tests = 0, fails = 0;\n#define test(_s) { printf(\"#%02d \", ++tests); printf(_s); }\n#define test_cond(_c) if(_c) printf(\"\\033[0;32mPASSED\\033[0;0m\\n\"); else {printf(\"\\033[0;31mFAILED\\033[0;0m\\n\"); fails++;}\n\nstatic long long usec(void) {\n struct timeval tv;\n gettimeofday(&tv,NULL);\n return (((long long)tv.tv_sec)*1000000)+tv.tv_usec;\n}\n\nstatic redisContext *select_database(redisContext *c) {\n redisReply *reply;\n\n /* Switch to DB 9 for testing, now that we know we can chat. */\n reply = redisCommand(c,\"SELECT 9\");\n assert(reply != NULL);\n freeReplyObject(reply);\n\n /* Make sure the DB is emtpy */\n reply = redisCommand(c,\"DBSIZE\");\n assert(reply != NULL);\n if (reply->type == REDIS_REPLY_INTEGER && reply->integer == 0) {\n /* Awesome, DB 9 is empty and we can continue. */\n freeReplyObject(reply);\n } else {\n printf(\"Database #9 is not empty, test can not continue\\n\");\n exit(1);\n }\n\n return c;\n}\n\nstatic void disconnect(redisContext *c) {\n redisReply *reply;\n\n /* Make sure we're on DB 9. */\n reply = redisCommand(c,\"SELECT 9\");\n assert(reply != NULL);\n freeReplyObject(reply);\n reply = redisCommand(c,\"FLUSHDB\");\n assert(reply != NULL);\n freeReplyObject(reply);\n\n /* Free the context as well. */\n redisFree(c);\n}\n\nstatic redisContext *connect(struct config config) {\n redisContext *c = NULL;\n\n if (config.type == CONN_TCP) {\n c = redisConnect(config.tcp.host, config.tcp.port);\n } else if (config.type == CONN_UNIX) {\n c = redisConnectUnix(config.unix.path);\n } else {\n assert(NULL);\n }\n\n if (c->err) {\n printf(\"Connection error: %s\\n\", c->errstr);\n exit(1);\n }\n\n return select_database(c);\n}\n\nstatic void test_format_commands(void) {\n char *cmd;\n int len;\n\n test(\"Format command without interpolation: \");\n len = redisFormatCommand(&cmd,\"SET foo bar\");\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$3\\r\\nbar\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(3+2));\n free(cmd);\n\n test(\"Format command with %%s string interpolation: \");\n len = redisFormatCommand(&cmd,\"SET %s %s\",\"foo\",\"bar\");\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$3\\r\\nbar\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(3+2));\n free(cmd);\n\n test(\"Format command with %%s and an empty string: \");\n len = redisFormatCommand(&cmd,\"SET %s %s\",\"foo\",\"\");\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$0\\r\\n\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(0+2));\n free(cmd);\n\n test(\"Format command with an empty string in between proper interpolations: \");\n len = redisFormatCommand(&cmd,\"SET %s %s\",\"\",\"foo\");\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$0\\r\\n\\r\\n$3\\r\\nfoo\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(0+2)+4+(3+2));\n free(cmd);\n\n test(\"Format command with %%b string interpolation: \");\n len = redisFormatCommand(&cmd,\"SET %b %b\",\"foo\",3,\"b\\0r\",3);\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$3\\r\\nb\\0r\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(3+2));\n free(cmd);\n\n test(\"Format command with %%b and an empty string: \");\n len = redisFormatCommand(&cmd,\"SET %b %b\",\"foo\",3,\"\",0);\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$0\\r\\n\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(0+2));\n free(cmd);\n\n test(\"Format command with literal %%: \");\n len = redisFormatCommand(&cmd,\"SET %% %%\");\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$1\\r\\n%\\r\\n$1\\r\\n%\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(1+2)+4+(1+2));\n free(cmd);\n\n /* Vararg width depends on the type. These tests make sure that the\n * width is correctly determined using the format and subsequent varargs\n * can correctly be interpolated. */\n#define INTEGER_WIDTH_TEST(fmt, type) do { \\\n type value = 123; \\\n test(\"Format command with printf-delegation (\" #type \"): \"); \\\n len = redisFormatCommand(&cmd,\"key:%08\" fmt \" str:%s\", value, \"hello\"); \\\n test_cond(strncmp(cmd,\"*2\\r\\n$12\\r\\nkey:00000123\\r\\n$9\\r\\nstr:hello\\r\\n\",len) == 0 && \\\n len == 4+5+(12+2)+4+(9+2)); \\\n free(cmd); \\\n} while(0)\n\n#define FLOAT_WIDTH_TEST(type) do { \\\n type value = 123.0; \\\n test(\"Format command with printf-delegation (\" #type \"): \"); \\\n len = redisFormatCommand(&cmd,\"key:%08.3f str:%s\", value, \"hello\"); \\\n test_cond(strncmp(cmd,\"*2\\r\\n$12\\r\\nkey:0123.000\\r\\n$9\\r\\nstr:hello\\r\\n\",len) == 0 && \\\n len == 4+5+(12+2)+4+(9+2)); \\\n free(cmd); \\\n} while(0)\n\n INTEGER_WIDTH_TEST(\"d\", int);\n INTEGER_WIDTH_TEST(\"hhd\", char);\n INTEGER_WIDTH_TEST(\"hd\", short);\n INTEGER_WIDTH_TEST(\"ld\", long);\n INTEGER_WIDTH_TEST(\"lld\", long long);\n INTEGER_WIDTH_TEST(\"u\", unsigned int);\n INTEGER_WIDTH_TEST(\"hhu\", unsigned char);\n INTEGER_WIDTH_TEST(\"hu\", unsigned short);\n INTEGER_WIDTH_TEST(\"lu\", unsigned long);\n INTEGER_WIDTH_TEST(\"llu\", unsigned long long);\n FLOAT_WIDTH_TEST(float);\n FLOAT_WIDTH_TEST(double);\n\n test(\"Format command with invalid printf format: \");\n len = redisFormatCommand(&cmd,\"key:%08p %b\",(void*)1234,\"foo\",3);\n test_cond(len == -1);\n\n const char *argv[3];\n argv[0] = \"SET\";\n argv[1] = \"foo\\0xxx\";\n argv[2] = \"bar\";\n size_t lens[3] = { 3, 7, 3 };\n int argc = 3;\n\n test(\"Format command by passing argc/argv without lengths: \");\n len = redisFormatCommandArgv(&cmd,argc,argv,NULL);\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nfoo\\r\\n$3\\r\\nbar\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(3+2)+4+(3+2));\n free(cmd);\n\n test(\"Format command by passing argc/argv with lengths: \");\n len = redisFormatCommandArgv(&cmd,argc,argv,lens);\n test_cond(strncmp(cmd,\"*3\\r\\n$3\\r\\nSET\\r\\n$7\\r\\nfoo\\0xxx\\r\\n$3\\r\\nbar\\r\\n\",len) == 0 &&\n len == 4+4+(3+2)+4+(7+2)+4+(3+2));\n free(cmd);\n}\n\nstatic void test_reply_reader(void) {\n redisReader *reader;\n void *reply;\n int ret;\n\n test(\"Error handling in reply parser: \");\n reader = redisReaderCreate();\n redisReaderFeed(reader,(char*)\"@foo\\r\\n\",6);\n ret = redisReaderGetReply(reader,NULL);\n test_cond(ret == REDIS_ERR &&\n strcasecmp(reader->errstr,\"Protocol error, got \\\"@\\\" as reply type byte\") == 0);\n redisReaderFree(reader);\n\n /* when the reply already contains multiple items, they must be free'd\n * on an error. valgrind will bark when this doesn't happen. */\n test(\"Memory cleanup in reply parser: \");\n reader = redisReaderCreate();\n redisReaderFeed(reader,(char*)\"*2\\r\\n\",4);\n redisReaderFeed(reader,(char*)\"$5\\r\\nhello\\r\\n\",11);\n redisReaderFeed(reader,(char*)\"@foo\\r\\n\",6);\n ret = redisReaderGetReply(reader,NULL);\n test_cond(ret == REDIS_ERR &&\n strcasecmp(reader->errstr,\"Protocol error, got \\\"@\\\" as reply type byte\") == 0);\n redisReaderFree(reader);\n\n test(\"Set error on nested multi bulks with depth > 2: \");\n reader = redisReaderCreate();\n redisReaderFeed(reader,(char*)\"*1\\r\\n\",4);\n redisReaderFeed(reader,(char*)\"*1\\r\\n\",4);\n redisReaderFeed(reader,(char*)\"*1\\r\\n\",4);\n redisReaderFeed(reader,(char*)\"*1\\r\\n\",4);\n ret = redisReaderGetReply(reader,NULL);\n test_cond(ret == REDIS_ERR &&\n strncasecmp(reader->errstr,\"No support for\",14) == 0);\n redisReaderFree(reader);\n\n test(\"Works with NULL functions for reply: \");\n reader = redisReaderCreate();\n reader->fn = NULL;\n redisReaderFeed(reader,(char*)\"+OK\\r\\n\",5);\n ret = redisReaderGetReply(reader,&reply);\n test_cond(ret == REDIS_OK && reply == (void*)REDIS_REPLY_STATUS);\n redisReaderFree(reader);\n\n test(\"Works when a single newline (\\\\r\\\\n) covers two calls to feed: \");\n reader = redisReaderCreate();\n reader->fn = NULL;\n redisReaderFeed(reader,(char*)\"+OK\\r\",4);\n ret = redisReaderGetReply(reader,&reply);\n assert(ret == REDIS_OK && reply == NULL);\n redisReaderFeed(reader,(char*)\"\\n\",1);\n ret = redisReaderGetReply(reader,&reply);\n test_cond(ret == REDIS_OK && reply == (void*)REDIS_REPLY_STATUS);\n redisReaderFree(reader);\n\n test(\"Don't reset state after protocol error: \");\n reader = redisReaderCreate();\n reader->fn = NULL;\n redisReaderFeed(reader,(char*)\"x\",1);\n ret = redisReaderGetReply(reader,&reply);\n assert(ret == REDIS_ERR);\n ret = redisReaderGetReply(reader,&reply);\n test_cond(ret == REDIS_ERR && reply == NULL);\n redisReaderFree(reader);\n\n /* Regression test for issue #45 on GitHub. */\n test(\"Don't do empty allocation for empty multi bulk: \");\n reader = redisReaderCreate();\n redisReaderFeed(reader,(char*)\"*0\\r\\n\",4);\n ret = redisReaderGetReply(reader,&reply);\n test_cond(ret == REDIS_OK &&\n ((redisReply*)reply)->type == REDIS_REPLY_ARRAY &&\n ((redisReply*)reply)->elements == 0);\n freeReplyObject(reply);\n redisReaderFree(reader);\n}\n\nstatic void test_blocking_connection_errors(void) {\n redisContext *c;\n\n test(\"Returns error when host cannot be resolved: \");\n c = redisConnect((char*)\"idontexist.local\", 6379);\n test_cond(c->err == REDIS_ERR_OTHER &&\n (strcmp(c->errstr,\"Name or service not known\") == 0 ||\n strcmp(c->errstr,\"Can't resolve: idontexist.local\") == 0));\n redisFree(c);\n\n test(\"Returns error when the port is not open: \");\n c = redisConnect((char*)\"localhost\", 1);\n test_cond(c->err == REDIS_ERR_IO &&\n strcmp(c->errstr,\"Connection refused\") == 0);\n redisFree(c);\n\n test(\"Returns error when the unix socket path doesn't accept connections: \");\n c = redisConnectUnix((char*)\"/tmp/idontexist.sock\");\n test_cond(c->err == REDIS_ERR_IO); /* Don't care about the message... */\n redisFree(c);\n}\n\nstatic void test_blocking_connection(struct config config) {\n redisContext *c;\n redisReply *reply;\n\n c = connect(config);\n\n test(\"Is able to deliver commands: \");\n reply = redisCommand(c,\"PING\");\n test_cond(reply->type == REDIS_REPLY_STATUS &&\n strcasecmp(reply->str,\"pong\") == 0)\n freeReplyObject(reply);\n\n test(\"Is a able to send commands verbatim: \");\n reply = redisCommand(c,\"SET foo bar\");\n test_cond (reply->type == REDIS_REPLY_STATUS &&\n strcasecmp(reply->str,\"ok\") == 0)\n freeReplyObject(reply);\n\n test(\"%%s String interpolation works: \");\n reply = redisCommand(c,\"SET %s %s\",\"foo\",\"hello world\");\n freeReplyObject(reply);\n reply = redisCommand(c,\"GET foo\");\n test_cond(reply->type == REDIS_REPLY_STRING &&\n strcmp(reply->str,\"hello world\") == 0);\n freeReplyObject(reply);\n\n test(\"%%b String interpolation works: \");\n reply = redisCommand(c,\"SET %b %b\",\"foo\",3,\"hello\\x00world\",11);\n freeReplyObject(reply);\n reply = redisCommand(c,\"GET foo\");\n test_cond(reply->type == REDIS_REPLY_STRING &&\n memcmp(reply->str,\"hello\\x00world\",11) == 0)\n\n test(\"Binary reply length is correct: \");\n test_cond(reply->len == 11)\n freeReplyObject(reply);\n\n test(\"Can parse nil replies: \");\n reply = redisCommand(c,\"GET nokey\");\n test_cond(reply->type == REDIS_REPLY_NIL)\n freeReplyObject(reply);\n\n /* test 7 */\n test(\"Can parse integer replies: \");\n reply = redisCommand(c,\"INCR mycounter\");\n test_cond(reply->type == REDIS_REPLY_INTEGER && reply->integer == 1)\n freeReplyObject(reply);\n\n test(\"Can parse multi bulk replies: \");\n freeReplyObject(redisCommand(c,\"LPUSH mylist foo\"));\n freeReplyObject(redisCommand(c,\"LPUSH mylist bar\"));\n reply = redisCommand(c,\"LRANGE mylist 0 -1\");\n test_cond(reply->type == REDIS_REPLY_ARRAY &&\n reply->elements == 2 &&\n !memcmp(reply->element[0]->str,\"bar\",3) &&\n !memcmp(reply->element[1]->str,\"foo\",3))\n freeReplyObject(reply);\n\n /* m/e with multi bulk reply *before* other reply.\n * specifically test ordering of reply items to parse. */\n test(\"Can handle nested multi bulk replies: \");\n freeReplyObject(redisCommand(c,\"MULTI\"));\n freeReplyObject(redisCommand(c,\"LRANGE mylist 0 -1\"));\n freeReplyObject(redisCommand(c,\"PING\"));\n reply = (redisCommand(c,\"EXEC\"));\n test_cond(reply->type == REDIS_REPLY_ARRAY &&\n reply->elements == 2 &&\n reply->element[0]->type == REDIS_REPLY_ARRAY &&\n reply->element[0]->elements == 2 &&\n !memcmp(reply->element[0]->element[0]->str,\"bar\",3) &&\n !memcmp(reply->element[0]->element[1]->str,\"foo\",3) &&\n reply->element[1]->type == REDIS_REPLY_STATUS &&\n strcasecmp(reply->element[1]->str,\"pong\") == 0);\n freeReplyObject(reply);\n\n disconnect(c);\n}\n\nstatic void test_blocking_io_errors(struct config config) {\n redisContext *c;\n redisReply *reply;\n void *_reply;\n int major, minor;\n\n /* Connect to target given by config. */\n c = connect(config);\n {\n /* Find out Redis version to determine the path for the next test */\n const char *field = \"redis_version:\";\n char *p, *eptr;\n\n reply = redisCommand(c,\"INFO\");\n p = strstr(reply->str,field);\n major = strtol(p+strlen(field),&eptr,10);\n p = eptr+1; /* char next to the first \".\" */\n minor = strtol(p,&eptr,10);\n freeReplyObject(reply);\n }\n\n test(\"Returns I/O error when the connection is lost: \");\n reply = redisCommand(c,\"QUIT\");\n if (major >= 2 && minor > 0) {\n /* > 2.0 returns OK on QUIT and read() should be issued once more\n * to know the descriptor is at EOF. */\n test_cond(strcasecmp(reply->str,\"OK\") == 0 &&\n redisGetReply(c,&_reply) == REDIS_ERR);\n freeReplyObject(reply);\n } else {\n test_cond(reply == NULL);\n }\n\n /* On 2.0, QUIT will cause the connection to be closed immediately and\n * the read(2) for the reply on QUIT will set the error to EOF.\n * On >2.0, QUIT will return with OK and another read(2) needed to be\n * issued to find out the socket was closed by the server. In both\n * conditions, the error will be set to EOF. */\n assert(c->err == REDIS_ERR_EOF &&\n strcmp(c->errstr,\"Server closed the connection\") == 0);\n redisFree(c);\n\n c = connect(config);\n test(\"Returns I/O error on socket timeout: \");\n struct timeval tv = { 0, 1000 };\n assert(redisSetTimeout(c,tv) == REDIS_OK);\n test_cond(redisGetReply(c,&_reply) == REDIS_ERR &&\n c->err == REDIS_ERR_IO && errno == EAGAIN);\n redisFree(c);\n}\n\nstatic void test_throughput(struct config config) {\n redisContext *c = connect(config);\n redisReply **replies;\n int i, num;\n long long t1, t2;\n\n test(\"Throughput:\\n\");\n for (i = 0; i < 500; i++)\n freeReplyObject(redisCommand(c,\"LPUSH mylist foo\"));\n\n num = 1000;\n replies = malloc(sizeof(redisReply*)*num);\n t1 = usec();\n for (i = 0; i < num; i++) {\n replies[i] = redisCommand(c,\"PING\");\n assert(replies[i] != NULL && replies[i]->type == REDIS_REPLY_STATUS);\n }\n t2 = usec();\n for (i = 0; i < num; i++) freeReplyObject(replies[i]);\n free(replies);\n printf(\"\\t(%dx PING: %.3fs)\\n\", num, (t2-t1)/1000000.0);\n\n replies = malloc(sizeof(redisReply*)*num);\n t1 = usec();\n for (i = 0; i < num; i++) {\n replies[i] = redisCommand(c,\"LRANGE mylist 0 499\");\n assert(replies[i] != NULL && replies[i]->type == REDIS_REPLY_ARRAY);\n assert(replies[i] != NULL && replies[i]->elements == 500);\n }\n t2 = usec();\n for (i = 0; i < num; i++) freeReplyObject(replies[i]);\n free(replies);\n printf(\"\\t(%dx LRANGE with 500 elements: %.3fs)\\n\", num, (t2-t1)/1000000.0);\n\n num = 10000;\n replies = malloc(sizeof(redisReply*)*num);\n for (i = 0; i < num; i++)\n redisAppendCommand(c,\"PING\");\n t1 = usec();\n for (i = 0; i < num; i++) {\n assert(redisGetReply(c, (void*)&replies[i]) == REDIS_OK);\n assert(replies[i] != NULL && replies[i]->type == REDIS_REPLY_STATUS);\n }\n t2 = usec();\n for (i = 0; i < num; i++) freeReplyObject(replies[i]);\n free(replies);\n printf(\"\\t(%dx PING (pipelined): %.3fs)\\n\", num, (t2-t1)/1000000.0);\n\n replies = malloc(sizeof(redisReply*)*num);\n for (i = 0; i < num; i++)\n redisAppendCommand(c,\"LRANGE mylist 0 499\");\n t1 = usec();\n for (i = 0; i < num; i++) {\n assert(redisGetReply(c, (void*)&replies[i]) == REDIS_OK);\n assert(replies[i] != NULL && replies[i]->type == REDIS_REPLY_ARRAY);\n assert(replies[i] != NULL && replies[i]->elements == 500);\n }\n t2 = usec();\n for (i = 0; i < num; i++) freeReplyObject(replies[i]);\n free(replies);\n printf(\"\\t(%dx LRANGE with 500 elements (pipelined): %.3fs)\\n\", num, (t2-t1)/1000000.0);\n\n disconnect(c);\n}\n\n// static long __test_callback_flags = 0;\n// static void __test_callback(redisContext *c, void *privdata) {\n// ((void)c);\n// /* Shift to detect execution order */\n// __test_callback_flags <<= 8;\n// __test_callback_flags |= (long)privdata;\n// }\n//\n// static void __test_reply_callback(redisContext *c, redisReply *reply, void *privdata) {\n// ((void)c);\n// /* Shift to detect execution order */\n// __test_callback_flags <<= 8;\n// __test_callback_flags |= (long)privdata;\n// if (reply) freeReplyObject(reply);\n// }\n//\n// static redisContext *__connect_nonblock() {\n// /* Reset callback flags */\n// __test_callback_flags = 0;\n// return redisConnectNonBlock(\"127.0.0.1\", port, NULL);\n// }\n//\n// static void test_nonblocking_connection() {\n// redisContext *c;\n// int wdone = 0;\n//\n// test(\"Calls command callback when command is issued: \");\n// c = __connect_nonblock();\n// redisSetCommandCallback(c,__test_callback,(void*)1);\n// redisCommand(c,\"PING\");\n// test_cond(__test_callback_flags == 1);\n// redisFree(c);\n//\n// test(\"Calls disconnect callback on redisDisconnect: \");\n// c = __connect_nonblock();\n// redisSetDisconnectCallback(c,__test_callback,(void*)2);\n// redisDisconnect(c);\n// test_cond(__test_callback_flags == 2);\n// redisFree(c);\n//\n// test(\"Calls disconnect callback and free callback on redisFree: \");\n// c = __connect_nonblock();\n// redisSetDisconnectCallback(c,__test_callback,(void*)2);\n// redisSetFreeCallback(c,__test_callback,(void*)4);\n// redisFree(c);\n// test_cond(__test_callback_flags == ((2 << 8) | 4));\n//\n// test(\"redisBufferWrite against empty write buffer: \");\n// c = __connect_nonblock();\n// test_cond(redisBufferWrite(c,&wdone) == REDIS_OK && wdone == 1);\n// redisFree(c);\n//\n// test(\"redisBufferWrite against not yet connected fd: \");\n// c = __connect_nonblock();\n// redisCommand(c,\"PING\");\n// test_cond(redisBufferWrite(c,NULL) == REDIS_ERR &&\n// strncmp(c->error,\"write:\",6) == 0);\n// redisFree(c);\n//\n// test(\"redisBufferWrite against closed fd: \");\n// c = __connect_nonblock();\n// redisCommand(c,\"PING\");\n// redisDisconnect(c);\n// test_cond(redisBufferWrite(c,NULL) == REDIS_ERR &&\n// strncmp(c->error,\"write:\",6) == 0);\n// redisFree(c);\n//\n// test(\"Process callbacks in the right sequence: \");\n// c = __connect_nonblock();\n// redisCommandWithCallback(c,__test_reply_callback,(void*)1,\"PING\");\n// redisCommandWithCallback(c,__test_reply_callback,(void*)2,\"PING\");\n// redisCommandWithCallback(c,__test_reply_callback,(void*)3,\"PING\");\n//\n// /* Write output buffer */\n// wdone = 0;\n// while(!wdone) {\n// usleep(500);\n// redisBufferWrite(c,&wdone);\n// }\n//\n// /* Read until at least one callback is executed (the 3 replies will\n// * arrive in a single packet, causing all callbacks to be executed in\n// * a single pass). */\n// while(__test_callback_flags == 0) {\n// assert(redisBufferRead(c) == REDIS_OK);\n// redisProcessCallbacks(c);\n// }\n// test_cond(__test_callback_flags == 0x010203);\n// redisFree(c);\n//\n// test(\"redisDisconnect executes pending callbacks with NULL reply: \");\n// c = __connect_nonblock();\n// redisSetDisconnectCallback(c,__test_callback,(void*)1);\n// redisCommandWithCallback(c,__test_reply_callback,(void*)2,\"PING\");\n// redisDisconnect(c);\n// test_cond(__test_callback_flags == 0x0201);\n// redisFree(c);\n// }\n\nint main(int argc, char **argv) {\n struct config cfg = {\n .tcp = {\n .host = \"127.0.0.1\",\n .port = 6379\n },\n .unix = {\n .path = \"/tmp/redis.sock\"\n }\n };\n int throughput = 1;\n\n /* Ignore broken pipe signal (for I/O error tests). */\n signal(SIGPIPE, SIG_IGN);\n\n /* Parse command line options. */\n argv++; argc--;\n while (argc) {\n if (argc >= 2 && !strcmp(argv[0],\"-h\")) {\n argv++; argc--;\n cfg.tcp.host = argv[0];\n } else if (argc >= 2 && !strcmp(argv[0],\"-p\")) {\n argv++; argc--;\n cfg.tcp.port = atoi(argv[0]);\n } else if (argc >= 2 && !strcmp(argv[0],\"-s\")) {\n argv++; argc--;\n cfg.unix.path = argv[0];\n } else if (argc >= 1 && !strcmp(argv[0],\"--skip-throughput\")) {\n throughput = 0;\n } else {\n fprintf(stderr, \"Invalid argument: %s\\n\", argv[0]);\n exit(1);\n }\n argv++; argc--;\n }\n\n test_format_commands();\n test_reply_reader();\n test_blocking_connection_errors();\n\n printf(\"\\nTesting against TCP connection (%s:%d):\\n\", cfg.tcp.host, cfg.tcp.port);\n cfg.type = CONN_TCP;\n test_blocking_connection(cfg);\n test_blocking_io_errors(cfg);\n if (throughput) test_throughput(cfg);\n\n printf(\"\\nTesting against Unix socket connection (%s):\\n\", cfg.unix.path);\n cfg.type = CONN_UNIX;\n test_blocking_connection(cfg);\n test_blocking_io_errors(cfg);\n if (throughput) test_throughput(cfg);\n\n if (fails) {\n printf(\"*** %d TESTS FAILED ***\\n\", fails);\n return 1;\n }\n\n printf(\"ALL TESTS PASSED\\n\");\n return 0;\n}\n" }, { "path": "deps/hiredis/zmalloc.h", "content": "/* Drop in replacement for zmalloc.h in order to just use libc malloc without\n * any wrappering. */\n\n#ifndef ZMALLOC_H\n#define ZMALLOC_H\n\n#define zmalloc malloc\n#define zrealloc realloc\n#define zcalloc(x) calloc(x,1)\n#define zfree free\n#define zstrdup strdup\n\n#endif\n" }, { "path": "deps/jemalloc/.gitignore", "content": "/autom4te.cache/\n/config.stamp\n/config.log\n/config.status\n/doc/html.xsl\n/doc/manpages.xsl\n/doc/jemalloc.xml\n/lib/\n/Makefile\n/include/jemalloc/internal/jemalloc_internal\\.h\n/include/jemalloc/internal/size_classes\\.h\n/include/jemalloc/jemalloc\\.h\n/include/jemalloc/jemalloc_defs\\.h\n/src/*.[od]\n/test/*.[od]\n/test/*.out\n/test/[a-zA-Z_]*\n!test/*.c\n!test/*.exp\n/bin/jemalloc.sh\n" }, { "path": "deps/jemalloc/COPYING", "content": "Unless otherwise specified, files in the jemalloc source distribution are\nsubject to the following license:\n--------------------------------------------------------------------------------\nCopyright (C) 2002-2012 Jason Evans .\nAll rights reserved.\nCopyright (C) 2007-2012 Mozilla Foundation. All rights reserved.\nCopyright (C) 2009-2012 Facebook, Inc. All rights reserved.\n\nRedistribution and use in source and binary forms, with or without\nmodification, are permitted provided that the following conditions are met:\n1. Redistributions of source code must retain the above copyright notice(s),\n this list of conditions and the following disclaimer.\n2. Redistributions in binary form must reproduce the above copyright notice(s),\n this list of conditions and the following disclaimer in the documentation\n and/or other materials provided with the distribution.\n\nTHIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS\nOR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF\nMERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO\nEVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,\nINCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\nLIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR\nPROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF\nLIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE\nOR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF\nADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n--------------------------------------------------------------------------------\n" }, { "path": "deps/jemalloc/ChangeLog", "content": "Following are change highlights associated with official releases. Important\nbug fixes are all mentioned, but internal enhancements are omitted here for\nbrevity (even though they are more fun to write about). Much more detail can be\nfound in the git revision history:\n\n http://www.canonware.com/cgi-bin/gitweb.cgi?p=jemalloc.git\n git://canonware.com/jemalloc.git\n\n* 3.2.0 (November 9, 2012)\n\n In addition to a couple of bug fixes, this version modifies page run\n allocation and dirty page purging algorithms in order to better control\n page-level virtual memory fragmentation.\n\n Incompatible changes:\n - Change the \"opt.lg_dirty_mult\" default from 5 to 3 (32:1 to 8:1).\n\n Bug fixes:\n - Fix dss/mmap allocation precedence code to use recyclable mmap memory only\n after primary dss allocation fails.\n - Fix deadlock in the \"arenas.purge\" mallctl. This regression was introduced\n in 3.1.0 by the addition of the \"arena..purge\" mallctl.\n\n* 3.1.0 (October 16, 2012)\n\n New features:\n - Auto-detect whether running inside Valgrind, thus removing the need to\n manually specify MALLOC_CONF=valgrind:true.\n - Add the \"arenas.extend\" mallctl, which allows applications to create\n manually managed arenas.\n - Add the ALLOCM_ARENA() flag for {,r,d}allocm().\n - Add the \"opt.dss\", \"arena..dss\", and \"stats.arenas..dss\" mallctls,\n which provide control over dss/mmap precedence.\n - Add the \"arena..purge\" mallctl, which obsoletes \"arenas.purge\".\n - Define LG_QUANTUM for hppa.\n\n Incompatible changes:\n - Disable tcache by default if running inside Valgrind, in order to avoid\n making unallocated objects appear reachable to Valgrind.\n - Drop const from malloc_usable_size() argument on Linux.\n\n Bug fixes:\n - Fix heap profiling crash if sampled object is freed via realloc(p, 0).\n - Remove const from __*_hook variable declarations, so that glibc can modify\n them during process forking.\n - Fix mlockall(2)/madvise(2) interaction.\n - Fix fork(2)-related deadlocks.\n - Fix error return value for \"thread.tcache.enabled\" mallctl.\n\n* 3.0.0 (May 11, 2012)\n\n Although this version adds some major new features, the primary focus is on\n internal code cleanup that facilitates maintainability and portability, most\n of which is not reflected in the ChangeLog. This is the first release to\n incorporate substantial contributions from numerous other developers, and the\n result is a more broadly useful allocator (see the git revision history for\n contribution details). Note that the license has been unified, thanks to\n Facebook granting a license under the same terms as the other copyright\n holders (see COPYING).\n\n New features:\n - Implement Valgrind support, redzones, and quarantine.\n - Add support for additional platforms:\n + FreeBSD\n + Mac OS X Lion\n + MinGW\n + Windows (no support yet for replacing the system malloc)\n - Add support for additional architectures:\n + MIPS\n + SH4\n + Tilera\n - Add support for cross compiling.\n - Add nallocm(), which rounds a request size up to the nearest size class\n without actually allocating.\n - Implement aligned_alloc() (blame C11).\n - Add the \"thread.tcache.enabled\" mallctl.\n - Add the \"opt.prof_final\" mallctl.\n - Update pprof (from gperftools 2.0).\n - Add the --with-mangling option.\n - Add the --disable-experimental option.\n - Add the --disable-munmap option, and make it the default on Linux.\n - Add the --enable-mremap option, which disables use of mremap(2) by default.\n\n Incompatible changes:\n - Enable stats by default.\n - Enable fill by default.\n - Disable lazy locking by default.\n - Rename the \"tcache.flush\" mallctl to \"thread.tcache.flush\".\n - Rename the \"arenas.pagesize\" mallctl to \"arenas.page\".\n - Change the \"opt.lg_prof_sample\" default from 0 to 19 (1 B to 512 KiB).\n - Change the \"opt.prof_accum\" default from true to false.\n\n Removed features:\n - Remove the swap feature, including the \"config.swap\", \"swap.avail\",\n \"swap.prezeroed\", \"swap.nfds\", and \"swap.fds\" mallctls.\n - Remove highruns statistics, including the\n \"stats.arenas..bins..highruns\" and\n \"stats.arenas..lruns..highruns\" mallctls.\n - As part of small size class refactoring, remove the \"opt.lg_[qc]space_max\",\n \"arenas.cacheline\", \"arenas.subpage\", \"arenas.[tqcs]space_{min,max}\", and\n \"arenas.[tqcs]bins\" mallctls.\n - Remove the \"arenas.chunksize\" mallctl.\n - Remove the \"opt.lg_prof_tcmax\" option.\n - Remove the \"opt.lg_prof_bt_max\" option.\n - Remove the \"opt.lg_tcache_gc_sweep\" option.\n - Remove the --disable-tiny option, including the \"config.tiny\" mallctl.\n - Remove the --enable-dynamic-page-shift configure option.\n - Remove the --enable-sysv configure option.\n\n Bug fixes:\n - Fix a statistics-related bug in the \"thread.arena\" mallctl that could cause\n invalid statistics and crashes.\n - Work around TLS deallocation via free() on Linux. This bug could cause\n write-after-free memory corruption.\n - Fix a potential deadlock that could occur during interval- and\n growth-triggered heap profile dumps.\n - Fix large calloc() zeroing bugs due to dropping chunk map unzeroed flags.\n - Fix chunk_alloc_dss() to stop claiming memory is zeroed. This bug could\n cause memory corruption and crashes with --enable-dss specified.\n - Fix fork-related bugs that could cause deadlock in children between fork\n and exec.\n - Fix malloc_stats_print() to honor 'b' and 'l' in the opts parameter.\n - Fix realloc(p, 0) to act like free(p).\n - Do not enforce minimum alignment in memalign().\n - Check for NULL pointer in malloc_usable_size().\n - Fix an off-by-one heap profile statistics bug that could be observed in\n interval- and growth-triggered heap profiles.\n - Fix the \"epoch\" mallctl to update cached stats even if the passed in epoch\n is 0.\n - Fix bin->runcur management to fix a layout policy bug. This bug did not\n affect correctness.\n - Fix a bug in choose_arena_hard() that potentially caused more arenas to be\n initialized than necessary.\n - Add missing \"opt.lg_tcache_max\" mallctl implementation.\n - Use glibc allocator hooks to make mixed allocator usage less likely.\n - Fix build issues for --disable-tcache.\n - Don't mangle pthread_create() when --with-private-namespace is specified.\n\n* 2.2.5 (November 14, 2011)\n\n Bug fixes:\n - Fix huge_ralloc() race when using mremap(2). This is a serious bug that\n could cause memory corruption and/or crashes.\n - Fix huge_ralloc() to maintain chunk statistics.\n - Fix malloc_stats_print(..., \"a\") output.\n\n* 2.2.4 (November 5, 2011)\n\n Bug fixes:\n - Initialize arenas_tsd before using it. This bug existed for 2.2.[0-3], as\n well as for --disable-tls builds in earlier releases.\n - Do not assume a 4 KiB page size in test/rallocm.c.\n\n* 2.2.3 (August 31, 2011)\n\n This version fixes numerous bugs related to heap profiling.\n\n Bug fixes:\n - Fix a prof-related race condition. This bug could cause memory corruption,\n but only occurred in non-default configurations (prof_accum:false).\n - Fix off-by-one backtracing issues (make sure that prof_alloc_prep() is\n excluded from backtraces).\n - Fix a prof-related bug in realloc() (only triggered by OOM errors).\n - Fix prof-related bugs in allocm() and rallocm().\n - Fix prof_tdata_cleanup() for --disable-tls builds.\n - Fix a relative include path, to fix objdir builds.\n\n* 2.2.2 (July 30, 2011)\n\n Bug fixes:\n - Fix a build error for --disable-tcache.\n - Fix assertions in arena_purge() (for real this time).\n - Add the --with-private-namespace option. This is a workaround for symbol\n conflicts that can inadvertently arise when using static libraries.\n\n* 2.2.1 (March 30, 2011)\n\n Bug fixes:\n - Implement atomic operations for x86/x64. This fixes compilation failures\n for versions of gcc that are still in wide use.\n - Fix an assertion in arena_purge().\n\n* 2.2.0 (March 22, 2011)\n\n This version incorporates several improvements to algorithms and data\n structures that tend to reduce fragmentation and increase speed.\n\n New features:\n - Add the \"stats.cactive\" mallctl.\n - Update pprof (from google-perftools 1.7).\n - Improve backtracing-related configuration logic, and add the\n --disable-prof-libgcc option.\n\n Bug fixes:\n - Change default symbol visibility from \"internal\", to \"hidden\", which\n decreases the overhead of library-internal function calls.\n - Fix symbol visibility so that it is also set on OS X.\n - Fix a build dependency regression caused by the introduction of the .pic.o\n suffix for PIC object files.\n - Add missing checks for mutex initialization failures.\n - Don't use libgcc-based backtracing except on x64, where it is known to work.\n - Fix deadlocks on OS X that were due to memory allocation in\n pthread_mutex_lock().\n - Heap profiling-specific fixes:\n + Fix memory corruption due to integer overflow in small region index\n computation, when using a small enough sample interval that profiling\n context pointers are stored in small run headers.\n + Fix a bootstrap ordering bug that only occurred with TLS disabled.\n + Fix a rallocm() rsize bug.\n + Fix error detection bugs for aligned memory allocation.\n\n* 2.1.3 (March 14, 2011)\n\n Bug fixes:\n - Fix a cpp logic regression (due to the \"thread.{de,}allocatedp\" mallctl fix\n for OS X in 2.1.2).\n - Fix a \"thread.arena\" mallctl bug.\n - Fix a thread cache stats merging bug.\n\n* 2.1.2 (March 2, 2011)\n\n Bug fixes:\n - Fix \"thread.{de,}allocatedp\" mallctl for OS X.\n - Add missing jemalloc.a to build system.\n\n* 2.1.1 (January 31, 2011)\n\n Bug fixes:\n - Fix aligned huge reallocation (affected allocm()).\n - Fix the ALLOCM_LG_ALIGN macro definition.\n - Fix a heap dumping deadlock.\n - Fix a \"thread.arena\" mallctl bug.\n\n* 2.1.0 (December 3, 2010)\n\n This version incorporates some optimizations that can't quite be considered\n bug fixes.\n\n New features:\n - Use Linux's mremap(2) for huge object reallocation when possible.\n - Avoid locking in mallctl*() when possible.\n - Add the \"thread.[de]allocatedp\" mallctl's.\n - Convert the manual page source from roff to DocBook, and generate both roff\n and HTML manuals.\n\n Bug fixes:\n - Fix a crash due to incorrect bootstrap ordering. This only impacted\n --enable-debug --enable-dss configurations.\n - Fix a minor statistics bug for mallctl(\"swap.avail\", ...).\n\n* 2.0.1 (October 29, 2010)\n\n Bug fixes:\n - Fix a race condition in heap profiling that could cause undefined behavior\n if \"opt.prof_accum\" were disabled.\n - Add missing mutex unlocks for some OOM error paths in the heap profiling\n code.\n - Fix a compilation error for non-C99 builds.\n\n* 2.0.0 (October 24, 2010)\n\n This version focuses on the experimental *allocm() API, and on improved\n run-time configuration/introspection. Nonetheless, numerous performance\n improvements are also included.\n\n New features:\n - Implement the experimental {,r,s,d}allocm() API, which provides a superset\n of the functionality available via malloc(), calloc(), posix_memalign(),\n realloc(), malloc_usable_size(), and free(). These functions can be used to\n allocate/reallocate aligned zeroed memory, ask for optional extra memory\n during reallocation, prevent object movement during reallocation, etc.\n - Replace JEMALLOC_OPTIONS/JEMALLOC_PROF_PREFIX with MALLOC_CONF, which is\n more human-readable, and more flexible. For example:\n JEMALLOC_OPTIONS=AJP\n is now:\n MALLOC_CONF=abort:true,fill:true,stats_print:true\n - Port to Apple OS X. Sponsored by Mozilla.\n - Make it possible for the application to control thread-->arena mappings via\n the \"thread.arena\" mallctl.\n - Add compile-time support for all TLS-related functionality via pthreads TSD.\n This is mainly of interest for OS X, which does not support TLS, but has a\n TSD implementation with similar performance.\n - Override memalign() and valloc() if they are provided by the system.\n - Add the \"arenas.purge\" mallctl, which can be used to synchronously purge all\n dirty unused pages.\n - Make cumulative heap profiling data optional, so that it is possible to\n limit the amount of memory consumed by heap profiling data structures.\n - Add per thread allocation counters that can be accessed via the\n \"thread.allocated\" and \"thread.deallocated\" mallctls.\n\n Incompatible changes:\n - Remove JEMALLOC_OPTIONS and malloc_options (see MALLOC_CONF above).\n - Increase default backtrace depth from 4 to 128 for heap profiling.\n - Disable interval-based profile dumps by default.\n\n Bug fixes:\n - Remove bad assertions in fork handler functions. These assertions could\n cause aborts for some combinations of configure settings.\n - Fix strerror_r() usage to deal with non-standard semantics in GNU libc.\n - Fix leak context reporting. This bug tended to cause the number of contexts\n to be underreported (though the reported number of objects and bytes were\n correct).\n - Fix a realloc() bug for large in-place growing reallocation. This bug could\n cause memory corruption, but it was hard to trigger.\n - Fix an allocation bug for small allocations that could be triggered if\n multiple threads raced to create a new run of backing pages.\n - Enhance the heap profiler to trigger samples based on usable size, rather\n than request size.\n - Fix a heap profiling bug due to sometimes losing track of requested object\n size for sampled objects.\n\n* 1.0.3 (August 12, 2010)\n\n Bug fixes:\n - Fix the libunwind-based implementation of stack backtracing (used for heap\n profiling). This bug could cause zero-length backtraces to be reported.\n - Add a missing mutex unlock in library initialization code. If multiple\n threads raced to initialize malloc, some of them could end up permanently\n blocked.\n\n* 1.0.2 (May 11, 2010)\n\n Bug fixes:\n - Fix junk filling of large objects, which could cause memory corruption.\n - Add MAP_NORESERVE support for chunk mapping, because otherwise virtual\n memory limits could cause swap file configuration to fail. Contributed by\n Jordan DeLong.\n\n* 1.0.1 (April 14, 2010)\n\n Bug fixes:\n - Fix compilation when --enable-fill is specified.\n - Fix threads-related profiling bugs that affected accuracy and caused memory\n to be leaked during thread exit.\n - Fix dirty page purging race conditions that could cause crashes.\n - Fix crash in tcache flushing code during thread destruction.\n\n* 1.0.0 (April 11, 2010)\n\n This release focuses on speed and run-time introspection. Numerous\n algorithmic improvements make this release substantially faster than its\n predecessors.\n\n New features:\n - Implement autoconf-based configuration system.\n - Add mallctl*(), for the purposes of introspection and run-time\n configuration.\n - Make it possible for the application to manually flush a thread's cache, via\n the \"tcache.flush\" mallctl.\n - Base maximum dirty page count on proportion of active memory.\n - Compute various addtional run-time statistics, including per size class\n statistics for large objects.\n - Expose malloc_stats_print(), which can be called repeatedly by the\n application.\n - Simplify the malloc_message() signature to only take one string argument,\n and incorporate an opaque data pointer argument for use by the application\n in combination with malloc_stats_print().\n - Add support for allocation backed by one or more swap files, and allow the\n application to disable over-commit if swap files are in use.\n - Implement allocation profiling and leak checking.\n\n Removed features:\n - Remove the dynamic arena rebalancing code, since thread-specific caching\n reduces its utility.\n\n Bug fixes:\n - Modify chunk allocation to work when address space layout randomization\n (ASLR) is in use.\n - Fix thread cleanup bugs related to TLS destruction.\n - Handle 0-size allocation requests in posix_memalign().\n - Fix a chunk leak. The leaked chunks were never touched, so this impacted\n virtual memory usage, but not physical memory usage.\n\n* linux_2008082[78]a (August 27/28, 2008)\n\n These snapshot releases are the simple result of incorporating Linux-specific\n support into the FreeBSD malloc sources.\n\n--------------------------------------------------------------------------------\nvim:filetype=text:textwidth=80\n" }, { "path": "deps/jemalloc/INSTALL", "content": "Building and installing jemalloc can be as simple as typing the following while\nin the root directory of the source tree:\n\n ./configure\n make\n make install\n\n=== Advanced configuration =====================================================\n\nThe 'configure' script supports numerous options that allow control of which\nfunctionality is enabled, where jemalloc is installed, etc. Optionally, pass\nany of the following arguments (not a definitive list) to 'configure':\n\n--help\n Print a definitive list of options.\n\n--prefix=\n Set the base directory in which to install. For example:\n\n ./configure --prefix=/usr/local\n\n will cause files to be installed into /usr/local/include, /usr/local/lib,\n and /usr/local/man.\n\n--with-rpath=\n Embed one or more library paths, so that libjemalloc can find the libraries\n it is linked to. This works only on ELF-based systems.\n\n--with-mangling=\n Mangle public symbols specified in which is a comma-separated list of\n name:mangled pairs.\n\n For example, to use ld's --wrap option as an alternative method for\n overriding libc's malloc implementation, specify something like:\n\n --with-mangling=malloc:__wrap_malloc,free:__wrap_free[...]\n\n Note that mangling happens prior to application of the prefix specified by\n --with-jemalloc-prefix, and mangled symbols are then ignored when applying\n the prefix.\n\n--with-jemalloc-prefix=\n Prefix all public APIs with . For example, if is\n \"prefix_\", API changes like the following occur:\n\n malloc() --> prefix_malloc()\n malloc_conf --> prefix_malloc_conf\n /etc/malloc.conf --> /etc/prefix_malloc.conf\n MALLOC_CONF --> PREFIX_MALLOC_CONF\n\n This makes it possible to use jemalloc at the same time as the system\n allocator, or even to use multiple copies of jemalloc simultaneously.\n\n By default, the prefix is \"\", except on OS X, where it is \"je_\". On OS X,\n jemalloc overlays the default malloc zone, but makes no attempt to actually\n replace the \"malloc\", \"calloc\", etc. symbols.\n\n--with-private-namespace=\n Prefix all library-private APIs with . For shared libraries,\n symbol visibility mechanisms prevent these symbols from being exported, but\n for static libraries, naming collisions are a real possibility. By\n default, the prefix is \"\" (empty string).\n\n--with-install-suffix=\n Append to the base name of all installed files, such that multiple\n versions of jemalloc can coexist in the same installation directory. For\n example, libjemalloc.so.0 becomes libjemalloc.so.0.\n\n--enable-cc-silence\n Enable code that silences non-useful compiler warnings. This is helpful\n when trying to tell serious warnings from those due to compiler\n limitations, but it potentially incurs a performance penalty.\n\n--enable-debug\n Enable assertions and validation code. This incurs a substantial\n performance hit, but is very useful during application development.\n\n--disable-stats\n Disable statistics gathering functionality. See the \"opt.stats_print\"\n option documentation for usage details.\n\n--enable-prof\n Enable heap profiling and leak detection functionality. See the \"opt.prof\"\n option documentation for usage details. When enabled, there are several\n approaches to backtracing, and the configure script chooses the first one\n in the following list that appears to function correctly:\n\n + libunwind (requires --enable-prof-libunwind)\n + libgcc (unless --disable-prof-libgcc)\n + gcc intrinsics (unless --disable-prof-gcc)\n\n--enable-prof-libunwind\n Use the libunwind library (http://www.nongnu.org/libunwind/) for stack\n backtracing.\n\n--disable-prof-libgcc\n Disable the use of libgcc's backtracing functionality.\n\n--disable-prof-gcc\n Disable the use of gcc intrinsics for backtracing.\n\n--with-static-libunwind=\n Statically link against the specified libunwind.a rather than dynamically\n linking with -lunwind.\n\n--disable-tcache\n Disable thread-specific caches for small objects. Objects are cached and\n released in bulk, thus reducing the total number of mutex operations. See\n the \"opt.tcache\" option for usage details.\n\n--enable-mremap\n Enable huge realloc() via mremap(2). mremap() is disabled by default\n because the flavor used is specific to Linux, which has a quirk in its\n virtual memory allocation algorithm that causes semi-permanent VM map holes\n under normal jemalloc operation.\n\n--disable-munmap\n Disable virtual memory deallocation via munmap(2); instead keep track of\n the virtual memory for later use. munmap() is disabled by default (i.e.\n --disable-munmap is implied) on Linux, which has a quirk in its virtual\n memory allocation algorithm that causes semi-permanent VM map holes under\n normal jemalloc operation.\n\n--enable-dss\n Enable support for page allocation/deallocation via sbrk(2), in addition to\n mmap(2).\n\n--disable-fill\n Disable support for junk/zero filling of memory, quarantine, and redzones.\n See the \"opt.junk\", \"opt.zero\", \"opt.quarantine\", and \"opt.redzone\" option\n documentation for usage details.\n\n--disable-valgrind\n Disable support for Valgrind.\n\n--disable-experimental\n Disable support for the experimental API (*allocm()).\n\n--enable-utrace\n Enable utrace(2)-based allocation tracing. This feature is not broadly\n portable (FreeBSD has it, but Linux and OS X do not).\n\n--enable-xmalloc\n Enable support for optional immediate termination due to out-of-memory\n errors, as is commonly implemented by \"xmalloc\" wrapper function for malloc.\n See the \"opt.xmalloc\" option documentation for usage details.\n\n--enable-lazy-lock\n Enable code that wraps pthread_create() to detect when an application\n switches from single-threaded to multi-threaded mode, so that it can avoid\n mutex locking/unlocking operations while in single-threaded mode. In\n practice, this feature usually has little impact on performance unless\n thread-specific caching is disabled.\n\n--disable-tls\n Disable thread-local storage (TLS), which allows for fast access to\n thread-local variables via the __thread keyword. If TLS is available,\n jemalloc uses it for several purposes.\n\n--with-xslroot=\n Specify where to find DocBook XSL stylesheets when building the\n documentation.\n\nThe following environment variables (not a definitive list) impact configure's\nbehavior:\n\nCFLAGS=\"?\"\n Pass these flags to the compiler. You probably shouldn't define this unless\n you know what you are doing. (Use EXTRA_CFLAGS instead.)\n\nEXTRA_CFLAGS=\"?\"\n Append these flags to CFLAGS. This makes it possible to add flags such as\n -Werror, while allowing the configure script to determine what other flags\n are appropriate for the specified configuration.\n\n The configure script specifically checks whether an optimization flag (-O*)\n is specified in EXTRA_CFLAGS, and refrains from specifying an optimization\n level if it finds that one has already been specified.\n\nCPPFLAGS=\"?\"\n Pass these flags to the C preprocessor. Note that CFLAGS is not passed to\n 'cpp' when 'configure' is looking for include files, so you must use\n CPPFLAGS instead if you need to help 'configure' find header files.\n\nLD_LIBRARY_PATH=\"?\"\n 'ld' uses this colon-separated list to find libraries.\n\nLDFLAGS=\"?\"\n Pass these flags when linking.\n\nPATH=\"?\"\n 'configure' uses this to find programs.\n\n=== Advanced compilation =======================================================\n\nTo build only parts of jemalloc, use the following targets:\n\n build_lib_shared\n build_lib_static\n build_lib\n build_doc_html\n build_doc_man\n build_doc\n\nTo install only parts of jemalloc, use the following targets:\n\n install_bin\n install_include\n install_lib_shared\n install_lib_static\n install_lib\n install_doc_html\n install_doc_man\n install_doc\n\nTo clean up build results to varying degrees, use the following make targets:\n\n clean\n distclean\n relclean\n\n=== Advanced installation ======================================================\n\nOptionally, define make variables when invoking make, including (not\nexclusively):\n\nINCLUDEDIR=\"?\"\n Use this as the installation prefix for header files.\n\nLIBDIR=\"?\"\n Use this as the installation prefix for libraries.\n\nMANDIR=\"?\"\n Use this as the installation prefix for man pages.\n\nDESTDIR=\"?\"\n Prepend DESTDIR to INCLUDEDIR, LIBDIR, DATADIR, and MANDIR. This is useful\n when installing to a different path than was specified via --prefix.\n\nCC=\"?\"\n Use this to invoke the C compiler.\n\nCFLAGS=\"?\"\n Pass these flags to the compiler.\n\nCPPFLAGS=\"?\"\n Pass these flags to the C preprocessor.\n\nLDFLAGS=\"?\"\n Pass these flags when linking.\n\nPATH=\"?\"\n Use this to search for programs used during configuration and building.\n\n=== Development ================================================================\n\nIf you intend to make non-trivial changes to jemalloc, use the 'autogen.sh'\nscript rather than 'configure'. This re-generates 'configure', enables\nconfiguration dependency rules, and enables re-generation of automatically\ngenerated source files.\n\nThe build system supports using an object directory separate from the source\ntree. For example, you can create an 'obj' directory, and from within that\ndirectory, issue configuration and build commands:\n\n autoconf\n mkdir obj\n cd obj\n ../configure --enable-autogen\n make\n\n=== Documentation ==============================================================\n\nThe manual page is generated in both html and roff formats. Any web browser\ncan be used to view the html manual. The roff manual page can be formatted\nprior to installation via the following command:\n\n nroff -man -t doc/jemalloc.3\n" }, { "path": "deps/jemalloc/Makefile.in", "content": "# Clear out all vpaths, then set just one (default vpath) for the main build\n# directory.\nvpath\nvpath % .\n\n# Clear the default suffixes, so that built-in rules are not used.\n.SUFFIXES :\n\nSHELL := /bin/sh\n\nCC := @CC@\n\n# Configuration parameters.\nDESTDIR =\nBINDIR := $(DESTDIR)@BINDIR@\nINCLUDEDIR := $(DESTDIR)@INCLUDEDIR@\nLIBDIR := $(DESTDIR)@LIBDIR@\nDATADIR := $(DESTDIR)@DATADIR@\nMANDIR := $(DESTDIR)@MANDIR@\nsrcroot := @srcroot@\nobjroot := @objroot@\nabs_srcroot := @abs_srcroot@\nabs_objroot := @abs_objroot@\n\n# Build parameters.\nCPPFLAGS := @CPPFLAGS@ -I$(srcroot)include -I$(objroot)include\nCFLAGS := @CFLAGS@\nLDFLAGS := @LDFLAGS@\nEXTRA_LDFLAGS := @EXTRA_LDFLAGS@\nLIBS := @LIBS@\nRPATH_EXTRA := @RPATH_EXTRA@\nSO := @so@\nIMPORTLIB := @importlib@\nO := @o@\nA := @a@\nEXE := @exe@\nLIBPREFIX := @libprefix@\nREV := @rev@\ninstall_suffix := @install_suffix@\nABI := @abi@\nXSLTPROC := @XSLTPROC@\nAUTOCONF := @AUTOCONF@\n_RPATH = @RPATH@\nRPATH = $(if $(1),$(call _RPATH,$(1)))\ncfghdrs_in := @cfghdrs_in@\ncfghdrs_out := @cfghdrs_out@\ncfgoutputs_in := @cfgoutputs_in@\ncfgoutputs_out := @cfgoutputs_out@\nenable_autogen := @enable_autogen@\nenable_experimental := @enable_experimental@\nDSO_LDFLAGS = @DSO_LDFLAGS@\nSOREV = @SOREV@\nPIC_CFLAGS = @PIC_CFLAGS@\nCTARGET = @CTARGET@\nLDTARGET = @LDTARGET@\nMKLIB = @MKLIB@\nCC_MM = @CC_MM@\n\nifeq (macho, $(ABI))\nTEST_LIBRARY_PATH := DYLD_FALLBACK_LIBRARY_PATH=\"$(objroot)lib\"\nelse\nifeq (pecoff, $(ABI))\nTEST_LIBRARY_PATH := PATH=\"$(PATH):$(objroot)lib\"\nelse\nTEST_LIBRARY_PATH :=\nendif\nendif\n\nLIBJEMALLOC := $(LIBPREFIX)jemalloc$(install_suffix)\n\n# Lists of files.\nBINS := $(srcroot)bin/pprof $(objroot)bin/jemalloc.sh\nCHDRS := $(objroot)include/jemalloc/jemalloc$(install_suffix).h \\\n\t$(objroot)include/jemalloc/jemalloc_defs$(install_suffix).h\nCSRCS := $(srcroot)src/jemalloc.c $(srcroot)src/arena.c $(srcroot)src/atomic.c \\\n\t$(srcroot)src/base.c $(srcroot)src/bitmap.c $(srcroot)src/chunk.c \\\n\t$(srcroot)src/chunk_dss.c $(srcroot)src/chunk_mmap.c \\\n\t$(srcroot)src/ckh.c $(srcroot)src/ctl.c $(srcroot)src/extent.c \\\n\t$(srcroot)src/hash.c $(srcroot)src/huge.c $(srcroot)src/mb.c \\\n\t$(srcroot)src/mutex.c $(srcroot)src/prof.c $(srcroot)src/quarantine.c \\\n\t$(srcroot)src/rtree.c $(srcroot)src/stats.c $(srcroot)src/tcache.c \\\n\t$(srcroot)src/util.c $(srcroot)src/tsd.c\nifeq (macho, $(ABI))\nCSRCS += $(srcroot)src/zone.c\nendif\nifeq ($(IMPORTLIB),$(SO))\nSTATIC_LIBS := $(objroot)lib/$(LIBJEMALLOC).$(A)\nendif\nifdef PIC_CFLAGS\nSTATIC_LIBS += $(objroot)lib/$(LIBJEMALLOC)_pic.$(A)\nelse\nSTATIC_LIBS += $(objroot)lib/$(LIBJEMALLOC)_s.$(A)\nendif\nDSOS := $(objroot)lib/$(LIBJEMALLOC).$(SOREV)\nifneq ($(SOREV),$(SO))\nDSOS += $(objroot)lib/$(LIBJEMALLOC).$(SO)\nendif\nMAN3 := $(objroot)doc/jemalloc$(install_suffix).3\nDOCS_XML := $(objroot)doc/jemalloc$(install_suffix).xml\nDOCS_HTML := $(DOCS_XML:$(objroot)%.xml=$(srcroot)%.html)\nDOCS_MAN3 := $(DOCS_XML:$(objroot)%.xml=$(srcroot)%.3)\nDOCS := $(DOCS_HTML) $(DOCS_MAN3)\nCTESTS := $(srcroot)test/aligned_alloc.c $(srcroot)test/allocated.c \\\n\t$(srcroot)test/ALLOCM_ARENA.c $(srcroot)test/bitmap.c \\\n\t$(srcroot)test/mremap.c $(srcroot)test/posix_memalign.c \\\n\t$(srcroot)test/thread_arena.c $(srcroot)test/thread_tcache_enabled.c\nifeq ($(enable_experimental), 1)\nCTESTS += $(srcroot)test/allocm.c $(srcroot)test/rallocm.c\nendif\n\nCOBJS := $(CSRCS:$(srcroot)%.c=$(objroot)%.$(O))\nCPICOBJS := $(CSRCS:$(srcroot)%.c=$(objroot)%.pic.$(O))\nCTESTOBJS := $(CTESTS:$(srcroot)%.c=$(objroot)%.$(O))\n\n.PHONY: all dist doc_html doc_man doc\n.PHONY: install_bin install_include install_lib\n.PHONY: install_html install_man install_doc install\n.PHONY: tests check clean distclean relclean\n\n.SECONDARY : $(CTESTOBJS)\n\n# Default target.\nall: build\n\ndist: build_doc\n\n$(srcroot)doc/%.html : $(objroot)doc/%.xml $(srcroot)doc/stylesheet.xsl $(objroot)doc/html.xsl\n\t$(XSLTPROC) -o $@ $(objroot)doc/html.xsl $<\n\n$(srcroot)doc/%.3 : $(objroot)doc/%.xml $(srcroot)doc/stylesheet.xsl $(objroot)doc/manpages.xsl\n\t$(XSLTPROC) -o $@ $(objroot)doc/manpages.xsl $<\n\nbuild_doc_html: $(DOCS_HTML)\nbuild_doc_man: $(DOCS_MAN3)\nbuild_doc: $(DOCS)\n\n#\n# Include generated dependency files.\n#\nifdef CC_MM\n-include $(COBJS:%.$(O)=%.d)\n-include $(CPICOBJS:%.$(O)=%.d)\n-include $(CTESTOBJS:%.$(O)=%.d)\nendif\n\n$(COBJS): $(objroot)src/%.$(O): $(srcroot)src/%.c\n$(CPICOBJS): $(objroot)src/%.pic.$(O): $(srcroot)src/%.c\n$(CPICOBJS): CFLAGS += $(PIC_CFLAGS)\n$(CTESTOBJS): $(objroot)test/%.$(O): $(srcroot)test/%.c\n$(CTESTOBJS): CPPFLAGS += -I$(objroot)test\nifneq ($(IMPORTLIB),$(SO))\n$(COBJS): CPPFLAGS += -DDLLEXPORT\nendif\n\nifndef CC_MM\n# Dependencies\nHEADER_DIRS = $(srcroot)include/jemalloc/internal \\\n\t$(objroot)include/jemalloc $(objroot)include/jemalloc/internal\nHEADERS = $(wildcard $(foreach dir,$(HEADER_DIRS),$(dir)/*.h))\n$(COBJS) $(CPICOBJS) $(CTESTOBJS): $(HEADERS)\n$(CTESTOBJS): $(objroot)test/jemalloc_test.h\nendif\n\n$(COBJS) $(CPICOBJS) $(CTESTOBJS): %.$(O):\n\t@mkdir -p $(@D)\n\t$(CC) $(CFLAGS) -c $(CPPFLAGS) $(CTARGET) $<\nifdef CC_MM\n\t@$(CC) -MM $(CPPFLAGS) -MT $@ -o $(@:%.$(O)=%.d) $<\nendif\n\nifneq ($(SOREV),$(SO))\n%.$(SO) : %.$(SOREV)\n\t@mkdir -p $(@D)\n\tln -sf $( $(objroot)$${t}.out 2>&1; \\\n\t\t\tif test -e \"$(srcroot)$${t}.exp\"; then \\\n\t\t\t\tdiff -w -u $(srcroot)$${t}.exp \\\n\t\t\t\t $(objroot)$${t}.out >/dev/null 2>&1; \\\n\t\t\t\tfail=$$?; \\\n\t\t\t\tif test \"$${fail}\" -eq \"1\" ; then \\\n\t\t\t\t\tfailures=`expr $${failures} + 1`; \\\n\t\t\t\t\techo \"*** FAIL ***\"; \\\n\t\t\t\telse \\\n\t\t\t\t\techo \"pass\"; \\\n\t\t\t\tfi; \\\n\t\t\telse \\\n\t\t\t\techo \"*** FAIL *** (.exp file is missing)\"; \\\n\t\t\t\tfailures=`expr $${failures} + 1`; \\\n\t\t\tfi; \\\n\t\tdone; \\\n\t\techo \"=========================================\"; \\\n\t\techo \"Failures: $${failures}/$${total}\"'\n\nclean:\n\trm -f $(COBJS)\n\trm -f $(CPICOBJS)\n\trm -f $(COBJS:%.$(O)=%.d)\n\trm -f $(CPICOBJS:%.$(O)=%.d)\n\trm -f $(CTESTOBJS:%.$(O)=%$(EXE))\n\trm -f $(CTESTOBJS)\n\trm -f $(CTESTOBJS:%.$(O)=%.d)\n\trm -f $(CTESTOBJS:%.$(O)=%.out)\n\trm -f $(DSOS) $(STATIC_LIBS)\n\ndistclean: clean\n\trm -rf $(objroot)autom4te.cache\n\trm -f $(objroot)config.log\n\trm -f $(objroot)config.status\n\trm -f $(objroot)config.stamp\n\trm -f $(cfghdrs_out)\n\trm -f $(cfgoutputs_out)\n\nrelclean: distclean\n\trm -f $(objroot)configure\n\trm -f $(srcroot)VERSION\n\trm -f $(DOCS_HTML)\n\trm -f $(DOCS_MAN3)\n\n#===============================================================================\n# Re-configuration rules.\n\nifeq ($(enable_autogen), 1)\n$(srcroot)configure : $(srcroot)configure.ac\n\tcd ./$(srcroot) && $(AUTOCONF)\n\n$(objroot)config.status : $(srcroot)configure\n\t./$(objroot)config.status --recheck\n\n$(srcroot)config.stamp.in : $(srcroot)configure.ac\n\techo stamp > $(srcroot)config.stamp.in\n\n$(objroot)config.stamp : $(cfgoutputs_in) $(cfghdrs_in) $(srcroot)configure\n\t./$(objroot)config.status\n\t@touch $@\n\n# There must be some action in order for make to re-read Makefile when it is\n# out of date.\n$(cfgoutputs_out) $(cfghdrs_out) : $(objroot)config.stamp\n\t@true\nendif\n" }, { "path": "deps/jemalloc/README", "content": "jemalloc is a general-purpose scalable concurrent malloc(3) implementation.\nThis distribution is a \"portable\" implementation that currently targets\nFreeBSD, Linux, Apple OS X, and MinGW. jemalloc is included as the default\nallocator in the FreeBSD and NetBSD operating systems, and it is used by the\nMozilla Firefox web browser on Microsoft Windows-related platforms. Depending\non your needs, one of the other divergent versions may suit your needs better\nthan this distribution.\n\nThe COPYING file contains copyright and licensing information.\n\nThe INSTALL file contains information on how to configure, build, and install\njemalloc.\n\nThe ChangeLog file contains a brief summary of changes for each release.\n\nURL: http://www.canonware.com/jemalloc/\n" }, { "path": "deps/jemalloc/VERSION", "content": "3.2.0-0-g87499f6748ebe4817571e817e9f680ccb5bf54a9\n" }, { "path": "deps/jemalloc/autogen.sh", "content": "#!/bin/sh\n\nfor i in autoconf; do\n echo \"$i\"\n $i\n if [ $? -ne 0 ]; then\n\techo \"Error $? in $i\"\n\texit 1\n fi\ndone\n\necho \"./configure --enable-autogen $@\"\n./configure --enable-autogen $@\nif [ $? -ne 0 ]; then\n echo \"Error $? in ./configure\"\n exit 1\nfi\n" }, { "path": "deps/jemalloc/bin/jemalloc.sh.in", "content": "#!/bin/sh\n\nprefix=@prefix@\nexec_prefix=@exec_prefix@\nlibdir=@libdir@\n\n@LD_PRELOAD_VAR@=${libdir}/libjemalloc.@SOREV@\nexport @LD_PRELOAD_VAR@\nexec \"$@\"\n" }, { "path": "deps/jemalloc/bin/pprof", "content": "#! /usr/bin/env perl\n\n# Copyright (c) 1998-2007, Google Inc.\n# All rights reserved.\n# \n# Redistribution and use in source and binary forms, with or without\n# modification, are permitted provided that the following conditions are\n# met:\n# \n# * Redistributions of source code must retain the above copyright\n# notice, this list of conditions and the following disclaimer.\n# * Redistributions in binary form must reproduce the above\n# copyright notice, this list of conditions and the following disclaimer\n# in the documentation and/or other materials provided with the\n# distribution.\n# * Neither the name of Google Inc. nor the names of its\n# contributors may be used to endorse or promote products derived from\n# this software without specific prior written permission.\n# \n# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\n# \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\n# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\n# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\n# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\n# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\n# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\n# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\n# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n\n# ---\n# Program for printing the profile generated by common/profiler.cc,\n# or by the heap profiler (common/debugallocation.cc)\n#\n# The profile contains a sequence of entries of the form:\n# \n# This program parses the profile, and generates user-readable\n# output.\n#\n# Examples:\n#\n# % tools/pprof \"program\" \"profile\"\n# Enters \"interactive\" mode\n#\n# % tools/pprof --text \"program\" \"profile\"\n# Generates one line per procedure\n#\n# % tools/pprof --gv \"program\" \"profile\"\n# Generates annotated call-graph and displays via \"gv\"\n#\n# % tools/pprof --gv --focus=Mutex \"program\" \"profile\"\n# Restrict to code paths that involve an entry that matches \"Mutex\"\n#\n# % tools/pprof --gv --focus=Mutex --ignore=string \"program\" \"profile\"\n# Restrict to code paths that involve an entry that matches \"Mutex\"\n# and does not match \"string\"\n#\n# % tools/pprof --list=IBF_CheckDocid \"program\" \"profile\"\n# Generates disassembly listing of all routines with at least one\n# sample that match the --list= pattern. The listing is\n# annotated with the flat and cumulative sample counts at each line.\n#\n# % tools/pprof --disasm=IBF_CheckDocid \"program\" \"profile\"\n# Generates disassembly listing of all routines with at least one\n# sample that match the --disasm= pattern. The listing is\n# annotated with the flat and cumulative sample counts at each PC value.\n#\n# TODO: Use color to indicate files?\n\nuse strict;\nuse warnings;\nuse Getopt::Long;\n\nmy $PPROF_VERSION = \"2.0\";\n\n# These are the object tools we use which can come from a\n# user-specified location using --tools, from the PPROF_TOOLS\n# environment variable, or from the environment.\nmy %obj_tool_map = (\n \"objdump\" => \"objdump\",\n \"nm\" => \"nm\",\n \"addr2line\" => \"addr2line\",\n \"c++filt\" => \"c++filt\",\n ## ConfigureObjTools may add architecture-specific entries:\n #\"nm_pdb\" => \"nm-pdb\", # for reading windows (PDB-format) executables\n #\"addr2line_pdb\" => \"addr2line-pdb\", # ditto\n #\"otool\" => \"otool\", # equivalent of objdump on OS X\n);\n# NOTE: these are lists, so you can put in commandline flags if you want.\nmy @DOT = (\"dot\"); # leave non-absolute, since it may be in /usr/local\nmy @GV = (\"gv\");\nmy @EVINCE = (\"evince\"); # could also be xpdf or perhaps acroread\nmy @KCACHEGRIND = (\"kcachegrind\");\nmy @PS2PDF = (\"ps2pdf\");\n# These are used for dynamic profiles\nmy @URL_FETCHER = (\"curl\", \"-s\");\n\n# These are the web pages that servers need to support for dynamic profiles\nmy $HEAP_PAGE = \"/pprof/heap\";\nmy $PROFILE_PAGE = \"/pprof/profile\"; # must support cgi-param \"?seconds=#\"\nmy $PMUPROFILE_PAGE = \"/pprof/pmuprofile(?:\\\\?.*)?\"; # must support cgi-param\n # ?seconds=#&event=x&period=n\nmy $GROWTH_PAGE = \"/pprof/growth\";\nmy $CONTENTION_PAGE = \"/pprof/contention\";\nmy $WALL_PAGE = \"/pprof/wall(?:\\\\?.*)?\"; # accepts options like namefilter\nmy $FILTEREDPROFILE_PAGE = \"/pprof/filteredprofile(?:\\\\?.*)?\";\nmy $CENSUSPROFILE_PAGE = \"/pprof/censusprofile(?:\\\\?.*)?\"; # must support cgi-param\n # \"?seconds=#\",\n # \"?tags_regexp=#\" and\n # \"?type=#\".\nmy $SYMBOL_PAGE = \"/pprof/symbol\"; # must support symbol lookup via POST\nmy $PROGRAM_NAME_PAGE = \"/pprof/cmdline\";\n\n# These are the web pages that can be named on the command line.\n# All the alternatives must begin with /.\nmy $PROFILES = \"($HEAP_PAGE|$PROFILE_PAGE|$PMUPROFILE_PAGE|\" .\n \"$GROWTH_PAGE|$CONTENTION_PAGE|$WALL_PAGE|\" .\n \"$FILTEREDPROFILE_PAGE|$CENSUSPROFILE_PAGE)\";\n\n# default binary name\nmy $UNKNOWN_BINARY = \"(unknown)\";\n\n# There is a pervasive dependency on the length (in hex characters,\n# i.e., nibbles) of an address, distinguishing between 32-bit and\n# 64-bit profiles. To err on the safe size, default to 64-bit here:\nmy $address_length = 16;\n\nmy $dev_null = \"/dev/null\";\nif (! -e $dev_null && $^O =~ /MSWin/) { # $^O is the OS perl was built for\n $dev_null = \"nul\";\n}\n\n# A list of paths to search for shared object files\nmy @prefix_list = ();\n\n# Special routine name that should not have any symbols.\n# Used as separator to parse \"addr2line -i\" output.\nmy $sep_symbol = '_fini';\nmy $sep_address = undef;\n\n##### Argument parsing #####\n\nsub usage_string {\n return < \n is a space separated list of profile names.\npprof [options] \n is a list of profile files where each file contains\n the necessary symbol mappings as well as profile data (likely generated\n with --raw).\npprof [options] \n is a remote form. Symbols are obtained from host:port$SYMBOL_PAGE\n\n Each name can be:\n /path/to/profile - a path to a profile file\n host:port[/] - a location of a service to get profile from\n\n The / can be $HEAP_PAGE, $PROFILE_PAGE, /pprof/pmuprofile,\n $GROWTH_PAGE, $CONTENTION_PAGE, /pprof/wall,\n $CENSUSPROFILE_PAGE, or /pprof/filteredprofile.\n For instance:\n pprof http://myserver.com:80$HEAP_PAGE\n If / is omitted, the service defaults to $PROFILE_PAGE (cpu profiling).\npprof --symbols \n Maps addresses to symbol names. In this mode, stdin should be a\n list of library mappings, in the same format as is found in the heap-\n and cpu-profile files (this loosely matches that of /proc/self/maps\n on linux), followed by a list of hex addresses to map, one per line.\n\n For more help with querying remote servers, including how to add the\n necessary server-side support code, see this filename (or one like it):\n\n /usr/doc/gperftools-$PPROF_VERSION/pprof_remote_servers.html\n\nOptions:\n --cum Sort by cumulative data\n --base= Subtract from before display\n --interactive Run in interactive mode (interactive \"help\" gives help) [default]\n --seconds= Length of time for dynamic profiles [default=30 secs]\n --add_lib= Read additional symbols and line info from the given library\n --lib_prefix= Comma separated list of library path prefixes\n\nReporting Granularity:\n --addresses Report at address level\n --lines Report at source line level\n --functions Report at function level [default]\n --files Report at source file level\n\nOutput type:\n --text Generate text report\n --callgrind Generate callgrind format to stdout\n --gv Generate Postscript and display\n --evince Generate PDF and display\n --web Generate SVG and display\n --list= Generate source listing of matching routines\n --disasm= Generate disassembly of matching routines\n --symbols Print demangled symbol names found at given addresses\n --dot Generate DOT file to stdout\n --ps Generate Postcript to stdout\n --pdf Generate PDF to stdout\n --svg Generate SVG to stdout\n --gif Generate GIF to stdout\n --raw Generate symbolized pprof data (useful with remote fetch)\n\nHeap-Profile Options:\n --inuse_space Display in-use (mega)bytes [default]\n --inuse_objects Display in-use objects\n --alloc_space Display allocated (mega)bytes\n --alloc_objects Display allocated objects\n --show_bytes Display space in bytes\n --drop_negative Ignore negative differences\n\nContention-profile options:\n --total_delay Display total delay at each region [default]\n --contentions Display number of delays at each region\n --mean_delay Display mean delay at each region\n\nCall-graph Options:\n --nodecount= Show at most so many nodes [default=80]\n --nodefraction= Hide nodes below *total [default=.005]\n --edgefraction= Hide edges below *total [default=.001]\n --maxdegree= Max incoming/outgoing edges per node [default=8]\n --focus= Focus on nodes matching \n --ignore= Ignore nodes matching \n --scale= Set GV scaling [default=0]\n --heapcheck Make nodes with non-0 object counts\n (i.e. direct leak generators) more visible\n\nMiscellaneous:\n --tools=[,...] \\$PATH for object tool pathnames\n --test Run unit tests\n --help This message\n --version Version information\n\nEnvironment Variables:\n PPROF_TMPDIR Profiles directory. Defaults to \\$HOME/pprof\n PPROF_TOOLS Prefix for object tools pathnames\n\nExamples:\n\npprof /bin/ls ls.prof\n Enters \"interactive\" mode\npprof --text /bin/ls ls.prof\n Outputs one line per procedure\npprof --web /bin/ls ls.prof\n Displays annotated call-graph in web browser\npprof --gv /bin/ls ls.prof\n Displays annotated call-graph via 'gv'\npprof --gv --focus=Mutex /bin/ls ls.prof\n Restricts to code paths including a .*Mutex.* entry\npprof --gv --focus=Mutex --ignore=string /bin/ls ls.prof\n Code paths including Mutex but not string\npprof --list=getdir /bin/ls ls.prof\n (Per-line) annotated source listing for getdir()\npprof --disasm=getdir /bin/ls ls.prof\n (Per-PC) annotated disassembly for getdir()\n\npprof http://localhost:1234/\n Enters \"interactive\" mode\npprof --text localhost:1234\n Outputs one line per procedure for localhost:1234\npprof --raw localhost:1234 > ./local.raw\npprof --text ./local.raw\n Fetches a remote profile for later analysis and then\n analyzes it in text mode.\nEOF\n}\n\nsub version_string {\n return < \\$main::opt_help,\n \"version!\" => \\$main::opt_version,\n \"cum!\" => \\$main::opt_cum,\n \"base=s\" => \\$main::opt_base,\n \"seconds=i\" => \\$main::opt_seconds,\n \"add_lib=s\" => \\$main::opt_lib,\n \"lib_prefix=s\" => \\$main::opt_lib_prefix,\n \"functions!\" => \\$main::opt_functions,\n \"lines!\" => \\$main::opt_lines,\n \"addresses!\" => \\$main::opt_addresses,\n \"files!\" => \\$main::opt_files,\n \"text!\" => \\$main::opt_text,\n \"callgrind!\" => \\$main::opt_callgrind,\n \"list=s\" => \\$main::opt_list,\n \"disasm=s\" => \\$main::opt_disasm,\n \"symbols!\" => \\$main::opt_symbols,\n \"gv!\" => \\$main::opt_gv,\n \"evince!\" => \\$main::opt_evince,\n \"web!\" => \\$main::opt_web,\n \"dot!\" => \\$main::opt_dot,\n \"ps!\" => \\$main::opt_ps,\n \"pdf!\" => \\$main::opt_pdf,\n \"svg!\" => \\$main::opt_svg,\n \"gif!\" => \\$main::opt_gif,\n \"raw!\" => \\$main::opt_raw,\n \"interactive!\" => \\$main::opt_interactive,\n \"nodecount=i\" => \\$main::opt_nodecount,\n \"nodefraction=f\" => \\$main::opt_nodefraction,\n \"edgefraction=f\" => \\$main::opt_edgefraction,\n \"maxdegree=i\" => \\$main::opt_maxdegree,\n \"focus=s\" => \\$main::opt_focus,\n \"ignore=s\" => \\$main::opt_ignore,\n \"scale=i\" => \\$main::opt_scale,\n \"heapcheck\" => \\$main::opt_heapcheck,\n \"inuse_space!\" => \\$main::opt_inuse_space,\n \"inuse_objects!\" => \\$main::opt_inuse_objects,\n \"alloc_space!\" => \\$main::opt_alloc_space,\n \"alloc_objects!\" => \\$main::opt_alloc_objects,\n \"show_bytes!\" => \\$main::opt_show_bytes,\n \"drop_negative!\" => \\$main::opt_drop_negative,\n \"total_delay!\" => \\$main::opt_total_delay,\n \"contentions!\" => \\$main::opt_contentions,\n \"mean_delay!\" => \\$main::opt_mean_delay,\n \"tools=s\" => \\$main::opt_tools,\n \"test!\" => \\$main::opt_test,\n \"debug!\" => \\$main::opt_debug,\n # Undocumented flags used only by unittests:\n \"test_stride=i\" => \\$main::opt_test_stride,\n ) || usage(\"Invalid option(s)\");\n\n # Deal with the standard --help and --version\n if ($main::opt_help) {\n print usage_string();\n exit(0);\n }\n\n if ($main::opt_version) {\n print version_string();\n exit(0);\n }\n\n # Disassembly/listing/symbols mode requires address-level info\n if ($main::opt_disasm || $main::opt_list || $main::opt_symbols) {\n $main::opt_functions = 0;\n $main::opt_lines = 0;\n $main::opt_addresses = 1;\n $main::opt_files = 0;\n }\n\n # Check heap-profiling flags\n if ($main::opt_inuse_space +\n $main::opt_inuse_objects +\n $main::opt_alloc_space +\n $main::opt_alloc_objects > 1) {\n usage(\"Specify at most on of --inuse/--alloc options\");\n }\n\n # Check output granularities\n my $grains =\n $main::opt_functions +\n $main::opt_lines +\n $main::opt_addresses +\n $main::opt_files +\n 0;\n if ($grains > 1) {\n usage(\"Only specify one output granularity option\");\n }\n if ($grains == 0) {\n $main::opt_functions = 1;\n }\n\n # Check output modes\n my $modes =\n $main::opt_text +\n $main::opt_callgrind +\n ($main::opt_list eq '' ? 0 : 1) +\n ($main::opt_disasm eq '' ? 0 : 1) +\n ($main::opt_symbols == 0 ? 0 : 1) +\n $main::opt_gv +\n $main::opt_evince +\n $main::opt_web +\n $main::opt_dot +\n $main::opt_ps +\n $main::opt_pdf +\n $main::opt_svg +\n $main::opt_gif +\n $main::opt_raw +\n $main::opt_interactive +\n 0;\n if ($modes > 1) {\n usage(\"Only specify one output mode\");\n }\n if ($modes == 0) {\n if (-t STDOUT) { # If STDOUT is a tty, activate interactive mode\n $main::opt_interactive = 1;\n } else {\n $main::opt_text = 1;\n }\n }\n\n if ($main::opt_test) {\n RunUnitTests();\n # Should not return\n exit(1);\n }\n\n # Binary name and profile arguments list\n $main::prog = \"\";\n @main::pfile_args = ();\n\n # Remote profiling without a binary (using $SYMBOL_PAGE instead)\n if (@ARGV > 0) {\n if (IsProfileURL($ARGV[0])) {\n $main::use_symbol_page = 1;\n } elsif (IsSymbolizedProfileFile($ARGV[0])) {\n $main::use_symbolized_profile = 1;\n $main::prog = $UNKNOWN_BINARY; # will be set later from the profile file\n }\n }\n\n if ($main::use_symbol_page || $main::use_symbolized_profile) {\n # We don't need a binary!\n my %disabled = ('--lines' => $main::opt_lines,\n '--disasm' => $main::opt_disasm);\n for my $option (keys %disabled) {\n usage(\"$option cannot be used without a binary\") if $disabled{$option};\n }\n # Set $main::prog later...\n scalar(@ARGV) || usage(\"Did not specify profile file\");\n } elsif ($main::opt_symbols) {\n # --symbols needs a binary-name (to run nm on, etc) but not profiles\n $main::prog = shift(@ARGV) || usage(\"Did not specify program\");\n } else {\n $main::prog = shift(@ARGV) || usage(\"Did not specify program\");\n scalar(@ARGV) || usage(\"Did not specify profile file\");\n }\n\n # Parse profile file/location arguments\n foreach my $farg (@ARGV) {\n if ($farg =~ m/(.*)\\@([0-9]+)(|\\/.*)$/ ) {\n my $machine = $1;\n my $num_machines = $2;\n my $path = $3;\n for (my $i = 0; $i < $num_machines; $i++) {\n unshift(@main::pfile_args, \"$i.$machine$path\");\n }\n } else {\n unshift(@main::pfile_args, $farg);\n }\n }\n\n if ($main::use_symbol_page) {\n unless (IsProfileURL($main::pfile_args[0])) {\n error(\"The first profile should be a remote form to use $SYMBOL_PAGE\\n\");\n }\n CheckSymbolPage();\n $main::prog = FetchProgramName();\n } elsif (!$main::use_symbolized_profile) { # may not need objtools!\n ConfigureObjTools($main::prog)\n }\n\n # Break the opt_lib_prefix into the prefix_list array\n @prefix_list = split (',', $main::opt_lib_prefix);\n\n # Remove trailing / from the prefixes, in the list to prevent\n # searching things like /my/path//lib/mylib.so\n foreach (@prefix_list) {\n s|/+$||;\n }\n}\n\nsub Main() {\n Init();\n $main::collected_profile = undef;\n @main::profile_files = ();\n $main::op_time = time();\n\n # Printing symbols is special and requires a lot less info that most.\n if ($main::opt_symbols) {\n PrintSymbols(*STDIN); # Get /proc/maps and symbols output from stdin\n return;\n }\n\n # Fetch all profile data\n FetchDynamicProfiles();\n\n # this will hold symbols that we read from the profile files\n my $symbol_map = {};\n\n # Read one profile, pick the last item on the list\n my $data = ReadProfile($main::prog, pop(@main::profile_files));\n my $profile = $data->{profile};\n my $pcs = $data->{pcs};\n my $libs = $data->{libs}; # Info about main program and shared libraries\n $symbol_map = MergeSymbols($symbol_map, $data->{symbols});\n\n # Add additional profiles, if available.\n if (scalar(@main::profile_files) > 0) {\n foreach my $pname (@main::profile_files) {\n my $data2 = ReadProfile($main::prog, $pname);\n $profile = AddProfile($profile, $data2->{profile});\n $pcs = AddPcs($pcs, $data2->{pcs});\n $symbol_map = MergeSymbols($symbol_map, $data2->{symbols});\n }\n }\n\n # Subtract base from profile, if specified\n if ($main::opt_base ne '') {\n my $base = ReadProfile($main::prog, $main::opt_base);\n $profile = SubtractProfile($profile, $base->{profile});\n $pcs = AddPcs($pcs, $base->{pcs});\n $symbol_map = MergeSymbols($symbol_map, $base->{symbols});\n }\n\n # Get total data in profile\n my $total = TotalProfile($profile);\n\n # Collect symbols\n my $symbols;\n if ($main::use_symbolized_profile) {\n $symbols = FetchSymbols($pcs, $symbol_map);\n } elsif ($main::use_symbol_page) {\n $symbols = FetchSymbols($pcs);\n } else {\n # TODO(csilvers): $libs uses the /proc/self/maps data from profile1,\n # which may differ from the data from subsequent profiles, especially\n # if they were run on different machines. Use appropriate libs for\n # each pc somehow.\n $symbols = ExtractSymbols($libs, $pcs);\n }\n\n # Remove uniniteresting stack items\n $profile = RemoveUninterestingFrames($symbols, $profile);\n\n # Focus?\n if ($main::opt_focus ne '') {\n $profile = FocusProfile($symbols, $profile, $main::opt_focus);\n }\n\n # Ignore?\n if ($main::opt_ignore ne '') {\n $profile = IgnoreProfile($symbols, $profile, $main::opt_ignore);\n }\n\n my $calls = ExtractCalls($symbols, $profile);\n\n # Reduce profiles to required output granularity, and also clean\n # each stack trace so a given entry exists at most once.\n my $reduced = ReduceProfile($symbols, $profile);\n\n # Get derived profiles\n my $flat = FlatProfile($reduced);\n my $cumulative = CumulativeProfile($reduced);\n\n # Print\n if (!$main::opt_interactive) {\n if ($main::opt_disasm) {\n PrintDisassembly($libs, $flat, $cumulative, $main::opt_disasm);\n } elsif ($main::opt_list) {\n PrintListing($total, $libs, $flat, $cumulative, $main::opt_list, 0);\n } elsif ($main::opt_text) {\n # Make sure the output is empty when have nothing to report\n # (only matters when --heapcheck is given but we must be\n # compatible with old branches that did not pass --heapcheck always):\n if ($total != 0) {\n printf(\"Total: %s %s\\n\", Unparse($total), Units());\n }\n PrintText($symbols, $flat, $cumulative, -1);\n } elsif ($main::opt_raw) {\n PrintSymbolizedProfile($symbols, $profile, $main::prog);\n } elsif ($main::opt_callgrind) {\n PrintCallgrind($calls);\n } else {\n if (PrintDot($main::prog, $symbols, $profile, $flat, $cumulative, $total)) {\n if ($main::opt_gv) {\n RunGV(TempName($main::next_tmpfile, \"ps\"), \"\");\n } elsif ($main::opt_evince) {\n RunEvince(TempName($main::next_tmpfile, \"pdf\"), \"\");\n } elsif ($main::opt_web) {\n my $tmp = TempName($main::next_tmpfile, \"svg\");\n RunWeb($tmp);\n # The command we run might hand the file name off\n # to an already running browser instance and then exit.\n # Normally, we'd remove $tmp on exit (right now),\n # but fork a child to remove $tmp a little later, so that the\n # browser has time to load it first.\n delete $main::tempnames{$tmp};\n if (fork() == 0) {\n sleep 5;\n unlink($tmp);\n exit(0);\n }\n }\n } else {\n cleanup();\n exit(1);\n }\n }\n } else {\n InteractiveMode($profile, $symbols, $libs, $total);\n }\n\n cleanup();\n exit(0);\n}\n\n##### Entry Point #####\n\nMain();\n\n# Temporary code to detect if we're running on a Goobuntu system.\n# These systems don't have the right stuff installed for the special\n# Readline libraries to work, so as a temporary workaround, we default\n# to using the normal stdio code, rather than the fancier readline-based\n# code\nsub ReadlineMightFail {\n if (-e '/lib/libtermcap.so.2') {\n return 0; # libtermcap exists, so readline should be okay\n } else {\n return 1;\n }\n}\n\nsub RunGV {\n my $fname = shift;\n my $bg = shift; # \"\" or \" &\" if we should run in background\n if (!system(ShellEscape(@GV, \"--version\") . \" >$dev_null 2>&1\")) {\n # Options using double dash are supported by this gv version.\n # Also, turn on noantialias to better handle bug in gv for\n # postscript files with large dimensions.\n # TODO: Maybe we should not pass the --noantialias flag\n # if the gv version is known to work properly without the flag.\n system(ShellEscape(@GV, \"--scale=$main::opt_scale\", \"--noantialias\", $fname)\n . $bg);\n } else {\n # Old gv version - only supports options that use single dash.\n print STDERR ShellEscape(@GV, \"-scale\", $main::opt_scale) . \"\\n\";\n system(ShellEscape(@GV, \"-scale\", \"$main::opt_scale\", $fname) . $bg);\n }\n}\n\nsub RunEvince {\n my $fname = shift;\n my $bg = shift; # \"\" or \" &\" if we should run in background\n system(ShellEscape(@EVINCE, $fname) . $bg);\n}\n\nsub RunWeb {\n my $fname = shift;\n print STDERR \"Loading web page file:///$fname\\n\";\n\n if (`uname` =~ /Darwin/) {\n # OS X: open will use standard preference for SVG files.\n system(\"/usr/bin/open\", $fname);\n return;\n }\n\n # Some kind of Unix; try generic symlinks, then specific browsers.\n # (Stop once we find one.)\n # Works best if the browser is already running.\n my @alt = (\n \"/etc/alternatives/gnome-www-browser\",\n \"/etc/alternatives/x-www-browser\",\n \"google-chrome\",\n \"firefox\",\n );\n foreach my $b (@alt) {\n if (system($b, $fname) == 0) {\n return;\n }\n }\n\n print STDERR \"Could not load web browser.\\n\";\n}\n\nsub RunKcachegrind {\n my $fname = shift;\n my $bg = shift; # \"\" or \" &\" if we should run in background\n print STDERR \"Starting '@KCACHEGRIND \" . $fname . $bg . \"'\\n\";\n system(ShellEscape(@KCACHEGRIND, $fname) . $bg);\n}\n\n\n##### Interactive helper routines #####\n\nsub InteractiveMode {\n $| = 1; # Make output unbuffered for interactive mode\n my ($orig_profile, $symbols, $libs, $total) = @_;\n\n print STDERR \"Welcome to pprof! For help, type 'help'.\\n\";\n\n # Use ReadLine if it's installed and input comes from a console.\n if ( -t STDIN &&\n !ReadlineMightFail() &&\n defined(eval {require Term::ReadLine}) ) {\n my $term = new Term::ReadLine 'pprof';\n while ( defined ($_ = $term->readline('(pprof) '))) {\n $term->addhistory($_) if /\\S/;\n if (!InteractiveCommand($orig_profile, $symbols, $libs, $total, $_)) {\n last; # exit when we get an interactive command to quit\n }\n }\n } else { # don't have readline\n while (1) {\n print STDERR \"(pprof) \";\n $_ = ;\n last if ! defined $_ ;\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n\n # Save some flags that might be reset by InteractiveCommand()\n my $save_opt_lines = $main::opt_lines;\n\n if (!InteractiveCommand($orig_profile, $symbols, $libs, $total, $_)) {\n last; # exit when we get an interactive command to quit\n }\n\n # Restore flags\n $main::opt_lines = $save_opt_lines;\n }\n }\n}\n\n# Takes two args: orig profile, and command to run.\n# Returns 1 if we should keep going, or 0 if we were asked to quit\nsub InteractiveCommand {\n my($orig_profile, $symbols, $libs, $total, $command) = @_;\n $_ = $command; # just to make future m//'s easier\n if (!defined($_)) {\n print STDERR \"\\n\";\n return 0;\n }\n if (m/^\\s*quit/) {\n return 0;\n }\n if (m/^\\s*help/) {\n InteractiveHelpMessage();\n return 1;\n }\n # Clear all the mode options -- mode is controlled by \"$command\"\n $main::opt_text = 0;\n $main::opt_callgrind = 0;\n $main::opt_disasm = 0;\n $main::opt_list = 0;\n $main::opt_gv = 0;\n $main::opt_evince = 0;\n $main::opt_cum = 0;\n\n if (m/^\\s*(text|top)(\\d*)\\s*(.*)/) {\n $main::opt_text = 1;\n\n my $line_limit = ($2 ne \"\") ? int($2) : 10;\n\n my $routine;\n my $ignore;\n ($routine, $ignore) = ParseInteractiveArgs($3);\n\n my $profile = ProcessProfile($total, $orig_profile, $symbols, \"\", $ignore);\n my $reduced = ReduceProfile($symbols, $profile);\n\n # Get derived profiles\n my $flat = FlatProfile($reduced);\n my $cumulative = CumulativeProfile($reduced);\n\n PrintText($symbols, $flat, $cumulative, $line_limit);\n return 1;\n }\n if (m/^\\s*callgrind\\s*([^ \\n]*)/) {\n $main::opt_callgrind = 1;\n\n # Get derived profiles\n my $calls = ExtractCalls($symbols, $orig_profile);\n my $filename = $1;\n if ( $1 eq '' ) {\n $filename = TempName($main::next_tmpfile, \"callgrind\");\n }\n PrintCallgrind($calls, $filename);\n if ( $1 eq '' ) {\n RunKcachegrind($filename, \" & \");\n $main::next_tmpfile++;\n }\n\n return 1;\n }\n if (m/^\\s*(web)?list\\s*(.+)/) {\n my $html = (defined($1) && ($1 eq \"web\"));\n $main::opt_list = 1;\n\n my $routine;\n my $ignore;\n ($routine, $ignore) = ParseInteractiveArgs($2);\n\n my $profile = ProcessProfile($total, $orig_profile, $symbols, \"\", $ignore);\n my $reduced = ReduceProfile($symbols, $profile);\n\n # Get derived profiles\n my $flat = FlatProfile($reduced);\n my $cumulative = CumulativeProfile($reduced);\n\n PrintListing($total, $libs, $flat, $cumulative, $routine, $html);\n return 1;\n }\n if (m/^\\s*disasm\\s*(.+)/) {\n $main::opt_disasm = 1;\n\n my $routine;\n my $ignore;\n ($routine, $ignore) = ParseInteractiveArgs($1);\n\n # Process current profile to account for various settings\n my $profile = ProcessProfile($total, $orig_profile, $symbols, \"\", $ignore);\n my $reduced = ReduceProfile($symbols, $profile);\n\n # Get derived profiles\n my $flat = FlatProfile($reduced);\n my $cumulative = CumulativeProfile($reduced);\n\n PrintDisassembly($libs, $flat, $cumulative, $routine);\n return 1;\n }\n if (m/^\\s*(gv|web|evince)\\s*(.*)/) {\n $main::opt_gv = 0;\n $main::opt_evince = 0;\n $main::opt_web = 0;\n if ($1 eq \"gv\") {\n $main::opt_gv = 1;\n } elsif ($1 eq \"evince\") {\n $main::opt_evince = 1;\n } elsif ($1 eq \"web\") {\n $main::opt_web = 1;\n }\n\n my $focus;\n my $ignore;\n ($focus, $ignore) = ParseInteractiveArgs($2);\n\n # Process current profile to account for various settings\n my $profile = ProcessProfile($total, $orig_profile, $symbols,\n $focus, $ignore);\n my $reduced = ReduceProfile($symbols, $profile);\n\n # Get derived profiles\n my $flat = FlatProfile($reduced);\n my $cumulative = CumulativeProfile($reduced);\n\n if (PrintDot($main::prog, $symbols, $profile, $flat, $cumulative, $total)) {\n if ($main::opt_gv) {\n RunGV(TempName($main::next_tmpfile, \"ps\"), \" &\");\n } elsif ($main::opt_evince) {\n RunEvince(TempName($main::next_tmpfile, \"pdf\"), \" &\");\n } elsif ($main::opt_web) {\n RunWeb(TempName($main::next_tmpfile, \"svg\"));\n }\n $main::next_tmpfile++;\n }\n return 1;\n }\n if (m/^\\s*$/) {\n return 1;\n }\n print STDERR \"Unknown command: try 'help'.\\n\";\n return 1;\n}\n\n\nsub ProcessProfile {\n my $total_count = shift;\n my $orig_profile = shift;\n my $symbols = shift;\n my $focus = shift;\n my $ignore = shift;\n\n # Process current profile to account for various settings\n my $profile = $orig_profile;\n printf(\"Total: %s %s\\n\", Unparse($total_count), Units());\n if ($focus ne '') {\n $profile = FocusProfile($symbols, $profile, $focus);\n my $focus_count = TotalProfile($profile);\n printf(\"After focusing on '%s': %s %s of %s (%0.1f%%)\\n\",\n $focus,\n Unparse($focus_count), Units(),\n Unparse($total_count), ($focus_count*100.0) / $total_count);\n }\n if ($ignore ne '') {\n $profile = IgnoreProfile($symbols, $profile, $ignore);\n my $ignore_count = TotalProfile($profile);\n printf(\"After ignoring '%s': %s %s of %s (%0.1f%%)\\n\",\n $ignore,\n Unparse($ignore_count), Units(),\n Unparse($total_count),\n ($ignore_count*100.0) / $total_count);\n }\n\n return $profile;\n}\n\nsub InteractiveHelpMessage {\n print STDERR <{$k};\n my @addrs = split(/\\n/, $k);\n if ($#addrs >= 0) {\n my $depth = $#addrs + 1;\n # int(foo / 2**32) is the only reliable way to get rid of bottom\n # 32 bits on both 32- and 64-bit systems.\n print pack('L*', $count & 0xFFFFFFFF, int($count / 2**32));\n print pack('L*', $depth & 0xFFFFFFFF, int($depth / 2**32));\n\n foreach my $full_addr (@addrs) {\n my $addr = $full_addr;\n $addr =~ s/0x0*//; # strip off leading 0x, zeroes\n if (length($addr) > 16) {\n print STDERR \"Invalid address in profile: $full_addr\\n\";\n next;\n }\n my $low_addr = substr($addr, -8); # get last 8 hex chars\n my $high_addr = substr($addr, -16, 8); # get up to 8 more hex chars\n print pack('L*', hex('0x' . $low_addr), hex('0x' . $high_addr));\n }\n }\n }\n}\n\n# Print symbols and profile data\nsub PrintSymbolizedProfile {\n my $symbols = shift;\n my $profile = shift;\n my $prog = shift;\n\n $SYMBOL_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $symbol_marker = $&;\n\n print '--- ', $symbol_marker, \"\\n\";\n if (defined($prog)) {\n print 'binary=', $prog, \"\\n\";\n }\n while (my ($pc, $name) = each(%{$symbols})) {\n my $sep = ' ';\n print '0x', $pc;\n # We have a list of function names, which include the inlined\n # calls. They are separated (and terminated) by --, which is\n # illegal in function names.\n for (my $j = 2; $j <= $#{$name}; $j += 3) {\n print $sep, $name->[$j];\n $sep = '--';\n }\n print \"\\n\";\n }\n print '---', \"\\n\";\n\n $PROFILE_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $profile_marker = $&;\n print '--- ', $profile_marker, \"\\n\";\n if (defined($main::collected_profile)) {\n # if used with remote fetch, simply dump the collected profile to output.\n open(SRC, \"<$main::collected_profile\");\n while () {\n print $_;\n }\n close(SRC);\n } else {\n # dump a cpu-format profile to standard out\n PrintProfileData($profile);\n }\n}\n\n# Print text output\nsub PrintText {\n my $symbols = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $line_limit = shift;\n\n my $total = TotalProfile($flat);\n\n # Which profile to sort by?\n my $s = $main::opt_cum ? $cumulative : $flat;\n\n my $running_sum = 0;\n my $lines = 0;\n foreach my $k (sort { GetEntry($s, $b) <=> GetEntry($s, $a) || $a cmp $b }\n keys(%{$cumulative})) {\n my $f = GetEntry($flat, $k);\n my $c = GetEntry($cumulative, $k);\n $running_sum += $f;\n\n my $sym = $k;\n if (exists($symbols->{$k})) {\n $sym = $symbols->{$k}->[0] . \" \" . $symbols->{$k}->[1];\n if ($main::opt_addresses) {\n $sym = $k . \" \" . $sym;\n }\n }\n\n if ($f != 0 || $c != 0) {\n printf(\"%8s %6s %6s %8s %6s %s\\n\",\n Unparse($f),\n Percent($f, $total),\n Percent($running_sum, $total),\n Unparse($c),\n Percent($c, $total),\n $sym);\n }\n $lines++;\n last if ($line_limit >= 0 && $lines >= $line_limit);\n }\n}\n\n# Callgrind format has a compression for repeated function and file\n# names. You show the name the first time, and just use its number\n# subsequently. This can cut down the file to about a third or a\n# quarter of its uncompressed size. $key and $val are the key/value\n# pair that would normally be printed by callgrind; $map is a map from\n# value to number.\nsub CompressedCGName {\n my($key, $val, $map) = @_;\n my $idx = $map->{$val};\n # For very short keys, providing an index hurts rather than helps.\n if (length($val) <= 3) {\n return \"$key=$val\\n\";\n } elsif (defined($idx)) {\n return \"$key=($idx)\\n\";\n } else {\n # scalar(keys $map) gives the number of items in the map.\n $idx = scalar(keys(%{$map})) + 1;\n $map->{$val} = $idx;\n return \"$key=($idx) $val\\n\";\n }\n}\n\n# Print the call graph in a way that's suiteable for callgrind.\nsub PrintCallgrind {\n my $calls = shift;\n my $filename;\n my %filename_to_index_map;\n my %fnname_to_index_map;\n\n if ($main::opt_interactive) {\n $filename = shift;\n print STDERR \"Writing callgrind file to '$filename'.\\n\"\n } else {\n $filename = \"&STDOUT\";\n }\n open(CG, \">$filename\");\n printf CG (\"events: Hits\\n\\n\");\n foreach my $call ( map { $_->[0] }\n sort { $a->[1] cmp $b ->[1] ||\n $a->[2] <=> $b->[2] }\n map { /([^:]+):(\\d+):([^ ]+)( -> ([^:]+):(\\d+):(.+))?/;\n [$_, $1, $2] }\n keys %$calls ) {\n my $count = int($calls->{$call});\n $call =~ /([^:]+):(\\d+):([^ ]+)( -> ([^:]+):(\\d+):(.+))?/;\n my ( $caller_file, $caller_line, $caller_function,\n $callee_file, $callee_line, $callee_function ) =\n ( $1, $2, $3, $5, $6, $7 );\n\n # TODO(csilvers): for better compression, collect all the\n # caller/callee_files and functions first, before printing\n # anything, and only compress those referenced more than once.\n printf CG CompressedCGName(\"fl\", $caller_file, \\%filename_to_index_map);\n printf CG CompressedCGName(\"fn\", $caller_function, \\%fnname_to_index_map);\n if (defined $6) {\n printf CG CompressedCGName(\"cfl\", $callee_file, \\%filename_to_index_map);\n printf CG CompressedCGName(\"cfn\", $callee_function, \\%fnname_to_index_map);\n printf CG (\"calls=$count $callee_line\\n\");\n }\n printf CG (\"$caller_line $count\\n\\n\");\n }\n}\n\n# Print disassembly for all all routines that match $main::opt_disasm\nsub PrintDisassembly {\n my $libs = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $disasm_opts = shift;\n\n my $total = TotalProfile($flat);\n\n foreach my $lib (@{$libs}) {\n my $symbol_table = GetProcedureBoundaries($lib->[0], $disasm_opts);\n my $offset = AddressSub($lib->[1], $lib->[3]);\n foreach my $routine (sort ByName keys(%{$symbol_table})) {\n my $start_addr = $symbol_table->{$routine}->[0];\n my $end_addr = $symbol_table->{$routine}->[1];\n # See if there are any samples in this routine\n my $length = hex(AddressSub($end_addr, $start_addr));\n my $addr = AddressAdd($start_addr, $offset);\n for (my $i = 0; $i < $length; $i++) {\n if (defined($cumulative->{$addr})) {\n PrintDisassembledFunction($lib->[0], $offset,\n $routine, $flat, $cumulative,\n $start_addr, $end_addr, $total);\n last;\n }\n $addr = AddressInc($addr);\n }\n }\n }\n}\n\n# Return reference to array of tuples of the form:\n# [start_address, filename, linenumber, instruction, limit_address]\n# E.g.,\n# [\"0x806c43d\", \"/foo/bar.cc\", 131, \"ret\", \"0x806c440\"]\nsub Disassemble {\n my $prog = shift;\n my $offset = shift;\n my $start_addr = shift;\n my $end_addr = shift;\n\n my $objdump = $obj_tool_map{\"objdump\"};\n my $cmd = ShellEscape($objdump, \"-C\", \"-d\", \"-l\", \"--no-show-raw-insn\",\n \"--start-address=0x$start_addr\",\n \"--stop-address=0x$end_addr\", $prog);\n open(OBJDUMP, \"$cmd |\") || error(\"$cmd: $!\\n\");\n my @result = ();\n my $filename = \"\";\n my $linenumber = -1;\n my $last = [\"\", \"\", \"\", \"\"];\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n chop;\n if (m|\\s*([^:\\s]+):(\\d+)\\s*$|) {\n # Location line of the form:\n # :\n $filename = $1;\n $linenumber = $2;\n } elsif (m/^ +([0-9a-f]+):\\s*(.*)/) {\n # Disassembly line -- zero-extend address to full length\n my $addr = HexExtend($1);\n my $k = AddressAdd($addr, $offset);\n $last->[4] = $k; # Store ending address for previous instruction\n $last = [$k, $filename, $linenumber, $2, $end_addr];\n push(@result, $last);\n }\n }\n close(OBJDUMP);\n return @result;\n}\n\n# The input file should contain lines of the form /proc/maps-like\n# output (same format as expected from the profiles) or that looks\n# like hex addresses (like \"0xDEADBEEF\"). We will parse all\n# /proc/maps output, and for all the hex addresses, we will output\n# \"short\" symbol names, one per line, in the same order as the input.\nsub PrintSymbols {\n my $maps_and_symbols_file = shift;\n\n # ParseLibraries expects pcs to be in a set. Fine by us...\n my @pclist = (); # pcs in sorted order\n my $pcs = {};\n my $map = \"\";\n foreach my $line (<$maps_and_symbols_file>) {\n $line =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n if ($line =~ /\\b(0x[0-9a-f]+)\\b/i) {\n push(@pclist, HexExtend($1));\n $pcs->{$pclist[-1]} = 1;\n } else {\n $map .= $line;\n }\n }\n\n my $libs = ParseLibraries($main::prog, $map, $pcs);\n my $symbols = ExtractSymbols($libs, $pcs);\n\n foreach my $pc (@pclist) {\n # ->[0] is the shortname, ->[2] is the full name\n print(($symbols->{$pc}->[0] || \"??\") . \"\\n\");\n }\n}\n\n\n# For sorting functions by name\nsub ByName {\n return ShortFunctionName($a) cmp ShortFunctionName($b);\n}\n\n# Print source-listing for all all routines that match $list_opts\nsub PrintListing {\n my $total = shift;\n my $libs = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $list_opts = shift;\n my $html = shift;\n\n my $output = \\*STDOUT;\n my $fname = \"\";\n\n if ($html) {\n # Arrange to write the output to a temporary file\n $fname = TempName($main::next_tmpfile, \"html\");\n $main::next_tmpfile++;\n if (!open(TEMP, \">$fname\")) {\n print STDERR \"$fname: $!\\n\";\n return;\n }\n $output = \\*TEMP;\n print $output HtmlListingHeader();\n printf $output (\"
%s
Total: %s %s
\\n\",\n $main::prog, Unparse($total), Units());\n }\n\n my $listed = 0;\n foreach my $lib (@{$libs}) {\n my $symbol_table = GetProcedureBoundaries($lib->[0], $list_opts);\n my $offset = AddressSub($lib->[1], $lib->[3]);\n foreach my $routine (sort ByName keys(%{$symbol_table})) {\n # Print if there are any samples in this routine\n my $start_addr = $symbol_table->{$routine}->[0];\n my $end_addr = $symbol_table->{$routine}->[1];\n my $length = hex(AddressSub($end_addr, $start_addr));\n my $addr = AddressAdd($start_addr, $offset);\n for (my $i = 0; $i < $length; $i++) {\n if (defined($cumulative->{$addr})) {\n $listed += PrintSource(\n $lib->[0], $offset,\n $routine, $flat, $cumulative,\n $start_addr, $end_addr,\n $html,\n $output);\n last;\n }\n $addr = AddressInc($addr);\n }\n }\n }\n\n if ($html) {\n if ($listed > 0) {\n print $output HtmlListingFooter();\n close($output);\n RunWeb($fname);\n } else {\n close($output);\n unlink($fname);\n }\n }\n}\n\nsub HtmlListingHeader {\n return <<'EOF';\n\n\n\nPprof listing\n\n\n\n\nEOF\n}\n\nsub HtmlListingFooter {\n return <<'EOF';\n\n\nEOF\n}\n\nsub HtmlEscape {\n my $text = shift;\n $text =~ s/&/&/g;\n $text =~ s//>/g;\n return $text;\n}\n\n# Returns the indentation of the line, if it has any non-whitespace\n# characters. Otherwise, returns -1.\nsub Indentation {\n my $line = shift;\n if (m/^(\\s*)\\S/) {\n return length($1);\n } else {\n return -1;\n }\n}\n\n# If the symbol table contains inlining info, Disassemble() may tag an\n# instruction with a location inside an inlined function. But for\n# source listings, we prefer to use the location in the function we\n# are listing. So use MapToSymbols() to fetch full location\n# information for each instruction and then pick out the first\n# location from a location list (location list contains callers before\n# callees in case of inlining).\n#\n# After this routine has run, each entry in $instructions contains:\n# [0] start address\n# [1] filename for function we are listing\n# [2] line number for function we are listing\n# [3] disassembly\n# [4] limit address\n# [5] most specific filename (may be different from [1] due to inlining)\n# [6] most specific line number (may be different from [2] due to inlining)\nsub GetTopLevelLineNumbers {\n my ($lib, $offset, $instructions) = @_;\n my $pcs = [];\n for (my $i = 0; $i <= $#{$instructions}; $i++) {\n push(@{$pcs}, $instructions->[$i]->[0]);\n }\n my $symbols = {};\n MapToSymbols($lib, $offset, $pcs, $symbols);\n for (my $i = 0; $i <= $#{$instructions}; $i++) {\n my $e = $instructions->[$i];\n push(@{$e}, $e->[1]);\n push(@{$e}, $e->[2]);\n my $addr = $e->[0];\n my $sym = $symbols->{$addr};\n if (defined($sym)) {\n if ($#{$sym} >= 2 && $sym->[1] =~ m/^(.*):(\\d+)$/) {\n $e->[1] = $1; # File name\n $e->[2] = $2; # Line number\n }\n }\n }\n}\n\n# Print source-listing for one routine\nsub PrintSource {\n my $prog = shift;\n my $offset = shift;\n my $routine = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $start_addr = shift;\n my $end_addr = shift;\n my $html = shift;\n my $output = shift;\n\n # Disassemble all instructions (just to get line numbers)\n my @instructions = Disassemble($prog, $offset, $start_addr, $end_addr);\n GetTopLevelLineNumbers($prog, $offset, \\@instructions);\n\n # Hack 1: assume that the first source file encountered in the\n # disassembly contains the routine\n my $filename = undef;\n for (my $i = 0; $i <= $#instructions; $i++) {\n if ($instructions[$i]->[2] >= 0) {\n $filename = $instructions[$i]->[1];\n last;\n }\n }\n if (!defined($filename)) {\n print STDERR \"no filename found in $routine\\n\";\n return 0;\n }\n\n # Hack 2: assume that the largest line number from $filename is the\n # end of the procedure. This is typically safe since if P1 contains\n # an inlined call to P2, then P2 usually occurs earlier in the\n # source file. If this does not work, we might have to compute a\n # density profile or just print all regions we find.\n my $lastline = 0;\n for (my $i = 0; $i <= $#instructions; $i++) {\n my $f = $instructions[$i]->[1];\n my $l = $instructions[$i]->[2];\n if (($f eq $filename) && ($l > $lastline)) {\n $lastline = $l;\n }\n }\n\n # Hack 3: assume the first source location from \"filename\" is the start of\n # the source code.\n my $firstline = 1;\n for (my $i = 0; $i <= $#instructions; $i++) {\n if ($instructions[$i]->[1] eq $filename) {\n $firstline = $instructions[$i]->[2];\n last;\n }\n }\n\n # Hack 4: Extend last line forward until its indentation is less than\n # the indentation we saw on $firstline\n my $oldlastline = $lastline;\n {\n if (!open(FILE, \"<$filename\")) {\n print STDERR \"$filename: $!\\n\";\n return 0;\n }\n my $l = 0;\n my $first_indentation = -1;\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n $l++;\n my $indent = Indentation($_);\n if ($l >= $firstline) {\n if ($first_indentation < 0 && $indent >= 0) {\n $first_indentation = $indent;\n last if ($first_indentation == 0);\n }\n }\n if ($l >= $lastline && $indent >= 0) {\n if ($indent >= $first_indentation) {\n $lastline = $l+1;\n } else {\n last;\n }\n }\n }\n close(FILE);\n }\n\n # Assign all samples to the range $firstline,$lastline,\n # Hack 4: If an instruction does not occur in the range, its samples\n # are moved to the next instruction that occurs in the range.\n my $samples1 = {}; # Map from line number to flat count\n my $samples2 = {}; # Map from line number to cumulative count\n my $running1 = 0; # Unassigned flat counts\n my $running2 = 0; # Unassigned cumulative counts\n my $total1 = 0; # Total flat counts\n my $total2 = 0; # Total cumulative counts\n my %disasm = (); # Map from line number to disassembly\n my $running_disasm = \"\"; # Unassigned disassembly\n my $skip_marker = \"---\\n\";\n if ($html) {\n $skip_marker = \"\";\n for (my $l = $firstline; $l <= $lastline; $l++) {\n $disasm{$l} = \"\";\n }\n }\n my $last_dis_filename = '';\n my $last_dis_linenum = -1;\n my $last_touched_line = -1; # To detect gaps in disassembly for a line\n foreach my $e (@instructions) {\n # Add up counts for all address that fall inside this instruction\n my $c1 = 0;\n my $c2 = 0;\n for (my $a = $e->[0]; $a lt $e->[4]; $a = AddressInc($a)) {\n $c1 += GetEntry($flat, $a);\n $c2 += GetEntry($cumulative, $a);\n }\n\n if ($html) {\n my $dis = sprintf(\" %6s %6s \\t\\t%8s: %s \",\n HtmlPrintNumber($c1),\n HtmlPrintNumber($c2),\n UnparseAddress($offset, $e->[0]),\n CleanDisassembly($e->[3]));\n \n # Append the most specific source line associated with this instruction\n if (length($dis) < 80) { $dis .= (' ' x (80 - length($dis))) };\n $dis = HtmlEscape($dis);\n my $f = $e->[5];\n my $l = $e->[6];\n if ($f ne $last_dis_filename) {\n $dis .= sprintf(\"%s:%d\", \n HtmlEscape(CleanFileName($f)), $l);\n } elsif ($l ne $last_dis_linenum) {\n # De-emphasize the unchanged file name portion\n $dis .= sprintf(\"%s\" .\n \":%d\", \n HtmlEscape(CleanFileName($f)), $l);\n } else {\n # De-emphasize the entire location\n $dis .= sprintf(\"%s:%d\", \n HtmlEscape(CleanFileName($f)), $l);\n }\n $last_dis_filename = $f;\n $last_dis_linenum = $l;\n $running_disasm .= $dis;\n $running_disasm .= \"\\n\";\n }\n\n $running1 += $c1;\n $running2 += $c2;\n $total1 += $c1;\n $total2 += $c2;\n my $file = $e->[1];\n my $line = $e->[2];\n if (($file eq $filename) &&\n ($line >= $firstline) &&\n ($line <= $lastline)) {\n # Assign all accumulated samples to this line\n AddEntry($samples1, $line, $running1);\n AddEntry($samples2, $line, $running2);\n $running1 = 0;\n $running2 = 0;\n if ($html) {\n if ($line != $last_touched_line && $disasm{$line} ne '') {\n $disasm{$line} .= \"\\n\";\n }\n $disasm{$line} .= $running_disasm;\n $running_disasm = '';\n $last_touched_line = $line;\n }\n }\n }\n\n # Assign any leftover samples to $lastline\n AddEntry($samples1, $lastline, $running1);\n AddEntry($samples2, $lastline, $running2);\n if ($html) {\n if ($lastline != $last_touched_line && $disasm{$lastline} ne '') {\n $disasm{$lastline} .= \"\\n\";\n }\n $disasm{$lastline} .= $running_disasm;\n }\n\n if ($html) {\n printf $output (\n \"

%s

%s\\n
\\n\" .\n      \"Total:%6s %6s (flat / cumulative %s)\\n\",\n      HtmlEscape(ShortFunctionName($routine)),\n      HtmlEscape(CleanFileName($filename)),\n      Unparse($total1),\n      Unparse($total2),\n      Units());\n  } else {\n    printf $output (\n      \"ROUTINE ====================== %s in %s\\n\" .\n      \"%6s %6s Total %s (flat / cumulative)\\n\",\n      ShortFunctionName($routine),\n      CleanFileName($filename),\n      Unparse($total1),\n      Unparse($total2),\n      Units());\n  }\n  if (!open(FILE, \"<$filename\")) {\n    print STDERR \"$filename: $!\\n\";\n    return 0;\n  }\n  my $l = 0;\n  while () {\n    s/\\r//g;         # turn windows-looking lines into unix-looking lines\n    $l++;\n    if ($l >= $firstline - 5 &&\n        (($l <= $oldlastline + 5) || ($l <= $lastline))) {\n      chop;\n      my $text = $_;\n      if ($l == $firstline) { print $output $skip_marker; }\n      my $n1 = GetEntry($samples1, $l);\n      my $n2 = GetEntry($samples2, $l);\n      if ($html) {\n        # Emit a span that has one of the following classes:\n        #    livesrc -- has samples\n        #    deadsrc -- has disassembly, but with no samples\n        #    nop     -- has no matching disasembly\n        # Also emit an optional span containing disassembly.\n        my $dis = $disasm{$l};\n        my $asm = \"\";\n        if (defined($dis) && $dis ne '') {\n          $asm = \"\" . $dis . \"\";\n        }\n        my $source_class = (($n1 + $n2 > 0) \n                            ? \"livesrc\" \n                            : (($asm ne \"\") ? \"deadsrc\" : \"nop\"));\n        printf $output (\n          \"%5d \" .\n          \"%6s %6s %s%s\\n\",\n          $l, $source_class,\n          HtmlPrintNumber($n1),\n          HtmlPrintNumber($n2),\n          HtmlEscape($text),\n          $asm);\n      } else {\n        printf $output(\n          \"%6s %6s %4d: %s\\n\",\n          UnparseAlt($n1),\n          UnparseAlt($n2),\n          $l,\n          $text);\n      }\n      if ($l == $lastline)  { print $output $skip_marker; }\n    };\n  }\n  close(FILE);\n  if ($html) {\n    print $output \"
\\n\";\n }\n return 1;\n}\n\n# Return the source line for the specified file/linenumber.\n# Returns undef if not found.\nsub SourceLine {\n my $file = shift;\n my $line = shift;\n\n # Look in cache\n if (!defined($main::source_cache{$file})) {\n if (100 < scalar keys(%main::source_cache)) {\n # Clear the cache when it gets too big\n $main::source_cache = ();\n }\n\n # Read all lines from the file\n if (!open(FILE, \"<$file\")) {\n print STDERR \"$file: $!\\n\";\n $main::source_cache{$file} = []; # Cache the negative result\n return undef;\n }\n my $lines = [];\n push(@{$lines}, \"\"); # So we can use 1-based line numbers as indices\n while () {\n push(@{$lines}, $_);\n }\n close(FILE);\n\n # Save the lines in the cache\n $main::source_cache{$file} = $lines;\n }\n\n my $lines = $main::source_cache{$file};\n if (($line < 0) || ($line > $#{$lines})) {\n return undef;\n } else {\n return $lines->[$line];\n }\n}\n\n# Print disassembly for one routine with interspersed source if available\nsub PrintDisassembledFunction {\n my $prog = shift;\n my $offset = shift;\n my $routine = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $start_addr = shift;\n my $end_addr = shift;\n my $total = shift;\n\n # Disassemble all instructions\n my @instructions = Disassemble($prog, $offset, $start_addr, $end_addr);\n\n # Make array of counts per instruction\n my @flat_count = ();\n my @cum_count = ();\n my $flat_total = 0;\n my $cum_total = 0;\n foreach my $e (@instructions) {\n # Add up counts for all address that fall inside this instruction\n my $c1 = 0;\n my $c2 = 0;\n for (my $a = $e->[0]; $a lt $e->[4]; $a = AddressInc($a)) {\n $c1 += GetEntry($flat, $a);\n $c2 += GetEntry($cumulative, $a);\n }\n push(@flat_count, $c1);\n push(@cum_count, $c2);\n $flat_total += $c1;\n $cum_total += $c2;\n }\n\n # Print header with total counts\n printf(\"ROUTINE ====================== %s\\n\" .\n \"%6s %6s %s (flat, cumulative) %.1f%% of total\\n\",\n ShortFunctionName($routine),\n Unparse($flat_total),\n Unparse($cum_total),\n Units(),\n ($cum_total * 100.0) / $total);\n\n # Process instructions in order\n my $current_file = \"\";\n for (my $i = 0; $i <= $#instructions; ) {\n my $e = $instructions[$i];\n\n # Print the new file name whenever we switch files\n if ($e->[1] ne $current_file) {\n $current_file = $e->[1];\n my $fname = $current_file;\n $fname =~ s|^\\./||; # Trim leading \"./\"\n\n # Shorten long file names\n if (length($fname) >= 58) {\n $fname = \"...\" . substr($fname, -55);\n }\n printf(\"-------------------- %s\\n\", $fname);\n }\n\n # TODO: Compute range of lines to print together to deal with\n # small reorderings.\n my $first_line = $e->[2];\n my $last_line = $first_line;\n my %flat_sum = ();\n my %cum_sum = ();\n for (my $l = $first_line; $l <= $last_line; $l++) {\n $flat_sum{$l} = 0;\n $cum_sum{$l} = 0;\n }\n\n # Find run of instructions for this range of source lines\n my $first_inst = $i;\n while (($i <= $#instructions) &&\n ($instructions[$i]->[2] >= $first_line) &&\n ($instructions[$i]->[2] <= $last_line)) {\n $e = $instructions[$i];\n $flat_sum{$e->[2]} += $flat_count[$i];\n $cum_sum{$e->[2]} += $cum_count[$i];\n $i++;\n }\n my $last_inst = $i - 1;\n\n # Print source lines\n for (my $l = $first_line; $l <= $last_line; $l++) {\n my $line = SourceLine($current_file, $l);\n if (!defined($line)) {\n $line = \"?\\n\";\n next;\n } else {\n $line =~ s/^\\s+//;\n }\n printf(\"%6s %6s %5d: %s\",\n UnparseAlt($flat_sum{$l}),\n UnparseAlt($cum_sum{$l}),\n $l,\n $line);\n }\n\n # Print disassembly\n for (my $x = $first_inst; $x <= $last_inst; $x++) {\n my $e = $instructions[$x];\n printf(\"%6s %6s %8s: %6s\\n\",\n UnparseAlt($flat_count[$x]),\n UnparseAlt($cum_count[$x]),\n UnparseAddress($offset, $e->[0]),\n CleanDisassembly($e->[3]));\n }\n }\n}\n\n# Print DOT graph\nsub PrintDot {\n my $prog = shift;\n my $symbols = shift;\n my $raw = shift;\n my $flat = shift;\n my $cumulative = shift;\n my $overall_total = shift;\n\n # Get total\n my $local_total = TotalProfile($flat);\n my $nodelimit = int($main::opt_nodefraction * $local_total);\n my $edgelimit = int($main::opt_edgefraction * $local_total);\n my $nodecount = $main::opt_nodecount;\n\n # Find nodes to include\n my @list = (sort { abs(GetEntry($cumulative, $b)) <=>\n abs(GetEntry($cumulative, $a))\n || $a cmp $b }\n keys(%{$cumulative}));\n my $last = $nodecount - 1;\n if ($last > $#list) {\n $last = $#list;\n }\n while (($last >= 0) &&\n (abs(GetEntry($cumulative, $list[$last])) <= $nodelimit)) {\n $last--;\n }\n if ($last < 0) {\n print STDERR \"No nodes to print\\n\";\n return 0;\n }\n\n if ($nodelimit > 0 || $edgelimit > 0) {\n printf STDERR (\"Dropping nodes with <= %s %s; edges with <= %s abs(%s)\\n\",\n Unparse($nodelimit), Units(),\n Unparse($edgelimit), Units());\n }\n\n # Open DOT output file\n my $output;\n my $escaped_dot = ShellEscape(@DOT);\n my $escaped_ps2pdf = ShellEscape(@PS2PDF);\n if ($main::opt_gv) {\n my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, \"ps\"));\n $output = \"| $escaped_dot -Tps2 >$escaped_outfile\";\n } elsif ($main::opt_evince) {\n my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, \"pdf\"));\n $output = \"| $escaped_dot -Tps2 | $escaped_ps2pdf - $escaped_outfile\";\n } elsif ($main::opt_ps) {\n $output = \"| $escaped_dot -Tps2\";\n } elsif ($main::opt_pdf) {\n $output = \"| $escaped_dot -Tps2 | $escaped_ps2pdf - -\";\n } elsif ($main::opt_web || $main::opt_svg) {\n # We need to post-process the SVG, so write to a temporary file always.\n my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, \"svg\"));\n $output = \"| $escaped_dot -Tsvg >$escaped_outfile\";\n } elsif ($main::opt_gif) {\n $output = \"| $escaped_dot -Tgif\";\n } else {\n $output = \">&STDOUT\";\n }\n open(DOT, $output) || error(\"$output: $!\\n\");\n\n # Title\n printf DOT (\"digraph \\\"%s; %s %s\\\" {\\n\",\n $prog,\n Unparse($overall_total),\n Units());\n if ($main::opt_pdf) {\n # The output is more printable if we set the page size for dot.\n printf DOT (\"size=\\\"8,11\\\"\\n\");\n }\n printf DOT (\"node [width=0.375,height=0.25];\\n\");\n\n # Print legend\n printf DOT (\"Legend [shape=box,fontsize=24,shape=plaintext,\" .\n \"label=\\\"%s\\\\l%s\\\\l%s\\\\l%s\\\\l%s\\\\l\\\"];\\n\",\n $prog,\n sprintf(\"Total %s: %s\", Units(), Unparse($overall_total)),\n sprintf(\"Focusing on: %s\", Unparse($local_total)),\n sprintf(\"Dropped nodes with <= %s abs(%s)\",\n Unparse($nodelimit), Units()),\n sprintf(\"Dropped edges with <= %s %s\",\n Unparse($edgelimit), Units())\n );\n\n # Print nodes\n my %node = ();\n my $nextnode = 1;\n foreach my $a (@list[0..$last]) {\n # Pick font size\n my $f = GetEntry($flat, $a);\n my $c = GetEntry($cumulative, $a);\n\n my $fs = 8;\n if ($local_total > 0) {\n $fs = 8 + (50.0 * sqrt(abs($f * 1.0 / $local_total)));\n }\n\n $node{$a} = $nextnode++;\n my $sym = $a;\n $sym =~ s/\\s+/\\\\n/g;\n $sym =~ s/::/\\\\n/g;\n\n # Extra cumulative info to print for non-leaves\n my $extra = \"\";\n if ($f != $c) {\n $extra = sprintf(\"\\\\rof %s (%s)\",\n Unparse($c),\n Percent($c, $local_total));\n }\n my $style = \"\";\n if ($main::opt_heapcheck) {\n if ($f > 0) {\n # make leak-causing nodes more visible (add a background)\n $style = \",style=filled,fillcolor=gray\"\n } elsif ($f < 0) {\n # make anti-leak-causing nodes (which almost never occur)\n # stand out as well (triple border)\n $style = \",peripheries=3\"\n }\n }\n\n printf DOT (\"N%d [label=\\\"%s\\\\n%s (%s)%s\\\\r\" .\n \"\\\",shape=box,fontsize=%.1f%s];\\n\",\n $node{$a},\n $sym,\n Unparse($f),\n Percent($f, $local_total),\n $extra,\n $fs,\n $style,\n );\n }\n\n # Get edges and counts per edge\n my %edge = ();\n my $n;\n my $fullname_to_shortname_map = {};\n FillFullnameToShortnameMap($symbols, $fullname_to_shortname_map);\n foreach my $k (keys(%{$raw})) {\n # TODO: omit low %age edges\n $n = $raw->{$k};\n my @translated = TranslateStack($symbols, $fullname_to_shortname_map, $k);\n for (my $i = 1; $i <= $#translated; $i++) {\n my $src = $translated[$i];\n my $dst = $translated[$i-1];\n #next if ($src eq $dst); # Avoid self-edges?\n if (exists($node{$src}) && exists($node{$dst})) {\n my $edge_label = \"$src\\001$dst\";\n if (!exists($edge{$edge_label})) {\n $edge{$edge_label} = 0;\n }\n $edge{$edge_label} += $n;\n }\n }\n }\n\n # Print edges (process in order of decreasing counts)\n my %indegree = (); # Number of incoming edges added per node so far\n my %outdegree = (); # Number of outgoing edges added per node so far\n foreach my $e (sort { $edge{$b} <=> $edge{$a} } keys(%edge)) {\n my @x = split(/\\001/, $e);\n $n = $edge{$e};\n\n # Initialize degree of kept incoming and outgoing edges if necessary\n my $src = $x[0];\n my $dst = $x[1];\n if (!exists($outdegree{$src})) { $outdegree{$src} = 0; }\n if (!exists($indegree{$dst})) { $indegree{$dst} = 0; }\n\n my $keep;\n if ($indegree{$dst} == 0) {\n # Keep edge if needed for reachability\n $keep = 1;\n } elsif (abs($n) <= $edgelimit) {\n # Drop if we are below --edgefraction\n $keep = 0;\n } elsif ($outdegree{$src} >= $main::opt_maxdegree ||\n $indegree{$dst} >= $main::opt_maxdegree) {\n # Keep limited number of in/out edges per node\n $keep = 0;\n } else {\n $keep = 1;\n }\n\n if ($keep) {\n $outdegree{$src}++;\n $indegree{$dst}++;\n\n # Compute line width based on edge count\n my $fraction = abs($local_total ? (3 * ($n / $local_total)) : 0);\n if ($fraction > 1) { $fraction = 1; }\n my $w = $fraction * 2;\n if ($w < 1 && ($main::opt_web || $main::opt_svg)) {\n # SVG output treats line widths < 1 poorly.\n $w = 1;\n }\n\n # Dot sometimes segfaults if given edge weights that are too large, so\n # we cap the weights at a large value\n my $edgeweight = abs($n) ** 0.7;\n if ($edgeweight > 100000) { $edgeweight = 100000; }\n $edgeweight = int($edgeweight);\n\n my $style = sprintf(\"setlinewidth(%f)\", $w);\n if ($x[1] =~ m/\\(inline\\)/) {\n $style .= \",dashed\";\n }\n\n # Use a slightly squashed function of the edge count as the weight\n printf DOT (\"N%s -> N%s [label=%s, weight=%d, style=\\\"%s\\\"];\\n\",\n $node{$x[0]},\n $node{$x[1]},\n Unparse($n),\n $edgeweight,\n $style);\n }\n }\n\n print DOT (\"}\\n\");\n close(DOT);\n\n if ($main::opt_web || $main::opt_svg) {\n # Rewrite SVG to be more usable inside web browser.\n RewriteSvg(TempName($main::next_tmpfile, \"svg\"));\n }\n\n return 1;\n}\n\nsub RewriteSvg {\n my $svgfile = shift;\n\n open(SVG, $svgfile) || die \"open temp svg: $!\";\n my @svg = ;\n close(SVG);\n unlink $svgfile;\n my $svg = join('', @svg);\n\n # Dot's SVG output is\n #\n # \n # \n # ...\n # \n # \n #\n # Change it to\n #\n # \n # $svg_javascript\n # \n # \n # ...\n # \n # \n # \n\n # Fix width, height; drop viewBox.\n $svg =~ s/(?s) above first \n my $svg_javascript = SvgJavascript();\n my $viewport = \"\\n\";\n $svg =~ s/ above .\n $svg =~ s/(.*)(<\\/svg>)/$1<\\/g>$2/;\n $svg =~ s/$svgfile\") || die \"open $svgfile: $!\";\n print SVG $svg;\n close(SVG);\n }\n}\n\nsub SvgJavascript {\n return <<'EOF';\n\nEOF\n}\n\n# Provides a map from fullname to shortname for cases where the\n# shortname is ambiguous. The symlist has both the fullname and\n# shortname for all symbols, which is usually fine, but sometimes --\n# such as overloaded functions -- two different fullnames can map to\n# the same shortname. In that case, we use the address of the\n# function to disambiguate the two. This function fills in a map that\n# maps fullnames to modified shortnames in such cases. If a fullname\n# is not present in the map, the 'normal' shortname provided by the\n# symlist is the appropriate one to use.\nsub FillFullnameToShortnameMap {\n my $symbols = shift;\n my $fullname_to_shortname_map = shift;\n my $shortnames_seen_once = {};\n my $shortnames_seen_more_than_once = {};\n\n foreach my $symlist (values(%{$symbols})) {\n # TODO(csilvers): deal with inlined symbols too.\n my $shortname = $symlist->[0];\n my $fullname = $symlist->[2];\n if ($fullname !~ /<[0-9a-fA-F]+>$/) { # fullname doesn't end in an address\n next; # the only collisions we care about are when addresses differ\n }\n if (defined($shortnames_seen_once->{$shortname}) &&\n $shortnames_seen_once->{$shortname} ne $fullname) {\n $shortnames_seen_more_than_once->{$shortname} = 1;\n } else {\n $shortnames_seen_once->{$shortname} = $fullname;\n }\n }\n\n foreach my $symlist (values(%{$symbols})) {\n my $shortname = $symlist->[0];\n my $fullname = $symlist->[2];\n # TODO(csilvers): take in a list of addresses we care about, and only\n # store in the map if $symlist->[1] is in that list. Saves space.\n next if defined($fullname_to_shortname_map->{$fullname});\n if (defined($shortnames_seen_more_than_once->{$shortname})) {\n if ($fullname =~ /<0*([^>]*)>$/) { # fullname has address at end of it\n $fullname_to_shortname_map->{$fullname} = \"$shortname\\@$1\";\n }\n }\n }\n}\n\n# Return a small number that identifies the argument.\n# Multiple calls with the same argument will return the same number.\n# Calls with different arguments will return different numbers.\nsub ShortIdFor {\n my $key = shift;\n my $id = $main::uniqueid{$key};\n if (!defined($id)) {\n $id = keys(%main::uniqueid) + 1;\n $main::uniqueid{$key} = $id;\n }\n return $id;\n}\n\n# Translate a stack of addresses into a stack of symbols\nsub TranslateStack {\n my $symbols = shift;\n my $fullname_to_shortname_map = shift;\n my $k = shift;\n\n my @addrs = split(/\\n/, $k);\n my @result = ();\n for (my $i = 0; $i <= $#addrs; $i++) {\n my $a = $addrs[$i];\n\n # Skip large addresses since they sometimes show up as fake entries on RH9\n if (length($a) > 8 && $a gt \"7fffffffffffffff\") {\n next;\n }\n\n if ($main::opt_disasm || $main::opt_list) {\n # We want just the address for the key\n push(@result, $a);\n next;\n }\n\n my $symlist = $symbols->{$a};\n if (!defined($symlist)) {\n $symlist = [$a, \"\", $a];\n }\n\n # We can have a sequence of symbols for a particular entry\n # (more than one symbol in the case of inlining). Callers\n # come before callees in symlist, so walk backwards since\n # the translated stack should contain callees before callers.\n for (my $j = $#{$symlist}; $j >= 2; $j -= 3) {\n my $func = $symlist->[$j-2];\n my $fileline = $symlist->[$j-1];\n my $fullfunc = $symlist->[$j];\n if (defined($fullname_to_shortname_map->{$fullfunc})) {\n $func = $fullname_to_shortname_map->{$fullfunc};\n }\n if ($j > 2) {\n $func = \"$func (inline)\";\n }\n\n # Do not merge nodes corresponding to Callback::Run since that\n # causes confusing cycles in dot display. Instead, we synthesize\n # a unique name for this frame per caller.\n if ($func =~ m/Callback.*::Run$/) {\n my $caller = ($i > 0) ? $addrs[$i-1] : 0;\n $func = \"Run#\" . ShortIdFor($caller);\n }\n\n if ($main::opt_addresses) {\n push(@result, \"$a $func $fileline\");\n } elsif ($main::opt_lines) {\n if ($func eq '??' && $fileline eq '??:0') {\n push(@result, \"$a\");\n } else {\n push(@result, \"$func $fileline\");\n }\n } elsif ($main::opt_functions) {\n if ($func eq '??') {\n push(@result, \"$a\");\n } else {\n push(@result, $func);\n }\n } elsif ($main::opt_files) {\n if ($fileline eq '??:0' || $fileline eq '') {\n push(@result, \"$a\");\n } else {\n my $f = $fileline;\n $f =~ s/:\\d+$//;\n push(@result, $f);\n }\n } else {\n push(@result, $a);\n last; # Do not print inlined info\n }\n }\n }\n\n # print join(\",\", @addrs), \" => \", join(\",\", @result), \"\\n\";\n return @result;\n}\n\n# Generate percent string for a number and a total\nsub Percent {\n my $num = shift;\n my $tot = shift;\n if ($tot != 0) {\n return sprintf(\"%.1f%%\", $num * 100.0 / $tot);\n } else {\n return ($num == 0) ? \"nan\" : (($num > 0) ? \"+inf\" : \"-inf\");\n }\n}\n\n# Generate pretty-printed form of number\nsub Unparse {\n my $num = shift;\n if ($main::profile_type eq 'heap' || $main::profile_type eq 'growth') {\n if ($main::opt_inuse_objects || $main::opt_alloc_objects) {\n return sprintf(\"%d\", $num);\n } else {\n if ($main::opt_show_bytes) {\n return sprintf(\"%d\", $num);\n } else {\n return sprintf(\"%.1f\", $num / 1048576.0);\n }\n }\n } elsif ($main::profile_type eq 'contention' && !$main::opt_contentions) {\n return sprintf(\"%.3f\", $num / 1e9); # Convert nanoseconds to seconds\n } else {\n return sprintf(\"%d\", $num);\n }\n}\n\n# Alternate pretty-printed form: 0 maps to \".\"\nsub UnparseAlt {\n my $num = shift;\n if ($num == 0) {\n return \".\";\n } else {\n return Unparse($num);\n }\n}\n\n# Alternate pretty-printed form: 0 maps to \"\"\nsub HtmlPrintNumber {\n my $num = shift;\n if ($num == 0) {\n return \"\";\n } else {\n return Unparse($num);\n }\n}\n\n# Return output units\nsub Units {\n if ($main::profile_type eq 'heap' || $main::profile_type eq 'growth') {\n if ($main::opt_inuse_objects || $main::opt_alloc_objects) {\n return \"objects\";\n } else {\n if ($main::opt_show_bytes) {\n return \"B\";\n } else {\n return \"MB\";\n }\n }\n } elsif ($main::profile_type eq 'contention' && !$main::opt_contentions) {\n return \"seconds\";\n } else {\n return \"samples\";\n }\n}\n\n##### Profile manipulation code #####\n\n# Generate flattened profile:\n# If count is charged to stack [a,b,c,d], in generated profile,\n# it will be charged to [a]\nsub FlatProfile {\n my $profile = shift;\n my $result = {};\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n if ($#addrs >= 0) {\n AddEntry($result, $addrs[0], $count);\n }\n }\n return $result;\n}\n\n# Generate cumulative profile:\n# If count is charged to stack [a,b,c,d], in generated profile,\n# it will be charged to [a], [b], [c], [d]\nsub CumulativeProfile {\n my $profile = shift;\n my $result = {};\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n foreach my $a (@addrs) {\n AddEntry($result, $a, $count);\n }\n }\n return $result;\n}\n\n# If the second-youngest PC on the stack is always the same, returns\n# that pc. Otherwise, returns undef.\nsub IsSecondPcAlwaysTheSame {\n my $profile = shift;\n\n my $second_pc = undef;\n foreach my $k (keys(%{$profile})) {\n my @addrs = split(/\\n/, $k);\n if ($#addrs < 1) {\n return undef;\n }\n if (not defined $second_pc) {\n $second_pc = $addrs[1];\n } else {\n if ($second_pc ne $addrs[1]) {\n return undef;\n }\n }\n }\n return $second_pc;\n}\n\nsub ExtractSymbolLocation {\n my $symbols = shift;\n my $address = shift;\n # 'addr2line' outputs \"??:0\" for unknown locations; we do the\n # same to be consistent.\n my $location = \"??:0:unknown\";\n if (exists $symbols->{$address}) {\n my $file = $symbols->{$address}->[1];\n if ($file eq \"?\") {\n $file = \"??:0\"\n }\n $location = $file . \":\" . $symbols->{$address}->[0];\n }\n return $location;\n}\n\n# Extracts a graph of calls.\nsub ExtractCalls {\n my $symbols = shift;\n my $profile = shift;\n\n my $calls = {};\n while( my ($stack_trace, $count) = each %$profile ) {\n my @address = split(/\\n/, $stack_trace);\n my $destination = ExtractSymbolLocation($symbols, $address[0]);\n AddEntry($calls, $destination, $count);\n for (my $i = 1; $i <= $#address; $i++) {\n my $source = ExtractSymbolLocation($symbols, $address[$i]);\n my $call = \"$source -> $destination\";\n AddEntry($calls, $call, $count);\n $destination = $source;\n }\n }\n\n return $calls;\n}\n\nsub RemoveUninterestingFrames {\n my $symbols = shift;\n my $profile = shift;\n\n # List of function names to skip\n my %skip = ();\n my $skip_regexp = 'NOMATCH';\n if ($main::profile_type eq 'heap' || $main::profile_type eq 'growth') {\n foreach my $name ('calloc',\n 'cfree',\n 'malloc',\n 'free',\n 'memalign',\n 'posix_memalign',\n 'pvalloc',\n 'valloc',\n 'realloc',\n 'tc_calloc',\n 'tc_cfree',\n 'tc_malloc',\n 'tc_free',\n 'tc_memalign',\n 'tc_posix_memalign',\n 'tc_pvalloc',\n 'tc_valloc',\n 'tc_realloc',\n 'tc_new',\n 'tc_delete',\n 'tc_newarray',\n 'tc_deletearray',\n 'tc_new_nothrow',\n 'tc_newarray_nothrow',\n 'do_malloc',\n '::do_malloc', # new name -- got moved to an unnamed ns\n '::do_malloc_or_cpp_alloc',\n 'DoSampledAllocation',\n 'simple_alloc::allocate',\n '__malloc_alloc_template::allocate',\n '__builtin_delete',\n '__builtin_new',\n '__builtin_vec_delete',\n '__builtin_vec_new',\n 'operator new',\n 'operator new[]',\n # The entry to our memory-allocation routines on OS X\n 'malloc_zone_malloc',\n 'malloc_zone_calloc',\n 'malloc_zone_valloc',\n 'malloc_zone_realloc',\n 'malloc_zone_memalign',\n 'malloc_zone_free',\n # These mark the beginning/end of our custom sections\n '__start_google_malloc',\n '__stop_google_malloc',\n '__start_malloc_hook',\n '__stop_malloc_hook') {\n $skip{$name} = 1;\n $skip{\"_\" . $name} = 1; # Mach (OS X) adds a _ prefix to everything\n }\n # TODO: Remove TCMalloc once everything has been\n # moved into the tcmalloc:: namespace and we have flushed\n # old code out of the system.\n $skip_regexp = \"TCMalloc|^tcmalloc::\";\n } elsif ($main::profile_type eq 'contention') {\n foreach my $vname ('base::RecordLockProfileData',\n 'base::SubmitMutexProfileData',\n 'base::SubmitSpinLockProfileData',\n 'Mutex::Unlock',\n 'Mutex::UnlockSlow',\n 'Mutex::ReaderUnlock',\n 'MutexLock::~MutexLock',\n 'SpinLock::Unlock',\n 'SpinLock::SlowUnlock',\n 'SpinLockHolder::~SpinLockHolder') {\n $skip{$vname} = 1;\n }\n } elsif ($main::profile_type eq 'cpu') {\n # Drop signal handlers used for CPU profile collection\n # TODO(dpeng): this should not be necessary; it's taken\n # care of by the general 2nd-pc mechanism below.\n foreach my $name ('ProfileData::Add', # historical\n 'ProfileData::prof_handler', # historical\n 'CpuProfiler::prof_handler',\n '__FRAME_END__',\n '__pthread_sighandler',\n '__restore') {\n $skip{$name} = 1;\n }\n } else {\n # Nothing skipped for unknown types\n }\n\n if ($main::profile_type eq 'cpu') {\n # If all the second-youngest program counters are the same,\n # this STRONGLY suggests that it is an artifact of measurement,\n # i.e., stack frames pushed by the CPU profiler signal handler.\n # Hence, we delete them.\n # (The topmost PC is read from the signal structure, not from\n # the stack, so it does not get involved.)\n while (my $second_pc = IsSecondPcAlwaysTheSame($profile)) {\n my $result = {};\n my $func = '';\n if (exists($symbols->{$second_pc})) {\n $second_pc = $symbols->{$second_pc}->[0];\n }\n print STDERR \"Removing $second_pc from all stack traces.\\n\";\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n splice @addrs, 1, 1;\n my $reduced_path = join(\"\\n\", @addrs);\n AddEntry($result, $reduced_path, $count);\n }\n $profile = $result;\n }\n }\n\n my $result = {};\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n my @path = ();\n foreach my $a (@addrs) {\n if (exists($symbols->{$a})) {\n my $func = $symbols->{$a}->[0];\n if ($skip{$func} || ($func =~ m/$skip_regexp/)) {\n next;\n }\n }\n push(@path, $a);\n }\n my $reduced_path = join(\"\\n\", @path);\n AddEntry($result, $reduced_path, $count);\n }\n return $result;\n}\n\n# Reduce profile to granularity given by user\nsub ReduceProfile {\n my $symbols = shift;\n my $profile = shift;\n my $result = {};\n my $fullname_to_shortname_map = {};\n FillFullnameToShortnameMap($symbols, $fullname_to_shortname_map);\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @translated = TranslateStack($symbols, $fullname_to_shortname_map, $k);\n my @path = ();\n my %seen = ();\n $seen{''} = 1; # So that empty keys are skipped\n foreach my $e (@translated) {\n # To avoid double-counting due to recursion, skip a stack-trace\n # entry if it has already been seen\n if (!$seen{$e}) {\n $seen{$e} = 1;\n push(@path, $e);\n }\n }\n my $reduced_path = join(\"\\n\", @path);\n AddEntry($result, $reduced_path, $count);\n }\n return $result;\n}\n\n# Does the specified symbol array match the regexp?\nsub SymbolMatches {\n my $sym = shift;\n my $re = shift;\n if (defined($sym)) {\n for (my $i = 0; $i < $#{$sym}; $i += 3) {\n if ($sym->[$i] =~ m/$re/ || $sym->[$i+1] =~ m/$re/) {\n return 1;\n }\n }\n }\n return 0;\n}\n\n# Focus only on paths involving specified regexps\nsub FocusProfile {\n my $symbols = shift;\n my $profile = shift;\n my $focus = shift;\n my $result = {};\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n foreach my $a (@addrs) {\n # Reply if it matches either the address/shortname/fileline\n if (($a =~ m/$focus/) || SymbolMatches($symbols->{$a}, $focus)) {\n AddEntry($result, $k, $count);\n last;\n }\n }\n }\n return $result;\n}\n\n# Focus only on paths not involving specified regexps\nsub IgnoreProfile {\n my $symbols = shift;\n my $profile = shift;\n my $ignore = shift;\n my $result = {};\n foreach my $k (keys(%{$profile})) {\n my $count = $profile->{$k};\n my @addrs = split(/\\n/, $k);\n my $matched = 0;\n foreach my $a (@addrs) {\n # Reply if it matches either the address/shortname/fileline\n if (($a =~ m/$ignore/) || SymbolMatches($symbols->{$a}, $ignore)) {\n $matched = 1;\n last;\n }\n }\n if (!$matched) {\n AddEntry($result, $k, $count);\n }\n }\n return $result;\n}\n\n# Get total count in profile\nsub TotalProfile {\n my $profile = shift;\n my $result = 0;\n foreach my $k (keys(%{$profile})) {\n $result += $profile->{$k};\n }\n return $result;\n}\n\n# Add A to B\nsub AddProfile {\n my $A = shift;\n my $B = shift;\n\n my $R = {};\n # add all keys in A\n foreach my $k (keys(%{$A})) {\n my $v = $A->{$k};\n AddEntry($R, $k, $v);\n }\n # add all keys in B\n foreach my $k (keys(%{$B})) {\n my $v = $B->{$k};\n AddEntry($R, $k, $v);\n }\n return $R;\n}\n\n# Merges symbol maps\nsub MergeSymbols {\n my $A = shift;\n my $B = shift;\n\n my $R = {};\n foreach my $k (keys(%{$A})) {\n $R->{$k} = $A->{$k};\n }\n if (defined($B)) {\n foreach my $k (keys(%{$B})) {\n $R->{$k} = $B->{$k};\n }\n }\n return $R;\n}\n\n\n# Add A to B\nsub AddPcs {\n my $A = shift;\n my $B = shift;\n\n my $R = {};\n # add all keys in A\n foreach my $k (keys(%{$A})) {\n $R->{$k} = 1\n }\n # add all keys in B\n foreach my $k (keys(%{$B})) {\n $R->{$k} = 1\n }\n return $R;\n}\n\n# Subtract B from A\nsub SubtractProfile {\n my $A = shift;\n my $B = shift;\n\n my $R = {};\n foreach my $k (keys(%{$A})) {\n my $v = $A->{$k} - GetEntry($B, $k);\n if ($v < 0 && $main::opt_drop_negative) {\n $v = 0;\n }\n AddEntry($R, $k, $v);\n }\n if (!$main::opt_drop_negative) {\n # Take care of when subtracted profile has more entries\n foreach my $k (keys(%{$B})) {\n if (!exists($A->{$k})) {\n AddEntry($R, $k, 0 - $B->{$k});\n }\n }\n }\n return $R;\n}\n\n# Get entry from profile; zero if not present\nsub GetEntry {\n my $profile = shift;\n my $k = shift;\n if (exists($profile->{$k})) {\n return $profile->{$k};\n } else {\n return 0;\n }\n}\n\n# Add entry to specified profile\nsub AddEntry {\n my $profile = shift;\n my $k = shift;\n my $n = shift;\n if (!exists($profile->{$k})) {\n $profile->{$k} = 0;\n }\n $profile->{$k} += $n;\n}\n\n# Add a stack of entries to specified profile, and add them to the $pcs\n# list.\nsub AddEntries {\n my $profile = shift;\n my $pcs = shift;\n my $stack = shift;\n my $count = shift;\n my @k = ();\n\n foreach my $e (split(/\\s+/, $stack)) {\n my $pc = HexExtend($e);\n $pcs->{$pc} = 1;\n push @k, $pc;\n }\n AddEntry($profile, (join \"\\n\", @k), $count);\n}\n\n##### Code to profile a server dynamically #####\n\nsub CheckSymbolPage {\n my $url = SymbolPageURL();\n my $command = ShellEscape(@URL_FETCHER, $url);\n open(SYMBOL, \"$command |\") or error($command);\n my $line = ;\n $line =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n close(SYMBOL);\n unless (defined($line)) {\n error(\"$url doesn't exist\\n\");\n }\n\n if ($line =~ /^num_symbols:\\s+(\\d+)$/) {\n if ($1 == 0) {\n error(\"Stripped binary. No symbols available.\\n\");\n }\n } else {\n error(\"Failed to get the number of symbols from $url\\n\");\n }\n}\n\nsub IsProfileURL {\n my $profile_name = shift;\n if (-f $profile_name) {\n printf STDERR \"Using local file $profile_name.\\n\";\n return 0;\n }\n return 1;\n}\n\nsub ParseProfileURL {\n my $profile_name = shift;\n\n if (!defined($profile_name) || $profile_name eq \"\") {\n return ();\n }\n\n # Split profile URL - matches all non-empty strings, so no test.\n $profile_name =~ m,^(https?://)?([^/]+)(.*?)(/|$PROFILES)?$,;\n\n my $proto = $1 || \"http://\";\n my $hostport = $2;\n my $prefix = $3;\n my $profile = $4 || \"/\";\n\n my $host = $hostport;\n $host =~ s/:.*//;\n\n my $baseurl = \"$proto$hostport$prefix\";\n return ($host, $baseurl, $profile);\n}\n\n# We fetch symbols from the first profile argument.\nsub SymbolPageURL {\n my ($host, $baseURL, $path) = ParseProfileURL($main::pfile_args[0]);\n return \"$baseURL$SYMBOL_PAGE\";\n}\n\nsub FetchProgramName() {\n my ($host, $baseURL, $path) = ParseProfileURL($main::pfile_args[0]);\n my $url = \"$baseURL$PROGRAM_NAME_PAGE\";\n my $command_line = ShellEscape(@URL_FETCHER, $url);\n open(CMDLINE, \"$command_line |\") or error($command_line);\n my $cmdline = ;\n $cmdline =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n close(CMDLINE);\n error(\"Failed to get program name from $url\\n\") unless defined($cmdline);\n $cmdline =~ s/\\x00.+//; # Remove argv[1] and latters.\n $cmdline =~ s!\\n!!g; # Remove LFs.\n return $cmdline;\n}\n\n# Gee, curl's -L (--location) option isn't reliable at least\n# with its 7.12.3 version. Curl will forget to post data if\n# there is a redirection. This function is a workaround for\n# curl. Redirection happens on borg hosts.\nsub ResolveRedirectionForCurl {\n my $url = shift;\n my $command_line = ShellEscape(@URL_FETCHER, \"--head\", $url);\n open(CMDLINE, \"$command_line |\") or error($command_line);\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n if (/^Location: (.*)/) {\n $url = $1;\n }\n }\n close(CMDLINE);\n return $url;\n}\n\n# Add a timeout flat to URL_FETCHER. Returns a new list.\nsub AddFetchTimeout {\n my $timeout = shift;\n my @fetcher = shift;\n if (defined($timeout)) {\n if (join(\" \", @fetcher) =~ m/\\bcurl -s/) {\n push(@fetcher, \"--max-time\", sprintf(\"%d\", $timeout));\n } elsif (join(\" \", @fetcher) =~ m/\\brpcget\\b/) {\n push(@fetcher, sprintf(\"--deadline=%d\", $timeout));\n }\n }\n return @fetcher;\n}\n\n# Reads a symbol map from the file handle name given as $1, returning\n# the resulting symbol map. Also processes variables relating to symbols.\n# Currently, the only variable processed is 'binary=' which updates\n# $main::prog to have the correct program name.\nsub ReadSymbols {\n my $in = shift;\n my $map = {};\n while (<$in>) {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n # Removes all the leading zeroes from the symbols, see comment below.\n if (m/^0x0*([0-9a-f]+)\\s+(.+)/) {\n $map->{$1} = $2;\n } elsif (m/^---/) {\n last;\n } elsif (m/^([a-z][^=]*)=(.*)$/ ) {\n my ($variable, $value) = ($1, $2);\n for ($variable, $value) {\n s/^\\s+//;\n s/\\s+$//;\n }\n if ($variable eq \"binary\") {\n if ($main::prog ne $UNKNOWN_BINARY && $main::prog ne $value) {\n printf STDERR (\"Warning: Mismatched binary name '%s', using '%s'.\\n\",\n $main::prog, $value);\n }\n $main::prog = $value;\n } else {\n printf STDERR (\"Ignoring unknown variable in symbols list: \" .\n \"'%s' = '%s'\\n\", $variable, $value);\n }\n }\n }\n return $map;\n}\n\n# Fetches and processes symbols to prepare them for use in the profile output\n# code. If the optional 'symbol_map' arg is not given, fetches symbols from\n# $SYMBOL_PAGE for all PC values found in profile. Otherwise, the raw symbols\n# are assumed to have already been fetched into 'symbol_map' and are simply\n# extracted and processed.\nsub FetchSymbols {\n my $pcset = shift;\n my $symbol_map = shift;\n\n my %seen = ();\n my @pcs = grep { !$seen{$_}++ } keys(%$pcset); # uniq\n\n if (!defined($symbol_map)) {\n my $post_data = join(\"+\", sort((map {\"0x\" . \"$_\"} @pcs)));\n\n open(POSTFILE, \">$main::tmpfile_sym\");\n print POSTFILE $post_data;\n close(POSTFILE);\n\n my $url = SymbolPageURL();\n\n my $command_line;\n if (join(\" \", @URL_FETCHER) =~ m/\\bcurl -s/) {\n $url = ResolveRedirectionForCurl($url);\n $command_line = ShellEscape(@URL_FETCHER, \"-d\", \"\\@$main::tmpfile_sym\",\n $url);\n } else {\n $command_line = (ShellEscape(@URL_FETCHER, \"--post\", $url)\n . \" < \" . ShellEscape($main::tmpfile_sym));\n }\n # We use c++filt in case $SYMBOL_PAGE gives us mangled symbols.\n my $escaped_cppfilt = ShellEscape($obj_tool_map{\"c++filt\"});\n open(SYMBOL, \"$command_line | $escaped_cppfilt |\") or error($command_line);\n $symbol_map = ReadSymbols(*SYMBOL{IO});\n close(SYMBOL);\n }\n\n my $symbols = {};\n foreach my $pc (@pcs) {\n my $fullname;\n # For 64 bits binaries, symbols are extracted with 8 leading zeroes.\n # Then /symbol reads the long symbols in as uint64, and outputs\n # the result with a \"0x%08llx\" format which get rid of the zeroes.\n # By removing all the leading zeroes in both $pc and the symbols from\n # /symbol, the symbols match and are retrievable from the map.\n my $shortpc = $pc;\n $shortpc =~ s/^0*//;\n # Each line may have a list of names, which includes the function\n # and also other functions it has inlined. They are separated (in\n # PrintSymbolizedProfile), by --, which is illegal in function names.\n my $fullnames;\n if (defined($symbol_map->{$shortpc})) {\n $fullnames = $symbol_map->{$shortpc};\n } else {\n $fullnames = \"0x\" . $pc; # Just use addresses\n }\n my $sym = [];\n $symbols->{$pc} = $sym;\n foreach my $fullname (split(\"--\", $fullnames)) {\n my $name = ShortFunctionName($fullname);\n push(@{$sym}, $name, \"?\", $fullname);\n }\n }\n return $symbols;\n}\n\nsub BaseName {\n my $file_name = shift;\n $file_name =~ s!^.*/!!; # Remove directory name\n return $file_name;\n}\n\nsub MakeProfileBaseName {\n my ($binary_name, $profile_name) = @_;\n my ($host, $baseURL, $path) = ParseProfileURL($profile_name);\n my $binary_shortname = BaseName($binary_name);\n return sprintf(\"%s.%s.%s\",\n $binary_shortname, $main::op_time, $host);\n}\n\nsub FetchDynamicProfile {\n my $binary_name = shift;\n my $profile_name = shift;\n my $fetch_name_only = shift;\n my $encourage_patience = shift;\n\n if (!IsProfileURL($profile_name)) {\n return $profile_name;\n } else {\n my ($host, $baseURL, $path) = ParseProfileURL($profile_name);\n if ($path eq \"\" || $path eq \"/\") {\n # Missing type specifier defaults to cpu-profile\n $path = $PROFILE_PAGE;\n }\n\n my $profile_file = MakeProfileBaseName($binary_name, $profile_name);\n\n my $url = \"$baseURL$path\";\n my $fetch_timeout = undef;\n if ($path =~ m/$PROFILE_PAGE|$PMUPROFILE_PAGE/) {\n if ($path =~ m/[?]/) {\n $url .= \"&\";\n } else {\n $url .= \"?\";\n }\n $url .= sprintf(\"seconds=%d\", $main::opt_seconds);\n $fetch_timeout = $main::opt_seconds * 1.01 + 60;\n } else {\n # For non-CPU profiles, we add a type-extension to\n # the target profile file name.\n my $suffix = $path;\n $suffix =~ s,/,.,g;\n $profile_file .= $suffix;\n }\n\n my $profile_dir = $ENV{\"PPROF_TMPDIR\"} || ($ENV{HOME} . \"/pprof\");\n if (! -d $profile_dir) {\n mkdir($profile_dir)\n || die(\"Unable to create profile directory $profile_dir: $!\\n\");\n }\n my $tmp_profile = \"$profile_dir/.tmp.$profile_file\";\n my $real_profile = \"$profile_dir/$profile_file\";\n\n if ($fetch_name_only > 0) {\n return $real_profile;\n }\n\n my @fetcher = AddFetchTimeout($fetch_timeout, @URL_FETCHER);\n my $cmd = ShellEscape(@fetcher, $url) . \" > \" . ShellEscape($tmp_profile);\n if ($path =~ m/$PROFILE_PAGE|$PMUPROFILE_PAGE|$CENSUSPROFILE_PAGE/){\n print STDERR \"Gathering CPU profile from $url for $main::opt_seconds seconds to\\n ${real_profile}\\n\";\n if ($encourage_patience) {\n print STDERR \"Be patient...\\n\";\n }\n } else {\n print STDERR \"Fetching $path profile from $url to\\n ${real_profile}\\n\";\n }\n\n (system($cmd) == 0) || error(\"Failed to get profile: $cmd: $!\\n\");\n (system(\"mv\", $tmp_profile, $real_profile) == 0) || error(\"Unable to rename profile\\n\");\n print STDERR \"Wrote profile to $real_profile\\n\";\n $main::collected_profile = $real_profile;\n return $main::collected_profile;\n }\n}\n\n# Collect profiles in parallel\nsub FetchDynamicProfiles {\n my $items = scalar(@main::pfile_args);\n my $levels = log($items) / log(2);\n\n if ($items == 1) {\n $main::profile_files[0] = FetchDynamicProfile($main::prog, $main::pfile_args[0], 0, 1);\n } else {\n # math rounding issues\n if ((2 ** $levels) < $items) {\n $levels++;\n }\n my $count = scalar(@main::pfile_args);\n for (my $i = 0; $i < $count; $i++) {\n $main::profile_files[$i] = FetchDynamicProfile($main::prog, $main::pfile_args[$i], 1, 0);\n }\n print STDERR \"Fetching $count profiles, Be patient...\\n\";\n FetchDynamicProfilesRecurse($levels, 0, 0);\n $main::collected_profile = join(\" \\\\\\n \", @main::profile_files);\n }\n}\n\n# Recursively fork a process to get enough processes\n# collecting profiles\nsub FetchDynamicProfilesRecurse {\n my $maxlevel = shift;\n my $level = shift;\n my $position = shift;\n\n if (my $pid = fork()) {\n $position = 0 | ($position << 1);\n TryCollectProfile($maxlevel, $level, $position);\n wait;\n } else {\n $position = 1 | ($position << 1);\n TryCollectProfile($maxlevel, $level, $position);\n cleanup();\n exit(0);\n }\n}\n\n# Collect a single profile\nsub TryCollectProfile {\n my $maxlevel = shift;\n my $level = shift;\n my $position = shift;\n\n if ($level >= ($maxlevel - 1)) {\n if ($position < scalar(@main::pfile_args)) {\n FetchDynamicProfile($main::prog, $main::pfile_args[$position], 0, 0);\n }\n } else {\n FetchDynamicProfilesRecurse($maxlevel, $level+1, $position);\n }\n}\n\n##### Parsing code #####\n\n# Provide a small streaming-read module to handle very large\n# cpu-profile files. Stream in chunks along a sliding window.\n# Provides an interface to get one 'slot', correctly handling\n# endian-ness differences. A slot is one 32-bit or 64-bit word\n# (depending on the input profile). We tell endianness and bit-size\n# for the profile by looking at the first 8 bytes: in cpu profiles,\n# the second slot is always 3 (we'll accept anything that's not 0).\nBEGIN {\n package CpuProfileStream;\n\n sub new {\n my ($class, $file, $fname) = @_;\n my $self = { file => $file,\n base => 0,\n stride => 512 * 1024, # must be a multiple of bitsize/8\n slots => [],\n unpack_code => \"\", # N for big-endian, V for little\n perl_is_64bit => 1, # matters if profile is 64-bit\n };\n bless $self, $class;\n # Let unittests adjust the stride\n if ($main::opt_test_stride > 0) {\n $self->{stride} = $main::opt_test_stride;\n }\n # Read the first two slots to figure out bitsize and endianness.\n my $slots = $self->{slots};\n my $str;\n read($self->{file}, $str, 8);\n # Set the global $address_length based on what we see here.\n # 8 is 32-bit (8 hexadecimal chars); 16 is 64-bit (16 hexadecimal chars).\n $address_length = ($str eq (chr(0)x8)) ? 16 : 8;\n if ($address_length == 8) {\n if (substr($str, 6, 2) eq chr(0)x2) {\n $self->{unpack_code} = 'V'; # Little-endian.\n } elsif (substr($str, 4, 2) eq chr(0)x2) {\n $self->{unpack_code} = 'N'; # Big-endian\n } else {\n ::error(\"$fname: header size >= 2**16\\n\");\n }\n @$slots = unpack($self->{unpack_code} . \"*\", $str);\n } else {\n # If we're a 64-bit profile, check if we're a 64-bit-capable\n # perl. Otherwise, each slot will be represented as a float\n # instead of an int64, losing precision and making all the\n # 64-bit addresses wrong. We won't complain yet, but will\n # later if we ever see a value that doesn't fit in 32 bits.\n my $has_q = 0;\n eval { $has_q = pack(\"Q\", \"1\") ? 1 : 1; };\n if (!$has_q) {\n $self->{perl_is_64bit} = 0;\n }\n read($self->{file}, $str, 8);\n if (substr($str, 4, 4) eq chr(0)x4) {\n # We'd love to use 'Q', but it's a) not universal, b) not endian-proof.\n $self->{unpack_code} = 'V'; # Little-endian.\n } elsif (substr($str, 0, 4) eq chr(0)x4) {\n $self->{unpack_code} = 'N'; # Big-endian\n } else {\n ::error(\"$fname: header size >= 2**32\\n\");\n }\n my @pair = unpack($self->{unpack_code} . \"*\", $str);\n # Since we know one of the pair is 0, it's fine to just add them.\n @$slots = (0, $pair[0] + $pair[1]);\n }\n return $self;\n }\n\n # Load more data when we access slots->get(X) which is not yet in memory.\n sub overflow {\n my ($self) = @_;\n my $slots = $self->{slots};\n $self->{base} += $#$slots + 1; # skip over data we're replacing\n my $str;\n read($self->{file}, $str, $self->{stride});\n if ($address_length == 8) { # the 32-bit case\n # This is the easy case: unpack provides 32-bit unpacking primitives.\n @$slots = unpack($self->{unpack_code} . \"*\", $str);\n } else {\n # We need to unpack 32 bits at a time and combine.\n my @b32_values = unpack($self->{unpack_code} . \"*\", $str);\n my @b64_values = ();\n for (my $i = 0; $i < $#b32_values; $i += 2) {\n # TODO(csilvers): if this is a 32-bit perl, the math below\n # could end up in a too-large int, which perl will promote\n # to a double, losing necessary precision. Deal with that.\n # Right now, we just die.\n my ($lo, $hi) = ($b32_values[$i], $b32_values[$i+1]);\n if ($self->{unpack_code} eq 'N') { # big-endian\n ($lo, $hi) = ($hi, $lo);\n }\n my $value = $lo + $hi * (2**32);\n if (!$self->{perl_is_64bit} && # check value is exactly represented\n (($value % (2**32)) != $lo || int($value / (2**32)) != $hi)) {\n ::error(\"Need a 64-bit perl to process this 64-bit profile.\\n\");\n }\n push(@b64_values, $value);\n }\n @$slots = @b64_values;\n }\n }\n\n # Access the i-th long in the file (logically), or -1 at EOF.\n sub get {\n my ($self, $idx) = @_;\n my $slots = $self->{slots};\n while ($#$slots >= 0) {\n if ($idx < $self->{base}) {\n # The only time we expect a reference to $slots[$i - something]\n # after referencing $slots[$i] is reading the very first header.\n # Since $stride > |header|, that shouldn't cause any lookback\n # errors. And everything after the header is sequential.\n print STDERR \"Unexpected look-back reading CPU profile\";\n return -1; # shrug, don't know what better to return\n } elsif ($idx > $self->{base} + $#$slots) {\n $self->overflow();\n } else {\n return $slots->[$idx - $self->{base}];\n }\n }\n # If we get here, $slots is [], which means we've reached EOF\n return -1; # unique since slots is supposed to hold unsigned numbers\n }\n}\n\n# Reads the top, 'header' section of a profile, and returns the last\n# line of the header, commonly called a 'header line'. The header\n# section of a profile consists of zero or more 'command' lines that\n# are instructions to pprof, which pprof executes when reading the\n# header. All 'command' lines start with a %. After the command\n# lines is the 'header line', which is a profile-specific line that\n# indicates what type of profile it is, and perhaps other global\n# information about the profile. For instance, here's a header line\n# for a heap profile:\n# heap profile: 53: 38236 [ 5525: 1284029] @ heapprofile\n# For historical reasons, the CPU profile does not contain a text-\n# readable header line. If the profile looks like a CPU profile,\n# this function returns \"\". If no header line could be found, this\n# function returns undef.\n#\n# The following commands are recognized:\n# %warn -- emit the rest of this line to stderr, prefixed by 'WARNING:'\n#\n# The input file should be in binmode.\nsub ReadProfileHeader {\n local *PROFILE = shift;\n my $firstchar = \"\";\n my $line = \"\";\n read(PROFILE, $firstchar, 1);\n seek(PROFILE, -1, 1); # unread the firstchar\n if ($firstchar !~ /[[:print:]]/) { # is not a text character\n return \"\";\n }\n while (defined($line = )) {\n $line =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n if ($line =~ /^%warn\\s+(.*)/) { # 'warn' command\n # Note this matches both '%warn blah\\n' and '%warn\\n'.\n print STDERR \"WARNING: $1\\n\"; # print the rest of the line\n } elsif ($line =~ /^%/) {\n print STDERR \"Ignoring unknown command from profile header: $line\";\n } else {\n # End of commands, must be the header line.\n return $line;\n }\n }\n return undef; # got to EOF without seeing a header line\n}\n\nsub IsSymbolizedProfileFile {\n my $file_name = shift;\n if (!(-e $file_name) || !(-r $file_name)) {\n return 0;\n }\n # Check if the file contains a symbol-section marker.\n open(TFILE, \"<$file_name\");\n binmode TFILE;\n my $firstline = ReadProfileHeader(*TFILE);\n close(TFILE);\n if (!$firstline) {\n return 0;\n }\n $SYMBOL_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $symbol_marker = $&;\n return $firstline =~ /^--- *$symbol_marker/;\n}\n\n# Parse profile generated by common/profiler.cc and return a reference\n# to a map:\n# $result->{version} Version number of profile file\n# $result->{period} Sampling period (in microseconds)\n# $result->{profile} Profile object\n# $result->{map} Memory map info from profile\n# $result->{pcs} Hash of all PC values seen, key is hex address\nsub ReadProfile {\n my $prog = shift;\n my $fname = shift;\n my $result; # return value\n\n $CONTENTION_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $contention_marker = $&;\n $GROWTH_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $growth_marker = $&;\n $SYMBOL_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $symbol_marker = $&;\n $PROFILE_PAGE =~ m,[^/]+$,; # matches everything after the last slash\n my $profile_marker = $&;\n\n # Look at first line to see if it is a heap or a CPU profile.\n # CPU profile may start with no header at all, and just binary data\n # (starting with \\0\\0\\0\\0) -- in that case, don't try to read the\n # whole firstline, since it may be gigabytes(!) of data.\n open(PROFILE, \"<$fname\") || error(\"$fname: $!\\n\");\n binmode PROFILE; # New perls do UTF-8 processing\n my $header = ReadProfileHeader(*PROFILE);\n if (!defined($header)) { # means \"at EOF\"\n error(\"Profile is empty.\\n\");\n }\n\n my $symbols;\n if ($header =~ m/^--- *$symbol_marker/o) {\n # Verify that the user asked for a symbolized profile\n if (!$main::use_symbolized_profile) {\n # we have both a binary and symbolized profiles, abort\n error(\"FATAL ERROR: Symbolized profile\\n $fname\\ncannot be used with \" .\n \"a binary arg. Try again without passing\\n $prog\\n\");\n }\n # Read the symbol section of the symbolized profile file.\n $symbols = ReadSymbols(*PROFILE{IO});\n # Read the next line to get the header for the remaining profile.\n $header = ReadProfileHeader(*PROFILE) || \"\";\n }\n\n $main::profile_type = '';\n if ($header =~ m/^heap profile:.*$growth_marker/o) {\n $main::profile_type = 'growth';\n $result = ReadHeapProfile($prog, *PROFILE, $header);\n } elsif ($header =~ m/^heap profile:/) {\n $main::profile_type = 'heap';\n $result = ReadHeapProfile($prog, *PROFILE, $header);\n } elsif ($header =~ m/^--- *$contention_marker/o) {\n $main::profile_type = 'contention';\n $result = ReadSynchProfile($prog, *PROFILE);\n } elsif ($header =~ m/^--- *Stacks:/) {\n print STDERR\n \"Old format contention profile: mistakenly reports \" .\n \"condition variable signals as lock contentions.\\n\";\n $main::profile_type = 'contention';\n $result = ReadSynchProfile($prog, *PROFILE);\n } elsif ($header =~ m/^--- *$profile_marker/) {\n # the binary cpu profile data starts immediately after this line\n $main::profile_type = 'cpu';\n $result = ReadCPUProfile($prog, $fname, *PROFILE);\n } else {\n if (defined($symbols)) {\n # a symbolized profile contains a format we don't recognize, bail out\n error(\"$fname: Cannot recognize profile section after symbols.\\n\");\n }\n # no ascii header present -- must be a CPU profile\n $main::profile_type = 'cpu';\n $result = ReadCPUProfile($prog, $fname, *PROFILE);\n }\n\n close(PROFILE);\n\n # if we got symbols along with the profile, return those as well\n if (defined($symbols)) {\n $result->{symbols} = $symbols;\n }\n\n return $result;\n}\n\n# Subtract one from caller pc so we map back to call instr.\n# However, don't do this if we're reading a symbolized profile\n# file, in which case the subtract-one was done when the file\n# was written.\n#\n# We apply the same logic to all readers, though ReadCPUProfile uses an\n# independent implementation.\nsub FixCallerAddresses {\n my $stack = shift;\n if ($main::use_symbolized_profile) {\n return $stack;\n } else {\n $stack =~ /(\\s)/;\n my $delimiter = $1;\n my @addrs = split(' ', $stack);\n my @fixedaddrs;\n $#fixedaddrs = $#addrs;\n if ($#addrs >= 0) {\n $fixedaddrs[0] = $addrs[0];\n }\n for (my $i = 1; $i <= $#addrs; $i++) {\n $fixedaddrs[$i] = AddressSub($addrs[$i], \"0x1\");\n }\n return join $delimiter, @fixedaddrs;\n }\n}\n\n# CPU profile reader\nsub ReadCPUProfile {\n my $prog = shift;\n my $fname = shift; # just used for logging\n local *PROFILE = shift;\n my $version;\n my $period;\n my $i;\n my $profile = {};\n my $pcs = {};\n\n # Parse string into array of slots.\n my $slots = CpuProfileStream->new(*PROFILE, $fname);\n\n # Read header. The current header version is a 5-element structure\n # containing:\n # 0: header count (always 0)\n # 1: header \"words\" (after this one: 3)\n # 2: format version (0)\n # 3: sampling period (usec)\n # 4: unused padding (always 0)\n if ($slots->get(0) != 0 ) {\n error(\"$fname: not a profile file, or old format profile file\\n\");\n }\n $i = 2 + $slots->get(1);\n $version = $slots->get(2);\n $period = $slots->get(3);\n # Do some sanity checking on these header values.\n if ($version > (2**32) || $period > (2**32) || $i > (2**32) || $i < 5) {\n error(\"$fname: not a profile file, or corrupted profile file\\n\");\n }\n\n # Parse profile\n while ($slots->get($i) != -1) {\n my $n = $slots->get($i++);\n my $d = $slots->get($i++);\n if ($d > (2**16)) { # TODO(csilvers): what's a reasonable max-stack-depth?\n my $addr = sprintf(\"0%o\", $i * ($address_length == 8 ? 4 : 8));\n print STDERR \"At index $i (address $addr):\\n\";\n error(\"$fname: stack trace depth >= 2**32\\n\");\n }\n if ($slots->get($i) == 0) {\n # End of profile data marker\n $i += $d;\n last;\n }\n\n # Make key out of the stack entries\n my @k = ();\n for (my $j = 0; $j < $d; $j++) {\n my $pc = $slots->get($i+$j);\n # Subtract one from caller pc so we map back to call instr.\n # However, don't do this if we're reading a symbolized profile\n # file, in which case the subtract-one was done when the file\n # was written.\n if ($j > 0 && !$main::use_symbolized_profile) {\n $pc--;\n }\n $pc = sprintf(\"%0*x\", $address_length, $pc);\n $pcs->{$pc} = 1;\n push @k, $pc;\n }\n\n AddEntry($profile, (join \"\\n\", @k), $n);\n $i += $d;\n }\n\n # Parse map\n my $map = '';\n seek(PROFILE, $i * 4, 0);\n read(PROFILE, $map, (stat PROFILE)[7]);\n\n my $r = {};\n $r->{version} = $version;\n $r->{period} = $period;\n $r->{profile} = $profile;\n $r->{libs} = ParseLibraries($prog, $map, $pcs);\n $r->{pcs} = $pcs;\n\n return $r;\n}\n\nsub ReadHeapProfile {\n my $prog = shift;\n local *PROFILE = shift;\n my $header = shift;\n\n my $index = 1;\n if ($main::opt_inuse_space) {\n $index = 1;\n } elsif ($main::opt_inuse_objects) {\n $index = 0;\n } elsif ($main::opt_alloc_space) {\n $index = 3;\n } elsif ($main::opt_alloc_objects) {\n $index = 2;\n }\n\n # Find the type of this profile. The header line looks like:\n # heap profile: 1246: 8800744 [ 1246: 8800744] @ /266053\n # There are two pairs , the first inuse objects/space, and the\n # second allocated objects/space. This is followed optionally by a profile\n # type, and if that is present, optionally by a sampling frequency.\n # For remote heap profiles (v1):\n # The interpretation of the sampling frequency is that the profiler, for\n # each sample, calculates a uniformly distributed random integer less than\n # the given value, and records the next sample after that many bytes have\n # been allocated. Therefore, the expected sample interval is half of the\n # given frequency. By default, if not specified, the expected sample\n # interval is 128KB. Only remote-heap-page profiles are adjusted for\n # sample size.\n # For remote heap profiles (v2):\n # The sampling frequency is the rate of a Poisson process. This means that\n # the probability of sampling an allocation of size X with sampling rate Y\n # is 1 - exp(-X/Y)\n # For version 2, a typical header line might look like this:\n # heap profile: 1922: 127792360 [ 1922: 127792360] @ _v2/524288\n # the trailing number (524288) is the sampling rate. (Version 1 showed\n # double the 'rate' here)\n my $sampling_algorithm = 0;\n my $sample_adjustment = 0;\n chomp($header);\n my $type = \"unknown\";\n if ($header =~ m\"^heap profile:\\s*(\\d+):\\s+(\\d+)\\s+\\[\\s*(\\d+):\\s+(\\d+)\\](\\s*@\\s*([^/]*)(/(\\d+))?)?\") {\n if (defined($6) && ($6 ne '')) {\n $type = $6;\n my $sample_period = $8;\n # $type is \"heapprofile\" for profiles generated by the\n # heap-profiler, and either \"heap\" or \"heap_v2\" for profiles\n # generated by sampling directly within tcmalloc. It can also\n # be \"growth\" for heap-growth profiles. The first is typically\n # found for profiles generated locally, and the others for\n # remote profiles.\n if (($type eq \"heapprofile\") || ($type !~ /heap/) ) {\n # No need to adjust for the sampling rate with heap-profiler-derived data\n $sampling_algorithm = 0;\n } elsif ($type =~ /_v2/) {\n $sampling_algorithm = 2; # version 2 sampling\n if (defined($sample_period) && ($sample_period ne '')) {\n $sample_adjustment = int($sample_period);\n }\n } else {\n $sampling_algorithm = 1; # version 1 sampling\n if (defined($sample_period) && ($sample_period ne '')) {\n $sample_adjustment = int($sample_period)/2;\n }\n }\n } else {\n # We detect whether or not this is a remote-heap profile by checking\n # that the total-allocated stats ($n2,$s2) are exactly the\n # same as the in-use stats ($n1,$s1). It is remotely conceivable\n # that a non-remote-heap profile may pass this check, but it is hard\n # to imagine how that could happen.\n # In this case it's so old it's guaranteed to be remote-heap version 1.\n my ($n1, $s1, $n2, $s2) = ($1, $2, $3, $4);\n if (($n1 == $n2) && ($s1 == $s2)) {\n # This is likely to be a remote-heap based sample profile\n $sampling_algorithm = 1;\n }\n }\n }\n\n if ($sampling_algorithm > 0) {\n # For remote-heap generated profiles, adjust the counts and sizes to\n # account for the sample rate (we sample once every 128KB by default).\n if ($sample_adjustment == 0) {\n # Turn on profile adjustment.\n $sample_adjustment = 128*1024;\n print STDERR \"Adjusting heap profiles for 1-in-128KB sampling rate\\n\";\n } else {\n printf STDERR (\"Adjusting heap profiles for 1-in-%d sampling rate\\n\",\n $sample_adjustment);\n }\n if ($sampling_algorithm > 1) {\n # We don't bother printing anything for the original version (version 1)\n printf STDERR \"Heap version $sampling_algorithm\\n\";\n }\n }\n\n my $profile = {};\n my $pcs = {};\n my $map = \"\";\n\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n if (/^MAPPED_LIBRARIES:/) {\n # Read the /proc/self/maps data\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n $map .= $_;\n }\n last;\n }\n\n if (/^--- Memory map:/) {\n # Read /proc/self/maps data as formatted by DumpAddressMap()\n my $buildvar = \"\";\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n # Parse \"build=\" specification if supplied\n if (m/^\\s*build=(.*)\\n/) {\n $buildvar = $1;\n }\n\n # Expand \"$build\" variable if available\n $_ =~ s/\\$build\\b/$buildvar/g;\n\n $map .= $_;\n }\n last;\n }\n\n # Read entry of the form:\n # : [: ] @ a1 a2 a3 ... an\n s/^\\s*//;\n s/\\s*$//;\n if (m/^\\s*(\\d+):\\s+(\\d+)\\s+\\[\\s*(\\d+):\\s+(\\d+)\\]\\s+@\\s+(.*)$/) {\n my $stack = $5;\n my ($n1, $s1, $n2, $s2) = ($1, $2, $3, $4);\n\n if ($sample_adjustment) {\n if ($sampling_algorithm == 2) {\n # Remote-heap version 2\n # The sampling frequency is the rate of a Poisson process.\n # This means that the probability of sampling an allocation of\n # size X with sampling rate Y is 1 - exp(-X/Y)\n if ($n1 != 0) {\n my $ratio = (($s1*1.0)/$n1)/($sample_adjustment);\n my $scale_factor = 1/(1 - exp(-$ratio));\n $n1 *= $scale_factor;\n $s1 *= $scale_factor;\n }\n if ($n2 != 0) {\n my $ratio = (($s2*1.0)/$n2)/($sample_adjustment);\n my $scale_factor = 1/(1 - exp(-$ratio));\n $n2 *= $scale_factor;\n $s2 *= $scale_factor;\n }\n } else {\n # Remote-heap version 1\n my $ratio;\n $ratio = (($s1*1.0)/$n1)/($sample_adjustment);\n if ($ratio < 1) {\n $n1 /= $ratio;\n $s1 /= $ratio;\n }\n $ratio = (($s2*1.0)/$n2)/($sample_adjustment);\n if ($ratio < 1) {\n $n2 /= $ratio;\n $s2 /= $ratio;\n }\n }\n }\n\n my @counts = ($n1, $s1, $n2, $s2);\n AddEntries($profile, $pcs, FixCallerAddresses($stack), $counts[$index]);\n }\n }\n\n my $r = {};\n $r->{version} = \"heap\";\n $r->{period} = 1;\n $r->{profile} = $profile;\n $r->{libs} = ParseLibraries($prog, $map, $pcs);\n $r->{pcs} = $pcs;\n return $r;\n}\n\nsub ReadSynchProfile {\n my $prog = shift;\n local *PROFILE = shift;\n my $header = shift;\n\n my $map = '';\n my $profile = {};\n my $pcs = {};\n my $sampling_period = 1;\n my $cyclespernanosec = 2.8; # Default assumption for old binaries\n my $seen_clockrate = 0;\n my $line;\n\n my $index = 0;\n if ($main::opt_total_delay) {\n $index = 0;\n } elsif ($main::opt_contentions) {\n $index = 1;\n } elsif ($main::opt_mean_delay) {\n $index = 2;\n }\n\n while ( $line = ) {\n $line =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n if ( $line =~ /^\\s*(\\d+)\\s+(\\d+) \\@\\s*(.*?)\\s*$/ ) {\n my ($cycles, $count, $stack) = ($1, $2, $3);\n\n # Convert cycles to nanoseconds\n $cycles /= $cyclespernanosec;\n\n # Adjust for sampling done by application\n $cycles *= $sampling_period;\n $count *= $sampling_period;\n\n my @values = ($cycles, $count, $cycles / $count);\n AddEntries($profile, $pcs, FixCallerAddresses($stack), $values[$index]);\n\n } elsif ( $line =~ /^(slow release).*thread \\d+ \\@\\s*(.*?)\\s*$/ ||\n $line =~ /^\\s*(\\d+) \\@\\s*(.*?)\\s*$/ ) {\n my ($cycles, $stack) = ($1, $2);\n if ($cycles !~ /^\\d+$/) {\n next;\n }\n\n # Convert cycles to nanoseconds\n $cycles /= $cyclespernanosec;\n\n # Adjust for sampling done by application\n $cycles *= $sampling_period;\n\n AddEntries($profile, $pcs, FixCallerAddresses($stack), $cycles);\n\n } elsif ( $line =~ m/^([a-z][^=]*)=(.*)$/ ) {\n my ($variable, $value) = ($1,$2);\n for ($variable, $value) {\n s/^\\s+//;\n s/\\s+$//;\n }\n if ($variable eq \"cycles/second\") {\n $cyclespernanosec = $value / 1e9;\n $seen_clockrate = 1;\n } elsif ($variable eq \"sampling period\") {\n $sampling_period = $value;\n } elsif ($variable eq \"ms since reset\") {\n # Currently nothing is done with this value in pprof\n # So we just silently ignore it for now\n } elsif ($variable eq \"discarded samples\") {\n # Currently nothing is done with this value in pprof\n # So we just silently ignore it for now\n } else {\n printf STDERR (\"Ignoring unnknown variable in /contention output: \" .\n \"'%s' = '%s'\\n\",$variable,$value);\n }\n } else {\n # Memory map entry\n $map .= $line;\n }\n }\n\n if (!$seen_clockrate) {\n printf STDERR (\"No cycles/second entry in profile; Guessing %.1f GHz\\n\",\n $cyclespernanosec);\n }\n\n my $r = {};\n $r->{version} = 0;\n $r->{period} = $sampling_period;\n $r->{profile} = $profile;\n $r->{libs} = ParseLibraries($prog, $map, $pcs);\n $r->{pcs} = $pcs;\n return $r;\n}\n\n# Given a hex value in the form \"0x1abcd\" or \"1abcd\", return either\n# \"0001abcd\" or \"000000000001abcd\", depending on the current (global)\n# address length.\nsub HexExtend {\n my $addr = shift;\n\n $addr =~ s/^(0x)?0*//;\n my $zeros_needed = $address_length - length($addr);\n if ($zeros_needed < 0) {\n printf STDERR \"Warning: address $addr is longer than address length $address_length\\n\";\n return $addr;\n }\n return (\"0\" x $zeros_needed) . $addr;\n}\n\n##### Symbol extraction #####\n\n# Aggressively search the lib_prefix values for the given library\n# If all else fails, just return the name of the library unmodified.\n# If the lib_prefix is \"/my/path,/other/path\" and $file is \"/lib/dir/mylib.so\"\n# it will search the following locations in this order, until it finds a file:\n# /my/path/lib/dir/mylib.so\n# /other/path/lib/dir/mylib.so\n# /my/path/dir/mylib.so\n# /other/path/dir/mylib.so\n# /my/path/mylib.so\n# /other/path/mylib.so\n# /lib/dir/mylib.so (returned as last resort)\nsub FindLibrary {\n my $file = shift;\n my $suffix = $file;\n\n # Search for the library as described above\n do {\n foreach my $prefix (@prefix_list) {\n my $fullpath = $prefix . $suffix;\n if (-e $fullpath) {\n return $fullpath;\n }\n }\n } while ($suffix =~ s|^/[^/]+/|/|);\n return $file;\n}\n\n# Return path to library with debugging symbols.\n# For libc libraries, the copy in /usr/lib/debug contains debugging symbols\nsub DebuggingLibrary {\n my $file = shift;\n if ($file =~ m|^/| && -f \"/usr/lib/debug$file\") {\n return \"/usr/lib/debug$file\";\n }\n return undef;\n}\n\n# Parse text section header of a library using objdump\nsub ParseTextSectionHeaderFromObjdump {\n my $lib = shift;\n\n my $size = undef;\n my $vma;\n my $file_offset;\n # Get objdump output from the library file to figure out how to\n # map between mapped addresses and addresses in the library.\n my $cmd = ShellEscape($obj_tool_map{\"objdump\"}, \"-h\", $lib);\n open(OBJDUMP, \"$cmd |\") || error(\"$cmd: $!\\n\");\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n # Idx Name Size VMA LMA File off Algn\n # 10 .text 00104b2c 420156f0 420156f0 000156f0 2**4\n # For 64-bit objects, VMA and LMA will be 16 hex digits, size and file\n # offset may still be 8. But AddressSub below will still handle that.\n my @x = split;\n if (($#x >= 6) && ($x[1] eq '.text')) {\n $size = $x[2];\n $vma = $x[3];\n $file_offset = $x[5];\n last;\n }\n }\n close(OBJDUMP);\n\n if (!defined($size)) {\n return undef;\n }\n\n my $r = {};\n $r->{size} = $size;\n $r->{vma} = $vma;\n $r->{file_offset} = $file_offset;\n\n return $r;\n}\n\n# Parse text section header of a library using otool (on OS X)\nsub ParseTextSectionHeaderFromOtool {\n my $lib = shift;\n\n my $size = undef;\n my $vma = undef;\n my $file_offset = undef;\n # Get otool output from the library file to figure out how to\n # map between mapped addresses and addresses in the library.\n my $command = ShellEscape($obj_tool_map{\"otool\"}, \"-l\", $lib);\n open(OTOOL, \"$command |\") || error(\"$command: $!\\n\");\n my $cmd = \"\";\n my $sectname = \"\";\n my $segname = \"\";\n foreach my $line () {\n $line =~ s/\\r//g; # turn windows-looking lines into unix-looking lines\n # Load command <#>\n # cmd LC_SEGMENT\n # [...]\n # Section\n # sectname __text\n # segname __TEXT\n # addr 0x000009f8\n # size 0x00018b9e\n # offset 2552\n # align 2^2 (4)\n # We will need to strip off the leading 0x from the hex addresses,\n # and convert the offset into hex.\n if ($line =~ /Load command/) {\n $cmd = \"\";\n $sectname = \"\";\n $segname = \"\";\n } elsif ($line =~ /Section/) {\n $sectname = \"\";\n $segname = \"\";\n } elsif ($line =~ /cmd (\\w+)/) {\n $cmd = $1;\n } elsif ($line =~ /sectname (\\w+)/) {\n $sectname = $1;\n } elsif ($line =~ /segname (\\w+)/) {\n $segname = $1;\n } elsif (!(($cmd eq \"LC_SEGMENT\" || $cmd eq \"LC_SEGMENT_64\") &&\n $sectname eq \"__text\" &&\n $segname eq \"__TEXT\")) {\n next;\n } elsif ($line =~ /\\baddr 0x([0-9a-fA-F]+)/) {\n $vma = $1;\n } elsif ($line =~ /\\bsize 0x([0-9a-fA-F]+)/) {\n $size = $1;\n } elsif ($line =~ /\\boffset ([0-9]+)/) {\n $file_offset = sprintf(\"%016x\", $1);\n }\n if (defined($vma) && defined($size) && defined($file_offset)) {\n last;\n }\n }\n close(OTOOL);\n\n if (!defined($vma) || !defined($size) || !defined($file_offset)) {\n return undef;\n }\n\n my $r = {};\n $r->{size} = $size;\n $r->{vma} = $vma;\n $r->{file_offset} = $file_offset;\n\n return $r;\n}\n\nsub ParseTextSectionHeader {\n # obj_tool_map(\"otool\") is only defined if we're in a Mach-O environment\n if (defined($obj_tool_map{\"otool\"})) {\n my $r = ParseTextSectionHeaderFromOtool(@_);\n if (defined($r)){\n return $r;\n }\n }\n # If otool doesn't work, or we don't have it, fall back to objdump\n return ParseTextSectionHeaderFromObjdump(@_);\n}\n\n# Split /proc/pid/maps dump into a list of libraries\nsub ParseLibraries {\n return if $main::use_symbol_page; # We don't need libraries info.\n my $prog = shift;\n my $map = shift;\n my $pcs = shift;\n\n my $result = [];\n my $h = \"[a-f0-9]+\";\n my $zero_offset = HexExtend(\"0\");\n\n my $buildvar = \"\";\n foreach my $l (split(\"\\n\", $map)) {\n if ($l =~ m/^\\s*build=(.*)$/) {\n $buildvar = $1;\n }\n\n my $start;\n my $finish;\n my $offset;\n my $lib;\n if ($l =~ /^($h)-($h)\\s+..x.\\s+($h)\\s+\\S+:\\S+\\s+\\d+\\s+(\\S+\\.(so|dll|dylib|bundle)((\\.\\d+)+\\w*(\\.\\d+){0,3})?)$/i) {\n # Full line from /proc/self/maps. Example:\n # 40000000-40015000 r-xp 00000000 03:01 12845071 /lib/ld-2.3.2.so\n $start = HexExtend($1);\n $finish = HexExtend($2);\n $offset = HexExtend($3);\n $lib = $4;\n $lib =~ s|\\\\|/|g; # turn windows-style paths into unix-style paths\n } elsif ($l =~ /^\\s*($h)-($h):\\s*(\\S+\\.so(\\.\\d+)*)/) {\n # Cooked line from DumpAddressMap. Example:\n # 40000000-40015000: /lib/ld-2.3.2.so\n $start = HexExtend($1);\n $finish = HexExtend($2);\n $offset = $zero_offset;\n $lib = $3;\n } else {\n next;\n }\n\n # Expand \"$build\" variable if available\n $lib =~ s/\\$build\\b/$buildvar/g;\n\n $lib = FindLibrary($lib);\n\n # Check for pre-relocated libraries, which use pre-relocated symbol tables\n # and thus require adjusting the offset that we'll use to translate\n # VM addresses into symbol table addresses.\n # Only do this if we're not going to fetch the symbol table from a\n # debugging copy of the library.\n if (!DebuggingLibrary($lib)) {\n my $text = ParseTextSectionHeader($lib);\n if (defined($text)) {\n my $vma_offset = AddressSub($text->{vma}, $text->{file_offset});\n $offset = AddressAdd($offset, $vma_offset);\n }\n }\n\n push(@{$result}, [$lib, $start, $finish, $offset]);\n }\n\n # Append special entry for additional library (not relocated)\n if ($main::opt_lib ne \"\") {\n my $text = ParseTextSectionHeader($main::opt_lib);\n if (defined($text)) {\n my $start = $text->{vma};\n my $finish = AddressAdd($start, $text->{size});\n\n push(@{$result}, [$main::opt_lib, $start, $finish, $start]);\n }\n }\n\n # Append special entry for the main program. This covers\n # 0..max_pc_value_seen, so that we assume pc values not found in one\n # of the library ranges will be treated as coming from the main\n # program binary.\n my $min_pc = HexExtend(\"0\");\n my $max_pc = $min_pc; # find the maximal PC value in any sample\n foreach my $pc (keys(%{$pcs})) {\n if (HexExtend($pc) gt $max_pc) { $max_pc = HexExtend($pc); }\n }\n push(@{$result}, [$prog, $min_pc, $max_pc, $zero_offset]);\n\n return $result;\n}\n\n# Add two hex addresses of length $address_length.\n# Run pprof --test for unit test if this is changed.\nsub AddressAdd {\n my $addr1 = shift;\n my $addr2 = shift;\n my $sum;\n\n if ($address_length == 8) {\n # Perl doesn't cope with wraparound arithmetic, so do it explicitly:\n $sum = (hex($addr1)+hex($addr2)) % (0x10000000 * 16);\n return sprintf(\"%08x\", $sum);\n\n } else {\n # Do the addition in 7-nibble chunks to trivialize carry handling.\n\n if ($main::opt_debug and $main::opt_test) {\n print STDERR \"AddressAdd $addr1 + $addr2 = \";\n }\n\n my $a1 = substr($addr1,-7);\n $addr1 = substr($addr1,0,-7);\n my $a2 = substr($addr2,-7);\n $addr2 = substr($addr2,0,-7);\n $sum = hex($a1) + hex($a2);\n my $c = 0;\n if ($sum > 0xfffffff) {\n $c = 1;\n $sum -= 0x10000000;\n }\n my $r = sprintf(\"%07x\", $sum);\n\n $a1 = substr($addr1,-7);\n $addr1 = substr($addr1,0,-7);\n $a2 = substr($addr2,-7);\n $addr2 = substr($addr2,0,-7);\n $sum = hex($a1) + hex($a2) + $c;\n $c = 0;\n if ($sum > 0xfffffff) {\n $c = 1;\n $sum -= 0x10000000;\n }\n $r = sprintf(\"%07x\", $sum) . $r;\n\n $sum = hex($addr1) + hex($addr2) + $c;\n if ($sum > 0xff) { $sum -= 0x100; }\n $r = sprintf(\"%02x\", $sum) . $r;\n\n if ($main::opt_debug and $main::opt_test) { print STDERR \"$r\\n\"; }\n\n return $r;\n }\n}\n\n\n# Subtract two hex addresses of length $address_length.\n# Run pprof --test for unit test if this is changed.\nsub AddressSub {\n my $addr1 = shift;\n my $addr2 = shift;\n my $diff;\n\n if ($address_length == 8) {\n # Perl doesn't cope with wraparound arithmetic, so do it explicitly:\n $diff = (hex($addr1)-hex($addr2)) % (0x10000000 * 16);\n return sprintf(\"%08x\", $diff);\n\n } else {\n # Do the addition in 7-nibble chunks to trivialize borrow handling.\n # if ($main::opt_debug) { print STDERR \"AddressSub $addr1 - $addr2 = \"; }\n\n my $a1 = hex(substr($addr1,-7));\n $addr1 = substr($addr1,0,-7);\n my $a2 = hex(substr($addr2,-7));\n $addr2 = substr($addr2,0,-7);\n my $b = 0;\n if ($a2 > $a1) {\n $b = 1;\n $a1 += 0x10000000;\n }\n $diff = $a1 - $a2;\n my $r = sprintf(\"%07x\", $diff);\n\n $a1 = hex(substr($addr1,-7));\n $addr1 = substr($addr1,0,-7);\n $a2 = hex(substr($addr2,-7)) + $b;\n $addr2 = substr($addr2,0,-7);\n $b = 0;\n if ($a2 > $a1) {\n $b = 1;\n $a1 += 0x10000000;\n }\n $diff = $a1 - $a2;\n $r = sprintf(\"%07x\", $diff) . $r;\n\n $a1 = hex($addr1);\n $a2 = hex($addr2) + $b;\n if ($a2 > $a1) { $a1 += 0x100; }\n $diff = $a1 - $a2;\n $r = sprintf(\"%02x\", $diff) . $r;\n\n # if ($main::opt_debug) { print STDERR \"$r\\n\"; }\n\n return $r;\n }\n}\n\n# Increment a hex addresses of length $address_length.\n# Run pprof --test for unit test if this is changed.\nsub AddressInc {\n my $addr = shift;\n my $sum;\n\n if ($address_length == 8) {\n # Perl doesn't cope with wraparound arithmetic, so do it explicitly:\n $sum = (hex($addr)+1) % (0x10000000 * 16);\n return sprintf(\"%08x\", $sum);\n\n } else {\n # Do the addition in 7-nibble chunks to trivialize carry handling.\n # We are always doing this to step through the addresses in a function,\n # and will almost never overflow the first chunk, so we check for this\n # case and exit early.\n\n # if ($main::opt_debug) { print STDERR \"AddressInc $addr1 = \"; }\n\n my $a1 = substr($addr,-7);\n $addr = substr($addr,0,-7);\n $sum = hex($a1) + 1;\n my $r = sprintf(\"%07x\", $sum);\n if ($sum <= 0xfffffff) {\n $r = $addr . $r;\n # if ($main::opt_debug) { print STDERR \"$r\\n\"; }\n return HexExtend($r);\n } else {\n $r = \"0000000\";\n }\n\n $a1 = substr($addr,-7);\n $addr = substr($addr,0,-7);\n $sum = hex($a1) + 1;\n $r = sprintf(\"%07x\", $sum) . $r;\n if ($sum <= 0xfffffff) {\n $r = $addr . $r;\n # if ($main::opt_debug) { print STDERR \"$r\\n\"; }\n return HexExtend($r);\n } else {\n $r = \"00000000000000\";\n }\n\n $sum = hex($addr) + 1;\n if ($sum > 0xff) { $sum -= 0x100; }\n $r = sprintf(\"%02x\", $sum) . $r;\n\n # if ($main::opt_debug) { print STDERR \"$r\\n\"; }\n return $r;\n }\n}\n\n# Extract symbols for all PC values found in profile\nsub ExtractSymbols {\n my $libs = shift;\n my $pcset = shift;\n\n my $symbols = {};\n\n # Map each PC value to the containing library. To make this faster,\n # we sort libraries by their starting pc value (highest first), and\n # advance through the libraries as we advance the pc. Sometimes the\n # addresses of libraries may overlap with the addresses of the main\n # binary, so to make sure the libraries 'win', we iterate over the\n # libraries in reverse order (which assumes the binary doesn't start\n # in the middle of a library, which seems a fair assumption).\n my @pcs = (sort { $a cmp $b } keys(%{$pcset})); # pcset is 0-extended strings\n foreach my $lib (sort {$b->[1] cmp $a->[1]} @{$libs}) {\n my $libname = $lib->[0];\n my $start = $lib->[1];\n my $finish = $lib->[2];\n my $offset = $lib->[3];\n\n # Get list of pcs that belong in this library.\n my $contained = [];\n my ($start_pc_index, $finish_pc_index);\n # Find smallest finish_pc_index such that $finish < $pc[$finish_pc_index].\n for ($finish_pc_index = $#pcs + 1; $finish_pc_index > 0;\n $finish_pc_index--) {\n last if $pcs[$finish_pc_index - 1] le $finish;\n }\n # Find smallest start_pc_index such that $start <= $pc[$start_pc_index].\n for ($start_pc_index = $finish_pc_index; $start_pc_index > 0;\n $start_pc_index--) {\n last if $pcs[$start_pc_index - 1] lt $start;\n }\n # This keeps PC values higher than $pc[$finish_pc_index] in @pcs,\n # in case there are overlaps in libraries and the main binary.\n @{$contained} = splice(@pcs, $start_pc_index,\n $finish_pc_index - $start_pc_index);\n # Map to symbols\n MapToSymbols($libname, AddressSub($start, $offset), $contained, $symbols);\n }\n\n return $symbols;\n}\n\n# Map list of PC values to symbols for a given image\nsub MapToSymbols {\n my $image = shift;\n my $offset = shift;\n my $pclist = shift;\n my $symbols = shift;\n\n my $debug = 0;\n\n # Ignore empty binaries\n if ($#{$pclist} < 0) { return; }\n\n # Figure out the addr2line command to use\n my $addr2line = $obj_tool_map{\"addr2line\"};\n my $cmd = ShellEscape($addr2line, \"-f\", \"-C\", \"-e\", $image);\n if (exists $obj_tool_map{\"addr2line_pdb\"}) {\n $addr2line = $obj_tool_map{\"addr2line_pdb\"};\n $cmd = ShellEscape($addr2line, \"--demangle\", \"-f\", \"-C\", \"-e\", $image);\n }\n\n # If \"addr2line\" isn't installed on the system at all, just use\n # nm to get what info we can (function names, but not line numbers).\n if (system(ShellEscape($addr2line, \"--help\") . \" >$dev_null 2>&1\") != 0) {\n MapSymbolsWithNM($image, $offset, $pclist, $symbols);\n return;\n }\n\n # \"addr2line -i\" can produce a variable number of lines per input\n # address, with no separator that allows us to tell when data for\n # the next address starts. So we find the address for a special\n # symbol (_fini) and interleave this address between all real\n # addresses passed to addr2line. The name of this special symbol\n # can then be used as a separator.\n $sep_address = undef; # May be filled in by MapSymbolsWithNM()\n my $nm_symbols = {};\n MapSymbolsWithNM($image, $offset, $pclist, $nm_symbols);\n if (defined($sep_address)) {\n # Only add \" -i\" to addr2line if the binary supports it.\n # addr2line --help returns 0, but not if it sees an unknown flag first.\n if (system(\"$cmd -i --help >$dev_null 2>&1\") == 0) {\n $cmd .= \" -i\";\n } else {\n $sep_address = undef; # no need for sep_address if we don't support -i\n }\n }\n\n # Make file with all PC values with intervening 'sep_address' so\n # that we can reliably detect the end of inlined function list\n open(ADDRESSES, \">$main::tmpfile_sym\") || error(\"$main::tmpfile_sym: $!\\n\");\n if ($debug) { print(\"---- $image ---\\n\"); }\n for (my $i = 0; $i <= $#{$pclist}; $i++) {\n # addr2line always reads hex addresses, and does not need '0x' prefix.\n if ($debug) { printf STDERR (\"%s\\n\", $pclist->[$i]); }\n printf ADDRESSES (\"%s\\n\", AddressSub($pclist->[$i], $offset));\n if (defined($sep_address)) {\n printf ADDRESSES (\"%s\\n\", $sep_address);\n }\n }\n close(ADDRESSES);\n if ($debug) {\n print(\"----\\n\");\n system(\"cat\", $main::tmpfile_sym);\n print(\"----\\n\");\n system(\"$cmd < \" . ShellEscape($main::tmpfile_sym));\n print(\"----\\n\");\n }\n\n open(SYMBOLS, \"$cmd <\" . ShellEscape($main::tmpfile_sym) . \" |\")\n || error(\"$cmd: $!\\n\");\n my $count = 0; # Index in pclist\n while () {\n # Read fullfunction and filelineinfo from next pair of lines\n s/\\r?\\n$//g;\n my $fullfunction = $_;\n $_ = ;\n s/\\r?\\n$//g;\n my $filelinenum = $_;\n\n if (defined($sep_address) && $fullfunction eq $sep_symbol) {\n # Terminating marker for data for this address\n $count++;\n next;\n }\n\n $filelinenum =~ s|\\\\|/|g; # turn windows-style paths into unix-style paths\n\n my $pcstr = $pclist->[$count];\n my $function = ShortFunctionName($fullfunction);\n my $nms = $nm_symbols->{$pcstr};\n if (defined($nms)) {\n if ($fullfunction eq '??') {\n # nm found a symbol for us.\n $function = $nms->[0];\n $fullfunction = $nms->[2];\n } else {\n\t# MapSymbolsWithNM tags each routine with its starting address,\n\t# useful in case the image has multiple occurrences of this\n\t# routine. (It uses a syntax that resembles template paramters,\n\t# that are automatically stripped out by ShortFunctionName().)\n\t# addr2line does not provide the same information. So we check\n\t# if nm disambiguated our symbol, and if so take the annotated\n\t# (nm) version of the routine-name. TODO(csilvers): this won't\n\t# catch overloaded, inlined symbols, which nm doesn't see.\n\t# Better would be to do a check similar to nm's, in this fn.\n\tif ($nms->[2] =~ m/^\\Q$function\\E/) { # sanity check it's the right fn\n\t $function = $nms->[0];\n\t $fullfunction = $nms->[2];\n\t}\n }\n }\n \n # Prepend to accumulated symbols for pcstr\n # (so that caller comes before callee)\n my $sym = $symbols->{$pcstr};\n if (!defined($sym)) {\n $sym = [];\n $symbols->{$pcstr} = $sym;\n }\n unshift(@{$sym}, $function, $filelinenum, $fullfunction);\n if ($debug) { printf STDERR (\"%s => [%s]\\n\", $pcstr, join(\" \", @{$sym})); }\n if (!defined($sep_address)) {\n # Inlining is off, so this entry ends immediately\n $count++;\n }\n }\n close(SYMBOLS);\n}\n\n# Use nm to map the list of referenced PCs to symbols. Return true iff we\n# are able to read procedure information via nm.\nsub MapSymbolsWithNM {\n my $image = shift;\n my $offset = shift;\n my $pclist = shift;\n my $symbols = shift;\n\n # Get nm output sorted by increasing address\n my $symbol_table = GetProcedureBoundaries($image, \".\");\n if (!%{$symbol_table}) {\n return 0;\n }\n # Start addresses are already the right length (8 or 16 hex digits).\n my @names = sort { $symbol_table->{$a}->[0] cmp $symbol_table->{$b}->[0] }\n keys(%{$symbol_table});\n\n if ($#names < 0) {\n # No symbols: just use addresses\n foreach my $pc (@{$pclist}) {\n my $pcstr = \"0x\" . $pc;\n $symbols->{$pc} = [$pcstr, \"?\", $pcstr];\n }\n return 0;\n }\n\n # Sort addresses so we can do a join against nm output\n my $index = 0;\n my $fullname = $names[0];\n my $name = ShortFunctionName($fullname);\n foreach my $pc (sort { $a cmp $b } @{$pclist}) {\n # Adjust for mapped offset\n my $mpc = AddressSub($pc, $offset);\n while (($index < $#names) && ($mpc ge $symbol_table->{$fullname}->[1])){\n $index++;\n $fullname = $names[$index];\n $name = ShortFunctionName($fullname);\n }\n if ($mpc lt $symbol_table->{$fullname}->[1]) {\n $symbols->{$pc} = [$name, \"?\", $fullname];\n } else {\n my $pcstr = \"0x\" . $pc;\n $symbols->{$pc} = [$pcstr, \"?\", $pcstr];\n }\n }\n return 1;\n}\n\nsub ShortFunctionName {\n my $function = shift;\n while ($function =~ s/\\([^()]*\\)(\\s*const)?//g) { } # Argument types\n while ($function =~ s/<[^<>]*>//g) { } # Remove template arguments\n $function =~ s/^.*\\s+(\\w+::)/$1/; # Remove leading type\n return $function;\n}\n\n# Trim overly long symbols found in disassembler output\nsub CleanDisassembly {\n my $d = shift;\n while ($d =~ s/\\([^()%]*\\)(\\s*const)?//g) { } # Argument types, not (%rax)\n while ($d =~ s/(\\w+)<[^<>]*>/$1/g) { } # Remove template arguments\n return $d;\n}\n\n# Clean file name for display\nsub CleanFileName {\n my ($f) = @_;\n $f =~ s|^/proc/self/cwd/||;\n $f =~ s|^\\./||;\n return $f;\n}\n\n# Make address relative to section and clean up for display\nsub UnparseAddress {\n my ($offset, $address) = @_;\n $address = AddressSub($address, $offset);\n $address =~ s/^0x//;\n $address =~ s/^0*//;\n return $address;\n}\n\n##### Miscellaneous #####\n\n# Find the right versions of the above object tools to use. The\n# argument is the program file being analyzed, and should be an ELF\n# 32-bit or ELF 64-bit executable file. The location of the tools\n# is determined by considering the following options in this order:\n# 1) --tools option, if set\n# 2) PPROF_TOOLS environment variable, if set\n# 3) the environment\nsub ConfigureObjTools {\n my $prog_file = shift;\n\n # Check for the existence of $prog_file because /usr/bin/file does not\n # predictably return error status in prod.\n (-e $prog_file) || error(\"$prog_file does not exist.\\n\");\n\n my $file_type = undef;\n if (-e \"/usr/bin/file\") {\n # Follow symlinks (at least for systems where \"file\" supports that).\n my $escaped_prog_file = ShellEscape($prog_file);\n $file_type = `/usr/bin/file -L $escaped_prog_file 2>$dev_null ||\n /usr/bin/file $escaped_prog_file`;\n } elsif ($^O == \"MSWin32\") {\n $file_type = \"MS Windows\";\n } else {\n print STDERR \"WARNING: Can't determine the file type of $prog_file\";\n }\n\n if ($file_type =~ /64-bit/) {\n # Change $address_length to 16 if the program file is ELF 64-bit.\n # We can't detect this from many (most?) heap or lock contention\n # profiles, since the actual addresses referenced are generally in low\n # memory even for 64-bit programs.\n $address_length = 16;\n }\n\n if ($file_type =~ /MS Windows/) {\n # For windows, we provide a version of nm and addr2line as part of\n # the opensource release, which is capable of parsing\n # Windows-style PDB executables. It should live in the path, or\n # in the same directory as pprof.\n $obj_tool_map{\"nm_pdb\"} = \"nm-pdb\";\n $obj_tool_map{\"addr2line_pdb\"} = \"addr2line-pdb\";\n }\n\n if ($file_type =~ /Mach-O/) {\n # OS X uses otool to examine Mach-O files, rather than objdump.\n $obj_tool_map{\"otool\"} = \"otool\";\n $obj_tool_map{\"addr2line\"} = \"false\"; # no addr2line\n $obj_tool_map{\"objdump\"} = \"false\"; # no objdump\n }\n\n # Go fill in %obj_tool_map with the pathnames to use:\n foreach my $tool (keys %obj_tool_map) {\n $obj_tool_map{$tool} = ConfigureTool($obj_tool_map{$tool});\n }\n}\n\n# Returns the path of a caller-specified object tool. If --tools or\n# PPROF_TOOLS are specified, then returns the full path to the tool\n# with that prefix. Otherwise, returns the path unmodified (which\n# means we will look for it on PATH).\nsub ConfigureTool {\n my $tool = shift;\n my $path;\n\n # --tools (or $PPROF_TOOLS) is a comma separated list, where each\n # item is either a) a pathname prefix, or b) a map of the form\n # :. First we look for an entry of type (b) for our\n # tool. If one is found, we use it. Otherwise, we consider all the\n # pathname prefixes in turn, until one yields an existing file. If\n # none does, we use a default path.\n my $tools = $main::opt_tools || $ENV{\"PPROF_TOOLS\"} || \"\";\n if ($tools =~ m/(,|^)\\Q$tool\\E:([^,]*)/) {\n $path = $2;\n # TODO(csilvers): sanity-check that $path exists? Hard if it's relative.\n } elsif ($tools ne '') {\n foreach my $prefix (split(',', $tools)) {\n next if ($prefix =~ /:/); # ignore \"tool:fullpath\" entries in the list\n if (-x $prefix . $tool) {\n $path = $prefix . $tool;\n last;\n }\n }\n if (!$path) {\n error(\"No '$tool' found with prefix specified by \" .\n \"--tools (or \\$PPROF_TOOLS) '$tools'\\n\");\n }\n } else {\n # ... otherwise use the version that exists in the same directory as\n # pprof. If there's nothing there, use $PATH.\n $0 =~ m,[^/]*$,; # this is everything after the last slash\n my $dirname = $`; # this is everything up to and including the last slash\n if (-x \"$dirname$tool\") {\n $path = \"$dirname$tool\";\n } else { \n $path = $tool;\n }\n }\n if ($main::opt_debug) { print STDERR \"Using '$path' for '$tool'.\\n\"; }\n return $path;\n}\n\nsub ShellEscape {\n my @escaped_words = ();\n foreach my $word (@_) {\n my $escaped_word = $word;\n if ($word =~ m![^a-zA-Z0-9/.,_=-]!) { # check for anything not in whitelist\n $escaped_word =~ s/'/'\\\\''/;\n $escaped_word = \"'$escaped_word'\";\n }\n push(@escaped_words, $escaped_word);\n }\n return join(\" \", @escaped_words);\n}\n\nsub cleanup {\n unlink($main::tmpfile_sym);\n unlink(keys %main::tempnames);\n\n # We leave any collected profiles in $HOME/pprof in case the user wants\n # to look at them later. We print a message informing them of this.\n if ((scalar(@main::profile_files) > 0) &&\n defined($main::collected_profile)) {\n if (scalar(@main::profile_files) == 1) {\n print STDERR \"Dynamically gathered profile is in $main::collected_profile\\n\";\n }\n print STDERR \"If you want to investigate this profile further, you can do:\\n\";\n print STDERR \"\\n\";\n print STDERR \" pprof \\\\\\n\";\n print STDERR \" $main::prog \\\\\\n\";\n print STDERR \" $main::collected_profile\\n\";\n print STDERR \"\\n\";\n }\n}\n\nsub sighandler {\n cleanup();\n exit(1);\n}\n\nsub error {\n my $msg = shift;\n print STDERR $msg;\n cleanup();\n exit(1);\n}\n\n\n# Run $nm_command and get all the resulting procedure boundaries whose\n# names match \"$regexp\" and returns them in a hashtable mapping from\n# procedure name to a two-element vector of [start address, end address]\nsub GetProcedureBoundariesViaNm {\n my $escaped_nm_command = shift; # shell-escaped\n my $regexp = shift;\n\n my $symbol_table = {};\n open(NM, \"$escaped_nm_command |\") || error(\"$escaped_nm_command: $!\\n\");\n my $last_start = \"0\";\n my $routine = \"\";\n while () {\n s/\\r//g; # turn windows-looking lines into unix-looking lines\n if (m/^\\s*([0-9a-f]+) (.) (..*)/) {\n my $start_val = $1;\n my $type = $2;\n my $this_routine = $3;\n\n # It's possible for two symbols to share the same address, if\n # one is a zero-length variable (like __start_google_malloc) or\n # one symbol is a weak alias to another (like __libc_malloc).\n # In such cases, we want to ignore all values except for the\n # actual symbol, which in nm-speak has type \"T\". The logic\n # below does this, though it's a bit tricky: what happens when\n # we have a series of lines with the same address, is the first\n # one gets queued up to be processed. However, it won't\n # *actually* be processed until later, when we read a line with\n # a different address. That means that as long as we're reading\n # lines with the same address, we have a chance to replace that\n # item in the queue, which we do whenever we see a 'T' entry --\n # that is, a line with type 'T'. If we never see a 'T' entry,\n # we'll just go ahead and process the first entry (which never\n # got touched in the queue), and ignore the others.\n if ($start_val eq $last_start && $type =~ /t/i) {\n # We are the 'T' symbol at this address, replace previous symbol.\n $routine = $this_routine;\n next;\n } elsif ($start_val eq $last_start) {\n # We're not the 'T' symbol at this address, so ignore us.\n next;\n }\n\n if ($this_routine eq $sep_symbol) {\n $sep_address = HexExtend($start_val);\n }\n\n # Tag this routine with the starting address in case the image\n # has multiple occurrences of this routine. We use a syntax\n # that resembles template paramters that are automatically\n # stripped out by ShortFunctionName()\n $this_routine .= \"<$start_val>\";\n\n if (defined($routine) && $routine =~ m/$regexp/) {\n $symbol_table->{$routine} = [HexExtend($last_start),\n HexExtend($start_val)];\n }\n $last_start = $start_val;\n $routine = $this_routine;\n } elsif (m/^Loaded image name: (.+)/) {\n # The win32 nm workalike emits information about the binary it is using.\n if ($main::opt_debug) { print STDERR \"Using Image $1\\n\"; }\n } elsif (m/^PDB file name: (.+)/) {\n # The win32 nm workalike emits information about the pdb it is using.\n if ($main::opt_debug) { print STDERR \"Using PDB $1\\n\"; }\n }\n }\n close(NM);\n # Handle the last line in the nm output. Unfortunately, we don't know\n # how big this last symbol is, because we don't know how big the file\n # is. For now, we just give it a size of 0.\n # TODO(csilvers): do better here.\n if (defined($routine) && $routine =~ m/$regexp/) {\n $symbol_table->{$routine} = [HexExtend($last_start),\n HexExtend($last_start)];\n }\n return $symbol_table;\n}\n\n# Gets the procedure boundaries for all routines in \"$image\" whose names\n# match \"$regexp\" and returns them in a hashtable mapping from procedure\n# name to a two-element vector of [start address, end address].\n# Will return an empty map if nm is not installed or not working properly.\nsub GetProcedureBoundaries {\n my $image = shift;\n my $regexp = shift;\n\n # If $image doesn't start with /, then put ./ in front of it. This works\n # around an obnoxious bug in our probing of nm -f behavior.\n # \"nm -f $image\" is supposed to fail on GNU nm, but if:\n #\n # a. $image starts with [BbSsPp] (for example, bin/foo/bar), AND\n # b. you have a.out in your current directory (a not uncommon occurence)\n #\n # then \"nm -f $image\" succeeds because -f only looks at the first letter of\n # the argument, which looks valid because it's [BbSsPp], and then since\n # there's no image provided, it looks for a.out and finds it.\n #\n # This regex makes sure that $image starts with . or /, forcing the -f\n # parsing to fail since . and / are not valid formats.\n $image =~ s#^[^/]#./$&#;\n\n # For libc libraries, the copy in /usr/lib/debug contains debugging symbols\n my $debugging = DebuggingLibrary($image);\n if ($debugging) {\n $image = $debugging;\n }\n\n my $nm = $obj_tool_map{\"nm\"};\n my $cppfilt = $obj_tool_map{\"c++filt\"};\n\n # nm can fail for two reasons: 1) $image isn't a debug library; 2) nm\n # binary doesn't support --demangle. In addition, for OS X we need\n # to use the -f flag to get 'flat' nm output (otherwise we don't sort\n # properly and get incorrect results). Unfortunately, GNU nm uses -f\n # in an incompatible way. So first we test whether our nm supports\n # --demangle and -f.\n my $demangle_flag = \"\";\n my $cppfilt_flag = \"\";\n my $to_devnull = \">$dev_null 2>&1\";\n if (system(ShellEscape($nm, \"--demangle\", \"image\") . $to_devnull) == 0) {\n # In this mode, we do \"nm --demangle \"\n $demangle_flag = \"--demangle\";\n $cppfilt_flag = \"\";\n } elsif (system(ShellEscape($cppfilt, $image) . $to_devnull) == 0) {\n # In this mode, we do \"nm | c++filt\"\n $cppfilt_flag = \" | \" . ShellEscape($cppfilt);\n };\n my $flatten_flag = \"\";\n if (system(ShellEscape($nm, \"-f\", $image) . $to_devnull) == 0) {\n $flatten_flag = \"-f\";\n }\n\n # Finally, in the case $imagie isn't a debug library, we try again with\n # -D to at least get *exported* symbols. If we can't use --demangle,\n # we use c++filt instead, if it exists on this system.\n my @nm_commands = (ShellEscape($nm, \"-n\", $flatten_flag, $demangle_flag,\n $image) . \" 2>$dev_null $cppfilt_flag\",\n ShellEscape($nm, \"-D\", \"-n\", $flatten_flag, $demangle_flag,\n $image) . \" 2>$dev_null $cppfilt_flag\",\n # 6nm is for Go binaries\n ShellEscape(\"6nm\", \"$image\") . \" 2>$dev_null | sort\",\n );\n\n # If the executable is an MS Windows PDB-format executable, we'll\n # have set up obj_tool_map(\"nm_pdb\"). In this case, we actually\n # want to use both unix nm and windows-specific nm_pdb, since\n # PDB-format executables can apparently include dwarf .o files.\n if (exists $obj_tool_map{\"nm_pdb\"}) {\n push(@nm_commands,\n ShellEscape($obj_tool_map{\"nm_pdb\"}, \"--demangle\", $image)\n . \" 2>$dev_null\");\n }\n\n foreach my $nm_command (@nm_commands) {\n my $symbol_table = GetProcedureBoundariesViaNm($nm_command, $regexp);\n return $symbol_table if (%{$symbol_table});\n }\n my $symbol_table = {};\n return $symbol_table;\n}\n\n\n# The test vectors for AddressAdd/Sub/Inc are 8-16-nibble hex strings.\n# To make them more readable, we add underscores at interesting places.\n# This routine removes the underscores, producing the canonical representation\n# used by pprof to represent addresses, particularly in the tested routines.\nsub CanonicalHex {\n my $arg = shift;\n return join '', (split '_',$arg);\n}\n\n\n# Unit test for AddressAdd:\nsub AddressAddUnitTest {\n my $test_data_8 = shift;\n my $test_data_16 = shift;\n my $error_count = 0;\n my $fail_count = 0;\n my $pass_count = 0;\n # print STDERR \"AddressAddUnitTest: \", 1+$#{$test_data_8}, \" tests\\n\";\n\n # First a few 8-nibble addresses. Note that this implementation uses\n # plain old arithmetic, so a quick sanity check along with verifying what\n # happens to overflow (we want it to wrap):\n $address_length = 8;\n foreach my $row (@{$test_data_8}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressAdd ($row->[0], $row->[1]);\n if ($sum ne $row->[2]) {\n printf STDERR \"ERROR: %s != %s + %s = %s\\n\", $sum,\n $row->[0], $row->[1], $row->[2];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressAdd 32-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count = $fail_count;\n $fail_count = 0;\n $pass_count = 0;\n\n # Now 16-nibble addresses.\n $address_length = 16;\n foreach my $row (@{$test_data_16}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressAdd (CanonicalHex($row->[0]), CanonicalHex($row->[1]));\n my $expected = join '', (split '_',$row->[2]);\n if ($sum ne CanonicalHex($row->[2])) {\n printf STDERR \"ERROR: %s != %s + %s = %s\\n\", $sum,\n $row->[0], $row->[1], $row->[2];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressAdd 64-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count += $fail_count;\n\n return $error_count;\n}\n\n\n# Unit test for AddressSub:\nsub AddressSubUnitTest {\n my $test_data_8 = shift;\n my $test_data_16 = shift;\n my $error_count = 0;\n my $fail_count = 0;\n my $pass_count = 0;\n # print STDERR \"AddressSubUnitTest: \", 1+$#{$test_data_8}, \" tests\\n\";\n\n # First a few 8-nibble addresses. Note that this implementation uses\n # plain old arithmetic, so a quick sanity check along with verifying what\n # happens to overflow (we want it to wrap):\n $address_length = 8;\n foreach my $row (@{$test_data_8}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressSub ($row->[0], $row->[1]);\n if ($sum ne $row->[3]) {\n printf STDERR \"ERROR: %s != %s - %s = %s\\n\", $sum,\n $row->[0], $row->[1], $row->[3];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressSub 32-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count = $fail_count;\n $fail_count = 0;\n $pass_count = 0;\n\n # Now 16-nibble addresses.\n $address_length = 16;\n foreach my $row (@{$test_data_16}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressSub (CanonicalHex($row->[0]), CanonicalHex($row->[1]));\n if ($sum ne CanonicalHex($row->[3])) {\n printf STDERR \"ERROR: %s != %s - %s = %s\\n\", $sum,\n $row->[0], $row->[1], $row->[3];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressSub 64-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count += $fail_count;\n\n return $error_count;\n}\n\n\n# Unit test for AddressInc:\nsub AddressIncUnitTest {\n my $test_data_8 = shift;\n my $test_data_16 = shift;\n my $error_count = 0;\n my $fail_count = 0;\n my $pass_count = 0;\n # print STDERR \"AddressIncUnitTest: \", 1+$#{$test_data_8}, \" tests\\n\";\n\n # First a few 8-nibble addresses. Note that this implementation uses\n # plain old arithmetic, so a quick sanity check along with verifying what\n # happens to overflow (we want it to wrap):\n $address_length = 8;\n foreach my $row (@{$test_data_8}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressInc ($row->[0]);\n if ($sum ne $row->[4]) {\n printf STDERR \"ERROR: %s != %s + 1 = %s\\n\", $sum,\n $row->[0], $row->[4];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressInc 32-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count = $fail_count;\n $fail_count = 0;\n $pass_count = 0;\n\n # Now 16-nibble addresses.\n $address_length = 16;\n foreach my $row (@{$test_data_16}) {\n if ($main::opt_debug and $main::opt_test) { print STDERR \"@{$row}\\n\"; }\n my $sum = AddressInc (CanonicalHex($row->[0]));\n if ($sum ne CanonicalHex($row->[4])) {\n printf STDERR \"ERROR: %s != %s + 1 = %s\\n\", $sum,\n $row->[0], $row->[4];\n ++$fail_count;\n } else {\n ++$pass_count;\n }\n }\n printf STDERR \"AddressInc 64-bit tests: %d passes, %d failures\\n\",\n $pass_count, $fail_count;\n $error_count += $fail_count;\n\n return $error_count;\n}\n\n\n# Driver for unit tests.\n# Currently just the address add/subtract/increment routines for 64-bit.\nsub RunUnitTests {\n my $error_count = 0;\n\n # This is a list of tuples [a, b, a+b, a-b, a+1]\n my $unit_test_data_8 = [\n [qw(aaaaaaaa 50505050 fafafafa 5a5a5a5a aaaaaaab)],\n [qw(50505050 aaaaaaaa fafafafa a5a5a5a6 50505051)],\n [qw(ffffffff aaaaaaaa aaaaaaa9 55555555 00000000)],\n [qw(00000001 ffffffff 00000000 00000002 00000002)],\n [qw(00000001 fffffff0 fffffff1 00000011 00000002)],\n ];\n my $unit_test_data_16 = [\n # The implementation handles data in 7-nibble chunks, so those are the\n # interesting boundaries.\n [qw(aaaaaaaa 50505050\n 00_000000f_afafafa 00_0000005_a5a5a5a 00_000000a_aaaaaab)],\n [qw(50505050 aaaaaaaa\n 00_000000f_afafafa ff_ffffffa_5a5a5a6 00_0000005_0505051)],\n [qw(ffffffff aaaaaaaa\n 00_000001a_aaaaaa9 00_0000005_5555555 00_0000010_0000000)],\n [qw(00000001 ffffffff\n 00_0000010_0000000 ff_ffffff0_0000002 00_0000000_0000002)],\n [qw(00000001 fffffff0\n 00_000000f_ffffff1 ff_ffffff0_0000011 00_0000000_0000002)],\n\n [qw(00_a00000a_aaaaaaa 50505050\n 00_a00000f_afafafa 00_a000005_a5a5a5a 00_a00000a_aaaaaab)],\n [qw(0f_fff0005_0505050 aaaaaaaa\n 0f_fff000f_afafafa 0f_ffefffa_5a5a5a6 0f_fff0005_0505051)],\n [qw(00_000000f_fffffff 01_800000a_aaaaaaa\n 01_800001a_aaaaaa9 fe_8000005_5555555 00_0000010_0000000)],\n [qw(00_0000000_0000001 ff_fffffff_fffffff\n 00_0000000_0000000 00_0000000_0000002 00_0000000_0000002)],\n [qw(00_0000000_0000001 ff_fffffff_ffffff0\n ff_fffffff_ffffff1 00_0000000_0000011 00_0000000_0000002)],\n ];\n\n $error_count += AddressAddUnitTest($unit_test_data_8, $unit_test_data_16);\n $error_count += AddressSubUnitTest($unit_test_data_8, $unit_test_data_16);\n $error_count += AddressIncUnitTest($unit_test_data_8, $unit_test_data_16);\n if ($error_count > 0) {\n print STDERR $error_count, \" errors: FAILED\\n\";\n } else {\n print STDERR \"PASS\\n\";\n }\n exit ($error_count);\n}\n" }, { "path": "deps/jemalloc/config.guess", "content": "#! /bin/sh\n# Attempt to guess a canonical system name.\n# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,\n# 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010,\n# 2011, 2012 Free Software Foundation, Inc.\n\ntimestamp='2012-02-10'\n\n# This file is free software; you can redistribute it and/or modify it\n# under the terms of the GNU General Public License as published by\n# the Free Software Foundation; either version 2 of the License, or\n# (at your option) any later version.\n#\n# This program is distributed in the hope that it will be useful, but\n# WITHOUT ANY WARRANTY; without even the implied warranty of\n# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU\n# General Public License for more details.\n#\n# You should have received a copy of the GNU General Public License\n# along with this program; if not, see .\n#\n# As a special exception to the GNU General Public License, if you\n# distribute this file as part of a program that contains a\n# configuration script generated by Autoconf, you may include it under\n# the same distribution terms that you use for the rest of that program.\n\n\n# Originally written by Per Bothner. Please send patches (context\n# diff format) to and include a ChangeLog\n# entry.\n#\n# This script attempts to guess a canonical system name similar to\n# config.sub. If it succeeds, it prints the system name on stdout, and\n# exits with 0. Otherwise, it exits with 1.\n#\n# You can get the latest version of this script from:\n# http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD\n\nme=`echo \"$0\" | sed -e 's,.*/,,'`\n\nusage=\"\\\nUsage: $0 [OPTION]\n\nOutput the configuration name of the system \\`$me' is run on.\n\nOperation modes:\n -h, --help print this help, then exit\n -t, --time-stamp print date of last modification, then exit\n -v, --version print version number, then exit\n\nReport bugs and patches to .\"\n\nversion=\"\\\nGNU config.guess ($timestamp)\n\nOriginally written by Per Bothner.\nCopyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,\n2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012\nFree Software Foundation, Inc.\n\nThis is free software; see the source for copying conditions. There is NO\nwarranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\"\n\nhelp=\"\nTry \\`$me --help' for more information.\"\n\n# Parse command line\nwhile test $# -gt 0 ; do\n case $1 in\n --time-stamp | --time* | -t )\n echo \"$timestamp\" ; exit ;;\n --version | -v )\n echo \"$version\" ; exit ;;\n --help | --h* | -h )\n echo \"$usage\"; exit ;;\n -- ) # Stop option processing\n shift; break ;;\n - )\t# Use stdin as input.\n break ;;\n -* )\n echo \"$me: invalid option $1$help\" >&2\n exit 1 ;;\n * )\n break ;;\n esac\ndone\n\nif test $# != 0; then\n echo \"$me: too many arguments$help\" >&2\n exit 1\nfi\n\ntrap 'exit 1' 1 2 15\n\n# CC_FOR_BUILD -- compiler used by this script. Note that the use of a\n# compiler to aid in system detection is discouraged as it requires\n# temporary files to be created and, as you can see below, it is a\n# headache to deal with in a portable fashion.\n\n# Historically, `CC_FOR_BUILD' used to be named `HOST_CC'. We still\n# use `HOST_CC' if defined, but it is deprecated.\n\n# Portable tmp directory creation inspired by the Autoconf team.\n\nset_cc_for_build='\ntrap \"exitcode=\\$?; (rm -f \\$tmpfiles 2>/dev/null; rmdir \\$tmp 2>/dev/null) && exit \\$exitcode\" 0 ;\ntrap \"rm -f \\$tmpfiles 2>/dev/null; rmdir \\$tmp 2>/dev/null; exit 1\" 1 2 13 15 ;\n: ${TMPDIR=/tmp} ;\n { tmp=`(umask 077 && mktemp -d \"$TMPDIR/cgXXXXXX\") 2>/dev/null` && test -n \"$tmp\" && test -d \"$tmp\" ; } ||\n { test -n \"$RANDOM\" && tmp=$TMPDIR/cg$$-$RANDOM && (umask 077 && mkdir $tmp) ; } ||\n { tmp=$TMPDIR/cg-$$ && (umask 077 && mkdir $tmp) && echo \"Warning: creating insecure temp directory\" >&2 ; } ||\n { echo \"$me: cannot create a temporary directory in $TMPDIR\" >&2 ; exit 1 ; } ;\ndummy=$tmp/dummy ;\ntmpfiles=\"$dummy.c $dummy.o $dummy.rel $dummy\" ;\ncase $CC_FOR_BUILD,$HOST_CC,$CC in\n ,,) echo \"int x;\" > $dummy.c ;\n\tfor c in cc gcc c89 c99 ; do\n\t if ($c -c -o $dummy.o $dummy.c) >/dev/null 2>&1 ; then\n\t CC_FOR_BUILD=\"$c\"; break ;\n\t fi ;\n\tdone ;\n\tif test x\"$CC_FOR_BUILD\" = x ; then\n\t CC_FOR_BUILD=no_compiler_found ;\n\tfi\n\t;;\n ,,*) CC_FOR_BUILD=$CC ;;\n ,*,*) CC_FOR_BUILD=$HOST_CC ;;\nesac ; set_cc_for_build= ;'\n\n# This is needed to find uname on a Pyramid OSx when run in the BSD universe.\n# (ghazi@noc.rutgers.edu 1994-08-24)\nif (test -f /.attbin/uname) >/dev/null 2>&1 ; then\n\tPATH=$PATH:/.attbin ; export PATH\nfi\n\nUNAME_MACHINE=`(uname -m) 2>/dev/null` || UNAME_MACHINE=unknown\nUNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown\nUNAME_SYSTEM=`(uname -s) 2>/dev/null` || UNAME_SYSTEM=unknown\nUNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown\n\n# Note: order is significant - the case branches are not exclusive.\n\ncase \"${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}\" in\n *:NetBSD:*:*)\n\t# NetBSD (nbsd) targets should (where applicable) match one or\n\t# more of the tuples: *-*-netbsdelf*, *-*-netbsdaout*,\n\t# *-*-netbsdecoff* and *-*-netbsd*. For targets that recently\n\t# switched to ELF, *-*-netbsd* would select the old\n\t# object file format. This provides both forward\n\t# compatibility and a consistent mechanism for selecting the\n\t# object file format.\n\t#\n\t# Note: NetBSD doesn't particularly care about the vendor\n\t# portion of the name. We always set it to \"unknown\".\n\tsysctl=\"sysctl -n hw.machine_arch\"\n\tUNAME_MACHINE_ARCH=`(/sbin/$sysctl 2>/dev/null || \\\n\t /usr/sbin/$sysctl 2>/dev/null || echo unknown)`\n\tcase \"${UNAME_MACHINE_ARCH}\" in\n\t armeb) machine=armeb-unknown ;;\n\t arm*) machine=arm-unknown ;;\n\t sh3el) machine=shl-unknown ;;\n\t sh3eb) machine=sh-unknown ;;\n\t sh5el) machine=sh5le-unknown ;;\n\t *) machine=${UNAME_MACHINE_ARCH}-unknown ;;\n\tesac\n\t# The Operating System including object format, if it has switched\n\t# to ELF recently, or will in the future.\n\tcase \"${UNAME_MACHINE_ARCH}\" in\n\t arm*|i386|m68k|ns32k|sh3*|sparc|vax)\n\t\teval $set_cc_for_build\n\t\tif echo __ELF__ | $CC_FOR_BUILD -E - 2>/dev/null \\\n\t\t\t| grep -q __ELF__\n\t\tthen\n\t\t # Once all utilities can be ECOFF (netbsdecoff) or a.out (netbsdaout).\n\t\t # Return netbsd for either. FIX?\n\t\t os=netbsd\n\t\telse\n\t\t os=netbsdelf\n\t\tfi\n\t\t;;\n\t *)\n\t\tos=netbsd\n\t\t;;\n\tesac\n\t# The OS release\n\t# Debian GNU/NetBSD machines have a different userland, and\n\t# thus, need a distinct triplet. However, they do not need\n\t# kernel version information, so it can be replaced with a\n\t# suitable tag, in the style of linux-gnu.\n\tcase \"${UNAME_VERSION}\" in\n\t Debian*)\n\t\trelease='-gnu'\n\t\t;;\n\t *)\n\t\trelease=`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\\./'`\n\t\t;;\n\tesac\n\t# Since CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM:\n\t# contains redundant information, the shorter form:\n\t# CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM is used.\n\techo \"${machine}-${os}${release}\"\n\texit ;;\n *:OpenBSD:*:*)\n\tUNAME_MACHINE_ARCH=`arch | sed 's/OpenBSD.//'`\n\techo ${UNAME_MACHINE_ARCH}-unknown-openbsd${UNAME_RELEASE}\n\texit ;;\n *:ekkoBSD:*:*)\n\techo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}\n\texit ;;\n *:SolidBSD:*:*)\n\techo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE}\n\texit ;;\n macppc:MirBSD:*:*)\n\techo powerpc-unknown-mirbsd${UNAME_RELEASE}\n\texit ;;\n *:MirBSD:*:*)\n\techo ${UNAME_MACHINE}-unknown-mirbsd${UNAME_RELEASE}\n\texit ;;\n alpha:OSF1:*:*)\n\tcase $UNAME_RELEASE in\n\t*4.0)\n\t\tUNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'`\n\t\t;;\n\t*5.*)\n\t\tUNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'`\n\t\t;;\n\tesac\n\t# According to Compaq, /usr/sbin/psrinfo has been available on\n\t# OSF/1 and Tru64 systems produced since 1995. I hope that\n\t# covers most systems running today. This code pipes the CPU\n\t# types through head -n 1, so we only detect the type of CPU 0.\n\tALPHA_CPU_TYPE=`/usr/sbin/psrinfo -v | sed -n -e 's/^ The alpha \\(.*\\) processor.*$/\\1/p' | head -n 1`\n\tcase \"$ALPHA_CPU_TYPE\" in\n\t \"EV4 (21064)\")\n\t\tUNAME_MACHINE=\"alpha\" ;;\n\t \"EV4.5 (21064)\")\n\t\tUNAME_MACHINE=\"alpha\" ;;\n\t \"LCA4 (21066/21068)\")\n\t\tUNAME_MACHINE=\"alpha\" ;;\n\t \"EV5 (21164)\")\n\t\tUNAME_MACHINE=\"alphaev5\" ;;\n\t \"EV5.6 (21164A)\")\n\t\tUNAME_MACHINE=\"alphaev56\" ;;\n\t \"EV5.6 (21164PC)\")\n\t\tUNAME_MACHINE=\"alphapca56\" ;;\n\t \"EV5.7 (21164PC)\")\n\t\tUNAME_MACHINE=\"alphapca57\" ;;\n\t \"EV6 (21264)\")\n\t\tUNAME_MACHINE=\"alphaev6\" ;;\n\t \"EV6.7 (21264A)\")\n\t\tUNAME_MACHINE=\"alphaev67\" ;;\n\t \"EV6.8CB (21264C)\")\n\t\tUNAME_MACHINE=\"alphaev68\" ;;\n\t \"EV6.8AL (21264B)\")\n\t\tUNAME_MACHINE=\"alphaev68\" ;;\n\t \"EV6.8CX (21264D)\")\n\t\tUNAME_MACHINE=\"alphaev68\" ;;\n\t \"EV6.9A (21264/EV69A)\")\n\t\tUNAME_MACHINE=\"alphaev69\" ;;\n\t \"EV7 (21364)\")\n\t\tUNAME_MACHINE=\"alphaev7\" ;;\n\t \"EV7.9 (21364A)\")\n\t\tUNAME_MACHINE=\"alphaev79\" ;;\n\tesac\n\t# A Pn.n version is a patched version.\n\t# A Vn.n version is a released version.\n\t# A Tn.n version is a released field test version.\n\t# A Xn.n version is an unreleased experimental baselevel.\n\t# 1.2 uses \"1.2\" for uname -r.\n\techo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[PVTX]//' | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`\n\t# Reset EXIT trap before exiting to avoid spurious non-zero exit code.\n\texitcode=$?\n\ttrap '' 0\n\texit $exitcode ;;\n Alpha\\ *:Windows_NT*:*)\n\t# How do we know it's Interix rather than the generic POSIX subsystem?\n\t# Should we change UNAME_MACHINE based on the output of uname instead\n\t# of the specific Alpha model?\n\techo alpha-pc-interix\n\texit ;;\n 21064:Windows_NT:50:3)\n\techo alpha-dec-winnt3.5\n\texit ;;\n Amiga*:UNIX_System_V:4.0:*)\n\techo m68k-unknown-sysv4\n\texit ;;\n *:[Aa]miga[Oo][Ss]:*:*)\n\techo ${UNAME_MACHINE}-unknown-amigaos\n\texit ;;\n *:[Mm]orph[Oo][Ss]:*:*)\n\techo ${UNAME_MACHINE}-unknown-morphos\n\texit ;;\n *:OS/390:*:*)\n\techo i370-ibm-openedition\n\texit ;;\n *:z/VM:*:*)\n\techo s390-ibm-zvmoe\n\texit ;;\n *:OS400:*:*)\n\techo powerpc-ibm-os400\n\texit ;;\n arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*)\n\techo arm-acorn-riscix${UNAME_RELEASE}\n\texit ;;\n arm:riscos:*:*|arm:RISCOS:*:*)\n\techo arm-unknown-riscos\n\texit ;;\n SR2?01:HI-UX/MPP:*:* | SR8000:HI-UX/MPP:*:*)\n\techo hppa1.1-hitachi-hiuxmpp\n\texit ;;\n Pyramid*:OSx*:*:* | MIS*:OSx*:*:* | MIS*:SMP_DC-OSx*:*:*)\n\t# akee@wpdis03.wpafb.af.mil (Earle F. Ake) contributed MIS and NILE.\n\tif test \"`(/bin/universe) 2>/dev/null`\" = att ; then\n\t\techo pyramid-pyramid-sysv3\n\telse\n\t\techo pyramid-pyramid-bsd\n\tfi\n\texit ;;\n NILE*:*:*:dcosx)\n\techo pyramid-pyramid-svr4\n\texit ;;\n DRS?6000:unix:4.0:6*)\n\techo sparc-icl-nx6\n\texit ;;\n DRS?6000:UNIX_SV:4.2*:7* | DRS?6000:isis:4.2*:7*)\n\tcase `/usr/bin/uname -p` in\n\t sparc) echo sparc-icl-nx7; exit ;;\n\tesac ;;\n s390x:SunOS:*:*)\n\techo ${UNAME_MACHINE}-ibm-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n sun4H:SunOS:5.*:*)\n\techo sparc-hal-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*)\n\techo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n i86pc:AuroraUX:5.*:* | i86xen:AuroraUX:5.*:*)\n\techo i386-pc-auroraux${UNAME_RELEASE}\n\texit ;;\n i86pc:SunOS:5.*:* | i86xen:SunOS:5.*:*)\n\teval $set_cc_for_build\n\tSUN_ARCH=\"i386\"\n\t# If there is a compiler, see if it is configured for 64-bit objects.\n\t# Note that the Sun cc does not turn __LP64__ into 1 like gcc does.\n\t# This test works for both compilers.\n\tif [ \"$CC_FOR_BUILD\" != 'no_compiler_found' ]; then\n\t if (echo '#ifdef __amd64'; echo IS_64BIT_ARCH; echo '#endif') | \\\n\t\t(CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \\\n\t\tgrep IS_64BIT_ARCH >/dev/null\n\t then\n\t\tSUN_ARCH=\"x86_64\"\n\t fi\n\tfi\n\techo ${SUN_ARCH}-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n sun4*:SunOS:6*:*)\n\t# According to config.sub, this is the proper way to canonicalize\n\t# SunOS6. Hard to guess exactly what SunOS6 will be like, but\n\t# it's likely to be more like Solaris than SunOS4.\n\techo sparc-sun-solaris3`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n sun4*:SunOS:*:*)\n\tcase \"`/usr/bin/arch -k`\" in\n\t Series*|S4*)\n\t\tUNAME_RELEASE=`uname -v`\n\t\t;;\n\tesac\n\t# Japanese Language versions have a version number like `4.1.3-JL'.\n\techo sparc-sun-sunos`echo ${UNAME_RELEASE}|sed -e 's/-/_/'`\n\texit ;;\n sun3*:SunOS:*:*)\n\techo m68k-sun-sunos${UNAME_RELEASE}\n\texit ;;\n sun*:*:4.2BSD:*)\n\tUNAME_RELEASE=`(sed 1q /etc/motd | awk '{print substr($5,1,3)}') 2>/dev/null`\n\ttest \"x${UNAME_RELEASE}\" = \"x\" && UNAME_RELEASE=3\n\tcase \"`/bin/arch`\" in\n\t sun3)\n\t\techo m68k-sun-sunos${UNAME_RELEASE}\n\t\t;;\n\t sun4)\n\t\techo sparc-sun-sunos${UNAME_RELEASE}\n\t\t;;\n\tesac\n\texit ;;\n aushp:SunOS:*:*)\n\techo sparc-auspex-sunos${UNAME_RELEASE}\n\texit ;;\n # The situation for MiNT is a little confusing. The machine name\n # can be virtually everything (everything which is not\n # \"atarist\" or \"atariste\" at least should have a processor\n # > m68000). The system name ranges from \"MiNT\" over \"FreeMiNT\"\n # to the lowercase version \"mint\" (or \"freemint\"). Finally\n # the system name \"TOS\" denotes a system which is actually not\n # MiNT. But MiNT is downward compatible to TOS, so this should\n # be no problem.\n atarist[e]:*MiNT:*:* | atarist[e]:*mint:*:* | atarist[e]:*TOS:*:*)\n\techo m68k-atari-mint${UNAME_RELEASE}\n\texit ;;\n atari*:*MiNT:*:* | atari*:*mint:*:* | atarist[e]:*TOS:*:*)\n\techo m68k-atari-mint${UNAME_RELEASE}\n\texit ;;\n *falcon*:*MiNT:*:* | *falcon*:*mint:*:* | *falcon*:*TOS:*:*)\n\techo m68k-atari-mint${UNAME_RELEASE}\n\texit ;;\n milan*:*MiNT:*:* | milan*:*mint:*:* | *milan*:*TOS:*:*)\n\techo m68k-milan-mint${UNAME_RELEASE}\n\texit ;;\n hades*:*MiNT:*:* | hades*:*mint:*:* | *hades*:*TOS:*:*)\n\techo m68k-hades-mint${UNAME_RELEASE}\n\texit ;;\n *:*MiNT:*:* | *:*mint:*:* | *:*TOS:*:*)\n\techo m68k-unknown-mint${UNAME_RELEASE}\n\texit ;;\n m68k:machten:*:*)\n\techo m68k-apple-machten${UNAME_RELEASE}\n\texit ;;\n powerpc:machten:*:*)\n\techo powerpc-apple-machten${UNAME_RELEASE}\n\texit ;;\n RISC*:Mach:*:*)\n\techo mips-dec-mach_bsd4.3\n\texit ;;\n RISC*:ULTRIX:*:*)\n\techo mips-dec-ultrix${UNAME_RELEASE}\n\texit ;;\n VAX*:ULTRIX*:*:*)\n\techo vax-dec-ultrix${UNAME_RELEASE}\n\texit ;;\n 2020:CLIX:*:* | 2430:CLIX:*:*)\n\techo clipper-intergraph-clix${UNAME_RELEASE}\n\texit ;;\n mips:*:*:UMIPS | mips:*:*:RISCos)\n\teval $set_cc_for_build\n\tsed 's/^\t//' << EOF >$dummy.c\n#ifdef __cplusplus\n#include /* for printf() prototype */\n\tint main (int argc, char *argv[]) {\n#else\n\tint main (argc, argv) int argc; char *argv[]; {\n#endif\n\t#if defined (host_mips) && defined (MIPSEB)\n\t#if defined (SYSTYPE_SYSV)\n\t printf (\"mips-mips-riscos%ssysv\\n\", argv[1]); exit (0);\n\t#endif\n\t#if defined (SYSTYPE_SVR4)\n\t printf (\"mips-mips-riscos%ssvr4\\n\", argv[1]); exit (0);\n\t#endif\n\t#if defined (SYSTYPE_BSD43) || defined(SYSTYPE_BSD)\n\t printf (\"mips-mips-riscos%sbsd\\n\", argv[1]); exit (0);\n\t#endif\n\t#endif\n\t exit (-1);\n\t}\nEOF\n\t$CC_FOR_BUILD -o $dummy $dummy.c &&\n\t dummyarg=`echo \"${UNAME_RELEASE}\" | sed -n 's/\\([0-9]*\\).*/\\1/p'` &&\n\t SYSTEM_NAME=`$dummy $dummyarg` &&\n\t { echo \"$SYSTEM_NAME\"; exit; }\n\techo mips-mips-riscos${UNAME_RELEASE}\n\texit ;;\n Motorola:PowerMAX_OS:*:*)\n\techo powerpc-motorola-powermax\n\texit ;;\n Motorola:*:4.3:PL8-*)\n\techo powerpc-harris-powermax\n\texit ;;\n Night_Hawk:*:*:PowerMAX_OS | Synergy:PowerMAX_OS:*:*)\n\techo powerpc-harris-powermax\n\texit ;;\n Night_Hawk:Power_UNIX:*:*)\n\techo powerpc-harris-powerunix\n\texit ;;\n m88k:CX/UX:7*:*)\n\techo m88k-harris-cxux7\n\texit ;;\n m88k:*:4*:R4*)\n\techo m88k-motorola-sysv4\n\texit ;;\n m88k:*:3*:R3*)\n\techo m88k-motorola-sysv3\n\texit ;;\n AViiON:dgux:*:*)\n\t# DG/UX returns AViiON for all architectures\n\tUNAME_PROCESSOR=`/usr/bin/uname -p`\n\tif [ $UNAME_PROCESSOR = mc88100 ] || [ $UNAME_PROCESSOR = mc88110 ]\n\tthen\n\t if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx ] || \\\n\t [ ${TARGET_BINARY_INTERFACE}x = x ]\n\t then\n\t\techo m88k-dg-dgux${UNAME_RELEASE}\n\t else\n\t\techo m88k-dg-dguxbcs${UNAME_RELEASE}\n\t fi\n\telse\n\t echo i586-dg-dgux${UNAME_RELEASE}\n\tfi\n\texit ;;\n M88*:DolphinOS:*:*)\t# DolphinOS (SVR3)\n\techo m88k-dolphin-sysv3\n\texit ;;\n M88*:*:R3*:*)\n\t# Delta 88k system running SVR3\n\techo m88k-motorola-sysv3\n\texit ;;\n XD88*:*:*:*) # Tektronix XD88 system running UTekV (SVR3)\n\techo m88k-tektronix-sysv3\n\texit ;;\n Tek43[0-9][0-9]:UTek:*:*) # Tektronix 4300 system running UTek (BSD)\n\techo m68k-tektronix-bsd\n\texit ;;\n *:IRIX*:*:*)\n\techo mips-sgi-irix`echo ${UNAME_RELEASE}|sed -e 's/-/_/g'`\n\texit ;;\n ????????:AIX?:[12].1:2) # AIX 2.2.1 or AIX 2.1.1 is RT/PC AIX.\n\techo romp-ibm-aix # uname -m gives an 8 hex-code CPU id\n\texit ;; # Note that: echo \"'`uname -s`'\" gives 'AIX '\n i*86:AIX:*:*)\n\techo i386-ibm-aix\n\texit ;;\n ia64:AIX:*:*)\n\tif [ -x /usr/bin/oslevel ] ; then\n\t\tIBM_REV=`/usr/bin/oslevel`\n\telse\n\t\tIBM_REV=${UNAME_VERSION}.${UNAME_RELEASE}\n\tfi\n\techo ${UNAME_MACHINE}-ibm-aix${IBM_REV}\n\texit ;;\n *:AIX:2:3)\n\tif grep bos325 /usr/include/stdio.h >/dev/null 2>&1; then\n\t\teval $set_cc_for_build\n\t\tsed 's/^\t\t//' << EOF >$dummy.c\n\t\t#include \n\n\t\tmain()\n\t\t\t{\n\t\t\tif (!__power_pc())\n\t\t\t\texit(1);\n\t\t\tputs(\"powerpc-ibm-aix3.2.5\");\n\t\t\texit(0);\n\t\t\t}\nEOF\n\t\tif $CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy`\n\t\tthen\n\t\t\techo \"$SYSTEM_NAME\"\n\t\telse\n\t\t\techo rs6000-ibm-aix3.2.5\n\t\tfi\n\telif grep bos324 /usr/include/stdio.h >/dev/null 2>&1; then\n\t\techo rs6000-ibm-aix3.2.4\n\telse\n\t\techo rs6000-ibm-aix3.2\n\tfi\n\texit ;;\n *:AIX:*:[4567])\n\tIBM_CPU_ID=`/usr/sbin/lsdev -C -c processor -S available | sed 1q | awk '{ print $1 }'`\n\tif /usr/sbin/lsattr -El ${IBM_CPU_ID} | grep ' POWER' >/dev/null 2>&1; then\n\t\tIBM_ARCH=rs6000\n\telse\n\t\tIBM_ARCH=powerpc\n\tfi\n\tif [ -x /usr/bin/oslevel ] ; then\n\t\tIBM_REV=`/usr/bin/oslevel`\n\telse\n\t\tIBM_REV=${UNAME_VERSION}.${UNAME_RELEASE}\n\tfi\n\techo ${IBM_ARCH}-ibm-aix${IBM_REV}\n\texit ;;\n *:AIX:*:*)\n\techo rs6000-ibm-aix\n\texit ;;\n ibmrt:4.4BSD:*|romp-ibm:BSD:*)\n\techo romp-ibm-bsd4.4\n\texit ;;\n ibmrt:*BSD:*|romp-ibm:BSD:*) # covers RT/PC BSD and\n\techo romp-ibm-bsd${UNAME_RELEASE} # 4.3 with uname added to\n\texit ;; # report: romp-ibm BSD 4.3\n *:BOSX:*:*)\n\techo rs6000-bull-bosx\n\texit ;;\n DPX/2?00:B.O.S.:*:*)\n\techo m68k-bull-sysv3\n\texit ;;\n 9000/[34]??:4.3bsd:1.*:*)\n\techo m68k-hp-bsd\n\texit ;;\n hp300:4.4BSD:*:* | 9000/[34]??:4.3bsd:2.*:*)\n\techo m68k-hp-bsd4.4\n\texit ;;\n 9000/[34678]??:HP-UX:*:*)\n\tHPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'`\n\tcase \"${UNAME_MACHINE}\" in\n\t 9000/31? ) HP_ARCH=m68000 ;;\n\t 9000/[34]?? ) HP_ARCH=m68k ;;\n\t 9000/[678][0-9][0-9])\n\t\tif [ -x /usr/bin/getconf ]; then\n\t\t sc_cpu_version=`/usr/bin/getconf SC_CPU_VERSION 2>/dev/null`\n\t\t sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null`\n\t\t case \"${sc_cpu_version}\" in\n\t\t 523) HP_ARCH=\"hppa1.0\" ;; # CPU_PA_RISC1_0\n\t\t 528) HP_ARCH=\"hppa1.1\" ;; # CPU_PA_RISC1_1\n\t\t 532) # CPU_PA_RISC2_0\n\t\t\tcase \"${sc_kernel_bits}\" in\n\t\t\t 32) HP_ARCH=\"hppa2.0n\" ;;\n\t\t\t 64) HP_ARCH=\"hppa2.0w\" ;;\n\t\t\t '') HP_ARCH=\"hppa2.0\" ;; # HP-UX 10.20\n\t\t\tesac ;;\n\t\t esac\n\t\tfi\n\t\tif [ \"${HP_ARCH}\" = \"\" ]; then\n\t\t eval $set_cc_for_build\n\t\t sed 's/^\t\t//' << EOF >$dummy.c\n\n\t\t#define _HPUX_SOURCE\n\t\t#include \n\t\t#include \n\n\t\tint main ()\n\t\t{\n\t\t#if defined(_SC_KERNEL_BITS)\n\t\t long bits = sysconf(_SC_KERNEL_BITS);\n\t\t#endif\n\t\t long cpu = sysconf (_SC_CPU_VERSION);\n\n\t\t switch (cpu)\n\t\t\t{\n\t\t\tcase CPU_PA_RISC1_0: puts (\"hppa1.0\"); break;\n\t\t\tcase CPU_PA_RISC1_1: puts (\"hppa1.1\"); break;\n\t\t\tcase CPU_PA_RISC2_0:\n\t\t#if defined(_SC_KERNEL_BITS)\n\t\t\t switch (bits)\n\t\t\t\t{\n\t\t\t\tcase 64: puts (\"hppa2.0w\"); break;\n\t\t\t\tcase 32: puts (\"hppa2.0n\"); break;\n\t\t\t\tdefault: puts (\"hppa2.0\"); break;\n\t\t\t\t} break;\n\t\t#else /* !defined(_SC_KERNEL_BITS) */\n\t\t\t puts (\"hppa2.0\"); break;\n\t\t#endif\n\t\t\tdefault: puts (\"hppa1.0\"); break;\n\t\t\t}\n\t\t exit (0);\n\t\t}\nEOF\n\t\t (CCOPTS= $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null) && HP_ARCH=`$dummy`\n\t\t test -z \"$HP_ARCH\" && HP_ARCH=hppa\n\t\tfi ;;\n\tesac\n\tif [ ${HP_ARCH} = \"hppa2.0w\" ]\n\tthen\n\t eval $set_cc_for_build\n\n\t # hppa2.0w-hp-hpux* has a 64-bit kernel and a compiler generating\n\t # 32-bit code. hppa64-hp-hpux* has the same kernel and a compiler\n\t # generating 64-bit code. GNU and HP use different nomenclature:\n\t #\n\t # $ CC_FOR_BUILD=cc ./config.guess\n\t # => hppa2.0w-hp-hpux11.23\n\t # $ CC_FOR_BUILD=\"cc +DA2.0w\" ./config.guess\n\t # => hppa64-hp-hpux11.23\n\n\t if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) |\n\t\tgrep -q __LP64__\n\t then\n\t\tHP_ARCH=\"hppa2.0w\"\n\t else\n\t\tHP_ARCH=\"hppa64\"\n\t fi\n\tfi\n\techo ${HP_ARCH}-hp-hpux${HPUX_REV}\n\texit ;;\n ia64:HP-UX:*:*)\n\tHPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'`\n\techo ia64-hp-hpux${HPUX_REV}\n\texit ;;\n 3050*:HI-UX:*:*)\n\teval $set_cc_for_build\n\tsed 's/^\t//' << EOF >$dummy.c\n\t#include \n\tint\n\tmain ()\n\t{\n\t long cpu = sysconf (_SC_CPU_VERSION);\n\t /* The order matters, because CPU_IS_HP_MC68K erroneously returns\n\t true for CPU_PA_RISC1_0. CPU_IS_PA_RISC returns correct\n\t results, however. */\n\t if (CPU_IS_PA_RISC (cpu))\n\t {\n\t switch (cpu)\n\t\t{\n\t\t case CPU_PA_RISC1_0: puts (\"hppa1.0-hitachi-hiuxwe2\"); break;\n\t\t case CPU_PA_RISC1_1: puts (\"hppa1.1-hitachi-hiuxwe2\"); break;\n\t\t case CPU_PA_RISC2_0: puts (\"hppa2.0-hitachi-hiuxwe2\"); break;\n\t\t default: puts (\"hppa-hitachi-hiuxwe2\"); break;\n\t\t}\n\t }\n\t else if (CPU_IS_HP_MC68K (cpu))\n\t puts (\"m68k-hitachi-hiuxwe2\");\n\t else puts (\"unknown-hitachi-hiuxwe2\");\n\t exit (0);\n\t}\nEOF\n\t$CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy` &&\n\t\t{ echo \"$SYSTEM_NAME\"; exit; }\n\techo unknown-hitachi-hiuxwe2\n\texit ;;\n 9000/7??:4.3bsd:*:* | 9000/8?[79]:4.3bsd:*:* )\n\techo hppa1.1-hp-bsd\n\texit ;;\n 9000/8??:4.3bsd:*:*)\n\techo hppa1.0-hp-bsd\n\texit ;;\n *9??*:MPE/iX:*:* | *3000*:MPE/iX:*:*)\n\techo hppa1.0-hp-mpeix\n\texit ;;\n hp7??:OSF1:*:* | hp8?[79]:OSF1:*:* )\n\techo hppa1.1-hp-osf\n\texit ;;\n hp8??:OSF1:*:*)\n\techo hppa1.0-hp-osf\n\texit ;;\n i*86:OSF1:*:*)\n\tif [ -x /usr/sbin/sysversion ] ; then\n\t echo ${UNAME_MACHINE}-unknown-osf1mk\n\telse\n\t echo ${UNAME_MACHINE}-unknown-osf1\n\tfi\n\texit ;;\n parisc*:Lites*:*:*)\n\techo hppa1.1-hp-lites\n\texit ;;\n C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*)\n\techo c1-convex-bsd\n\texit ;;\n C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*)\n\tif getsysinfo -f scalar_acc\n\tthen echo c32-convex-bsd\n\telse echo c2-convex-bsd\n\tfi\n\texit ;;\n C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*)\n\techo c34-convex-bsd\n\texit ;;\n C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*)\n\techo c38-convex-bsd\n\texit ;;\n C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*)\n\techo c4-convex-bsd\n\texit ;;\n CRAY*Y-MP:*:*:*)\n\techo ymp-cray-unicos${UNAME_RELEASE} | sed -e 's/\\.[^.]*$/.X/'\n\texit ;;\n CRAY*[A-Z]90:*:*:*)\n\techo ${UNAME_MACHINE}-cray-unicos${UNAME_RELEASE} \\\n\t| sed -e 's/CRAY.*\\([A-Z]90\\)/\\1/' \\\n\t -e y/ABCDEFGHIJKLMNOPQRSTUVWXYZ/abcdefghijklmnopqrstuvwxyz/ \\\n\t -e 's/\\.[^.]*$/.X/'\n\texit ;;\n CRAY*TS:*:*:*)\n\techo t90-cray-unicos${UNAME_RELEASE} | sed -e 's/\\.[^.]*$/.X/'\n\texit ;;\n CRAY*T3E:*:*:*)\n\techo alphaev5-cray-unicosmk${UNAME_RELEASE} | sed -e 's/\\.[^.]*$/.X/'\n\texit ;;\n CRAY*SV1:*:*:*)\n\techo sv1-cray-unicos${UNAME_RELEASE} | sed -e 's/\\.[^.]*$/.X/'\n\texit ;;\n *:UNICOS/mp:*:*)\n\techo craynv-cray-unicosmp${UNAME_RELEASE} | sed -e 's/\\.[^.]*$/.X/'\n\texit ;;\n F30[01]:UNIX_System_V:*:* | F700:UNIX_System_V:*:*)\n\tFUJITSU_PROC=`uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`\n\tFUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\\///'`\n\tFUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'`\n\techo \"${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}\"\n\texit ;;\n 5000:UNIX_System_V:4.*:*)\n\tFUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\\///'`\n\tFUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'`\n\techo \"sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}\"\n\texit ;;\n i*86:BSD/386:*:* | i*86:BSD/OS:*:* | *:Ascend\\ Embedded/OS:*:*)\n\techo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE}\n\texit ;;\n sparc*:BSD/OS:*:*)\n\techo sparc-unknown-bsdi${UNAME_RELEASE}\n\texit ;;\n *:BSD/OS:*:*)\n\techo ${UNAME_MACHINE}-unknown-bsdi${UNAME_RELEASE}\n\texit ;;\n *:FreeBSD:*:*)\n\tUNAME_PROCESSOR=`/usr/bin/uname -p`\n\tcase ${UNAME_PROCESSOR} in\n\t amd64)\n\t\techo x86_64-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;\n\t *)\n\t\techo ${UNAME_PROCESSOR}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;\n\tesac\n\texit ;;\n i*:CYGWIN*:*)\n\techo ${UNAME_MACHINE}-pc-cygwin\n\texit ;;\n *:MINGW*:*)\n\techo ${UNAME_MACHINE}-pc-mingw32\n\texit ;;\n i*:MSYS*:*)\n\techo ${UNAME_MACHINE}-pc-msys\n\texit ;;\n i*:windows32*:*)\n\t# uname -m includes \"-pc\" on this system.\n\techo ${UNAME_MACHINE}-mingw32\n\texit ;;\n i*:PW*:*)\n\techo ${UNAME_MACHINE}-pc-pw32\n\texit ;;\n *:Interix*:*)\n\tcase ${UNAME_MACHINE} in\n\t x86)\n\t\techo i586-pc-interix${UNAME_RELEASE}\n\t\texit ;;\n\t authenticamd | genuineintel | EM64T)\n\t\techo x86_64-unknown-interix${UNAME_RELEASE}\n\t\texit ;;\n\t IA64)\n\t\techo ia64-unknown-interix${UNAME_RELEASE}\n\t\texit ;;\n\tesac ;;\n [345]86:Windows_95:* | [345]86:Windows_98:* | [345]86:Windows_NT:*)\n\techo i${UNAME_MACHINE}-pc-mks\n\texit ;;\n 8664:Windows_NT:*)\n\techo x86_64-pc-mks\n\texit ;;\n i*:Windows_NT*:* | Pentium*:Windows_NT*:*)\n\t# How do we know it's Interix rather than the generic POSIX subsystem?\n\t# It also conflicts with pre-2.0 versions of AT&T UWIN. Should we\n\t# UNAME_MACHINE based on the output of uname instead of i386?\n\techo i586-pc-interix\n\texit ;;\n i*:UWIN*:*)\n\techo ${UNAME_MACHINE}-pc-uwin\n\texit ;;\n amd64:CYGWIN*:*:* | x86_64:CYGWIN*:*:*)\n\techo x86_64-unknown-cygwin\n\texit ;;\n p*:CYGWIN*:*)\n\techo powerpcle-unknown-cygwin\n\texit ;;\n prep*:SunOS:5.*:*)\n\techo powerpcle-unknown-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`\n\texit ;;\n *:GNU:*:*)\n\t# the GNU system\n\techo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-gnu`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'`\n\texit ;;\n *:GNU/*:*:*)\n\t# other systems with GNU libc and userland\n\techo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-gnu\n\texit ;;\n i*86:Minix:*:*)\n\techo ${UNAME_MACHINE}-pc-minix\n\texit ;;\n aarch64:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n aarch64_be:Linux:*:*)\n\tUNAME_MACHINE=aarch64_be\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n alpha:Linux:*:*)\n\tcase `sed -n '/^cpu model/s/^.*: \\(.*\\)/\\1/p' < /proc/cpuinfo` in\n\t EV5) UNAME_MACHINE=alphaev5 ;;\n\t EV56) UNAME_MACHINE=alphaev56 ;;\n\t PCA56) UNAME_MACHINE=alphapca56 ;;\n\t PCA57) UNAME_MACHINE=alphapca56 ;;\n\t EV6) UNAME_MACHINE=alphaev6 ;;\n\t EV67) UNAME_MACHINE=alphaev67 ;;\n\t EV68*) UNAME_MACHINE=alphaev68 ;;\n\tesac\n\tobjdump --private-headers /bin/sh | grep -q ld.so.1\n\tif test \"$?\" = 0 ; then LIBC=\"libc1\" ; else LIBC=\"\" ; fi\n\techo ${UNAME_MACHINE}-unknown-linux-gnu${LIBC}\n\texit ;;\n arm*:Linux:*:*)\n\teval $set_cc_for_build\n\tif echo __ARM_EABI__ | $CC_FOR_BUILD -E - 2>/dev/null \\\n\t | grep -q __ARM_EABI__\n\tthen\n\t echo ${UNAME_MACHINE}-unknown-linux-gnu\n\telse\n\t if echo __ARM_PCS_VFP | $CC_FOR_BUILD -E - 2>/dev/null \\\n\t\t| grep -q __ARM_PCS_VFP\n\t then\n\t\techo ${UNAME_MACHINE}-unknown-linux-gnueabi\n\t else\n\t\techo ${UNAME_MACHINE}-unknown-linux-gnueabihf\n\t fi\n\tfi\n\texit ;;\n avr32*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n cris:Linux:*:*)\n\techo ${UNAME_MACHINE}-axis-linux-gnu\n\texit ;;\n crisv32:Linux:*:*)\n\techo ${UNAME_MACHINE}-axis-linux-gnu\n\texit ;;\n frv:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n hexagon:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n i*86:Linux:*:*)\n\tLIBC=gnu\n\teval $set_cc_for_build\n\tsed 's/^\t//' << EOF >$dummy.c\n\t#ifdef __dietlibc__\n\tLIBC=dietlibc\n\t#endif\nEOF\n\teval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^LIBC'`\n\techo \"${UNAME_MACHINE}-pc-linux-${LIBC}\"\n\texit ;;\n ia64:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n m32r*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n m68*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n mips:Linux:*:* | mips64:Linux:*:*)\n\teval $set_cc_for_build\n\tsed 's/^\t//' << EOF >$dummy.c\n\t#undef CPU\n\t#undef ${UNAME_MACHINE}\n\t#undef ${UNAME_MACHINE}el\n\t#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)\n\tCPU=${UNAME_MACHINE}el\n\t#else\n\t#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)\n\tCPU=${UNAME_MACHINE}\n\t#else\n\tCPU=\n\t#endif\n\t#endif\nEOF\n\teval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^CPU'`\n\ttest x\"${CPU}\" != x && { echo \"${CPU}-unknown-linux-gnu\"; exit; }\n\t;;\n or32:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n padre:Linux:*:*)\n\techo sparc-unknown-linux-gnu\n\texit ;;\n parisc64:Linux:*:* | hppa64:Linux:*:*)\n\techo hppa64-unknown-linux-gnu\n\texit ;;\n parisc:Linux:*:* | hppa:Linux:*:*)\n\t# Look for CPU level\n\tcase `grep '^cpu[^a-z]*:' /proc/cpuinfo 2>/dev/null | cut -d' ' -f2` in\n\t PA7*) echo hppa1.1-unknown-linux-gnu ;;\n\t PA8*) echo hppa2.0-unknown-linux-gnu ;;\n\t *) echo hppa-unknown-linux-gnu ;;\n\tesac\n\texit ;;\n ppc64:Linux:*:*)\n\techo powerpc64-unknown-linux-gnu\n\texit ;;\n ppc:Linux:*:*)\n\techo powerpc-unknown-linux-gnu\n\texit ;;\n s390:Linux:*:* | s390x:Linux:*:*)\n\techo ${UNAME_MACHINE}-ibm-linux\n\texit ;;\n sh64*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n sh*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n sparc:Linux:*:* | sparc64:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n tile*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n vax:Linux:*:*)\n\techo ${UNAME_MACHINE}-dec-linux-gnu\n\texit ;;\n x86_64:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n xtensa*:Linux:*:*)\n\techo ${UNAME_MACHINE}-unknown-linux-gnu\n\texit ;;\n i*86:DYNIX/ptx:4*:*)\n\t# ptx 4.0 does uname -s correctly, with DYNIX/ptx in there.\n\t# earlier versions are messed up and put the nodename in both\n\t# sysname and nodename.\n\techo i386-sequent-sysv4\n\texit ;;\n i*86:UNIX_SV:4.2MP:2.*)\n\t# Unixware is an offshoot of SVR4, but it has its own version\n\t# number series starting with 2...\n\t# I am not positive that other SVR4 systems won't match this,\n\t# I just have to hope. -- rms.\n\t# Use sysv4.2uw... so that sysv4* matches it.\n\techo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION}\n\texit ;;\n i*86:OS/2:*:*)\n\t# If we were able to find `uname', then EMX Unix compatibility\n\t# is probably installed.\n\techo ${UNAME_MACHINE}-pc-os2-emx\n\texit ;;\n i*86:XTS-300:*:STOP)\n\techo ${UNAME_MACHINE}-unknown-stop\n\texit ;;\n i*86:atheos:*:*)\n\techo ${UNAME_MACHINE}-unknown-atheos\n\texit ;;\n i*86:syllable:*:*)\n\techo ${UNAME_MACHINE}-pc-syllable\n\texit ;;\n i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.[02]*:*)\n\techo i386-unknown-lynxos${UNAME_RELEASE}\n\texit ;;\n i*86:*DOS:*:*)\n\techo ${UNAME_MACHINE}-pc-msdosdjgpp\n\texit ;;\n i*86:*:4.*:* | i*86:SYSTEM_V:4.*:*)\n\tUNAME_REL=`echo ${UNAME_RELEASE} | sed 's/\\/MP$//'`\n\tif grep Novell /usr/include/link.h >/dev/null 2>/dev/null; then\n\t\techo ${UNAME_MACHINE}-univel-sysv${UNAME_REL}\n\telse\n\t\techo ${UNAME_MACHINE}-pc-sysv${UNAME_REL}\n\tfi\n\texit ;;\n i*86:*:5:[678]*)\n\t# UnixWare 7.x, OpenUNIX and OpenServer 6.\n\tcase `/bin/uname -X | grep \"^Machine\"` in\n\t *486*)\t UNAME_MACHINE=i486 ;;\n\t *Pentium)\t UNAME_MACHINE=i586 ;;\n\t *Pent*|*Celeron) UNAME_MACHINE=i686 ;;\n\tesac\n\techo ${UNAME_MACHINE}-unknown-sysv${UNAME_RELEASE}${UNAME_SYSTEM}${UNAME_VERSION}\n\texit ;;\n i*86:*:3.2:*)\n\tif test -f /usr/options/cb.name; then\n\t\tUNAME_REL=`sed -n 's/.*Version //p' /dev/null >/dev/null ; then\n\t\tUNAME_REL=`(/bin/uname -X|grep Release|sed -e 's/.*= //')`\n\t\t(/bin/uname -X|grep i80486 >/dev/null) && UNAME_MACHINE=i486\n\t\t(/bin/uname -X|grep '^Machine.*Pentium' >/dev/null) \\\n\t\t\t&& UNAME_MACHINE=i586\n\t\t(/bin/uname -X|grep '^Machine.*Pent *II' >/dev/null) \\\n\t\t\t&& UNAME_MACHINE=i686\n\t\t(/bin/uname -X|grep '^Machine.*Pentium Pro' >/dev/null) \\\n\t\t\t&& UNAME_MACHINE=i686\n\t\techo ${UNAME_MACHINE}-pc-sco$UNAME_REL\n\telse\n\t\techo ${UNAME_MACHINE}-pc-sysv32\n\tfi\n\texit ;;\n pc:*:*:*)\n\t# Left here for compatibility:\n\t# uname -m prints for DJGPP always 'pc', but it prints nothing about\n\t# the processor, so we play safe by assuming i586.\n\t# Note: whatever this is, it MUST be the same as what config.sub\n\t# prints for the \"djgpp\" host, or else GDB configury will decide that\n\t# this is a cross-build.\n\techo i586-pc-msdosdjgpp\n\texit ;;\n Intel:Mach:3*:*)\n\techo i386-pc-mach3\n\texit ;;\n paragon:*:*:*)\n\techo i860-intel-osf1\n\texit ;;\n i860:*:4.*:*) # i860-SVR4\n\tif grep Stardent /usr/include/sys/uadmin.h >/dev/null 2>&1 ; then\n\t echo i860-stardent-sysv${UNAME_RELEASE} # Stardent Vistra i860-SVR4\n\telse # Add other i860-SVR4 vendors below as they are discovered.\n\t echo i860-unknown-sysv${UNAME_RELEASE} # Unknown i860-SVR4\n\tfi\n\texit ;;\n mini*:CTIX:SYS*5:*)\n\t# \"miniframe\"\n\techo m68010-convergent-sysv\n\texit ;;\n mc68k:UNIX:SYSTEM5:3.51m)\n\techo m68k-convergent-sysv\n\texit ;;\n M680?0:D-NIX:5.3:*)\n\techo m68k-diab-dnix\n\texit ;;\n M68*:*:R3V[5678]*:*)\n\ttest -r /sysV68 && { echo 'm68k-motorola-sysv'; exit; } ;;\n 3[345]??:*:4.0:3.0 | 3[34]??A:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 3[34]??/*:*:4.0:3.0 | 4400:*:4.0:3.0 | 4850:*:4.0:3.0 | SKA40:*:4.0:3.0 | SDS2:*:4.0:3.0 | SHG2:*:4.0:3.0 | S7501*:*:4.0:3.0)\n\tOS_REL=''\n\ttest -r /etc/.relid \\\n\t&& OS_REL=.`sed -n 's/[^ ]* [^ ]* \\([0-9][0-9]\\).*/\\1/p' < /etc/.relid`\n\t/bin/uname -p 2>/dev/null | grep 86 >/dev/null \\\n\t && { echo i486-ncr-sysv4.3${OS_REL}; exit; }\n\t/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \\\n\t && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;;\n 3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*)\n\t/bin/uname -p 2>/dev/null | grep 86 >/dev/null \\\n\t && { echo i486-ncr-sysv4; exit; } ;;\n NCR*:*:4.2:* | MPRAS*:*:4.2:*)\n\tOS_REL='.3'\n\ttest -r /etc/.relid \\\n\t && OS_REL=.`sed -n 's/[^ ]* [^ ]* \\([0-9][0-9]\\).*/\\1/p' < /etc/.relid`\n\t/bin/uname -p 2>/dev/null | grep 86 >/dev/null \\\n\t && { echo i486-ncr-sysv4.3${OS_REL}; exit; }\n\t/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \\\n\t && { echo i586-ncr-sysv4.3${OS_REL}; exit; }\n\t/bin/uname -p 2>/dev/null | /bin/grep pteron >/dev/null \\\n\t && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;;\n m68*:LynxOS:2.*:* | m68*:LynxOS:3.0*:*)\n\techo m68k-unknown-lynxos${UNAME_RELEASE}\n\texit ;;\n mc68030:UNIX_System_V:4.*:*)\n\techo m68k-atari-sysv4\n\texit ;;\n TSUNAMI:LynxOS:2.*:*)\n\techo sparc-unknown-lynxos${UNAME_RELEASE}\n\texit ;;\n rs6000:LynxOS:2.*:*)\n\techo rs6000-unknown-lynxos${UNAME_RELEASE}\n\texit ;;\n PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.[02]*:*)\n\techo powerpc-unknown-lynxos${UNAME_RELEASE}\n\texit ;;\n SM[BE]S:UNIX_SV:*:*)\n\techo mips-dde-sysv${UNAME_RELEASE}\n\texit ;;\n RM*:ReliantUNIX-*:*:*)\n\techo mips-sni-sysv4\n\texit ;;\n RM*:SINIX-*:*:*)\n\techo mips-sni-sysv4\n\texit ;;\n *:SINIX-*:*:*)\n\tif uname -p 2>/dev/null >/dev/null ; then\n\t\tUNAME_MACHINE=`(uname -p) 2>/dev/null`\n\t\techo ${UNAME_MACHINE}-sni-sysv4\n\telse\n\t\techo ns32k-sni-sysv\n\tfi\n\texit ;;\n PENTIUM:*:4.0*:*)\t# Unisys `ClearPath HMP IX 4000' SVR4/MP effort\n\t\t\t# says \n\techo i586-unisys-sysv4\n\texit ;;\n *:UNIX_System_V:4*:FTX*)\n\t# From Gerald Hewes .\n\t# How about differentiating between stratus architectures? -djm\n\techo hppa1.1-stratus-sysv4\n\texit ;;\n *:*:*:FTX*)\n\t# From seanf@swdc.stratus.com.\n\techo i860-stratus-sysv4\n\texit ;;\n i*86:VOS:*:*)\n\t# From Paul.Green@stratus.com.\n\techo ${UNAME_MACHINE}-stratus-vos\n\texit ;;\n *:VOS:*:*)\n\t# From Paul.Green@stratus.com.\n\techo hppa1.1-stratus-vos\n\texit ;;\n mc68*:A/UX:*:*)\n\techo m68k-apple-aux${UNAME_RELEASE}\n\texit ;;\n news*:NEWS-OS:6*:*)\n\techo mips-sony-newsos6\n\texit ;;\n R[34]000:*System_V*:*:* | R4000:UNIX_SYSV:*:* | R*000:UNIX_SV:*:*)\n\tif [ -d /usr/nec ]; then\n\t\techo mips-nec-sysv${UNAME_RELEASE}\n\telse\n\t\techo mips-unknown-sysv${UNAME_RELEASE}\n\tfi\n\texit ;;\n BeBox:BeOS:*:*)\t# BeOS running on hardware made by Be, PPC only.\n\techo powerpc-be-beos\n\texit ;;\n BeMac:BeOS:*:*)\t# BeOS running on Mac or Mac clone, PPC only.\n\techo powerpc-apple-beos\n\texit ;;\n BePC:BeOS:*:*)\t# BeOS running on Intel PC compatible.\n\techo i586-pc-beos\n\texit ;;\n BePC:Haiku:*:*)\t# Haiku running on Intel PC compatible.\n\techo i586-pc-haiku\n\texit ;;\n SX-4:SUPER-UX:*:*)\n\techo sx4-nec-superux${UNAME_RELEASE}\n\texit ;;\n SX-5:SUPER-UX:*:*)\n\techo sx5-nec-superux${UNAME_RELEASE}\n\texit ;;\n SX-6:SUPER-UX:*:*)\n\techo sx6-nec-superux${UNAME_RELEASE}\n\texit ;;\n SX-7:SUPER-UX:*:*)\n\techo sx7-nec-superux${UNAME_RELEASE}\n\texit ;;\n SX-8:SUPER-UX:*:*)\n\techo sx8-nec-superux${UNAME_RELEASE}\n\texit ;;\n SX-8R:SUPER-UX:*:*)\n\techo sx8r-nec-superux${UNAME_RELEASE}\n\texit ;;\n Power*:Rhapsody:*:*)\n\techo powerpc-apple-rhapsody${UNAME_RELEASE}\n\texit ;;\n *:Rhapsody:*:*)\n\techo ${UNAME_MACHINE}-apple-rhapsody${UNAME_RELEASE}\n\texit ;;\n *:Darwin:*:*)\n\tUNAME_PROCESSOR=`uname -p` || UNAME_PROCESSOR=unknown\n\tcase $UNAME_PROCESSOR in\n\t i386)\n\t\teval $set_cc_for_build\n\t\tif [ \"$CC_FOR_BUILD\" != 'no_compiler_found' ]; then\n\t\t if (echo '#ifdef __LP64__'; echo IS_64BIT_ARCH; echo '#endif') | \\\n\t\t (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \\\n\t\t grep IS_64BIT_ARCH >/dev/null\n\t\t then\n\t\t UNAME_PROCESSOR=\"x86_64\"\n\t\t fi\n\t\tfi ;;\n\t unknown) UNAME_PROCESSOR=powerpc ;;\n\tesac\n\techo ${UNAME_PROCESSOR}-apple-darwin${UNAME_RELEASE}\n\texit ;;\n *:procnto*:*:* | *:QNX:[0123456789]*:*)\n\tUNAME_PROCESSOR=`uname -p`\n\tif test \"$UNAME_PROCESSOR\" = \"x86\"; then\n\t\tUNAME_PROCESSOR=i386\n\t\tUNAME_MACHINE=pc\n\tfi\n\techo ${UNAME_PROCESSOR}-${UNAME_MACHINE}-nto-qnx${UNAME_RELEASE}\n\texit ;;\n *:QNX:*:4*)\n\techo i386-pc-qnx\n\texit ;;\n NEO-?:NONSTOP_KERNEL:*:*)\n\techo neo-tandem-nsk${UNAME_RELEASE}\n\texit ;;\n NSE-?:NONSTOP_KERNEL:*:*)\n\techo nse-tandem-nsk${UNAME_RELEASE}\n\texit ;;\n NSR-?:NONSTOP_KERNEL:*:*)\n\techo nsr-tandem-nsk${UNAME_RELEASE}\n\texit ;;\n *:NonStop-UX:*:*)\n\techo mips-compaq-nonstopux\n\texit ;;\n BS2000:POSIX*:*:*)\n\techo bs2000-siemens-sysv\n\texit ;;\n DS/*:UNIX_System_V:*:*)\n\techo ${UNAME_MACHINE}-${UNAME_SYSTEM}-${UNAME_RELEASE}\n\texit ;;\n *:Plan9:*:*)\n\t# \"uname -m\" is not consistent, so use $cputype instead. 386\n\t# is converted to i386 for consistency with other x86\n\t# operating systems.\n\tif test \"$cputype\" = \"386\"; then\n\t UNAME_MACHINE=i386\n\telse\n\t UNAME_MACHINE=\"$cputype\"\n\tfi\n\techo ${UNAME_MACHINE}-unknown-plan9\n\texit ;;\n *:TOPS-10:*:*)\n\techo pdp10-unknown-tops10\n\texit ;;\n *:TENEX:*:*)\n\techo pdp10-unknown-tenex\n\texit ;;\n KS10:TOPS-20:*:* | KL10:TOPS-20:*:* | TYPE4:TOPS-20:*:*)\n\techo pdp10-dec-tops20\n\texit ;;\n XKL-1:TOPS-20:*:* | TYPE5:TOPS-20:*:*)\n\techo pdp10-xkl-tops20\n\texit ;;\n *:TOPS-20:*:*)\n\techo pdp10-unknown-tops20\n\texit ;;\n *:ITS:*:*)\n\techo pdp10-unknown-its\n\texit ;;\n SEI:*:*:SEIUX)\n\techo mips-sei-seiux${UNAME_RELEASE}\n\texit ;;\n *:DragonFly:*:*)\n\techo ${UNAME_MACHINE}-unknown-dragonfly`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`\n\texit ;;\n *:*VMS:*:*)\n\tUNAME_MACHINE=`(uname -p) 2>/dev/null`\n\tcase \"${UNAME_MACHINE}\" in\n\t A*) echo alpha-dec-vms ; exit ;;\n\t I*) echo ia64-dec-vms ; exit ;;\n\t V*) echo vax-dec-vms ; exit ;;\n\tesac ;;\n *:XENIX:*:SysV)\n\techo i386-pc-xenix\n\texit ;;\n i*86:skyos:*:*)\n\techo ${UNAME_MACHINE}-pc-skyos`echo ${UNAME_RELEASE}` | sed -e 's/ .*$//'\n\texit ;;\n i*86:rdos:*:*)\n\techo ${UNAME_MACHINE}-pc-rdos\n\texit ;;\n i*86:AROS:*:*)\n\techo ${UNAME_MACHINE}-pc-aros\n\texit ;;\n x86_64:VMkernel:*:*)\n\techo ${UNAME_MACHINE}-unknown-esx\n\texit ;;\nesac\n\n#echo '(No uname command or uname output not recognized.)' 1>&2\n#echo \"${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}\" 1>&2\n\neval $set_cc_for_build\ncat >$dummy.c <\n# include \n#endif\nmain ()\n{\n#if defined (sony)\n#if defined (MIPSEB)\n /* BFD wants \"bsd\" instead of \"newsos\". Perhaps BFD should be changed,\n I don't know.... */\n printf (\"mips-sony-bsd\\n\"); exit (0);\n#else\n#include \n printf (\"m68k-sony-newsos%s\\n\",\n#ifdef NEWSOS4\n\t\"4\"\n#else\n\t\"\"\n#endif\n\t); exit (0);\n#endif\n#endif\n\n#if defined (__arm) && defined (__acorn) && defined (__unix)\n printf (\"arm-acorn-riscix\\n\"); exit (0);\n#endif\n\n#if defined (hp300) && !defined (hpux)\n printf (\"m68k-hp-bsd\\n\"); exit (0);\n#endif\n\n#if defined (NeXT)\n#if !defined (__ARCHITECTURE__)\n#define __ARCHITECTURE__ \"m68k\"\n#endif\n int version;\n version=`(hostinfo | sed -n 's/.*NeXT Mach \\([0-9]*\\).*/\\1/p') 2>/dev/null`;\n if (version < 4)\n printf (\"%s-next-nextstep%d\\n\", __ARCHITECTURE__, version);\n else\n printf (\"%s-next-openstep%d\\n\", __ARCHITECTURE__, version);\n exit (0);\n#endif\n\n#if defined (MULTIMAX) || defined (n16)\n#if defined (UMAXV)\n printf (\"ns32k-encore-sysv\\n\"); exit (0);\n#else\n#if defined (CMU)\n printf (\"ns32k-encore-mach\\n\"); exit (0);\n#else\n printf (\"ns32k-encore-bsd\\n\"); exit (0);\n#endif\n#endif\n#endif\n\n#if defined (__386BSD__)\n printf (\"i386-pc-bsd\\n\"); exit (0);\n#endif\n\n#if defined (sequent)\n#if defined (i386)\n printf (\"i386-sequent-dynix\\n\"); exit (0);\n#endif\n#if defined (ns32000)\n printf (\"ns32k-sequent-dynix\\n\"); exit (0);\n#endif\n#endif\n\n#if defined (_SEQUENT_)\n struct utsname un;\n\n uname(&un);\n\n if (strncmp(un.version, \"V2\", 2) == 0) {\n\tprintf (\"i386-sequent-ptx2\\n\"); exit (0);\n }\n if (strncmp(un.version, \"V1\", 2) == 0) { /* XXX is V1 correct? */\n\tprintf (\"i386-sequent-ptx1\\n\"); exit (0);\n }\n printf (\"i386-sequent-ptx\\n\"); exit (0);\n\n#endif\n\n#if defined (vax)\n# if !defined (ultrix)\n# include \n# if defined (BSD)\n# if BSD == 43\n printf (\"vax-dec-bsd4.3\\n\"); exit (0);\n# else\n# if BSD == 199006\n printf (\"vax-dec-bsd4.3reno\\n\"); exit (0);\n# else\n printf (\"vax-dec-bsd\\n\"); exit (0);\n# endif\n# endif\n# else\n printf (\"vax-dec-bsd\\n\"); exit (0);\n# endif\n# else\n printf (\"vax-dec-ultrix\\n\"); exit (0);\n# endif\n#endif\n\n#if defined (alliant) && defined (i860)\n printf (\"i860-alliant-bsd\\n\"); exit (0);\n#endif\n\n exit (1);\n}\nEOF\n\n$CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null && SYSTEM_NAME=`$dummy` &&\n\t{ echo \"$SYSTEM_NAME\"; exit; }\n\n# Apollos put the system type in the environment.\n\ntest -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit; }\n\n# Convex versions that predate uname can use getsysinfo(1)\n\nif [ -x /usr/convex/getsysinfo ]\nthen\n case `getsysinfo -f cpu_type` in\n c1*)\n\techo c1-convex-bsd\n\texit ;;\n c2*)\n\tif getsysinfo -f scalar_acc\n\tthen echo c32-convex-bsd\n\telse echo c2-convex-bsd\n\tfi\n\texit ;;\n c34*)\n\techo c34-convex-bsd\n\texit ;;\n c38*)\n\techo c38-convex-bsd\n\texit ;;\n c4*)\n\techo c4-convex-bsd\n\texit ;;\n esac\nfi\n\ncat >&2 < in order to provide the needed\ninformation to handle your system.\n\nconfig.guess timestamp = $timestamp\n\nuname -m = `(uname -m) 2>/dev/null || echo unknown`\nuname -r = `(uname -r) 2>/dev/null || echo unknown`\nuname -s = `(uname -s) 2>/dev/null || echo unknown`\nuname -v = `(uname -v) 2>/dev/null || echo unknown`\n\n/usr/bin/uname -p = `(/usr/bin/uname -p) 2>/dev/null`\n/bin/uname -X = `(/bin/uname -X) 2>/dev/null`\n\nhostinfo = `(hostinfo) 2>/dev/null`\n/bin/universe = `(/bin/universe) 2>/dev/null`\n/usr/bin/arch -k = `(/usr/bin/arch -k) 2>/dev/null`\n/bin/arch = `(/bin/arch) 2>/dev/null`\n/usr/bin/oslevel = `(/usr/bin/oslevel) 2>/dev/null`\n/usr/convex/getsysinfo = `(/usr/convex/getsysinfo) 2>/dev/null`\n\nUNAME_MACHINE = ${UNAME_MACHINE}\nUNAME_RELEASE = ${UNAME_RELEASE}\nUNAME_SYSTEM = ${UNAME_SYSTEM}\nUNAME_VERSION = ${UNAME_VERSION}\nEOF\n\nexit 1\n\n# Local variables:\n# eval: (add-hook 'write-file-hooks 'time-stamp)\n# time-stamp-start: \"timestamp='\"\n# time-stamp-format: \"%:y-%02m-%02d\"\n# time-stamp-end: \"'\"\n# End:\n" }, { "path": "deps/jemalloc/config.stamp.in", "content": "" }, { "path": "deps/jemalloc/config.sub", "content": "#! /bin/sh\n# Configuration validation subroutine script.\n# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,\n# 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010,\n# 2011, 2012 Free Software Foundation, Inc.\n\ntimestamp='2012-02-10'\n\n# This file is (in principle) common to ALL GNU software.\n# The presence of a machine in this file suggests that SOME GNU software\n# can handle that machine. It does not imply ALL GNU software can.\n#\n# This file is free software; you can redistribute it and/or modify\n# it under the terms of the GNU General Public License as published by\n# the Free Software Foundation; either version 2 of the License, or\n# (at your option) any later version.\n#\n# This program is distributed in the hope that it will be useful,\n# but WITHOUT ANY WARRANTY; without even the implied warranty of\n# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\n# GNU General Public License for more details.\n#\n# You should have received a copy of the GNU General Public License\n# along with this program; if not, see .\n#\n# As a special exception to the GNU General Public License, if you\n# distribute this file as part of a program that contains a\n# configuration script generated by Autoconf, you may include it under\n# the same distribution terms that you use for the rest of that program.\n\n\n# Please send patches to . Submit a context\n# diff and a properly formatted GNU ChangeLog entry.\n#\n# Configuration subroutine to validate and canonicalize a configuration type.\n# Supply the specified configuration type as an argument.\n# If it is invalid, we print an error message on stderr and exit with code 1.\n# Otherwise, we print the canonical config type on stdout and succeed.\n\n# You can get the latest version of this script from:\n# http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD\n\n# This file is supposed to be the same for all GNU packages\n# and recognize all the CPU types, system types and aliases\n# that are meaningful with *any* GNU software.\n# Each package is responsible for reporting which valid configurations\n# it does not support. The user should be able to distinguish\n# a failure to support a valid configuration from a meaningless\n# configuration.\n\n# The goal of this file is to map all the various variations of a given\n# machine specification into a single specification in the form:\n#\tCPU_TYPE-MANUFACTURER-OPERATING_SYSTEM\n# or in some cases, the newer four-part form:\n#\tCPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM\n# It is wrong to echo any other type of specification.\n\nme=`echo \"$0\" | sed -e 's,.*/,,'`\n\nusage=\"\\\nUsage: $0 [OPTION] CPU-MFR-OPSYS\n $0 [OPTION] ALIAS\n\nCanonicalize a configuration name.\n\nOperation modes:\n -h, --help print this help, then exit\n -t, --time-stamp print date of last modification, then exit\n -v, --version print version number, then exit\n\nReport bugs and patches to .\"\n\nversion=\"\\\nGNU config.sub ($timestamp)\n\nCopyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,\n2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012\nFree Software Foundation, Inc.\n\nThis is free software; see the source for copying conditions. There is NO\nwarranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\"\n\nhelp=\"\nTry \\`$me --help' for more information.\"\n\n# Parse command line\nwhile test $# -gt 0 ; do\n case $1 in\n --time-stamp | --time* | -t )\n echo \"$timestamp\" ; exit ;;\n --version | -v )\n echo \"$version\" ; exit ;;\n --help | --h* | -h )\n echo \"$usage\"; exit ;;\n -- ) # Stop option processing\n shift; break ;;\n - )\t# Use stdin as input.\n break ;;\n -* )\n echo \"$me: invalid option $1$help\"\n exit 1 ;;\n\n *local*)\n # First pass through any local machine types.\n echo $1\n exit ;;\n\n * )\n break ;;\n esac\ndone\n\ncase $# in\n 0) echo \"$me: missing argument$help\" >&2\n exit 1;;\n 1) ;;\n *) echo \"$me: too many arguments$help\" >&2\n exit 1;;\nesac\n\n# Separate what the user gave into CPU-COMPANY and OS or KERNEL-OS (if any).\n# Here we must recognize all the valid KERNEL-OS combinations.\nmaybe_os=`echo $1 | sed 's/^\\(.*\\)-\\([^-]*-[^-]*\\)$/\\2/'`\ncase $maybe_os in\n nto-qnx* | linux-gnu* | linux-android* | linux-dietlibc | linux-newlib* | \\\n linux-uclibc* | uclinux-uclibc* | uclinux-gnu* | kfreebsd*-gnu* | \\\n knetbsd*-gnu* | netbsd*-gnu* | \\\n kopensolaris*-gnu* | \\\n storm-chaos* | os2-emx* | rtmk-nova*)\n os=-$maybe_os\n basic_machine=`echo $1 | sed 's/^\\(.*\\)-\\([^-]*-[^-]*\\)$/\\1/'`\n ;;\n android-linux)\n os=-linux-android\n basic_machine=`echo $1 | sed 's/^\\(.*\\)-\\([^-]*-[^-]*\\)$/\\1/'`-unknown\n ;;\n *)\n basic_machine=`echo $1 | sed 's/-[^-]*$//'`\n if [ $basic_machine != $1 ]\n then os=`echo $1 | sed 's/.*-/-/'`\n else os=; fi\n ;;\nesac\n\n### Let's recognize common machines as not being operating systems so\n### that things like config.sub decstation-3100 work. We also\n### recognize some manufacturers as not being operating systems, so we\n### can provide default operating systems below.\ncase $os in\n\t-sun*os*)\n\t\t# Prevent following clause from handling this invalid input.\n\t\t;;\n\t-dec* | -mips* | -sequent* | -encore* | -pc532* | -sgi* | -sony* | \\\n\t-att* | -7300* | -3300* | -delta* | -motorola* | -sun[234]* | \\\n\t-unicom* | -ibm* | -next | -hp | -isi* | -apollo | -altos* | \\\n\t-convergent* | -ncr* | -news | -32* | -3600* | -3100* | -hitachi* |\\\n\t-c[123]* | -convex* | -sun | -crds | -omron* | -dg | -ultra | -tti* | \\\n\t-harris | -dolphin | -highlevel | -gould | -cbm | -ns | -masscomp | \\\n\t-apple | -axis | -knuth | -cray | -microblaze)\n\t\tos=\n\t\tbasic_machine=$1\n\t\t;;\n\t-bluegene*)\n\t\tos=-cnk\n\t\t;;\n\t-sim | -cisco | -oki | -wec | -winbond)\n\t\tos=\n\t\tbasic_machine=$1\n\t\t;;\n\t-scout)\n\t\t;;\n\t-wrs)\n\t\tos=-vxworks\n\t\tbasic_machine=$1\n\t\t;;\n\t-chorusos*)\n\t\tos=-chorusos\n\t\tbasic_machine=$1\n\t\t;;\n\t-chorusrdb)\n\t\tos=-chorusrdb\n\t\tbasic_machine=$1\n\t\t;;\n\t-hiux*)\n\t\tos=-hiuxwe2\n\t\t;;\n\t-sco6)\n\t\tos=-sco5v6\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco5)\n\t\tos=-sco3.2v5\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco4)\n\t\tos=-sco3.2v4\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco3.2.[4-9]*)\n\t\tos=`echo $os | sed -e 's/sco3.2./sco3.2v/'`\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco3.2v[4-9]*)\n\t\t# Don't forget version if it is 3.2v4 or newer.\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco5v6*)\n\t\t# Don't forget version if it is 3.2v4 or newer.\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-sco*)\n\t\tos=-sco3.2v2\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-udk*)\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-isc)\n\t\tos=-isc2.2\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-clix*)\n\t\tbasic_machine=clipper-intergraph\n\t\t;;\n\t-isc*)\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`\n\t\t;;\n\t-lynx*)\n\t\tos=-lynxos\n\t\t;;\n\t-ptx*)\n\t\tbasic_machine=`echo $1 | sed -e 's/86-.*/86-sequent/'`\n\t\t;;\n\t-windowsnt*)\n\t\tos=`echo $os | sed -e 's/windowsnt/winnt/'`\n\t\t;;\n\t-psos*)\n\t\tos=-psos\n\t\t;;\n\t-mint | -mint[0-9]*)\n\t\tbasic_machine=m68k-atari\n\t\tos=-mint\n\t\t;;\nesac\n\n# Decode aliases for certain CPU-COMPANY combinations.\ncase $basic_machine in\n\t# Recognize the basic CPU types without company name.\n\t# Some are omitted here because they have special meanings below.\n\t1750a | 580 \\\n\t| a29k \\\n\t| aarch64 | aarch64_be \\\n\t| alpha | alphaev[4-8] | alphaev56 | alphaev6[78] | alphapca5[67] \\\n\t| alpha64 | alpha64ev[4-8] | alpha64ev56 | alpha64ev6[78] | alpha64pca5[67] \\\n\t| am33_2.0 \\\n\t| arc | arm | arm[bl]e | arme[lb] | armv[2345] | armv[345][lb] | avr | avr32 \\\n | be32 | be64 \\\n\t| bfin \\\n\t| c4x | clipper \\\n\t| d10v | d30v | dlx | dsp16xx \\\n\t| epiphany \\\n\t| fido | fr30 | frv \\\n\t| h8300 | h8500 | hppa | hppa1.[01] | hppa2.0 | hppa2.0[nw] | hppa64 \\\n\t| hexagon \\\n\t| i370 | i860 | i960 | ia64 \\\n\t| ip2k | iq2000 \\\n\t| le32 | le64 \\\n\t| lm32 \\\n\t| m32c | m32r | m32rle | m68000 | m68k | m88k \\\n\t| maxq | mb | microblaze | mcore | mep | metag \\\n\t| mips | mipsbe | mipseb | mipsel | mipsle \\\n\t| mips16 \\\n\t| mips64 | mips64el \\\n\t| mips64octeon | mips64octeonel \\\n\t| mips64orion | mips64orionel \\\n\t| mips64r5900 | mips64r5900el \\\n\t| mips64vr | mips64vrel \\\n\t| mips64vr4100 | mips64vr4100el \\\n\t| mips64vr4300 | mips64vr4300el \\\n\t| mips64vr5000 | mips64vr5000el \\\n\t| mips64vr5900 | mips64vr5900el \\\n\t| mipsisa32 | mipsisa32el \\\n\t| mipsisa32r2 | mipsisa32r2el \\\n\t| mipsisa64 | mipsisa64el \\\n\t| mipsisa64r2 | mipsisa64r2el \\\n\t| mipsisa64sb1 | mipsisa64sb1el \\\n\t| mipsisa64sr71k | mipsisa64sr71kel \\\n\t| mipstx39 | mipstx39el \\\n\t| mn10200 | mn10300 \\\n\t| moxie \\\n\t| mt \\\n\t| msp430 \\\n\t| nds32 | nds32le | nds32be \\\n\t| nios | nios2 \\\n\t| ns16k | ns32k \\\n\t| open8 \\\n\t| or32 \\\n\t| pdp10 | pdp11 | pj | pjl \\\n\t| powerpc | powerpc64 | powerpc64le | powerpcle \\\n\t| pyramid \\\n\t| rl78 | rx \\\n\t| score \\\n\t| sh | sh[1234] | sh[24]a | sh[24]aeb | sh[23]e | sh[34]eb | sheb | shbe | shle | sh[1234]le | sh3ele \\\n\t| sh64 | sh64le \\\n\t| sparc | sparc64 | sparc64b | sparc64v | sparc86x | sparclet | sparclite \\\n\t| sparcv8 | sparcv9 | sparcv9b | sparcv9v \\\n\t| spu \\\n\t| tahoe | tic4x | tic54x | tic55x | tic6x | tic80 | tron \\\n\t| ubicom32 \\\n\t| v850 | v850e | v850e1 | v850e2 | v850es | v850e2v3 \\\n\t| we32k \\\n\t| x86 | xc16x | xstormy16 | xtensa \\\n\t| z8k | z80)\n\t\tbasic_machine=$basic_machine-unknown\n\t\t;;\n\tc54x)\n\t\tbasic_machine=tic54x-unknown\n\t\t;;\n\tc55x)\n\t\tbasic_machine=tic55x-unknown\n\t\t;;\n\tc6x)\n\t\tbasic_machine=tic6x-unknown\n\t\t;;\n\tm6811 | m68hc11 | m6812 | m68hc12 | m68hcs12x | picochip)\n\t\tbasic_machine=$basic_machine-unknown\n\t\tos=-none\n\t\t;;\n\tm88110 | m680[12346]0 | m683?2 | m68360 | m5200 | v70 | w65 | z8k)\n\t\t;;\n\tms1)\n\t\tbasic_machine=mt-unknown\n\t\t;;\n\n\tstrongarm | thumb | xscale)\n\t\tbasic_machine=arm-unknown\n\t\t;;\n\txgate)\n\t\tbasic_machine=$basic_machine-unknown\n\t\tos=-none\n\t\t;;\n\txscaleeb)\n\t\tbasic_machine=armeb-unknown\n\t\t;;\n\n\txscaleel)\n\t\tbasic_machine=armel-unknown\n\t\t;;\n\n\t# We use `pc' rather than `unknown'\n\t# because (1) that's what they normally are, and\n\t# (2) the word \"unknown\" tends to confuse beginning users.\n\ti*86 | x86_64)\n\t basic_machine=$basic_machine-pc\n\t ;;\n\t# Object if more than one company name word.\n\t*-*-*)\n\t\techo Invalid configuration \\`$1\\': machine \\`$basic_machine\\' not recognized 1>&2\n\t\texit 1\n\t\t;;\n\t# Recognize the basic CPU types with company name.\n\t580-* \\\n\t| a29k-* \\\n\t| aarch64-* | aarch64_be-* \\\n\t| alpha-* | alphaev[4-8]-* | alphaev56-* | alphaev6[78]-* \\\n\t| alpha64-* | alpha64ev[4-8]-* | alpha64ev56-* | alpha64ev6[78]-* \\\n\t| alphapca5[67]-* | alpha64pca5[67]-* | arc-* \\\n\t| arm-* | armbe-* | armle-* | armeb-* | armv*-* \\\n\t| avr-* | avr32-* \\\n\t| be32-* | be64-* \\\n\t| bfin-* | bs2000-* \\\n\t| c[123]* | c30-* | [cjt]90-* | c4x-* \\\n\t| clipper-* | craynv-* | cydra-* \\\n\t| d10v-* | d30v-* | dlx-* \\\n\t| elxsi-* \\\n\t| f30[01]-* | f700-* | fido-* | fr30-* | frv-* | fx80-* \\\n\t| h8300-* | h8500-* \\\n\t| hppa-* | hppa1.[01]-* | hppa2.0-* | hppa2.0[nw]-* | hppa64-* \\\n\t| hexagon-* \\\n\t| i*86-* | i860-* | i960-* | ia64-* \\\n\t| ip2k-* | iq2000-* \\\n\t| le32-* | le64-* \\\n\t| lm32-* \\\n\t| m32c-* | m32r-* | m32rle-* \\\n\t| m68000-* | m680[012346]0-* | m68360-* | m683?2-* | m68k-* \\\n\t| m88110-* | m88k-* | maxq-* | mcore-* | metag-* | microblaze-* \\\n\t| mips-* | mipsbe-* | mipseb-* | mipsel-* | mipsle-* \\\n\t| mips16-* \\\n\t| mips64-* | mips64el-* \\\n\t| mips64octeon-* | mips64octeonel-* \\\n\t| mips64orion-* | mips64orionel-* \\\n\t| mips64r5900-* | mips64r5900el-* \\\n\t| mips64vr-* | mips64vrel-* \\\n\t| mips64vr4100-* | mips64vr4100el-* \\\n\t| mips64vr4300-* | mips64vr4300el-* \\\n\t| mips64vr5000-* | mips64vr5000el-* \\\n\t| mips64vr5900-* | mips64vr5900el-* \\\n\t| mipsisa32-* | mipsisa32el-* \\\n\t| mipsisa32r2-* | mipsisa32r2el-* \\\n\t| mipsisa64-* | mipsisa64el-* \\\n\t| mipsisa64r2-* | mipsisa64r2el-* \\\n\t| mipsisa64sb1-* | mipsisa64sb1el-* \\\n\t| mipsisa64sr71k-* | mipsisa64sr71kel-* \\\n\t| mipstx39-* | mipstx39el-* \\\n\t| mmix-* \\\n\t| mt-* \\\n\t| msp430-* \\\n\t| nds32-* | nds32le-* | nds32be-* \\\n\t| nios-* | nios2-* \\\n\t| none-* | np1-* | ns16k-* | ns32k-* \\\n\t| open8-* \\\n\t| orion-* \\\n\t| pdp10-* | pdp11-* | pj-* | pjl-* | pn-* | power-* \\\n\t| powerpc-* | powerpc64-* | powerpc64le-* | powerpcle-* \\\n\t| pyramid-* \\\n\t| rl78-* | romp-* | rs6000-* | rx-* \\\n\t| sh-* | sh[1234]-* | sh[24]a-* | sh[24]aeb-* | sh[23]e-* | sh[34]eb-* | sheb-* | shbe-* \\\n\t| shle-* | sh[1234]le-* | sh3ele-* | sh64-* | sh64le-* \\\n\t| sparc-* | sparc64-* | sparc64b-* | sparc64v-* | sparc86x-* | sparclet-* \\\n\t| sparclite-* \\\n\t| sparcv8-* | sparcv9-* | sparcv9b-* | sparcv9v-* | sv1-* | sx?-* \\\n\t| tahoe-* \\\n\t| tic30-* | tic4x-* | tic54x-* | tic55x-* | tic6x-* | tic80-* \\\n\t| tile*-* \\\n\t| tron-* \\\n\t| ubicom32-* \\\n\t| v850-* | v850e-* | v850e1-* | v850es-* | v850e2-* | v850e2v3-* \\\n\t| vax-* \\\n\t| we32k-* \\\n\t| x86-* | x86_64-* | xc16x-* | xps100-* \\\n\t| xstormy16-* | xtensa*-* \\\n\t| ymp-* \\\n\t| z8k-* | z80-*)\n\t\t;;\n\t# Recognize the basic CPU types without company name, with glob match.\n\txtensa*)\n\t\tbasic_machine=$basic_machine-unknown\n\t\t;;\n\t# Recognize the various machine names and aliases which stand\n\t# for a CPU type and a company and sometimes even an OS.\n\t386bsd)\n\t\tbasic_machine=i386-unknown\n\t\tos=-bsd\n\t\t;;\n\t3b1 | 7300 | 7300-att | att-7300 | pc7300 | safari | unixpc)\n\t\tbasic_machine=m68000-att\n\t\t;;\n\t3b*)\n\t\tbasic_machine=we32k-att\n\t\t;;\n\ta29khif)\n\t\tbasic_machine=a29k-amd\n\t\tos=-udi\n\t\t;;\n\tabacus)\n\t\tbasic_machine=abacus-unknown\n\t\t;;\n\tadobe68k)\n\t\tbasic_machine=m68010-adobe\n\t\tos=-scout\n\t\t;;\n\talliant | fx80)\n\t\tbasic_machine=fx80-alliant\n\t\t;;\n\taltos | altos3068)\n\t\tbasic_machine=m68k-altos\n\t\t;;\n\tam29k)\n\t\tbasic_machine=a29k-none\n\t\tos=-bsd\n\t\t;;\n\tamd64)\n\t\tbasic_machine=x86_64-pc\n\t\t;;\n\tamd64-*)\n\t\tbasic_machine=x86_64-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tamdahl)\n\t\tbasic_machine=580-amdahl\n\t\tos=-sysv\n\t\t;;\n\tamiga | amiga-*)\n\t\tbasic_machine=m68k-unknown\n\t\t;;\n\tamigaos | amigados)\n\t\tbasic_machine=m68k-unknown\n\t\tos=-amigaos\n\t\t;;\n\tamigaunix | amix)\n\t\tbasic_machine=m68k-unknown\n\t\tos=-sysv4\n\t\t;;\n\tapollo68)\n\t\tbasic_machine=m68k-apollo\n\t\tos=-sysv\n\t\t;;\n\tapollo68bsd)\n\t\tbasic_machine=m68k-apollo\n\t\tos=-bsd\n\t\t;;\n\taros)\n\t\tbasic_machine=i386-pc\n\t\tos=-aros\n\t\t;;\n\taux)\n\t\tbasic_machine=m68k-apple\n\t\tos=-aux\n\t\t;;\n\tbalance)\n\t\tbasic_machine=ns32k-sequent\n\t\tos=-dynix\n\t\t;;\n\tblackfin)\n\t\tbasic_machine=bfin-unknown\n\t\tos=-linux\n\t\t;;\n\tblackfin-*)\n\t\tbasic_machine=bfin-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\tos=-linux\n\t\t;;\n\tbluegene*)\n\t\tbasic_machine=powerpc-ibm\n\t\tos=-cnk\n\t\t;;\n\tc54x-*)\n\t\tbasic_machine=tic54x-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tc55x-*)\n\t\tbasic_machine=tic55x-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tc6x-*)\n\t\tbasic_machine=tic6x-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tc90)\n\t\tbasic_machine=c90-cray\n\t\tos=-unicos\n\t\t;;\n\tcegcc)\n\t\tbasic_machine=arm-unknown\n\t\tos=-cegcc\n\t\t;;\n\tconvex-c1)\n\t\tbasic_machine=c1-convex\n\t\tos=-bsd\n\t\t;;\n\tconvex-c2)\n\t\tbasic_machine=c2-convex\n\t\tos=-bsd\n\t\t;;\n\tconvex-c32)\n\t\tbasic_machine=c32-convex\n\t\tos=-bsd\n\t\t;;\n\tconvex-c34)\n\t\tbasic_machine=c34-convex\n\t\tos=-bsd\n\t\t;;\n\tconvex-c38)\n\t\tbasic_machine=c38-convex\n\t\tos=-bsd\n\t\t;;\n\tcray | j90)\n\t\tbasic_machine=j90-cray\n\t\tos=-unicos\n\t\t;;\n\tcraynv)\n\t\tbasic_machine=craynv-cray\n\t\tos=-unicosmp\n\t\t;;\n\tcr16 | cr16-*)\n\t\tbasic_machine=cr16-unknown\n\t\tos=-elf\n\t\t;;\n\tcrds | unos)\n\t\tbasic_machine=m68k-crds\n\t\t;;\n\tcrisv32 | crisv32-* | etraxfs*)\n\t\tbasic_machine=crisv32-axis\n\t\t;;\n\tcris | cris-* | etrax*)\n\t\tbasic_machine=cris-axis\n\t\t;;\n\tcrx)\n\t\tbasic_machine=crx-unknown\n\t\tos=-elf\n\t\t;;\n\tda30 | da30-*)\n\t\tbasic_machine=m68k-da30\n\t\t;;\n\tdecstation | decstation-3100 | pmax | pmax-* | pmin | dec3100 | decstatn)\n\t\tbasic_machine=mips-dec\n\t\t;;\n\tdecsystem10* | dec10*)\n\t\tbasic_machine=pdp10-dec\n\t\tos=-tops10\n\t\t;;\n\tdecsystem20* | dec20*)\n\t\tbasic_machine=pdp10-dec\n\t\tos=-tops20\n\t\t;;\n\tdelta | 3300 | motorola-3300 | motorola-delta \\\n\t | 3300-motorola | delta-motorola)\n\t\tbasic_machine=m68k-motorola\n\t\t;;\n\tdelta88)\n\t\tbasic_machine=m88k-motorola\n\t\tos=-sysv3\n\t\t;;\n\tdicos)\n\t\tbasic_machine=i686-pc\n\t\tos=-dicos\n\t\t;;\n\tdjgpp)\n\t\tbasic_machine=i586-pc\n\t\tos=-msdosdjgpp\n\t\t;;\n\tdpx20 | dpx20-*)\n\t\tbasic_machine=rs6000-bull\n\t\tos=-bosx\n\t\t;;\n\tdpx2* | dpx2*-bull)\n\t\tbasic_machine=m68k-bull\n\t\tos=-sysv3\n\t\t;;\n\tebmon29k)\n\t\tbasic_machine=a29k-amd\n\t\tos=-ebmon\n\t\t;;\n\telxsi)\n\t\tbasic_machine=elxsi-elxsi\n\t\tos=-bsd\n\t\t;;\n\tencore | umax | mmax)\n\t\tbasic_machine=ns32k-encore\n\t\t;;\n\tes1800 | OSE68k | ose68k | ose | OSE)\n\t\tbasic_machine=m68k-ericsson\n\t\tos=-ose\n\t\t;;\n\tfx2800)\n\t\tbasic_machine=i860-alliant\n\t\t;;\n\tgenix)\n\t\tbasic_machine=ns32k-ns\n\t\t;;\n\tgmicro)\n\t\tbasic_machine=tron-gmicro\n\t\tos=-sysv\n\t\t;;\n\tgo32)\n\t\tbasic_machine=i386-pc\n\t\tos=-go32\n\t\t;;\n\th3050r* | hiux*)\n\t\tbasic_machine=hppa1.1-hitachi\n\t\tos=-hiuxwe2\n\t\t;;\n\th8300hms)\n\t\tbasic_machine=h8300-hitachi\n\t\tos=-hms\n\t\t;;\n\th8300xray)\n\t\tbasic_machine=h8300-hitachi\n\t\tos=-xray\n\t\t;;\n\th8500hms)\n\t\tbasic_machine=h8500-hitachi\n\t\tos=-hms\n\t\t;;\n\tharris)\n\t\tbasic_machine=m88k-harris\n\t\tos=-sysv3\n\t\t;;\n\thp300-*)\n\t\tbasic_machine=m68k-hp\n\t\t;;\n\thp300bsd)\n\t\tbasic_machine=m68k-hp\n\t\tos=-bsd\n\t\t;;\n\thp300hpux)\n\t\tbasic_machine=m68k-hp\n\t\tos=-hpux\n\t\t;;\n\thp3k9[0-9][0-9] | hp9[0-9][0-9])\n\t\tbasic_machine=hppa1.0-hp\n\t\t;;\n\thp9k2[0-9][0-9] | hp9k31[0-9])\n\t\tbasic_machine=m68000-hp\n\t\t;;\n\thp9k3[2-9][0-9])\n\t\tbasic_machine=m68k-hp\n\t\t;;\n\thp9k6[0-9][0-9] | hp6[0-9][0-9])\n\t\tbasic_machine=hppa1.0-hp\n\t\t;;\n\thp9k7[0-79][0-9] | hp7[0-79][0-9])\n\t\tbasic_machine=hppa1.1-hp\n\t\t;;\n\thp9k78[0-9] | hp78[0-9])\n\t\t# FIXME: really hppa2.0-hp\n\t\tbasic_machine=hppa1.1-hp\n\t\t;;\n\thp9k8[67]1 | hp8[67]1 | hp9k80[24] | hp80[24] | hp9k8[78]9 | hp8[78]9 | hp9k893 | hp893)\n\t\t# FIXME: really hppa2.0-hp\n\t\tbasic_machine=hppa1.1-hp\n\t\t;;\n\thp9k8[0-9][13679] | hp8[0-9][13679])\n\t\tbasic_machine=hppa1.1-hp\n\t\t;;\n\thp9k8[0-9][0-9] | hp8[0-9][0-9])\n\t\tbasic_machine=hppa1.0-hp\n\t\t;;\n\thppa-next)\n\t\tos=-nextstep3\n\t\t;;\n\thppaosf)\n\t\tbasic_machine=hppa1.1-hp\n\t\tos=-osf\n\t\t;;\n\thppro)\n\t\tbasic_machine=hppa1.1-hp\n\t\tos=-proelf\n\t\t;;\n\ti370-ibm* | ibm*)\n\t\tbasic_machine=i370-ibm\n\t\t;;\n\ti*86v32)\n\t\tbasic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`\n\t\tos=-sysv32\n\t\t;;\n\ti*86v4*)\n\t\tbasic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`\n\t\tos=-sysv4\n\t\t;;\n\ti*86v)\n\t\tbasic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`\n\t\tos=-sysv\n\t\t;;\n\ti*86sol2)\n\t\tbasic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`\n\t\tos=-solaris2\n\t\t;;\n\ti386mach)\n\t\tbasic_machine=i386-mach\n\t\tos=-mach\n\t\t;;\n\ti386-vsta | vsta)\n\t\tbasic_machine=i386-unknown\n\t\tos=-vsta\n\t\t;;\n\tiris | iris4d)\n\t\tbasic_machine=mips-sgi\n\t\tcase $os in\n\t\t -irix*)\n\t\t\t;;\n\t\t *)\n\t\t\tos=-irix4\n\t\t\t;;\n\t\tesac\n\t\t;;\n\tisi68 | isi)\n\t\tbasic_machine=m68k-isi\n\t\tos=-sysv\n\t\t;;\n\tm68knommu)\n\t\tbasic_machine=m68k-unknown\n\t\tos=-linux\n\t\t;;\n\tm68knommu-*)\n\t\tbasic_machine=m68k-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\tos=-linux\n\t\t;;\n\tm88k-omron*)\n\t\tbasic_machine=m88k-omron\n\t\t;;\n\tmagnum | m3230)\n\t\tbasic_machine=mips-mips\n\t\tos=-sysv\n\t\t;;\n\tmerlin)\n\t\tbasic_machine=ns32k-utek\n\t\tos=-sysv\n\t\t;;\n\tmicroblaze)\n\t\tbasic_machine=microblaze-xilinx\n\t\t;;\n\tmingw32)\n\t\tbasic_machine=i386-pc\n\t\tos=-mingw32\n\t\t;;\n\tmingw32ce)\n\t\tbasic_machine=arm-unknown\n\t\tos=-mingw32ce\n\t\t;;\n\tminiframe)\n\t\tbasic_machine=m68000-convergent\n\t\t;;\n\t*mint | -mint[0-9]* | *MiNT | *MiNT[0-9]*)\n\t\tbasic_machine=m68k-atari\n\t\tos=-mint\n\t\t;;\n\tmips3*-*)\n\t\tbasic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`\n\t\t;;\n\tmips3*)\n\t\tbasic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`-unknown\n\t\t;;\n\tmonitor)\n\t\tbasic_machine=m68k-rom68k\n\t\tos=-coff\n\t\t;;\n\tmorphos)\n\t\tbasic_machine=powerpc-unknown\n\t\tos=-morphos\n\t\t;;\n\tmsdos)\n\t\tbasic_machine=i386-pc\n\t\tos=-msdos\n\t\t;;\n\tms1-*)\n\t\tbasic_machine=`echo $basic_machine | sed -e 's/ms1-/mt-/'`\n\t\t;;\n\tmsys)\n\t\tbasic_machine=i386-pc\n\t\tos=-msys\n\t\t;;\n\tmvs)\n\t\tbasic_machine=i370-ibm\n\t\tos=-mvs\n\t\t;;\n\tnacl)\n\t\tbasic_machine=le32-unknown\n\t\tos=-nacl\n\t\t;;\n\tncr3000)\n\t\tbasic_machine=i486-ncr\n\t\tos=-sysv4\n\t\t;;\n\tnetbsd386)\n\t\tbasic_machine=i386-unknown\n\t\tos=-netbsd\n\t\t;;\n\tnetwinder)\n\t\tbasic_machine=armv4l-rebel\n\t\tos=-linux\n\t\t;;\n\tnews | news700 | news800 | news900)\n\t\tbasic_machine=m68k-sony\n\t\tos=-newsos\n\t\t;;\n\tnews1000)\n\t\tbasic_machine=m68030-sony\n\t\tos=-newsos\n\t\t;;\n\tnews-3600 | risc-news)\n\t\tbasic_machine=mips-sony\n\t\tos=-newsos\n\t\t;;\n\tnecv70)\n\t\tbasic_machine=v70-nec\n\t\tos=-sysv\n\t\t;;\n\tnext | m*-next )\n\t\tbasic_machine=m68k-next\n\t\tcase $os in\n\t\t -nextstep* )\n\t\t\t;;\n\t\t -ns2*)\n\t\t os=-nextstep2\n\t\t\t;;\n\t\t *)\n\t\t os=-nextstep3\n\t\t\t;;\n\t\tesac\n\t\t;;\n\tnh3000)\n\t\tbasic_machine=m68k-harris\n\t\tos=-cxux\n\t\t;;\n\tnh[45]000)\n\t\tbasic_machine=m88k-harris\n\t\tos=-cxux\n\t\t;;\n\tnindy960)\n\t\tbasic_machine=i960-intel\n\t\tos=-nindy\n\t\t;;\n\tmon960)\n\t\tbasic_machine=i960-intel\n\t\tos=-mon960\n\t\t;;\n\tnonstopux)\n\t\tbasic_machine=mips-compaq\n\t\tos=-nonstopux\n\t\t;;\n\tnp1)\n\t\tbasic_machine=np1-gould\n\t\t;;\n\tneo-tandem)\n\t\tbasic_machine=neo-tandem\n\t\t;;\n\tnse-tandem)\n\t\tbasic_machine=nse-tandem\n\t\t;;\n\tnsr-tandem)\n\t\tbasic_machine=nsr-tandem\n\t\t;;\n\top50n-* | op60c-*)\n\t\tbasic_machine=hppa1.1-oki\n\t\tos=-proelf\n\t\t;;\n\topenrisc | openrisc-*)\n\t\tbasic_machine=or32-unknown\n\t\t;;\n\tos400)\n\t\tbasic_machine=powerpc-ibm\n\t\tos=-os400\n\t\t;;\n\tOSE68000 | ose68000)\n\t\tbasic_machine=m68000-ericsson\n\t\tos=-ose\n\t\t;;\n\tos68k)\n\t\tbasic_machine=m68k-none\n\t\tos=-os68k\n\t\t;;\n\tpa-hitachi)\n\t\tbasic_machine=hppa1.1-hitachi\n\t\tos=-hiuxwe2\n\t\t;;\n\tparagon)\n\t\tbasic_machine=i860-intel\n\t\tos=-osf\n\t\t;;\n\tparisc)\n\t\tbasic_machine=hppa-unknown\n\t\tos=-linux\n\t\t;;\n\tparisc-*)\n\t\tbasic_machine=hppa-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\tos=-linux\n\t\t;;\n\tpbd)\n\t\tbasic_machine=sparc-tti\n\t\t;;\n\tpbb)\n\t\tbasic_machine=m68k-tti\n\t\t;;\n\tpc532 | pc532-*)\n\t\tbasic_machine=ns32k-pc532\n\t\t;;\n\tpc98)\n\t\tbasic_machine=i386-pc\n\t\t;;\n\tpc98-*)\n\t\tbasic_machine=i386-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tpentium | p5 | k5 | k6 | nexgen | viac3)\n\t\tbasic_machine=i586-pc\n\t\t;;\n\tpentiumpro | p6 | 6x86 | athlon | athlon_*)\n\t\tbasic_machine=i686-pc\n\t\t;;\n\tpentiumii | pentium2 | pentiumiii | pentium3)\n\t\tbasic_machine=i686-pc\n\t\t;;\n\tpentium4)\n\t\tbasic_machine=i786-pc\n\t\t;;\n\tpentium-* | p5-* | k5-* | k6-* | nexgen-* | viac3-*)\n\t\tbasic_machine=i586-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tpentiumpro-* | p6-* | 6x86-* | athlon-*)\n\t\tbasic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tpentiumii-* | pentium2-* | pentiumiii-* | pentium3-*)\n\t\tbasic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tpentium4-*)\n\t\tbasic_machine=i786-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tpn)\n\t\tbasic_machine=pn-gould\n\t\t;;\n\tpower)\tbasic_machine=power-ibm\n\t\t;;\n\tppc | ppcbe)\tbasic_machine=powerpc-unknown\n\t\t;;\n\tppc-* | ppcbe-*)\n\t\tbasic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tppcle | powerpclittle | ppc-le | powerpc-little)\n\t\tbasic_machine=powerpcle-unknown\n\t\t;;\n\tppcle-* | powerpclittle-*)\n\t\tbasic_machine=powerpcle-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tppc64)\tbasic_machine=powerpc64-unknown\n\t\t;;\n\tppc64-*) basic_machine=powerpc64-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tppc64le | powerpc64little | ppc64-le | powerpc64-little)\n\t\tbasic_machine=powerpc64le-unknown\n\t\t;;\n\tppc64le-* | powerpc64little-*)\n\t\tbasic_machine=powerpc64le-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tps2)\n\t\tbasic_machine=i386-ibm\n\t\t;;\n\tpw32)\n\t\tbasic_machine=i586-unknown\n\t\tos=-pw32\n\t\t;;\n\trdos)\n\t\tbasic_machine=i386-pc\n\t\tos=-rdos\n\t\t;;\n\trom68k)\n\t\tbasic_machine=m68k-rom68k\n\t\tos=-coff\n\t\t;;\n\trm[46]00)\n\t\tbasic_machine=mips-siemens\n\t\t;;\n\trtpc | rtpc-*)\n\t\tbasic_machine=romp-ibm\n\t\t;;\n\ts390 | s390-*)\n\t\tbasic_machine=s390-ibm\n\t\t;;\n\ts390x | s390x-*)\n\t\tbasic_machine=s390x-ibm\n\t\t;;\n\tsa29200)\n\t\tbasic_machine=a29k-amd\n\t\tos=-udi\n\t\t;;\n\tsb1)\n\t\tbasic_machine=mipsisa64sb1-unknown\n\t\t;;\n\tsb1el)\n\t\tbasic_machine=mipsisa64sb1el-unknown\n\t\t;;\n\tsde)\n\t\tbasic_machine=mipsisa32-sde\n\t\tos=-elf\n\t\t;;\n\tsei)\n\t\tbasic_machine=mips-sei\n\t\tos=-seiux\n\t\t;;\n\tsequent)\n\t\tbasic_machine=i386-sequent\n\t\t;;\n\tsh)\n\t\tbasic_machine=sh-hitachi\n\t\tos=-hms\n\t\t;;\n\tsh5el)\n\t\tbasic_machine=sh5le-unknown\n\t\t;;\n\tsh64)\n\t\tbasic_machine=sh64-unknown\n\t\t;;\n\tsparclite-wrs | simso-wrs)\n\t\tbasic_machine=sparclite-wrs\n\t\tos=-vxworks\n\t\t;;\n\tsps7)\n\t\tbasic_machine=m68k-bull\n\t\tos=-sysv2\n\t\t;;\n\tspur)\n\t\tbasic_machine=spur-unknown\n\t\t;;\n\tst2000)\n\t\tbasic_machine=m68k-tandem\n\t\t;;\n\tstratus)\n\t\tbasic_machine=i860-stratus\n\t\tos=-sysv4\n\t\t;;\n\tstrongarm-* | thumb-*)\n\t\tbasic_machine=arm-`echo $basic_machine | sed 's/^[^-]*-//'`\n\t\t;;\n\tsun2)\n\t\tbasic_machine=m68000-sun\n\t\t;;\n\tsun2os3)\n\t\tbasic_machine=m68000-sun\n\t\tos=-sunos3\n\t\t;;\n\tsun2os4)\n\t\tbasic_machine=m68000-sun\n\t\tos=-sunos4\n\t\t;;\n\tsun3os3)\n\t\tbasic_machine=m68k-sun\n\t\tos=-sunos3\n\t\t;;\n\tsun3os4)\n\t\tbasic_machine=m68k-sun\n\t\tos=-sunos4\n\t\t;;\n\tsun4os3)\n\t\tbasic_machine=sparc-sun\n\t\tos=-sunos3\n\t\t;;\n\tsun4os4)\n\t\tbasic_machine=sparc-sun\n\t\tos=-sunos4\n\t\t;;\n\tsun4sol2)\n\t\tbasic_machine=sparc-sun\n\t\tos=-solaris2\n\t\t;;\n\tsun3 | sun3-*)\n\t\tbasic_machine=m68k-sun\n\t\t;;\n\tsun4)\n\t\tbasic_machine=sparc-sun\n\t\t;;\n\tsun386 | sun386i | roadrunner)\n\t\tbasic_machine=i386-sun\n\t\t;;\n\tsv1)\n\t\tbasic_machine=sv1-cray\n\t\tos=-unicos\n\t\t;;\n\tsymmetry)\n\t\tbasic_machine=i386-sequent\n\t\tos=-dynix\n\t\t;;\n\tt3e)\n\t\tbasic_machine=alphaev5-cray\n\t\tos=-unicos\n\t\t;;\n\tt90)\n\t\tbasic_machine=t90-cray\n\t\tos=-unicos\n\t\t;;\n\ttile*)\n\t\tbasic_machine=$basic_machine-unknown\n\t\tos=-linux-gnu\n\t\t;;\n\ttx39)\n\t\tbasic_machine=mipstx39-unknown\n\t\t;;\n\ttx39el)\n\t\tbasic_machine=mipstx39el-unknown\n\t\t;;\n\ttoad1)\n\t\tbasic_machine=pdp10-xkl\n\t\tos=-tops20\n\t\t;;\n\ttower | tower-32)\n\t\tbasic_machine=m68k-ncr\n\t\t;;\n\ttpf)\n\t\tbasic_machine=s390x-ibm\n\t\tos=-tpf\n\t\t;;\n\tudi29k)\n\t\tbasic_machine=a29k-amd\n\t\tos=-udi\n\t\t;;\n\tultra3)\n\t\tbasic_machine=a29k-nyu\n\t\tos=-sym1\n\t\t;;\n\tv810 | necv810)\n\t\tbasic_machine=v810-nec\n\t\tos=-none\n\t\t;;\n\tvaxv)\n\t\tbasic_machine=vax-dec\n\t\tos=-sysv\n\t\t;;\n\tvms)\n\t\tbasic_machine=vax-dec\n\t\tos=-vms\n\t\t;;\n\tvpp*|vx|vx-*)\n\t\tbasic_machine=f301-fujitsu\n\t\t;;\n\tvxworks960)\n\t\tbasic_machine=i960-wrs\n\t\tos=-vxworks\n\t\t;;\n\tvxworks68)\n\t\tbasic_machine=m68k-wrs\n\t\tos=-vxworks\n\t\t;;\n\tvxworks29k)\n\t\tbasic_machine=a29k-wrs\n\t\tos=-vxworks\n\t\t;;\n\tw65*)\n\t\tbasic_machine=w65-wdc\n\t\tos=-none\n\t\t;;\n\tw89k-*)\n\t\tbasic_machine=hppa1.1-winbond\n\t\tos=-proelf\n\t\t;;\n\txbox)\n\t\tbasic_machine=i686-pc\n\t\tos=-mingw32\n\t\t;;\n\txps | xps100)\n\t\tbasic_machine=xps100-honeywell\n\t\t;;\n\txscale-* | xscalee[bl]-*)\n\t\tbasic_machine=`echo $basic_machine | sed 's/^xscale/arm/'`\n\t\t;;\n\tymp)\n\t\tbasic_machine=ymp-cray\n\t\tos=-unicos\n\t\t;;\n\tz8k-*-coff)\n\t\tbasic_machine=z8k-unknown\n\t\tos=-sim\n\t\t;;\n\tz80-*-coff)\n\t\tbasic_machine=z80-unknown\n\t\tos=-sim\n\t\t;;\n\tnone)\n\t\tbasic_machine=none-none\n\t\tos=-none\n\t\t;;\n\n# Here we handle the default manufacturer of certain CPU types. It is in\n# some cases the only manufacturer, in others, it is the most popular.\n\tw89k)\n\t\tbasic_machine=hppa1.1-winbond\n\t\t;;\n\top50n)\n\t\tbasic_machine=hppa1.1-oki\n\t\t;;\n\top60c)\n\t\tbasic_machine=hppa1.1-oki\n\t\t;;\n\tromp)\n\t\tbasic_machine=romp-ibm\n\t\t;;\n\tmmix)\n\t\tbasic_machine=mmix-knuth\n\t\t;;\n\trs6000)\n\t\tbasic_machine=rs6000-ibm\n\t\t;;\n\tvax)\n\t\tbasic_machine=vax-dec\n\t\t;;\n\tpdp10)\n\t\t# there are many clones, so DEC is not a safe bet\n\t\tbasic_machine=pdp10-unknown\n\t\t;;\n\tpdp11)\n\t\tbasic_machine=pdp11-dec\n\t\t;;\n\twe32k)\n\t\tbasic_machine=we32k-att\n\t\t;;\n\tsh[1234] | sh[24]a | sh[24]aeb | sh[34]eb | sh[1234]le | sh[23]ele)\n\t\tbasic_machine=sh-unknown\n\t\t;;\n\tsparc | sparcv8 | sparcv9 | sparcv9b | sparcv9v)\n\t\tbasic_machine=sparc-sun\n\t\t;;\n\tcydra)\n\t\tbasic_machine=cydra-cydrome\n\t\t;;\n\torion)\n\t\tbasic_machine=orion-highlevel\n\t\t;;\n\torion105)\n\t\tbasic_machine=clipper-highlevel\n\t\t;;\n\tmac | mpw | mac-mpw)\n\t\tbasic_machine=m68k-apple\n\t\t;;\n\tpmac | pmac-mpw)\n\t\tbasic_machine=powerpc-apple\n\t\t;;\n\t*-unknown)\n\t\t# Make sure to match an already-canonicalized machine name.\n\t\t;;\n\t*)\n\t\techo Invalid configuration \\`$1\\': machine \\`$basic_machine\\' not recognized 1>&2\n\t\texit 1\n\t\t;;\nesac\n\n# Here we canonicalize certain aliases for manufacturers.\ncase $basic_machine in\n\t*-digital*)\n\t\tbasic_machine=`echo $basic_machine | sed 's/digital.*/dec/'`\n\t\t;;\n\t*-commodore*)\n\t\tbasic_machine=`echo $basic_machine | sed 's/commodore.*/cbm/'`\n\t\t;;\n\t*)\n\t\t;;\nesac\n\n# Decode manufacturer-specific aliases for certain operating systems.\n\nif [ x\"$os\" != x\"\" ]\nthen\ncase $os in\n\t# First match some system type aliases\n\t# that might get confused with valid system types.\n\t# -solaris* is a basic system type, with this one exception.\n\t-auroraux)\n\t\tos=-auroraux\n\t\t;;\n\t-solaris1 | -solaris1.*)\n\t\tos=`echo $os | sed -e 's|solaris1|sunos4|'`\n\t\t;;\n\t-solaris)\n\t\tos=-solaris2\n\t\t;;\n\t-svr4*)\n\t\tos=-sysv4\n\t\t;;\n\t-unixware*)\n\t\tos=-sysv4.2uw\n\t\t;;\n\t-gnu/linux*)\n\t\tos=`echo $os | sed -e 's|gnu/linux|linux-gnu|'`\n\t\t;;\n\t# First accept the basic system types.\n\t# The portable systems comes first.\n\t# Each alternative MUST END IN A *, to match a version number.\n\t# -sysv* is not here because it comes later, after sysvr4.\n\t-gnu* | -bsd* | -mach* | -minix* | -genix* | -ultrix* | -irix* \\\n\t | -*vms* | -sco* | -esix* | -isc* | -aix* | -cnk* | -sunos | -sunos[34]*\\\n\t | -hpux* | -unos* | -osf* | -luna* | -dgux* | -auroraux* | -solaris* \\\n\t | -sym* | -kopensolaris* \\\n\t | -amigaos* | -amigados* | -msdos* | -newsos* | -unicos* | -aof* \\\n\t | -aos* | -aros* \\\n\t | -nindy* | -vxsim* | -vxworks* | -ebmon* | -hms* | -mvs* \\\n\t | -clix* | -riscos* | -uniplus* | -iris* | -rtu* | -xenix* \\\n\t | -hiux* | -386bsd* | -knetbsd* | -mirbsd* | -netbsd* \\\n\t | -openbsd* | -solidbsd* \\\n\t | -ekkobsd* | -kfreebsd* | -freebsd* | -riscix* | -lynxos* \\\n\t | -bosx* | -nextstep* | -cxux* | -aout* | -elf* | -oabi* \\\n\t | -ptx* | -coff* | -ecoff* | -winnt* | -domain* | -vsta* \\\n\t | -udi* | -eabi* | -lites* | -ieee* | -go32* | -aux* \\\n\t | -chorusos* | -chorusrdb* | -cegcc* \\\n\t | -cygwin* | -msys* | -pe* | -psos* | -moss* | -proelf* | -rtems* \\\n\t | -mingw32* | -linux-gnu* | -linux-android* \\\n\t | -linux-newlib* | -linux-uclibc* \\\n\t | -uxpv* | -beos* | -mpeix* | -udk* \\\n\t | -interix* | -uwin* | -mks* | -rhapsody* | -darwin* | -opened* \\\n\t | -openstep* | -oskit* | -conix* | -pw32* | -nonstopux* \\\n\t | -storm-chaos* | -tops10* | -tenex* | -tops20* | -its* \\\n\t | -os2* | -vos* | -palmos* | -uclinux* | -nucleus* \\\n\t | -morphos* | -superux* | -rtmk* | -rtmk-nova* | -windiss* \\\n\t | -powermax* | -dnix* | -nx6 | -nx7 | -sei* | -dragonfly* \\\n\t | -skyos* | -haiku* | -rdos* | -toppers* | -drops* | -es*)\n\t# Remember, each alternative MUST END IN *, to match a version number.\n\t\t;;\n\t-qnx*)\n\t\tcase $basic_machine in\n\t\t x86-* | i*86-*)\n\t\t\t;;\n\t\t *)\n\t\t\tos=-nto$os\n\t\t\t;;\n\t\tesac\n\t\t;;\n\t-nto-qnx*)\n\t\t;;\n\t-nto*)\n\t\tos=`echo $os | sed -e 's|nto|nto-qnx|'`\n\t\t;;\n\t-sim | -es1800* | -hms* | -xray | -os68k* | -none* | -v88r* \\\n\t | -windows* | -osx | -abug | -netware* | -os9* | -beos* | -haiku* \\\n\t | -macos* | -mpw* | -magic* | -mmixware* | -mon960* | -lnews*)\n\t\t;;\n\t-mac*)\n\t\tos=`echo $os | sed -e 's|mac|macos|'`\n\t\t;;\n\t-linux-dietlibc)\n\t\tos=-linux-dietlibc\n\t\t;;\n\t-linux*)\n\t\tos=`echo $os | sed -e 's|linux|linux-gnu|'`\n\t\t;;\n\t-sunos5*)\n\t\tos=`echo $os | sed -e 's|sunos5|solaris2|'`\n\t\t;;\n\t-sunos6*)\n\t\tos=`echo $os | sed -e 's|sunos6|solaris3|'`\n\t\t;;\n\t-opened*)\n\t\tos=-openedition\n\t\t;;\n\t-os400*)\n\t\tos=-os400\n\t\t;;\n\t-wince*)\n\t\tos=-wince\n\t\t;;\n\t-osfrose*)\n\t\tos=-osfrose\n\t\t;;\n\t-osf*)\n\t\tos=-osf\n\t\t;;\n\t-utek*)\n\t\tos=-bsd\n\t\t;;\n\t-dynix*)\n\t\tos=-bsd\n\t\t;;\n\t-acis*)\n\t\tos=-aos\n\t\t;;\n\t-atheos*)\n\t\tos=-atheos\n\t\t;;\n\t-syllable*)\n\t\tos=-syllable\n\t\t;;\n\t-386bsd)\n\t\tos=-bsd\n\t\t;;\n\t-ctix* | -uts*)\n\t\tos=-sysv\n\t\t;;\n\t-nova*)\n\t\tos=-rtmk-nova\n\t\t;;\n\t-ns2 )\n\t\tos=-nextstep2\n\t\t;;\n\t-nsk*)\n\t\tos=-nsk\n\t\t;;\n\t# Preserve the version number of sinix5.\n\t-sinix5.*)\n\t\tos=`echo $os | sed -e 's|sinix|sysv|'`\n\t\t;;\n\t-sinix*)\n\t\tos=-sysv4\n\t\t;;\n\t-tpf*)\n\t\tos=-tpf\n\t\t;;\n\t-triton*)\n\t\tos=-sysv3\n\t\t;;\n\t-oss*)\n\t\tos=-sysv3\n\t\t;;\n\t-svr4)\n\t\tos=-sysv4\n\t\t;;\n\t-svr3)\n\t\tos=-sysv3\n\t\t;;\n\t-sysvr4)\n\t\tos=-sysv4\n\t\t;;\n\t# This must come after -sysvr4.\n\t-sysv*)\n\t\t;;\n\t-ose*)\n\t\tos=-ose\n\t\t;;\n\t-es1800*)\n\t\tos=-ose\n\t\t;;\n\t-xenix)\n\t\tos=-xenix\n\t\t;;\n\t-*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*)\n\t\tos=-mint\n\t\t;;\n\t-aros*)\n\t\tos=-aros\n\t\t;;\n\t-kaos*)\n\t\tos=-kaos\n\t\t;;\n\t-zvmoe)\n\t\tos=-zvmoe\n\t\t;;\n\t-dicos*)\n\t\tos=-dicos\n\t\t;;\n\t-nacl*)\n\t\t;;\n\t-none)\n\t\t;;\n\t*)\n\t\t# Get rid of the `-' at the beginning of $os.\n\t\tos=`echo $os | sed 's/[^-]*-//'`\n\t\techo Invalid configuration \\`$1\\': system \\`$os\\' not recognized 1>&2\n\t\texit 1\n\t\t;;\nesac\nelse\n\n# Here we handle the default operating systems that come with various machines.\n# The value should be what the vendor currently ships out the door with their\n# machine or put another way, the most popular os provided with the machine.\n\n# Note that if you're going to try to match \"-MANUFACTURER\" here (say,\n# \"-sun\"), then you have to tell the case statement up towards the top\n# that MANUFACTURER isn't an operating system. Otherwise, code above\n# will signal an error saying that MANUFACTURER isn't an operating\n# system, and we'll never get to this point.\n\ncase $basic_machine in\n\tscore-*)\n\t\tos=-elf\n\t\t;;\n\tspu-*)\n\t\tos=-elf\n\t\t;;\n\t*-acorn)\n\t\tos=-riscix1.2\n\t\t;;\n\tarm*-rebel)\n\t\tos=-linux\n\t\t;;\n\tarm*-semi)\n\t\tos=-aout\n\t\t;;\n\tc4x-* | tic4x-*)\n\t\tos=-coff\n\t\t;;\n\ttic54x-*)\n\t\tos=-coff\n\t\t;;\n\ttic55x-*)\n\t\tos=-coff\n\t\t;;\n\ttic6x-*)\n\t\tos=-coff\n\t\t;;\n\t# This must come before the *-dec entry.\n\tpdp10-*)\n\t\tos=-tops20\n\t\t;;\n\tpdp11-*)\n\t\tos=-none\n\t\t;;\n\t*-dec | vax-*)\n\t\tos=-ultrix4.2\n\t\t;;\n\tm68*-apollo)\n\t\tos=-domain\n\t\t;;\n\ti386-sun)\n\t\tos=-sunos4.0.2\n\t\t;;\n\tm68000-sun)\n\t\tos=-sunos3\n\t\t;;\n\tm68*-cisco)\n\t\tos=-aout\n\t\t;;\n\tmep-*)\n\t\tos=-elf\n\t\t;;\n\tmips*-cisco)\n\t\tos=-elf\n\t\t;;\n\tmips*-*)\n\t\tos=-elf\n\t\t;;\n\tor32-*)\n\t\tos=-coff\n\t\t;;\n\t*-tti)\t# must be before sparc entry or we get the wrong os.\n\t\tos=-sysv3\n\t\t;;\n\tsparc-* | *-sun)\n\t\tos=-sunos4.1.1\n\t\t;;\n\t*-be)\n\t\tos=-beos\n\t\t;;\n\t*-haiku)\n\t\tos=-haiku\n\t\t;;\n\t*-ibm)\n\t\tos=-aix\n\t\t;;\n\t*-knuth)\n\t\tos=-mmixware\n\t\t;;\n\t*-wec)\n\t\tos=-proelf\n\t\t;;\n\t*-winbond)\n\t\tos=-proelf\n\t\t;;\n\t*-oki)\n\t\tos=-proelf\n\t\t;;\n\t*-hp)\n\t\tos=-hpux\n\t\t;;\n\t*-hitachi)\n\t\tos=-hiux\n\t\t;;\n\ti860-* | *-att | *-ncr | *-altos | *-motorola | *-convergent)\n\t\tos=-sysv\n\t\t;;\n\t*-cbm)\n\t\tos=-amigaos\n\t\t;;\n\t*-dg)\n\t\tos=-dgux\n\t\t;;\n\t*-dolphin)\n\t\tos=-sysv3\n\t\t;;\n\tm68k-ccur)\n\t\tos=-rtu\n\t\t;;\n\tm88k-omron*)\n\t\tos=-luna\n\t\t;;\n\t*-next )\n\t\tos=-nextstep\n\t\t;;\n\t*-sequent)\n\t\tos=-ptx\n\t\t;;\n\t*-crds)\n\t\tos=-unos\n\t\t;;\n\t*-ns)\n\t\tos=-genix\n\t\t;;\n\ti370-*)\n\t\tos=-mvs\n\t\t;;\n\t*-next)\n\t\tos=-nextstep3\n\t\t;;\n\t*-gould)\n\t\tos=-sysv\n\t\t;;\n\t*-highlevel)\n\t\tos=-bsd\n\t\t;;\n\t*-encore)\n\t\tos=-bsd\n\t\t;;\n\t*-sgi)\n\t\tos=-irix\n\t\t;;\n\t*-siemens)\n\t\tos=-sysv4\n\t\t;;\n\t*-masscomp)\n\t\tos=-rtu\n\t\t;;\n\tf30[01]-fujitsu | f700-fujitsu)\n\t\tos=-uxpv\n\t\t;;\n\t*-rom68k)\n\t\tos=-coff\n\t\t;;\n\t*-*bug)\n\t\tos=-coff\n\t\t;;\n\t*-apple)\n\t\tos=-macos\n\t\t;;\n\t*-atari*)\n\t\tos=-mint\n\t\t;;\n\t*)\n\t\tos=-none\n\t\t;;\nesac\nfi\n\n# Here we handle the case where we know the os, and the CPU type, but not the\n# manufacturer. We pick the logical manufacturer.\nvendor=unknown\ncase $basic_machine in\n\t*-unknown)\n\t\tcase $os in\n\t\t\t-riscix*)\n\t\t\t\tvendor=acorn\n\t\t\t\t;;\n\t\t\t-sunos*)\n\t\t\t\tvendor=sun\n\t\t\t\t;;\n\t\t\t-cnk*|-aix*)\n\t\t\t\tvendor=ibm\n\t\t\t\t;;\n\t\t\t-beos*)\n\t\t\t\tvendor=be\n\t\t\t\t;;\n\t\t\t-hpux*)\n\t\t\t\tvendor=hp\n\t\t\t\t;;\n\t\t\t-mpeix*)\n\t\t\t\tvendor=hp\n\t\t\t\t;;\n\t\t\t-hiux*)\n\t\t\t\tvendor=hitachi\n\t\t\t\t;;\n\t\t\t-unos*)\n\t\t\t\tvendor=crds\n\t\t\t\t;;\n\t\t\t-dgux*)\n\t\t\t\tvendor=dg\n\t\t\t\t;;\n\t\t\t-luna*)\n\t\t\t\tvendor=omron\n\t\t\t\t;;\n\t\t\t-genix*)\n\t\t\t\tvendor=ns\n\t\t\t\t;;\n\t\t\t-mvs* | -opened*)\n\t\t\t\tvendor=ibm\n\t\t\t\t;;\n\t\t\t-os400*)\n\t\t\t\tvendor=ibm\n\t\t\t\t;;\n\t\t\t-ptx*)\n\t\t\t\tvendor=sequent\n\t\t\t\t;;\n\t\t\t-tpf*)\n\t\t\t\tvendor=ibm\n\t\t\t\t;;\n\t\t\t-vxsim* | -vxworks* | -windiss*)\n\t\t\t\tvendor=wrs\n\t\t\t\t;;\n\t\t\t-aux*)\n\t\t\t\tvendor=apple\n\t\t\t\t;;\n\t\t\t-hms*)\n\t\t\t\tvendor=hitachi\n\t\t\t\t;;\n\t\t\t-mpw* | -macos*)\n\t\t\t\tvendor=apple\n\t\t\t\t;;\n\t\t\t-*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*)\n\t\t\t\tvendor=atari\n\t\t\t\t;;\n\t\t\t-vos*)\n\t\t\t\tvendor=stratus\n\t\t\t\t;;\n\t\tesac\n\t\tbasic_machine=`echo $basic_machine | sed \"s/unknown/$vendor/\"`\n\t\t;;\nesac\n\necho $basic_machine$os\nexit\n\n# Local variables:\n# eval: (add-hook 'write-file-hooks 'time-stamp)\n# time-stamp-start: \"timestamp='\"\n# time-stamp-format: \"%:y-%02m-%02d\"\n# time-stamp-end: \"'\"\n# End:\n" }, { "path": "deps/jemalloc/configure", "content": "#! /bin/sh\n# Guess values for system-dependent variables and create Makefiles.\n# Generated by GNU Autoconf 2.68.\n#\n#\n# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,\n# 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010 Free Software\n# Foundation, Inc.\n#\n#\n# This configure script is free software; the Free Software Foundation\n# gives unlimited permission to copy, distribute and modify it.\n## -------------------- ##\n## M4sh Initialization. ##\n## -------------------- ##\n\n# Be more Bourne compatible\nDUALCASE=1; export DUALCASE # for MKS sh\nif test -n \"${ZSH_VERSION+set}\" && (emulate sh) >/dev/null 2>&1; then :\n emulate sh\n NULLCMD=:\n # Pre-4.2 versions of Zsh do word splitting on ${1+\"$@\"}, which\n # is contrary to our usage. Disable this feature.\n alias -g '${1+\"$@\"}'='\"$@\"'\n setopt NO_GLOB_SUBST\nelse\n case `(set -o) 2>/dev/null` in #(\n *posix*) :\n set -o posix ;; #(\n *) :\n ;;\nesac\nfi\n\n\nas_nl='\n'\nexport as_nl\n# Printing a long string crashes Solaris 7 /usr/bin/printf.\nas_echo='\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\'\nas_echo=$as_echo$as_echo$as_echo$as_echo$as_echo\nas_echo=$as_echo$as_echo$as_echo$as_echo$as_echo$as_echo\n# Prefer a ksh shell builtin over an external printf program on Solaris,\n# but without wasting forks for bash or zsh.\nif test -z \"$BASH_VERSION$ZSH_VERSION\" \\\n && (test \"X`print -r -- $as_echo`\" = \"X$as_echo\") 2>/dev/null; then\n as_echo='print -r --'\n as_echo_n='print -rn --'\nelif (test \"X`printf %s $as_echo`\" = \"X$as_echo\") 2>/dev/null; then\n as_echo='printf %s\\n'\n as_echo_n='printf %s'\nelse\n if test \"X`(/usr/ucb/echo -n -n $as_echo) 2>/dev/null`\" = \"X-n $as_echo\"; then\n as_echo_body='eval /usr/ucb/echo -n \"$1$as_nl\"'\n as_echo_n='/usr/ucb/echo -n'\n else\n as_echo_body='eval expr \"X$1\" : \"X\\\\(.*\\\\)\"'\n as_echo_n_body='eval\n arg=$1;\n case $arg in #(\n *\"$as_nl\"*)\n\texpr \"X$arg\" : \"X\\\\(.*\\\\)$as_nl\";\n\targ=`expr \"X$arg\" : \".*$as_nl\\\\(.*\\\\)\"`;;\n esac;\n expr \"X$arg\" : \"X\\\\(.*\\\\)\" | tr -d \"$as_nl\"\n '\n export as_echo_n_body\n as_echo_n='sh -c $as_echo_n_body as_echo'\n fi\n export as_echo_body\n as_echo='sh -c $as_echo_body as_echo'\nfi\n\n# The user is always right.\nif test \"${PATH_SEPARATOR+set}\" != set; then\n PATH_SEPARATOR=:\n (PATH='/bin;/bin'; FPATH=$PATH; sh -c :) >/dev/null 2>&1 && {\n (PATH='/bin:/bin'; FPATH=$PATH; sh -c :) >/dev/null 2>&1 ||\n PATH_SEPARATOR=';'\n }\nfi\n\n\n# IFS\n# We need space, tab and new line, in precisely that order. Quoting is\n# there to prevent editors from complaining about space-tab.\n# (If _AS_PATH_WALK were called with IFS unset, it would disable word\n# splitting by setting IFS to empty value.)\nIFS=\" \"\"\t$as_nl\"\n\n# Find who we are. Look in the path if we contain no directory separator.\nas_myself=\ncase $0 in #((\n *[\\\\/]* ) as_myself=$0 ;;\n *) as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n test -r \"$as_dir/$0\" && as_myself=$as_dir/$0 && break\n done\nIFS=$as_save_IFS\n\n ;;\nesac\n# We did not find ourselves, most probably we were run as `sh COMMAND'\n# in which case we are not to be found in the path.\nif test \"x$as_myself\" = x; then\n as_myself=$0\nfi\nif test ! -f \"$as_myself\"; then\n $as_echo \"$as_myself: error: cannot find myself; rerun with an absolute file name\" >&2\n exit 1\nfi\n\n# Unset variables that we do not need and which cause bugs (e.g. in\n# pre-3.0 UWIN ksh). But do not cause bugs in bash 2.01; the \"|| exit 1\"\n# suppresses any \"Segmentation fault\" message there. '((' could\n# trigger a bug in pdksh 5.2.14.\nfor as_var in BASH_ENV ENV MAIL MAILPATH\ndo eval test x\\${$as_var+set} = xset \\\n && ( (unset $as_var) || exit 1) >/dev/null 2>&1 && unset $as_var || :\ndone\nPS1='$ '\nPS2='> '\nPS4='+ '\n\n# NLS nuisances.\nLC_ALL=C\nexport LC_ALL\nLANGUAGE=C\nexport LANGUAGE\n\n# CDPATH.\n(unset CDPATH) >/dev/null 2>&1 && unset CDPATH\n\nif test \"x$CONFIG_SHELL\" = x; then\n as_bourne_compatible=\"if test -n \\\"\\${ZSH_VERSION+set}\\\" && (emulate sh) >/dev/null 2>&1; then :\n emulate sh\n NULLCMD=:\n # Pre-4.2 versions of Zsh do word splitting on \\${1+\\\"\\$@\\\"}, which\n # is contrary to our usage. Disable this feature.\n alias -g '\\${1+\\\"\\$@\\\"}'='\\\"\\$@\\\"'\n setopt NO_GLOB_SUBST\nelse\n case \\`(set -o) 2>/dev/null\\` in #(\n *posix*) :\n set -o posix ;; #(\n *) :\n ;;\nesac\nfi\n\"\n as_required=\"as_fn_return () { (exit \\$1); }\nas_fn_success () { as_fn_return 0; }\nas_fn_failure () { as_fn_return 1; }\nas_fn_ret_success () { return 0; }\nas_fn_ret_failure () { return 1; }\n\nexitcode=0\nas_fn_success || { exitcode=1; echo as_fn_success failed.; }\nas_fn_failure && { exitcode=1; echo as_fn_failure succeeded.; }\nas_fn_ret_success || { exitcode=1; echo as_fn_ret_success failed.; }\nas_fn_ret_failure && { exitcode=1; echo as_fn_ret_failure succeeded.; }\nif ( set x; as_fn_ret_success y && test x = \\\"\\$1\\\" ); then :\n\nelse\n exitcode=1; echo positional parameters were not saved.\nfi\ntest x\\$exitcode = x0 || exit 1\"\n as_suggested=\" as_lineno_1=\";as_suggested=$as_suggested$LINENO;as_suggested=$as_suggested\" as_lineno_1a=\\$LINENO\n as_lineno_2=\";as_suggested=$as_suggested$LINENO;as_suggested=$as_suggested\" as_lineno_2a=\\$LINENO\n eval 'test \\\"x\\$as_lineno_1'\\$as_run'\\\" != \\\"x\\$as_lineno_2'\\$as_run'\\\" &&\n test \\\"x\\`expr \\$as_lineno_1'\\$as_run' + 1\\`\\\" = \\\"x\\$as_lineno_2'\\$as_run'\\\"' || exit 1\ntest \\$(( 1 + 1 )) = 2 || exit 1\"\n if (eval \"$as_required\") 2>/dev/null; then :\n as_have_required=yes\nelse\n as_have_required=no\nfi\n if test x$as_have_required = xyes && (eval \"$as_suggested\") 2>/dev/null; then :\n\nelse\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nas_found=false\nfor as_dir in /bin$PATH_SEPARATOR/usr/bin$PATH_SEPARATOR$PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n as_found=:\n case $as_dir in #(\n\t /*)\n\t for as_base in sh bash ksh sh5; do\n\t # Try only shells that exist, to save several forks.\n\t as_shell=$as_dir/$as_base\n\t if { test -f \"$as_shell\" || test -f \"$as_shell.exe\"; } &&\n\t\t { $as_echo \"$as_bourne_compatible\"\"$as_required\" | as_run=a \"$as_shell\"; } 2>/dev/null; then :\n CONFIG_SHELL=$as_shell as_have_required=yes\n\t\t if { $as_echo \"$as_bourne_compatible\"\"$as_suggested\" | as_run=a \"$as_shell\"; } 2>/dev/null; then :\n break 2\nfi\nfi\n\t done;;\n esac\n as_found=false\ndone\n$as_found || { if { test -f \"$SHELL\" || test -f \"$SHELL.exe\"; } &&\n\t { $as_echo \"$as_bourne_compatible\"\"$as_required\" | as_run=a \"$SHELL\"; } 2>/dev/null; then :\n CONFIG_SHELL=$SHELL as_have_required=yes\nfi; }\nIFS=$as_save_IFS\n\n\n if test \"x$CONFIG_SHELL\" != x; then :\n # We cannot yet assume a decent shell, so we have to provide a\n\t# neutralization value for shells without unset; and this also\n\t# works around shells that cannot unset nonexistent variables.\n\t# Preserve -v and -x to the replacement shell.\n\tBASH_ENV=/dev/null\n\tENV=/dev/null\n\t(unset BASH_ENV) >/dev/null 2>&1 && unset BASH_ENV ENV\n\texport CONFIG_SHELL\n\tcase $- in # ((((\n\t *v*x* | *x*v* ) as_opts=-vx ;;\n\t *v* ) as_opts=-v ;;\n\t *x* ) as_opts=-x ;;\n\t * ) as_opts= ;;\n\tesac\n\texec \"$CONFIG_SHELL\" $as_opts \"$as_myself\" ${1+\"$@\"}\nfi\n\n if test x$as_have_required = xno; then :\n $as_echo \"$0: This script requires a shell more modern than all\"\n $as_echo \"$0: the shells that I found on your system.\"\n if test x${ZSH_VERSION+set} = xset ; then\n $as_echo \"$0: In particular, zsh $ZSH_VERSION has bugs and should\"\n $as_echo \"$0: be upgraded to zsh 4.3.4 or later.\"\n else\n $as_echo \"$0: Please tell bug-autoconf@gnu.org about your system,\n$0: including any error possibly output before this\n$0: message. Then install a modern shell, or manually run\n$0: the script under such a shell if you do have one.\"\n fi\n exit 1\nfi\nfi\nfi\nSHELL=${CONFIG_SHELL-/bin/sh}\nexport SHELL\n# Unset more variables known to interfere with behavior of common tools.\nCLICOLOR_FORCE= GREP_OPTIONS=\nunset CLICOLOR_FORCE GREP_OPTIONS\n\n## --------------------- ##\n## M4sh Shell Functions. ##\n## --------------------- ##\n# as_fn_unset VAR\n# ---------------\n# Portably unset VAR.\nas_fn_unset ()\n{\n { eval $1=; unset $1;}\n}\nas_unset=as_fn_unset\n\n# as_fn_set_status STATUS\n# -----------------------\n# Set $? to STATUS, without forking.\nas_fn_set_status ()\n{\n return $1\n} # as_fn_set_status\n\n# as_fn_exit STATUS\n# -----------------\n# Exit the shell with STATUS, even in a \"trap 0\" or \"set -e\" context.\nas_fn_exit ()\n{\n set +e\n as_fn_set_status $1\n exit $1\n} # as_fn_exit\n\n# as_fn_mkdir_p\n# -------------\n# Create \"$as_dir\" as a directory, including parents if necessary.\nas_fn_mkdir_p ()\n{\n\n case $as_dir in #(\n -*) as_dir=./$as_dir;;\n esac\n test -d \"$as_dir\" || eval $as_mkdir_p || {\n as_dirs=\n while :; do\n case $as_dir in #(\n *\\'*) as_qdir=`$as_echo \"$as_dir\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"`;; #'(\n *) as_qdir=$as_dir;;\n esac\n as_dirs=\"'$as_qdir' $as_dirs\"\n as_dir=`$as_dirname -- \"$as_dir\" ||\n$as_expr X\"$as_dir\" : 'X\\(.*[^/]\\)//*[^/][^/]*/*$' \\| \\\n\t X\"$as_dir\" : 'X\\(//\\)[^/]' \\| \\\n\t X\"$as_dir\" : 'X\\(//\\)$' \\| \\\n\t X\"$as_dir\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X\"$as_dir\" |\n sed '/^X\\(.*[^/]\\)\\/\\/*[^/][^/]*\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)[^/].*/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n test -d \"$as_dir\" && break\n done\n test -z \"$as_dirs\" || eval \"mkdir $as_dirs\"\n } || test -d \"$as_dir\" || as_fn_error $? \"cannot create directory $as_dir\"\n\n\n} # as_fn_mkdir_p\n# as_fn_append VAR VALUE\n# ----------------------\n# Append the text in VALUE to the end of the definition contained in VAR. Take\n# advantage of any shell optimizations that allow amortized linear growth over\n# repeated appends, instead of the typical quadratic growth present in naive\n# implementations.\nif (eval \"as_var=1; as_var+=2; test x\\$as_var = x12\") 2>/dev/null; then :\n eval 'as_fn_append ()\n {\n eval $1+=\\$2\n }'\nelse\n as_fn_append ()\n {\n eval $1=\\$$1\\$2\n }\nfi # as_fn_append\n\n# as_fn_arith ARG...\n# ------------------\n# Perform arithmetic evaluation on the ARGs, and store the result in the\n# global $as_val. Take advantage of shells that can avoid forks. The arguments\n# must be portable across $(()) and expr.\nif (eval \"test \\$(( 1 + 1 )) = 2\") 2>/dev/null; then :\n eval 'as_fn_arith ()\n {\n as_val=$(( $* ))\n }'\nelse\n as_fn_arith ()\n {\n as_val=`expr \"$@\" || test $? -eq 1`\n }\nfi # as_fn_arith\n\n\n# as_fn_error STATUS ERROR [LINENO LOG_FD]\n# ----------------------------------------\n# Output \"`basename $0`: error: ERROR\" to stderr. If LINENO and LOG_FD are\n# provided, also output the error to LOG_FD, referencing LINENO. Then exit the\n# script with STATUS, using 1 if that was 0.\nas_fn_error ()\n{\n as_status=$1; test $as_status -eq 0 && as_status=1\n if test \"$4\"; then\n as_lineno=${as_lineno-\"$3\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n $as_echo \"$as_me:${as_lineno-$LINENO}: error: $2\" >&$4\n fi\n $as_echo \"$as_me: error: $2\" >&2\n as_fn_exit $as_status\n} # as_fn_error\n\nif expr a : '\\(a\\)' >/dev/null 2>&1 &&\n test \"X`expr 00001 : '.*\\(...\\)'`\" = X001; then\n as_expr=expr\nelse\n as_expr=false\nfi\n\nif (basename -- /) >/dev/null 2>&1 && test \"X`basename -- / 2>&1`\" = \"X/\"; then\n as_basename=basename\nelse\n as_basename=false\nfi\n\nif (as_dir=`dirname -- /` && test \"X$as_dir\" = X/) >/dev/null 2>&1; then\n as_dirname=dirname\nelse\n as_dirname=false\nfi\n\nas_me=`$as_basename -- \"$0\" ||\n$as_expr X/\"$0\" : '.*/\\([^/][^/]*\\)/*$' \\| \\\n\t X\"$0\" : 'X\\(//\\)$' \\| \\\n\t X\"$0\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X/\"$0\" |\n sed '/^.*\\/\\([^/][^/]*\\)\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\/\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\/\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n\n# Avoid depending upon Character Ranges.\nas_cr_letters='abcdefghijklmnopqrstuvwxyz'\nas_cr_LETTERS='ABCDEFGHIJKLMNOPQRSTUVWXYZ'\nas_cr_Letters=$as_cr_letters$as_cr_LETTERS\nas_cr_digits='0123456789'\nas_cr_alnum=$as_cr_Letters$as_cr_digits\n\n\n as_lineno_1=$LINENO as_lineno_1a=$LINENO\n as_lineno_2=$LINENO as_lineno_2a=$LINENO\n eval 'test \"x$as_lineno_1'$as_run'\" != \"x$as_lineno_2'$as_run'\" &&\n test \"x`expr $as_lineno_1'$as_run' + 1`\" = \"x$as_lineno_2'$as_run'\"' || {\n # Blame Lee E. McMahon (1931-1989) for sed's syntax. :-)\n sed -n '\n p\n /[$]LINENO/=\n ' <$as_myself |\n sed '\n s/[$]LINENO.*/&-/\n t lineno\n b\n :lineno\n N\n :loop\n s/[$]LINENO\\([^'$as_cr_alnum'_].*\\n\\)\\(.*\\)/\\2\\1\\2/\n t loop\n s/-\\n.*//\n ' >$as_me.lineno &&\n chmod +x \"$as_me.lineno\" ||\n { $as_echo \"$as_me: error: cannot create $as_me.lineno; rerun with a POSIX shell\" >&2; as_fn_exit 1; }\n\n # Don't try to exec as it changes $[0], causing all sort of problems\n # (the dirname of $[0] is not the place where we might find the\n # original and so on. Autoconf is especially sensitive to this).\n . \"./$as_me.lineno\"\n # Exit status is that of the last command.\n exit\n}\n\nECHO_C= ECHO_N= ECHO_T=\ncase `echo -n x` in #(((((\n-n*)\n case `echo 'xy\\c'` in\n *c*) ECHO_T='\t';;\t# ECHO_T is single tab character.\n xy) ECHO_C='\\c';;\n *) echo `echo ksh88 bug on AIX 6.1` > /dev/null\n ECHO_T='\t';;\n esac;;\n*)\n ECHO_N='-n';;\nesac\n\nrm -f conf$$ conf$$.exe conf$$.file\nif test -d conf$$.dir; then\n rm -f conf$$.dir/conf$$.file\nelse\n rm -f conf$$.dir\n mkdir conf$$.dir 2>/dev/null\nfi\nif (echo >conf$$.file) 2>/dev/null; then\n if ln -s conf$$.file conf$$ 2>/dev/null; then\n as_ln_s='ln -s'\n # ... but there are two gotchas:\n # 1) On MSYS, both `ln -s file dir' and `ln file dir' fail.\n # 2) DJGPP < 2.04 has no symlinks; `ln -s' creates a wrapper executable.\n # In both cases, we have to default to `cp -p'.\n ln -s conf$$.file conf$$.dir 2>/dev/null && test ! -f conf$$.exe ||\n as_ln_s='cp -p'\n elif ln conf$$.file conf$$ 2>/dev/null; then\n as_ln_s=ln\n else\n as_ln_s='cp -p'\n fi\nelse\n as_ln_s='cp -p'\nfi\nrm -f conf$$ conf$$.exe conf$$.dir/conf$$.file conf$$.file\nrmdir conf$$.dir 2>/dev/null\n\nif mkdir -p . 2>/dev/null; then\n as_mkdir_p='mkdir -p \"$as_dir\"'\nelse\n test -d ./-p && rmdir ./-p\n as_mkdir_p=false\nfi\n\nif test -x / >/dev/null 2>&1; then\n as_test_x='test -x'\nelse\n if ls -dL / >/dev/null 2>&1; then\n as_ls_L_option=L\n else\n as_ls_L_option=\n fi\n as_test_x='\n eval sh -c '\\''\n if test -d \"$1\"; then\n\ttest -d \"$1/.\";\n else\n\tcase $1 in #(\n\t-*)set \"./$1\";;\n\tesac;\n\tcase `ls -ld'$as_ls_L_option' \"$1\" 2>/dev/null` in #((\n\t???[sx]*):;;*)false;;esac;fi\n '\\'' sh\n '\nfi\nas_executable_p=$as_test_x\n\n# Sed expression to map a string onto a valid CPP name.\nas_tr_cpp=\"eval sed 'y%*$as_cr_letters%P$as_cr_LETTERS%;s%[^_$as_cr_alnum]%_%g'\"\n\n# Sed expression to map a string onto a valid variable name.\nas_tr_sh=\"eval sed 'y%*+%pp%;s%[^_$as_cr_alnum]%_%g'\"\n\n\ntest -n \"$DJDIR\" || exec 7<&0 &1\n\n# Name of the host.\n# hostname on some systems (SVR3.2, old GNU/Linux) returns a bogus exit status,\n# so uname gets run too.\nac_hostname=`(hostname || uname -n) 2>/dev/null | sed 1q`\n\n#\n# Initializations.\n#\nac_default_prefix=/usr/local\nac_clean_files=\nac_config_libobj_dir=.\nLIBOBJS=\ncross_compiling=no\nsubdirs=\nMFLAGS=\nMAKEFLAGS=\n\n# Identity of this package.\nPACKAGE_NAME=\nPACKAGE_TARNAME=\nPACKAGE_VERSION=\nPACKAGE_STRING=\nPACKAGE_BUGREPORT=\nPACKAGE_URL=\n\nac_unique_file=\"Makefile.in\"\n# Factoring default headers for most tests.\nac_includes_default=\"\\\n#include \n#ifdef HAVE_SYS_TYPES_H\n# include \n#endif\n#ifdef HAVE_SYS_STAT_H\n# include \n#endif\n#ifdef STDC_HEADERS\n# include \n# include \n#else\n# ifdef HAVE_STDLIB_H\n# include \n# endif\n#endif\n#ifdef HAVE_STRING_H\n# if !defined STDC_HEADERS && defined HAVE_MEMORY_H\n# include \n# endif\n# include \n#endif\n#ifdef HAVE_STRINGS_H\n# include \n#endif\n#ifdef HAVE_INTTYPES_H\n# include \n#endif\n#ifdef HAVE_STDINT_H\n# include \n#endif\n#ifdef HAVE_UNISTD_H\n# include \n#endif\"\n\nac_subst_vars='LTLIBOBJS\nLIBOBJS\ncfgoutputs_out\ncfgoutputs_in\ncfghdrs_out\ncfghdrs_in\nenable_tls\nenable_lazy_lock\njemalloc_version_gid\njemalloc_version_nrev\njemalloc_version_bugfix\njemalloc_version_minor\njemalloc_version_major\njemalloc_version\nenable_xmalloc\nenable_valgrind\nenable_utrace\nenable_fill\nenable_dss\nenable_munmap\nenable_mremap\nenable_tcache\nenable_prof\nenable_stats\nenable_debug\ninstall_suffix\nenable_experimental\nAUTOCONF\nLD\nAR\nRANLIB\nINSTALL_DATA\nINSTALL_SCRIPT\nINSTALL_PROGRAM\nenable_autogen\nRPATH_EXTRA\nCC_MM\nMKLIB\nLDTARGET\nCTARGET\nPIC_CFLAGS\nSOREV\nEXTRA_LDFLAGS\nDSO_LDFLAGS\nlibprefix\nexe\na\no\nimportlib\nso\nLD_PRELOAD_VAR\nRPATH\nabi\nhost_os\nhost_vendor\nhost_cpu\nhost\nbuild_os\nbuild_vendor\nbuild_cpu\nbuild\nEGREP\nGREP\nCPP\nOBJEXT\nEXEEXT\nac_ct_CC\nCPPFLAGS\nLDFLAGS\nCFLAGS\nCC\nXSLROOT\nXSLTPROC\nMANDIR\nDATADIR\nLIBDIR\nINCLUDEDIR\nBINDIR\nPREFIX\nabs_objroot\nobjroot\nabs_srcroot\nsrcroot\nrev\ntarget_alias\nhost_alias\nbuild_alias\nLIBS\nECHO_T\nECHO_N\nECHO_C\nDEFS\nmandir\nlocaledir\nlibdir\npsdir\npdfdir\ndvidir\nhtmldir\ninfodir\ndocdir\noldincludedir\nincludedir\nlocalstatedir\nsharedstatedir\nsysconfdir\ndatadir\ndatarootdir\nlibexecdir\nsbindir\nbindir\nprogram_transform_name\nprefix\nexec_prefix\nPACKAGE_URL\nPACKAGE_BUGREPORT\nPACKAGE_STRING\nPACKAGE_VERSION\nPACKAGE_TARNAME\nPACKAGE_NAME\nPATH_SEPARATOR\nSHELL'\nac_subst_files=''\nac_user_opts='\nenable_option_checking\nwith_xslroot\nwith_rpath\nenable_autogen\nenable_experimental\nwith_mangling\nwith_jemalloc_prefix\nwith_private_namespace\nwith_install_suffix\nenable_cc_silence\nenable_debug\nenable_stats\nenable_prof\nenable_prof_libunwind\nwith_static_libunwind\nenable_prof_libgcc\nenable_prof_gcc\nenable_tcache\nenable_mremap\nenable_munmap\nenable_dss\nenable_fill\nenable_utrace\nenable_valgrind\nenable_xmalloc\nenable_lazy_lock\nenable_tls\n'\n ac_precious_vars='build_alias\nhost_alias\ntarget_alias\nCC\nCFLAGS\nLDFLAGS\nLIBS\nCPPFLAGS\nCPP'\n\n\n# Initialize some variables set by options.\nac_init_help=\nac_init_version=false\nac_unrecognized_opts=\nac_unrecognized_sep=\n# The variables have the same names as the options, with\n# dashes changed to underlines.\ncache_file=/dev/null\nexec_prefix=NONE\nno_create=\nno_recursion=\nprefix=NONE\nprogram_prefix=NONE\nprogram_suffix=NONE\nprogram_transform_name=s,x,x,\nsilent=\nsite=\nsrcdir=\nverbose=\nx_includes=NONE\nx_libraries=NONE\n\n# Installation directory options.\n# These are left unexpanded so users can \"make install exec_prefix=/foo\"\n# and all the variables that are supposed to be based on exec_prefix\n# by default will actually change.\n# Use braces instead of parens because sh, perl, etc. also accept them.\n# (The list follows the same order as the GNU Coding Standards.)\nbindir='${exec_prefix}/bin'\nsbindir='${exec_prefix}/sbin'\nlibexecdir='${exec_prefix}/libexec'\ndatarootdir='${prefix}/share'\ndatadir='${datarootdir}'\nsysconfdir='${prefix}/etc'\nsharedstatedir='${prefix}/com'\nlocalstatedir='${prefix}/var'\nincludedir='${prefix}/include'\noldincludedir='/usr/include'\ndocdir='${datarootdir}/doc/${PACKAGE}'\ninfodir='${datarootdir}/info'\nhtmldir='${docdir}'\ndvidir='${docdir}'\npdfdir='${docdir}'\npsdir='${docdir}'\nlibdir='${exec_prefix}/lib'\nlocaledir='${datarootdir}/locale'\nmandir='${datarootdir}/man'\n\nac_prev=\nac_dashdash=\nfor ac_option\ndo\n # If the previous option needs an argument, assign it.\n if test -n \"$ac_prev\"; then\n eval $ac_prev=\\$ac_option\n ac_prev=\n continue\n fi\n\n case $ac_option in\n *=?*) ac_optarg=`expr \"X$ac_option\" : '[^=]*=\\(.*\\)'` ;;\n *=) ac_optarg= ;;\n *) ac_optarg=yes ;;\n esac\n\n # Accept the important Cygnus configure options, so we can diagnose typos.\n\n case $ac_dashdash$ac_option in\n --)\n ac_dashdash=yes ;;\n\n -bindir | --bindir | --bindi | --bind | --bin | --bi)\n ac_prev=bindir ;;\n -bindir=* | --bindir=* | --bindi=* | --bind=* | --bin=* | --bi=*)\n bindir=$ac_optarg ;;\n\n -build | --build | --buil | --bui | --bu)\n ac_prev=build_alias ;;\n -build=* | --build=* | --buil=* | --bui=* | --bu=*)\n build_alias=$ac_optarg ;;\n\n -cache-file | --cache-file | --cache-fil | --cache-fi \\\n | --cache-f | --cache- | --cache | --cach | --cac | --ca | --c)\n ac_prev=cache_file ;;\n -cache-file=* | --cache-file=* | --cache-fil=* | --cache-fi=* \\\n | --cache-f=* | --cache-=* | --cache=* | --cach=* | --cac=* | --ca=* | --c=*)\n cache_file=$ac_optarg ;;\n\n --config-cache | -C)\n cache_file=config.cache ;;\n\n -datadir | --datadir | --datadi | --datad)\n ac_prev=datadir ;;\n -datadir=* | --datadir=* | --datadi=* | --datad=*)\n datadir=$ac_optarg ;;\n\n -datarootdir | --datarootdir | --datarootdi | --datarootd | --dataroot \\\n | --dataroo | --dataro | --datar)\n ac_prev=datarootdir ;;\n -datarootdir=* | --datarootdir=* | --datarootdi=* | --datarootd=* \\\n | --dataroot=* | --dataroo=* | --dataro=* | --datar=*)\n datarootdir=$ac_optarg ;;\n\n -disable-* | --disable-*)\n ac_useropt=`expr \"x$ac_option\" : 'x-*disable-\\(.*\\)'`\n # Reject names that are not valid shell variable names.\n expr \"x$ac_useropt\" : \".*[^-+._$as_cr_alnum]\" >/dev/null &&\n as_fn_error $? \"invalid feature name: $ac_useropt\"\n ac_useropt_orig=$ac_useropt\n ac_useropt=`$as_echo \"$ac_useropt\" | sed 's/[-+.]/_/g'`\n case $ac_user_opts in\n *\"\n\"enable_$ac_useropt\"\n\"*) ;;\n *) ac_unrecognized_opts=\"$ac_unrecognized_opts$ac_unrecognized_sep--disable-$ac_useropt_orig\"\n\t ac_unrecognized_sep=', ';;\n esac\n eval enable_$ac_useropt=no ;;\n\n -docdir | --docdir | --docdi | --doc | --do)\n ac_prev=docdir ;;\n -docdir=* | --docdir=* | --docdi=* | --doc=* | --do=*)\n docdir=$ac_optarg ;;\n\n -dvidir | --dvidir | --dvidi | --dvid | --dvi | --dv)\n ac_prev=dvidir ;;\n -dvidir=* | --dvidir=* | --dvidi=* | --dvid=* | --dvi=* | --dv=*)\n dvidir=$ac_optarg ;;\n\n -enable-* | --enable-*)\n ac_useropt=`expr \"x$ac_option\" : 'x-*enable-\\([^=]*\\)'`\n # Reject names that are not valid shell variable names.\n expr \"x$ac_useropt\" : \".*[^-+._$as_cr_alnum]\" >/dev/null &&\n as_fn_error $? \"invalid feature name: $ac_useropt\"\n ac_useropt_orig=$ac_useropt\n ac_useropt=`$as_echo \"$ac_useropt\" | sed 's/[-+.]/_/g'`\n case $ac_user_opts in\n *\"\n\"enable_$ac_useropt\"\n\"*) ;;\n *) ac_unrecognized_opts=\"$ac_unrecognized_opts$ac_unrecognized_sep--enable-$ac_useropt_orig\"\n\t ac_unrecognized_sep=', ';;\n esac\n eval enable_$ac_useropt=\\$ac_optarg ;;\n\n -exec-prefix | --exec_prefix | --exec-prefix | --exec-prefi \\\n | --exec-pref | --exec-pre | --exec-pr | --exec-p | --exec- \\\n | --exec | --exe | --ex)\n ac_prev=exec_prefix ;;\n -exec-prefix=* | --exec_prefix=* | --exec-prefix=* | --exec-prefi=* \\\n | --exec-pref=* | --exec-pre=* | --exec-pr=* | --exec-p=* | --exec-=* \\\n | --exec=* | --exe=* | --ex=*)\n exec_prefix=$ac_optarg ;;\n\n -gas | --gas | --ga | --g)\n # Obsolete; use --with-gas.\n with_gas=yes ;;\n\n -help | --help | --hel | --he | -h)\n ac_init_help=long ;;\n -help=r* | --help=r* | --hel=r* | --he=r* | -hr*)\n ac_init_help=recursive ;;\n -help=s* | --help=s* | --hel=s* | --he=s* | -hs*)\n ac_init_help=short ;;\n\n -host | --host | --hos | --ho)\n ac_prev=host_alias ;;\n -host=* | --host=* | --hos=* | --ho=*)\n host_alias=$ac_optarg ;;\n\n -htmldir | --htmldir | --htmldi | --htmld | --html | --htm | --ht)\n ac_prev=htmldir ;;\n -htmldir=* | --htmldir=* | --htmldi=* | --htmld=* | --html=* | --htm=* \\\n | --ht=*)\n htmldir=$ac_optarg ;;\n\n -includedir | --includedir | --includedi | --included | --include \\\n | --includ | --inclu | --incl | --inc)\n ac_prev=includedir ;;\n -includedir=* | --includedir=* | --includedi=* | --included=* | --include=* \\\n | --includ=* | --inclu=* | --incl=* | --inc=*)\n includedir=$ac_optarg ;;\n\n -infodir | --infodir | --infodi | --infod | --info | --inf)\n ac_prev=infodir ;;\n -infodir=* | --infodir=* | --infodi=* | --infod=* | --info=* | --inf=*)\n infodir=$ac_optarg ;;\n\n -libdir | --libdir | --libdi | --libd)\n ac_prev=libdir ;;\n -libdir=* | --libdir=* | --libdi=* | --libd=*)\n libdir=$ac_optarg ;;\n\n -libexecdir | --libexecdir | --libexecdi | --libexecd | --libexec \\\n | --libexe | --libex | --libe)\n ac_prev=libexecdir ;;\n -libexecdir=* | --libexecdir=* | --libexecdi=* | --libexecd=* | --libexec=* \\\n | --libexe=* | --libex=* | --libe=*)\n libexecdir=$ac_optarg ;;\n\n -localedir | --localedir | --localedi | --localed | --locale)\n ac_prev=localedir ;;\n -localedir=* | --localedir=* | --localedi=* | --localed=* | --locale=*)\n localedir=$ac_optarg ;;\n\n -localstatedir | --localstatedir | --localstatedi | --localstated \\\n | --localstate | --localstat | --localsta | --localst | --locals)\n ac_prev=localstatedir ;;\n -localstatedir=* | --localstatedir=* | --localstatedi=* | --localstated=* \\\n | --localstate=* | --localstat=* | --localsta=* | --localst=* | --locals=*)\n localstatedir=$ac_optarg ;;\n\n -mandir | --mandir | --mandi | --mand | --man | --ma | --m)\n ac_prev=mandir ;;\n -mandir=* | --mandir=* | --mandi=* | --mand=* | --man=* | --ma=* | --m=*)\n mandir=$ac_optarg ;;\n\n -nfp | --nfp | --nf)\n # Obsolete; use --without-fp.\n with_fp=no ;;\n\n -no-create | --no-create | --no-creat | --no-crea | --no-cre \\\n | --no-cr | --no-c | -n)\n no_create=yes ;;\n\n -no-recursion | --no-recursion | --no-recursio | --no-recursi \\\n | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r)\n no_recursion=yes ;;\n\n -oldincludedir | --oldincludedir | --oldincludedi | --oldincluded \\\n | --oldinclude | --oldinclud | --oldinclu | --oldincl | --oldinc \\\n | --oldin | --oldi | --old | --ol | --o)\n ac_prev=oldincludedir ;;\n -oldincludedir=* | --oldincludedir=* | --oldincludedi=* | --oldincluded=* \\\n | --oldinclude=* | --oldinclud=* | --oldinclu=* | --oldincl=* | --oldinc=* \\\n | --oldin=* | --oldi=* | --old=* | --ol=* | --o=*)\n oldincludedir=$ac_optarg ;;\n\n -prefix | --prefix | --prefi | --pref | --pre | --pr | --p)\n ac_prev=prefix ;;\n -prefix=* | --prefix=* | --prefi=* | --pref=* | --pre=* | --pr=* | --p=*)\n prefix=$ac_optarg ;;\n\n -program-prefix | --program-prefix | --program-prefi | --program-pref \\\n | --program-pre | --program-pr | --program-p)\n ac_prev=program_prefix ;;\n -program-prefix=* | --program-prefix=* | --program-prefi=* \\\n | --program-pref=* | --program-pre=* | --program-pr=* | --program-p=*)\n program_prefix=$ac_optarg ;;\n\n -program-suffix | --program-suffix | --program-suffi | --program-suff \\\n | --program-suf | --program-su | --program-s)\n ac_prev=program_suffix ;;\n -program-suffix=* | --program-suffix=* | --program-suffi=* \\\n | --program-suff=* | --program-suf=* | --program-su=* | --program-s=*)\n program_suffix=$ac_optarg ;;\n\n -program-transform-name | --program-transform-name \\\n | --program-transform-nam | --program-transform-na \\\n | --program-transform-n | --program-transform- \\\n | --program-transform | --program-transfor \\\n | --program-transfo | --program-transf \\\n | --program-trans | --program-tran \\\n | --progr-tra | --program-tr | --program-t)\n ac_prev=program_transform_name ;;\n -program-transform-name=* | --program-transform-name=* \\\n | --program-transform-nam=* | --program-transform-na=* \\\n | --program-transform-n=* | --program-transform-=* \\\n | --program-transform=* | --program-transfor=* \\\n | --program-transfo=* | --program-transf=* \\\n | --program-trans=* | --program-tran=* \\\n | --progr-tra=* | --program-tr=* | --program-t=*)\n program_transform_name=$ac_optarg ;;\n\n -pdfdir | --pdfdir | --pdfdi | --pdfd | --pdf | --pd)\n ac_prev=pdfdir ;;\n -pdfdir=* | --pdfdir=* | --pdfdi=* | --pdfd=* | --pdf=* | --pd=*)\n pdfdir=$ac_optarg ;;\n\n -psdir | --psdir | --psdi | --psd | --ps)\n ac_prev=psdir ;;\n -psdir=* | --psdir=* | --psdi=* | --psd=* | --ps=*)\n psdir=$ac_optarg ;;\n\n -q | -quiet | --quiet | --quie | --qui | --qu | --q \\\n | -silent | --silent | --silen | --sile | --sil)\n silent=yes ;;\n\n -sbindir | --sbindir | --sbindi | --sbind | --sbin | --sbi | --sb)\n ac_prev=sbindir ;;\n -sbindir=* | --sbindir=* | --sbindi=* | --sbind=* | --sbin=* \\\n | --sbi=* | --sb=*)\n sbindir=$ac_optarg ;;\n\n -sharedstatedir | --sharedstatedir | --sharedstatedi \\\n | --sharedstated | --sharedstate | --sharedstat | --sharedsta \\\n | --sharedst | --shareds | --shared | --share | --shar \\\n | --sha | --sh)\n ac_prev=sharedstatedir ;;\n -sharedstatedir=* | --sharedstatedir=* | --sharedstatedi=* \\\n | --sharedstated=* | --sharedstate=* | --sharedstat=* | --sharedsta=* \\\n | --sharedst=* | --shareds=* | --shared=* | --share=* | --shar=* \\\n | --sha=* | --sh=*)\n sharedstatedir=$ac_optarg ;;\n\n -site | --site | --sit)\n ac_prev=site ;;\n -site=* | --site=* | --sit=*)\n site=$ac_optarg ;;\n\n -srcdir | --srcdir | --srcdi | --srcd | --src | --sr)\n ac_prev=srcdir ;;\n -srcdir=* | --srcdir=* | --srcdi=* | --srcd=* | --src=* | --sr=*)\n srcdir=$ac_optarg ;;\n\n -sysconfdir | --sysconfdir | --sysconfdi | --sysconfd | --sysconf \\\n | --syscon | --sysco | --sysc | --sys | --sy)\n ac_prev=sysconfdir ;;\n -sysconfdir=* | --sysconfdir=* | --sysconfdi=* | --sysconfd=* | --sysconf=* \\\n | --syscon=* | --sysco=* | --sysc=* | --sys=* | --sy=*)\n sysconfdir=$ac_optarg ;;\n\n -target | --target | --targe | --targ | --tar | --ta | --t)\n ac_prev=target_alias ;;\n -target=* | --target=* | --targe=* | --targ=* | --tar=* | --ta=* | --t=*)\n target_alias=$ac_optarg ;;\n\n -v | -verbose | --verbose | --verbos | --verbo | --verb)\n verbose=yes ;;\n\n -version | --version | --versio | --versi | --vers | -V)\n ac_init_version=: ;;\n\n -with-* | --with-*)\n ac_useropt=`expr \"x$ac_option\" : 'x-*with-\\([^=]*\\)'`\n # Reject names that are not valid shell variable names.\n expr \"x$ac_useropt\" : \".*[^-+._$as_cr_alnum]\" >/dev/null &&\n as_fn_error $? \"invalid package name: $ac_useropt\"\n ac_useropt_orig=$ac_useropt\n ac_useropt=`$as_echo \"$ac_useropt\" | sed 's/[-+.]/_/g'`\n case $ac_user_opts in\n *\"\n\"with_$ac_useropt\"\n\"*) ;;\n *) ac_unrecognized_opts=\"$ac_unrecognized_opts$ac_unrecognized_sep--with-$ac_useropt_orig\"\n\t ac_unrecognized_sep=', ';;\n esac\n eval with_$ac_useropt=\\$ac_optarg ;;\n\n -without-* | --without-*)\n ac_useropt=`expr \"x$ac_option\" : 'x-*without-\\(.*\\)'`\n # Reject names that are not valid shell variable names.\n expr \"x$ac_useropt\" : \".*[^-+._$as_cr_alnum]\" >/dev/null &&\n as_fn_error $? \"invalid package name: $ac_useropt\"\n ac_useropt_orig=$ac_useropt\n ac_useropt=`$as_echo \"$ac_useropt\" | sed 's/[-+.]/_/g'`\n case $ac_user_opts in\n *\"\n\"with_$ac_useropt\"\n\"*) ;;\n *) ac_unrecognized_opts=\"$ac_unrecognized_opts$ac_unrecognized_sep--without-$ac_useropt_orig\"\n\t ac_unrecognized_sep=', ';;\n esac\n eval with_$ac_useropt=no ;;\n\n --x)\n # Obsolete; use --with-x.\n with_x=yes ;;\n\n -x-includes | --x-includes | --x-include | --x-includ | --x-inclu \\\n | --x-incl | --x-inc | --x-in | --x-i)\n ac_prev=x_includes ;;\n -x-includes=* | --x-includes=* | --x-include=* | --x-includ=* | --x-inclu=* \\\n | --x-incl=* | --x-inc=* | --x-in=* | --x-i=*)\n x_includes=$ac_optarg ;;\n\n -x-libraries | --x-libraries | --x-librarie | --x-librari \\\n | --x-librar | --x-libra | --x-libr | --x-lib | --x-li | --x-l)\n ac_prev=x_libraries ;;\n -x-libraries=* | --x-libraries=* | --x-librarie=* | --x-librari=* \\\n | --x-librar=* | --x-libra=* | --x-libr=* | --x-lib=* | --x-li=* | --x-l=*)\n x_libraries=$ac_optarg ;;\n\n -*) as_fn_error $? \"unrecognized option: \\`$ac_option'\nTry \\`$0 --help' for more information\"\n ;;\n\n *=*)\n ac_envvar=`expr \"x$ac_option\" : 'x\\([^=]*\\)='`\n # Reject names that are not valid shell variable names.\n case $ac_envvar in #(\n '' | [0-9]* | *[!_$as_cr_alnum]* )\n as_fn_error $? \"invalid variable name: \\`$ac_envvar'\" ;;\n esac\n eval $ac_envvar=\\$ac_optarg\n export $ac_envvar ;;\n\n *)\n # FIXME: should be removed in autoconf 3.0.\n $as_echo \"$as_me: WARNING: you should use --build, --host, --target\" >&2\n expr \"x$ac_option\" : \".*[^-._$as_cr_alnum]\" >/dev/null &&\n $as_echo \"$as_me: WARNING: invalid host type: $ac_option\" >&2\n : \"${build_alias=$ac_option} ${host_alias=$ac_option} ${target_alias=$ac_option}\"\n ;;\n\n esac\ndone\n\nif test -n \"$ac_prev\"; then\n ac_option=--`echo $ac_prev | sed 's/_/-/g'`\n as_fn_error $? \"missing argument to $ac_option\"\nfi\n\nif test -n \"$ac_unrecognized_opts\"; then\n case $enable_option_checking in\n no) ;;\n fatal) as_fn_error $? \"unrecognized options: $ac_unrecognized_opts\" ;;\n *) $as_echo \"$as_me: WARNING: unrecognized options: $ac_unrecognized_opts\" >&2 ;;\n esac\nfi\n\n# Check all directory arguments for consistency.\nfor ac_var in\texec_prefix prefix bindir sbindir libexecdir datarootdir \\\n\t\tdatadir sysconfdir sharedstatedir localstatedir includedir \\\n\t\toldincludedir docdir infodir htmldir dvidir pdfdir psdir \\\n\t\tlibdir localedir mandir\ndo\n eval ac_val=\\$$ac_var\n # Remove trailing slashes.\n case $ac_val in\n */ )\n ac_val=`expr \"X$ac_val\" : 'X\\(.*[^/]\\)' \\| \"X$ac_val\" : 'X\\(.*\\)'`\n eval $ac_var=\\$ac_val;;\n esac\n # Be sure to have absolute directory names.\n case $ac_val in\n [\\\\/$]* | ?:[\\\\/]* ) continue;;\n NONE | '' ) case $ac_var in *prefix ) continue;; esac;;\n esac\n as_fn_error $? \"expected an absolute directory name for --$ac_var: $ac_val\"\ndone\n\n# There might be people who depend on the old broken behavior: `$host'\n# used to hold the argument of --host etc.\n# FIXME: To remove some day.\nbuild=$build_alias\nhost=$host_alias\ntarget=$target_alias\n\n# FIXME: To remove some day.\nif test \"x$host_alias\" != x; then\n if test \"x$build_alias\" = x; then\n cross_compiling=maybe\n $as_echo \"$as_me: WARNING: if you wanted to set the --build type, don't use --host.\n If a cross compiler is detected then cross compile mode will be used\" >&2\n elif test \"x$build_alias\" != \"x$host_alias\"; then\n cross_compiling=yes\n fi\nfi\n\nac_tool_prefix=\ntest -n \"$host_alias\" && ac_tool_prefix=$host_alias-\n\ntest \"$silent\" = yes && exec 6>/dev/null\n\n\nac_pwd=`pwd` && test -n \"$ac_pwd\" &&\nac_ls_di=`ls -di .` &&\nac_pwd_ls_di=`cd \"$ac_pwd\" && ls -di .` ||\n as_fn_error $? \"working directory cannot be determined\"\ntest \"X$ac_ls_di\" = \"X$ac_pwd_ls_di\" ||\n as_fn_error $? \"pwd does not report name of working directory\"\n\n\n# Find the source files, if location was not specified.\nif test -z \"$srcdir\"; then\n ac_srcdir_defaulted=yes\n # Try the directory containing this script, then the parent directory.\n ac_confdir=`$as_dirname -- \"$as_myself\" ||\n$as_expr X\"$as_myself\" : 'X\\(.*[^/]\\)//*[^/][^/]*/*$' \\| \\\n\t X\"$as_myself\" : 'X\\(//\\)[^/]' \\| \\\n\t X\"$as_myself\" : 'X\\(//\\)$' \\| \\\n\t X\"$as_myself\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X\"$as_myself\" |\n sed '/^X\\(.*[^/]\\)\\/\\/*[^/][^/]*\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)[^/].*/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n srcdir=$ac_confdir\n if test ! -r \"$srcdir/$ac_unique_file\"; then\n srcdir=..\n fi\nelse\n ac_srcdir_defaulted=no\nfi\nif test ! -r \"$srcdir/$ac_unique_file\"; then\n test \"$ac_srcdir_defaulted\" = yes && srcdir=\"$ac_confdir or ..\"\n as_fn_error $? \"cannot find sources ($ac_unique_file) in $srcdir\"\nfi\nac_msg=\"sources are in $srcdir, but \\`cd $srcdir' does not work\"\nac_abs_confdir=`(\n\tcd \"$srcdir\" && test -r \"./$ac_unique_file\" || as_fn_error $? \"$ac_msg\"\n\tpwd)`\n# When building in place, set srcdir=.\nif test \"$ac_abs_confdir\" = \"$ac_pwd\"; then\n srcdir=.\nfi\n# Remove unnecessary trailing slashes from srcdir.\n# Double slashes in file names in object file debugging info\n# mess up M-x gdb in Emacs.\ncase $srcdir in\n*/) srcdir=`expr \"X$srcdir\" : 'X\\(.*[^/]\\)' \\| \"X$srcdir\" : 'X\\(.*\\)'`;;\nesac\nfor ac_var in $ac_precious_vars; do\n eval ac_env_${ac_var}_set=\\${${ac_var}+set}\n eval ac_env_${ac_var}_value=\\$${ac_var}\n eval ac_cv_env_${ac_var}_set=\\${${ac_var}+set}\n eval ac_cv_env_${ac_var}_value=\\$${ac_var}\ndone\n\n#\n# Report the --help message.\n#\nif test \"$ac_init_help\" = \"long\"; then\n # Omit some internal or obsolete options to make the list less imposing.\n # This message is too long to be a string in the A/UX 3.1 sh.\n cat <<_ACEOF\n\\`configure' configures this package to adapt to many kinds of systems.\n\nUsage: $0 [OPTION]... [VAR=VALUE]...\n\nTo assign environment variables (e.g., CC, CFLAGS...), specify them as\nVAR=VALUE. See below for descriptions of some of the useful variables.\n\nDefaults for the options are specified in brackets.\n\nConfiguration:\n -h, --help display this help and exit\n --help=short display options specific to this package\n --help=recursive display the short help of all the included packages\n -V, --version display version information and exit\n -q, --quiet, --silent do not print \\`checking ...' messages\n --cache-file=FILE cache test results in FILE [disabled]\n -C, --config-cache alias for \\`--cache-file=config.cache'\n -n, --no-create do not create output files\n --srcdir=DIR find the sources in DIR [configure dir or \\`..']\n\nInstallation directories:\n --prefix=PREFIX install architecture-independent files in PREFIX\n [$ac_default_prefix]\n --exec-prefix=EPREFIX install architecture-dependent files in EPREFIX\n [PREFIX]\n\nBy default, \\`make install' will install all the files in\n\\`$ac_default_prefix/bin', \\`$ac_default_prefix/lib' etc. You can specify\nan installation prefix other than \\`$ac_default_prefix' using \\`--prefix',\nfor instance \\`--prefix=\\$HOME'.\n\nFor better control, use the options below.\n\nFine tuning of the installation directories:\n --bindir=DIR user executables [EPREFIX/bin]\n --sbindir=DIR system admin executables [EPREFIX/sbin]\n --libexecdir=DIR program executables [EPREFIX/libexec]\n --sysconfdir=DIR read-only single-machine data [PREFIX/etc]\n --sharedstatedir=DIR modifiable architecture-independent data [PREFIX/com]\n --localstatedir=DIR modifiable single-machine data [PREFIX/var]\n --libdir=DIR object code libraries [EPREFIX/lib]\n --includedir=DIR C header files [PREFIX/include]\n --oldincludedir=DIR C header files for non-gcc [/usr/include]\n --datarootdir=DIR read-only arch.-independent data root [PREFIX/share]\n --datadir=DIR read-only architecture-independent data [DATAROOTDIR]\n --infodir=DIR info documentation [DATAROOTDIR/info]\n --localedir=DIR locale-dependent data [DATAROOTDIR/locale]\n --mandir=DIR man documentation [DATAROOTDIR/man]\n --docdir=DIR documentation root [DATAROOTDIR/doc/PACKAGE]\n --htmldir=DIR html documentation [DOCDIR]\n --dvidir=DIR dvi documentation [DOCDIR]\n --pdfdir=DIR pdf documentation [DOCDIR]\n --psdir=DIR ps documentation [DOCDIR]\n_ACEOF\n\n cat <<\\_ACEOF\n\nSystem types:\n --build=BUILD configure for building on BUILD [guessed]\n --host=HOST cross-compile to build programs to run on HOST [BUILD]\n_ACEOF\nfi\n\nif test -n \"$ac_init_help\"; then\n\n cat <<\\_ACEOF\n\nOptional Features:\n --disable-option-checking ignore unrecognized --enable/--with options\n --disable-FEATURE do not include FEATURE (same as --enable-FEATURE=no)\n --enable-FEATURE[=ARG] include FEATURE [ARG=yes]\n --enable-autogen Automatically regenerate configure output\n --disable-experimental Disable support for the experimental API\n --enable-cc-silence Silence irrelevant compiler warnings\n --enable-debug Build debugging code\n --disable-stats Disable statistics calculation/reporting\n --enable-prof Enable allocation profiling\n --enable-prof-libunwind Use libunwind for backtracing\n --disable-prof-libgcc Do not use libgcc for backtracing\n --disable-prof-gcc Do not use gcc intrinsics for backtracing\n --disable-tcache Disable per thread caches\n --enable-mremap Enable mremap(2) for huge realloc()\n --disable-munmap Disable VM deallocation via munmap(2)\n --enable-dss Enable allocation from DSS\n --disable-fill Disable support for junk/zero filling, quarantine,\n and redzones\n --enable-utrace Enable utrace(2)-based tracing\n --disable-valgrind Disable support for Valgrind\n --enable-xmalloc Support xmalloc option\n --enable-lazy-lock Enable lazy locking (only lock when multi-threaded)\n --disable-tls Disable thread-local storage (__thread keyword)\n\nOptional Packages:\n --with-PACKAGE[=ARG] use PACKAGE [ARG=yes]\n --without-PACKAGE do not use PACKAGE (same as --with-PACKAGE=no)\n --with-xslroot= XSL stylesheet root path\n --with-rpath= Colon-separated rpath (ELF systems only)\n --with-mangling= Mangle symbols in \n --with-jemalloc-prefix=\n Prefix to prepend to all public APIs\n --with-private-namespace=\n Prefix to prepend to all library-private APIs\n --with-install-suffix=\n Suffix to append to all installed files\n --with-static-libunwind=\n Path to static libunwind library; use rather than\n dynamically linking\n\nSome influential environment variables:\n CC C compiler command\n CFLAGS C compiler flags\n LDFLAGS linker flags, e.g. -L if you have libraries in a\n nonstandard directory \n LIBS libraries to pass to the linker, e.g. -l\n CPPFLAGS (Objective) C/C++ preprocessor flags, e.g. -I if\n you have headers in a nonstandard directory \n CPP C preprocessor\n\nUse these variables to override the choices made by `configure' or to help\nit to find libraries and programs with nonstandard names/locations.\n\nReport bugs to the package provider.\n_ACEOF\nac_status=$?\nfi\n\nif test \"$ac_init_help\" = \"recursive\"; then\n # If there are subdirs, report their specific --help.\n for ac_dir in : $ac_subdirs_all; do test \"x$ac_dir\" = x: && continue\n test -d \"$ac_dir\" ||\n { cd \"$srcdir\" && ac_pwd=`pwd` && srcdir=. && test -d \"$ac_dir\"; } ||\n continue\n ac_builddir=.\n\ncase \"$ac_dir\" in\n.) ac_dir_suffix= ac_top_builddir_sub=. ac_top_build_prefix= ;;\n*)\n ac_dir_suffix=/`$as_echo \"$ac_dir\" | sed 's|^\\.[\\\\/]||'`\n # A \"..\" for each directory in $ac_dir_suffix.\n ac_top_builddir_sub=`$as_echo \"$ac_dir_suffix\" | sed 's|/[^\\\\/]*|/..|g;s|/||'`\n case $ac_top_builddir_sub in\n \"\") ac_top_builddir_sub=. ac_top_build_prefix= ;;\n *) ac_top_build_prefix=$ac_top_builddir_sub/ ;;\n esac ;;\nesac\nac_abs_top_builddir=$ac_pwd\nac_abs_builddir=$ac_pwd$ac_dir_suffix\n# for backward compatibility:\nac_top_builddir=$ac_top_build_prefix\n\ncase $srcdir in\n .) # We are building in place.\n ac_srcdir=.\n ac_top_srcdir=$ac_top_builddir_sub\n ac_abs_top_srcdir=$ac_pwd ;;\n [\\\\/]* | ?:[\\\\/]* ) # Absolute name.\n ac_srcdir=$srcdir$ac_dir_suffix;\n ac_top_srcdir=$srcdir\n ac_abs_top_srcdir=$srcdir ;;\n *) # Relative name.\n ac_srcdir=$ac_top_build_prefix$srcdir$ac_dir_suffix\n ac_top_srcdir=$ac_top_build_prefix$srcdir\n ac_abs_top_srcdir=$ac_pwd/$srcdir ;;\nesac\nac_abs_srcdir=$ac_abs_top_srcdir$ac_dir_suffix\n\n cd \"$ac_dir\" || { ac_status=$?; continue; }\n # Check for guested configure.\n if test -f \"$ac_srcdir/configure.gnu\"; then\n echo &&\n $SHELL \"$ac_srcdir/configure.gnu\" --help=recursive\n elif test -f \"$ac_srcdir/configure\"; then\n echo &&\n $SHELL \"$ac_srcdir/configure\" --help=recursive\n else\n $as_echo \"$as_me: WARNING: no configuration information is in $ac_dir\" >&2\n fi || ac_status=$?\n cd \"$ac_pwd\" || { ac_status=$?; break; }\n done\nfi\n\ntest -n \"$ac_init_help\" && exit $ac_status\nif $ac_init_version; then\n cat <<\\_ACEOF\nconfigure\ngenerated by GNU Autoconf 2.68\n\nCopyright (C) 2010 Free Software Foundation, Inc.\nThis configure script is free software; the Free Software Foundation\ngives unlimited permission to copy, distribute and modify it.\n_ACEOF\n exit\nfi\n\n## ------------------------ ##\n## Autoconf initialization. ##\n## ------------------------ ##\n\n# ac_fn_c_try_compile LINENO\n# --------------------------\n# Try to compile conftest.$ac_ext, and return whether this succeeded.\nac_fn_c_try_compile ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n rm -f conftest.$ac_objext\n if { { ac_try=\"$ac_compile\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_compile\") 2>conftest.err\n ac_status=$?\n if test -s conftest.err; then\n grep -v '^ *+' conftest.err >conftest.er1\n cat conftest.er1 >&5\n mv -f conftest.er1 conftest.err\n fi\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; } && {\n\t test -z \"$ac_c_werror_flag\" ||\n\t test ! -s conftest.err\n } && test -s conftest.$ac_objext; then :\n ac_retval=0\nelse\n $as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n\tac_retval=1\nfi\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n as_fn_set_status $ac_retval\n\n} # ac_fn_c_try_compile\n\n# ac_fn_c_try_cpp LINENO\n# ----------------------\n# Try to preprocess conftest.$ac_ext, and return whether this succeeded.\nac_fn_c_try_cpp ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n if { { ac_try=\"$ac_cpp conftest.$ac_ext\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_cpp conftest.$ac_ext\") 2>conftest.err\n ac_status=$?\n if test -s conftest.err; then\n grep -v '^ *+' conftest.err >conftest.er1\n cat conftest.er1 >&5\n mv -f conftest.er1 conftest.err\n fi\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; } > conftest.i && {\n\t test -z \"$ac_c_preproc_warn_flag$ac_c_werror_flag\" ||\n\t test ! -s conftest.err\n }; then :\n ac_retval=0\nelse\n $as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n ac_retval=1\nfi\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n as_fn_set_status $ac_retval\n\n} # ac_fn_c_try_cpp\n\n# ac_fn_c_try_run LINENO\n# ----------------------\n# Try to link conftest.$ac_ext, and return whether this succeeded. Assumes\n# that executables *can* be run.\nac_fn_c_try_run ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n if { { ac_try=\"$ac_link\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_link\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; } && { ac_try='./conftest$ac_exeext'\n { { case \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_try\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }; }; then :\n ac_retval=0\nelse\n $as_echo \"$as_me: program exited with status $ac_status\" >&5\n $as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n ac_retval=$ac_status\nfi\n rm -rf conftest.dSYM conftest_ipa8_conftest.oo\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n as_fn_set_status $ac_retval\n\n} # ac_fn_c_try_run\n\n# ac_fn_c_compute_int LINENO EXPR VAR INCLUDES\n# --------------------------------------------\n# Tries to find the compile-time value of EXPR in a program that includes\n# INCLUDES, setting VAR accordingly. Returns whether the value could be\n# computed\nac_fn_c_compute_int ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n if test \"$cross_compiling\" = yes; then\n # Depending upon the size, compute the lo and hi bounds.\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nstatic int test_array [1 - 2 * !(($2) >= 0)];\ntest_array [0] = 0\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_lo=0 ac_mid=0\n while :; do\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nstatic int test_array [1 - 2 * !(($2) <= $ac_mid)];\ntest_array [0] = 0\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_hi=$ac_mid; break\nelse\n as_fn_arith $ac_mid + 1 && ac_lo=$as_val\n\t\t\tif test $ac_lo -le $ac_mid; then\n\t\t\t ac_lo= ac_hi=\n\t\t\t break\n\t\t\tfi\n\t\t\tas_fn_arith 2 '*' $ac_mid + 1 && ac_mid=$as_val\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n done\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nstatic int test_array [1 - 2 * !(($2) < 0)];\ntest_array [0] = 0\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_hi=-1 ac_mid=-1\n while :; do\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nstatic int test_array [1 - 2 * !(($2) >= $ac_mid)];\ntest_array [0] = 0\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_lo=$ac_mid; break\nelse\n as_fn_arith '(' $ac_mid ')' - 1 && ac_hi=$as_val\n\t\t\tif test $ac_mid -le $ac_hi; then\n\t\t\t ac_lo= ac_hi=\n\t\t\t break\n\t\t\tfi\n\t\t\tas_fn_arith 2 '*' $ac_mid && ac_mid=$as_val\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n done\nelse\n ac_lo= ac_hi=\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n# Binary search between lo and hi bounds.\nwhile test \"x$ac_lo\" != \"x$ac_hi\"; do\n as_fn_arith '(' $ac_hi - $ac_lo ')' / 2 + $ac_lo && ac_mid=$as_val\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nstatic int test_array [1 - 2 * !(($2) <= $ac_mid)];\ntest_array [0] = 0\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_hi=$ac_mid\nelse\n as_fn_arith '(' $ac_mid ')' + 1 && ac_lo=$as_val\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\ndone\ncase $ac_lo in #((\n?*) eval \"$3=\\$ac_lo\"; ac_retval=0 ;;\n'') ac_retval=1 ;;\nesac\n else\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nstatic long int longval () { return $2; }\nstatic unsigned long int ulongval () { return $2; }\n#include \n#include \nint\nmain ()\n{\n\n FILE *f = fopen (\"conftest.val\", \"w\");\n if (! f)\n return 1;\n if (($2) < 0)\n {\n long int i = longval ();\n if (i != ($2))\n\treturn 1;\n fprintf (f, \"%ld\", i);\n }\n else\n {\n unsigned long int i = ulongval ();\n if (i != ($2))\n\treturn 1;\n fprintf (f, \"%lu\", i);\n }\n /* Do not output a trailing newline, as this causes \\r\\n confusion\n on some platforms. */\n return ferror (f) || fclose (f) != 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_run \"$LINENO\"; then :\n echo >>conftest.val; read $3 &5\n$as_echo_n \"checking for $2... \" >&6; }\nif eval \\${$3+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\n#include <$2>\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n eval \"$3=yes\"\nelse\n eval \"$3=no\"\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\neval ac_res=\\$$3\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_res\" >&5\n$as_echo \"$ac_res\" >&6; }\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n\n} # ac_fn_c_check_header_compile\n\n# ac_fn_c_try_link LINENO\n# -----------------------\n# Try to link conftest.$ac_ext, and return whether this succeeded.\nac_fn_c_try_link ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n rm -f conftest.$ac_objext conftest$ac_exeext\n if { { ac_try=\"$ac_link\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_link\") 2>conftest.err\n ac_status=$?\n if test -s conftest.err; then\n grep -v '^ *+' conftest.err >conftest.er1\n cat conftest.er1 >&5\n mv -f conftest.er1 conftest.err\n fi\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; } && {\n\t test -z \"$ac_c_werror_flag\" ||\n\t test ! -s conftest.err\n } && test -s conftest$ac_exeext && {\n\t test \"$cross_compiling\" = yes ||\n\t $as_test_x conftest$ac_exeext\n }; then :\n ac_retval=0\nelse\n $as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n\tac_retval=1\nfi\n # Delete the IPA/IPO (Inter Procedural Analysis/Optimization) information\n # created by the PGI compiler (conftest_ipa8_conftest.oo), as it would\n # interfere with the next link command; also delete a directory that is\n # left behind by Apple's compiler. We do this before executing the actions.\n rm -rf conftest.dSYM conftest_ipa8_conftest.oo\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n as_fn_set_status $ac_retval\n\n} # ac_fn_c_try_link\n\n# ac_fn_c_check_func LINENO FUNC VAR\n# ----------------------------------\n# Tests whether FUNC exists, setting the cache variable VAR accordingly\nac_fn_c_check_func ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $2\" >&5\n$as_echo_n \"checking for $2... \" >&6; }\nif eval \\${$3+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n/* Define $2 to an innocuous variant, in case declares $2.\n For example, HP-UX 11i declares gettimeofday. */\n#define $2 innocuous_$2\n\n/* System header to define __stub macros and hopefully few prototypes,\n which can conflict with char $2 (); below.\n Prefer to if __STDC__ is defined, since\n exists even on freestanding compilers. */\n\n#ifdef __STDC__\n# include \n#else\n# include \n#endif\n\n#undef $2\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar $2 ();\n/* The GNU C library defines this for functions which it implements\n to always fail with ENOSYS. Some functions are actually named\n something starting with __ and the normal name is an alias. */\n#if defined __stub_$2 || defined __stub___$2\nchoke me\n#endif\n\nint\nmain ()\n{\nreturn $2 ();\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n eval \"$3=yes\"\nelse\n eval \"$3=no\"\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\neval ac_res=\\$$3\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_res\" >&5\n$as_echo \"$ac_res\" >&6; }\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n\n} # ac_fn_c_check_func\n\n# ac_fn_c_check_header_mongrel LINENO HEADER VAR INCLUDES\n# -------------------------------------------------------\n# Tests whether HEADER exists, giving a warning if it cannot be compiled using\n# the include files in INCLUDES and setting the cache variable VAR\n# accordingly.\nac_fn_c_check_header_mongrel ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n if eval \\${$3+:} false; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $2\" >&5\n$as_echo_n \"checking for $2... \" >&6; }\nif eval \\${$3+:} false; then :\n $as_echo_n \"(cached) \" >&6\nfi\neval ac_res=\\$$3\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_res\" >&5\n$as_echo \"$ac_res\" >&6; }\nelse\n # Is the header compilable?\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking $2 usability\" >&5\n$as_echo_n \"checking $2 usability... \" >&6; }\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\n#include <$2>\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_header_compiler=yes\nelse\n ac_header_compiler=no\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_header_compiler\" >&5\n$as_echo \"$ac_header_compiler\" >&6; }\n\n# Is the header present?\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking $2 presence\" >&5\n$as_echo_n \"checking $2 presence... \" >&6; }\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include <$2>\n_ACEOF\nif ac_fn_c_try_cpp \"$LINENO\"; then :\n ac_header_preproc=yes\nelse\n ac_header_preproc=no\nfi\nrm -f conftest.err conftest.i conftest.$ac_ext\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_header_preproc\" >&5\n$as_echo \"$ac_header_preproc\" >&6; }\n\n# So? What about this header?\ncase $ac_header_compiler:$ac_header_preproc:$ac_c_preproc_warn_flag in #((\n yes:no: )\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: accepted by the compiler, rejected by the preprocessor!\" >&5\n$as_echo \"$as_me: WARNING: $2: accepted by the compiler, rejected by the preprocessor!\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: proceeding with the compiler's result\" >&5\n$as_echo \"$as_me: WARNING: $2: proceeding with the compiler's result\" >&2;}\n ;;\n no:yes:* )\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: present but cannot be compiled\" >&5\n$as_echo \"$as_me: WARNING: $2: present but cannot be compiled\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: check for missing prerequisite headers?\" >&5\n$as_echo \"$as_me: WARNING: $2: check for missing prerequisite headers?\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: see the Autoconf documentation\" >&5\n$as_echo \"$as_me: WARNING: $2: see the Autoconf documentation\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: section \\\"Present But Cannot Be Compiled\\\"\" >&5\n$as_echo \"$as_me: WARNING: $2: section \\\"Present But Cannot Be Compiled\\\"\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $2: proceeding with the compiler's result\" >&5\n$as_echo \"$as_me: WARNING: $2: proceeding with the compiler's result\" >&2;}\n ;;\nesac\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $2\" >&5\n$as_echo_n \"checking for $2... \" >&6; }\nif eval \\${$3+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n eval \"$3=\\$ac_header_compiler\"\nfi\neval ac_res=\\$$3\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_res\" >&5\n$as_echo \"$ac_res\" >&6; }\nfi\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n\n} # ac_fn_c_check_header_mongrel\n\n# ac_fn_c_check_type LINENO TYPE VAR INCLUDES\n# -------------------------------------------\n# Tests whether TYPE exists after having included INCLUDES, setting cache\n# variable VAR accordingly.\nac_fn_c_check_type ()\n{\n as_lineno=${as_lineno-\"$1\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $2\" >&5\n$as_echo_n \"checking for $2... \" >&6; }\nif eval \\${$3+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n eval \"$3=no\"\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nif (sizeof ($2))\n\t return 0;\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n$4\nint\nmain ()\n{\nif (sizeof (($2)))\n\t return 0;\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n\nelse\n eval \"$3=yes\"\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\neval ac_res=\\$$3\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_res\" >&5\n$as_echo \"$ac_res\" >&6; }\n eval $as_lineno_stack; ${as_lineno_stack:+:} unset as_lineno\n\n} # ac_fn_c_check_type\ncat >config.log <<_ACEOF\nThis file contains any messages produced by compilers while\nrunning configure, to aid debugging if configure makes a mistake.\n\nIt was created by $as_me, which was\ngenerated by GNU Autoconf 2.68. Invocation command line was\n\n $ $0 $@\n\n_ACEOF\nexec 5>>config.log\n{\ncat <<_ASUNAME\n## --------- ##\n## Platform. ##\n## --------- ##\n\nhostname = `(hostname || uname -n) 2>/dev/null | sed 1q`\nuname -m = `(uname -m) 2>/dev/null || echo unknown`\nuname -r = `(uname -r) 2>/dev/null || echo unknown`\nuname -s = `(uname -s) 2>/dev/null || echo unknown`\nuname -v = `(uname -v) 2>/dev/null || echo unknown`\n\n/usr/bin/uname -p = `(/usr/bin/uname -p) 2>/dev/null || echo unknown`\n/bin/uname -X = `(/bin/uname -X) 2>/dev/null || echo unknown`\n\n/bin/arch = `(/bin/arch) 2>/dev/null || echo unknown`\n/usr/bin/arch -k = `(/usr/bin/arch -k) 2>/dev/null || echo unknown`\n/usr/convex/getsysinfo = `(/usr/convex/getsysinfo) 2>/dev/null || echo unknown`\n/usr/bin/hostinfo = `(/usr/bin/hostinfo) 2>/dev/null || echo unknown`\n/bin/machine = `(/bin/machine) 2>/dev/null || echo unknown`\n/usr/bin/oslevel = `(/usr/bin/oslevel) 2>/dev/null || echo unknown`\n/bin/universe = `(/bin/universe) 2>/dev/null || echo unknown`\n\n_ASUNAME\n\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n $as_echo \"PATH: $as_dir\"\n done\nIFS=$as_save_IFS\n\n} >&5\n\ncat >&5 <<_ACEOF\n\n\n## ----------- ##\n## Core tests. ##\n## ----------- ##\n\n_ACEOF\n\n\n# Keep a trace of the command line.\n# Strip out --no-create and --no-recursion so they do not pile up.\n# Strip out --silent because we don't want to record it for future runs.\n# Also quote any args containing shell meta-characters.\n# Make two passes to allow for proper duplicate-argument suppression.\nac_configure_args=\nac_configure_args0=\nac_configure_args1=\nac_must_keep_next=false\nfor ac_pass in 1 2\ndo\n for ac_arg\n do\n case $ac_arg in\n -no-create | --no-c* | -n | -no-recursion | --no-r*) continue ;;\n -q | -quiet | --quiet | --quie | --qui | --qu | --q \\\n | -silent | --silent | --silen | --sile | --sil)\n continue ;;\n *\\'*)\n ac_arg=`$as_echo \"$ac_arg\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"` ;;\n esac\n case $ac_pass in\n 1) as_fn_append ac_configure_args0 \" '$ac_arg'\" ;;\n 2)\n as_fn_append ac_configure_args1 \" '$ac_arg'\"\n if test $ac_must_keep_next = true; then\n\tac_must_keep_next=false # Got value, back to normal.\n else\n\tcase $ac_arg in\n\t *=* | --config-cache | -C | -disable-* | --disable-* \\\n\t | -enable-* | --enable-* | -gas | --g* | -nfp | --nf* \\\n\t | -q | -quiet | --q* | -silent | --sil* | -v | -verb* \\\n\t | -with-* | --with-* | -without-* | --without-* | --x)\n\t case \"$ac_configure_args0 \" in\n\t \"$ac_configure_args1\"*\" '$ac_arg' \"* ) continue ;;\n\t esac\n\t ;;\n\t -* ) ac_must_keep_next=true ;;\n\tesac\n fi\n as_fn_append ac_configure_args \" '$ac_arg'\"\n ;;\n esac\n done\ndone\n{ ac_configure_args0=; unset ac_configure_args0;}\n{ ac_configure_args1=; unset ac_configure_args1;}\n\n# When interrupted or exit'd, cleanup temporary files, and complete\n# config.log. We remove comments because anyway the quotes in there\n# would cause problems or look ugly.\n# WARNING: Use '\\'' to represent an apostrophe within the trap.\n# WARNING: Do not start the trap code with a newline, due to a FreeBSD 4.0 bug.\ntrap 'exit_status=$?\n # Save into config.log some information that might help in debugging.\n {\n echo\n\n $as_echo \"## ---------------- ##\n## Cache variables. ##\n## ---------------- ##\"\n echo\n # The following way of writing the cache mishandles newlines in values,\n(\n for ac_var in `(set) 2>&1 | sed -n '\\''s/^\\([a-zA-Z_][a-zA-Z0-9_]*\\)=.*/\\1/p'\\''`; do\n eval ac_val=\\$$ac_var\n case $ac_val in #(\n *${as_nl}*)\n case $ac_var in #(\n *_cv_*) { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: cache variable $ac_var contains a newline\" >&5\n$as_echo \"$as_me: WARNING: cache variable $ac_var contains a newline\" >&2;} ;;\n esac\n case $ac_var in #(\n _ | IFS | as_nl) ;; #(\n BASH_ARGV | BASH_SOURCE) eval $ac_var= ;; #(\n *) { eval $ac_var=; unset $ac_var;} ;;\n esac ;;\n esac\n done\n (set) 2>&1 |\n case $as_nl`(ac_space='\\'' '\\''; set) 2>&1` in #(\n *${as_nl}ac_space=\\ *)\n sed -n \\\n\t\"s/'\\''/'\\''\\\\\\\\'\\'''\\''/g;\n\t s/^\\\\([_$as_cr_alnum]*_cv_[_$as_cr_alnum]*\\\\)=\\\\(.*\\\\)/\\\\1='\\''\\\\2'\\''/p\"\n ;; #(\n *)\n sed -n \"/^[_$as_cr_alnum]*_cv_[_$as_cr_alnum]*=/p\"\n ;;\n esac |\n sort\n)\n echo\n\n $as_echo \"## ----------------- ##\n## Output variables. ##\n## ----------------- ##\"\n echo\n for ac_var in $ac_subst_vars\n do\n eval ac_val=\\$$ac_var\n case $ac_val in\n *\\'\\''*) ac_val=`$as_echo \"$ac_val\" | sed \"s/'\\''/'\\''\\\\\\\\\\\\\\\\'\\'''\\''/g\"`;;\n esac\n $as_echo \"$ac_var='\\''$ac_val'\\''\"\n done | sort\n echo\n\n if test -n \"$ac_subst_files\"; then\n $as_echo \"## ------------------- ##\n## File substitutions. ##\n## ------------------- ##\"\n echo\n for ac_var in $ac_subst_files\n do\n\teval ac_val=\\$$ac_var\n\tcase $ac_val in\n\t*\\'\\''*) ac_val=`$as_echo \"$ac_val\" | sed \"s/'\\''/'\\''\\\\\\\\\\\\\\\\'\\'''\\''/g\"`;;\n\tesac\n\t$as_echo \"$ac_var='\\''$ac_val'\\''\"\n done | sort\n echo\n fi\n\n if test -s confdefs.h; then\n $as_echo \"## ----------- ##\n## confdefs.h. ##\n## ----------- ##\"\n echo\n cat confdefs.h\n echo\n fi\n test \"$ac_signal\" != 0 &&\n $as_echo \"$as_me: caught signal $ac_signal\"\n $as_echo \"$as_me: exit $exit_status\"\n } >&5\n rm -f core *.core core.conftest.* &&\n rm -f -r conftest* confdefs* conf$$* $ac_clean_files &&\n exit $exit_status\n' 0\nfor ac_signal in 1 2 13 15; do\n trap 'ac_signal='$ac_signal'; as_fn_exit 1' $ac_signal\ndone\nac_signal=0\n\n# confdefs.h avoids OS command line length limits that DEFS can exceed.\nrm -f -r conftest* confdefs.h\n\n$as_echo \"/* confdefs.h */\" > confdefs.h\n\n# Predefined preprocessor variables.\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_NAME \"$PACKAGE_NAME\"\n_ACEOF\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_TARNAME \"$PACKAGE_TARNAME\"\n_ACEOF\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_VERSION \"$PACKAGE_VERSION\"\n_ACEOF\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_STRING \"$PACKAGE_STRING\"\n_ACEOF\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_BUGREPORT \"$PACKAGE_BUGREPORT\"\n_ACEOF\n\ncat >>confdefs.h <<_ACEOF\n#define PACKAGE_URL \"$PACKAGE_URL\"\n_ACEOF\n\n\n# Let the site file select an alternate cache file if it wants to.\n# Prefer an explicitly selected file to automatically selected ones.\nac_site_file1=NONE\nac_site_file2=NONE\nif test -n \"$CONFIG_SITE\"; then\n # We do not want a PATH search for config.site.\n case $CONFIG_SITE in #((\n -*) ac_site_file1=./$CONFIG_SITE;;\n */*) ac_site_file1=$CONFIG_SITE;;\n *) ac_site_file1=./$CONFIG_SITE;;\n esac\nelif test \"x$prefix\" != xNONE; then\n ac_site_file1=$prefix/share/config.site\n ac_site_file2=$prefix/etc/config.site\nelse\n ac_site_file1=$ac_default_prefix/share/config.site\n ac_site_file2=$ac_default_prefix/etc/config.site\nfi\nfor ac_site_file in \"$ac_site_file1\" \"$ac_site_file2\"\ndo\n test \"x$ac_site_file\" = xNONE && continue\n if test /dev/null != \"$ac_site_file\" && test -r \"$ac_site_file\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: loading site script $ac_site_file\" >&5\n$as_echo \"$as_me: loading site script $ac_site_file\" >&6;}\n sed 's/^/| /' \"$ac_site_file\" >&5\n . \"$ac_site_file\" \\\n || { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"failed to load site script $ac_site_file\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n fi\ndone\n\nif test -r \"$cache_file\"; then\n # Some versions of bash will fail to source /dev/null (special files\n # actually), so we avoid doing that. DJGPP emulates it as a regular file.\n if test /dev/null != \"$cache_file\" && test -f \"$cache_file\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: loading cache $cache_file\" >&5\n$as_echo \"$as_me: loading cache $cache_file\" >&6;}\n case $cache_file in\n [\\\\/]* | ?:[\\\\/]* ) . \"$cache_file\";;\n *) . \"./$cache_file\";;\n esac\n fi\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: creating cache $cache_file\" >&5\n$as_echo \"$as_me: creating cache $cache_file\" >&6;}\n >$cache_file\nfi\n\n# Check that the precious variables saved in the cache have kept the same\n# value.\nac_cache_corrupted=false\nfor ac_var in $ac_precious_vars; do\n eval ac_old_set=\\$ac_cv_env_${ac_var}_set\n eval ac_new_set=\\$ac_env_${ac_var}_set\n eval ac_old_val=\\$ac_cv_env_${ac_var}_value\n eval ac_new_val=\\$ac_env_${ac_var}_value\n case $ac_old_set,$ac_new_set in\n set,)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: error: \\`$ac_var' was set to \\`$ac_old_val' in the previous run\" >&5\n$as_echo \"$as_me: error: \\`$ac_var' was set to \\`$ac_old_val' in the previous run\" >&2;}\n ac_cache_corrupted=: ;;\n ,set)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: error: \\`$ac_var' was not set in the previous run\" >&5\n$as_echo \"$as_me: error: \\`$ac_var' was not set in the previous run\" >&2;}\n ac_cache_corrupted=: ;;\n ,);;\n *)\n if test \"x$ac_old_val\" != \"x$ac_new_val\"; then\n\t# differences in whitespace do not lead to failure.\n\tac_old_val_w=`echo x $ac_old_val`\n\tac_new_val_w=`echo x $ac_new_val`\n\tif test \"$ac_old_val_w\" != \"$ac_new_val_w\"; then\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: error: \\`$ac_var' has changed since the previous run:\" >&5\n$as_echo \"$as_me: error: \\`$ac_var' has changed since the previous run:\" >&2;}\n\t ac_cache_corrupted=:\n\telse\n\t { $as_echo \"$as_me:${as_lineno-$LINENO}: warning: ignoring whitespace changes in \\`$ac_var' since the previous run:\" >&5\n$as_echo \"$as_me: warning: ignoring whitespace changes in \\`$ac_var' since the previous run:\" >&2;}\n\t eval $ac_var=\\$ac_old_val\n\tfi\n\t{ $as_echo \"$as_me:${as_lineno-$LINENO}: former value: \\`$ac_old_val'\" >&5\n$as_echo \"$as_me: former value: \\`$ac_old_val'\" >&2;}\n\t{ $as_echo \"$as_me:${as_lineno-$LINENO}: current value: \\`$ac_new_val'\" >&5\n$as_echo \"$as_me: current value: \\`$ac_new_val'\" >&2;}\n fi;;\n esac\n # Pass precious variables to config.status.\n if test \"$ac_new_set\" = set; then\n case $ac_new_val in\n *\\'*) ac_arg=$ac_var=`$as_echo \"$ac_new_val\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"` ;;\n *) ac_arg=$ac_var=$ac_new_val ;;\n esac\n case \" $ac_configure_args \" in\n *\" '$ac_arg' \"*) ;; # Avoid dups. Use of quotes ensures accuracy.\n *) as_fn_append ac_configure_args \" '$ac_arg'\" ;;\n esac\n fi\ndone\nif $ac_cache_corrupted; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\n { $as_echo \"$as_me:${as_lineno-$LINENO}: error: changes in the environment can compromise the build\" >&5\n$as_echo \"$as_me: error: changes in the environment can compromise the build\" >&2;}\n as_fn_error $? \"run \\`make distclean' and/or \\`rm $cache_file' and start over\" \"$LINENO\" 5\nfi\n## -------------------- ##\n## Main body of script. ##\n## -------------------- ##\n\nac_ext=c\nac_cpp='$CPP $CPPFLAGS'\nac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'\nac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'\nac_compiler_gnu=$ac_cv_c_compiler_gnu\n\n\n\n\n\n\n\n\n\nrev=1\n\n\nsrcroot=$srcdir\nif test \"x${srcroot}\" = \"x.\" ; then\n srcroot=\"\"\nelse\n srcroot=\"${srcroot}/\"\nfi\n\nabs_srcroot=\"`cd \\\"${srcdir}\\\"; pwd`/\"\n\n\nobjroot=\"\"\n\nabs_objroot=\"`pwd`/\"\n\n\nif test \"x$prefix\" = \"xNONE\" ; then\n prefix=\"/usr/local\"\nfi\nif test \"x$exec_prefix\" = \"xNONE\" ; then\n exec_prefix=$prefix\nfi\nPREFIX=$prefix\n\nBINDIR=`eval echo $bindir`\nBINDIR=`eval echo $BINDIR`\n\nINCLUDEDIR=`eval echo $includedir`\nINCLUDEDIR=`eval echo $INCLUDEDIR`\n\nLIBDIR=`eval echo $libdir`\nLIBDIR=`eval echo $LIBDIR`\n\nDATADIR=`eval echo $datadir`\nDATADIR=`eval echo $DATADIR`\n\nMANDIR=`eval echo $mandir`\nMANDIR=`eval echo $MANDIR`\n\n\n# Extract the first word of \"xsltproc\", so it can be a program name with args.\nset dummy xsltproc; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_path_XSLTPROC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n case $XSLTPROC in\n [\\\\/]* | ?:[\\\\/]*)\n ac_cv_path_XSLTPROC=\"$XSLTPROC\" # Let the user override the test with a path.\n ;;\n *)\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_path_XSLTPROC=\"$as_dir/$ac_word$ac_exec_ext\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\n ;;\nesac\nfi\nXSLTPROC=$ac_cv_path_XSLTPROC\nif test -n \"$XSLTPROC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $XSLTPROC\" >&5\n$as_echo \"$XSLTPROC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\nif test -d \"/usr/share/xml/docbook/stylesheet/docbook-xsl\" ; then\n DEFAULT_XSLROOT=\"/usr/share/xml/docbook/stylesheet/docbook-xsl\"\nelif test -d \"/usr/share/sgml/docbook/xsl-stylesheets\" ; then\n DEFAULT_XSLROOT=\"/usr/share/sgml/docbook/xsl-stylesheets\"\nelse\n DEFAULT_XSLROOT=\"\"\nfi\n\n# Check whether --with-xslroot was given.\nif test \"${with_xslroot+set}\" = set; then :\n withval=$with_xslroot;\nif test \"x$with_xslroot\" = \"xno\" ; then\n XSLROOT=\"${DEFAULT_XSLROOT}\"\nelse\n XSLROOT=\"${with_xslroot}\"\nfi\n\nelse\n XSLROOT=\"${DEFAULT_XSLROOT}\"\n\nfi\n\n\n\nCFLAGS=$CFLAGS\nac_ext=c\nac_cpp='$CPP $CPPFLAGS'\nac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'\nac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'\nac_compiler_gnu=$ac_cv_c_compiler_gnu\nif test -n \"$ac_tool_prefix\"; then\n # Extract the first word of \"${ac_tool_prefix}gcc\", so it can be a program name with args.\nset dummy ${ac_tool_prefix}gcc; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$CC\"; then\n ac_cv_prog_CC=\"$CC\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_CC=\"${ac_tool_prefix}gcc\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nCC=$ac_cv_prog_CC\nif test -n \"$CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $CC\" >&5\n$as_echo \"$CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\nfi\nif test -z \"$ac_cv_prog_CC\"; then\n ac_ct_CC=$CC\n # Extract the first word of \"gcc\", so it can be a program name with args.\nset dummy gcc; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_ac_ct_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$ac_ct_CC\"; then\n ac_cv_prog_ac_ct_CC=\"$ac_ct_CC\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_ac_ct_CC=\"gcc\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nac_ct_CC=$ac_cv_prog_ac_ct_CC\nif test -n \"$ac_ct_CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_ct_CC\" >&5\n$as_echo \"$ac_ct_CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n if test \"x$ac_ct_CC\" = x; then\n CC=\"\"\n else\n case $cross_compiling:$ac_tool_warned in\nyes:)\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: using cross tools not prefixed with host triplet\" >&5\n$as_echo \"$as_me: WARNING: using cross tools not prefixed with host triplet\" >&2;}\nac_tool_warned=yes ;;\nesac\n CC=$ac_ct_CC\n fi\nelse\n CC=\"$ac_cv_prog_CC\"\nfi\n\nif test -z \"$CC\"; then\n if test -n \"$ac_tool_prefix\"; then\n # Extract the first word of \"${ac_tool_prefix}cc\", so it can be a program name with args.\nset dummy ${ac_tool_prefix}cc; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$CC\"; then\n ac_cv_prog_CC=\"$CC\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_CC=\"${ac_tool_prefix}cc\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nCC=$ac_cv_prog_CC\nif test -n \"$CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $CC\" >&5\n$as_echo \"$CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n fi\nfi\nif test -z \"$CC\"; then\n # Extract the first word of \"cc\", so it can be a program name with args.\nset dummy cc; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$CC\"; then\n ac_cv_prog_CC=\"$CC\" # Let the user override the test.\nelse\n ac_prog_rejected=no\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n if test \"$as_dir/$ac_word$ac_exec_ext\" = \"/usr/ucb/cc\"; then\n ac_prog_rejected=yes\n continue\n fi\n ac_cv_prog_CC=\"cc\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nif test $ac_prog_rejected = yes; then\n # We found a bogon in the path, so make sure we never use it.\n set dummy $ac_cv_prog_CC\n shift\n if test $# != 0; then\n # We chose a different compiler from the bogus one.\n # However, it has the same basename, so the bogon will be chosen\n # first if we set CC to just the basename; use the full file name.\n shift\n ac_cv_prog_CC=\"$as_dir/$ac_word${1+' '}$@\"\n fi\nfi\nfi\nfi\nCC=$ac_cv_prog_CC\nif test -n \"$CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $CC\" >&5\n$as_echo \"$CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\nfi\nif test -z \"$CC\"; then\n if test -n \"$ac_tool_prefix\"; then\n for ac_prog in cl.exe\n do\n # Extract the first word of \"$ac_tool_prefix$ac_prog\", so it can be a program name with args.\nset dummy $ac_tool_prefix$ac_prog; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$CC\"; then\n ac_cv_prog_CC=\"$CC\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_CC=\"$ac_tool_prefix$ac_prog\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nCC=$ac_cv_prog_CC\nif test -n \"$CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $CC\" >&5\n$as_echo \"$CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n test -n \"$CC\" && break\n done\nfi\nif test -z \"$CC\"; then\n ac_ct_CC=$CC\n for ac_prog in cl.exe\ndo\n # Extract the first word of \"$ac_prog\", so it can be a program name with args.\nset dummy $ac_prog; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_ac_ct_CC+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$ac_ct_CC\"; then\n ac_cv_prog_ac_ct_CC=\"$ac_ct_CC\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_ac_ct_CC=\"$ac_prog\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nac_ct_CC=$ac_cv_prog_ac_ct_CC\nif test -n \"$ac_ct_CC\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_ct_CC\" >&5\n$as_echo \"$ac_ct_CC\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n test -n \"$ac_ct_CC\" && break\ndone\n\n if test \"x$ac_ct_CC\" = x; then\n CC=\"\"\n else\n case $cross_compiling:$ac_tool_warned in\nyes:)\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: using cross tools not prefixed with host triplet\" >&5\n$as_echo \"$as_me: WARNING: using cross tools not prefixed with host triplet\" >&2;}\nac_tool_warned=yes ;;\nesac\n CC=$ac_ct_CC\n fi\nfi\n\nfi\n\n\ntest -z \"$CC\" && { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"no acceptable C compiler found in \\$PATH\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n\n# Provide some information about the compiler.\n$as_echo \"$as_me:${as_lineno-$LINENO}: checking for C compiler version\" >&5\nset X $ac_compile\nac_compiler=$2\nfor ac_option in --version -v -V -qversion; do\n { { ac_try=\"$ac_compiler $ac_option >&5\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_compiler $ac_option >&5\") 2>conftest.err\n ac_status=$?\n if test -s conftest.err; then\n sed '10a\\\n... rest of stderr output deleted ...\n 10q' conftest.err >conftest.er1\n cat conftest.er1 >&5\n fi\n rm -f conftest.er1 conftest.err\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }\ndone\n\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nac_clean_files_save=$ac_clean_files\nac_clean_files=\"$ac_clean_files a.out a.out.dSYM a.exe b.out\"\n# Try to create an executable without -o first, disregard a.out.\n# It will help us diagnose broken compilers, and finding out an intuition\n# of exeext.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether the C compiler works\" >&5\n$as_echo_n \"checking whether the C compiler works... \" >&6; }\nac_link_default=`$as_echo \"$ac_link\" | sed 's/ -o *conftest[^ ]*//'`\n\n# The possible output files:\nac_files=\"a.out conftest.exe conftest a.exe a_out.exe b.out conftest.*\"\n\nac_rmfiles=\nfor ac_file in $ac_files\ndo\n case $ac_file in\n *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.dSYM | *.o | *.obj ) ;;\n * ) ac_rmfiles=\"$ac_rmfiles $ac_file\";;\n esac\ndone\nrm -f $ac_rmfiles\n\nif { { ac_try=\"$ac_link_default\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_link_default\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }; then :\n # Autoconf-2.13 could set the ac_cv_exeext variable to `no'.\n# So ignore a value of `no', otherwise this would lead to `EXEEXT = no'\n# in a Makefile. We should not override ac_cv_exeext if it was cached,\n# so that the user can short-circuit this test for compilers unknown to\n# Autoconf.\nfor ac_file in $ac_files ''\ndo\n test -f \"$ac_file\" || continue\n case $ac_file in\n *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.dSYM | *.o | *.obj )\n\t;;\n [ab].out )\n\t# We found the default executable, but exeext='' is most\n\t# certainly right.\n\tbreak;;\n *.* )\n\tif test \"${ac_cv_exeext+set}\" = set && test \"$ac_cv_exeext\" != no;\n\tthen :; else\n\t ac_cv_exeext=`expr \"$ac_file\" : '[^.]*\\(\\..*\\)'`\n\tfi\n\t# We set ac_cv_exeext here because the later test for it is not\n\t# safe: cross compilers may not add the suffix if given an `-o'\n\t# argument, so we may need to know it at that point already.\n\t# Even if this section looks crufty: it has the advantage of\n\t# actually working.\n\tbreak;;\n * )\n\tbreak;;\n esac\ndone\ntest \"$ac_cv_exeext\" = no && ac_cv_exeext=\n\nelse\n ac_file=''\nfi\nif test -z \"$ac_file\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n$as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n{ { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error 77 \"C compiler cannot create executables\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for C compiler default output file name\" >&5\n$as_echo_n \"checking for C compiler default output file name... \" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_file\" >&5\n$as_echo \"$ac_file\" >&6; }\nac_exeext=$ac_cv_exeext\n\nrm -f -r a.out a.out.dSYM a.exe conftest$ac_cv_exeext b.out\nac_clean_files=$ac_clean_files_save\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for suffix of executables\" >&5\n$as_echo_n \"checking for suffix of executables... \" >&6; }\nif { { ac_try=\"$ac_link\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_link\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }; then :\n # If both `conftest.exe' and `conftest' are `present' (well, observable)\n# catch `conftest.exe'. For instance with Cygwin, `ls conftest' will\n# work properly (i.e., refer to `conftest.exe'), while it won't with\n# `rm'.\nfor ac_file in conftest.exe conftest conftest.*; do\n test -f \"$ac_file\" || continue\n case $ac_file in\n *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.dSYM | *.o | *.obj ) ;;\n *.* ) ac_cv_exeext=`expr \"$ac_file\" : '[^.]*\\(\\..*\\)'`\n\t break;;\n * ) break;;\n esac\ndone\nelse\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"cannot compute suffix of executables: cannot compile and link\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\nfi\nrm -f conftest conftest$ac_cv_exeext\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_exeext\" >&5\n$as_echo \"$ac_cv_exeext\" >&6; }\n\nrm -f conftest.$ac_ext\nEXEEXT=$ac_cv_exeext\nac_exeext=$EXEEXT\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nFILE *f = fopen (\"conftest.out\", \"w\");\n return ferror (f) || fclose (f) != 0;\n\n ;\n return 0;\n}\n_ACEOF\nac_clean_files=\"$ac_clean_files conftest.out\"\n# Check that the compiler produces executables we can run. If not, either\n# the compiler is broken, or we cross compile.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether we are cross compiling\" >&5\n$as_echo_n \"checking whether we are cross compiling... \" >&6; }\nif test \"$cross_compiling\" != yes; then\n { { ac_try=\"$ac_link\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_link\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }\n if { ac_try='./conftest$ac_cv_exeext'\n { { case \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_try\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }; }; then\n cross_compiling=no\n else\n if test \"$cross_compiling\" = maybe; then\n\tcross_compiling=yes\n else\n\t{ { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"cannot run C compiled programs.\nIf you meant to cross compile, use \\`--host'.\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n fi\n fi\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $cross_compiling\" >&5\n$as_echo \"$cross_compiling\" >&6; }\n\nrm -f conftest.$ac_ext conftest$ac_cv_exeext conftest.out\nac_clean_files=$ac_clean_files_save\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for suffix of object files\" >&5\n$as_echo_n \"checking for suffix of object files... \" >&6; }\nif ${ac_cv_objext+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nrm -f conftest.o conftest.obj\nif { { ac_try=\"$ac_compile\"\ncase \"(($ac_try\" in\n *\\\"* | *\\`* | *\\\\*) ac_try_echo=\\$ac_try;;\n *) ac_try_echo=$ac_try;;\nesac\neval ac_try_echo=\"\\\"\\$as_me:${as_lineno-$LINENO}: $ac_try_echo\\\"\"\n$as_echo \"$ac_try_echo\"; } >&5\n (eval \"$ac_compile\") 2>&5\n ac_status=$?\n $as_echo \"$as_me:${as_lineno-$LINENO}: \\$? = $ac_status\" >&5\n test $ac_status = 0; }; then :\n for ac_file in conftest.o conftest.obj conftest.*; do\n test -f \"$ac_file\" || continue;\n case $ac_file in\n *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.dSYM ) ;;\n *) ac_cv_objext=`expr \"$ac_file\" : '.*\\.\\(.*\\)'`\n break;;\n esac\ndone\nelse\n $as_echo \"$as_me: failed program was:\" >&5\nsed 's/^/| /' conftest.$ac_ext >&5\n\n{ { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"cannot compute suffix of object files: cannot compile\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\nfi\nrm -f conftest.$ac_cv_objext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_objext\" >&5\n$as_echo \"$ac_cv_objext\" >&6; }\nOBJEXT=$ac_cv_objext\nac_objext=$OBJEXT\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether we are using the GNU C compiler\" >&5\n$as_echo_n \"checking whether we are using the GNU C compiler... \" >&6; }\nif ${ac_cv_c_compiler_gnu+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n#ifndef __GNUC__\n choke me\n#endif\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_compiler_gnu=yes\nelse\n ac_compiler_gnu=no\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nac_cv_c_compiler_gnu=$ac_compiler_gnu\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_c_compiler_gnu\" >&5\n$as_echo \"$ac_cv_c_compiler_gnu\" >&6; }\nif test $ac_compiler_gnu = yes; then\n GCC=yes\nelse\n GCC=\nfi\nac_test_CFLAGS=${CFLAGS+set}\nac_save_CFLAGS=$CFLAGS\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether $CC accepts -g\" >&5\n$as_echo_n \"checking whether $CC accepts -g... \" >&6; }\nif ${ac_cv_prog_cc_g+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_save_c_werror_flag=$ac_c_werror_flag\n ac_c_werror_flag=yes\n ac_cv_prog_cc_g=no\n CFLAGS=\"-g\"\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_cv_prog_cc_g=yes\nelse\n CFLAGS=\"\"\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n\nelse\n ac_c_werror_flag=$ac_save_c_werror_flag\n\t CFLAGS=\"-g\"\n\t cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_cv_prog_cc_g=yes\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n ac_c_werror_flag=$ac_save_c_werror_flag\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_prog_cc_g\" >&5\n$as_echo \"$ac_cv_prog_cc_g\" >&6; }\nif test \"$ac_test_CFLAGS\" = set; then\n CFLAGS=$ac_save_CFLAGS\nelif test $ac_cv_prog_cc_g = yes; then\n if test \"$GCC\" = yes; then\n CFLAGS=\"-g -O2\"\n else\n CFLAGS=\"-g\"\n fi\nelse\n if test \"$GCC\" = yes; then\n CFLAGS=\"-O2\"\n else\n CFLAGS=\n fi\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $CC option to accept ISO C89\" >&5\n$as_echo_n \"checking for $CC option to accept ISO C89... \" >&6; }\nif ${ac_cv_prog_cc_c89+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_cv_prog_cc_c89=no\nac_save_CC=$CC\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n#include \n#include \n#include \n/* Most of the following tests are stolen from RCS 5.7's src/conf.sh. */\nstruct buf { int x; };\nFILE * (*rcsopen) (struct buf *, struct stat *, int);\nstatic char *e (p, i)\n char **p;\n int i;\n{\n return p[i];\n}\nstatic char *f (char * (*g) (char **, int), char **p, ...)\n{\n char *s;\n va_list v;\n va_start (v,p);\n s = g (p, va_arg (v,int));\n va_end (v);\n return s;\n}\n\n/* OSF 4.0 Compaq cc is some sort of almost-ANSI by default. It has\n function prototypes and stuff, but not '\\xHH' hex character constants.\n These don't provoke an error unfortunately, instead are silently treated\n as 'x'. The following induces an error, until -std is added to get\n proper ANSI mode. Curiously '\\x00'!='x' always comes out true, for an\n array size at least. It's necessary to write '\\x00'==0 to get something\n that's true only with -std. */\nint osf4_cc_array ['\\x00' == 0 ? 1 : -1];\n\n/* IBM C 6 for AIX is almost-ANSI by default, but it replaces macro parameters\n inside strings and character constants. */\n#define FOO(x) 'x'\nint xlc6_cc_array[FOO(a) == 'x' ? 1 : -1];\n\nint test (int i, double x);\nstruct s1 {int (*f) (int a);};\nstruct s2 {int (*f) (double a);};\nint pairnames (int, char **, FILE *(*)(struct buf *, struct stat *, int), int, int);\nint argc;\nchar **argv;\nint\nmain ()\n{\nreturn f (e, argv, 0) != argv[0] || f (e, argv, 1) != argv[1];\n ;\n return 0;\n}\n_ACEOF\nfor ac_arg in '' -qlanglvl=extc89 -qlanglvl=ansi -std \\\n\t-Ae \"-Aa -D_HPUX_SOURCE\" \"-Xc -D__EXTENSIONS__\"\ndo\n CC=\"$ac_save_CC $ac_arg\"\n if ac_fn_c_try_compile \"$LINENO\"; then :\n ac_cv_prog_cc_c89=$ac_arg\nfi\nrm -f core conftest.err conftest.$ac_objext\n test \"x$ac_cv_prog_cc_c89\" != \"xno\" && break\ndone\nrm -f conftest.$ac_ext\nCC=$ac_save_CC\n\nfi\n# AC_CACHE_VAL\ncase \"x$ac_cv_prog_cc_c89\" in\n x)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: none needed\" >&5\n$as_echo \"none needed\" >&6; } ;;\n xno)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: unsupported\" >&5\n$as_echo \"unsupported\" >&6; } ;;\n *)\n CC=\"$CC $ac_cv_prog_cc_c89\"\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_prog_cc_c89\" >&5\n$as_echo \"$ac_cv_prog_cc_c89\" >&6; } ;;\nesac\nif test \"x$ac_cv_prog_cc_c89\" != xno; then :\n\nfi\n\nac_ext=c\nac_cpp='$CPP $CPPFLAGS'\nac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'\nac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'\nac_compiler_gnu=$ac_cv_c_compiler_gnu\n\nif test \"x$GCC\" != \"xyes\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler is MSVC\" >&5\n$as_echo_n \"checking whether compiler is MSVC... \" >&6; }\nif ${je_cv_msvc+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n\n#ifndef _MSC_VER\n int fail-1;\n#endif\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n je_cv_msvc=yes\nelse\n je_cv_msvc=no\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_msvc\" >&5\n$as_echo \"$je_cv_msvc\" >&6; }\nfi\n\nif test \"x$CFLAGS\" = \"x\" ; then\n no_CFLAGS=\"yes\"\n if test \"x$GCC\" = \"xyes\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -std=gnu99\" >&5\n$as_echo_n \"checking whether compiler supports -std=gnu99... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-std=gnu99\"\nelse\n CFLAGS=\"${CFLAGS} -std=gnu99\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -Wall\" >&5\n$as_echo_n \"checking whether compiler supports -Wall... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-Wall\"\nelse\n CFLAGS=\"${CFLAGS} -Wall\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -pipe\" >&5\n$as_echo_n \"checking whether compiler supports -pipe... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-pipe\"\nelse\n CFLAGS=\"${CFLAGS} -pipe\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -g3\" >&5\n$as_echo_n \"checking whether compiler supports -g3... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-g3\"\nelse\n CFLAGS=\"${CFLAGS} -g3\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n elif test \"x$je_cv_msvc\" = \"xyes\" ; then\n CC=\"$CC -nologo\"\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -Zi\" >&5\n$as_echo_n \"checking whether compiler supports -Zi... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-Zi\"\nelse\n CFLAGS=\"${CFLAGS} -Zi\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -MT\" >&5\n$as_echo_n \"checking whether compiler supports -MT... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-MT\"\nelse\n CFLAGS=\"${CFLAGS} -MT\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -W3\" >&5\n$as_echo_n \"checking whether compiler supports -W3... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-W3\"\nelse\n CFLAGS=\"${CFLAGS} -W3\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n CPPFLAGS=\"$CPPFLAGS -I${srcroot}/include/msvc_compat\"\n fi\nfi\nif test \"x$EXTRA_CFLAGS\" != \"x\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports $EXTRA_CFLAGS\" >&5\n$as_echo_n \"checking whether compiler supports $EXTRA_CFLAGS... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"$EXTRA_CFLAGS\"\nelse\n CFLAGS=\"${CFLAGS} $EXTRA_CFLAGS\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\nfi\nac_ext=c\nac_cpp='$CPP $CPPFLAGS'\nac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'\nac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'\nac_compiler_gnu=$ac_cv_c_compiler_gnu\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking how to run the C preprocessor\" >&5\n$as_echo_n \"checking how to run the C preprocessor... \" >&6; }\n# On Suns, sometimes $CPP names a directory.\nif test -n \"$CPP\" && test -d \"$CPP\"; then\n CPP=\nfi\nif test -z \"$CPP\"; then\n if ${ac_cv_prog_CPP+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n # Double quotes because CPP needs to be expanded\n for CPP in \"$CC -E\" \"$CC -E -traditional-cpp\" \"/lib/cpp\"\n do\n ac_preproc_ok=false\nfor ac_c_preproc_warn_flag in '' yes\ndo\n # Use a header file that comes with gcc, so configuring glibc\n # with a fresh cross-compiler works.\n # Prefer to if __STDC__ is defined, since\n # exists even on freestanding compilers.\n # On the NeXT, cc -E runs the code through the compiler's parser,\n # not just through cpp. \"Syntax error\" is here to catch this case.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#ifdef __STDC__\n# include \n#else\n# include \n#endif\n\t\t Syntax error\n_ACEOF\nif ac_fn_c_try_cpp \"$LINENO\"; then :\n\nelse\n # Broken: fails on valid input.\ncontinue\nfi\nrm -f conftest.err conftest.i conftest.$ac_ext\n\n # OK, works on sane cases. Now check whether nonexistent headers\n # can be detected and how.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n_ACEOF\nif ac_fn_c_try_cpp \"$LINENO\"; then :\n # Broken: success on invalid input.\ncontinue\nelse\n # Passes both tests.\nac_preproc_ok=:\nbreak\nfi\nrm -f conftest.err conftest.i conftest.$ac_ext\n\ndone\n# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.\nrm -f conftest.i conftest.err conftest.$ac_ext\nif $ac_preproc_ok; then :\n break\nfi\n\n done\n ac_cv_prog_CPP=$CPP\n\nfi\n CPP=$ac_cv_prog_CPP\nelse\n ac_cv_prog_CPP=$CPP\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $CPP\" >&5\n$as_echo \"$CPP\" >&6; }\nac_preproc_ok=false\nfor ac_c_preproc_warn_flag in '' yes\ndo\n # Use a header file that comes with gcc, so configuring glibc\n # with a fresh cross-compiler works.\n # Prefer to if __STDC__ is defined, since\n # exists even on freestanding compilers.\n # On the NeXT, cc -E runs the code through the compiler's parser,\n # not just through cpp. \"Syntax error\" is here to catch this case.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#ifdef __STDC__\n# include \n#else\n# include \n#endif\n\t\t Syntax error\n_ACEOF\nif ac_fn_c_try_cpp \"$LINENO\"; then :\n\nelse\n # Broken: fails on valid input.\ncontinue\nfi\nrm -f conftest.err conftest.i conftest.$ac_ext\n\n # OK, works on sane cases. Now check whether nonexistent headers\n # can be detected and how.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n_ACEOF\nif ac_fn_c_try_cpp \"$LINENO\"; then :\n # Broken: success on invalid input.\ncontinue\nelse\n # Passes both tests.\nac_preproc_ok=:\nbreak\nfi\nrm -f conftest.err conftest.i conftest.$ac_ext\n\ndone\n# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.\nrm -f conftest.i conftest.err conftest.$ac_ext\nif $ac_preproc_ok; then :\n\nelse\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"C preprocessor \\\"$CPP\\\" fails sanity check\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\nfi\n\nac_ext=c\nac_cpp='$CPP $CPPFLAGS'\nac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'\nac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'\nac_compiler_gnu=$ac_cv_c_compiler_gnu\n\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for grep that handles long lines and -e\" >&5\n$as_echo_n \"checking for grep that handles long lines and -e... \" >&6; }\nif ${ac_cv_path_GREP+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -z \"$GREP\"; then\n ac_path_GREP_found=false\n # Loop through the user's path and test for each of PROGNAME-LIST\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH$PATH_SEPARATOR/usr/xpg4/bin\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_prog in grep ggrep; do\n for ac_exec_ext in '' $ac_executable_extensions; do\n ac_path_GREP=\"$as_dir/$ac_prog$ac_exec_ext\"\n { test -f \"$ac_path_GREP\" && $as_test_x \"$ac_path_GREP\"; } || continue\n# Check for GNU ac_path_GREP and select it if it is found.\n # Check for GNU $ac_path_GREP\ncase `\"$ac_path_GREP\" --version 2>&1` in\n*GNU*)\n ac_cv_path_GREP=\"$ac_path_GREP\" ac_path_GREP_found=:;;\n*)\n ac_count=0\n $as_echo_n 0123456789 >\"conftest.in\"\n while :\n do\n cat \"conftest.in\" \"conftest.in\" >\"conftest.tmp\"\n mv \"conftest.tmp\" \"conftest.in\"\n cp \"conftest.in\" \"conftest.nl\"\n $as_echo 'GREP' >> \"conftest.nl\"\n \"$ac_path_GREP\" -e 'GREP$' -e '-(cannot match)-' < \"conftest.nl\" >\"conftest.out\" 2>/dev/null || break\n diff \"conftest.out\" \"conftest.nl\" >/dev/null 2>&1 || break\n as_fn_arith $ac_count + 1 && ac_count=$as_val\n if test $ac_count -gt ${ac_path_GREP_max-0}; then\n # Best one so far, save it but keep looking for a better one\n ac_cv_path_GREP=\"$ac_path_GREP\"\n ac_path_GREP_max=$ac_count\n fi\n # 10*(2^10) chars as input seems more than enough\n test $ac_count -gt 10 && break\n done\n rm -f conftest.in conftest.tmp conftest.nl conftest.out;;\nesac\n\n $ac_path_GREP_found && break 3\n done\n done\n done\nIFS=$as_save_IFS\n if test -z \"$ac_cv_path_GREP\"; then\n as_fn_error $? \"no acceptable grep could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin\" \"$LINENO\" 5\n fi\nelse\n ac_cv_path_GREP=$GREP\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_path_GREP\" >&5\n$as_echo \"$ac_cv_path_GREP\" >&6; }\n GREP=\"$ac_cv_path_GREP\"\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for egrep\" >&5\n$as_echo_n \"checking for egrep... \" >&6; }\nif ${ac_cv_path_EGREP+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if echo a | $GREP -E '(a|b)' >/dev/null 2>&1\n then ac_cv_path_EGREP=\"$GREP -E\"\n else\n if test -z \"$EGREP\"; then\n ac_path_EGREP_found=false\n # Loop through the user's path and test for each of PROGNAME-LIST\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH$PATH_SEPARATOR/usr/xpg4/bin\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_prog in egrep; do\n for ac_exec_ext in '' $ac_executable_extensions; do\n ac_path_EGREP=\"$as_dir/$ac_prog$ac_exec_ext\"\n { test -f \"$ac_path_EGREP\" && $as_test_x \"$ac_path_EGREP\"; } || continue\n# Check for GNU ac_path_EGREP and select it if it is found.\n # Check for GNU $ac_path_EGREP\ncase `\"$ac_path_EGREP\" --version 2>&1` in\n*GNU*)\n ac_cv_path_EGREP=\"$ac_path_EGREP\" ac_path_EGREP_found=:;;\n*)\n ac_count=0\n $as_echo_n 0123456789 >\"conftest.in\"\n while :\n do\n cat \"conftest.in\" \"conftest.in\" >\"conftest.tmp\"\n mv \"conftest.tmp\" \"conftest.in\"\n cp \"conftest.in\" \"conftest.nl\"\n $as_echo 'EGREP' >> \"conftest.nl\"\n \"$ac_path_EGREP\" 'EGREP$' < \"conftest.nl\" >\"conftest.out\" 2>/dev/null || break\n diff \"conftest.out\" \"conftest.nl\" >/dev/null 2>&1 || break\n as_fn_arith $ac_count + 1 && ac_count=$as_val\n if test $ac_count -gt ${ac_path_EGREP_max-0}; then\n # Best one so far, save it but keep looking for a better one\n ac_cv_path_EGREP=\"$ac_path_EGREP\"\n ac_path_EGREP_max=$ac_count\n fi\n # 10*(2^10) chars as input seems more than enough\n test $ac_count -gt 10 && break\n done\n rm -f conftest.in conftest.tmp conftest.nl conftest.out;;\nesac\n\n $ac_path_EGREP_found && break 3\n done\n done\n done\nIFS=$as_save_IFS\n if test -z \"$ac_cv_path_EGREP\"; then\n as_fn_error $? \"no acceptable egrep could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin\" \"$LINENO\" 5\n fi\nelse\n ac_cv_path_EGREP=$EGREP\nfi\n\n fi\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_path_EGREP\" >&5\n$as_echo \"$ac_cv_path_EGREP\" >&6; }\n EGREP=\"$ac_cv_path_EGREP\"\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for ANSI C header files\" >&5\n$as_echo_n \"checking for ANSI C header files... \" >&6; }\nif ${ac_cv_header_stdc+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n#include \n#include \n#include \n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_cv_header_stdc=yes\nelse\n ac_cv_header_stdc=no\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\nif test $ac_cv_header_stdc = yes; then\n # SunOS 4.x string.h does not declare mem*, contrary to ANSI.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n\n_ACEOF\nif (eval \"$ac_cpp conftest.$ac_ext\") 2>&5 |\n $EGREP \"memchr\" >/dev/null 2>&1; then :\n\nelse\n ac_cv_header_stdc=no\nfi\nrm -f conftest*\n\nfi\n\nif test $ac_cv_header_stdc = yes; then\n # ISC 2.0.2 stdlib.h does not declare free, contrary to ANSI.\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n\n_ACEOF\nif (eval \"$ac_cpp conftest.$ac_ext\") 2>&5 |\n $EGREP \"free\" >/dev/null 2>&1; then :\n\nelse\n ac_cv_header_stdc=no\nfi\nrm -f conftest*\n\nfi\n\nif test $ac_cv_header_stdc = yes; then\n # /bin/cc in Irix-4.0.5 gets non-ANSI ctype macros unless using -ansi.\n if test \"$cross_compiling\" = yes; then :\n :\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \n#include \n#if ((' ' & 0x0FF) == 0x020)\n# define ISLOWER(c) ('a' <= (c) && (c) <= 'z')\n# define TOUPPER(c) (ISLOWER(c) ? 'A' + ((c) - 'a') : (c))\n#else\n# define ISLOWER(c) \\\n\t\t (('a' <= (c) && (c) <= 'i') \\\n\t\t || ('j' <= (c) && (c) <= 'r') \\\n\t\t || ('s' <= (c) && (c) <= 'z'))\n# define TOUPPER(c) (ISLOWER(c) ? ((c) | 0x40) : (c))\n#endif\n\n#define XOR(e, f) (((e) && !(f)) || (!(e) && (f)))\nint\nmain ()\n{\n int i;\n for (i = 0; i < 256; i++)\n if (XOR (islower (i), ISLOWER (i))\n\t|| toupper (i) != TOUPPER (i))\n return 2;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_run \"$LINENO\"; then :\n\nelse\n ac_cv_header_stdc=no\nfi\nrm -f core *.core core.conftest.* gmon.out bb.out conftest$ac_exeext \\\n conftest.$ac_objext conftest.beam conftest.$ac_ext\nfi\n\nfi\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_header_stdc\" >&5\n$as_echo \"$ac_cv_header_stdc\" >&6; }\nif test $ac_cv_header_stdc = yes; then\n\n$as_echo \"#define STDC_HEADERS 1\" >>confdefs.h\n\nfi\n\n# On IRIX 5.3, sys/types and inttypes.h are conflicting.\nfor ac_header in sys/types.h sys/stat.h stdlib.h string.h memory.h strings.h \\\n\t\t inttypes.h stdint.h unistd.h\ndo :\n as_ac_Header=`$as_echo \"ac_cv_header_$ac_header\" | $as_tr_sh`\nac_fn_c_check_header_compile \"$LINENO\" \"$ac_header\" \"$as_ac_Header\" \"$ac_includes_default\n\"\nif eval test \\\"x\\$\"$as_ac_Header\"\\\" = x\"yes\"; then :\n cat >>confdefs.h <<_ACEOF\n#define `$as_echo \"HAVE_$ac_header\" | $as_tr_cpp` 1\n_ACEOF\n\nfi\n\ndone\n\n\n# The cast to long int works around a bug in the HP C Compiler\n# version HP92453-01 B.11.11.23709.GP, which incorrectly rejects\n# declarations like `int a3[[(sizeof (unsigned char)) >= 0]];'.\n# This bug is HP SR number 8606223364.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking size of void *\" >&5\n$as_echo_n \"checking size of void *... \" >&6; }\nif ${ac_cv_sizeof_void_p+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if ac_fn_c_compute_int \"$LINENO\" \"(long int) (sizeof (void *))\" \"ac_cv_sizeof_void_p\" \"$ac_includes_default\"; then :\n\nelse\n if test \"$ac_cv_type_void_p\" = yes; then\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error 77 \"cannot compute sizeof (void *)\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n else\n ac_cv_sizeof_void_p=0\n fi\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_sizeof_void_p\" >&5\n$as_echo \"$ac_cv_sizeof_void_p\" >&6; }\n\n\n\ncat >>confdefs.h <<_ACEOF\n#define SIZEOF_VOID_P $ac_cv_sizeof_void_p\n_ACEOF\n\n\nif test \"x${ac_cv_sizeof_void_p}\" = \"x8\" ; then\n LG_SIZEOF_PTR=3\nelif test \"x${ac_cv_sizeof_void_p}\" = \"x4\" ; then\n LG_SIZEOF_PTR=2\nelse\n as_fn_error $? \"Unsupported pointer size: ${ac_cv_sizeof_void_p}\" \"$LINENO\" 5\nfi\ncat >>confdefs.h <<_ACEOF\n#define LG_SIZEOF_PTR $LG_SIZEOF_PTR\n_ACEOF\n\n\n# The cast to long int works around a bug in the HP C Compiler\n# version HP92453-01 B.11.11.23709.GP, which incorrectly rejects\n# declarations like `int a3[[(sizeof (unsigned char)) >= 0]];'.\n# This bug is HP SR number 8606223364.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking size of int\" >&5\n$as_echo_n \"checking size of int... \" >&6; }\nif ${ac_cv_sizeof_int+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if ac_fn_c_compute_int \"$LINENO\" \"(long int) (sizeof (int))\" \"ac_cv_sizeof_int\" \"$ac_includes_default\"; then :\n\nelse\n if test \"$ac_cv_type_int\" = yes; then\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error 77 \"cannot compute sizeof (int)\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n else\n ac_cv_sizeof_int=0\n fi\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_sizeof_int\" >&5\n$as_echo \"$ac_cv_sizeof_int\" >&6; }\n\n\n\ncat >>confdefs.h <<_ACEOF\n#define SIZEOF_INT $ac_cv_sizeof_int\n_ACEOF\n\n\nif test \"x${ac_cv_sizeof_int}\" = \"x8\" ; then\n LG_SIZEOF_INT=3\nelif test \"x${ac_cv_sizeof_int}\" = \"x4\" ; then\n LG_SIZEOF_INT=2\nelse\n as_fn_error $? \"Unsupported int size: ${ac_cv_sizeof_int}\" \"$LINENO\" 5\nfi\ncat >>confdefs.h <<_ACEOF\n#define LG_SIZEOF_INT $LG_SIZEOF_INT\n_ACEOF\n\n\n# The cast to long int works around a bug in the HP C Compiler\n# version HP92453-01 B.11.11.23709.GP, which incorrectly rejects\n# declarations like `int a3[[(sizeof (unsigned char)) >= 0]];'.\n# This bug is HP SR number 8606223364.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking size of long\" >&5\n$as_echo_n \"checking size of long... \" >&6; }\nif ${ac_cv_sizeof_long+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if ac_fn_c_compute_int \"$LINENO\" \"(long int) (sizeof (long))\" \"ac_cv_sizeof_long\" \"$ac_includes_default\"; then :\n\nelse\n if test \"$ac_cv_type_long\" = yes; then\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error 77 \"cannot compute sizeof (long)\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n else\n ac_cv_sizeof_long=0\n fi\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_sizeof_long\" >&5\n$as_echo \"$ac_cv_sizeof_long\" >&6; }\n\n\n\ncat >>confdefs.h <<_ACEOF\n#define SIZEOF_LONG $ac_cv_sizeof_long\n_ACEOF\n\n\nif test \"x${ac_cv_sizeof_long}\" = \"x8\" ; then\n LG_SIZEOF_LONG=3\nelif test \"x${ac_cv_sizeof_long}\" = \"x4\" ; then\n LG_SIZEOF_LONG=2\nelse\n as_fn_error $? \"Unsupported long size: ${ac_cv_sizeof_long}\" \"$LINENO\" 5\nfi\ncat >>confdefs.h <<_ACEOF\n#define LG_SIZEOF_LONG $LG_SIZEOF_LONG\n_ACEOF\n\n\n# The cast to long int works around a bug in the HP C Compiler\n# version HP92453-01 B.11.11.23709.GP, which incorrectly rejects\n# declarations like `int a3[[(sizeof (unsigned char)) >= 0]];'.\n# This bug is HP SR number 8606223364.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking size of intmax_t\" >&5\n$as_echo_n \"checking size of intmax_t... \" >&6; }\nif ${ac_cv_sizeof_intmax_t+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if ac_fn_c_compute_int \"$LINENO\" \"(long int) (sizeof (intmax_t))\" \"ac_cv_sizeof_intmax_t\" \"$ac_includes_default\"; then :\n\nelse\n if test \"$ac_cv_type_intmax_t\" = yes; then\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error 77 \"cannot compute sizeof (intmax_t)\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\n else\n ac_cv_sizeof_intmax_t=0\n fi\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_sizeof_intmax_t\" >&5\n$as_echo \"$ac_cv_sizeof_intmax_t\" >&6; }\n\n\n\ncat >>confdefs.h <<_ACEOF\n#define SIZEOF_INTMAX_T $ac_cv_sizeof_intmax_t\n_ACEOF\n\n\nif test \"x${ac_cv_sizeof_intmax_t}\" = \"x16\" ; then\n LG_SIZEOF_INTMAX_T=4\nelif test \"x${ac_cv_sizeof_intmax_t}\" = \"x8\" ; then\n LG_SIZEOF_INTMAX_T=3\nelif test \"x${ac_cv_sizeof_intmax_t}\" = \"x4\" ; then\n LG_SIZEOF_INTMAX_T=2\nelse\n as_fn_error $? \"Unsupported intmax_t size: ${ac_cv_sizeof_intmax_t}\" \"$LINENO\" 5\nfi\ncat >>confdefs.h <<_ACEOF\n#define LG_SIZEOF_INTMAX_T $LG_SIZEOF_INTMAX_T\n_ACEOF\n\n\nac_aux_dir=\nfor ac_dir in \"$srcdir\" \"$srcdir/..\" \"$srcdir/../..\"; do\n if test -f \"$ac_dir/install-sh\"; then\n ac_aux_dir=$ac_dir\n ac_install_sh=\"$ac_aux_dir/install-sh -c\"\n break\n elif test -f \"$ac_dir/install.sh\"; then\n ac_aux_dir=$ac_dir\n ac_install_sh=\"$ac_aux_dir/install.sh -c\"\n break\n elif test -f \"$ac_dir/shtool\"; then\n ac_aux_dir=$ac_dir\n ac_install_sh=\"$ac_aux_dir/shtool install -c\"\n break\n fi\ndone\nif test -z \"$ac_aux_dir\"; then\n as_fn_error $? \"cannot find install-sh, install.sh, or shtool in \\\"$srcdir\\\" \\\"$srcdir/..\\\" \\\"$srcdir/../..\\\"\" \"$LINENO\" 5\nfi\n\n# These three variables are undocumented and unsupported,\n# and are intended to be withdrawn in a future Autoconf release.\n# They can cause serious problems if a builder's source tree is in a directory\n# whose full name contains unusual characters.\nac_config_guess=\"$SHELL $ac_aux_dir/config.guess\" # Please don't use this var.\nac_config_sub=\"$SHELL $ac_aux_dir/config.sub\" # Please don't use this var.\nac_configure=\"$SHELL $ac_aux_dir/configure\" # Please don't use this var.\n\n\n# Make sure we can run config.sub.\n$SHELL \"$ac_aux_dir/config.sub\" sun4 >/dev/null 2>&1 ||\n as_fn_error $? \"cannot run $SHELL $ac_aux_dir/config.sub\" \"$LINENO\" 5\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking build system type\" >&5\n$as_echo_n \"checking build system type... \" >&6; }\nif ${ac_cv_build+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_build_alias=$build_alias\ntest \"x$ac_build_alias\" = x &&\n ac_build_alias=`$SHELL \"$ac_aux_dir/config.guess\"`\ntest \"x$ac_build_alias\" = x &&\n as_fn_error $? \"cannot guess build type; you must specify one\" \"$LINENO\" 5\nac_cv_build=`$SHELL \"$ac_aux_dir/config.sub\" $ac_build_alias` ||\n as_fn_error $? \"$SHELL $ac_aux_dir/config.sub $ac_build_alias failed\" \"$LINENO\" 5\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_build\" >&5\n$as_echo \"$ac_cv_build\" >&6; }\ncase $ac_cv_build in\n*-*-*) ;;\n*) as_fn_error $? \"invalid value of canonical build\" \"$LINENO\" 5;;\nesac\nbuild=$ac_cv_build\nac_save_IFS=$IFS; IFS='-'\nset x $ac_cv_build\nshift\nbuild_cpu=$1\nbuild_vendor=$2\nshift; shift\n# Remember, the first character of IFS is used to create $*,\n# except with old shells:\nbuild_os=$*\nIFS=$ac_save_IFS\ncase $build_os in *\\ *) build_os=`echo \"$build_os\" | sed 's/ /-/g'`;; esac\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking host system type\" >&5\n$as_echo_n \"checking host system type... \" >&6; }\nif ${ac_cv_host+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test \"x$host_alias\" = x; then\n ac_cv_host=$ac_cv_build\nelse\n ac_cv_host=`$SHELL \"$ac_aux_dir/config.sub\" $host_alias` ||\n as_fn_error $? \"$SHELL $ac_aux_dir/config.sub $host_alias failed\" \"$LINENO\" 5\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_host\" >&5\n$as_echo \"$ac_cv_host\" >&6; }\ncase $ac_cv_host in\n*-*-*) ;;\n*) as_fn_error $? \"invalid value of canonical host\" \"$LINENO\" 5;;\nesac\nhost=$ac_cv_host\nac_save_IFS=$IFS; IFS='-'\nset x $ac_cv_host\nshift\nhost_cpu=$1\nhost_vendor=$2\nshift; shift\n# Remember, the first character of IFS is used to create $*,\n# except with old shells:\nhost_os=$*\nIFS=$ac_save_IFS\ncase $host_os in *\\ *) host_os=`echo \"$host_os\" | sed 's/ /-/g'`;; esac\n\n\nCPU_SPINWAIT=\"\"\ncase \"${host_cpu}\" in\n i[345]86)\n\t;;\n i686)\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether __asm__ is compilable\" >&5\n$as_echo_n \"checking whether __asm__ is compilable... \" >&6; }\nif ${je_cv_asm+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n__asm__ volatile(\"pause\"); return 0;\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_asm=yes\nelse\n je_cv_asm=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_asm\" >&5\n$as_echo \"$je_cv_asm\" >&6; }\n\n\tif test \"x${je_cv_asm}\" = \"xyes\" ; then\n\t CPU_SPINWAIT='__asm__ volatile(\"pause\")'\n\tfi\n\t;;\n x86_64)\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether __asm__ syntax is compilable\" >&5\n$as_echo_n \"checking whether __asm__ syntax is compilable... \" >&6; }\nif ${je_cv_asm+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\n__asm__ volatile(\"pause\"); return 0;\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_asm=yes\nelse\n je_cv_asm=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_asm\" >&5\n$as_echo \"$je_cv_asm\" >&6; }\n\n\tif test \"x${je_cv_asm}\" = \"xyes\" ; then\n\t CPU_SPINWAIT='__asm__ volatile(\"pause\")'\n\tfi\n\t;;\n *)\n\t;;\nesac\ncat >>confdefs.h <<_ACEOF\n#define CPU_SPINWAIT $CPU_SPINWAIT\n_ACEOF\n\n\nLD_PRELOAD_VAR=\"LD_PRELOAD\"\nso=\"so\"\nimportlib=\"${so}\"\no=\"$ac_objext\"\na=\"a\"\nexe=\"$ac_exeext\"\nlibprefix=\"lib\"\nDSO_LDFLAGS='-shared -Wl,-soname,$(@F)'\nRPATH='-Wl,-rpath,$(1)'\nSOREV=\"${so}.${rev}\"\nPIC_CFLAGS='-fPIC -DPIC'\nCTARGET='-o $@'\nLDTARGET='-o $@'\nEXTRA_LDFLAGS=\nMKLIB='ar crus $@'\nCC_MM=1\n\ndefault_munmap=\"1\"\nJEMALLOC_USABLE_SIZE_CONST=\"const\"\ncase \"${host}\" in\n *-*-darwin*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"macho\"\n\t$as_echo \"#define JEMALLOC_PURGE_MADVISE_FREE \" >>confdefs.h\n\n\tRPATH=\"\"\n\tLD_PRELOAD_VAR=\"DYLD_INSERT_LIBRARIES\"\n\tso=\"dylib\"\n\timportlib=\"${so}\"\n\tforce_tls=\"0\"\n\tDSO_LDFLAGS='-shared -Wl,-dylib_install_name,$(@F)'\n\tSOREV=\"${rev}.${so}\"\n\t;;\n *-*-freebsd*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"elf\"\n\t$as_echo \"#define JEMALLOC_PURGE_MADVISE_FREE \" >>confdefs.h\n\n\tforce_lazy_lock=\"1\"\n\t;;\n *-*-linux*)\n\tCFLAGS=\"$CFLAGS\"\n\tCPPFLAGS=\"$CPPFLAGS -D_GNU_SOURCE\"\n\tabi=\"elf\"\n\t$as_echo \"#define JEMALLOC_PURGE_MADVISE_DONTNEED \" >>confdefs.h\n\n\t$as_echo \"#define JEMALLOC_THREADED_INIT \" >>confdefs.h\n\n\tJEMALLOC_USABLE_SIZE_CONST=\"\"\n\tdefault_munmap=\"0\"\n\t;;\n *-*-netbsd*)\n\t{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking ABI\" >&5\n$as_echo_n \"checking ABI... \" >&6; }\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#ifdef __ELF__\n/* ELF */\n#else\n#error aout\n#endif\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n CFLAGS=\"$CFLAGS\"; abi=\"elf\"\nelse\n abi=\"aout\"\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\t{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $abi\" >&5\n$as_echo \"$abi\" >&6; }\n\t$as_echo \"#define JEMALLOC_PURGE_MADVISE_FREE \" >>confdefs.h\n\n\t;;\n *-*-solaris2*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"elf\"\n\tRPATH='-Wl,-R,$(1)'\n\t\tCPPFLAGS=\"$CPPFLAGS -D_POSIX_PTHREAD_SEMANTICS\"\n\tLIBS=\"$LIBS -lposix4 -lsocket -lnsl\"\n\t;;\n *-ibm-aix*)\n\tif \"$LG_SIZEOF_PTR\" = \"8\"; then\n\t \t LD_PRELOAD_VAR=\"LDR_PRELOAD64\"\n\telse\n\t \t LD_PRELOAD_VAR=\"LDR_PRELOAD\"\n\tfi\n\tabi=\"xcoff\"\n\t;;\n *-*-mingw*)\n\tabi=\"pecoff\"\n\tforce_tls=\"0\"\n\tRPATH=\"\"\n\tso=\"dll\"\n\tif test \"x$je_cv_msvc\" = \"xyes\" ; then\n\t importlib=\"lib\"\n\t DSO_LDFLAGS=\"-LD\"\n\t EXTRA_LDFLAGS=\"-link -DEBUG\"\n\t CTARGET='-Fo$@'\n\t LDTARGET='-Fe$@'\n\t MKLIB='lib -nologo -out:$@'\n\t CC_MM=\n else\n\t importlib=\"${so}\"\n\t DSO_LDFLAGS=\"-shared\"\n\tfi\n\ta=\"lib\"\n\tlibprefix=\"\"\n\tSOREV=\"${so}\"\n\tPIC_CFLAGS=\"\"\n\t;;\n *)\n\t{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: Unsupported operating system: ${host}\" >&5\n$as_echo \"Unsupported operating system: ${host}\" >&6; }\n\tabi=\"elf\"\n\t;;\nesac\ncat >>confdefs.h <<_ACEOF\n#define JEMALLOC_USABLE_SIZE_CONST $JEMALLOC_USABLE_SIZE_CONST\n_ACEOF\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nif test \"x$abi\" != \"xpecoff\"; then\n LIBS=\"$LIBS -lm\"\nfi\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether __attribute__ syntax is compilable\" >&5\n$as_echo_n \"checking whether __attribute__ syntax is compilable... \" >&6; }\nif ${je_cv_attribute+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\nstatic __attribute__((unused)) void foo(void){}\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_attribute=yes\nelse\n je_cv_attribute=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_attribute\" >&5\n$as_echo \"$je_cv_attribute\" >&6; }\n\nif test \"x${je_cv_attribute}\" = \"xyes\" ; then\n $as_echo \"#define JEMALLOC_HAVE_ATTR \" >>confdefs.h\n\n if test \"x${GCC}\" = \"xyes\" -a \"x${abi}\" = \"xelf\"; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -fvisibility=hidden\" >&5\n$as_echo_n \"checking whether compiler supports -fvisibility=hidden... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-fvisibility=hidden\"\nelse\n CFLAGS=\"${CFLAGS} -fvisibility=hidden\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n fi\nfi\nSAVED_CFLAGS=\"${CFLAGS}\"\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -Werror\" >&5\n$as_echo_n \"checking whether compiler supports -Werror... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-Werror\"\nelse\n CFLAGS=\"${CFLAGS} -Werror\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether tls_model attribute is compilable\" >&5\n$as_echo_n \"checking whether tls_model attribute is compilable... \" >&6; }\nif ${je_cv_tls_model+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\nint\nmain ()\n{\nstatic __thread int\n __attribute__((tls_model(\"initial-exec\"))) foo;\n foo = 0;\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_tls_model=yes\nelse\n je_cv_tls_model=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_tls_model\" >&5\n$as_echo \"$je_cv_tls_model\" >&6; }\n\nCFLAGS=\"${SAVED_CFLAGS}\"\nif test \"x${je_cv_tls_model}\" = \"xyes\" ; then\n $as_echo \"#define JEMALLOC_TLS_MODEL __attribute__((tls_model(\\\"initial-exec\\\")))\" >>confdefs.h\n\nelse\n $as_echo \"#define JEMALLOC_TLS_MODEL \" >>confdefs.h\n\nfi\n\n\n# Check whether --with-rpath was given.\nif test \"${with_rpath+set}\" = set; then :\n withval=$with_rpath; if test \"x$with_rpath\" = \"xno\" ; then\n RPATH_EXTRA=\nelse\n RPATH_EXTRA=\"`echo $with_rpath | tr \\\":\\\" \\\" \\\"`\"\nfi\nelse\n RPATH_EXTRA=\n\nfi\n\n\n\n# Check whether --enable-autogen was given.\nif test \"${enable_autogen+set}\" = set; then :\n enableval=$enable_autogen; if test \"x$enable_autogen\" = \"xno\" ; then\n enable_autogen=\"0\"\nelse\n enable_autogen=\"1\"\nfi\n\nelse\n enable_autogen=\"0\"\n\nfi\n\n\n\n# Find a good install program. We prefer a C program (faster),\n# so one script is as good as another. But avoid the broken or\n# incompatible versions:\n# SysV /etc/install, /usr/sbin/install\n# SunOS /usr/etc/install\n# IRIX /sbin/install\n# AIX /bin/install\n# AmigaOS /C/install, which installs bootblocks on floppy discs\n# AIX 4 /usr/bin/installbsd, which doesn't work without a -g flag\n# AFS /usr/afsws/bin/install, which mishandles nonexistent args\n# SVR4 /usr/ucb/install, which tries to use the nonexistent group \"staff\"\n# OS/2's system install, which has a completely different semantic\n# ./install, which can be erroneously created by make from ./install.sh.\n# Reject install programs that cannot install multiple files.\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for a BSD-compatible install\" >&5\n$as_echo_n \"checking for a BSD-compatible install... \" >&6; }\nif test -z \"$INSTALL\"; then\nif ${ac_cv_path_install+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n # Account for people who put trailing slashes in PATH elements.\ncase $as_dir/ in #((\n ./ | .// | /[cC]/* | \\\n /etc/* | /usr/sbin/* | /usr/etc/* | /sbin/* | /usr/afsws/bin/* | \\\n ?:[\\\\/]os2[\\\\/]install[\\\\/]* | ?:[\\\\/]OS2[\\\\/]INSTALL[\\\\/]* | \\\n /usr/ucb/* ) ;;\n *)\n # OSF1 and SCO ODT 3.0 have their own names for install.\n # Don't use installbsd from OSF since it installs stuff as root\n # by default.\n for ac_prog in ginstall scoinst install; do\n for ac_exec_ext in '' $ac_executable_extensions; do\n\tif { test -f \"$as_dir/$ac_prog$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_prog$ac_exec_ext\"; }; then\n\t if test $ac_prog = install &&\n\t grep dspmsg \"$as_dir/$ac_prog$ac_exec_ext\" >/dev/null 2>&1; then\n\t # AIX install. It has an incompatible calling convention.\n\t :\n\t elif test $ac_prog = install &&\n\t grep pwplus \"$as_dir/$ac_prog$ac_exec_ext\" >/dev/null 2>&1; then\n\t # program-specific install script used by HP pwplus--don't use.\n\t :\n\t else\n\t rm -rf conftest.one conftest.two conftest.dir\n\t echo one > conftest.one\n\t echo two > conftest.two\n\t mkdir conftest.dir\n\t if \"$as_dir/$ac_prog$ac_exec_ext\" -c conftest.one conftest.two \"`pwd`/conftest.dir\" &&\n\t test -s conftest.one && test -s conftest.two &&\n\t test -s conftest.dir/conftest.one &&\n\t test -s conftest.dir/conftest.two\n\t then\n\t ac_cv_path_install=\"$as_dir/$ac_prog$ac_exec_ext -c\"\n\t break 3\n\t fi\n\t fi\n\tfi\n done\n done\n ;;\nesac\n\n done\nIFS=$as_save_IFS\n\nrm -rf conftest.one conftest.two conftest.dir\n\nfi\n if test \"${ac_cv_path_install+set}\" = set; then\n INSTALL=$ac_cv_path_install\n else\n # As a last resort, use the slow shell script. Don't cache a\n # value for INSTALL within a source directory, because that will\n # break other packages using the cache if that directory is\n # removed, or if the value is a relative name.\n INSTALL=$ac_install_sh\n fi\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $INSTALL\" >&5\n$as_echo \"$INSTALL\" >&6; }\n\n# Use test -z because SunOS4 sh mishandles braces in ${var-val}.\n# It thinks the first close brace ends the variable substitution.\ntest -z \"$INSTALL_PROGRAM\" && INSTALL_PROGRAM='${INSTALL}'\n\ntest -z \"$INSTALL_SCRIPT\" && INSTALL_SCRIPT='${INSTALL}'\n\ntest -z \"$INSTALL_DATA\" && INSTALL_DATA='${INSTALL} -m 644'\n\nif test -n \"$ac_tool_prefix\"; then\n # Extract the first word of \"${ac_tool_prefix}ranlib\", so it can be a program name with args.\nset dummy ${ac_tool_prefix}ranlib; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_RANLIB+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$RANLIB\"; then\n ac_cv_prog_RANLIB=\"$RANLIB\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_RANLIB=\"${ac_tool_prefix}ranlib\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nRANLIB=$ac_cv_prog_RANLIB\nif test -n \"$RANLIB\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $RANLIB\" >&5\n$as_echo \"$RANLIB\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\nfi\nif test -z \"$ac_cv_prog_RANLIB\"; then\n ac_ct_RANLIB=$RANLIB\n # Extract the first word of \"ranlib\", so it can be a program name with args.\nset dummy ranlib; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_prog_ac_ct_RANLIB+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test -n \"$ac_ct_RANLIB\"; then\n ac_cv_prog_ac_ct_RANLIB=\"$ac_ct_RANLIB\" # Let the user override the test.\nelse\nas_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_prog_ac_ct_RANLIB=\"ranlib\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\nfi\nfi\nac_ct_RANLIB=$ac_cv_prog_ac_ct_RANLIB\nif test -n \"$ac_ct_RANLIB\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_ct_RANLIB\" >&5\n$as_echo \"$ac_ct_RANLIB\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n if test \"x$ac_ct_RANLIB\" = x; then\n RANLIB=\":\"\n else\n case $cross_compiling:$ac_tool_warned in\nyes:)\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: using cross tools not prefixed with host triplet\" >&5\n$as_echo \"$as_me: WARNING: using cross tools not prefixed with host triplet\" >&2;}\nac_tool_warned=yes ;;\nesac\n RANLIB=$ac_ct_RANLIB\n fi\nelse\n RANLIB=\"$ac_cv_prog_RANLIB\"\nfi\n\n# Extract the first word of \"ar\", so it can be a program name with args.\nset dummy ar; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_path_AR+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n case $AR in\n [\\\\/]* | ?:[\\\\/]*)\n ac_cv_path_AR=\"$AR\" # Let the user override the test with a path.\n ;;\n *)\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_path_AR=\"$as_dir/$ac_word$ac_exec_ext\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\n ;;\nesac\nfi\nAR=$ac_cv_path_AR\nif test -n \"$AR\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $AR\" >&5\n$as_echo \"$AR\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n# Extract the first word of \"ld\", so it can be a program name with args.\nset dummy ld; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_path_LD+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n case $LD in\n [\\\\/]* | ?:[\\\\/]*)\n ac_cv_path_LD=\"$LD\" # Let the user override the test with a path.\n ;;\n *)\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_path_LD=\"$as_dir/$ac_word$ac_exec_ext\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\n ;;\nesac\nfi\nLD=$ac_cv_path_LD\nif test -n \"$LD\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $LD\" >&5\n$as_echo \"$LD\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n# Extract the first word of \"autoconf\", so it can be a program name with args.\nset dummy autoconf; ac_word=$2\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for $ac_word\" >&5\n$as_echo_n \"checking for $ac_word... \" >&6; }\nif ${ac_cv_path_AUTOCONF+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n case $AUTOCONF in\n [\\\\/]* | ?:[\\\\/]*)\n ac_cv_path_AUTOCONF=\"$AUTOCONF\" # Let the user override the test with a path.\n ;;\n *)\n as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n for ac_exec_ext in '' $ac_executable_extensions; do\n if { test -f \"$as_dir/$ac_word$ac_exec_ext\" && $as_test_x \"$as_dir/$ac_word$ac_exec_ext\"; }; then\n ac_cv_path_AUTOCONF=\"$as_dir/$ac_word$ac_exec_ext\"\n $as_echo \"$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext\" >&5\n break 2\n fi\ndone\n done\nIFS=$as_save_IFS\n\n ;;\nesac\nfi\nAUTOCONF=$ac_cv_path_AUTOCONF\nif test -n \"$AUTOCONF\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $AUTOCONF\" >&5\n$as_echo \"$AUTOCONF\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\nfi\n\n\n\npublic_syms=\"malloc_conf malloc_message malloc calloc posix_memalign aligned_alloc realloc free malloc_usable_size malloc_stats_print mallctl mallctlnametomib mallctlbymib\"\n\nac_fn_c_check_func \"$LINENO\" \"memalign\" \"ac_cv_func_memalign\"\nif test \"x$ac_cv_func_memalign\" = xyes; then :\n $as_echo \"#define JEMALLOC_OVERRIDE_MEMALIGN \" >>confdefs.h\n\n\t public_syms=\"${public_syms} memalign\"\nfi\n\nac_fn_c_check_func \"$LINENO\" \"valloc\" \"ac_cv_func_valloc\"\nif test \"x$ac_cv_func_valloc\" = xyes; then :\n $as_echo \"#define JEMALLOC_OVERRIDE_VALLOC \" >>confdefs.h\n\n\t public_syms=\"${public_syms} valloc\"\nfi\n\n\n# Check whether --enable-experimental was given.\nif test \"${enable_experimental+set}\" = set; then :\n enableval=$enable_experimental; if test \"x$enable_experimental\" = \"xno\" ; then\n enable_experimental=\"0\"\nelse\n enable_experimental=\"1\"\nfi\n\nelse\n enable_experimental=\"1\"\n\nfi\n\nif test \"x$enable_experimental\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_EXPERIMENTAL \" >>confdefs.h\n\n public_syms=\"${public_syms} allocm dallocm nallocm rallocm sallocm\"\nfi\n\n\n\n# Check whether --with-mangling was given.\nif test \"${with_mangling+set}\" = set; then :\n withval=$with_mangling; mangling_map=\"$with_mangling\"\nelse\n mangling_map=\"\"\nfi\n\nfor nm in `echo ${mangling_map} |tr ',' ' '` ; do\n k=\"`echo ${nm} |tr ':' ' ' |awk '{print $1}'`\"\n n=\"je_${k}\"\n m=`echo ${nm} |tr ':' ' ' |awk '{print $2}'`\n cat >>confdefs.h <<_ACEOF\n#define ${n} ${m}\n_ACEOF\n\n public_syms=`for sym in ${public_syms}; do echo \"${sym}\"; done |grep -v \"^${k}\\$\" |tr '\\n' ' '`\ndone\n\n\n# Check whether --with-jemalloc_prefix was given.\nif test \"${with_jemalloc_prefix+set}\" = set; then :\n withval=$with_jemalloc_prefix; JEMALLOC_PREFIX=\"$with_jemalloc_prefix\"\nelse\n if test \"x$abi\" != \"xmacho\" -a \"x$abi\" != \"xpecoff\"; then\n JEMALLOC_PREFIX=\"\"\nelse\n JEMALLOC_PREFIX=\"je_\"\nfi\n\nfi\n\nif test \"x$JEMALLOC_PREFIX\" != \"x\" ; then\n JEMALLOC_CPREFIX=`echo ${JEMALLOC_PREFIX} | tr \"a-z\" \"A-Z\"`\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_PREFIX \"$JEMALLOC_PREFIX\"\n_ACEOF\n\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_CPREFIX \"$JEMALLOC_CPREFIX\"\n_ACEOF\n\nfi\nfor stem in ${public_syms}; do\n n=\"je_${stem}\"\n m=\"${JEMALLOC_PREFIX}${stem}\"\n cat >>confdefs.h <<_ACEOF\n#define ${n} ${m}\n_ACEOF\n\ndone\n\n\n# Check whether --with-private_namespace was given.\nif test \"${with_private_namespace+set}\" = set; then :\n withval=$with_private_namespace; JEMALLOC_PRIVATE_NAMESPACE=\"$with_private_namespace\"\nelse\n JEMALLOC_PRIVATE_NAMESPACE=\"\"\n\nfi\n\ncat >>confdefs.h <<_ACEOF\n#define JEMALLOC_PRIVATE_NAMESPACE \"$JEMALLOC_PRIVATE_NAMESPACE\"\n_ACEOF\n\nif test \"x$JEMALLOC_PRIVATE_NAMESPACE\" != \"x\" ; then\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_N(string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix) ${JEMALLOC_PRIVATE_NAMESPACE}##string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix\n_ACEOF\n\nelse\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_N(string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix) string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix\n_ACEOF\n\nfi\n\n\n# Check whether --with-install_suffix was given.\nif test \"${with_install_suffix+set}\" = set; then :\n withval=$with_install_suffix; INSTALL_SUFFIX=\"$with_install_suffix\"\nelse\n INSTALL_SUFFIX=\n\nfi\n\ninstall_suffix=\"$INSTALL_SUFFIX\"\n\n\ncfgoutputs_in=\"${srcroot}Makefile.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/html.xsl.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/manpages.xsl.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/jemalloc.xml.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}include/jemalloc/jemalloc.h.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}include/jemalloc/internal/jemalloc_internal.h.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}test/jemalloc_test.h.in\"\n\ncfgoutputs_out=\"Makefile\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/html.xsl\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/manpages.xsl\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/jemalloc${install_suffix}.xml\"\ncfgoutputs_out=\"${cfgoutputs_out} include/jemalloc/jemalloc${install_suffix}.h\"\ncfgoutputs_out=\"${cfgoutputs_out} include/jemalloc/internal/jemalloc_internal.h\"\ncfgoutputs_out=\"${cfgoutputs_out} test/jemalloc_test.h\"\n\ncfgoutputs_tup=\"Makefile\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/html.xsl:doc/html.xsl.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/manpages.xsl:doc/manpages.xsl.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/jemalloc${install_suffix}.xml:doc/jemalloc.xml.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} include/jemalloc/jemalloc${install_suffix}.h:include/jemalloc/jemalloc.h.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} include/jemalloc/internal/jemalloc_internal.h\"\ncfgoutputs_tup=\"${cfgoutputs_tup} test/jemalloc_test.h:test/jemalloc_test.h.in\"\n\ncfghdrs_in=\"${srcroot}include/jemalloc/jemalloc_defs.h.in\"\ncfghdrs_in=\"${cfghdrs_in} ${srcroot}include/jemalloc/internal/size_classes.sh\"\n\ncfghdrs_out=\"include/jemalloc/jemalloc_defs${install_suffix}.h\"\ncfghdrs_out=\"${cfghdrs_out} include/jemalloc/internal/size_classes.h\"\n\ncfghdrs_tup=\"include/jemalloc/jemalloc_defs${install_suffix}.h:include/jemalloc/jemalloc_defs.h.in\"\n\n# Check whether --enable-cc-silence was given.\nif test \"${enable_cc_silence+set}\" = set; then :\n enableval=$enable_cc_silence; if test \"x$enable_cc_silence\" = \"xno\" ; then\n enable_cc_silence=\"0\"\nelse\n enable_cc_silence=\"1\"\nfi\n\nelse\n enable_cc_silence=\"0\"\n\nfi\n\nif test \"x$enable_cc_silence\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_CC_SILENCE \" >>confdefs.h\n\nfi\n\n# Check whether --enable-debug was given.\nif test \"${enable_debug+set}\" = set; then :\n enableval=$enable_debug; if test \"x$enable_debug\" = \"xno\" ; then\n enable_debug=\"0\"\nelse\n enable_debug=\"1\"\nfi\n\nelse\n enable_debug=\"0\"\n\nfi\n\nif test \"x$enable_debug\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_DEBUG \" >>confdefs.h\n\n $as_echo \"#define JEMALLOC_IVSALLOC \" >>confdefs.h\n\nfi\n\n\nif test \"x$enable_debug\" = \"x0\" -a \"x$no_CFLAGS\" = \"xyes\" ; then\n optimize=\"no\"\n echo \"$EXTRA_CFLAGS\" | grep \"\\-O\" >/dev/null || optimize=\"yes\"\n if test \"x${optimize}\" = \"xyes\" ; then\n if test \"x$GCC\" = \"xyes\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -O3\" >&5\n$as_echo_n \"checking whether compiler supports -O3... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-O3\"\nelse\n CFLAGS=\"${CFLAGS} -O3\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -funroll-loops\" >&5\n$as_echo_n \"checking whether compiler supports -funroll-loops... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-funroll-loops\"\nelse\n CFLAGS=\"${CFLAGS} -funroll-loops\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n elif test \"x$je_cv_msvc\" = \"xyes\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -O2\" >&5\n$as_echo_n \"checking whether compiler supports -O2... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-O2\"\nelse\n CFLAGS=\"${CFLAGS} -O2\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n else\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether compiler supports -O\" >&5\n$as_echo_n \"checking whether compiler supports -O... \" >&6; }\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"-O\"\nelse\n CFLAGS=\"${CFLAGS} -O\"\nfi\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n\nint\nmain ()\n{\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n CFLAGS=\"${TCFLAGS}\"\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n\n fi\n fi\nfi\n\n# Check whether --enable-stats was given.\nif test \"${enable_stats+set}\" = set; then :\n enableval=$enable_stats; if test \"x$enable_stats\" = \"xno\" ; then\n enable_stats=\"0\"\nelse\n enable_stats=\"1\"\nfi\n\nelse\n enable_stats=\"1\"\n\nfi\n\nif test \"x$enable_stats\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_STATS \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-prof was given.\nif test \"${enable_prof+set}\" = set; then :\n enableval=$enable_prof; if test \"x$enable_prof\" = \"xno\" ; then\n enable_prof=\"0\"\nelse\n enable_prof=\"1\"\nfi\n\nelse\n enable_prof=\"0\"\n\nfi\n\nif test \"x$enable_prof\" = \"x1\" ; then\n backtrace_method=\"\"\nelse\n backtrace_method=\"N/A\"\nfi\n\n# Check whether --enable-prof-libunwind was given.\nif test \"${enable_prof_libunwind+set}\" = set; then :\n enableval=$enable_prof_libunwind; if test \"x$enable_prof_libunwind\" = \"xno\" ; then\n enable_prof_libunwind=\"0\"\nelse\n enable_prof_libunwind=\"1\"\nfi\n\nelse\n enable_prof_libunwind=\"0\"\n\nfi\n\n\n# Check whether --with-static_libunwind was given.\nif test \"${with_static_libunwind+set}\" = set; then :\n withval=$with_static_libunwind; if test \"x$with_static_libunwind\" = \"xno\" ; then\n LUNWIND=\"-lunwind\"\nelse\n if test ! -f \"$with_static_libunwind\" ; then\n as_fn_error $? \"Static libunwind not found: $with_static_libunwind\" \"$LINENO\" 5\n fi\n LUNWIND=\"$with_static_libunwind\"\nfi\nelse\n LUNWIND=\"-lunwind\"\n\nfi\n\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_libunwind\" = \"x1\" ; then\n for ac_header in libunwind.h\ndo :\n ac_fn_c_check_header_mongrel \"$LINENO\" \"libunwind.h\" \"ac_cv_header_libunwind_h\" \"$ac_includes_default\"\nif test \"x$ac_cv_header_libunwind_h\" = xyes; then :\n cat >>confdefs.h <<_ACEOF\n#define HAVE_LIBUNWIND_H 1\n_ACEOF\n\nelse\n enable_prof_libunwind=\"0\"\nfi\n\ndone\n\n if test \"x$LUNWIND\" = \"x-lunwind\" ; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for backtrace in -lunwind\" >&5\n$as_echo_n \"checking for backtrace in -lunwind... \" >&6; }\nif ${ac_cv_lib_unwind_backtrace+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_check_lib_save_LIBS=$LIBS\nLIBS=\"-lunwind $LIBS\"\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar backtrace ();\nint\nmain ()\n{\nreturn backtrace ();\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n ac_cv_lib_unwind_backtrace=yes\nelse\n ac_cv_lib_unwind_backtrace=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nLIBS=$ac_check_lib_save_LIBS\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_unwind_backtrace\" >&5\n$as_echo \"$ac_cv_lib_unwind_backtrace\" >&6; }\nif test \"x$ac_cv_lib_unwind_backtrace\" = xyes; then :\n LIBS=\"$LIBS $LUNWIND\"\nelse\n enable_prof_libunwind=\"0\"\nfi\n\n else\n LIBS=\"$LIBS $LUNWIND\"\n fi\n if test \"x${enable_prof_libunwind}\" = \"x1\" ; then\n backtrace_method=\"libunwind\"\n $as_echo \"#define JEMALLOC_PROF_LIBUNWIND \" >>confdefs.h\n\n fi\nfi\n\n# Check whether --enable-prof-libgcc was given.\nif test \"${enable_prof_libgcc+set}\" = set; then :\n enableval=$enable_prof_libgcc; if test \"x$enable_prof_libgcc\" = \"xno\" ; then\n enable_prof_libgcc=\"0\"\nelse\n enable_prof_libgcc=\"1\"\nfi\n\nelse\n enable_prof_libgcc=\"1\"\n\nfi\n\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_libgcc\" = \"x1\" \\\n -a \"x$GCC\" = \"xyes\" ; then\n for ac_header in unwind.h\ndo :\n ac_fn_c_check_header_mongrel \"$LINENO\" \"unwind.h\" \"ac_cv_header_unwind_h\" \"$ac_includes_default\"\nif test \"x$ac_cv_header_unwind_h\" = xyes; then :\n cat >>confdefs.h <<_ACEOF\n#define HAVE_UNWIND_H 1\n_ACEOF\n\nelse\n enable_prof_libgcc=\"0\"\nfi\n\ndone\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for _Unwind_Backtrace in -lgcc\" >&5\n$as_echo_n \"checking for _Unwind_Backtrace in -lgcc... \" >&6; }\nif ${ac_cv_lib_gcc__Unwind_Backtrace+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_check_lib_save_LIBS=$LIBS\nLIBS=\"-lgcc $LIBS\"\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar _Unwind_Backtrace ();\nint\nmain ()\n{\nreturn _Unwind_Backtrace ();\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n ac_cv_lib_gcc__Unwind_Backtrace=yes\nelse\n ac_cv_lib_gcc__Unwind_Backtrace=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nLIBS=$ac_check_lib_save_LIBS\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_gcc__Unwind_Backtrace\" >&5\n$as_echo \"$ac_cv_lib_gcc__Unwind_Backtrace\" >&6; }\nif test \"x$ac_cv_lib_gcc__Unwind_Backtrace\" = xyes; then :\n LIBS=\"$LIBS -lgcc\"\nelse\n enable_prof_libgcc=\"0\"\nfi\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking libgcc-based backtracing reliability on ${host_cpu}\" >&5\n$as_echo_n \"checking libgcc-based backtracing reliability on ${host_cpu}... \" >&6; }\n case \"${host_cpu}\" in\n i[3456]86)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: unreliable\" >&5\n$as_echo \"unreliable\" >&6; }\n enable_prof_libgcc=\"0\";\n ;;\n x86_64)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: reliable\" >&5\n$as_echo \"reliable\" >&6; }\n ;;\n *)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: unreliable\" >&5\n$as_echo \"unreliable\" >&6; }\n enable_prof_libgcc=\"0\";\n ;;\n esac\n if test \"x${enable_prof_libgcc}\" = \"x1\" ; then\n backtrace_method=\"libgcc\"\n $as_echo \"#define JEMALLOC_PROF_LIBGCC \" >>confdefs.h\n\n fi\nelse\n enable_prof_libgcc=\"0\"\nfi\n\n# Check whether --enable-prof-gcc was given.\nif test \"${enable_prof_gcc+set}\" = set; then :\n enableval=$enable_prof_gcc; if test \"x$enable_prof_gcc\" = \"xno\" ; then\n enable_prof_gcc=\"0\"\nelse\n enable_prof_gcc=\"1\"\nfi\n\nelse\n enable_prof_gcc=\"1\"\n\nfi\n\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_gcc\" = \"x1\" \\\n -a \"x$GCC\" = \"xyes\" ; then\n backtrace_method=\"gcc intrinsics\"\n $as_echo \"#define JEMALLOC_PROF_GCC \" >>confdefs.h\n\nelse\n enable_prof_gcc=\"0\"\nfi\n\nif test \"x$backtrace_method\" = \"x\" ; then\n backtrace_method=\"none (disabling profiling)\"\n enable_prof=\"0\"\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking configured backtracing method\" >&5\n$as_echo_n \"checking configured backtracing method... \" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $backtrace_method\" >&5\n$as_echo \"$backtrace_method\" >&6; }\nif test \"x$enable_prof\" = \"x1\" ; then\n if test \"x${force_tls}\" = \"x0\" ; then\n as_fn_error $? \"Heap profiling requires TLS\" \"$LINENO\" 5;\n fi\n force_tls=\"1\"\n $as_echo \"#define JEMALLOC_PROF \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-tcache was given.\nif test \"${enable_tcache+set}\" = set; then :\n enableval=$enable_tcache; if test \"x$enable_tcache\" = \"xno\" ; then\n enable_tcache=\"0\"\nelse\n enable_tcache=\"1\"\nfi\n\nelse\n enable_tcache=\"1\"\n\nfi\n\nif test \"x$enable_tcache\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_TCACHE \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-mremap was given.\nif test \"${enable_mremap+set}\" = set; then :\n enableval=$enable_mremap; if test \"x$enable_mremap\" = \"xno\" ; then\n enable_mremap=\"0\"\nelse\n enable_mremap=\"1\"\nfi\n\nelse\n enable_mremap=\"0\"\n\nfi\n\nif test \"x$enable_mremap\" = \"x1\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether mremap(...MREMAP_FIXED...) is compilable\" >&5\n$as_echo_n \"checking whether mremap(...MREMAP_FIXED...) is compilable... \" >&6; }\nif ${je_cv_mremap_fixed+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#define _GNU_SOURCE\n#include \n\nint\nmain ()\n{\n\nvoid *p = mremap((void *)0, 0, 0, MREMAP_MAYMOVE|MREMAP_FIXED, (void *)0);\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_mremap_fixed=yes\nelse\n je_cv_mremap_fixed=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_mremap_fixed\" >&5\n$as_echo \"$je_cv_mremap_fixed\" >&6; }\n\n if test \"x${je_cv_mremap_fixed}\" = \"xno\" ; then\n enable_mremap=\"0\"\n fi\nfi\nif test \"x$enable_mremap\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_MREMAP \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-munmap was given.\nif test \"${enable_munmap+set}\" = set; then :\n enableval=$enable_munmap; if test \"x$enable_munmap\" = \"xno\" ; then\n enable_munmap=\"0\"\nelse\n enable_munmap=\"1\"\nfi\n\nelse\n enable_munmap=\"${default_munmap}\"\n\nfi\n\nif test \"x$enable_munmap\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_MUNMAP \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-dss was given.\nif test \"${enable_dss+set}\" = set; then :\n enableval=$enable_dss; if test \"x$enable_dss\" = \"xno\" ; then\n enable_dss=\"0\"\nelse\n enable_dss=\"1\"\nfi\n\nelse\n enable_dss=\"0\"\n\nfi\n\nac_fn_c_check_func \"$LINENO\" \"sbrk\" \"ac_cv_func_sbrk\"\nif test \"x$ac_cv_func_sbrk\" = xyes; then :\n have_sbrk=\"1\"\nelse\n have_sbrk=\"0\"\nfi\n\nif test \"x$have_sbrk\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_HAVE_SBRK \" >>confdefs.h\n\nelse\n enable_dss=\"0\"\nfi\n\nif test \"x$enable_dss\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_DSS \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-fill was given.\nif test \"${enable_fill+set}\" = set; then :\n enableval=$enable_fill; if test \"x$enable_fill\" = \"xno\" ; then\n enable_fill=\"0\"\nelse\n enable_fill=\"1\"\nfi\n\nelse\n enable_fill=\"1\"\n\nfi\n\nif test \"x$enable_fill\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_FILL \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-utrace was given.\nif test \"${enable_utrace+set}\" = set; then :\n enableval=$enable_utrace; if test \"x$enable_utrace\" = \"xno\" ; then\n enable_utrace=\"0\"\nelse\n enable_utrace=\"1\"\nfi\n\nelse\n enable_utrace=\"0\"\n\nfi\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether utrace(2) is compilable\" >&5\n$as_echo_n \"checking whether utrace(2) is compilable... \" >&6; }\nif ${je_cv_utrace+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n#include \n#include \n#include \n\nint\nmain ()\n{\n\n\tutrace((void *)0, 0);\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_utrace=yes\nelse\n je_cv_utrace=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_utrace\" >&5\n$as_echo \"$je_cv_utrace\" >&6; }\n\nif test \"x${je_cv_utrace}\" = \"xno\" ; then\n enable_utrace=\"0\"\nfi\nif test \"x$enable_utrace\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_UTRACE \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-valgrind was given.\nif test \"${enable_valgrind+set}\" = set; then :\n enableval=$enable_valgrind; if test \"x$enable_valgrind\" = \"xno\" ; then\n enable_valgrind=\"0\"\nelse\n enable_valgrind=\"1\"\nfi\n\nelse\n enable_valgrind=\"1\"\n\nfi\n\nif test \"x$enable_valgrind\" = \"x1\" ; then\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether valgrind is compilable\" >&5\n$as_echo_n \"checking whether valgrind is compilable... \" >&6; }\nif ${je_cv_valgrind+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n\n#if !defined(VALGRIND_RESIZEINPLACE_BLOCK)\n# error \"Incompatible Valgrind version\"\n#endif\n\nint\nmain ()\n{\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_valgrind=yes\nelse\n je_cv_valgrind=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_valgrind\" >&5\n$as_echo \"$je_cv_valgrind\" >&6; }\n\n if test \"x${je_cv_valgrind}\" = \"xno\" ; then\n enable_valgrind=\"0\"\n fi\n if test \"x$enable_valgrind\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_VALGRIND \" >>confdefs.h\n\n fi\nfi\n\n\n# Check whether --enable-xmalloc was given.\nif test \"${enable_xmalloc+set}\" = set; then :\n enableval=$enable_xmalloc; if test \"x$enable_xmalloc\" = \"xno\" ; then\n enable_xmalloc=\"0\"\nelse\n enable_xmalloc=\"1\"\nfi\n\nelse\n enable_xmalloc=\"0\"\n\nfi\n\nif test \"x$enable_xmalloc\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_XMALLOC \" >>confdefs.h\n\nfi\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking STATIC_PAGE_SHIFT\" >&5\n$as_echo_n \"checking STATIC_PAGE_SHIFT... \" >&6; }\nif ${je_cv_static_page_shift+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n if test \"$cross_compiling\" = yes; then :\n { { $as_echo \"$as_me:${as_lineno-$LINENO}: error: in \\`$ac_pwd':\" >&5\n$as_echo \"$as_me: error: in \\`$ac_pwd':\" >&2;}\nas_fn_error $? \"cannot run test program while cross compiling\nSee \\`config.log' for more details\" \"$LINENO\" 5; }\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#ifdef _WIN32\n#include \n#else\n#include \n#endif\n#include \n\nint\nmain ()\n{\n\n long result;\n FILE *f;\n\n#ifdef _WIN32\n SYSTEM_INFO si;\n GetSystemInfo(&si);\n result = si.dwPageSize;\n#else\n result = sysconf(_SC_PAGESIZE);\n#endif\n if (result == -1) {\n\treturn 1;\n }\n result = ffsl(result) - 1;\n\n f = fopen(\"conftest.out\", \"w\");\n if (f == NULL) {\n\treturn 1;\n }\n fprintf(f, \"%u\\n\", result);\n fclose(f);\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_run \"$LINENO\"; then :\n je_cv_static_page_shift=`cat conftest.out`\nelse\n je_cv_static_page_shift=undefined\nfi\nrm -f core *.core core.conftest.* gmon.out bb.out conftest$ac_exeext \\\n conftest.$ac_objext conftest.beam conftest.$ac_ext\nfi\n\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_static_page_shift\" >&5\n$as_echo \"$je_cv_static_page_shift\" >&6; }\n\nif test \"x$je_cv_static_page_shift\" != \"xundefined\"; then\n cat >>confdefs.h <<_ACEOF\n#define STATIC_PAGE_SHIFT $je_cv_static_page_shift\n_ACEOF\n\nelse\n as_fn_error $? \"cannot determine value for STATIC_PAGE_SHIFT\" \"$LINENO\" 5\nfi\n\n\nif test -d \"${srcroot}.git\" ; then\n git describe --long --abbrev=40 > ${srcroot}VERSION\nfi\njemalloc_version=`cat ${srcroot}VERSION`\njemalloc_version_major=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print $1}'`\njemalloc_version_minor=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print $2}'`\njemalloc_version_bugfix=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print $3}'`\njemalloc_version_nrev=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print $4}'`\njemalloc_version_gid=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print $5}'`\n\n\n\n\n\n\n\n\nif test \"x$abi\" != \"xpecoff\" ; then\n for ac_header in pthread.h\ndo :\n ac_fn_c_check_header_mongrel \"$LINENO\" \"pthread.h\" \"ac_cv_header_pthread_h\" \"$ac_includes_default\"\nif test \"x$ac_cv_header_pthread_h\" = xyes; then :\n cat >>confdefs.h <<_ACEOF\n#define HAVE_PTHREAD_H 1\n_ACEOF\n\nelse\n as_fn_error $? \"pthread.h is missing\" \"$LINENO\" 5\nfi\n\ndone\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for pthread_create in -lpthread\" >&5\n$as_echo_n \"checking for pthread_create in -lpthread... \" >&6; }\nif ${ac_cv_lib_pthread_pthread_create+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_check_lib_save_LIBS=$LIBS\nLIBS=\"-lpthread $LIBS\"\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar pthread_create ();\nint\nmain ()\n{\nreturn pthread_create ();\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n ac_cv_lib_pthread_pthread_create=yes\nelse\n ac_cv_lib_pthread_pthread_create=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nLIBS=$ac_check_lib_save_LIBS\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_pthread_pthread_create\" >&5\n$as_echo \"$ac_cv_lib_pthread_pthread_create\" >&6; }\nif test \"x$ac_cv_lib_pthread_pthread_create\" = xyes; then :\n LIBS=\"$LIBS -lpthread\"\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for library containing pthread_create\" >&5\n$as_echo_n \"checking for library containing pthread_create... \" >&6; }\nif ${ac_cv_search_pthread_create+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_func_search_save_LIBS=$LIBS\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar pthread_create ();\nint\nmain ()\n{\nreturn pthread_create ();\n ;\n return 0;\n}\n_ACEOF\nfor ac_lib in '' ; do\n if test -z \"$ac_lib\"; then\n ac_res=\"none required\"\n else\n ac_res=-l$ac_lib\n LIBS=\"-l$ac_lib $ac_func_search_save_LIBS\"\n fi\n if ac_fn_c_try_link \"$LINENO\"; then :\n ac_cv_search_pthread_create=$ac_res\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext\n if ${ac_cv_search_pthread_create+:} false; then :\n break\nfi\ndone\nif ${ac_cv_search_pthread_create+:} false; then :\n\nelse\n ac_cv_search_pthread_create=no\nfi\nrm conftest.$ac_ext\nLIBS=$ac_func_search_save_LIBS\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_search_pthread_create\" >&5\n$as_echo \"$ac_cv_search_pthread_create\" >&6; }\nac_res=$ac_cv_search_pthread_create\nif test \"$ac_res\" != no; then :\n test \"$ac_res\" = \"none required\" || LIBS=\"$ac_res $LIBS\"\n\nelse\n as_fn_error $? \"libpthread is missing\" \"$LINENO\" 5\nfi\n\nfi\n\nfi\n\nCPPFLAGS=\"$CPPFLAGS -D_REENTRANT\"\n\nac_fn_c_check_func \"$LINENO\" \"_malloc_thread_cleanup\" \"ac_cv_func__malloc_thread_cleanup\"\nif test \"x$ac_cv_func__malloc_thread_cleanup\" = xyes; then :\n have__malloc_thread_cleanup=\"1\"\nelse\n have__malloc_thread_cleanup=\"0\"\n\nfi\n\nif test \"x$have__malloc_thread_cleanup\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_MALLOC_THREAD_CLEANUP \" >>confdefs.h\n\n force_tls=\"1\"\nfi\n\nac_fn_c_check_func \"$LINENO\" \"_pthread_mutex_init_calloc_cb\" \"ac_cv_func__pthread_mutex_init_calloc_cb\"\nif test \"x$ac_cv_func__pthread_mutex_init_calloc_cb\" = xyes; then :\n have__pthread_mutex_init_calloc_cb=\"1\"\nelse\n have__pthread_mutex_init_calloc_cb=\"0\"\n\nfi\n\nif test \"x$have__pthread_mutex_init_calloc_cb\" = \"x1\" ; then\n $as_echo \"#define JEMALLOC_MUTEX_INIT_CB 1\" >>confdefs.h\n\nfi\n\n# Check whether --enable-lazy_lock was given.\nif test \"${enable_lazy_lock+set}\" = set; then :\n enableval=$enable_lazy_lock; if test \"x$enable_lazy_lock\" = \"xno\" ; then\n enable_lazy_lock=\"0\"\nelse\n enable_lazy_lock=\"1\"\nfi\n\nelse\n enable_lazy_lock=\"0\"\n\nfi\n\nif test \"x$enable_lazy_lock\" = \"x0\" -a \"x${force_lazy_lock}\" = \"x1\" ; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: Forcing lazy-lock to avoid allocator/threading bootstrap issues\" >&5\n$as_echo \"Forcing lazy-lock to avoid allocator/threading bootstrap issues\" >&6; }\n enable_lazy_lock=\"1\"\nfi\nif test \"x$enable_lazy_lock\" = \"x1\" ; then\n if test \"x$abi\" != \"xpecoff\" ; then\n for ac_header in dlfcn.h\ndo :\n ac_fn_c_check_header_mongrel \"$LINENO\" \"dlfcn.h\" \"ac_cv_header_dlfcn_h\" \"$ac_includes_default\"\nif test \"x$ac_cv_header_dlfcn_h\" = xyes; then :\n cat >>confdefs.h <<_ACEOF\n#define HAVE_DLFCN_H 1\n_ACEOF\n\nelse\n as_fn_error $? \"dlfcn.h is missing\" \"$LINENO\" 5\nfi\n\ndone\n\n ac_fn_c_check_func \"$LINENO\" \"dlsym\" \"ac_cv_func_dlsym\"\nif test \"x$ac_cv_func_dlsym\" = xyes; then :\n\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking for dlsym in -ldl\" >&5\n$as_echo_n \"checking for dlsym in -ldl... \" >&6; }\nif ${ac_cv_lib_dl_dlsym+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n ac_check_lib_save_LIBS=$LIBS\nLIBS=\"-ldl $LIBS\"\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n/* Override any GCC internal prototype to avoid an error.\n Use char because int might match the return type of a GCC\n builtin and then its argument prototype would still apply. */\n#ifdef __cplusplus\nextern \"C\"\n#endif\nchar dlsym ();\nint\nmain ()\n{\nreturn dlsym ();\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n ac_cv_lib_dl_dlsym=yes\nelse\n ac_cv_lib_dl_dlsym=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nLIBS=$ac_check_lib_save_LIBS\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_lib_dl_dlsym\" >&5\n$as_echo \"$ac_cv_lib_dl_dlsym\" >&6; }\nif test \"x$ac_cv_lib_dl_dlsym\" = xyes; then :\n LIBS=\"$LIBS -ldl\"\nelse\n as_fn_error $? \"libdl is missing\" \"$LINENO\" 5\nfi\n\n\nfi\n\n fi\n $as_echo \"#define JEMALLOC_LAZY_LOCK \" >>confdefs.h\n\nfi\n\n\n# Check whether --enable-tls was given.\nif test \"${enable_tls+set}\" = set; then :\n enableval=$enable_tls; if test \"x$enable_tls\" = \"xno\" ; then\n enable_tls=\"0\"\nelse\n enable_tls=\"1\"\nfi\n\nelse\n enable_tls=\"1\"\n\nfi\n\nif test \"x${enable_tls}\" = \"x0\" -a \"x${force_tls}\" = \"x1\" ; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: Forcing TLS to avoid allocator/threading bootstrap issues\" >&5\n$as_echo \"Forcing TLS to avoid allocator/threading bootstrap issues\" >&6; }\n enable_tls=\"1\"\nfi\nif test \"x${enable_tls}\" = \"x1\" -a \"x${force_tls}\" = \"x0\" ; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: Forcing no TLS to avoid allocator/threading bootstrap issues\" >&5\n$as_echo \"Forcing no TLS to avoid allocator/threading bootstrap issues\" >&6; }\n enable_tls=\"0\"\nfi\nif test \"x${enable_tls}\" = \"x1\" ; then\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for TLS\" >&5\n$as_echo_n \"checking for TLS... \" >&6; }\ncat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n __thread int x;\n\nint\nmain ()\n{\n\n x = 42;\n\n return 0;\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: yes\" >&5\n$as_echo \"yes\" >&6; }\nelse\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: no\" >&5\n$as_echo \"no\" >&6; }\n enable_tls=\"0\"\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\n\nif test \"x${enable_tls}\" = \"x1\" ; then\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_TLS\n_ACEOF\n\nelif test \"x${force_tls}\" = \"x1\" ; then\n as_fn_error $? \"Failed to configure TLS, which is mandatory for correct function\" \"$LINENO\" 5\nfi\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether a program using ffsl is compilable\" >&5\n$as_echo_n \"checking whether a program using ffsl is compilable... \" >&6; }\nif ${je_cv_function_ffsl+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n\nint\nmain ()\n{\n\n\t{\n\t\tint rv = ffsl(0x08);\n\t}\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_function_ffsl=yes\nelse\n je_cv_function_ffsl=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_function_ffsl\" >&5\n$as_echo \"$je_cv_function_ffsl\" >&6; }\n\nif test \"x${je_cv_function_ffsl}\" != \"xyes\" ; then\n as_fn_error $? \"Cannot build without ffsl(3)\" \"$LINENO\" 5\nfi\n\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether atomic(9) is compilable\" >&5\n$as_echo_n \"checking whether atomic(9) is compilable... \" >&6; }\nif ${je_cv_atomic9+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n#include \n\nint\nmain ()\n{\n\n\t{\n\t\tuint32_t x32 = 0;\n\t\tvolatile uint32_t *x32p = &x32;\n\t\tatomic_fetchadd_32(x32p, 1);\n\t}\n\t{\n\t\tunsigned long xlong = 0;\n\t\tvolatile unsigned long *xlongp = &xlong;\n\t\tatomic_fetchadd_long(xlongp, 1);\n\t}\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_atomic9=yes\nelse\n je_cv_atomic9=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_atomic9\" >&5\n$as_echo \"$je_cv_atomic9\" >&6; }\n\nif test \"x${je_cv_atomic9}\" = \"xyes\" ; then\n $as_echo \"#define JEMALLOC_ATOMIC9 1\" >>confdefs.h\n\nfi\n\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether Darwin OSAtomic*() is compilable\" >&5\n$as_echo_n \"checking whether Darwin OSAtomic*() is compilable... \" >&6; }\nif ${je_cv_osatomic+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n\nint\nmain ()\n{\n\n\t{\n\t\tint32_t x32 = 0;\n\t\tvolatile int32_t *x32p = &x32;\n\t\tOSAtomicAdd32(1, x32p);\n\t}\n\t{\n\t\tint64_t x64 = 0;\n\t\tvolatile int64_t *x64p = &x64;\n\t\tOSAtomicAdd64(1, x64p);\n\t}\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_osatomic=yes\nelse\n je_cv_osatomic=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_osatomic\" >&5\n$as_echo \"$je_cv_osatomic\" >&6; }\n\nif test \"x${je_cv_osatomic}\" = \"xyes\" ; then\n $as_echo \"#define JEMALLOC_OSATOMIC \" >>confdefs.h\n\nfi\n\n\n\n\nif test \"x${je_cv_atomic9}\" != \"xyes\" -a \"x${je_cv_osatomic}\" != \"xyes\" ; then\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether to force 32-bit __sync_{add,sub}_and_fetch()\" >&5\n$as_echo_n \"checking whether to force 32-bit __sync_{add,sub}_and_fetch()... \" >&6; }\nif ${je_cv_sync_compare_and_swap_4+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n #include \n\nint\nmain ()\n{\n\n #ifndef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4\n {\n uint32_t x32 = 0;\n __sync_add_and_fetch(&x32, 42);\n __sync_sub_and_fetch(&x32, 1);\n }\n #else\n #error __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4 is defined, no need to force\n #endif\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_sync_compare_and_swap_4=yes\nelse\n je_cv_sync_compare_and_swap_4=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_sync_compare_and_swap_4\" >&5\n$as_echo \"$je_cv_sync_compare_and_swap_4\" >&6; }\n\n if test \"x${je_cv_sync_compare_and_swap_4}\" = \"xyes\" ; then\n $as_echo \"#define JE_FORCE_SYNC_COMPARE_AND_SWAP_4 \" >>confdefs.h\n\n fi\n\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether to force 64-bit __sync_{add,sub}_and_fetch()\" >&5\n$as_echo_n \"checking whether to force 64-bit __sync_{add,sub}_and_fetch()... \" >&6; }\nif ${je_cv_sync_compare_and_swap_8+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n #include \n\nint\nmain ()\n{\n\n #ifndef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8\n {\n uint64_t x64 = 0;\n __sync_add_and_fetch(&x64, 42);\n __sync_sub_and_fetch(&x64, 1);\n }\n #else\n #error __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 is defined, no need to force\n #endif\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_sync_compare_and_swap_8=yes\nelse\n je_cv_sync_compare_and_swap_8=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_sync_compare_and_swap_8\" >&5\n$as_echo \"$je_cv_sync_compare_and_swap_8\" >&6; }\n\n if test \"x${je_cv_sync_compare_and_swap_8}\" = \"xyes\" ; then\n $as_echo \"#define JE_FORCE_SYNC_COMPARE_AND_SWAP_8 \" >>confdefs.h\n\n fi\n\nfi\n\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking whether Darwin OSSpin*() is compilable\" >&5\n$as_echo_n \"checking whether Darwin OSSpin*() is compilable... \" >&6; }\nif ${je_cv_osspin+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#include \n\nint\nmain ()\n{\n\n\tOSSpinLock lock = 0;\n\tOSSpinLockLock(&lock);\n\tOSSpinLockUnlock(&lock);\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_link \"$LINENO\"; then :\n je_cv_osspin=yes\nelse\n je_cv_osspin=no\nfi\nrm -f core conftest.err conftest.$ac_objext \\\n conftest$ac_exeext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $je_cv_osspin\" >&5\n$as_echo \"$je_cv_osspin\" >&6; }\n\nif test \"x${je_cv_osspin}\" = \"xyes\" ; then\n $as_echo \"#define JEMALLOC_OSSPIN \" >>confdefs.h\n\nfi\n\n\nif test \"x${abi}\" = \"xmacho\" ; then\n $as_echo \"#define JEMALLOC_IVSALLOC \" >>confdefs.h\n\n $as_echo \"#define JEMALLOC_ZONE \" >>confdefs.h\n\n\n { $as_echo \"$as_me:${as_lineno-$LINENO}: checking malloc zone version\" >&5\n$as_echo_n \"checking malloc zone version... \" >&6; }\n\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_zone_t) == sizeof(void *) * 14 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=3\nelse\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_zone_t) == sizeof(void *) * 15 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=5\nelse\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_zone_t) == sizeof(void *) * 16 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_introspection_t) == sizeof(void *) * 9 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=6\nelse\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_introspection_t) == sizeof(void *) * 13 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=7\nelse\n JEMALLOC_ZONE_VERSION=\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nelse\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_zone_t) == sizeof(void *) * 17 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=8\nelse\n\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n#include \nint\nmain ()\n{\nstatic foo[sizeof(malloc_zone_t) > sizeof(void *) * 17 ? 1 : -1]\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n JEMALLOC_ZONE_VERSION=9\nelse\n JEMALLOC_ZONE_VERSION=\n\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\n if test \"x${JEMALLOC_ZONE_VERSION}\" = \"x\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: unsupported\" >&5\n$as_echo \"unsupported\" >&6; }\n as_fn_error $? \"Unsupported malloc zone version\" \"$LINENO\" 5\n fi\n if test \"${JEMALLOC_ZONE_VERSION}\" = 9; then\n JEMALLOC_ZONE_VERSION=8\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: > 8\" >&5\n$as_echo \"> 8\" >&6; }\n else\n { $as_echo \"$as_me:${as_lineno-$LINENO}: result: $JEMALLOC_ZONE_VERSION\" >&5\n$as_echo \"$JEMALLOC_ZONE_VERSION\" >&6; }\n fi\n cat >>confdefs.h <<_ACEOF\n#define JEMALLOC_ZONE_VERSION $JEMALLOC_ZONE_VERSION\n_ACEOF\n\nfi\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: checking for stdbool.h that conforms to C99\" >&5\n$as_echo_n \"checking for stdbool.h that conforms to C99... \" >&6; }\nif ${ac_cv_header_stdbool_h+:} false; then :\n $as_echo_n \"(cached) \" >&6\nelse\n cat confdefs.h - <<_ACEOF >conftest.$ac_ext\n/* end confdefs.h. */\n\n#include \n#ifndef bool\n \"error: bool is not defined\"\n#endif\n#ifndef false\n \"error: false is not defined\"\n#endif\n#if false\n \"error: false is not 0\"\n#endif\n#ifndef true\n \"error: true is not defined\"\n#endif\n#if true != 1\n \"error: true is not 1\"\n#endif\n#ifndef __bool_true_false_are_defined\n \"error: __bool_true_false_are_defined is not defined\"\n#endif\n\n\tstruct s { _Bool s: 1; _Bool t; } s;\n\n\tchar a[true == 1 ? 1 : -1];\n\tchar b[false == 0 ? 1 : -1];\n\tchar c[__bool_true_false_are_defined == 1 ? 1 : -1];\n\tchar d[(bool) 0.5 == true ? 1 : -1];\n\t/* See body of main program for 'e'. */\n\tchar f[(_Bool) 0.0 == false ? 1 : -1];\n\tchar g[true];\n\tchar h[sizeof (_Bool)];\n\tchar i[sizeof s.t];\n\tenum { j = false, k = true, l = false * true, m = true * 256 };\n\t/* The following fails for\n\t HP aC++/ANSI C B3910B A.05.55 [Dec 04 2003]. */\n\t_Bool n[m];\n\tchar o[sizeof n == m * sizeof n[0] ? 1 : -1];\n\tchar p[-1 - (_Bool) 0 < 0 && -1 - (bool) 0 < 0 ? 1 : -1];\n\t/* Catch a bug in an HP-UX C compiler. See\n\t http://gcc.gnu.org/ml/gcc-patches/2003-12/msg02303.html\n\t http://lists.gnu.org/archive/html/bug-coreutils/2005-11/msg00161.html\n\t */\n\t_Bool q = true;\n\t_Bool *pq = &q;\n\nint\nmain ()\n{\n\n\tbool e = &s;\n\t*pq |= q;\n\t*pq |= ! q;\n\t/* Refer to every declared value, to avoid compiler optimizations. */\n\treturn (!a + !b + !c + !d + !e + !f + !g + !h + !i + !!j + !k + !!l\n\t\t+ !m + !n + !o + !p + !q + !pq);\n\n ;\n return 0;\n}\n_ACEOF\nif ac_fn_c_try_compile \"$LINENO\"; then :\n ac_cv_header_stdbool_h=yes\nelse\n ac_cv_header_stdbool_h=no\nfi\nrm -f core conftest.err conftest.$ac_objext conftest.$ac_ext\nfi\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: $ac_cv_header_stdbool_h\" >&5\n$as_echo \"$ac_cv_header_stdbool_h\" >&6; }\nac_fn_c_check_type \"$LINENO\" \"_Bool\" \"ac_cv_type__Bool\" \"$ac_includes_default\"\nif test \"x$ac_cv_type__Bool\" = xyes; then :\n\ncat >>confdefs.h <<_ACEOF\n#define HAVE__BOOL 1\n_ACEOF\n\n\nfi\n\nif test $ac_cv_header_stdbool_h = yes; then\n\n$as_echo \"#define HAVE_STDBOOL_H 1\" >>confdefs.h\n\nfi\n\n\nac_config_commands=\"$ac_config_commands include/jemalloc/internal/size_classes.h\"\n\n\n\n\nac_config_headers=\"$ac_config_headers $cfghdrs_tup\"\n\n\nac_config_files=\"$ac_config_files $cfgoutputs_tup config.stamp bin/jemalloc.sh\"\n\n\n\ncat >confcache <<\\_ACEOF\n# This file is a shell script that caches the results of configure\n# tests run on this system so they can be shared between configure\n# scripts and configure runs, see configure's option --config-cache.\n# It is not useful on other systems. If it contains results you don't\n# want to keep, you may remove or edit it.\n#\n# config.status only pays attention to the cache file if you give it\n# the --recheck option to rerun configure.\n#\n# `ac_cv_env_foo' variables (set or unset) will be overridden when\n# loading this file, other *unset* `ac_cv_foo' will be assigned the\n# following values.\n\n_ACEOF\n\n# The following way of writing the cache mishandles newlines in values,\n# but we know of no workaround that is simple, portable, and efficient.\n# So, we kill variables containing newlines.\n# Ultrix sh set writes to stderr and can't be redirected directly,\n# and sets the high bit in the cache file unless we assign to the vars.\n(\n for ac_var in `(set) 2>&1 | sed -n 's/^\\([a-zA-Z_][a-zA-Z0-9_]*\\)=.*/\\1/p'`; do\n eval ac_val=\\$$ac_var\n case $ac_val in #(\n *${as_nl}*)\n case $ac_var in #(\n *_cv_*) { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: cache variable $ac_var contains a newline\" >&5\n$as_echo \"$as_me: WARNING: cache variable $ac_var contains a newline\" >&2;} ;;\n esac\n case $ac_var in #(\n _ | IFS | as_nl) ;; #(\n BASH_ARGV | BASH_SOURCE) eval $ac_var= ;; #(\n *) { eval $ac_var=; unset $ac_var;} ;;\n esac ;;\n esac\n done\n\n (set) 2>&1 |\n case $as_nl`(ac_space=' '; set) 2>&1` in #(\n *${as_nl}ac_space=\\ *)\n # `set' does not quote correctly, so add quotes: double-quote\n # substitution turns \\\\\\\\ into \\\\, and sed turns \\\\ into \\.\n sed -n \\\n\t\"s/'/'\\\\\\\\''/g;\n\t s/^\\\\([_$as_cr_alnum]*_cv_[_$as_cr_alnum]*\\\\)=\\\\(.*\\\\)/\\\\1='\\\\2'/p\"\n ;; #(\n *)\n # `set' quotes correctly as required by POSIX, so do not add quotes.\n sed -n \"/^[_$as_cr_alnum]*_cv_[_$as_cr_alnum]*=/p\"\n ;;\n esac |\n sort\n) |\n sed '\n /^ac_cv_env_/b end\n t clear\n :clear\n s/^\\([^=]*\\)=\\(.*[{}].*\\)$/test \"${\\1+set}\" = set || &/\n t end\n s/^\\([^=]*\\)=\\(.*\\)$/\\1=${\\1=\\2}/\n :end' >>confcache\nif diff \"$cache_file\" confcache >/dev/null 2>&1; then :; else\n if test -w \"$cache_file\"; then\n if test \"x$cache_file\" != \"x/dev/null\"; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: updating cache $cache_file\" >&5\n$as_echo \"$as_me: updating cache $cache_file\" >&6;}\n if test ! -f \"$cache_file\" || test -h \"$cache_file\"; then\n\tcat confcache >\"$cache_file\"\n else\n case $cache_file in #(\n */* | ?:*)\n\t mv -f confcache \"$cache_file\"$$ &&\n\t mv -f \"$cache_file\"$$ \"$cache_file\" ;; #(\n *)\n\t mv -f confcache \"$cache_file\" ;;\n\tesac\n fi\n fi\n else\n { $as_echo \"$as_me:${as_lineno-$LINENO}: not updating unwritable cache $cache_file\" >&5\n$as_echo \"$as_me: not updating unwritable cache $cache_file\" >&6;}\n fi\nfi\nrm -f confcache\n\ntest \"x$prefix\" = xNONE && prefix=$ac_default_prefix\n# Let make expand exec_prefix.\ntest \"x$exec_prefix\" = xNONE && exec_prefix='${prefix}'\n\nDEFS=-DHAVE_CONFIG_H\n\nac_libobjs=\nac_ltlibobjs=\nU=\nfor ac_i in : $LIBOBJS; do test \"x$ac_i\" = x: && continue\n # 1. Remove the extension, and $U if already installed.\n ac_script='s/\\$U\\././;s/\\.o$//;s/\\.obj$//'\n ac_i=`$as_echo \"$ac_i\" | sed \"$ac_script\"`\n # 2. Prepend LIBOBJDIR. When used with automake>=1.10 LIBOBJDIR\n # will be set to the directory where LIBOBJS objects are built.\n as_fn_append ac_libobjs \" \\${LIBOBJDIR}$ac_i\\$U.$ac_objext\"\n as_fn_append ac_ltlibobjs \" \\${LIBOBJDIR}$ac_i\"'$U.lo'\ndone\nLIBOBJS=$ac_libobjs\n\nLTLIBOBJS=$ac_ltlibobjs\n\n\n\n: \"${CONFIG_STATUS=./config.status}\"\nac_write_fail=0\nac_clean_files_save=$ac_clean_files\nac_clean_files=\"$ac_clean_files $CONFIG_STATUS\"\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: creating $CONFIG_STATUS\" >&5\n$as_echo \"$as_me: creating $CONFIG_STATUS\" >&6;}\nas_write_fail=0\ncat >$CONFIG_STATUS <<_ASEOF || as_write_fail=1\n#! $SHELL\n# Generated by $as_me.\n# Run this file to recreate the current configuration.\n# Compiler output produced by configure, useful for debugging\n# configure, is in config.log if it exists.\n\ndebug=false\nac_cs_recheck=false\nac_cs_silent=false\n\nSHELL=\\${CONFIG_SHELL-$SHELL}\nexport SHELL\n_ASEOF\ncat >>$CONFIG_STATUS <<\\_ASEOF || as_write_fail=1\n## -------------------- ##\n## M4sh Initialization. ##\n## -------------------- ##\n\n# Be more Bourne compatible\nDUALCASE=1; export DUALCASE # for MKS sh\nif test -n \"${ZSH_VERSION+set}\" && (emulate sh) >/dev/null 2>&1; then :\n emulate sh\n NULLCMD=:\n # Pre-4.2 versions of Zsh do word splitting on ${1+\"$@\"}, which\n # is contrary to our usage. Disable this feature.\n alias -g '${1+\"$@\"}'='\"$@\"'\n setopt NO_GLOB_SUBST\nelse\n case `(set -o) 2>/dev/null` in #(\n *posix*) :\n set -o posix ;; #(\n *) :\n ;;\nesac\nfi\n\n\nas_nl='\n'\nexport as_nl\n# Printing a long string crashes Solaris 7 /usr/bin/printf.\nas_echo='\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\'\nas_echo=$as_echo$as_echo$as_echo$as_echo$as_echo\nas_echo=$as_echo$as_echo$as_echo$as_echo$as_echo$as_echo\n# Prefer a ksh shell builtin over an external printf program on Solaris,\n# but without wasting forks for bash or zsh.\nif test -z \"$BASH_VERSION$ZSH_VERSION\" \\\n && (test \"X`print -r -- $as_echo`\" = \"X$as_echo\") 2>/dev/null; then\n as_echo='print -r --'\n as_echo_n='print -rn --'\nelif (test \"X`printf %s $as_echo`\" = \"X$as_echo\") 2>/dev/null; then\n as_echo='printf %s\\n'\n as_echo_n='printf %s'\nelse\n if test \"X`(/usr/ucb/echo -n -n $as_echo) 2>/dev/null`\" = \"X-n $as_echo\"; then\n as_echo_body='eval /usr/ucb/echo -n \"$1$as_nl\"'\n as_echo_n='/usr/ucb/echo -n'\n else\n as_echo_body='eval expr \"X$1\" : \"X\\\\(.*\\\\)\"'\n as_echo_n_body='eval\n arg=$1;\n case $arg in #(\n *\"$as_nl\"*)\n\texpr \"X$arg\" : \"X\\\\(.*\\\\)$as_nl\";\n\targ=`expr \"X$arg\" : \".*$as_nl\\\\(.*\\\\)\"`;;\n esac;\n expr \"X$arg\" : \"X\\\\(.*\\\\)\" | tr -d \"$as_nl\"\n '\n export as_echo_n_body\n as_echo_n='sh -c $as_echo_n_body as_echo'\n fi\n export as_echo_body\n as_echo='sh -c $as_echo_body as_echo'\nfi\n\n# The user is always right.\nif test \"${PATH_SEPARATOR+set}\" != set; then\n PATH_SEPARATOR=:\n (PATH='/bin;/bin'; FPATH=$PATH; sh -c :) >/dev/null 2>&1 && {\n (PATH='/bin:/bin'; FPATH=$PATH; sh -c :) >/dev/null 2>&1 ||\n PATH_SEPARATOR=';'\n }\nfi\n\n\n# IFS\n# We need space, tab and new line, in precisely that order. Quoting is\n# there to prevent editors from complaining about space-tab.\n# (If _AS_PATH_WALK were called with IFS unset, it would disable word\n# splitting by setting IFS to empty value.)\nIFS=\" \"\"\t$as_nl\"\n\n# Find who we are. Look in the path if we contain no directory separator.\nas_myself=\ncase $0 in #((\n *[\\\\/]* ) as_myself=$0 ;;\n *) as_save_IFS=$IFS; IFS=$PATH_SEPARATOR\nfor as_dir in $PATH\ndo\n IFS=$as_save_IFS\n test -z \"$as_dir\" && as_dir=.\n test -r \"$as_dir/$0\" && as_myself=$as_dir/$0 && break\n done\nIFS=$as_save_IFS\n\n ;;\nesac\n# We did not find ourselves, most probably we were run as `sh COMMAND'\n# in which case we are not to be found in the path.\nif test \"x$as_myself\" = x; then\n as_myself=$0\nfi\nif test ! -f \"$as_myself\"; then\n $as_echo \"$as_myself: error: cannot find myself; rerun with an absolute file name\" >&2\n exit 1\nfi\n\n# Unset variables that we do not need and which cause bugs (e.g. in\n# pre-3.0 UWIN ksh). But do not cause bugs in bash 2.01; the \"|| exit 1\"\n# suppresses any \"Segmentation fault\" message there. '((' could\n# trigger a bug in pdksh 5.2.14.\nfor as_var in BASH_ENV ENV MAIL MAILPATH\ndo eval test x\\${$as_var+set} = xset \\\n && ( (unset $as_var) || exit 1) >/dev/null 2>&1 && unset $as_var || :\ndone\nPS1='$ '\nPS2='> '\nPS4='+ '\n\n# NLS nuisances.\nLC_ALL=C\nexport LC_ALL\nLANGUAGE=C\nexport LANGUAGE\n\n# CDPATH.\n(unset CDPATH) >/dev/null 2>&1 && unset CDPATH\n\n\n# as_fn_error STATUS ERROR [LINENO LOG_FD]\n# ----------------------------------------\n# Output \"`basename $0`: error: ERROR\" to stderr. If LINENO and LOG_FD are\n# provided, also output the error to LOG_FD, referencing LINENO. Then exit the\n# script with STATUS, using 1 if that was 0.\nas_fn_error ()\n{\n as_status=$1; test $as_status -eq 0 && as_status=1\n if test \"$4\"; then\n as_lineno=${as_lineno-\"$3\"} as_lineno_stack=as_lineno_stack=$as_lineno_stack\n $as_echo \"$as_me:${as_lineno-$LINENO}: error: $2\" >&$4\n fi\n $as_echo \"$as_me: error: $2\" >&2\n as_fn_exit $as_status\n} # as_fn_error\n\n\n# as_fn_set_status STATUS\n# -----------------------\n# Set $? to STATUS, without forking.\nas_fn_set_status ()\n{\n return $1\n} # as_fn_set_status\n\n# as_fn_exit STATUS\n# -----------------\n# Exit the shell with STATUS, even in a \"trap 0\" or \"set -e\" context.\nas_fn_exit ()\n{\n set +e\n as_fn_set_status $1\n exit $1\n} # as_fn_exit\n\n# as_fn_unset VAR\n# ---------------\n# Portably unset VAR.\nas_fn_unset ()\n{\n { eval $1=; unset $1;}\n}\nas_unset=as_fn_unset\n# as_fn_append VAR VALUE\n# ----------------------\n# Append the text in VALUE to the end of the definition contained in VAR. Take\n# advantage of any shell optimizations that allow amortized linear growth over\n# repeated appends, instead of the typical quadratic growth present in naive\n# implementations.\nif (eval \"as_var=1; as_var+=2; test x\\$as_var = x12\") 2>/dev/null; then :\n eval 'as_fn_append ()\n {\n eval $1+=\\$2\n }'\nelse\n as_fn_append ()\n {\n eval $1=\\$$1\\$2\n }\nfi # as_fn_append\n\n# as_fn_arith ARG...\n# ------------------\n# Perform arithmetic evaluation on the ARGs, and store the result in the\n# global $as_val. Take advantage of shells that can avoid forks. The arguments\n# must be portable across $(()) and expr.\nif (eval \"test \\$(( 1 + 1 )) = 2\") 2>/dev/null; then :\n eval 'as_fn_arith ()\n {\n as_val=$(( $* ))\n }'\nelse\n as_fn_arith ()\n {\n as_val=`expr \"$@\" || test $? -eq 1`\n }\nfi # as_fn_arith\n\n\nif expr a : '\\(a\\)' >/dev/null 2>&1 &&\n test \"X`expr 00001 : '.*\\(...\\)'`\" = X001; then\n as_expr=expr\nelse\n as_expr=false\nfi\n\nif (basename -- /) >/dev/null 2>&1 && test \"X`basename -- / 2>&1`\" = \"X/\"; then\n as_basename=basename\nelse\n as_basename=false\nfi\n\nif (as_dir=`dirname -- /` && test \"X$as_dir\" = X/) >/dev/null 2>&1; then\n as_dirname=dirname\nelse\n as_dirname=false\nfi\n\nas_me=`$as_basename -- \"$0\" ||\n$as_expr X/\"$0\" : '.*/\\([^/][^/]*\\)/*$' \\| \\\n\t X\"$0\" : 'X\\(//\\)$' \\| \\\n\t X\"$0\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X/\"$0\" |\n sed '/^.*\\/\\([^/][^/]*\\)\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\/\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\/\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n\n# Avoid depending upon Character Ranges.\nas_cr_letters='abcdefghijklmnopqrstuvwxyz'\nas_cr_LETTERS='ABCDEFGHIJKLMNOPQRSTUVWXYZ'\nas_cr_Letters=$as_cr_letters$as_cr_LETTERS\nas_cr_digits='0123456789'\nas_cr_alnum=$as_cr_Letters$as_cr_digits\n\nECHO_C= ECHO_N= ECHO_T=\ncase `echo -n x` in #(((((\n-n*)\n case `echo 'xy\\c'` in\n *c*) ECHO_T='\t';;\t# ECHO_T is single tab character.\n xy) ECHO_C='\\c';;\n *) echo `echo ksh88 bug on AIX 6.1` > /dev/null\n ECHO_T='\t';;\n esac;;\n*)\n ECHO_N='-n';;\nesac\n\nrm -f conf$$ conf$$.exe conf$$.file\nif test -d conf$$.dir; then\n rm -f conf$$.dir/conf$$.file\nelse\n rm -f conf$$.dir\n mkdir conf$$.dir 2>/dev/null\nfi\nif (echo >conf$$.file) 2>/dev/null; then\n if ln -s conf$$.file conf$$ 2>/dev/null; then\n as_ln_s='ln -s'\n # ... but there are two gotchas:\n # 1) On MSYS, both `ln -s file dir' and `ln file dir' fail.\n # 2) DJGPP < 2.04 has no symlinks; `ln -s' creates a wrapper executable.\n # In both cases, we have to default to `cp -p'.\n ln -s conf$$.file conf$$.dir 2>/dev/null && test ! -f conf$$.exe ||\n as_ln_s='cp -p'\n elif ln conf$$.file conf$$ 2>/dev/null; then\n as_ln_s=ln\n else\n as_ln_s='cp -p'\n fi\nelse\n as_ln_s='cp -p'\nfi\nrm -f conf$$ conf$$.exe conf$$.dir/conf$$.file conf$$.file\nrmdir conf$$.dir 2>/dev/null\n\n\n# as_fn_mkdir_p\n# -------------\n# Create \"$as_dir\" as a directory, including parents if necessary.\nas_fn_mkdir_p ()\n{\n\n case $as_dir in #(\n -*) as_dir=./$as_dir;;\n esac\n test -d \"$as_dir\" || eval $as_mkdir_p || {\n as_dirs=\n while :; do\n case $as_dir in #(\n *\\'*) as_qdir=`$as_echo \"$as_dir\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"`;; #'(\n *) as_qdir=$as_dir;;\n esac\n as_dirs=\"'$as_qdir' $as_dirs\"\n as_dir=`$as_dirname -- \"$as_dir\" ||\n$as_expr X\"$as_dir\" : 'X\\(.*[^/]\\)//*[^/][^/]*/*$' \\| \\\n\t X\"$as_dir\" : 'X\\(//\\)[^/]' \\| \\\n\t X\"$as_dir\" : 'X\\(//\\)$' \\| \\\n\t X\"$as_dir\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X\"$as_dir\" |\n sed '/^X\\(.*[^/]\\)\\/\\/*[^/][^/]*\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)[^/].*/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n test -d \"$as_dir\" && break\n done\n test -z \"$as_dirs\" || eval \"mkdir $as_dirs\"\n } || test -d \"$as_dir\" || as_fn_error $? \"cannot create directory $as_dir\"\n\n\n} # as_fn_mkdir_p\nif mkdir -p . 2>/dev/null; then\n as_mkdir_p='mkdir -p \"$as_dir\"'\nelse\n test -d ./-p && rmdir ./-p\n as_mkdir_p=false\nfi\n\nif test -x / >/dev/null 2>&1; then\n as_test_x='test -x'\nelse\n if ls -dL / >/dev/null 2>&1; then\n as_ls_L_option=L\n else\n as_ls_L_option=\n fi\n as_test_x='\n eval sh -c '\\''\n if test -d \"$1\"; then\n\ttest -d \"$1/.\";\n else\n\tcase $1 in #(\n\t-*)set \"./$1\";;\n\tesac;\n\tcase `ls -ld'$as_ls_L_option' \"$1\" 2>/dev/null` in #((\n\t???[sx]*):;;*)false;;esac;fi\n '\\'' sh\n '\nfi\nas_executable_p=$as_test_x\n\n# Sed expression to map a string onto a valid CPP name.\nas_tr_cpp=\"eval sed 'y%*$as_cr_letters%P$as_cr_LETTERS%;s%[^_$as_cr_alnum]%_%g'\"\n\n# Sed expression to map a string onto a valid variable name.\nas_tr_sh=\"eval sed 'y%*+%pp%;s%[^_$as_cr_alnum]%_%g'\"\n\n\nexec 6>&1\n## ----------------------------------- ##\n## Main body of $CONFIG_STATUS script. ##\n## ----------------------------------- ##\n_ASEOF\ntest $as_write_fail = 0 && chmod +x $CONFIG_STATUS || ac_write_fail=1\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n# Save the log message, to keep $0 and so on meaningful, and to\n# report actual input values of CONFIG_FILES etc. instead of their\n# values after options handling.\nac_log=\"\nThis file was extended by $as_me, which was\ngenerated by GNU Autoconf 2.68. Invocation command line was\n\n CONFIG_FILES = $CONFIG_FILES\n CONFIG_HEADERS = $CONFIG_HEADERS\n CONFIG_LINKS = $CONFIG_LINKS\n CONFIG_COMMANDS = $CONFIG_COMMANDS\n $ $0 $@\n\non `(hostname || uname -n) 2>/dev/null | sed 1q`\n\"\n\n_ACEOF\n\ncase $ac_config_files in *\"\n\"*) set x $ac_config_files; shift; ac_config_files=$*;;\nesac\n\ncase $ac_config_headers in *\"\n\"*) set x $ac_config_headers; shift; ac_config_headers=$*;;\nesac\n\n\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\n# Files that config.status was made for.\nconfig_files=\"$ac_config_files\"\nconfig_headers=\"$ac_config_headers\"\nconfig_commands=\"$ac_config_commands\"\n\n_ACEOF\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\nac_cs_usage=\"\\\n\\`$as_me' instantiates files and other configuration actions\nfrom templates according to the current configuration. Unless the files\nand actions are specified as TAGs, all are instantiated by default.\n\nUsage: $0 [OPTION]... [TAG]...\n\n -h, --help print this help, then exit\n -V, --version print version number and configuration settings, then exit\n --config print configuration, then exit\n -q, --quiet, --silent\n do not print progress messages\n -d, --debug don't remove temporary files\n --recheck update $as_me by reconfiguring in the same conditions\n --file=FILE[:TEMPLATE]\n instantiate the configuration file FILE\n --header=FILE[:TEMPLATE]\n instantiate the configuration header FILE\n\nConfiguration files:\n$config_files\n\nConfiguration headers:\n$config_headers\n\nConfiguration commands:\n$config_commands\n\nReport bugs to the package provider.\"\n\n_ACEOF\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\nac_cs_config=\"`$as_echo \"$ac_configure_args\" | sed 's/^ //; s/[\\\\\"\"\\`\\$]/\\\\\\\\&/g'`\"\nac_cs_version=\"\\\\\nconfig.status\nconfigured by $0, generated by GNU Autoconf 2.68,\n with options \\\\\"\\$ac_cs_config\\\\\"\n\nCopyright (C) 2010 Free Software Foundation, Inc.\nThis config.status script is free software; the Free Software Foundation\ngives unlimited permission to copy, distribute and modify it.\"\n\nac_pwd='$ac_pwd'\nsrcdir='$srcdir'\nINSTALL='$INSTALL'\ntest -n \"\\$AWK\" || AWK=awk\n_ACEOF\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n# The default lists apply if the user does not specify any file.\nac_need_defaults=:\nwhile test $# != 0\ndo\n case $1 in\n --*=?*)\n ac_option=`expr \"X$1\" : 'X\\([^=]*\\)='`\n ac_optarg=`expr \"X$1\" : 'X[^=]*=\\(.*\\)'`\n ac_shift=:\n ;;\n --*=)\n ac_option=`expr \"X$1\" : 'X\\([^=]*\\)='`\n ac_optarg=\n ac_shift=:\n ;;\n *)\n ac_option=$1\n ac_optarg=$2\n ac_shift=shift\n ;;\n esac\n\n case $ac_option in\n # Handling of the options.\n -recheck | --recheck | --rechec | --reche | --rech | --rec | --re | --r)\n ac_cs_recheck=: ;;\n --version | --versio | --versi | --vers | --ver | --ve | --v | -V )\n $as_echo \"$ac_cs_version\"; exit ;;\n --config | --confi | --conf | --con | --co | --c )\n $as_echo \"$ac_cs_config\"; exit ;;\n --debug | --debu | --deb | --de | --d | -d )\n debug=: ;;\n --file | --fil | --fi | --f )\n $ac_shift\n case $ac_optarg in\n *\\'*) ac_optarg=`$as_echo \"$ac_optarg\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"` ;;\n '') as_fn_error $? \"missing file argument\" ;;\n esac\n as_fn_append CONFIG_FILES \" '$ac_optarg'\"\n ac_need_defaults=false;;\n --header | --heade | --head | --hea )\n $ac_shift\n case $ac_optarg in\n *\\'*) ac_optarg=`$as_echo \"$ac_optarg\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"` ;;\n esac\n as_fn_append CONFIG_HEADERS \" '$ac_optarg'\"\n ac_need_defaults=false;;\n --he | --h)\n # Conflict between --help and --header\n as_fn_error $? \"ambiguous option: \\`$1'\nTry \\`$0 --help' for more information.\";;\n --help | --hel | -h )\n $as_echo \"$ac_cs_usage\"; exit ;;\n -q | -quiet | --quiet | --quie | --qui | --qu | --q \\\n | -silent | --silent | --silen | --sile | --sil | --si | --s)\n ac_cs_silent=: ;;\n\n # This is an error.\n -*) as_fn_error $? \"unrecognized option: \\`$1'\nTry \\`$0 --help' for more information.\" ;;\n\n *) as_fn_append ac_config_targets \" $1\"\n ac_need_defaults=false ;;\n\n esac\n shift\ndone\n\nac_configure_extra_args=\n\nif $ac_cs_silent; then\n exec 6>/dev/null\n ac_configure_extra_args=\"$ac_configure_extra_args --silent\"\nfi\n\n_ACEOF\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\nif \\$ac_cs_recheck; then\n set X '$SHELL' '$0' $ac_configure_args \\$ac_configure_extra_args --no-create --no-recursion\n shift\n \\$as_echo \"running CONFIG_SHELL=$SHELL \\$*\" >&6\n CONFIG_SHELL='$SHELL'\n export CONFIG_SHELL\n exec \"\\$@\"\nfi\n\n_ACEOF\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\nexec 5>>config.log\n{\n echo\n sed 'h;s/./-/g;s/^.../## /;s/...$/ ##/;p;x;p;x' <<_ASBOX\n## Running $as_me. ##\n_ASBOX\n $as_echo \"$ac_log\"\n} >&5\n\n_ACEOF\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\n_ACEOF\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n\n# Handling of arguments.\nfor ac_config_target in $ac_config_targets\ndo\n case $ac_config_target in\n \"include/jemalloc/internal/size_classes.h\") CONFIG_COMMANDS=\"$CONFIG_COMMANDS include/jemalloc/internal/size_classes.h\" ;;\n \"$cfghdrs_tup\") CONFIG_HEADERS=\"$CONFIG_HEADERS $cfghdrs_tup\" ;;\n \"$cfgoutputs_tup\") CONFIG_FILES=\"$CONFIG_FILES $cfgoutputs_tup\" ;;\n \"config.stamp\") CONFIG_FILES=\"$CONFIG_FILES config.stamp\" ;;\n \"bin/jemalloc.sh\") CONFIG_FILES=\"$CONFIG_FILES bin/jemalloc.sh\" ;;\n\n *) as_fn_error $? \"invalid argument: \\`$ac_config_target'\" \"$LINENO\" 5;;\n esac\ndone\n\n\n# If the user did not use the arguments to specify the items to instantiate,\n# then the envvar interface is used. Set only those that are not.\n# We use the long form for the default assignment because of an extremely\n# bizarre bug on SunOS 4.1.3.\nif $ac_need_defaults; then\n test \"${CONFIG_FILES+set}\" = set || CONFIG_FILES=$config_files\n test \"${CONFIG_HEADERS+set}\" = set || CONFIG_HEADERS=$config_headers\n test \"${CONFIG_COMMANDS+set}\" = set || CONFIG_COMMANDS=$config_commands\nfi\n\n# Have a temporary directory for convenience. Make it in the build tree\n# simply because there is no reason against having it here, and in addition,\n# creating and moving files from /tmp can sometimes cause problems.\n# Hook for its removal unless debugging.\n# Note that there is a small window in which the directory will not be cleaned:\n# after its creation but before its name has been assigned to `$tmp'.\n$debug ||\n{\n tmp= ac_tmp=\n trap 'exit_status=$?\n : \"${ac_tmp:=$tmp}\"\n { test ! -d \"$ac_tmp\" || rm -fr \"$ac_tmp\"; } && exit $exit_status\n' 0\n trap 'as_fn_exit 1' 1 2 13 15\n}\n# Create a (secure) tmp directory for tmp files.\n\n{\n tmp=`(umask 077 && mktemp -d \"./confXXXXXX\") 2>/dev/null` &&\n test -d \"$tmp\"\n} ||\n{\n tmp=./conf$$-$RANDOM\n (umask 077 && mkdir \"$tmp\")\n} || as_fn_error $? \"cannot create a temporary directory in .\" \"$LINENO\" 5\nac_tmp=$tmp\n\n# Set up the scripts for CONFIG_FILES section.\n# No need to generate them if there are no CONFIG_FILES.\n# This happens for instance with `./config.status config.h'.\nif test -n \"$CONFIG_FILES\"; then\n\n\nac_cr=`echo X | tr X '\\015'`\n# On cygwin, bash can eat \\r inside `` if the user requested igncr.\n# But we know of no other shell where ac_cr would be empty at this\n# point, so we can use a bashism as a fallback.\nif test \"x$ac_cr\" = x; then\n eval ac_cr=\\$\\'\\\\r\\'\nfi\nac_cs_awk_cr=`$AWK 'BEGIN { print \"a\\rb\" }' /dev/null`\nif test \"$ac_cs_awk_cr\" = \"a${ac_cr}b\"; then\n ac_cs_awk_cr='\\\\r'\nelse\n ac_cs_awk_cr=$ac_cr\nfi\n\necho 'BEGIN {' >\"$ac_tmp/subs1.awk\" &&\n_ACEOF\n\n\n{\n echo \"cat >conf$$subs.awk <<_ACEOF\" &&\n echo \"$ac_subst_vars\" | sed 's/.*/&!$&$ac_delim/' &&\n echo \"_ACEOF\"\n} >conf$$subs.sh ||\n as_fn_error $? \"could not make $CONFIG_STATUS\" \"$LINENO\" 5\nac_delim_num=`echo \"$ac_subst_vars\" | grep -c '^'`\nac_delim='%!_!# '\nfor ac_last_try in false false false false false :; do\n . ./conf$$subs.sh ||\n as_fn_error $? \"could not make $CONFIG_STATUS\" \"$LINENO\" 5\n\n ac_delim_n=`sed -n \"s/.*$ac_delim\\$/X/p\" conf$$subs.awk | grep -c X`\n if test $ac_delim_n = $ac_delim_num; then\n break\n elif $ac_last_try; then\n as_fn_error $? \"could not make $CONFIG_STATUS\" \"$LINENO\" 5\n else\n ac_delim=\"$ac_delim!$ac_delim _$ac_delim!! \"\n fi\ndone\nrm -f conf$$subs.sh\n\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\ncat >>\"\\$ac_tmp/subs1.awk\" <<\\\\_ACAWK &&\n_ACEOF\nsed -n '\nh\ns/^/S[\"/; s/!.*/\"]=/\np\ng\ns/^[^!]*!//\n:repl\nt repl\ns/'\"$ac_delim\"'$//\nt delim\n:nl\nh\ns/\\(.\\{148\\}\\)..*/\\1/\nt more1\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\\\\n\"\\\\/\np\nn\nb repl\n:more1\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\"\\\\/\np\ng\ns/.\\{148\\}//\nt nl\n:delim\nh\ns/\\(.\\{148\\}\\)..*/\\1/\nt more2\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\"/\np\nb\n:more2\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\"\\\\/\np\ng\ns/.\\{148\\}//\nt delim\n' >$CONFIG_STATUS || ac_write_fail=1\nrm -f conf$$subs.awk\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\n_ACAWK\ncat >>\"\\$ac_tmp/subs1.awk\" <<_ACAWK &&\n for (key in S) S_is_set[key] = 1\n FS = \"\u0007\"\n\n}\n{\n line = $ 0\n nfields = split(line, field, \"@\")\n substed = 0\n len = length(field[1])\n for (i = 2; i < nfields; i++) {\n key = field[i]\n keylen = length(key)\n if (S_is_set[key]) {\n value = S[key]\n line = substr(line, 1, len) \"\" value \"\" substr(line, len + keylen + 3)\n len += length(value) + length(field[++i])\n substed = 1\n } else\n len += 1 + keylen\n }\n\n print line\n}\n\n_ACAWK\n_ACEOF\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\nif sed \"s/$ac_cr//\" < /dev/null > /dev/null 2>&1; then\n sed \"s/$ac_cr\\$//; s/$ac_cr/$ac_cs_awk_cr/g\"\nelse\n cat\nfi < \"$ac_tmp/subs1.awk\" > \"$ac_tmp/subs.awk\" \\\n || as_fn_error $? \"could not setup config files machinery\" \"$LINENO\" 5\n_ACEOF\n\n# VPATH may cause trouble with some makes, so we remove sole $(srcdir),\n# ${srcdir} and @srcdir@ entries from VPATH if srcdir is \".\", strip leading and\n# trailing colons and then remove the whole line if VPATH becomes empty\n# (actually we leave an empty line to preserve line numbers).\nif test \"x$srcdir\" = x.; then\n ac_vpsub='/^[\t ]*VPATH[\t ]*=[\t ]*/{\nh\ns///\ns/^/:/\ns/[\t ]*$/:/\ns/:\\$(srcdir):/:/g\ns/:\\${srcdir}:/:/g\ns/:@srcdir@:/:/g\ns/^:*//\ns/:*$//\nx\ns/\\(=[\t ]*\\).*/\\1/\nG\ns/\\n//\ns/^[^=]*=[\t ]*$//\n}'\nfi\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\nfi # test -n \"$CONFIG_FILES\"\n\n# Set up the scripts for CONFIG_HEADERS section.\n# No need to generate them if there are no CONFIG_HEADERS.\n# This happens for instance with `./config.status Makefile'.\nif test -n \"$CONFIG_HEADERS\"; then\ncat >\"$ac_tmp/defines.awk\" <<\\_ACAWK ||\nBEGIN {\n_ACEOF\n\n# Transform confdefs.h into an awk script `defines.awk', embedded as\n# here-document in config.status, that substitutes the proper values into\n# config.h.in to produce config.h.\n\n# Create a delimiter string that does not exist in confdefs.h, to ease\n# handling of long lines.\nac_delim='%!_!# '\nfor ac_last_try in false false :; do\n ac_tt=`sed -n \"/$ac_delim/p\" confdefs.h`\n if test -z \"$ac_tt\"; then\n break\n elif $ac_last_try; then\n as_fn_error $? \"could not make $CONFIG_HEADERS\" \"$LINENO\" 5\n else\n ac_delim=\"$ac_delim!$ac_delim _$ac_delim!! \"\n fi\ndone\n\n# For the awk script, D is an array of macro values keyed by name,\n# likewise P contains macro parameters if any. Preserve backslash\n# newline sequences.\n\nac_word_re=[_$as_cr_Letters][_$as_cr_alnum]*\nsed -n '\ns/.\\{148\\}/&'\"$ac_delim\"'/g\nt rset\n:rset\ns/^[\t ]*#[\t ]*define[\t ][\t ]*/ /\nt def\nd\n:def\ns/\\\\$//\nt bsnl\ns/[\"\\\\]/\\\\&/g\ns/^ \\('\"$ac_word_re\"'\\)\\(([^()]*)\\)[\t ]*\\(.*\\)/P[\"\\1\"]=\"\\2\"\\\nD[\"\\1\"]=\" \\3\"/p\ns/^ \\('\"$ac_word_re\"'\\)[\t ]*\\(.*\\)/D[\"\\1\"]=\" \\2\"/p\nd\n:bsnl\ns/[\"\\\\]/\\\\&/g\ns/^ \\('\"$ac_word_re\"'\\)\\(([^()]*)\\)[\t ]*\\(.*\\)/P[\"\\1\"]=\"\\2\"\\\nD[\"\\1\"]=\" \\3\\\\\\\\\\\\n\"\\\\/p\nt cont\ns/^ \\('\"$ac_word_re\"'\\)[\t ]*\\(.*\\)/D[\"\\1\"]=\" \\2\\\\\\\\\\\\n\"\\\\/p\nt cont\nd\n:cont\nn\ns/.\\{148\\}/&'\"$ac_delim\"'/g\nt clear\n:clear\ns/\\\\$//\nt bsnlc\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\"/p\nd\n:bsnlc\ns/[\"\\\\]/\\\\&/g; s/^/\"/; s/$/\\\\\\\\\\\\n\"\\\\/p\nb cont\n' >$CONFIG_STATUS || ac_write_fail=1\n\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\n for (key in D) D_is_set[key] = 1\n FS = \"\u0007\"\n}\n/^[\\t ]*#[\\t ]*(define|undef)[\\t ]+$ac_word_re([\\t (]|\\$)/ {\n line = \\$ 0\n split(line, arg, \" \")\n if (arg[1] == \"#\") {\n defundef = arg[2]\n mac1 = arg[3]\n } else {\n defundef = substr(arg[1], 2)\n mac1 = arg[2]\n }\n split(mac1, mac2, \"(\") #)\n macro = mac2[1]\n prefix = substr(line, 1, index(line, defundef) - 1)\n if (D_is_set[macro]) {\n # Preserve the white space surrounding the \"#\".\n print prefix \"define\", macro P[macro] D[macro]\n next\n } else {\n # Replace #undef with comments. This is necessary, for example,\n # in the case of _POSIX_SOURCE, which is predefined and required\n # on some systems where configure will not decide to define it.\n if (defundef == \"undef\") {\n print \"/*\", prefix defundef, macro, \"*/\"\n next\n }\n }\n}\n{ print }\n_ACAWK\n_ACEOF\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n as_fn_error $? \"could not setup config headers machinery\" \"$LINENO\" 5\nfi # test -n \"$CONFIG_HEADERS\"\n\n\neval set X \" :F $CONFIG_FILES :H $CONFIG_HEADERS :C $CONFIG_COMMANDS\"\nshift\nfor ac_tag\ndo\n case $ac_tag in\n :[FHLC]) ac_mode=$ac_tag; continue;;\n esac\n case $ac_mode$ac_tag in\n :[FHL]*:*);;\n :L* | :C*:*) as_fn_error $? \"invalid tag \\`$ac_tag'\" \"$LINENO\" 5;;\n :[FH]-) ac_tag=-:-;;\n :[FH]*) ac_tag=$ac_tag:$ac_tag.in;;\n esac\n ac_save_IFS=$IFS\n IFS=:\n set x $ac_tag\n IFS=$ac_save_IFS\n shift\n ac_file=$1\n shift\n\n case $ac_mode in\n :L) ac_source=$1;;\n :[FH])\n ac_file_inputs=\n for ac_f\n do\n case $ac_f in\n -) ac_f=\"$ac_tmp/stdin\";;\n *) # Look for the file first in the build tree, then in the source tree\n\t # (if the path is not absolute). The absolute path cannot be DOS-style,\n\t # because $ac_f cannot contain `:'.\n\t test -f \"$ac_f\" ||\n\t case $ac_f in\n\t [\\\\/$]*) false;;\n\t *) test -f \"$srcdir/$ac_f\" && ac_f=\"$srcdir/$ac_f\";;\n\t esac ||\n\t as_fn_error 1 \"cannot find input file: \\`$ac_f'\" \"$LINENO\" 5;;\n esac\n case $ac_f in *\\'*) ac_f=`$as_echo \"$ac_f\" | sed \"s/'/'\\\\\\\\\\\\\\\\''/g\"`;; esac\n as_fn_append ac_file_inputs \" '$ac_f'\"\n done\n\n # Let's still pretend it is `configure' which instantiates (i.e., don't\n # use $as_me), people would be surprised to read:\n # /* config.h. Generated by config.status. */\n configure_input='Generated from '`\n\t $as_echo \"$*\" | sed 's|^[^:]*/||;s|:[^:]*/|, |g'\n\t`' by configure.'\n if test x\"$ac_file\" != x-; then\n configure_input=\"$ac_file. $configure_input\"\n { $as_echo \"$as_me:${as_lineno-$LINENO}: creating $ac_file\" >&5\n$as_echo \"$as_me: creating $ac_file\" >&6;}\n fi\n # Neutralize special characters interpreted by sed in replacement strings.\n case $configure_input in #(\n *\\&* | *\\|* | *\\\\* )\n ac_sed_conf_input=`$as_echo \"$configure_input\" |\n sed 's/[\\\\\\\\&|]/\\\\\\\\&/g'`;; #(\n *) ac_sed_conf_input=$configure_input;;\n esac\n\n case $ac_tag in\n *:-:* | *:-) cat >\"$ac_tmp/stdin\" \\\n || as_fn_error $? \"could not create $ac_file\" \"$LINENO\" 5 ;;\n esac\n ;;\n esac\n\n ac_dir=`$as_dirname -- \"$ac_file\" ||\n$as_expr X\"$ac_file\" : 'X\\(.*[^/]\\)//*[^/][^/]*/*$' \\| \\\n\t X\"$ac_file\" : 'X\\(//\\)[^/]' \\| \\\n\t X\"$ac_file\" : 'X\\(//\\)$' \\| \\\n\t X\"$ac_file\" : 'X\\(/\\)' \\| . 2>/dev/null ||\n$as_echo X\"$ac_file\" |\n sed '/^X\\(.*[^/]\\)\\/\\/*[^/][^/]*\\/*$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)[^/].*/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\/\\)$/{\n\t s//\\1/\n\t q\n\t }\n\t /^X\\(\\/\\).*/{\n\t s//\\1/\n\t q\n\t }\n\t s/.*/./; q'`\n as_dir=\"$ac_dir\"; as_fn_mkdir_p\n ac_builddir=.\n\ncase \"$ac_dir\" in\n.) ac_dir_suffix= ac_top_builddir_sub=. ac_top_build_prefix= ;;\n*)\n ac_dir_suffix=/`$as_echo \"$ac_dir\" | sed 's|^\\.[\\\\/]||'`\n # A \"..\" for each directory in $ac_dir_suffix.\n ac_top_builddir_sub=`$as_echo \"$ac_dir_suffix\" | sed 's|/[^\\\\/]*|/..|g;s|/||'`\n case $ac_top_builddir_sub in\n \"\") ac_top_builddir_sub=. ac_top_build_prefix= ;;\n *) ac_top_build_prefix=$ac_top_builddir_sub/ ;;\n esac ;;\nesac\nac_abs_top_builddir=$ac_pwd\nac_abs_builddir=$ac_pwd$ac_dir_suffix\n# for backward compatibility:\nac_top_builddir=$ac_top_build_prefix\n\ncase $srcdir in\n .) # We are building in place.\n ac_srcdir=.\n ac_top_srcdir=$ac_top_builddir_sub\n ac_abs_top_srcdir=$ac_pwd ;;\n [\\\\/]* | ?:[\\\\/]* ) # Absolute name.\n ac_srcdir=$srcdir$ac_dir_suffix;\n ac_top_srcdir=$srcdir\n ac_abs_top_srcdir=$srcdir ;;\n *) # Relative name.\n ac_srcdir=$ac_top_build_prefix$srcdir$ac_dir_suffix\n ac_top_srcdir=$ac_top_build_prefix$srcdir\n ac_abs_top_srcdir=$ac_pwd/$srcdir ;;\nesac\nac_abs_srcdir=$ac_abs_top_srcdir$ac_dir_suffix\n\n\n case $ac_mode in\n :F)\n #\n # CONFIG_FILE\n #\n\n case $INSTALL in\n [\\\\/$]* | ?:[\\\\/]* ) ac_INSTALL=$INSTALL ;;\n *) ac_INSTALL=$ac_top_build_prefix$INSTALL ;;\n esac\n_ACEOF\n\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n# If the template does not know about datarootdir, expand it.\n# FIXME: This hack should be removed a few years after 2.60.\nac_datarootdir_hack=; ac_datarootdir_seen=\nac_sed_dataroot='\n/datarootdir/ {\n p\n q\n}\n/@datadir@/p\n/@docdir@/p\n/@infodir@/p\n/@localedir@/p\n/@mandir@/p'\ncase `eval \"sed -n \\\"\\$ac_sed_dataroot\\\" $ac_file_inputs\"` in\n*datarootdir*) ac_datarootdir_seen=yes;;\n*@datadir@*|*@docdir@*|*@infodir@*|*@localedir@*|*@mandir@*)\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $ac_file_inputs seems to ignore the --datarootdir setting\" >&5\n$as_echo \"$as_me: WARNING: $ac_file_inputs seems to ignore the --datarootdir setting\" >&2;}\n_ACEOF\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\n ac_datarootdir_hack='\n s&@datadir@&$datadir&g\n s&@docdir@&$docdir&g\n s&@infodir@&$infodir&g\n s&@localedir@&$localedir&g\n s&@mandir@&$mandir&g\n s&\\\\\\${datarootdir}&$datarootdir&g' ;;\nesac\n_ACEOF\n\n# Neutralize VPATH when `$srcdir' = `.'.\n# Shell code in configure.ac might set extrasub.\n# FIXME: do we really want to maintain this feature?\ncat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1\nac_sed_extra=\"$ac_vpsub\n$extrasub\n_ACEOF\ncat >>$CONFIG_STATUS <<\\_ACEOF || ac_write_fail=1\n:t\n/@[a-zA-Z_][a-zA-Z_0-9]*@/!b\ns|@configure_input@|$ac_sed_conf_input|;t t\ns&@top_builddir@&$ac_top_builddir_sub&;t t\ns&@top_build_prefix@&$ac_top_build_prefix&;t t\ns&@srcdir@&$ac_srcdir&;t t\ns&@abs_srcdir@&$ac_abs_srcdir&;t t\ns&@top_srcdir@&$ac_top_srcdir&;t t\ns&@abs_top_srcdir@&$ac_abs_top_srcdir&;t t\ns&@builddir@&$ac_builddir&;t t\ns&@abs_builddir@&$ac_abs_builddir&;t t\ns&@abs_top_builddir@&$ac_abs_top_builddir&;t t\ns&@INSTALL@&$ac_INSTALL&;t t\n$ac_datarootdir_hack\n\"\neval sed \\\"\\$ac_sed_extra\\\" \"$ac_file_inputs\" | $AWK -f \"$ac_tmp/subs.awk\" \\\n >$ac_tmp/out || as_fn_error $? \"could not create $ac_file\" \"$LINENO\" 5\n\ntest -z \"$ac_datarootdir_hack$ac_datarootdir_seen\" &&\n { ac_out=`sed -n '/\\${datarootdir}/p' \"$ac_tmp/out\"`; test -n \"$ac_out\"; } &&\n { ac_out=`sed -n '/^[\t ]*datarootdir[\t ]*:*=/p' \\\n \"$ac_tmp/out\"`; test -z \"$ac_out\"; } &&\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: $ac_file contains a reference to the variable \\`datarootdir'\nwhich seems to be undefined. Please make sure it is defined\" >&5\n$as_echo \"$as_me: WARNING: $ac_file contains a reference to the variable \\`datarootdir'\nwhich seems to be undefined. Please make sure it is defined\" >&2;}\n\n rm -f \"$ac_tmp/stdin\"\n case $ac_file in\n -) cat \"$ac_tmp/out\" && rm -f \"$ac_tmp/out\";;\n *) rm -f \"$ac_file\" && mv \"$ac_tmp/out\" \"$ac_file\";;\n esac \\\n || as_fn_error $? \"could not create $ac_file\" \"$LINENO\" 5\n ;;\n :H)\n #\n # CONFIG_HEADER\n #\n if test x\"$ac_file\" != x-; then\n {\n $as_echo \"/* $configure_input */\" \\\n && eval '$AWK -f \"$ac_tmp/defines.awk\"' \"$ac_file_inputs\"\n } >\"$ac_tmp/config.h\" \\\n || as_fn_error $? \"could not create $ac_file\" \"$LINENO\" 5\n if diff \"$ac_file\" \"$ac_tmp/config.h\" >/dev/null 2>&1; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: $ac_file is unchanged\" >&5\n$as_echo \"$as_me: $ac_file is unchanged\" >&6;}\n else\n rm -f \"$ac_file\"\n mv \"$ac_tmp/config.h\" \"$ac_file\" \\\n\t|| as_fn_error $? \"could not create $ac_file\" \"$LINENO\" 5\n fi\n else\n $as_echo \"/* $configure_input */\" \\\n && eval '$AWK -f \"$ac_tmp/defines.awk\"' \"$ac_file_inputs\" \\\n || as_fn_error $? \"could not create -\" \"$LINENO\" 5\n fi\n ;;\n\n :C) { $as_echo \"$as_me:${as_lineno-$LINENO}: executing $ac_file commands\" >&5\n$as_echo \"$as_me: executing $ac_file commands\" >&6;}\n ;;\n esac\n\n\n case $ac_file$ac_mode in\n \"include/jemalloc/internal/size_classes.h\":C)\n mkdir -p \"include/jemalloc/internal\"\n \"${srcdir}/include/jemalloc/internal/size_classes.sh\" > \"${objroot}include/jemalloc/internal/size_classes.h\"\n ;;\n\n esac\ndone # for ac_tag\n\n\nas_fn_exit 0\n_ACEOF\nac_clean_files=$ac_clean_files_save\n\ntest $ac_write_fail = 0 ||\n as_fn_error $? \"write failure creating $CONFIG_STATUS\" \"$LINENO\" 5\n\n\n# configure is writing to config.log, and then calls config.status.\n# config.status does its own redirection, appending to config.log.\n# Unfortunately, on DOS this fails, as config.log is still kept open\n# by configure, so config.status won't be able to write to it; its\n# output is simply discarded. So we exec the FD to /dev/null,\n# effectively closing config.log, so it can be properly (re)opened and\n# appended to by config.status. When coming back to configure, we\n# need to make the FD available again.\nif test \"$no_create\" != yes; then\n ac_cs_success=:\n ac_config_status_args=\n test \"$silent\" = yes &&\n ac_config_status_args=\"$ac_config_status_args --quiet\"\n exec 5>/dev/null\n $SHELL $CONFIG_STATUS $ac_config_status_args || ac_cs_success=false\n exec 5>>config.log\n # Use ||, not &&, to avoid exiting from the if with $? = 1, which\n # would make configure fail if this is the last instruction.\n $ac_cs_success || as_fn_exit 1\nfi\nif test -n \"$ac_unrecognized_opts\" && test \"$enable_option_checking\" != no; then\n { $as_echo \"$as_me:${as_lineno-$LINENO}: WARNING: unrecognized options: $ac_unrecognized_opts\" >&5\n$as_echo \"$as_me: WARNING: unrecognized options: $ac_unrecognized_opts\" >&2;}\nfi\n\n\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: ===============================================================================\" >&5\n$as_echo \"===============================================================================\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: jemalloc version : ${jemalloc_version}\" >&5\n$as_echo \"jemalloc version : ${jemalloc_version}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: library revision : ${rev}\" >&5\n$as_echo \"library revision : ${rev}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: \" >&5\n$as_echo \"\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: CC : ${CC}\" >&5\n$as_echo \"CC : ${CC}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: CPPFLAGS : ${CPPFLAGS}\" >&5\n$as_echo \"CPPFLAGS : ${CPPFLAGS}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: CFLAGS : ${CFLAGS}\" >&5\n$as_echo \"CFLAGS : ${CFLAGS}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: LDFLAGS : ${LDFLAGS}\" >&5\n$as_echo \"LDFLAGS : ${LDFLAGS}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: LIBS : ${LIBS}\" >&5\n$as_echo \"LIBS : ${LIBS}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: RPATH_EXTRA : ${RPATH_EXTRA}\" >&5\n$as_echo \"RPATH_EXTRA : ${RPATH_EXTRA}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: \" >&5\n$as_echo \"\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: XSLTPROC : ${XSLTPROC}\" >&5\n$as_echo \"XSLTPROC : ${XSLTPROC}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: XSLROOT : ${XSLROOT}\" >&5\n$as_echo \"XSLROOT : ${XSLROOT}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: \" >&5\n$as_echo \"\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: PREFIX : ${PREFIX}\" >&5\n$as_echo \"PREFIX : ${PREFIX}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: BINDIR : ${BINDIR}\" >&5\n$as_echo \"BINDIR : ${BINDIR}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: INCLUDEDIR : ${INCLUDEDIR}\" >&5\n$as_echo \"INCLUDEDIR : ${INCLUDEDIR}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: LIBDIR : ${LIBDIR}\" >&5\n$as_echo \"LIBDIR : ${LIBDIR}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: DATADIR : ${DATADIR}\" >&5\n$as_echo \"DATADIR : ${DATADIR}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: MANDIR : ${MANDIR}\" >&5\n$as_echo \"MANDIR : ${MANDIR}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: \" >&5\n$as_echo \"\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: srcroot : ${srcroot}\" >&5\n$as_echo \"srcroot : ${srcroot}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: abs_srcroot : ${abs_srcroot}\" >&5\n$as_echo \"abs_srcroot : ${abs_srcroot}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: objroot : ${objroot}\" >&5\n$as_echo \"objroot : ${objroot}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: abs_objroot : ${abs_objroot}\" >&5\n$as_echo \"abs_objroot : ${abs_objroot}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: \" >&5\n$as_echo \"\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: JEMALLOC_PREFIX : ${JEMALLOC_PREFIX}\" >&5\n$as_echo \"JEMALLOC_PREFIX : ${JEMALLOC_PREFIX}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: JEMALLOC_PRIVATE_NAMESPACE\" >&5\n$as_echo \"JEMALLOC_PRIVATE_NAMESPACE\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: : ${JEMALLOC_PRIVATE_NAMESPACE}\" >&5\n$as_echo \" : ${JEMALLOC_PRIVATE_NAMESPACE}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: install_suffix : ${install_suffix}\" >&5\n$as_echo \"install_suffix : ${install_suffix}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: autogen : ${enable_autogen}\" >&5\n$as_echo \"autogen : ${enable_autogen}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: experimental : ${enable_experimental}\" >&5\n$as_echo \"experimental : ${enable_experimental}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: cc-silence : ${enable_cc_silence}\" >&5\n$as_echo \"cc-silence : ${enable_cc_silence}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: debug : ${enable_debug}\" >&5\n$as_echo \"debug : ${enable_debug}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: stats : ${enable_stats}\" >&5\n$as_echo \"stats : ${enable_stats}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: prof : ${enable_prof}\" >&5\n$as_echo \"prof : ${enable_prof}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: prof-libunwind : ${enable_prof_libunwind}\" >&5\n$as_echo \"prof-libunwind : ${enable_prof_libunwind}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: prof-libgcc : ${enable_prof_libgcc}\" >&5\n$as_echo \"prof-libgcc : ${enable_prof_libgcc}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: prof-gcc : ${enable_prof_gcc}\" >&5\n$as_echo \"prof-gcc : ${enable_prof_gcc}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: tcache : ${enable_tcache}\" >&5\n$as_echo \"tcache : ${enable_tcache}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: fill : ${enable_fill}\" >&5\n$as_echo \"fill : ${enable_fill}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: utrace : ${enable_utrace}\" >&5\n$as_echo \"utrace : ${enable_utrace}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: valgrind : ${enable_valgrind}\" >&5\n$as_echo \"valgrind : ${enable_valgrind}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: xmalloc : ${enable_xmalloc}\" >&5\n$as_echo \"xmalloc : ${enable_xmalloc}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: mremap : ${enable_mremap}\" >&5\n$as_echo \"mremap : ${enable_mremap}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: munmap : ${enable_munmap}\" >&5\n$as_echo \"munmap : ${enable_munmap}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: dss : ${enable_dss}\" >&5\n$as_echo \"dss : ${enable_dss}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: lazy_lock : ${enable_lazy_lock}\" >&5\n$as_echo \"lazy_lock : ${enable_lazy_lock}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: tls : ${enable_tls}\" >&5\n$as_echo \"tls : ${enable_tls}\" >&6; }\n{ $as_echo \"$as_me:${as_lineno-$LINENO}: result: ===============================================================================\" >&5\n$as_echo \"===============================================================================\" >&6; }\n" }, { "path": "deps/jemalloc/configure.ac", "content": "dnl Process this file with autoconf to produce a configure script.\nAC_INIT([Makefile.in])\n\ndnl ============================================================================\ndnl Custom macro definitions.\n\ndnl JE_CFLAGS_APPEND(cflag)\nAC_DEFUN([JE_CFLAGS_APPEND],\n[\nAC_MSG_CHECKING([whether compiler supports $1])\nTCFLAGS=\"${CFLAGS}\"\nif test \"x${CFLAGS}\" = \"x\" ; then\n CFLAGS=\"$1\"\nelse\n CFLAGS=\"${CFLAGS} $1\"\nfi\nAC_COMPILE_IFELSE([AC_LANG_PROGRAM(\n[[\n]], [[\n return 0;\n]])],\n AC_MSG_RESULT([yes]),\n AC_MSG_RESULT([no])\n [CFLAGS=\"${TCFLAGS}\"]\n)\n])\n\ndnl JE_COMPILABLE(label, hcode, mcode, rvar)\ndnl \ndnl Use AC_LINK_IFELSE() rather than AC_COMPILE_IFELSE() so that linker errors\ndnl cause failure.\nAC_DEFUN([JE_COMPILABLE],\n[\nAC_CACHE_CHECK([whether $1 is compilable],\n [$4],\n [AC_LINK_IFELSE([AC_LANG_PROGRAM([$2],\n [$3])],\n [$4=yes],\n [$4=no])])\n])\n\ndnl ============================================================================\n\ndnl Library revision.\nrev=1\nAC_SUBST([rev])\n\nsrcroot=$srcdir\nif test \"x${srcroot}\" = \"x.\" ; then\n srcroot=\"\"\nelse\n srcroot=\"${srcroot}/\"\nfi\nAC_SUBST([srcroot])\nabs_srcroot=\"`cd \\\"${srcdir}\\\"; pwd`/\"\nAC_SUBST([abs_srcroot])\n\nobjroot=\"\"\nAC_SUBST([objroot])\nabs_objroot=\"`pwd`/\"\nAC_SUBST([abs_objroot])\n\ndnl Munge install path variables.\nif test \"x$prefix\" = \"xNONE\" ; then\n prefix=\"/usr/local\"\nfi\nif test \"x$exec_prefix\" = \"xNONE\" ; then\n exec_prefix=$prefix\nfi\nPREFIX=$prefix\nAC_SUBST([PREFIX])\nBINDIR=`eval echo $bindir`\nBINDIR=`eval echo $BINDIR`\nAC_SUBST([BINDIR])\nINCLUDEDIR=`eval echo $includedir`\nINCLUDEDIR=`eval echo $INCLUDEDIR`\nAC_SUBST([INCLUDEDIR])\nLIBDIR=`eval echo $libdir`\nLIBDIR=`eval echo $LIBDIR`\nAC_SUBST([LIBDIR])\nDATADIR=`eval echo $datadir`\nDATADIR=`eval echo $DATADIR`\nAC_SUBST([DATADIR])\nMANDIR=`eval echo $mandir`\nMANDIR=`eval echo $MANDIR`\nAC_SUBST([MANDIR])\n\ndnl Support for building documentation.\nAC_PATH_PROG([XSLTPROC], [xsltproc], , [$PATH])\nif test -d \"/usr/share/xml/docbook/stylesheet/docbook-xsl\" ; then\n DEFAULT_XSLROOT=\"/usr/share/xml/docbook/stylesheet/docbook-xsl\"\nelif test -d \"/usr/share/sgml/docbook/xsl-stylesheets\" ; then\n DEFAULT_XSLROOT=\"/usr/share/sgml/docbook/xsl-stylesheets\"\nelse\n dnl Documentation building will fail if this default gets used.\n DEFAULT_XSLROOT=\"\"\nfi\nAC_ARG_WITH([xslroot],\n [AS_HELP_STRING([--with-xslroot=], [XSL stylesheet root path])], [\nif test \"x$with_xslroot\" = \"xno\" ; then\n XSLROOT=\"${DEFAULT_XSLROOT}\"\nelse\n XSLROOT=\"${with_xslroot}\"\nfi\n],\n XSLROOT=\"${DEFAULT_XSLROOT}\"\n)\nAC_SUBST([XSLROOT])\n\ndnl If CFLAGS isn't defined, set CFLAGS to something reasonable. Otherwise,\ndnl just prevent autoconf from molesting CFLAGS.\nCFLAGS=$CFLAGS\nAC_PROG_CC\nif test \"x$GCC\" != \"xyes\" ; then\n AC_CACHE_CHECK([whether compiler is MSVC],\n [je_cv_msvc],\n [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],\n [\n#ifndef _MSC_VER\n int fail[-1];\n#endif\n])],\n [je_cv_msvc=yes],\n [je_cv_msvc=no])])\nfi\n\nif test \"x$CFLAGS\" = \"x\" ; then\n no_CFLAGS=\"yes\"\n if test \"x$GCC\" = \"xyes\" ; then\n JE_CFLAGS_APPEND([-std=gnu99])\n JE_CFLAGS_APPEND([-Wall])\n JE_CFLAGS_APPEND([-pipe])\n JE_CFLAGS_APPEND([-g3])\n elif test \"x$je_cv_msvc\" = \"xyes\" ; then\n CC=\"$CC -nologo\"\n JE_CFLAGS_APPEND([-Zi])\n JE_CFLAGS_APPEND([-MT])\n JE_CFLAGS_APPEND([-W3])\n CPPFLAGS=\"$CPPFLAGS -I${srcroot}/include/msvc_compat\"\n fi\nfi\ndnl Append EXTRA_CFLAGS to CFLAGS, if defined.\nif test \"x$EXTRA_CFLAGS\" != \"x\" ; then\n JE_CFLAGS_APPEND([$EXTRA_CFLAGS])\nfi\nAC_PROG_CPP\n\nAC_CHECK_SIZEOF([void *])\nif test \"x${ac_cv_sizeof_void_p}\" = \"x8\" ; then\n LG_SIZEOF_PTR=3\nelif test \"x${ac_cv_sizeof_void_p}\" = \"x4\" ; then\n LG_SIZEOF_PTR=2\nelse\n AC_MSG_ERROR([Unsupported pointer size: ${ac_cv_sizeof_void_p}])\nfi\nAC_DEFINE_UNQUOTED([LG_SIZEOF_PTR], [$LG_SIZEOF_PTR])\n\nAC_CHECK_SIZEOF([int])\nif test \"x${ac_cv_sizeof_int}\" = \"x8\" ; then\n LG_SIZEOF_INT=3\nelif test \"x${ac_cv_sizeof_int}\" = \"x4\" ; then\n LG_SIZEOF_INT=2\nelse\n AC_MSG_ERROR([Unsupported int size: ${ac_cv_sizeof_int}])\nfi\nAC_DEFINE_UNQUOTED([LG_SIZEOF_INT], [$LG_SIZEOF_INT])\n\nAC_CHECK_SIZEOF([long])\nif test \"x${ac_cv_sizeof_long}\" = \"x8\" ; then\n LG_SIZEOF_LONG=3\nelif test \"x${ac_cv_sizeof_long}\" = \"x4\" ; then\n LG_SIZEOF_LONG=2\nelse\n AC_MSG_ERROR([Unsupported long size: ${ac_cv_sizeof_long}])\nfi\nAC_DEFINE_UNQUOTED([LG_SIZEOF_LONG], [$LG_SIZEOF_LONG])\n\nAC_CHECK_SIZEOF([intmax_t])\nif test \"x${ac_cv_sizeof_intmax_t}\" = \"x16\" ; then\n LG_SIZEOF_INTMAX_T=4\nelif test \"x${ac_cv_sizeof_intmax_t}\" = \"x8\" ; then\n LG_SIZEOF_INTMAX_T=3\nelif test \"x${ac_cv_sizeof_intmax_t}\" = \"x4\" ; then\n LG_SIZEOF_INTMAX_T=2\nelse\n AC_MSG_ERROR([Unsupported intmax_t size: ${ac_cv_sizeof_intmax_t}])\nfi\nAC_DEFINE_UNQUOTED([LG_SIZEOF_INTMAX_T], [$LG_SIZEOF_INTMAX_T])\n\nAC_CANONICAL_HOST\ndnl CPU-specific settings.\nCPU_SPINWAIT=\"\"\ncase \"${host_cpu}\" in\n i[[345]]86)\n\t;;\n i686)\n\tJE_COMPILABLE([__asm__], [], [[__asm__ volatile(\"pause\"); return 0;]],\n\t [je_cv_asm])\n\tif test \"x${je_cv_asm}\" = \"xyes\" ; then\n\t CPU_SPINWAIT='__asm__ volatile(\"pause\")'\n\tfi\n\t;;\n x86_64)\n\tJE_COMPILABLE([__asm__ syntax], [],\n\t [[__asm__ volatile(\"pause\"); return 0;]], [je_cv_asm])\n\tif test \"x${je_cv_asm}\" = \"xyes\" ; then\n\t CPU_SPINWAIT='__asm__ volatile(\"pause\")'\n\tfi\n\t;;\n *)\n\t;;\nesac\nAC_DEFINE_UNQUOTED([CPU_SPINWAIT], [$CPU_SPINWAIT])\n\nLD_PRELOAD_VAR=\"LD_PRELOAD\"\nso=\"so\"\nimportlib=\"${so}\"\no=\"$ac_objext\"\na=\"a\"\nexe=\"$ac_exeext\"\nlibprefix=\"lib\"\nDSO_LDFLAGS='-shared -Wl,-soname,$(@F)'\nRPATH='-Wl,-rpath,$(1)'\nSOREV=\"${so}.${rev}\"\nPIC_CFLAGS='-fPIC -DPIC'\nCTARGET='-o $@'\nLDTARGET='-o $@'\nEXTRA_LDFLAGS=\nMKLIB='ar crus $@'\nCC_MM=1\n\ndnl Platform-specific settings. abi and RPATH can probably be determined\ndnl programmatically, but doing so is error-prone, which makes it generally\ndnl not worth the trouble.\ndnl \ndnl Define cpp macros in CPPFLAGS, rather than doing AC_DEFINE(macro), since the\ndnl definitions need to be seen before any headers are included, which is a pain\ndnl to make happen otherwise.\ndefault_munmap=\"1\"\nJEMALLOC_USABLE_SIZE_CONST=\"const\"\ncase \"${host}\" in\n *-*-darwin*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"macho\"\n\tAC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ])\n\tRPATH=\"\"\n\tLD_PRELOAD_VAR=\"DYLD_INSERT_LIBRARIES\"\n\tso=\"dylib\"\n\timportlib=\"${so}\"\n\tforce_tls=\"0\"\n\tDSO_LDFLAGS='-shared -Wl,-dylib_install_name,$(@F)'\n\tSOREV=\"${rev}.${so}\"\n\t;;\n *-*-freebsd*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"elf\"\n\tAC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ])\n\tforce_lazy_lock=\"1\"\n\t;;\n *-*-linux*)\n\tCFLAGS=\"$CFLAGS\"\n\tCPPFLAGS=\"$CPPFLAGS -D_GNU_SOURCE\"\n\tabi=\"elf\"\n\tAC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED], [ ])\n\tAC_DEFINE([JEMALLOC_THREADED_INIT], [ ])\n\tJEMALLOC_USABLE_SIZE_CONST=\"\"\n\tdefault_munmap=\"0\"\n\t;;\n *-*-netbsd*)\n\tAC_MSG_CHECKING([ABI])\n AC_COMPILE_IFELSE([AC_LANG_PROGRAM(\n[[#ifdef __ELF__\n/* ELF */\n#else\n#error aout\n#endif\n]])],\n [CFLAGS=\"$CFLAGS\"; abi=\"elf\"],\n [abi=\"aout\"])\n\tAC_MSG_RESULT([$abi])\n\tAC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ])\n\t;;\n *-*-solaris2*)\n\tCFLAGS=\"$CFLAGS\"\n\tabi=\"elf\"\n\tRPATH='-Wl,-R,$(1)'\n\tdnl Solaris needs this for sigwait().\n\tCPPFLAGS=\"$CPPFLAGS -D_POSIX_PTHREAD_SEMANTICS\"\n\tLIBS=\"$LIBS -lposix4 -lsocket -lnsl\"\n\t;;\n *-ibm-aix*)\n\tif \"$LG_SIZEOF_PTR\" = \"8\"; then\n\t dnl 64bit AIX\n\t LD_PRELOAD_VAR=\"LDR_PRELOAD64\"\n\telse\n\t dnl 32bit AIX\n\t LD_PRELOAD_VAR=\"LDR_PRELOAD\"\n\tfi\n\tabi=\"xcoff\"\n\t;;\n *-*-mingw*)\n\tabi=\"pecoff\"\n\tforce_tls=\"0\"\n\tRPATH=\"\"\n\tso=\"dll\"\n\tif test \"x$je_cv_msvc\" = \"xyes\" ; then\n\t importlib=\"lib\"\n\t DSO_LDFLAGS=\"-LD\"\n\t EXTRA_LDFLAGS=\"-link -DEBUG\"\n\t CTARGET='-Fo$@'\n\t LDTARGET='-Fe$@'\n\t MKLIB='lib -nologo -out:$@'\n\t CC_MM=\n else\n\t importlib=\"${so}\"\n\t DSO_LDFLAGS=\"-shared\"\n\tfi\n\ta=\"lib\"\n\tlibprefix=\"\"\n\tSOREV=\"${so}\"\n\tPIC_CFLAGS=\"\"\n\t;;\n *)\n\tAC_MSG_RESULT([Unsupported operating system: ${host}])\n\tabi=\"elf\"\n\t;;\nesac\nAC_DEFINE_UNQUOTED([JEMALLOC_USABLE_SIZE_CONST], [$JEMALLOC_USABLE_SIZE_CONST])\nAC_SUBST([abi])\nAC_SUBST([RPATH])\nAC_SUBST([LD_PRELOAD_VAR])\nAC_SUBST([so])\nAC_SUBST([importlib])\nAC_SUBST([o])\nAC_SUBST([a])\nAC_SUBST([exe])\nAC_SUBST([libprefix])\nAC_SUBST([DSO_LDFLAGS])\nAC_SUBST([EXTRA_LDFLAGS])\nAC_SUBST([SOREV])\nAC_SUBST([PIC_CFLAGS])\nAC_SUBST([CTARGET])\nAC_SUBST([LDTARGET])\nAC_SUBST([MKLIB])\nAC_SUBST([CC_MM])\n\nif test \"x$abi\" != \"xpecoff\"; then\n dnl Heap profiling uses the log(3) function.\n LIBS=\"$LIBS -lm\"\nfi\n\nJE_COMPILABLE([__attribute__ syntax],\n [static __attribute__((unused)) void foo(void){}],\n [],\n [je_cv_attribute])\nif test \"x${je_cv_attribute}\" = \"xyes\" ; then\n AC_DEFINE([JEMALLOC_HAVE_ATTR], [ ])\n if test \"x${GCC}\" = \"xyes\" -a \"x${abi}\" = \"xelf\"; then\n JE_CFLAGS_APPEND([-fvisibility=hidden])\n fi\nfi\ndnl Check for tls_model attribute support (clang 3.0 still lacks support).\nSAVED_CFLAGS=\"${CFLAGS}\"\nJE_CFLAGS_APPEND([-Werror])\nJE_COMPILABLE([tls_model attribute], [],\n [static __thread int\n __attribute__((tls_model(\"initial-exec\"))) foo;\n foo = 0;],\n [je_cv_tls_model])\nCFLAGS=\"${SAVED_CFLAGS}\"\nif test \"x${je_cv_tls_model}\" = \"xyes\" ; then\n AC_DEFINE([JEMALLOC_TLS_MODEL],\n [__attribute__((tls_model(\"initial-exec\")))])\nelse\n AC_DEFINE([JEMALLOC_TLS_MODEL], [ ])\nfi\n\ndnl Support optional additions to rpath.\nAC_ARG_WITH([rpath],\n [AS_HELP_STRING([--with-rpath=], [Colon-separated rpath (ELF systems only)])],\nif test \"x$with_rpath\" = \"xno\" ; then\n RPATH_EXTRA=\nelse\n RPATH_EXTRA=\"`echo $with_rpath | tr \\\":\\\" \\\" \\\"`\"\nfi,\n RPATH_EXTRA=\n)\nAC_SUBST([RPATH_EXTRA])\n\ndnl Disable rules that do automatic regeneration of configure output by default.\nAC_ARG_ENABLE([autogen],\n [AS_HELP_STRING([--enable-autogen], [Automatically regenerate configure output])],\nif test \"x$enable_autogen\" = \"xno\" ; then\n enable_autogen=\"0\"\nelse\n enable_autogen=\"1\"\nfi\n,\nenable_autogen=\"0\"\n)\nAC_SUBST([enable_autogen])\n\nAC_PROG_INSTALL\nAC_PROG_RANLIB\nAC_PATH_PROG([AR], [ar], , [$PATH])\nAC_PATH_PROG([LD], [ld], , [$PATH])\nAC_PATH_PROG([AUTOCONF], [autoconf], , [$PATH])\n\npublic_syms=\"malloc_conf malloc_message malloc calloc posix_memalign aligned_alloc realloc free malloc_usable_size malloc_stats_print mallctl mallctlnametomib mallctlbymib\"\n\ndnl Check for allocator-related functions that should be wrapped.\nAC_CHECK_FUNC([memalign],\n\t [AC_DEFINE([JEMALLOC_OVERRIDE_MEMALIGN], [ ])\n\t public_syms=\"${public_syms} memalign\"])\nAC_CHECK_FUNC([valloc],\n\t [AC_DEFINE([JEMALLOC_OVERRIDE_VALLOC], [ ])\n\t public_syms=\"${public_syms} valloc\"])\n\ndnl Support the experimental API by default.\nAC_ARG_ENABLE([experimental],\n [AS_HELP_STRING([--disable-experimental],\n [Disable support for the experimental API])],\n[if test \"x$enable_experimental\" = \"xno\" ; then\n enable_experimental=\"0\"\nelse\n enable_experimental=\"1\"\nfi\n],\n[enable_experimental=\"1\"]\n)\nif test \"x$enable_experimental\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_EXPERIMENTAL], [ ])\n public_syms=\"${public_syms} allocm dallocm nallocm rallocm sallocm\"\nfi\nAC_SUBST([enable_experimental])\n\ndnl Perform no name mangling by default.\nAC_ARG_WITH([mangling],\n [AS_HELP_STRING([--with-mangling=], [Mangle symbols in ])],\n [mangling_map=\"$with_mangling\"], [mangling_map=\"\"])\nfor nm in `echo ${mangling_map} |tr ',' ' '` ; do\n k=\"`echo ${nm} |tr ':' ' ' |awk '{print $1}'`\"\n n=\"je_${k}\"\n m=`echo ${nm} |tr ':' ' ' |awk '{print $2}'`\n AC_DEFINE_UNQUOTED([${n}], [${m}])\n dnl Remove key from public_syms so that it isn't redefined later.\n public_syms=`for sym in ${public_syms}; do echo \"${sym}\"; done |grep -v \"^${k}\\$\" |tr '\\n' ' '`\ndone\n\ndnl Do not prefix public APIs by default.\nAC_ARG_WITH([jemalloc_prefix],\n [AS_HELP_STRING([--with-jemalloc-prefix=], [Prefix to prepend to all public APIs])],\n [JEMALLOC_PREFIX=\"$with_jemalloc_prefix\"],\n [if test \"x$abi\" != \"xmacho\" -a \"x$abi\" != \"xpecoff\"; then\n JEMALLOC_PREFIX=\"\"\nelse\n JEMALLOC_PREFIX=\"je_\"\nfi]\n)\nif test \"x$JEMALLOC_PREFIX\" != \"x\" ; then\n JEMALLOC_CPREFIX=`echo ${JEMALLOC_PREFIX} | tr \"a-z\" \"A-Z\"`\n AC_DEFINE_UNQUOTED([JEMALLOC_PREFIX], [\"$JEMALLOC_PREFIX\"])\n AC_DEFINE_UNQUOTED([JEMALLOC_CPREFIX], [\"$JEMALLOC_CPREFIX\"])\nfi\ndnl Generate macros to rename public symbols. All public symbols are prefixed\ndnl with je_ in the source code, so these macro definitions are needed even if\ndnl --with-jemalloc-prefix wasn't specified.\nfor stem in ${public_syms}; do\n n=\"je_${stem}\"\n m=\"${JEMALLOC_PREFIX}${stem}\"\n AC_DEFINE_UNQUOTED([${n}], [${m}])\ndone\n\ndnl Do not mangle library-private APIs by default.\nAC_ARG_WITH([private_namespace],\n [AS_HELP_STRING([--with-private-namespace=], [Prefix to prepend to all library-private APIs])],\n [JEMALLOC_PRIVATE_NAMESPACE=\"$with_private_namespace\"],\n [JEMALLOC_PRIVATE_NAMESPACE=\"\"]\n)\nAC_DEFINE_UNQUOTED([JEMALLOC_PRIVATE_NAMESPACE], [\"$JEMALLOC_PRIVATE_NAMESPACE\"])\nif test \"x$JEMALLOC_PRIVATE_NAMESPACE\" != \"x\" ; then\n AC_DEFINE_UNQUOTED([JEMALLOC_N(string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix)], [${JEMALLOC_PRIVATE_NAMESPACE}##string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix])\nelse\n AC_DEFINE_UNQUOTED([JEMALLOC_N(string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix)], [string_that_no_one_should_want_to_use_as_a_jemalloc_private_namespace_prefix])\nfi\n\ndnl Do not add suffix to installed files by default.\nAC_ARG_WITH([install_suffix],\n [AS_HELP_STRING([--with-install-suffix=], [Suffix to append to all installed files])],\n [INSTALL_SUFFIX=\"$with_install_suffix\"],\n [INSTALL_SUFFIX=]\n)\ninstall_suffix=\"$INSTALL_SUFFIX\"\nAC_SUBST([install_suffix])\n\ncfgoutputs_in=\"${srcroot}Makefile.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/html.xsl.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/manpages.xsl.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}doc/jemalloc.xml.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}include/jemalloc/jemalloc.h.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}include/jemalloc/internal/jemalloc_internal.h.in\"\ncfgoutputs_in=\"${cfgoutputs_in} ${srcroot}test/jemalloc_test.h.in\"\n\ncfgoutputs_out=\"Makefile\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/html.xsl\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/manpages.xsl\"\ncfgoutputs_out=\"${cfgoutputs_out} doc/jemalloc${install_suffix}.xml\"\ncfgoutputs_out=\"${cfgoutputs_out} include/jemalloc/jemalloc${install_suffix}.h\"\ncfgoutputs_out=\"${cfgoutputs_out} include/jemalloc/internal/jemalloc_internal.h\"\ncfgoutputs_out=\"${cfgoutputs_out} test/jemalloc_test.h\"\n\ncfgoutputs_tup=\"Makefile\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/html.xsl:doc/html.xsl.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/manpages.xsl:doc/manpages.xsl.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} doc/jemalloc${install_suffix}.xml:doc/jemalloc.xml.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} include/jemalloc/jemalloc${install_suffix}.h:include/jemalloc/jemalloc.h.in\"\ncfgoutputs_tup=\"${cfgoutputs_tup} include/jemalloc/internal/jemalloc_internal.h\"\ncfgoutputs_tup=\"${cfgoutputs_tup} test/jemalloc_test.h:test/jemalloc_test.h.in\"\n\ncfghdrs_in=\"${srcroot}include/jemalloc/jemalloc_defs.h.in\"\ncfghdrs_in=\"${cfghdrs_in} ${srcroot}include/jemalloc/internal/size_classes.sh\"\n\ncfghdrs_out=\"include/jemalloc/jemalloc_defs${install_suffix}.h\"\ncfghdrs_out=\"${cfghdrs_out} include/jemalloc/internal/size_classes.h\"\n\ncfghdrs_tup=\"include/jemalloc/jemalloc_defs${install_suffix}.h:include/jemalloc/jemalloc_defs.h.in\"\n\ndnl Do not silence irrelevant compiler warnings by default, since enabling this\ndnl option incurs a performance penalty.\nAC_ARG_ENABLE([cc-silence],\n [AS_HELP_STRING([--enable-cc-silence],\n [Silence irrelevant compiler warnings])],\n[if test \"x$enable_cc_silence\" = \"xno\" ; then\n enable_cc_silence=\"0\"\nelse\n enable_cc_silence=\"1\"\nfi\n],\n[enable_cc_silence=\"0\"]\n)\nif test \"x$enable_cc_silence\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_CC_SILENCE], [ ])\nfi\n\ndnl Do not compile with debugging by default.\nAC_ARG_ENABLE([debug],\n [AS_HELP_STRING([--enable-debug], [Build debugging code])],\n[if test \"x$enable_debug\" = \"xno\" ; then\n enable_debug=\"0\"\nelse\n enable_debug=\"1\"\nfi\n],\n[enable_debug=\"0\"]\n)\nif test \"x$enable_debug\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_DEBUG], [ ])\n AC_DEFINE([JEMALLOC_IVSALLOC], [ ])\nfi\nAC_SUBST([enable_debug])\n\ndnl Only optimize if not debugging.\nif test \"x$enable_debug\" = \"x0\" -a \"x$no_CFLAGS\" = \"xyes\" ; then\n dnl Make sure that an optimization flag was not specified in EXTRA_CFLAGS.\n optimize=\"no\"\n echo \"$EXTRA_CFLAGS\" | grep \"\\-O\" >/dev/null || optimize=\"yes\"\n if test \"x${optimize}\" = \"xyes\" ; then\n if test \"x$GCC\" = \"xyes\" ; then\n JE_CFLAGS_APPEND([-O3])\n JE_CFLAGS_APPEND([-funroll-loops])\n elif test \"x$je_cv_msvc\" = \"xyes\" ; then\n JE_CFLAGS_APPEND([-O2])\n else\n JE_CFLAGS_APPEND([-O])\n fi\n fi\nfi\n\ndnl Enable statistics calculation by default.\nAC_ARG_ENABLE([stats],\n [AS_HELP_STRING([--disable-stats],\n [Disable statistics calculation/reporting])],\n[if test \"x$enable_stats\" = \"xno\" ; then\n enable_stats=\"0\"\nelse\n enable_stats=\"1\"\nfi\n],\n[enable_stats=\"1\"]\n)\nif test \"x$enable_stats\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_STATS], [ ])\nfi\nAC_SUBST([enable_stats])\n\ndnl Do not enable profiling by default.\nAC_ARG_ENABLE([prof],\n [AS_HELP_STRING([--enable-prof], [Enable allocation profiling])],\n[if test \"x$enable_prof\" = \"xno\" ; then\n enable_prof=\"0\"\nelse\n enable_prof=\"1\"\nfi\n],\n[enable_prof=\"0\"]\n)\nif test \"x$enable_prof\" = \"x1\" ; then\n backtrace_method=\"\"\nelse\n backtrace_method=\"N/A\"\nfi\n\nAC_ARG_ENABLE([prof-libunwind],\n [AS_HELP_STRING([--enable-prof-libunwind], [Use libunwind for backtracing])],\n[if test \"x$enable_prof_libunwind\" = \"xno\" ; then\n enable_prof_libunwind=\"0\"\nelse\n enable_prof_libunwind=\"1\"\nfi\n],\n[enable_prof_libunwind=\"0\"]\n)\nAC_ARG_WITH([static_libunwind],\n [AS_HELP_STRING([--with-static-libunwind=],\n [Path to static libunwind library; use rather than dynamically linking])],\nif test \"x$with_static_libunwind\" = \"xno\" ; then\n LUNWIND=\"-lunwind\"\nelse\n if test ! -f \"$with_static_libunwind\" ; then\n AC_MSG_ERROR([Static libunwind not found: $with_static_libunwind])\n fi\n LUNWIND=\"$with_static_libunwind\"\nfi,\n LUNWIND=\"-lunwind\"\n)\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_libunwind\" = \"x1\" ; then\n AC_CHECK_HEADERS([libunwind.h], , [enable_prof_libunwind=\"0\"])\n if test \"x$LUNWIND\" = \"x-lunwind\" ; then\n AC_CHECK_LIB([unwind], [backtrace], [LIBS=\"$LIBS $LUNWIND\"],\n [enable_prof_libunwind=\"0\"])\n else\n LIBS=\"$LIBS $LUNWIND\"\n fi\n if test \"x${enable_prof_libunwind}\" = \"x1\" ; then\n backtrace_method=\"libunwind\"\n AC_DEFINE([JEMALLOC_PROF_LIBUNWIND], [ ])\n fi\nfi\n\nAC_ARG_ENABLE([prof-libgcc],\n [AS_HELP_STRING([--disable-prof-libgcc],\n [Do not use libgcc for backtracing])],\n[if test \"x$enable_prof_libgcc\" = \"xno\" ; then\n enable_prof_libgcc=\"0\"\nelse\n enable_prof_libgcc=\"1\"\nfi\n],\n[enable_prof_libgcc=\"1\"]\n)\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_libgcc\" = \"x1\" \\\n -a \"x$GCC\" = \"xyes\" ; then\n AC_CHECK_HEADERS([unwind.h], , [enable_prof_libgcc=\"0\"])\n AC_CHECK_LIB([gcc], [_Unwind_Backtrace], [LIBS=\"$LIBS -lgcc\"], [enable_prof_libgcc=\"0\"])\n dnl The following is conservative, in that it only has entries for CPUs on\n dnl which jemalloc has been tested.\n AC_MSG_CHECKING([libgcc-based backtracing reliability on ${host_cpu}])\n case \"${host_cpu}\" in\n i[[3456]]86)\n AC_MSG_RESULT([unreliable])\n enable_prof_libgcc=\"0\";\n ;;\n x86_64)\n AC_MSG_RESULT([reliable])\n ;;\n *)\n AC_MSG_RESULT([unreliable])\n enable_prof_libgcc=\"0\";\n ;;\n esac\n if test \"x${enable_prof_libgcc}\" = \"x1\" ; then\n backtrace_method=\"libgcc\"\n AC_DEFINE([JEMALLOC_PROF_LIBGCC], [ ])\n fi\nelse\n enable_prof_libgcc=\"0\"\nfi\n\nAC_ARG_ENABLE([prof-gcc],\n [AS_HELP_STRING([--disable-prof-gcc],\n [Do not use gcc intrinsics for backtracing])],\n[if test \"x$enable_prof_gcc\" = \"xno\" ; then\n enable_prof_gcc=\"0\"\nelse\n enable_prof_gcc=\"1\"\nfi\n],\n[enable_prof_gcc=\"1\"]\n)\nif test \"x$backtrace_method\" = \"x\" -a \"x$enable_prof_gcc\" = \"x1\" \\\n -a \"x$GCC\" = \"xyes\" ; then\n backtrace_method=\"gcc intrinsics\"\n AC_DEFINE([JEMALLOC_PROF_GCC], [ ])\nelse\n enable_prof_gcc=\"0\"\nfi\n\nif test \"x$backtrace_method\" = \"x\" ; then\n backtrace_method=\"none (disabling profiling)\"\n enable_prof=\"0\"\nfi\nAC_MSG_CHECKING([configured backtracing method])\nAC_MSG_RESULT([$backtrace_method])\nif test \"x$enable_prof\" = \"x1\" ; then\n if test \"x${force_tls}\" = \"x0\" ; then\n AC_MSG_ERROR([Heap profiling requires TLS]);\n fi\n force_tls=\"1\"\n AC_DEFINE([JEMALLOC_PROF], [ ])\nfi\nAC_SUBST([enable_prof])\n\ndnl Enable thread-specific caching by default.\nAC_ARG_ENABLE([tcache],\n [AS_HELP_STRING([--disable-tcache], [Disable per thread caches])],\n[if test \"x$enable_tcache\" = \"xno\" ; then\n enable_tcache=\"0\"\nelse\n enable_tcache=\"1\"\nfi\n],\n[enable_tcache=\"1\"]\n)\nif test \"x$enable_tcache\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_TCACHE], [ ])\nfi\nAC_SUBST([enable_tcache])\n\ndnl Disable mremap() for huge realloc() by default.\nAC_ARG_ENABLE([mremap],\n [AS_HELP_STRING([--enable-mremap], [Enable mremap(2) for huge realloc()])],\n[if test \"x$enable_mremap\" = \"xno\" ; then\n enable_mremap=\"0\"\nelse\n enable_mremap=\"1\"\nfi\n],\n[enable_mremap=\"0\"]\n)\nif test \"x$enable_mremap\" = \"x1\" ; then\n JE_COMPILABLE([mremap(...MREMAP_FIXED...)], [\n#define _GNU_SOURCE\n#include \n], [\nvoid *p = mremap((void *)0, 0, 0, MREMAP_MAYMOVE|MREMAP_FIXED, (void *)0);\n], [je_cv_mremap_fixed])\n if test \"x${je_cv_mremap_fixed}\" = \"xno\" ; then\n enable_mremap=\"0\"\n fi\nfi\nif test \"x$enable_mremap\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_MREMAP], [ ])\nfi\nAC_SUBST([enable_mremap])\n\ndnl Enable VM deallocation via munmap() by default.\nAC_ARG_ENABLE([munmap],\n [AS_HELP_STRING([--disable-munmap], [Disable VM deallocation via munmap(2)])],\n[if test \"x$enable_munmap\" = \"xno\" ; then\n enable_munmap=\"0\"\nelse\n enable_munmap=\"1\"\nfi\n],\n[enable_munmap=\"${default_munmap}\"]\n)\nif test \"x$enable_munmap\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_MUNMAP], [ ])\nfi\nAC_SUBST([enable_munmap])\n\ndnl Do not enable allocation from DSS by default.\nAC_ARG_ENABLE([dss],\n [AS_HELP_STRING([--enable-dss], [Enable allocation from DSS])],\n[if test \"x$enable_dss\" = \"xno\" ; then\n enable_dss=\"0\"\nelse\n enable_dss=\"1\"\nfi\n],\n[enable_dss=\"0\"]\n)\ndnl Check whether the BSD/SUSv1 sbrk() exists. If not, disable DSS support.\nAC_CHECK_FUNC([sbrk], [have_sbrk=\"1\"], [have_sbrk=\"0\"])\nif test \"x$have_sbrk\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_HAVE_SBRK], [ ])\nelse\n enable_dss=\"0\"\nfi\n\nif test \"x$enable_dss\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_DSS], [ ])\nfi\nAC_SUBST([enable_dss])\n\ndnl Support the junk/zero filling option by default.\nAC_ARG_ENABLE([fill],\n [AS_HELP_STRING([--disable-fill],\n [Disable support for junk/zero filling, quarantine, and redzones])],\n[if test \"x$enable_fill\" = \"xno\" ; then\n enable_fill=\"0\"\nelse\n enable_fill=\"1\"\nfi\n],\n[enable_fill=\"1\"]\n)\nif test \"x$enable_fill\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_FILL], [ ])\nfi\nAC_SUBST([enable_fill])\n\ndnl Disable utrace(2)-based tracing by default.\nAC_ARG_ENABLE([utrace],\n [AS_HELP_STRING([--enable-utrace], [Enable utrace(2)-based tracing])],\n[if test \"x$enable_utrace\" = \"xno\" ; then\n enable_utrace=\"0\"\nelse\n enable_utrace=\"1\"\nfi\n],\n[enable_utrace=\"0\"]\n)\nJE_COMPILABLE([utrace(2)], [\n#include \n#include \n#include \n#include \n#include \n], [\n\tutrace((void *)0, 0);\n], [je_cv_utrace])\nif test \"x${je_cv_utrace}\" = \"xno\" ; then\n enable_utrace=\"0\"\nfi\nif test \"x$enable_utrace\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_UTRACE], [ ])\nfi\nAC_SUBST([enable_utrace])\n\ndnl Support Valgrind by default.\nAC_ARG_ENABLE([valgrind],\n [AS_HELP_STRING([--disable-valgrind], [Disable support for Valgrind])],\n[if test \"x$enable_valgrind\" = \"xno\" ; then\n enable_valgrind=\"0\"\nelse\n enable_valgrind=\"1\"\nfi\n],\n[enable_valgrind=\"1\"]\n)\nif test \"x$enable_valgrind\" = \"x1\" ; then\n JE_COMPILABLE([valgrind], [\n#include \n#include \n\n#if !defined(VALGRIND_RESIZEINPLACE_BLOCK)\n# error \"Incompatible Valgrind version\"\n#endif\n], [], [je_cv_valgrind])\n if test \"x${je_cv_valgrind}\" = \"xno\" ; then\n enable_valgrind=\"0\"\n fi\n if test \"x$enable_valgrind\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_VALGRIND], [ ])\n fi\nfi\nAC_SUBST([enable_valgrind])\n\ndnl Do not support the xmalloc option by default.\nAC_ARG_ENABLE([xmalloc],\n [AS_HELP_STRING([--enable-xmalloc], [Support xmalloc option])],\n[if test \"x$enable_xmalloc\" = \"xno\" ; then\n enable_xmalloc=\"0\"\nelse\n enable_xmalloc=\"1\"\nfi\n],\n[enable_xmalloc=\"0\"]\n)\nif test \"x$enable_xmalloc\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_XMALLOC], [ ])\nfi\nAC_SUBST([enable_xmalloc])\n\nAC_CACHE_CHECK([STATIC_PAGE_SHIFT],\n [je_cv_static_page_shift],\n AC_RUN_IFELSE([AC_LANG_PROGRAM(\n[[\n#include \n#ifdef _WIN32\n#include \n#else\n#include \n#endif\n#include \n]],\n[[\n long result;\n FILE *f;\n\n#ifdef _WIN32\n SYSTEM_INFO si;\n GetSystemInfo(&si);\n result = si.dwPageSize;\n#else\n result = sysconf(_SC_PAGESIZE);\n#endif\n if (result == -1) {\n\treturn 1;\n }\n result = ffsl(result) - 1;\n\n f = fopen(\"conftest.out\", \"w\");\n if (f == NULL) {\n\treturn 1;\n }\n fprintf(f, \"%u\\n\", result);\n fclose(f);\n\n return 0;\n]])],\n [je_cv_static_page_shift=`cat conftest.out`],\n [je_cv_static_page_shift=undefined]))\n\nif test \"x$je_cv_static_page_shift\" != \"xundefined\"; then\n AC_DEFINE_UNQUOTED([STATIC_PAGE_SHIFT], [$je_cv_static_page_shift])\nelse\n AC_MSG_ERROR([cannot determine value for STATIC_PAGE_SHIFT])\nfi\n\ndnl ============================================================================\ndnl jemalloc configuration.\ndnl \n\ndnl Set VERSION if source directory has an embedded git repository.\nif test -d \"${srcroot}.git\" ; then\n git describe --long --abbrev=40 > ${srcroot}VERSION\nfi\njemalloc_version=`cat ${srcroot}VERSION`\njemalloc_version_major=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print [$]1}'`\njemalloc_version_minor=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print [$]2}'`\njemalloc_version_bugfix=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print [$]3}'`\njemalloc_version_nrev=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print [$]4}'`\njemalloc_version_gid=`echo ${jemalloc_version} | tr \".g-\" \" \" | awk '{print [$]5}'`\nAC_SUBST([jemalloc_version])\nAC_SUBST([jemalloc_version_major])\nAC_SUBST([jemalloc_version_minor])\nAC_SUBST([jemalloc_version_bugfix])\nAC_SUBST([jemalloc_version_nrev])\nAC_SUBST([jemalloc_version_gid])\n\ndnl ============================================================================\ndnl Configure pthreads.\n\nif test \"x$abi\" != \"xpecoff\" ; then\n AC_CHECK_HEADERS([pthread.h], , [AC_MSG_ERROR([pthread.h is missing])])\n dnl Some systems may embed pthreads functionality in libc; check for libpthread\n dnl first, but try libc too before failing.\n AC_CHECK_LIB([pthread], [pthread_create], [LIBS=\"$LIBS -lpthread\"],\n [AC_SEARCH_LIBS([pthread_create], , ,\n AC_MSG_ERROR([libpthread is missing]))])\nfi\n\nCPPFLAGS=\"$CPPFLAGS -D_REENTRANT\"\n\ndnl Check whether the BSD-specific _malloc_thread_cleanup() exists. If so, use\ndnl it rather than pthreads TSD cleanup functions to support cleanup during\ndnl thread exit, in order to avoid pthreads library recursion during\ndnl bootstrapping.\nAC_CHECK_FUNC([_malloc_thread_cleanup],\n [have__malloc_thread_cleanup=\"1\"],\n [have__malloc_thread_cleanup=\"0\"]\n )\nif test \"x$have__malloc_thread_cleanup\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_MALLOC_THREAD_CLEANUP], [ ])\n force_tls=\"1\"\nfi\n\ndnl Check whether the BSD-specific _pthread_mutex_init_calloc_cb() exists. If\ndnl so, mutex initialization causes allocation, and we need to implement this\ndnl callback function in order to prevent recursive allocation.\nAC_CHECK_FUNC([_pthread_mutex_init_calloc_cb],\n [have__pthread_mutex_init_calloc_cb=\"1\"],\n [have__pthread_mutex_init_calloc_cb=\"0\"]\n )\nif test \"x$have__pthread_mutex_init_calloc_cb\" = \"x1\" ; then\n AC_DEFINE([JEMALLOC_MUTEX_INIT_CB])\nfi\n\ndnl Disable lazy locking by default.\nAC_ARG_ENABLE([lazy_lock],\n [AS_HELP_STRING([--enable-lazy-lock],\n [Enable lazy locking (only lock when multi-threaded)])],\n[if test \"x$enable_lazy_lock\" = \"xno\" ; then\n enable_lazy_lock=\"0\"\nelse\n enable_lazy_lock=\"1\"\nfi\n],\n[enable_lazy_lock=\"0\"]\n)\nif test \"x$enable_lazy_lock\" = \"x0\" -a \"x${force_lazy_lock}\" = \"x1\" ; then\n AC_MSG_RESULT([Forcing lazy-lock to avoid allocator/threading bootstrap issues])\n enable_lazy_lock=\"1\"\nfi\nif test \"x$enable_lazy_lock\" = \"x1\" ; then\n if test \"x$abi\" != \"xpecoff\" ; then\n AC_CHECK_HEADERS([dlfcn.h], , [AC_MSG_ERROR([dlfcn.h is missing])])\n AC_CHECK_FUNC([dlsym], [],\n [AC_CHECK_LIB([dl], [dlsym], [LIBS=\"$LIBS -ldl\"],\n [AC_MSG_ERROR([libdl is missing])])\n ])\n fi\n AC_DEFINE([JEMALLOC_LAZY_LOCK], [ ])\nfi\nAC_SUBST([enable_lazy_lock])\n\nAC_ARG_ENABLE([tls],\n [AS_HELP_STRING([--disable-tls], [Disable thread-local storage (__thread keyword)])],\nif test \"x$enable_tls\" = \"xno\" ; then\n enable_tls=\"0\"\nelse\n enable_tls=\"1\"\nfi\n,\nenable_tls=\"1\"\n)\nif test \"x${enable_tls}\" = \"x0\" -a \"x${force_tls}\" = \"x1\" ; then\n AC_MSG_RESULT([Forcing TLS to avoid allocator/threading bootstrap issues])\n enable_tls=\"1\"\nfi\nif test \"x${enable_tls}\" = \"x1\" -a \"x${force_tls}\" = \"x0\" ; then\n AC_MSG_RESULT([Forcing no TLS to avoid allocator/threading bootstrap issues])\n enable_tls=\"0\"\nfi\nif test \"x${enable_tls}\" = \"x1\" ; then\nAC_MSG_CHECKING([for TLS])\nAC_COMPILE_IFELSE([AC_LANG_PROGRAM(\n[[\n __thread int x;\n]], [[\n x = 42;\n\n return 0;\n]])],\n AC_MSG_RESULT([yes]),\n AC_MSG_RESULT([no])\n enable_tls=\"0\")\nfi\nAC_SUBST([enable_tls])\nif test \"x${enable_tls}\" = \"x1\" ; then\n AC_DEFINE_UNQUOTED([JEMALLOC_TLS], [ ])\nelif test \"x${force_tls}\" = \"x1\" ; then\n AC_MSG_ERROR([Failed to configure TLS, which is mandatory for correct function])\nfi\n\ndnl ============================================================================\ndnl Check for ffsl(3), and fail if not found. This function exists on all\ndnl platforms that jemalloc currently has a chance of functioning on without\ndnl modification.\nJE_COMPILABLE([a program using ffsl], [\n#include \n#include \n], [\n\t{\n\t\tint rv = ffsl(0x08);\n\t}\n], [je_cv_function_ffsl])\nif test \"x${je_cv_function_ffsl}\" != \"xyes\" ; then\n AC_MSG_ERROR([Cannot build without ffsl(3)])\nfi\n\ndnl ============================================================================\ndnl Check for atomic(9) operations as provided on FreeBSD.\n\nJE_COMPILABLE([atomic(9)], [\n#include \n#include \n#include \n], [\n\t{\n\t\tuint32_t x32 = 0;\n\t\tvolatile uint32_t *x32p = &x32;\n\t\tatomic_fetchadd_32(x32p, 1);\n\t}\n\t{\n\t\tunsigned long xlong = 0;\n\t\tvolatile unsigned long *xlongp = &xlong;\n\t\tatomic_fetchadd_long(xlongp, 1);\n\t}\n], [je_cv_atomic9])\nif test \"x${je_cv_atomic9}\" = \"xyes\" ; then\n AC_DEFINE([JEMALLOC_ATOMIC9])\nfi\n\ndnl ============================================================================\ndnl Check for atomic(3) operations as provided on Darwin.\n\nJE_COMPILABLE([Darwin OSAtomic*()], [\n#include \n#include \n], [\n\t{\n\t\tint32_t x32 = 0;\n\t\tvolatile int32_t *x32p = &x32;\n\t\tOSAtomicAdd32(1, x32p);\n\t}\n\t{\n\t\tint64_t x64 = 0;\n\t\tvolatile int64_t *x64p = &x64;\n\t\tOSAtomicAdd64(1, x64p);\n\t}\n], [je_cv_osatomic])\nif test \"x${je_cv_osatomic}\" = \"xyes\" ; then\n AC_DEFINE([JEMALLOC_OSATOMIC], [ ])\nfi\n\ndnl ============================================================================\ndnl Check whether __sync_{add,sub}_and_fetch() are available despite\ndnl __GCC_HAVE_SYNC_COMPARE_AND_SWAP_n macros being undefined.\n\nAC_DEFUN([JE_SYNC_COMPARE_AND_SWAP_CHECK],[\n AC_CACHE_CHECK([whether to force $1-bit __sync_{add,sub}_and_fetch()],\n [je_cv_sync_compare_and_swap_$2],\n [AC_LINK_IFELSE([AC_LANG_PROGRAM([\n #include \n ],\n [\n #ifndef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_$2\n {\n uint$1_t x$1 = 0;\n __sync_add_and_fetch(&x$1, 42);\n __sync_sub_and_fetch(&x$1, 1);\n }\n #else\n #error __GCC_HAVE_SYNC_COMPARE_AND_SWAP_$2 is defined, no need to force\n #endif\n ])],\n [je_cv_sync_compare_and_swap_$2=yes],\n [je_cv_sync_compare_and_swap_$2=no])])\n\n if test \"x${je_cv_sync_compare_and_swap_$2}\" = \"xyes\" ; then\n AC_DEFINE([JE_FORCE_SYNC_COMPARE_AND_SWAP_$2], [ ])\n fi\n])\n\nif test \"x${je_cv_atomic9}\" != \"xyes\" -a \"x${je_cv_osatomic}\" != \"xyes\" ; then\n JE_SYNC_COMPARE_AND_SWAP_CHECK(32, 4)\n JE_SYNC_COMPARE_AND_SWAP_CHECK(64, 8)\nfi\n\ndnl ============================================================================\ndnl Check for spinlock(3) operations as provided on Darwin.\n\nJE_COMPILABLE([Darwin OSSpin*()], [\n#include \n#include \n], [\n\tOSSpinLock lock = 0;\n\tOSSpinLockLock(&lock);\n\tOSSpinLockUnlock(&lock);\n], [je_cv_osspin])\nif test \"x${je_cv_osspin}\" = \"xyes\" ; then\n AC_DEFINE([JEMALLOC_OSSPIN], [ ])\nfi\n\ndnl ============================================================================\ndnl Darwin-related configuration.\n\nif test \"x${abi}\" = \"xmacho\" ; then\n AC_DEFINE([JEMALLOC_IVSALLOC], [ ])\n AC_DEFINE([JEMALLOC_ZONE], [ ])\n\n dnl The szone version jumped from 3 to 6 between the OS X 10.5.x and 10.6\n dnl releases. malloc_zone_t and malloc_introspection_t have new fields in\n dnl 10.6, which is the only source-level indication of the change.\n AC_MSG_CHECKING([malloc zone version])\n AC_DEFUN([JE_ZONE_PROGRAM],\n [AC_LANG_PROGRAM(\n [#include ],\n [static foo[[sizeof($1) $2 sizeof(void *) * $3 ? 1 : -1]]]\n )])\n\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,14)],[JEMALLOC_ZONE_VERSION=3],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,15)],[JEMALLOC_ZONE_VERSION=5],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,16)],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_introspection_t,==,9)],[JEMALLOC_ZONE_VERSION=6],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_introspection_t,==,13)],[JEMALLOC_ZONE_VERSION=7],[JEMALLOC_ZONE_VERSION=]\n )])],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,17)],[JEMALLOC_ZONE_VERSION=8],[\n AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,>,17)],[JEMALLOC_ZONE_VERSION=9],[JEMALLOC_ZONE_VERSION=]\n )])])])])\n if test \"x${JEMALLOC_ZONE_VERSION}\" = \"x\"; then\n AC_MSG_RESULT([unsupported])\n AC_MSG_ERROR([Unsupported malloc zone version])\n fi\n if test \"${JEMALLOC_ZONE_VERSION}\" = 9; then\n JEMALLOC_ZONE_VERSION=8\n AC_MSG_RESULT([> 8])\n else\n AC_MSG_RESULT([$JEMALLOC_ZONE_VERSION])\n fi\n AC_DEFINE_UNQUOTED(JEMALLOC_ZONE_VERSION, [$JEMALLOC_ZONE_VERSION])\nfi\n\ndnl ============================================================================\ndnl Check for typedefs, structures, and compiler characteristics.\nAC_HEADER_STDBOOL\n\nAC_CONFIG_COMMANDS([include/jemalloc/internal/size_classes.h], [\n mkdir -p \"include/jemalloc/internal\"\n \"${srcdir}/include/jemalloc/internal/size_classes.sh\" > \"${objroot}include/jemalloc/internal/size_classes.h\"\n])\n\ndnl Process .in files.\nAC_SUBST([cfghdrs_in])\nAC_SUBST([cfghdrs_out])\nAC_CONFIG_HEADERS([$cfghdrs_tup])\n\ndnl ============================================================================\ndnl Generate outputs.\nAC_CONFIG_FILES([$cfgoutputs_tup config.stamp bin/jemalloc.sh])\nAC_SUBST([cfgoutputs_in])\nAC_SUBST([cfgoutputs_out])\nAC_OUTPUT\n\ndnl ============================================================================\ndnl Print out the results of configuration.\nAC_MSG_RESULT([===============================================================================])\nAC_MSG_RESULT([jemalloc version : ${jemalloc_version}])\nAC_MSG_RESULT([library revision : ${rev}])\nAC_MSG_RESULT([])\nAC_MSG_RESULT([CC : ${CC}])\nAC_MSG_RESULT([CPPFLAGS : ${CPPFLAGS}])\nAC_MSG_RESULT([CFLAGS : ${CFLAGS}])\nAC_MSG_RESULT([LDFLAGS : ${LDFLAGS}])\nAC_MSG_RESULT([LIBS : ${LIBS}])\nAC_MSG_RESULT([RPATH_EXTRA : ${RPATH_EXTRA}])\nAC_MSG_RESULT([])\nAC_MSG_RESULT([XSLTPROC : ${XSLTPROC}])\nAC_MSG_RESULT([XSLROOT : ${XSLROOT}])\nAC_MSG_RESULT([])\nAC_MSG_RESULT([PREFIX : ${PREFIX}])\nAC_MSG_RESULT([BINDIR : ${BINDIR}])\nAC_MSG_RESULT([INCLUDEDIR : ${INCLUDEDIR}])\nAC_MSG_RESULT([LIBDIR : ${LIBDIR}])\nAC_MSG_RESULT([DATADIR : ${DATADIR}])\nAC_MSG_RESULT([MANDIR : ${MANDIR}])\nAC_MSG_RESULT([])\nAC_MSG_RESULT([srcroot : ${srcroot}])\nAC_MSG_RESULT([abs_srcroot : ${abs_srcroot}])\nAC_MSG_RESULT([objroot : ${objroot}])\nAC_MSG_RESULT([abs_objroot : ${abs_objroot}])\nAC_MSG_RESULT([])\nAC_MSG_RESULT([JEMALLOC_PREFIX : ${JEMALLOC_PREFIX}])\nAC_MSG_RESULT([JEMALLOC_PRIVATE_NAMESPACE])\nAC_MSG_RESULT([ : ${JEMALLOC_PRIVATE_NAMESPACE}])\nAC_MSG_RESULT([install_suffix : ${install_suffix}])\nAC_MSG_RESULT([autogen : ${enable_autogen}])\nAC_MSG_RESULT([experimental : ${enable_experimental}])\nAC_MSG_RESULT([cc-silence : ${enable_cc_silence}])\nAC_MSG_RESULT([debug : ${enable_debug}])\nAC_MSG_RESULT([stats : ${enable_stats}])\nAC_MSG_RESULT([prof : ${enable_prof}])\nAC_MSG_RESULT([prof-libunwind : ${enable_prof_libunwind}])\nAC_MSG_RESULT([prof-libgcc : ${enable_prof_libgcc}])\nAC_MSG_RESULT([prof-gcc : ${enable_prof_gcc}])\nAC_MSG_RESULT([tcache : ${enable_tcache}])\nAC_MSG_RESULT([fill : ${enable_fill}])\nAC_MSG_RESULT([utrace : ${enable_utrace}])\nAC_MSG_RESULT([valgrind : ${enable_valgrind}])\nAC_MSG_RESULT([xmalloc : ${enable_xmalloc}])\nAC_MSG_RESULT([mremap : ${enable_mremap}])\nAC_MSG_RESULT([munmap : ${enable_munmap}])\nAC_MSG_RESULT([dss : ${enable_dss}])\nAC_MSG_RESULT([lazy_lock : ${enable_lazy_lock}])\nAC_MSG_RESULT([tls : ${enable_tls}])\nAC_MSG_RESULT([===============================================================================])\n" }, { "path": "deps/jemalloc/doc/html.xsl.in", "content": "\n \n \n\n" }, { "path": "deps/jemalloc/doc/jemalloc.3", "content": "'\\\" t\n.\\\" Title: JEMALLOC\n.\\\" Author: Jason Evans\n.\\\" Generator: DocBook XSL Stylesheets v1.76.1 \n.\\\" Date: 11/09/2012\n.\\\" Manual: User Manual\n.\\\" Source: jemalloc 3.2.0-0-g87499f6748ebe4817571e817e9f680ccb5bf54a9\n.\\\" Language: English\n.\\\"\n.TH \"JEMALLOC\" \"3\" \"11/09/2012\" \"jemalloc 3.2.0-0-g87499f6748eb\" \"User Manual\"\n.\\\" -----------------------------------------------------------------\n.\\\" * Define some portability stuff\n.\\\" -----------------------------------------------------------------\n.\\\" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n.\\\" http://bugs.debian.org/507673\n.\\\" http://lists.gnu.org/archive/html/groff/2009-02/msg00013.html\n.\\\" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n.ie \\n(.g .ds Aq \\(aq\n.el .ds Aq '\n.\\\" -----------------------------------------------------------------\n.\\\" * set default formatting\n.\\\" -----------------------------------------------------------------\n.\\\" disable hyphenation\n.nh\n.\\\" disable justification (adjust text to left margin only)\n.ad l\n.\\\" -----------------------------------------------------------------\n.\\\" * MAIN CONTENT STARTS HERE *\n.\\\" -----------------------------------------------------------------\n.SH \"NAME\"\njemalloc \\- general purpose memory allocation functions\n.SH \"LIBRARY\"\n.PP\nThis manual describes jemalloc 3\\&.2\\&.0\\-0\\-g87499f6748ebe4817571e817e9f680ccb5bf54a9\\&. More information can be found at the\n\\m[blue]\\fBjemalloc website\\fR\\m[]\\&\\s-2\\u[1]\\d\\s+2\\&.\n.SH \"SYNOPSIS\"\n.sp\n.ft B\n.nf\n#include \n#include \n.fi\n.ft\n.SS \"Standard API\"\n.HP \\w'void\\ *malloc('u\n.BI \"void *malloc(size_t\\ \" \"size\" \");\"\n.HP \\w'void\\ *calloc('u\n.BI \"void *calloc(size_t\\ \" \"number\" \", size_t\\ \" \"size\" \");\"\n.HP \\w'int\\ posix_memalign('u\n.BI \"int posix_memalign(void\\ **\" \"ptr\" \", size_t\\ \" \"alignment\" \", size_t\\ \" \"size\" \");\"\n.HP \\w'void\\ *aligned_alloc('u\n.BI \"void *aligned_alloc(size_t\\ \" \"alignment\" \", size_t\\ \" \"size\" \");\"\n.HP \\w'void\\ *realloc('u\n.BI \"void *realloc(void\\ *\" \"ptr\" \", size_t\\ \" \"size\" \");\"\n.HP \\w'void\\ free('u\n.BI \"void free(void\\ *\" \"ptr\" \");\"\n.SS \"Non\\-standard API\"\n.HP \\w'size_t\\ malloc_usable_size('u\n.BI \"size_t malloc_usable_size(const\\ void\\ *\" \"ptr\" \");\"\n.HP \\w'void\\ malloc_stats_print('u\n.BI \"void malloc_stats_print(void\\ \" \"(*write_cb)\" \"\\ (void\\ *,\\ const\\ char\\ *), void\\ *\" \"cbopaque\" \", const\\ char\\ *\" \"opts\" \");\"\n.HP \\w'int\\ mallctl('u\n.BI \"int mallctl(const\\ char\\ *\" \"name\" \", void\\ *\" \"oldp\" \", size_t\\ *\" \"oldlenp\" \", void\\ *\" \"newp\" \", size_t\\ \" \"newlen\" \");\"\n.HP \\w'int\\ mallctlnametomib('u\n.BI \"int mallctlnametomib(const\\ char\\ *\" \"name\" \", size_t\\ *\" \"mibp\" \", size_t\\ *\" \"miblenp\" \");\"\n.HP \\w'int\\ mallctlbymib('u\n.BI \"int mallctlbymib(const\\ size_t\\ *\" \"mib\" \", size_t\\ \" \"miblen\" \", void\\ *\" \"oldp\" \", size_t\\ *\" \"oldlenp\" \", void\\ *\" \"newp\" \", size_t\\ \" \"newlen\" \");\"\n.HP \\w'void\\ (*malloc_message)('u\n.BI \"void (*malloc_message)(void\\ *\" \"cbopaque\" \", const\\ char\\ *\" \"s\" \");\"\n.PP\nconst char *\\fImalloc_conf\\fR;\n.SS \"Experimental API\"\n.HP \\w'int\\ allocm('u\n.BI \"int allocm(void\\ **\" \"ptr\" \", size_t\\ *\" \"rsize\" \", size_t\\ \" \"size\" \", int\\ \" \"flags\" \");\"\n.HP \\w'int\\ rallocm('u\n.BI \"int rallocm(void\\ **\" \"ptr\" \", size_t\\ *\" \"rsize\" \", size_t\\ \" \"size\" \", size_t\\ \" \"extra\" \", int\\ \" \"flags\" \");\"\n.HP \\w'int\\ sallocm('u\n.BI \"int sallocm(const\\ void\\ *\" \"ptr\" \", size_t\\ *\" \"rsize\" \", int\\ \" \"flags\" \");\"\n.HP \\w'int\\ dallocm('u\n.BI \"int dallocm(void\\ *\" \"ptr\" \", int\\ \" \"flags\" \");\"\n.HP \\w'int\\ nallocm('u\n.BI \"int nallocm(size_t\\ *\" \"rsize\" \", size_t\\ \" \"size\" \", int\\ \" \"flags\" \");\"\n.SH \"DESCRIPTION\"\n.SS \"Standard API\"\n.PP\nThe\n\\fBmalloc\\fR\\fB\\fR\nfunction allocates\n\\fIsize\\fR\nbytes of uninitialized memory\\&. The allocated space is suitably aligned (after possible pointer coercion) for storage of any type of object\\&.\n.PP\nThe\n\\fBcalloc\\fR\\fB\\fR\nfunction allocates space for\n\\fInumber\\fR\nobjects, each\n\\fIsize\\fR\nbytes in length\\&. The result is identical to calling\n\\fBmalloc\\fR\\fB\\fR\nwith an argument of\n\\fInumber\\fR\n*\n\\fIsize\\fR, with the exception that the allocated memory is explicitly initialized to zero bytes\\&.\n.PP\nThe\n\\fBposix_memalign\\fR\\fB\\fR\nfunction allocates\n\\fIsize\\fR\nbytes of memory such that the allocation\\*(Aqs base address is an even multiple of\n\\fIalignment\\fR, and returns the allocation in the value pointed to by\n\\fIptr\\fR\\&. The requested\n\\fIalignment\\fR\nmust be a power of 2 at least as large as\nsizeof(\\fBvoid *\\fR)\\&.\n.PP\nThe\n\\fBaligned_alloc\\fR\\fB\\fR\nfunction allocates\n\\fIsize\\fR\nbytes of memory such that the allocation\\*(Aqs base address is an even multiple of\n\\fIalignment\\fR\\&. The requested\n\\fIalignment\\fR\nmust be a power of 2\\&. Behavior is undefined if\n\\fIsize\\fR\nis not an integral multiple of\n\\fIalignment\\fR\\&.\n.PP\nThe\n\\fBrealloc\\fR\\fB\\fR\nfunction changes the size of the previously allocated memory referenced by\n\\fIptr\\fR\nto\n\\fIsize\\fR\nbytes\\&. The contents of the memory are unchanged up to the lesser of the new and old sizes\\&. If the new size is larger, the contents of the newly allocated portion of the memory are undefined\\&. Upon success, the memory referenced by\n\\fIptr\\fR\nis freed and a pointer to the newly allocated memory is returned\\&. Note that\n\\fBrealloc\\fR\\fB\\fR\nmay move the memory allocation, resulting in a different return value than\n\\fIptr\\fR\\&. If\n\\fIptr\\fR\nis\n\\fBNULL\\fR, the\n\\fBrealloc\\fR\\fB\\fR\nfunction behaves identically to\n\\fBmalloc\\fR\\fB\\fR\nfor the specified size\\&.\n.PP\nThe\n\\fBfree\\fR\\fB\\fR\nfunction causes the allocated memory referenced by\n\\fIptr\\fR\nto be made available for future allocations\\&. If\n\\fIptr\\fR\nis\n\\fBNULL\\fR, no action occurs\\&.\n.SS \"Non\\-standard API\"\n.PP\nThe\n\\fBmalloc_usable_size\\fR\\fB\\fR\nfunction returns the usable size of the allocation pointed to by\n\\fIptr\\fR\\&. The return value may be larger than the size that was requested during allocation\\&. The\n\\fBmalloc_usable_size\\fR\\fB\\fR\nfunction is not a mechanism for in\\-place\n\\fBrealloc\\fR\\fB\\fR; rather it is provided solely as a tool for introspection purposes\\&. Any discrepancy between the requested allocation size and the size reported by\n\\fBmalloc_usable_size\\fR\\fB\\fR\nshould not be depended on, since such behavior is entirely implementation\\-dependent\\&.\n.PP\nThe\n\\fBmalloc_stats_print\\fR\\fB\\fR\nfunction writes human\\-readable summary statistics via the\n\\fIwrite_cb\\fR\ncallback function pointer and\n\\fIcbopaque\\fR\ndata passed to\n\\fIwrite_cb\\fR, or\n\\fBmalloc_message\\fR\\fB\\fR\nif\n\\fIwrite_cb\\fR\nis\n\\fBNULL\\fR\\&. This function can be called repeatedly\\&. General information that never changes during execution can be omitted by specifying \"g\" as a character within the\n\\fIopts\\fR\nstring\\&. Note that\n\\fBmalloc_message\\fR\\fB\\fR\nuses the\n\\fBmallctl*\\fR\\fB\\fR\nfunctions internally, so inconsistent statistics can be reported if multiple threads use these functions simultaneously\\&. If\n\\fB\\-\\-enable\\-stats\\fR\nis specified during configuration, \\(lqm\\(rq and \\(lqa\\(rq can be specified to omit merged arena and per arena statistics, respectively; \\(lqb\\(rq and \\(lql\\(rq can be specified to omit per size class statistics for bins and large objects, respectively\\&. Unrecognized characters are silently ignored\\&. Note that thread caching may prevent some statistics from being completely up to date, since extra locking would be required to merge counters that track thread cache operations\\&.\n.PP\nThe\n\\fBmallctl\\fR\\fB\\fR\nfunction provides a general interface for introspecting the memory allocator, as well as setting modifiable parameters and triggering actions\\&. The period\\-separated\n\\fIname\\fR\nargument specifies a location in a tree\\-structured namespace; see the\nMALLCTL NAMESPACE\nsection for documentation on the tree contents\\&. To read a value, pass a pointer via\n\\fIoldp\\fR\nto adequate space to contain the value, and a pointer to its length via\n\\fIoldlenp\\fR; otherwise pass\n\\fBNULL\\fR\nand\n\\fBNULL\\fR\\&. Similarly, to write a value, pass a pointer to the value via\n\\fInewp\\fR, and its length via\n\\fInewlen\\fR; otherwise pass\n\\fBNULL\\fR\nand\n\\fB0\\fR\\&.\n.PP\nThe\n\\fBmallctlnametomib\\fR\\fB\\fR\nfunction provides a way to avoid repeated name lookups for applications that repeatedly query the same portion of the namespace, by translating a name to a \\(lqManagement Information Base\\(rq (MIB) that can be passed repeatedly to\n\\fBmallctlbymib\\fR\\fB\\fR\\&. Upon successful return from\n\\fBmallctlnametomib\\fR\\fB\\fR,\n\\fImibp\\fR\ncontains an array of\n\\fI*miblenp\\fR\nintegers, where\n\\fI*miblenp\\fR\nis the lesser of the number of components in\n\\fIname\\fR\nand the input value of\n\\fI*miblenp\\fR\\&. Thus it is possible to pass a\n\\fI*miblenp\\fR\nthat is smaller than the number of period\\-separated name components, which results in a partial MIB that can be used as the basis for constructing a complete MIB\\&. For name components that are integers (e\\&.g\\&. the 2 in\n\"arenas\\&.bin\\&.2\\&.size\"), the corresponding MIB component will always be that integer\\&. Therefore, it is legitimate to construct code like the following:\n.sp\n.if n \\{\\\n.RS 4\n.\\}\n.nf\nunsigned nbins, i;\n\nint mib[4];\nsize_t len, miblen;\n\nlen = sizeof(nbins);\nmallctl(\"arenas\\&.nbins\", &nbins, &len, NULL, 0);\n\nmiblen = 4;\nmallnametomib(\"arenas\\&.bin\\&.0\\&.size\", mib, &miblen);\nfor (i = 0; i < nbins; i++) {\n\tsize_t bin_size;\n\n\tmib[2] = i;\n\tlen = sizeof(bin_size);\n\tmallctlbymib(mib, miblen, &bin_size, &len, NULL, 0);\n\t/* Do something with bin_size\\&.\\&.\\&. */\n}\n.fi\n.if n \\{\\\n.RE\n.\\}\n.SS \"Experimental API\"\n.PP\nThe experimental API is subject to change or removal without regard for backward compatibility\\&. If\n\\fB\\-\\-disable\\-experimental\\fR\nis specified during configuration, the experimental API is omitted\\&.\n.PP\nThe\n\\fBallocm\\fR\\fB\\fR,\n\\fBrallocm\\fR\\fB\\fR,\n\\fBsallocm\\fR\\fB\\fR,\n\\fBdallocm\\fR\\fB\\fR, and\n\\fBnallocm\\fR\\fB\\fR\nfunctions all have a\n\\fIflags\\fR\nargument that can be used to specify options\\&. The functions only check the options that are contextually relevant\\&. Use bitwise or (|) operations to specify one or more of the following:\n.PP\n\\fBALLOCM_LG_ALIGN(\\fR\\fB\\fIla\\fR\\fR\\fB) \\fR\n.RS 4\nAlign the memory allocation to start at an address that is a multiple of\n(1 << \\fIla\\fR)\\&. This macro does not validate that\n\\fIla\\fR\nis within the valid range\\&.\n.RE\n.PP\n\\fBALLOCM_ALIGN(\\fR\\fB\\fIa\\fR\\fR\\fB) \\fR\n.RS 4\nAlign the memory allocation to start at an address that is a multiple of\n\\fIa\\fR, where\n\\fIa\\fR\nis a power of two\\&. This macro does not validate that\n\\fIa\\fR\nis a power of 2\\&.\n.RE\n.PP\n\\fBALLOCM_ZERO\\fR\n.RS 4\nInitialize newly allocated memory to contain zero bytes\\&. In the growing reallocation case, the real size prior to reallocation defines the boundary between untouched bytes and those that are initialized to contain zero bytes\\&. If this option is absent, newly allocated memory is uninitialized\\&.\n.RE\n.PP\n\\fBALLOCM_NO_MOVE\\fR\n.RS 4\nFor reallocation, fail rather than moving the object\\&. This constraint can apply to both growth and shrinkage\\&.\n.RE\n.PP\n\\fBALLOCM_ARENA(\\fR\\fB\\fIa\\fR\\fR\\fB) \\fR\n.RS 4\nUse the arena specified by the index\n\\fIa\\fR\\&. This macro does not validate that\n\\fIa\\fR\nspecifies an arena in the valid range\\&.\n.RE\n.PP\nThe\n\\fBallocm\\fR\\fB\\fR\nfunction allocates at least\n\\fIsize\\fR\nbytes of memory, sets\n\\fI*ptr\\fR\nto the base address of the allocation, and sets\n\\fI*rsize\\fR\nto the real size of the allocation if\n\\fIrsize\\fR\nis not\n\\fBNULL\\fR\\&. Behavior is undefined if\n\\fIsize\\fR\nis\n\\fB0\\fR\\&.\n.PP\nThe\n\\fBrallocm\\fR\\fB\\fR\nfunction resizes the allocation at\n\\fI*ptr\\fR\nto be at least\n\\fIsize\\fR\nbytes, sets\n\\fI*ptr\\fR\nto the base address of the allocation if it moved, and sets\n\\fI*rsize\\fR\nto the real size of the allocation if\n\\fIrsize\\fR\nis not\n\\fBNULL\\fR\\&. If\n\\fIextra\\fR\nis non\\-zero, an attempt is made to resize the allocation to be at least\n\\fIsize\\fR + \\fIextra\\fR)\nbytes, though inability to allocate the extra byte(s) will not by itself result in failure\\&. Behavior is undefined if\n\\fIsize\\fR\nis\n\\fB0\\fR, or if\n(\\fIsize\\fR + \\fIextra\\fR > \\fBSIZE_T_MAX\\fR)\\&.\n.PP\nThe\n\\fBsallocm\\fR\\fB\\fR\nfunction sets\n\\fI*rsize\\fR\nto the real size of the allocation\\&.\n.PP\nThe\n\\fBdallocm\\fR\\fB\\fR\nfunction causes the memory referenced by\n\\fIptr\\fR\nto be made available for future allocations\\&.\n.PP\nThe\n\\fBnallocm\\fR\\fB\\fR\nfunction allocates no memory, but it performs the same size computation as the\n\\fBallocm\\fR\\fB\\fR\nfunction, and if\n\\fIrsize\\fR\nis not\n\\fBNULL\\fR\nit sets\n\\fI*rsize\\fR\nto the real size of the allocation that would result from the equivalent\n\\fBallocm\\fR\\fB\\fR\nfunction call\\&. Behavior is undefined if\n\\fIsize\\fR\nis\n\\fB0\\fR\\&.\n.SH \"TUNING\"\n.PP\nOnce, when the first call is made to one of the memory allocation routines, the allocator initializes its internals based in part on various options that can be specified at compile\\- or run\\-time\\&.\n.PP\nThe string pointed to by the global variable\n\\fImalloc_conf\\fR, the \\(lqname\\(rq of the file referenced by the symbolic link named\n/etc/malloc\\&.conf, and the value of the environment variable\n\\fBMALLOC_CONF\\fR, will be interpreted, in that order, from left to right as options\\&.\n.PP\nAn options string is a comma\\-separated list of option:value pairs\\&. There is one key corresponding to each\n\"opt\\&.*\"\nmallctl (see the\nMALLCTL NAMESPACE\nsection for options documentation)\\&. For example,\nabort:true,narenas:1\nsets the\n\"opt\\&.abort\"\nand\n\"opt\\&.narenas\"\noptions\\&. Some options have boolean values (true/false), others have integer values (base 8, 10, or 16, depending on prefix), and yet others have raw string values\\&.\n.SH \"IMPLEMENTATION NOTES\"\n.PP\nTraditionally, allocators have used\n\\fBsbrk\\fR(2)\nto obtain memory, which is suboptimal for several reasons, including race conditions, increased fragmentation, and artificial limitations on maximum usable memory\\&. If\n\\fB\\-\\-enable\\-dss\\fR\nis specified during configuration, this allocator uses both\n\\fBmmap\\fR(2)\nand\n\\fBsbrk\\fR(2), in that order of preference; otherwise only\n\\fBmmap\\fR(2)\nis used\\&.\n.PP\nThis allocator uses multiple arenas in order to reduce lock contention for threaded programs on multi\\-processor systems\\&. This works well with regard to threading scalability, but incurs some costs\\&. There is a small fixed per\\-arena overhead, and additionally, arenas manage memory completely independently of each other, which means a small fixed increase in overall memory fragmentation\\&. These overheads are not generally an issue, given the number of arenas normally used\\&. Note that using substantially more arenas than the default is not likely to improve performance, mainly due to reduced cache performance\\&. However, it may make sense to reduce the number of arenas if an application does not make much use of the allocation functions\\&.\n.PP\nIn addition to multiple arenas, unless\n\\fB\\-\\-disable\\-tcache\\fR\nis specified during configuration, this allocator supports thread\\-specific caching for small and large objects, in order to make it possible to completely avoid synchronization for most allocation requests\\&. Such caching allows very fast allocation in the common case, but it increases memory usage and fragmentation, since a bounded number of objects can remain allocated in each thread cache\\&.\n.PP\nMemory is conceptually broken into equal\\-sized chunks, where the chunk size is a power of two that is greater than the page size\\&. Chunks are always aligned to multiples of the chunk size\\&. This alignment makes it possible to find metadata for user objects very quickly\\&.\n.PP\nUser objects are broken into three categories according to size: small, large, and huge\\&. Small objects are smaller than one page\\&. Large objects are smaller than the chunk size\\&. Huge objects are a multiple of the chunk size\\&. Small and large objects are managed by arenas; huge objects are managed separately in a single data structure that is shared by all threads\\&. Huge objects are used by applications infrequently enough that this single data structure is not a scalability issue\\&.\n.PP\nEach chunk that is managed by an arena tracks its contents as runs of contiguous pages (unused, backing a set of small objects, or backing one large object)\\&. The combination of chunk alignment and chunk page maps makes it possible to determine all metadata regarding small and large allocations in constant time\\&.\n.PP\nSmall objects are managed in groups by page runs\\&. Each run maintains a frontier and free list to track which regions are in use\\&. Allocation requests that are no more than half the quantum (8 or 16, depending on architecture) are rounded up to the nearest power of two that is at least\nsizeof(\\fBdouble\\fR)\\&. All other small object size classes are multiples of the quantum, spaced such that internal fragmentation is limited to approximately 25% for all but the smallest size classes\\&. Allocation requests that are larger than the maximum small size class, but small enough to fit in an arena\\-managed chunk (see the\n\"opt\\&.lg_chunk\"\noption), are rounded up to the nearest run size\\&. Allocation requests that are too large to fit in an arena\\-managed chunk are rounded up to the nearest multiple of the chunk size\\&.\n.PP\nAllocations are packed tightly together, which can be an issue for multi\\-threaded applications\\&. If you need to assure that allocations do not suffer from cacheline sharing, round your allocation requests up to the nearest multiple of the cacheline size, or specify cacheline alignment when allocating\\&.\n.PP\nAssuming 4 MiB chunks, 4 KiB pages, and a 16\\-byte quantum on a 64\\-bit system, the size classes in each category are as shown in\nTable 1\\&.\n.sp\n.it 1 an-trap\n.nr an-no-space-flag 1\n.nr an-break-flag 1\n.br\n.B Table\\ \\&1.\\ \\&Size classes\n.TS\nallbox tab(:);\nlB rB lB.\nT{\nCategory\nT}:T{\nSpacing\nT}:T{\nSize\nT}\n.T&\nl r l\n^ r l\n^ r l\n^ r l\n^ r l\n^ r l\n^ r l\nl r l\nl r l.\nT{\nSmall\nT}:T{\nlg\nT}:T{\n[8]\nT}\n:T{\n16\nT}:T{\n[16, 32, 48, \\&.\\&.\\&., 128]\nT}\n:T{\n32\nT}:T{\n[160, 192, 224, 256]\nT}\n:T{\n64\nT}:T{\n[320, 384, 448, 512]\nT}\n:T{\n128\nT}:T{\n[640, 768, 896, 1024]\nT}\n:T{\n256\nT}:T{\n[1280, 1536, 1792, 2048]\nT}\n:T{\n512\nT}:T{\n[2560, 3072, 3584]\nT}\nT{\nLarge\nT}:T{\n4 KiB\nT}:T{\n[4 KiB, 8 KiB, 12 KiB, \\&.\\&.\\&., 4072 KiB]\nT}\nT{\nHuge\nT}:T{\n4 MiB\nT}:T{\n[4 MiB, 8 MiB, 12 MiB, \\&.\\&.\\&.]\nT}\n.TE\n.sp 1\n.SH \"MALLCTL NAMESPACE\"\n.PP\nThe following names are defined in the namespace accessible via the\n\\fBmallctl*\\fR\\fB\\fR\nfunctions\\&. Value types are specified in parentheses, their readable/writable statuses are encoded as\nrw,\nr\\-,\n\\-w, or\n\\-\\-, and required build configuration flags follow, if any\\&. A name element encoded as\n\nor\n\nindicates an integer component, where the integer varies from 0 to some upper value that must be determined via introspection\\&. In the case of\n\"stats\\&.arenas\\&.\\&.*\",\n\nequal to\n\"arenas\\&.narenas\"\ncan be used to access the summation of statistics from all arenas\\&. Take special note of the\n\"epoch\"\nmallctl, which controls refreshing of cached dynamic statistics\\&.\n.PP\n\"version\" (\\fBconst char *\\fR) r\\-\n.RS 4\nReturn the jemalloc version string\\&.\n.RE\n.PP\n\"epoch\" (\\fBuint64_t\\fR) rw\n.RS 4\nIf a value is passed in, refresh the data from which the\n\\fBmallctl*\\fR\\fB\\fR\nfunctions report values, and increment the epoch\\&. Return the current epoch\\&. This is useful for detecting whether another thread caused a refresh\\&.\n.RE\n.PP\n\"config\\&.debug\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-debug\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.dss\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-dss\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.fill\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-fill\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.lazy_lock\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-lazy\\-lock\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.mremap\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-mremap\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.munmap\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-munmap\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.prof\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-prof\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.prof_libgcc\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-disable\\-prof\\-libgcc\\fR\nwas not specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.prof_libunwind\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-prof\\-libunwind\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.stats\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-stats\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.tcache\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-disable\\-tcache\\fR\nwas not specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.tls\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-disable\\-tls\\fR\nwas not specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.utrace\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-utrace\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.valgrind\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-valgrind\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"config\\&.xmalloc\" (\\fBbool\\fR) r\\-\n.RS 4\n\\fB\\-\\-enable\\-xmalloc\\fR\nwas specified during build configuration\\&.\n.RE\n.PP\n\"opt\\&.abort\" (\\fBbool\\fR) r\\-\n.RS 4\nAbort\\-on\\-warning enabled/disabled\\&. If true, most warnings are fatal\\&. The process will call\n\\fBabort\\fR(3)\nin these cases\\&. This option is disabled by default unless\n\\fB\\-\\-enable\\-debug\\fR\nis specified during configuration, in which case it is enabled by default\\&.\n.RE\n.PP\n\"opt\\&.lg_chunk\" (\\fBsize_t\\fR) r\\-\n.RS 4\nVirtual memory chunk size (log base 2)\\&. The default chunk size is 4 MiB (2^22)\\&.\n.RE\n.PP\n\"opt\\&.dss\" (\\fBconst char *\\fR) r\\-\n.RS 4\ndss (\\fBsbrk\\fR(2)) allocation precedence as related to\n\\fBmmap\\fR(2)\nallocation\\&. The following settings are supported: \\(lqdisabled\\(rq, \\(lqprimary\\(rq, and \\(lqsecondary\\(rq (default)\\&.\n.RE\n.PP\n\"opt\\&.narenas\" (\\fBsize_t\\fR) r\\-\n.RS 4\nMaximum number of arenas to use for automatic multiplexing of threads and arenas\\&. The default is four times the number of CPUs, or one if there is a single CPU\\&.\n.RE\n.PP\n\"opt\\&.lg_dirty_mult\" (\\fBssize_t\\fR) r\\-\n.RS 4\nPer\\-arena minimum ratio (log base 2) of active to dirty pages\\&. Some dirty unused pages may be allowed to accumulate, within the limit set by the ratio (or one chunk worth of dirty pages, whichever is greater), before informing the kernel about some of those pages via\n\\fBmadvise\\fR(2)\nor a similar system call\\&. This provides the kernel with sufficient information to recycle dirty pages if physical memory becomes scarce and the pages remain unused\\&. The default minimum ratio is 8:1 (2^3:1); an option value of \\-1 will disable dirty page purging\\&.\n.RE\n.PP\n\"opt\\&.stats_print\" (\\fBbool\\fR) r\\-\n.RS 4\nEnable/disable statistics printing at exit\\&. If enabled, the\n\\fBmalloc_stats_print\\fR\\fB\\fR\nfunction is called at program exit via an\n\\fBatexit\\fR(3)\nfunction\\&. If\n\\fB\\-\\-enable\\-stats\\fR\nis specified during configuration, this has the potential to cause deadlock for a multi\\-threaded process that exits while one or more threads are executing in the memory allocation functions\\&. Therefore, this option should only be used with care; it is primarily intended as a performance tuning aid during application development\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.junk\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-fill\\fR]\n.RS 4\nJunk filling enabled/disabled\\&. If enabled, each byte of uninitialized allocated memory will be initialized to\n0xa5\\&. All deallocated memory will be initialized to\n0x5a\\&. This is intended for debugging and will impact performance negatively\\&. This option is disabled by default unless\n\\fB\\-\\-enable\\-debug\\fR\nis specified during configuration, in which case it is enabled by default unless running inside\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2\\&.\n.RE\n.PP\n\"opt\\&.quarantine\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-fill\\fR]\n.RS 4\nPer thread quarantine size in bytes\\&. If non\\-zero, each thread maintains a FIFO object quarantine that stores up to the specified number of bytes of memory\\&. The quarantined memory is not freed until it is released from quarantine, though it is immediately junk\\-filled if the\n\"opt\\&.junk\"\noption is enabled\\&. This feature is of particular use in combination with\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2, which can detect attempts to access quarantined objects\\&. This is intended for debugging and will impact performance negatively\\&. The default quarantine size is 0 unless running inside Valgrind, in which case the default is 16 MiB\\&.\n.RE\n.PP\n\"opt\\&.redzone\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-fill\\fR]\n.RS 4\nRedzones enabled/disabled\\&. If enabled, small allocations have redzones before and after them\\&. Furthermore, if the\n\"opt\\&.junk\"\noption is enabled, the redzones are checked for corruption during deallocation\\&. However, the primary intended purpose of this feature is to be used in combination with\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2, which needs redzones in order to do effective buffer overflow/underflow detection\\&. This option is intended for debugging and will impact performance negatively\\&. This option is disabled by default unless running inside Valgrind\\&.\n.RE\n.PP\n\"opt\\&.zero\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-fill\\fR]\n.RS 4\nZero filling enabled/disabled\\&. If enabled, each byte of uninitialized allocated memory will be initialized to 0\\&. Note that this initialization only happens once for each byte, so\n\\fBrealloc\\fR\\fB\\fR\nand\n\\fBrallocm\\fR\\fB\\fR\ncalls do not zero memory that was previously allocated\\&. This is intended for debugging and will impact performance negatively\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.utrace\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-utrace\\fR]\n.RS 4\nAllocation tracing based on\n\\fButrace\\fR(2)\nenabled/disabled\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.valgrind\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-valgrind\\fR]\n.RS 4\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2\nsupport enabled/disabled\\&. This option is vestigal because jemalloc auto\\-detects whether it is running inside Valgrind\\&. This option is disabled by default, unless running inside Valgrind\\&.\n.RE\n.PP\n\"opt\\&.xmalloc\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-xmalloc\\fR]\n.RS 4\nAbort\\-on\\-out\\-of\\-memory enabled/disabled\\&. If enabled, rather than returning failure for any allocation function, display a diagnostic message on\n\\fBSTDERR_FILENO\\fR\nand cause the program to drop core (using\n\\fBabort\\fR(3))\\&. If an application is designed to depend on this behavior, set the option at compile time by including the following in the source code:\n.sp\n.if n \\{\\\n.RS 4\n.\\}\n.nf\nmalloc_conf = \"xmalloc:true\";\n.fi\n.if n \\{\\\n.RE\n.\\}\n.sp\nThis option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.tcache\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nThread\\-specific caching enabled/disabled\\&. When there are multiple threads, each thread uses a thread\\-specific cache for objects up to a certain size\\&. Thread\\-specific caching allows many allocations to be satisfied without performing any thread synchronization, at the cost of increased memory use\\&. See the\n\"opt\\&.lg_tcache_max\"\noption for related tuning information\\&. This option is enabled by default unless running inside\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2\\&.\n.RE\n.PP\n\"opt\\&.lg_tcache_max\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nMaximum size class (log base 2) to cache in the thread\\-specific cache\\&. At a minimum, all small size classes are cached, and at a maximum all large size classes are cached\\&. The default maximum is 32 KiB (2^15)\\&.\n.RE\n.PP\n\"opt\\&.prof\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nMemory profiling enabled/disabled\\&. If enabled, profile memory allocation activity\\&. See the\n\"opt\\&.prof_active\"\noption for on\\-the\\-fly activation/deactivation\\&. See the\n\"opt\\&.lg_prof_sample\"\noption for probabilistic sampling control\\&. See the\n\"opt\\&.prof_accum\"\noption for control of cumulative sample reporting\\&. See the\n\"opt\\&.lg_prof_interval\"\noption for information on interval\\-triggered profile dumping, the\n\"opt\\&.prof_gdump\"\noption for information on high\\-water\\-triggered profile dumping, and the\n\"opt\\&.prof_final\"\noption for final profile dumping\\&. Profile output is compatible with the included\n\\fBpprof\\fR\nPerl script, which originates from the\n\\m[blue]\\fBgperftools package\\fR\\m[]\\&\\s-2\\u[3]\\d\\s+2\\&.\n.RE\n.PP\n\"opt\\&.prof_prefix\" (\\fBconst char *\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nFilename prefix for profile dumps\\&. If the prefix is set to the empty string, no automatic dumps will occur; this is primarily useful for disabling the automatic final heap dump (which also disables leak reporting, if enabled)\\&. The default prefix is\njeprof\\&.\n.RE\n.PP\n\"opt\\&.prof_active\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nProfiling activated/deactivated\\&. This is a secondary control mechanism that makes it possible to start the application with profiling enabled (see the\n\"opt\\&.prof\"\noption) but inactive, then toggle profiling at any time during program execution with the\n\"prof\\&.active\"\nmallctl\\&. This option is enabled by default\\&.\n.RE\n.PP\n\"opt\\&.lg_prof_sample\" (\\fBssize_t\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nAverage interval (log base 2) between allocation samples, as measured in bytes of allocation activity\\&. Increasing the sampling interval decreases profile fidelity, but also decreases the computational overhead\\&. The default sample interval is 512 KiB (2^19 B)\\&.\n.RE\n.PP\n\"opt\\&.prof_accum\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nReporting of cumulative object/byte counts in profile dumps enabled/disabled\\&. If this option is enabled, every unique backtrace must be stored for the duration of execution\\&. Depending on the application, this can impose a large memory overhead, and the cumulative counts are not always of interest\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.lg_prof_interval\" (\\fBssize_t\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nAverage interval (log base 2) between memory profile dumps, as measured in bytes of allocation activity\\&. The actual interval between dumps may be sporadic because decentralized allocation counters are used to avoid synchronization bottlenecks\\&. Profiles are dumped to files named according to the pattern\n\\&.\\&.\\&.i\\&.heap, where\n\nis controlled by the\n\"opt\\&.prof_prefix\"\noption\\&. By default, interval\\-triggered profile dumping is disabled (encoded as \\-1)\\&.\n.RE\n.PP\n\"opt\\&.prof_gdump\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nTrigger a memory profile dump every time the total virtual memory exceeds the previous maximum\\&. Profiles are dumped to files named according to the pattern\n\\&.\\&.\\&.u\\&.heap, where\n\nis controlled by the\n\"opt\\&.prof_prefix\"\noption\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"opt\\&.prof_final\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nUse an\n\\fBatexit\\fR(3)\nfunction to dump final memory usage to a file named according to the pattern\n\\&.\\&.\\&.f\\&.heap, where\n\nis controlled by the\n\"opt\\&.prof_prefix\"\noption\\&. This option is enabled by default\\&.\n.RE\n.PP\n\"opt\\&.prof_leak\" (\\fBbool\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nLeak reporting enabled/disabled\\&. If enabled, use an\n\\fBatexit\\fR(3)\nfunction to report memory leaks detected by allocation sampling\\&. See the\n\"opt\\&.prof\"\noption for information on analyzing heap profile output\\&. This option is disabled by default\\&.\n.RE\n.PP\n\"thread\\&.arena\" (\\fBunsigned\\fR) rw\n.RS 4\nGet or set the arena associated with the calling thread\\&. If the specified arena was not initialized beforehand (see the\n\"arenas\\&.initialized\"\nmallctl), it will be automatically initialized as a side effect of calling this interface\\&.\n.RE\n.PP\n\"thread\\&.allocated\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nGet the total number of bytes ever allocated by the calling thread\\&. This counter has the potential to wrap around; it is up to the application to appropriately interpret the counter in such cases\\&.\n.RE\n.PP\n\"thread\\&.allocatedp\" (\\fBuint64_t *\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nGet a pointer to the the value that is returned by the\n\"thread\\&.allocated\"\nmallctl\\&. This is useful for avoiding the overhead of repeated\n\\fBmallctl*\\fR\\fB\\fR\ncalls\\&.\n.RE\n.PP\n\"thread\\&.deallocated\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nGet the total number of bytes ever deallocated by the calling thread\\&. This counter has the potential to wrap around; it is up to the application to appropriately interpret the counter in such cases\\&.\n.RE\n.PP\n\"thread\\&.deallocatedp\" (\\fBuint64_t *\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nGet a pointer to the the value that is returned by the\n\"thread\\&.deallocated\"\nmallctl\\&. This is useful for avoiding the overhead of repeated\n\\fBmallctl*\\fR\\fB\\fR\ncalls\\&.\n.RE\n.PP\n\"thread\\&.tcache\\&.enabled\" (\\fBbool\\fR) rw [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nEnable/disable calling thread\\*(Aqs tcache\\&. The tcache is implicitly flushed as a side effect of becoming disabled (see\n\"thread\\&.tcache\\&.flush\")\\&.\n.RE\n.PP\n\"thread\\&.tcache\\&.flush\" (\\fBvoid\\fR) \\-\\- [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nFlush calling thread\\*(Aqs tcache\\&. This interface releases all cached objects and internal data structures associated with the calling thread\\*(Aqs thread\\-specific cache\\&. Ordinarily, this interface need not be called, since automatic periodic incremental garbage collection occurs, and the thread cache is automatically discarded when a thread exits\\&. However, garbage collection is triggered by allocation activity, so it is possible for a thread that stops allocating/deallocating to retain its cache indefinitely, in which case the developer may find manual flushing useful\\&.\n.RE\n.PP\n\"arena\\&.\\&.purge\" (\\fBunsigned\\fR) \\-\\-\n.RS 4\nPurge unused dirty pages for arena , or for all arenas if equals\n\"arenas\\&.narenas\"\\&.\n.RE\n.PP\n\"arena\\&.\\&.dss\" (\\fBconst char *\\fR) rw\n.RS 4\nSet the precedence of dss allocation as related to mmap allocation for arena , or for all arenas if equals\n\"arenas\\&.narenas\"\\&. See\n\"opt\\&.dss\"\nfor supported settings\\&.\n.RE\n.PP\n\"arenas\\&.narenas\" (\\fBunsigned\\fR) r\\-\n.RS 4\nCurrent limit on number of arenas\\&.\n.RE\n.PP\n\"arenas\\&.initialized\" (\\fBbool *\\fR) r\\-\n.RS 4\nAn array of\n\"arenas\\&.narenas\"\nbooleans\\&. Each boolean indicates whether the corresponding arena is initialized\\&.\n.RE\n.PP\n\"arenas\\&.quantum\" (\\fBsize_t\\fR) r\\-\n.RS 4\nQuantum size\\&.\n.RE\n.PP\n\"arenas\\&.page\" (\\fBsize_t\\fR) r\\-\n.RS 4\nPage size\\&.\n.RE\n.PP\n\"arenas\\&.tcache_max\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nMaximum thread\\-cached size class\\&.\n.RE\n.PP\n\"arenas\\&.nbins\" (\\fBunsigned\\fR) r\\-\n.RS 4\nNumber of bin size classes\\&.\n.RE\n.PP\n\"arenas\\&.nhbins\" (\\fBunsigned\\fR) r\\- [\\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nTotal number of thread cache bin size classes\\&.\n.RE\n.PP\n\"arenas\\&.bin\\&.\\&.size\" (\\fBsize_t\\fR) r\\-\n.RS 4\nMaximum size supported by size class\\&.\n.RE\n.PP\n\"arenas\\&.bin\\&.\\&.nregs\" (\\fBuint32_t\\fR) r\\-\n.RS 4\nNumber of regions per page run\\&.\n.RE\n.PP\n\"arenas\\&.bin\\&.\\&.run_size\" (\\fBsize_t\\fR) r\\-\n.RS 4\nNumber of bytes per page run\\&.\n.RE\n.PP\n\"arenas\\&.nlruns\" (\\fBsize_t\\fR) r\\-\n.RS 4\nTotal number of large size classes\\&.\n.RE\n.PP\n\"arenas\\&.lrun\\&.\\&.size\" (\\fBsize_t\\fR) r\\-\n.RS 4\nMaximum size supported by this large size class\\&.\n.RE\n.PP\n\"arenas\\&.purge\" (\\fBunsigned\\fR) \\-w\n.RS 4\nPurge unused dirty pages for the specified arena, or for all arenas if none is specified\\&.\n.RE\n.PP\n\"arenas\\&.extend\" (\\fBunsigned\\fR) r\\-\n.RS 4\nExtend the array of arenas by appending a new arena, and returning the new arena index\\&.\n.RE\n.PP\n\"prof\\&.active\" (\\fBbool\\fR) rw [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nControl whether sampling is currently active\\&. See the\n\"opt\\&.prof_active\"\noption for additional information\\&.\n.RE\n.PP\n\"prof\\&.dump\" (\\fBconst char *\\fR) \\-w [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nDump a memory profile to the specified file, or if NULL is specified, to a file according to the pattern\n\\&.\\&.\\&.m\\&.heap, where\n\nis controlled by the\n\"opt\\&.prof_prefix\"\noption\\&.\n.RE\n.PP\n\"prof\\&.interval\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-prof\\fR]\n.RS 4\nAverage number of bytes allocated between inverval\\-based profile dumps\\&. See the\n\"opt\\&.lg_prof_interval\"\noption for additional information\\&.\n.RE\n.PP\n\"stats\\&.cactive\" (\\fBsize_t *\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nPointer to a counter that contains an approximate count of the current number of bytes in active pages\\&. The estimate may be high, but never low, because each arena rounds up to the nearest multiple of the chunk size when computing its contribution to the counter\\&. Note that the\n\"epoch\"\nmallctl has no bearing on this counter\\&. Furthermore, counter consistency is maintained via atomic operations, so it is necessary to use an atomic operation in order to guarantee a consistent read when dereferencing the pointer\\&.\n.RE\n.PP\n\"stats\\&.allocated\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nTotal number of bytes allocated by the application\\&.\n.RE\n.PP\n\"stats\\&.active\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nTotal number of bytes in active pages allocated by the application\\&. This is a multiple of the page size, and greater than or equal to\n\"stats\\&.allocated\"\\&. This does not include\n\"stats\\&.arenas\\&.\\&.pdirty\"\nand pages entirely devoted to allocator metadata\\&.\n.RE\n.PP\n\"stats\\&.mapped\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nTotal number of bytes in chunks mapped on behalf of the application\\&. This is a multiple of the chunk size, and is at least as large as\n\"stats\\&.active\"\\&. This does not include inactive chunks\\&.\n.RE\n.PP\n\"stats\\&.chunks\\&.current\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nTotal number of chunks actively mapped on behalf of the application\\&. This does not include inactive chunks\\&.\n.RE\n.PP\n\"stats\\&.chunks\\&.total\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of chunks allocated\\&.\n.RE\n.PP\n\"stats\\&.chunks\\&.high\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nMaximum number of active chunks at any time thus far\\&.\n.RE\n.PP\n\"stats\\&.huge\\&.allocated\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of bytes currently allocated by huge objects\\&.\n.RE\n.PP\n\"stats\\&.huge\\&.nmalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of huge allocation requests\\&.\n.RE\n.PP\n\"stats\\&.huge\\&.ndalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of huge deallocation requests\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.dss\" (\\fBconst char *\\fR) r\\-\n.RS 4\ndss (\\fBsbrk\\fR(2)) allocation precedence as related to\n\\fBmmap\\fR(2)\nallocation\\&. See\n\"opt\\&.dss\"\nfor details\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.nthreads\" (\\fBunsigned\\fR) r\\-\n.RS 4\nNumber of threads currently assigned to arena\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.pactive\" (\\fBsize_t\\fR) r\\-\n.RS 4\nNumber of pages in active runs\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.pdirty\" (\\fBsize_t\\fR) r\\-\n.RS 4\nNumber of pages within unused runs that are potentially dirty, and for which\n\\fBmadvise\\fR\\fB\\fI\\&.\\&.\\&.\\fR\\fR\\fB \\fR\\fB\\fI\\fBMADV_DONTNEED\\fR\\fR\\fR\nor similar has not been called\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.mapped\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of mapped bytes\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.npurge\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of dirty page purge sweeps performed\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.nmadvise\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of\n\\fBmadvise\\fR\\fB\\fI\\&.\\&.\\&.\\fR\\fR\\fB \\fR\\fB\\fI\\fBMADV_DONTNEED\\fR\\fR\\fR\nor similar calls made to purge dirty pages\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.npurged\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of pages purged\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.small\\&.allocated\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of bytes currently allocated by small objects\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.small\\&.nmalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocation requests served by small bins\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.small\\&.ndalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of small objects returned to bins\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.small\\&.nrequests\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of small allocation requests\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.large\\&.allocated\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nNumber of bytes currently allocated by large objects\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.large\\&.nmalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of large allocation requests served directly by the arena\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.large\\&.ndalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of large deallocation requests served directly by the arena\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.large\\&.nrequests\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of large allocation requests\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.allocated\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCurrent number of bytes allocated by bin\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nmalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocations served by bin\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.ndalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocations returned to bin\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nrequests\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocation requests\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nfills\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR \\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nCumulative number of tcache fills\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nflushes\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR \\fB\\-\\-enable\\-tcache\\fR]\n.RS 4\nCumulative number of tcache flushes\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nruns\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of runs created\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.nreruns\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of times the current run from which to allocate changed\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.bins\\&.\\&.curruns\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCurrent number of runs\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.lruns\\&.\\&.nmalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocation requests for this size class served directly by the arena\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.lruns\\&.\\&.ndalloc\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of deallocation requests for this size class served directly by the arena\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.lruns\\&.\\&.nrequests\" (\\fBuint64_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCumulative number of allocation requests for this size class\\&.\n.RE\n.PP\n\"stats\\&.arenas\\&.\\&.lruns\\&.\\&.curruns\" (\\fBsize_t\\fR) r\\- [\\fB\\-\\-enable\\-stats\\fR]\n.RS 4\nCurrent number of runs for this size class\\&.\n.RE\n.SH \"DEBUGGING MALLOC PROBLEMS\"\n.PP\nWhen debugging, it is a good idea to configure/build jemalloc with the\n\\fB\\-\\-enable\\-debug\\fR\nand\n\\fB\\-\\-enable\\-fill\\fR\noptions, and recompile the program with suitable options and symbols for debugger support\\&. When so configured, jemalloc incorporates a wide variety of run\\-time assertions that catch application errors such as double\\-free, write\\-after\\-free, etc\\&.\n.PP\nPrograms often accidentally depend on \\(lquninitialized\\(rq memory actually being filled with zero bytes\\&. Junk filling (see the\n\"opt\\&.junk\"\noption) tends to expose such bugs in the form of obviously incorrect results and/or coredumps\\&. Conversely, zero filling (see the\n\"opt\\&.zero\"\noption) eliminates the symptoms of such bugs\\&. Between these two options, it is usually possible to quickly detect, diagnose, and eliminate such bugs\\&.\n.PP\nThis implementation does not provide much detail about the problems it detects, because the performance impact for storing such information would be prohibitive\\&. However, jemalloc does integrate with the most excellent\n\\m[blue]\\fBValgrind\\fR\\m[]\\&\\s-2\\u[2]\\d\\s+2\ntool if the\n\\fB\\-\\-enable\\-valgrind\\fR\nconfiguration option is enabled\\&.\n.SH \"DIAGNOSTIC MESSAGES\"\n.PP\nIf any of the memory allocation/deallocation functions detect an error or warning condition, a message will be printed to file descriptor\n\\fBSTDERR_FILENO\\fR\\&. Errors will result in the process dumping core\\&. If the\n\"opt\\&.abort\"\noption is set, most warnings are treated as errors\\&.\n.PP\nThe\n\\fImalloc_message\\fR\nvariable allows the programmer to override the function which emits the text strings forming the errors and warnings if for some reason the\n\\fBSTDERR_FILENO\\fR\nfile descriptor is not suitable for this\\&.\n\\fBmalloc_message\\fR\\fB\\fR\ntakes the\n\\fIcbopaque\\fR\npointer argument that is\n\\fBNULL\\fR\nunless overridden by the arguments in a call to\n\\fBmalloc_stats_print\\fR\\fB\\fR, followed by a string pointer\\&. Please note that doing anything which tries to allocate memory in this function is likely to result in a crash or deadlock\\&.\n.PP\nAll messages are prefixed by \\(lq:\\(rq\\&.\n.SH \"RETURN VALUES\"\n.SS \"Standard API\"\n.PP\nThe\n\\fBmalloc\\fR\\fB\\fR\nand\n\\fBcalloc\\fR\\fB\\fR\nfunctions return a pointer to the allocated memory if successful; otherwise a\n\\fBNULL\\fR\npointer is returned and\n\\fIerrno\\fR\nis set to\nENOMEM\\&.\n.PP\nThe\n\\fBposix_memalign\\fR\\fB\\fR\nfunction returns the value 0 if successful; otherwise it returns an error value\\&. The\n\\fBposix_memalign\\fR\\fB\\fR\nfunction will fail if:\n.PP\nEINVAL\n.RS 4\nThe\n\\fIalignment\\fR\nparameter is not a power of 2 at least as large as\nsizeof(\\fBvoid *\\fR)\\&.\n.RE\n.PP\nENOMEM\n.RS 4\nMemory allocation error\\&.\n.RE\n.PP\nThe\n\\fBaligned_alloc\\fR\\fB\\fR\nfunction returns a pointer to the allocated memory if successful; otherwise a\n\\fBNULL\\fR\npointer is returned and\n\\fIerrno\\fR\nis set\\&. The\n\\fBaligned_alloc\\fR\\fB\\fR\nfunction will fail if:\n.PP\nEINVAL\n.RS 4\nThe\n\\fIalignment\\fR\nparameter is not a power of 2\\&.\n.RE\n.PP\nENOMEM\n.RS 4\nMemory allocation error\\&.\n.RE\n.PP\nThe\n\\fBrealloc\\fR\\fB\\fR\nfunction returns a pointer, possibly identical to\n\\fIptr\\fR, to the allocated memory if successful; otherwise a\n\\fBNULL\\fR\npointer is returned, and\n\\fIerrno\\fR\nis set to\nENOMEM\nif the error was the result of an allocation failure\\&. The\n\\fBrealloc\\fR\\fB\\fR\nfunction always leaves the original buffer intact when an error occurs\\&.\n.PP\nThe\n\\fBfree\\fR\\fB\\fR\nfunction returns no value\\&.\n.SS \"Non\\-standard API\"\n.PP\nThe\n\\fBmalloc_usable_size\\fR\\fB\\fR\nfunction returns the usable size of the allocation pointed to by\n\\fIptr\\fR\\&.\n.PP\nThe\n\\fBmallctl\\fR\\fB\\fR,\n\\fBmallctlnametomib\\fR\\fB\\fR, and\n\\fBmallctlbymib\\fR\\fB\\fR\nfunctions return 0 on success; otherwise they return an error value\\&. The functions will fail if:\n.PP\nEINVAL\n.RS 4\n\\fInewp\\fR\nis not\n\\fBNULL\\fR, and\n\\fInewlen\\fR\nis too large or too small\\&. Alternatively,\n\\fI*oldlenp\\fR\nis too large or too small; in this case as much data as possible are read despite the error\\&.\n.RE\n.PP\nENOMEM\n.RS 4\n\\fI*oldlenp\\fR\nis too short to hold the requested value\\&.\n.RE\n.PP\nENOENT\n.RS 4\n\\fIname\\fR\nor\n\\fImib\\fR\nspecifies an unknown/invalid value\\&.\n.RE\n.PP\nEPERM\n.RS 4\nAttempt to read or write void value, or attempt to write read\\-only value\\&.\n.RE\n.PP\nEAGAIN\n.RS 4\nA memory allocation failure occurred\\&.\n.RE\n.PP\nEFAULT\n.RS 4\nAn interface with side effects failed in some way not directly related to\n\\fBmallctl*\\fR\\fB\\fR\nread/write processing\\&.\n.RE\n.SS \"Experimental API\"\n.PP\nThe\n\\fBallocm\\fR\\fB\\fR,\n\\fBrallocm\\fR\\fB\\fR,\n\\fBsallocm\\fR\\fB\\fR,\n\\fBdallocm\\fR\\fB\\fR, and\n\\fBnallocm\\fR\\fB\\fR\nfunctions return\n\\fBALLOCM_SUCCESS\\fR\non success; otherwise they return an error value\\&. The\n\\fBallocm\\fR\\fB\\fR,\n\\fBrallocm\\fR\\fB\\fR, and\n\\fBnallocm\\fR\\fB\\fR\nfunctions will fail if:\n.PP\nALLOCM_ERR_OOM\n.RS 4\nOut of memory\\&. Insufficient contiguous memory was available to service the allocation request\\&. The\n\\fBallocm\\fR\\fB\\fR\nfunction additionally sets\n\\fI*ptr\\fR\nto\n\\fBNULL\\fR, whereas the\n\\fBrallocm\\fR\\fB\\fR\nfunction leaves\n\\fB*ptr\\fR\nunmodified\\&.\n.RE\nThe\n\\fBrallocm\\fR\\fB\\fR\nfunction will also fail if:\n.PP\nALLOCM_ERR_NOT_MOVED\n.RS 4\n\\fBALLOCM_NO_MOVE\\fR\nwas specified, but the reallocation request could not be serviced without moving the object\\&.\n.RE\n.SH \"ENVIRONMENT\"\n.PP\nThe following environment variable affects the execution of the allocation functions:\n.PP\n\\fBMALLOC_CONF\\fR\n.RS 4\nIf the environment variable\n\\fBMALLOC_CONF\\fR\nis set, the characters it contains will be interpreted as options\\&.\n.RE\n.SH \"EXAMPLES\"\n.PP\nTo dump core whenever a problem occurs:\n.sp\n.if n \\{\\\n.RS 4\n.\\}\n.nf\nln \\-s \\*(Aqabort:true\\*(Aq /etc/malloc\\&.conf\n.fi\n.if n \\{\\\n.RE\n.\\}\n.PP\nTo specify in the source a chunk size that is 16 MiB:\n.sp\n.if n \\{\\\n.RS 4\n.\\}\n.nf\nmalloc_conf = \"lg_chunk:24\";\n.fi\n.if n \\{\\\n.RE\n.\\}\n.SH \"SEE ALSO\"\n.PP\n\\fBmadvise\\fR(2),\n\\fBmmap\\fR(2),\n\\fBsbrk\\fR(2),\n\\fButrace\\fR(2),\n\\fBalloca\\fR(3),\n\\fBatexit\\fR(3),\n\\fBgetpagesize\\fR(3)\n.SH \"STANDARDS\"\n.PP\nThe\n\\fBmalloc\\fR\\fB\\fR,\n\\fBcalloc\\fR\\fB\\fR,\n\\fBrealloc\\fR\\fB\\fR, and\n\\fBfree\\fR\\fB\\fR\nfunctions conform to ISO/IEC 9899:1990 (\\(lqISO C90\\(rq)\\&.\n.PP\nThe\n\\fBposix_memalign\\fR\\fB\\fR\nfunction conforms to IEEE Std 1003\\&.1\\-2001 (\\(lqPOSIX\\&.1\\(rq)\\&.\n.SH \"AUTHOR\"\n.PP\n\\fBJason Evans\\fR\n.RS 4\n.RE\n.SH \"NOTES\"\n.IP \" 1.\" 4\njemalloc website\n.RS 4\n\\%http://www.canonware.com/jemalloc/\n.RE\n.IP \" 2.\" 4\nValgrind\n.RS 4\n\\%http://valgrind.org/\n.RE\n.IP \" 3.\" 4\ngperftools package\n.RS 4\n\\%http://code.google.com/p/gperftools/\n.RE\n" }, { "path": "deps/jemalloc/doc/jemalloc.html", "content": "JEMALLOC

Name

jemalloc — general purpose memory allocation functions

LIBRARY

This manual describes jemalloc 3.2.0-0-g87499f6748ebe4817571e817e9f680ccb5bf54a9. More information\n can be found at the jemalloc website.

SYNOPSIS

#include <stdlib.h>\n#include <jemalloc/jemalloc.h>

Standard API

void *malloc(size_t size);
void *calloc(size_t number,
size_t size);
int posix_memalign(void **ptr,
size_t alignment,
size_t size);
void *aligned_alloc(size_t alignment,
size_t size);
void *realloc(void *ptr,
size_t size);
void free(void *ptr);

Non-standard API

size_t malloc_usable_size(const void *ptr);
void malloc_stats_print(void (*write_cb)\n (void *, const char *)\n ,
void *cbopaque,
const char *opts);
int mallctl(const char *name,
void *oldp,
size_t *oldlenp,
void *newp,
size_t newlen);
int mallctlnametomib(const char *name,
size_t *mibp,
size_t *miblenp);
int mallctlbymib(const size_t *mib,
size_t miblen,
void *oldp,
size_t *oldlenp,
void *newp,
size_t newlen);
void (*malloc_message)(void *cbopaque,
const char *s);

const char *malloc_conf;

Experimental API

int allocm(void **ptr,
size_t *rsize,
size_t size,
int flags);
int rallocm(void **ptr,
size_t *rsize,
size_t size,
size_t extra,
int flags);
int sallocm(const void *ptr,
size_t *rsize,
int flags);
int dallocm(void *ptr,
int flags);
int nallocm(size_t *rsize,
size_t size,
int flags);

DESCRIPTION

Standard API

The malloc() function allocates\n size bytes of uninitialized memory. The allocated\n space is suitably aligned (after possible pointer coercion) for storage\n of any type of object.

The calloc() function allocates\n space for number objects, each\n size bytes in length. The result is identical to\n calling malloc() with an argument of\n number * size, with the\n exception that the allocated memory is explicitly initialized to zero\n bytes.

The posix_memalign() function\n allocates size bytes of memory such that the\n allocation's base address is an even multiple of\n alignment, and returns the allocation in the value\n pointed to by ptr. The requested\n alignment must be a power of 2 at least as large\n as sizeof(void *).

The aligned_alloc() function\n allocates size bytes of memory such that the\n allocation's base address is an even multiple of\n alignment. The requested\n alignment must be a power of 2. Behavior is\n undefined if size is not an integral multiple of\n alignment.

The realloc() function changes the\n size of the previously allocated memory referenced by\n ptr to size bytes. The\n contents of the memory are unchanged up to the lesser of the new and old\n sizes. If the new size is larger, the contents of the newly allocated\n portion of the memory are undefined. Upon success, the memory referenced\n by ptr is freed and a pointer to the newly\n allocated memory is returned. Note that\n realloc() may move the memory allocation,\n resulting in a different return value than ptr.\n If ptr is NULL, the\n realloc() function behaves identically to\n malloc() for the specified size.

The free() function causes the\n allocated memory referenced by ptr to be made\n available for future allocations. If ptr is\n NULL, no action occurs.

Non-standard API

The malloc_usable_size() function\n returns the usable size of the allocation pointed to by\n ptr. The return value may be larger than the size\n that was requested during allocation. The\n malloc_usable_size() function is not a\n mechanism for in-place realloc(); rather\n it is provided solely as a tool for introspection purposes. Any\n discrepancy between the requested allocation size and the size reported\n by malloc_usable_size() should not be\n depended on, since such behavior is entirely implementation-dependent.\n

The malloc_stats_print() function\n writes human-readable summary statistics via the\n write_cb callback function pointer and\n cbopaque data passed to\n write_cb, or\n malloc_message() if\n write_cb is NULL. This\n function can be called repeatedly. General information that never\n changes during execution can be omitted by specifying \"g\" as a character\n within the opts string. Note that\n malloc_message() uses the\n mallctl*() functions internally, so\n inconsistent statistics can be reported if multiple threads use these\n functions simultaneously. If --enable-stats is\n specified during configuration, “m” and “a” can\n be specified to omit merged arena and per arena statistics, respectively;\n “b” and “l” can be specified to omit per size\n class statistics for bins and large objects, respectively. Unrecognized\n characters are silently ignored. Note that thread caching may prevent\n some statistics from being completely up to date, since extra locking\n would be required to merge counters that track thread cache operations.\n

The mallctl() function provides a\n general interface for introspecting the memory allocator, as well as\n setting modifiable parameters and triggering actions. The\n period-separated name argument specifies a\n location in a tree-structured namespace; see the MALLCTL NAMESPACE section for\n documentation on the tree contents. To read a value, pass a pointer via\n oldp to adequate space to contain the value, and a\n pointer to its length via oldlenp; otherwise pass\n NULL and NULL. Similarly, to\n write a value, pass a pointer to the value via\n newp, and its length via\n newlen; otherwise pass NULL\n and 0.

The mallctlnametomib() function\n provides a way to avoid repeated name lookups for applications that\n repeatedly query the same portion of the namespace, by translating a name\n to a “Management Information Base” (MIB) that can be passed\n repeatedly to mallctlbymib(). Upon\n successful return from mallctlnametomib(),\n mibp contains an array of\n *miblenp integers, where\n *miblenp is the lesser of the number of components\n in name and the input value of\n *miblenp. Thus it is possible to pass a\n *miblenp that is smaller than the number of\n period-separated name components, which results in a partial MIB that can\n be used as the basis for constructing a complete MIB. For name\n components that are integers (e.g. the 2 in\n \n \"arenas.bin.2.size\"\n ),\n the corresponding MIB component will always be that integer. Therefore,\n it is legitimate to construct code like the following: 

\nunsigned nbins, i;\n\nint mib[4];\nsize_t len, miblen;\n\nlen = sizeof(nbins);\nmallctl(\"arenas.nbins\", &nbins, &len, NULL, 0);\n\nmiblen = 4;\nmallnametomib(\"arenas.bin.0.size\", mib, &miblen);\nfor (i = 0; i < nbins; i++) {\n\tsize_t bin_size;\n\n\tmib[2] = i;\n\tlen = sizeof(bin_size);\n\tmallctlbymib(mib, miblen, &bin_size, &len, NULL, 0);\n\t/* Do something with bin_size... */\n}

Experimental API

The experimental API is subject to change or removal without regard\n for backward compatibility. If --disable-experimental\n is specified during configuration, the experimental API is\n omitted.

The allocm(),\n rallocm(),\n sallocm(),\n dallocm(), and\n nallocm() functions all have a\n flags argument that can be used to specify\n options. The functions only check the options that are contextually\n relevant. Use bitwise or (|) operations to\n specify one or more of the following:\n

ALLOCM_LG_ALIGN(la)\n

Align the memory allocation to start at an address\n that is a multiple of (1 <<\n la). This macro does not validate\n that la is within the valid\n range.

ALLOCM_ALIGN(a)\n

Align the memory allocation to start at an address\n that is a multiple of a, where\n a is a power of two. This macro does not\n validate that a is a power of 2.\n

ALLOCM_ZERO

Initialize newly allocated memory to contain zero\n bytes. In the growing reallocation case, the real size prior to\n reallocation defines the boundary between untouched bytes and those\n that are initialized to contain zero bytes. If this option is\n absent, newly allocated memory is uninitialized.

ALLOCM_NO_MOVE

For reallocation, fail rather than moving the\n object. This constraint can apply to both growth and\n shrinkage.

ALLOCM_ARENA(a)\n

Use the arena specified by the index\n a. This macro does not validate that\n a specifies an arena in the valid\n range.

\n

The allocm() function allocates at\n least size bytes of memory, sets\n *ptr to the base address of the allocation, and\n sets *rsize to the real size of the allocation if\n rsize is not NULL. Behavior\n is undefined if size is\n 0.

The rallocm() function resizes the\n allocation at *ptr to be at least\n size bytes, sets *ptr to\n the base address of the allocation if it moved, and sets\n *rsize to the real size of the allocation if\n rsize is not NULL. If\n extra is non-zero, an attempt is made to resize\n the allocation to be at least size +\n extra) bytes, though inability to allocate\n the extra byte(s) will not by itself result in failure. Behavior is\n undefined if size is 0, or if\n (size +\n extra >\n SIZE_T_MAX).

The sallocm() function sets\n *rsize to the real size of the allocation.

The dallocm() function causes the\n memory referenced by ptr to be made available for\n future allocations.

The nallocm() function allocates no\n memory, but it performs the same size computation as the\n allocm() function, and if\n rsize is not NULL it sets\n *rsize to the real size of the allocation that\n would result from the equivalent allocm()\n function call. Behavior is undefined if\n size is 0.

TUNING

Once, when the first call is made to one of the memory allocation\n routines, the allocator initializes its internals based in part on various\n options that can be specified at compile- or run-time.

The string pointed to by the global variable\n malloc_conf, the “name” of the file\n referenced by the symbolic link named /etc/malloc.conf, and the value of the\n environment variable MALLOC_CONF, will be interpreted, in\n that order, from left to right as options.

An options string is a comma-separated list of option:value pairs.\n There is one key corresponding to each \n \"opt.*\"\n mallctl (see the MALLCTL NAMESPACE section for options\n documentation). For example, abort:true,narenas:1 sets\n the \n \"opt.abort\"\n and \n \"opt.narenas\"\n options. Some\n options have boolean values (true/false), others have integer values (base\n 8, 10, or 16, depending on prefix), and yet others have raw string\n values.

IMPLEMENTATION NOTES

Traditionally, allocators have used\n sbrk(2) to obtain memory, which is\n suboptimal for several reasons, including race conditions, increased\n fragmentation, and artificial limitations on maximum usable memory. If\n --enable-dss is specified during configuration, this\n allocator uses both mmap(2) and\n sbrk(2), in that order of preference;\n otherwise only mmap(2) is used.

This allocator uses multiple arenas in order to reduce lock\n contention for threaded programs on multi-processor systems. This works\n well with regard to threading scalability, but incurs some costs. There is\n a small fixed per-arena overhead, and additionally, arenas manage memory\n completely independently of each other, which means a small fixed increase\n in overall memory fragmentation. These overheads are not generally an\n issue, given the number of arenas normally used. Note that using\n substantially more arenas than the default is not likely to improve\n performance, mainly due to reduced cache performance. However, it may make\n sense to reduce the number of arenas if an application does not make much\n use of the allocation functions.

In addition to multiple arenas, unless\n --disable-tcache is specified during configuration, this\n allocator supports thread-specific caching for small and large objects, in\n order to make it possible to completely avoid synchronization for most\n allocation requests. Such caching allows very fast allocation in the\n common case, but it increases memory usage and fragmentation, since a\n bounded number of objects can remain allocated in each thread cache.

Memory is conceptually broken into equal-sized chunks, where the\n chunk size is a power of two that is greater than the page size. Chunks\n are always aligned to multiples of the chunk size. This alignment makes it\n possible to find metadata for user objects very quickly.

User objects are broken into three categories according to size:\n small, large, and huge. Small objects are smaller than one page. Large\n objects are smaller than the chunk size. Huge objects are a multiple of\n the chunk size. Small and large objects are managed by arenas; huge\n objects are managed separately in a single data structure that is shared by\n all threads. Huge objects are used by applications infrequently enough\n that this single data structure is not a scalability issue.

Each chunk that is managed by an arena tracks its contents as runs of\n contiguous pages (unused, backing a set of small objects, or backing one\n large object). The combination of chunk alignment and chunk page maps\n makes it possible to determine all metadata regarding small and large\n allocations in constant time.

Small objects are managed in groups by page runs. Each run maintains\n a frontier and free list to track which regions are in use. Allocation\n requests that are no more than half the quantum (8 or 16, depending on\n architecture) are rounded up to the nearest power of two that is at least\n sizeof(double). All other small\n object size classes are multiples of the quantum, spaced such that internal\n fragmentation is limited to approximately 25% for all but the smallest size\n classes. Allocation requests that are larger than the maximum small size\n class, but small enough to fit in an arena-managed chunk (see the \n \"opt.lg_chunk\"\n option), are\n rounded up to the nearest run size. Allocation requests that are too large\n to fit in an arena-managed chunk are rounded up to the nearest multiple of\n the chunk size.

Allocations are packed tightly together, which can be an issue for\n multi-threaded applications. If you need to assure that allocations do not\n suffer from cacheline sharing, round your allocation requests up to the\n nearest multiple of the cacheline size, or specify cacheline alignment when\n allocating.

Assuming 4 MiB chunks, 4 KiB pages, and a 16-byte quantum on a 64-bit\n system, the size classes in each category are as shown in Table 1.

Table1.Size classes

CategorySpacingSize
Smalllg[8]
16[16, 32, 48, ..., 128]
32[160, 192, 224, 256]
64[320, 384, 448, 512]
128[640, 768, 896, 1024]
256[1280, 1536, 1792, 2048]
512[2560, 3072, 3584]
Large4 KiB[4 KiB, 8 KiB, 12 KiB, ..., 4072 KiB]
Huge4 MiB[4 MiB, 8 MiB, 12 MiB, ...]

MALLCTL NAMESPACE

The following names are defined in the namespace accessible via the\n mallctl*() functions. Value types are\n specified in parentheses, their readable/writable statuses are encoded as\n rw, r-, -w, or\n --, and required build configuration flags follow, if\n any. A name element encoded as <i> or\n <j> indicates an integer component, where the\n integer varies from 0 to some upper value that must be determined via\n introspection. In the case of \n \"stats.arenas.<i>.*\"\n ,\n <i> equal to \n \"arenas.narenas\"\n can be\n used to access the summation of statistics from all arenas. Take special\n note of the \n \"epoch\"\n mallctl,\n which controls refreshing of cached dynamic statistics.

\n \n \"version\"\n \n (const char *)\n r-\n

Return the jemalloc version string.

\n \n \"epoch\"\n \n (uint64_t)\n rw\n

If a value is passed in, refresh the data from which\n the mallctl*() functions report values,\n and increment the epoch. Return the current epoch. This is useful for\n detecting whether another thread caused a refresh.

\n \n \"config.debug\"\n \n (bool)\n r-\n

--enable-debug was specified during\n build configuration.

\n \n \"config.dss\"\n \n (bool)\n r-\n

--enable-dss was specified during\n build configuration.

\n \n \"config.fill\"\n \n (bool)\n r-\n

--enable-fill was specified during\n build configuration.

\n \n \"config.lazy_lock\"\n \n (bool)\n r-\n

--enable-lazy-lock was specified\n during build configuration.

\n \n \"config.mremap\"\n \n (bool)\n r-\n

--enable-mremap was specified during\n build configuration.

\n \n \"config.munmap\"\n \n (bool)\n r-\n

--enable-munmap was specified during\n build configuration.

\n \n \"config.prof\"\n \n (bool)\n r-\n

--enable-prof was specified during\n build configuration.

\n \n \"config.prof_libgcc\"\n \n (bool)\n r-\n

--disable-prof-libgcc was not\n specified during build configuration.

\n \n \"config.prof_libunwind\"\n \n (bool)\n r-\n

--enable-prof-libunwind was specified\n during build configuration.

\n \n \"config.stats\"\n \n (bool)\n r-\n

--enable-stats was specified during\n build configuration.

\n \n \"config.tcache\"\n \n (bool)\n r-\n

--disable-tcache was not specified\n during build configuration.

\n \n \"config.tls\"\n \n (bool)\n r-\n

--disable-tls was not specified during\n build configuration.

\n \n \"config.utrace\"\n \n (bool)\n r-\n

--enable-utrace was specified during\n build configuration.

\n \n \"config.valgrind\"\n \n (bool)\n r-\n

--enable-valgrind was specified during\n build configuration.

\n \n \"config.xmalloc\"\n \n (bool)\n r-\n

--enable-xmalloc was specified during\n build configuration.

\n \n \"opt.abort\"\n \n (bool)\n r-\n

Abort-on-warning enabled/disabled. If true, most\n warnings are fatal. The process will call\n abort(3) in these cases. This option is\n disabled by default unless --enable-debug is\n specified during configuration, in which case it is enabled by default.\n

\n \n \"opt.lg_chunk\"\n \n (size_t)\n r-\n

Virtual memory chunk size (log base 2). The default\n chunk size is 4 MiB (2^22).

\n \n \"opt.dss\"\n \n (const char *)\n r-\n

dss (sbrk(2)) allocation precedence as\n related to mmap(2) allocation. The following\n settings are supported: “disabled”, “primary”,\n and “secondary” (default).

\n \n \"opt.narenas\"\n \n (size_t)\n r-\n

Maximum number of arenas to use for automatic\n multiplexing of threads and arenas. The default is four times the\n number of CPUs, or one if there is a single CPU.

\n \n \"opt.lg_dirty_mult\"\n \n (ssize_t)\n r-\n

Per-arena minimum ratio (log base 2) of active to dirty\n pages. Some dirty unused pages may be allowed to accumulate, within\n the limit set by the ratio (or one chunk worth of dirty pages,\n whichever is greater), before informing the kernel about some of those\n pages via madvise(2) or a similar system call. This\n provides the kernel with sufficient information to recycle dirty pages\n if physical memory becomes scarce and the pages remain unused. The\n default minimum ratio is 8:1 (2^3:1); an option value of -1 will\n disable dirty page purging.

\n \n \"opt.stats_print\"\n \n (bool)\n r-\n

Enable/disable statistics printing at exit. If\n enabled, the malloc_stats_print()\n function is called at program exit via an\n atexit(3) function. If\n --enable-stats is specified during configuration, this\n has the potential to cause deadlock for a multi-threaded process that\n exits while one or more threads are executing in the memory allocation\n functions. Therefore, this option should only be used with care; it is\n primarily intended as a performance tuning aid during application\n development. This option is disabled by default.

\n \n \"opt.junk\"\n \n (bool)\n r-\n [--enable-fill]\n

Junk filling enabled/disabled. If enabled, each byte\n of uninitialized allocated memory will be initialized to\n 0xa5. All deallocated memory will be initialized to\n 0x5a. This is intended for debugging and will\n impact performance negatively. This option is disabled by default\n unless --enable-debug is specified during\n configuration, in which case it is enabled by default unless running\n inside Valgrind.

\n \n \"opt.quarantine\"\n \n (size_t)\n r-\n [--enable-fill]\n

Per thread quarantine size in bytes. If non-zero, each\n thread maintains a FIFO object quarantine that stores up to the\n specified number of bytes of memory. The quarantined memory is not\n freed until it is released from quarantine, though it is immediately\n junk-filled if the \n \"opt.junk\"\n option is\n enabled. This feature is of particular use in combination with Valgrind, which can detect attempts\n to access quarantined objects. This is intended for debugging and will\n impact performance negatively. The default quarantine size is 0 unless\n running inside Valgrind, in which case the default is 16\n MiB.

\n \n \"opt.redzone\"\n \n (bool)\n r-\n [--enable-fill]\n

Redzones enabled/disabled. If enabled, small\n allocations have redzones before and after them. Furthermore, if the\n \n \"opt.junk\"\n option is\n enabled, the redzones are checked for corruption during deallocation.\n However, the primary intended purpose of this feature is to be used in\n combination with Valgrind,\n which needs redzones in order to do effective buffer overflow/underflow\n detection. This option is intended for debugging and will impact\n performance negatively. This option is disabled by\n default unless running inside Valgrind.

\n \n \"opt.zero\"\n \n (bool)\n r-\n [--enable-fill]\n

Zero filling enabled/disabled. If enabled, each byte\n of uninitialized allocated memory will be initialized to 0. Note that\n this initialization only happens once for each byte, so\n realloc() and\n rallocm() calls do not zero memory that\n was previously allocated. This is intended for debugging and will\n impact performance negatively. This option is disabled by default.\n

\n \n \"opt.utrace\"\n \n (bool)\n r-\n [--enable-utrace]\n

Allocation tracing based on\n utrace(2) enabled/disabled. This option\n is disabled by default.

\n \n \"opt.valgrind\"\n \n (bool)\n r-\n [--enable-valgrind]\n

Valgrind\n support enabled/disabled. This option is vestigal because jemalloc\n auto-detects whether it is running inside Valgrind. This option is\n disabled by default, unless running inside Valgrind.

\n \n \"opt.xmalloc\"\n \n (bool)\n r-\n [--enable-xmalloc]\n

Abort-on-out-of-memory enabled/disabled. If enabled,\n rather than returning failure for any allocation function, display a\n diagnostic message on STDERR_FILENO and cause the\n program to drop core (using\n abort(3)). If an application is\n designed to depend on this behavior, set the option at compile time by\n including the following in the source code:\n 

\nmalloc_conf = \"xmalloc:true\";

\n This option is disabled by default.

\n \n \"opt.tcache\"\n \n (bool)\n r-\n [--enable-tcache]\n

Thread-specific caching enabled/disabled. When there\n are multiple threads, each thread uses a thread-specific cache for\n objects up to a certain size. Thread-specific caching allows many\n allocations to be satisfied without performing any thread\n synchronization, at the cost of increased memory use. See the\n \n \"opt.lg_tcache_max\"\n \n option for related tuning information. This option is enabled by\n default unless running inside Valgrind.

\n \n \"opt.lg_tcache_max\"\n \n (size_t)\n r-\n [--enable-tcache]\n

Maximum size class (log base 2) to cache in the\n thread-specific cache. At a minimum, all small size classes are\n cached, and at a maximum all large size classes are cached. The\n default maximum is 32 KiB (2^15).

\n \n \"opt.prof\"\n \n (bool)\n r-\n [--enable-prof]\n

Memory profiling enabled/disabled. If enabled, profile\n memory allocation activity. See the \n \"opt.prof_active\"\n \n option for on-the-fly activation/deactivation. See the \n \"opt.lg_prof_sample\"\n \n option for probabilistic sampling control. See the \n \"opt.prof_accum\"\n \n option for control of cumulative sample reporting. See the \n \"opt.lg_prof_interval\"\n \n option for information on interval-triggered profile dumping, the \n \"opt.prof_gdump\"\n \n option for information on high-water-triggered profile dumping, and the\n \n \"opt.prof_final\"\n \n option for final profile dumping. Profile output is compatible with\n the included pprof Perl script, which originates\n from the gperftools\n package.

\n \n \"opt.prof_prefix\"\n \n (const char *)\n r-\n [--enable-prof]\n

Filename prefix for profile dumps. If the prefix is\n set to the empty string, no automatic dumps will occur; this is\n primarily useful for disabling the automatic final heap dump (which\n also disables leak reporting, if enabled). The default prefix is\n jeprof.

\n \n \"opt.prof_active\"\n \n (bool)\n r-\n [--enable-prof]\n

Profiling activated/deactivated. This is a secondary\n control mechanism that makes it possible to start the application with\n profiling enabled (see the \n \"opt.prof\"\n option) but\n inactive, then toggle profiling at any time during program execution\n with the \n \"prof.active\"\n mallctl.\n This option is enabled by default.

\n \n \"opt.lg_prof_sample\"\n \n (ssize_t)\n r-\n [--enable-prof]\n

Average interval (log base 2) between allocation\n samples, as measured in bytes of allocation activity. Increasing the\n sampling interval decreases profile fidelity, but also decreases the\n computational overhead. The default sample interval is 512 KiB (2^19\n B).

\n \n \"opt.prof_accum\"\n \n (bool)\n r-\n [--enable-prof]\n

Reporting of cumulative object/byte counts in profile\n dumps enabled/disabled. If this option is enabled, every unique\n backtrace must be stored for the duration of execution. Depending on\n the application, this can impose a large memory overhead, and the\n cumulative counts are not always of interest. This option is disabled\n by default.

\n \n \"opt.lg_prof_interval\"\n \n (ssize_t)\n r-\n [--enable-prof]\n

Average interval (log base 2) between memory profile\n dumps, as measured in bytes of allocation activity. The actual\n interval between dumps may be sporadic because decentralized allocation\n counters are used to avoid synchronization bottlenecks. Profiles are\n dumped to files named according to the pattern\n <prefix>.<pid>.<seq>.i<iseq>.heap,\n where <prefix> is controlled by the\n \n \"opt.prof_prefix\"\n \n option. By default, interval-triggered profile dumping is disabled\n (encoded as -1).\n

\n \n \"opt.prof_gdump\"\n \n (bool)\n r-\n [--enable-prof]\n

Trigger a memory profile dump every time the total\n virtual memory exceeds the previous maximum. Profiles are dumped to\n files named according to the pattern\n <prefix>.<pid>.<seq>.u<useq>.heap,\n where <prefix> is controlled by the \n \"opt.prof_prefix\"\n \n option. This option is disabled by default.

\n \n \"opt.prof_final\"\n \n (bool)\n r-\n [--enable-prof]\n

Use an\n atexit(3) function to dump final memory\n usage to a file named according to the pattern\n <prefix>.<pid>.<seq>.f.heap,\n where <prefix> is controlled by the \n \"opt.prof_prefix\"\n \n option. This option is enabled by default.

\n \n \"opt.prof_leak\"\n \n (bool)\n r-\n [--enable-prof]\n

Leak reporting enabled/disabled. If enabled, use an\n atexit(3) function to report memory leaks\n detected by allocation sampling. See the\n \n \"opt.prof\"\n option for\n information on analyzing heap profile output. This option is disabled\n by default.

\n \n \"thread.arena\"\n \n (unsigned)\n rw\n

Get or set the arena associated with the calling\n thread. If the specified arena was not initialized beforehand (see the\n \n \"arenas.initialized\"\n \n mallctl), it will be automatically initialized as a side effect of\n calling this interface.

\n \n \"thread.allocated\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Get the total number of bytes ever allocated by the\n calling thread. This counter has the potential to wrap around; it is\n up to the application to appropriately interpret the counter in such\n cases.

\n \n \"thread.allocatedp\"\n \n (uint64_t *)\n r-\n [--enable-stats]\n

Get a pointer to the the value that is returned by the\n \n \"thread.allocated\"\n \n mallctl. This is useful for avoiding the overhead of repeated\n mallctl*() calls.

\n \n \"thread.deallocated\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Get the total number of bytes ever deallocated by the\n calling thread. This counter has the potential to wrap around; it is\n up to the application to appropriately interpret the counter in such\n cases.

\n \n \"thread.deallocatedp\"\n \n (uint64_t *)\n r-\n [--enable-stats]\n

Get a pointer to the the value that is returned by the\n \n \"thread.deallocated\"\n \n mallctl. This is useful for avoiding the overhead of repeated\n mallctl*() calls.

\n \n \"thread.tcache.enabled\"\n \n (bool)\n rw\n [--enable-tcache]\n

Enable/disable calling thread's tcache. The tcache is\n implicitly flushed as a side effect of becoming\n disabled (see \n \"thread.tcache.flush\"\n ).\n

\n \n \"thread.tcache.flush\"\n \n (void)\n --\n [--enable-tcache]\n

Flush calling thread's tcache. This interface releases\n all cached objects and internal data structures associated with the\n calling thread's thread-specific cache. Ordinarily, this interface\n need not be called, since automatic periodic incremental garbage\n collection occurs, and the thread cache is automatically discarded when\n a thread exits. However, garbage collection is triggered by allocation\n activity, so it is possible for a thread that stops\n allocating/deallocating to retain its cache indefinitely, in which case\n the developer may find manual flushing useful.

\n \n \"arena.<i>.purge\"\n \n (unsigned)\n --\n

Purge unused dirty pages for arena <i>, or for\n all arenas if <i> equals \n \"arenas.narenas\"\n .\n

\n \n \"arena.<i>.dss\"\n \n (const char *)\n rw\n

Set the precedence of dss allocation as related to mmap\n allocation for arena <i>, or for all arenas if <i> equals\n \n \"arenas.narenas\"\n . See\n \n \"opt.dss\"\n for supported\n settings.\n

\n \n \"arenas.narenas\"\n \n (unsigned)\n r-\n

Current limit on number of arenas.

\n \n \"arenas.initialized\"\n \n (bool *)\n r-\n

An array of \n \"arenas.narenas\"\n \n booleans. Each boolean indicates whether the corresponding arena is\n initialized.

\n \n \"arenas.quantum\"\n \n (size_t)\n r-\n

Quantum size.

\n \n \"arenas.page\"\n \n (size_t)\n r-\n

Page size.

\n \n \"arenas.tcache_max\"\n \n (size_t)\n r-\n [--enable-tcache]\n

Maximum thread-cached size class.

\n \n \"arenas.nbins\"\n \n (unsigned)\n r-\n

Number of bin size classes.

\n \n \"arenas.nhbins\"\n \n (unsigned)\n r-\n [--enable-tcache]\n

Total number of thread cache bin size\n classes.

\n \n \"arenas.bin.<i>.size\"\n \n (size_t)\n r-\n

Maximum size supported by size class.

\n \n \"arenas.bin.<i>.nregs\"\n \n (uint32_t)\n r-\n

Number of regions per page run.

\n \n \"arenas.bin.<i>.run_size\"\n \n (size_t)\n r-\n

Number of bytes per page run.

\n \n \"arenas.nlruns\"\n \n (size_t)\n r-\n

Total number of large size classes.

\n \n \"arenas.lrun.<i>.size\"\n \n (size_t)\n r-\n

Maximum size supported by this large size\n class.

\n \n \"arenas.purge\"\n \n (unsigned)\n -w\n

Purge unused dirty pages for the specified arena, or\n for all arenas if none is specified.

\n \n \"arenas.extend\"\n \n (unsigned)\n r-\n

Extend the array of arenas by appending a new arena,\n and returning the new arena index.

\n \n \"prof.active\"\n \n (bool)\n rw\n [--enable-prof]\n

Control whether sampling is currently active. See the\n \n \"opt.prof_active\"\n \n option for additional information.\n

\n \n \"prof.dump\"\n \n (const char *)\n -w\n [--enable-prof]\n

Dump a memory profile to the specified file, or if NULL\n is specified, to a file according to the pattern\n <prefix>.<pid>.<seq>.m<mseq>.heap,\n where <prefix> is controlled by the\n \n \"opt.prof_prefix\"\n \n option.

\n \n \"prof.interval\"\n \n (uint64_t)\n r-\n [--enable-prof]\n

Average number of bytes allocated between\n inverval-based profile dumps. See the\n \n \"opt.lg_prof_interval\"\n \n option for additional information.

\n \n \"stats.cactive\"\n \n (size_t *)\n r-\n [--enable-stats]\n

Pointer to a counter that contains an approximate count\n of the current number of bytes in active pages. The estimate may be\n high, but never low, because each arena rounds up to the nearest\n multiple of the chunk size when computing its contribution to the\n counter. Note that the \n \"epoch\"\n mallctl has no bearing\n on this counter. Furthermore, counter consistency is maintained via\n atomic operations, so it is necessary to use an atomic operation in\n order to guarantee a consistent read when dereferencing the pointer.\n

\n \n \"stats.allocated\"\n \n (size_t)\n r-\n [--enable-stats]\n

Total number of bytes allocated by the\n application.

\n \n \"stats.active\"\n \n (size_t)\n r-\n [--enable-stats]\n

Total number of bytes in active pages allocated by the\n application. This is a multiple of the page size, and greater than or\n equal to \n \"stats.allocated\"\n .\n This does not include \n \n \"stats.arenas.<i>.pdirty\"\n and pages\n entirely devoted to allocator metadata.

\n \n \"stats.mapped\"\n \n (size_t)\n r-\n [--enable-stats]\n

Total number of bytes in chunks mapped on behalf of the\n application. This is a multiple of the chunk size, and is at least as\n large as \n \"stats.active\"\n . This\n does not include inactive chunks.

\n \n \"stats.chunks.current\"\n \n (size_t)\n r-\n [--enable-stats]\n

Total number of chunks actively mapped on behalf of the\n application. This does not include inactive chunks.\n

\n \n \"stats.chunks.total\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of chunks allocated.

\n \n \"stats.chunks.high\"\n \n (size_t)\n r-\n [--enable-stats]\n

Maximum number of active chunks at any time thus far.\n

\n \n \"stats.huge.allocated\"\n \n (size_t)\n r-\n [--enable-stats]\n

Number of bytes currently allocated by huge objects.\n

\n \n \"stats.huge.nmalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of huge allocation requests.\n

\n \n \"stats.huge.ndalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of huge deallocation requests.\n

\n \n \"stats.arenas.<i>.dss\"\n \n (const char *)\n r-\n

dss (sbrk(2)) allocation precedence as\n related to mmap(2) allocation. See \n \"opt.dss\"\n for details.\n

\n \n \"stats.arenas.<i>.nthreads\"\n \n (unsigned)\n r-\n

Number of threads currently assigned to\n arena.

\n \n \"stats.arenas.<i>.pactive\"\n \n (size_t)\n r-\n

Number of pages in active runs.

\n \n \"stats.arenas.<i>.pdirty\"\n \n (size_t)\n r-\n

Number of pages within unused runs that are potentially\n dirty, and for which madvise(..., \n MADV_DONTNEED) or\n similar has not been called.

\n \n \"stats.arenas.<i>.mapped\"\n \n (size_t)\n r-\n [--enable-stats]\n

Number of mapped bytes.

\n \n \"stats.arenas.<i>.npurge\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Number of dirty page purge sweeps performed.\n

\n \n \"stats.arenas.<i>.nmadvise\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Number of madvise(..., \n MADV_DONTNEED) or\n similar calls made to purge dirty pages.

\n \n \"stats.arenas.<i>.npurged\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Number of pages purged.

\n \n \"stats.arenas.<i>.small.allocated\"\n \n (size_t)\n r-\n [--enable-stats]\n

Number of bytes currently allocated by small objects.\n

\n \n \"stats.arenas.<i>.small.nmalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocation requests served by\n small bins.

\n \n \"stats.arenas.<i>.small.ndalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of small objects returned to bins.\n

\n \n \"stats.arenas.<i>.small.nrequests\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of small allocation requests.\n

\n \n \"stats.arenas.<i>.large.allocated\"\n \n (size_t)\n r-\n [--enable-stats]\n

Number of bytes currently allocated by large objects.\n

\n \n \"stats.arenas.<i>.large.nmalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of large allocation requests served\n directly by the arena.

\n \n \"stats.arenas.<i>.large.ndalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of large deallocation requests served\n directly by the arena.

\n \n \"stats.arenas.<i>.large.nrequests\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of large allocation requests.\n

\n \n \"stats.arenas.<i>.bins.<j>.allocated\"\n \n (size_t)\n r-\n [--enable-stats]\n

Current number of bytes allocated by\n bin.

\n \n \"stats.arenas.<i>.bins.<j>.nmalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocations served by bin.\n

\n \n \"stats.arenas.<i>.bins.<j>.ndalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocations returned to bin.\n

\n \n \"stats.arenas.<i>.bins.<j>.nrequests\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocation\n requests.

\n \n \"stats.arenas.<i>.bins.<j>.nfills\"\n \n (uint64_t)\n r-\n [--enable-stats --enable-tcache]\n

Cumulative number of tcache fills.

\n \n \"stats.arenas.<i>.bins.<j>.nflushes\"\n \n (uint64_t)\n r-\n [--enable-stats --enable-tcache]\n

Cumulative number of tcache flushes.

\n \n \"stats.arenas.<i>.bins.<j>.nruns\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of runs created.

\n \n \"stats.arenas.<i>.bins.<j>.nreruns\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of times the current run from which\n to allocate changed.

\n \n \"stats.arenas.<i>.bins.<j>.curruns\"\n \n (size_t)\n r-\n [--enable-stats]\n

Current number of runs.

\n \n \"stats.arenas.<i>.lruns.<j>.nmalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocation requests for this size\n class served directly by the arena.

\n \n \"stats.arenas.<i>.lruns.<j>.ndalloc\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of deallocation requests for this\n size class served directly by the arena.

\n \n \"stats.arenas.<i>.lruns.<j>.nrequests\"\n \n (uint64_t)\n r-\n [--enable-stats]\n

Cumulative number of allocation requests for this size\n class.

\n \n \"stats.arenas.<i>.lruns.<j>.curruns\"\n \n (size_t)\n r-\n [--enable-stats]\n

Current number of runs for this size class.\n

DEBUGGING MALLOC PROBLEMS

When debugging, it is a good idea to configure/build jemalloc with\n the --enable-debug and --enable-fill\n options, and recompile the program with suitable options and symbols for\n debugger support. When so configured, jemalloc incorporates a wide variety\n of run-time assertions that catch application errors such as double-free,\n write-after-free, etc.

Programs often accidentally depend on “uninitialized”\n memory actually being filled with zero bytes. Junk filling\n (see the \n \"opt.junk\"\n \n option) tends to expose such bugs in the form of obviously incorrect\n results and/or coredumps. Conversely, zero\n filling (see the \n \"opt.zero\"\n option) eliminates\n the symptoms of such bugs. Between these two options, it is usually\n possible to quickly detect, diagnose, and eliminate such bugs.

This implementation does not provide much detail about the problems\n it detects, because the performance impact for storing such information\n would be prohibitive. However, jemalloc does integrate with the most\n excellent Valgrind tool if the\n --enable-valgrind configuration option is enabled.

DIAGNOSTIC MESSAGES

If any of the memory allocation/deallocation functions detect an\n error or warning condition, a message will be printed to file descriptor\n STDERR_FILENO. Errors will result in the process\n dumping core. If the \n \"opt.abort\"\n option is set, most\n warnings are treated as errors.

The malloc_message variable allows the programmer\n to override the function which emits the text strings forming the errors\n and warnings if for some reason the STDERR_FILENO file\n descriptor is not suitable for this.\n malloc_message() takes the\n cbopaque pointer argument that is\n NULL unless overridden by the arguments in a call to\n malloc_stats_print(), followed by a string\n pointer. Please note that doing anything which tries to allocate memory in\n this function is likely to result in a crash or deadlock.

All messages are prefixed by\n “<jemalloc>: ”.

RETURN VALUES

Standard API

The malloc() and\n calloc() functions return a pointer to the\n allocated memory if successful; otherwise a NULL\n pointer is returned and errno is set to\n ENOMEM.

The posix_memalign() function\n returns the value 0 if successful; otherwise it returns an error value.\n The posix_memalign() function will fail\n if:\n

EINVAL

The alignment parameter is\n not a power of 2 at least as large as\n sizeof(void *).\n

ENOMEM

Memory allocation error.

\n

The aligned_alloc() function returns\n a pointer to the allocated memory if successful; otherwise a\n NULL pointer is returned and\n errno is set. The\n aligned_alloc() function will fail if:\n

EINVAL

The alignment parameter is\n not a power of 2.\n

ENOMEM

Memory allocation error.

\n

The realloc() function returns a\n pointer, possibly identical to ptr, to the\n allocated memory if successful; otherwise a NULL\n pointer is returned, and errno is set to\n ENOMEM if the error was the result of an\n allocation failure. The realloc()\n function always leaves the original buffer intact when an error occurs.\n

The free() function returns no\n value.

Non-standard API

The malloc_usable_size() function\n returns the usable size of the allocation pointed to by\n ptr.

The mallctl(),\n mallctlnametomib(), and\n mallctlbymib() functions return 0 on\n success; otherwise they return an error value. The functions will fail\n if:\n

EINVAL

newp is not\n NULL, and newlen is too\n large or too small. Alternatively, *oldlenp\n is too large or too small; in this case as much data as possible\n are read despite the error.

ENOMEM

*oldlenp is too short to\n hold the requested value.

ENOENT

name or\n mib specifies an unknown/invalid\n value.

EPERM

Attempt to read or write void value, or attempt to\n write read-only value.

EAGAIN

A memory allocation failure\n occurred.

EFAULT

An interface with side effects failed in some way\n not directly related to mallctl*()\n read/write processing.

\n

Experimental API

The allocm(),\n rallocm(),\n sallocm(),\n dallocm(), and\n nallocm() functions return\n ALLOCM_SUCCESS on success; otherwise they return an\n error value. The allocm(),\n rallocm(), and\n nallocm() functions will fail if:\n

ALLOCM_ERR_OOM

Out of memory. Insufficient contiguous memory was\n available to service the allocation request. The\n allocm() function additionally sets\n *ptr to NULL, whereas\n the rallocm() function leaves\n *ptr unmodified.

\n The rallocm() function will also\n fail if:\n

ALLOCM_ERR_NOT_MOVED

ALLOCM_NO_MOVE was specified,\n but the reallocation request could not be serviced without moving\n the object.

\n

ENVIRONMENT

The following environment variable affects the execution of the\n allocation functions:\n

MALLOC_CONF

If the environment variable\n MALLOC_CONF is set, the characters it contains\n will be interpreted as options.

\n

EXAMPLES

To dump core whenever a problem occurs:\n 

ln -s 'abort:true' /etc/malloc.conf

\n

To specify in the source a chunk size that is 16 MiB:\n 

\nmalloc_conf = \"lg_chunk:24\";

SEE ALSO

madvise(2),\n mmap(2),\n sbrk(2),\n utrace(2),\n alloca(3),\n atexit(3),\n getpagesize(3)

STANDARDS

The malloc(),\n calloc(),\n realloc(), and\n free() functions conform to ISO/IEC\n 9899:1990 (“ISO C90”).

The posix_memalign() function conforms\n to IEEE Std 1003.1-2001 (“POSIX.1”).

\n" }, { "path": "deps/jemalloc/doc/jemalloc.xml.in", "content": "\n\n\n\n\n \n User Manual\n jemalloc\n @jemalloc_version@\n \n \n Jason\n Evans\n Author\n \n \n \n \n JEMALLOC\n 3\n \n \n jemalloc\n jemalloc\n \n general purpose memory allocation functions\n \n \n LIBRARY\n This manual describes jemalloc @jemalloc_version@. More information\n can be found at the jemalloc website.\n \n \n SYNOPSIS\n \n #include <stdlib.h>\n#include <jemalloc/jemalloc.h>\n \n Standard API\n \n void *malloc\n size_t size\n \n \n void *calloc\n size_t number\n size_t size\n \n \n int posix_memalign\n void **ptr\n size_t alignment\n size_t size\n \n \n void *aligned_alloc\n size_t alignment\n size_t size\n \n \n void *realloc\n void *ptr\n size_t size\n \n \n void free\n void *ptr\n \n \n \n Non-standard API\n \n size_t malloc_usable_size\n const void *ptr\n \n \n void malloc_stats_print\n void (*write_cb)\n void *, const char *\n \n void *cbopaque\n const char *opts\n \n \n int mallctl\n const char *name\n void *oldp\n size_t *oldlenp\n void *newp\n size_t newlen\n \n \n int mallctlnametomib\n const char *name\n size_t *mibp\n size_t *miblenp\n \n \n int mallctlbymib\n const size_t *mib\n size_t miblen\n void *oldp\n size_t *oldlenp\n void *newp\n size_t newlen\n \n \n void (*malloc_message)\n void *cbopaque\n const char *s\n \n const char *malloc_conf;\n \n \n Experimental API\n \n int allocm\n void **ptr\n size_t *rsize\n size_t size\n int flags\n \n \n int rallocm\n void **ptr\n size_t *rsize\n size_t size\n size_t extra\n int flags\n \n \n int sallocm\n const void *ptr\n size_t *rsize\n int flags\n \n \n int dallocm\n void *ptr\n int flags\n \n \n int nallocm\n size_t *rsize\n size_t size\n int flags\n \n \n \n \n \n DESCRIPTION\n \n Standard API\n\n The malloc function allocates\n size bytes of uninitialized memory. The allocated\n space is suitably aligned (after possible pointer coercion) for storage\n of any type of object.\n\n The calloc function allocates\n space for number objects, each\n size bytes in length. The result is identical to\n calling malloc with an argument of\n number * size, with the\n exception that the allocated memory is explicitly initialized to zero\n bytes.\n\n The posix_memalign function\n allocates size bytes of memory such that the\n allocation's base address is an even multiple of\n alignment, and returns the allocation in the value\n pointed to by ptr. The requested\n alignment must be a power of 2 at least as large\n as sizeof(void *).\n\n The aligned_alloc function\n allocates size bytes of memory such that the\n allocation's base address is an even multiple of\n alignment. The requested\n alignment must be a power of 2. Behavior is\n undefined if size is not an integral multiple of\n alignment.\n\n The realloc function changes the\n size of the previously allocated memory referenced by\n ptr to size bytes. The\n contents of the memory are unchanged up to the lesser of the new and old\n sizes. If the new size is larger, the contents of the newly allocated\n portion of the memory are undefined. Upon success, the memory referenced\n by ptr is freed and a pointer to the newly\n allocated memory is returned. Note that\n realloc may move the memory allocation,\n resulting in a different return value than ptr.\n If ptr is NULL, the\n realloc function behaves identically to\n malloc for the specified size.\n\n The free function causes the\n allocated memory referenced by ptr to be made\n available for future allocations. If ptr is\n NULL, no action occurs.\n \n \n Non-standard API\n\n The malloc_usable_size function\n returns the usable size of the allocation pointed to by\n ptr. The return value may be larger than the size\n that was requested during allocation. The\n malloc_usable_size function is not a\n mechanism for in-place realloc; rather\n it is provided solely as a tool for introspection purposes. Any\n discrepancy between the requested allocation size and the size reported\n by malloc_usable_size should not be\n depended on, since such behavior is entirely implementation-dependent.\n \n\n The malloc_stats_print function\n writes human-readable summary statistics via the\n write_cb callback function pointer and\n cbopaque data passed to\n write_cb, or\n malloc_message if\n write_cb is NULL. This\n function can be called repeatedly. General information that never\n changes during execution can be omitted by specifying \"g\" as a character\n within the opts string. Note that\n malloc_message uses the\n mallctl* functions internally, so\n inconsistent statistics can be reported if multiple threads use these\n functions simultaneously. If is\n specified during configuration, “m” and “a” can\n be specified to omit merged arena and per arena statistics, respectively;\n “b” and “l” can be specified to omit per size\n class statistics for bins and large objects, respectively. Unrecognized\n characters are silently ignored. Note that thread caching may prevent\n some statistics from being completely up to date, since extra locking\n would be required to merge counters that track thread cache operations.\n \n\n The mallctl function provides a\n general interface for introspecting the memory allocator, as well as\n setting modifiable parameters and triggering actions. The\n period-separated name argument specifies a\n location in a tree-structured namespace; see the section for\n documentation on the tree contents. To read a value, pass a pointer via\n oldp to adequate space to contain the value, and a\n pointer to its length via oldlenp; otherwise pass\n NULL and NULL. Similarly, to\n write a value, pass a pointer to the value via\n newp, and its length via\n newlen; otherwise pass NULL\n and 0.\n\n The mallctlnametomib function\n provides a way to avoid repeated name lookups for applications that\n repeatedly query the same portion of the namespace, by translating a name\n to a “Management Information Base” (MIB) that can be passed\n repeatedly to mallctlbymib. Upon\n successful return from mallctlnametomib,\n mibp contains an array of\n *miblenp integers, where\n *miblenp is the lesser of the number of components\n in name and the input value of\n *miblenp. Thus it is possible to pass a\n *miblenp that is smaller than the number of\n period-separated name components, which results in a partial MIB that can\n be used as the basis for constructing a complete MIB. For name\n components that are integers (e.g. the 2 in\n arenas.bin.2.size),\n the corresponding MIB component will always be that integer. Therefore,\n it is legitimate to construct code like the following: \n \n \n Experimental API\n The experimental API is subject to change or removal without regard\n for backward compatibility. If \n is specified during configuration, the experimental API is\n omitted.\n\n The allocm,\n rallocm,\n sallocm,\n dallocm, and\n nallocm functions all have a\n flags argument that can be used to specify\n options. The functions only check the options that are contextually\n relevant. Use bitwise or (|) operations to\n specify one or more of the following:\n \n \n ALLOCM_LG_ALIGN(la)\n \n\n Align the memory allocation to start at an address\n that is a multiple of (1 <<\n la). This macro does not validate\n that la is within the valid\n range.\n \n \n ALLOCM_ALIGN(a)\n \n\n Align the memory allocation to start at an address\n that is a multiple of a, where\n a is a power of two. This macro does not\n validate that a is a power of 2.\n \n \n \n ALLOCM_ZERO\n\n Initialize newly allocated memory to contain zero\n bytes. In the growing reallocation case, the real size prior to\n reallocation defines the boundary between untouched bytes and those\n that are initialized to contain zero bytes. If this option is\n absent, newly allocated memory is uninitialized.\n \n \n ALLOCM_NO_MOVE\n\n For reallocation, fail rather than moving the\n object. This constraint can apply to both growth and\n shrinkage.\n \n \n ALLOCM_ARENA(a)\n \n\n Use the arena specified by the index\n a. This macro does not validate that\n a specifies an arena in the valid\n range.\n \n \n \n\n The allocm function allocates at\n least size bytes of memory, sets\n *ptr to the base address of the allocation, and\n sets *rsize to the real size of the allocation if\n rsize is not NULL. Behavior\n is undefined if size is\n 0.\n\n The rallocm function resizes the\n allocation at *ptr to be at least\n size bytes, sets *ptr to\n the base address of the allocation if it moved, and sets\n *rsize to the real size of the allocation if\n rsize is not NULL. If\n extra is non-zero, an attempt is made to resize\n the allocation to be at least size +\n extra) bytes, though inability to allocate\n the extra byte(s) will not by itself result in failure. Behavior is\n undefined if size is 0, or if\n (size +\n extra >\n SIZE_T_MAX).\n\n The sallocm function sets\n *rsize to the real size of the allocation.\n\n The dallocm function causes the\n memory referenced by ptr to be made available for\n future allocations.\n\n The nallocm function allocates no\n memory, but it performs the same size computation as the\n allocm function, and if\n rsize is not NULL it sets\n *rsize to the real size of the allocation that\n would result from the equivalent allocm\n function call. Behavior is undefined if\n size is 0.\n \n \n \n TUNING\n Once, when the first call is made to one of the memory allocation\n routines, the allocator initializes its internals based in part on various\n options that can be specified at compile- or run-time.\n\n The string pointed to by the global variable\n malloc_conf, the “name” of the file\n referenced by the symbolic link named /etc/malloc.conf, and the value of the\n environment variable MALLOC_CONF, will be interpreted, in\n that order, from left to right as options.\n\n An options string is a comma-separated list of option:value pairs.\n There is one key corresponding to each opt.* mallctl (see the section for options\n documentation). For example, abort:true,narenas:1 sets\n the opt.abort and opt.narenas options. Some\n options have boolean values (true/false), others have integer values (base\n 8, 10, or 16, depending on prefix), and yet others have raw string\n values.\n \n \n IMPLEMENTATION NOTES\n Traditionally, allocators have used\n sbrk\n 2 to obtain memory, which is\n suboptimal for several reasons, including race conditions, increased\n fragmentation, and artificial limitations on maximum usable memory. If\n is specified during configuration, this\n allocator uses both mmap\n 2 and\n sbrk\n 2, in that order of preference;\n otherwise only mmap\n 2 is used.\n\n This allocator uses multiple arenas in order to reduce lock\n contention for threaded programs on multi-processor systems. This works\n well with regard to threading scalability, but incurs some costs. There is\n a small fixed per-arena overhead, and additionally, arenas manage memory\n completely independently of each other, which means a small fixed increase\n in overall memory fragmentation. These overheads are not generally an\n issue, given the number of arenas normally used. Note that using\n substantially more arenas than the default is not likely to improve\n performance, mainly due to reduced cache performance. However, it may make\n sense to reduce the number of arenas if an application does not make much\n use of the allocation functions.\n\n In addition to multiple arenas, unless\n is specified during configuration, this\n allocator supports thread-specific caching for small and large objects, in\n order to make it possible to completely avoid synchronization for most\n allocation requests. Such caching allows very fast allocation in the\n common case, but it increases memory usage and fragmentation, since a\n bounded number of objects can remain allocated in each thread cache.\n\n Memory is conceptually broken into equal-sized chunks, where the\n chunk size is a power of two that is greater than the page size. Chunks\n are always aligned to multiples of the chunk size. This alignment makes it\n possible to find metadata for user objects very quickly.\n\n User objects are broken into three categories according to size:\n small, large, and huge. Small objects are smaller than one page. Large\n objects are smaller than the chunk size. Huge objects are a multiple of\n the chunk size. Small and large objects are managed by arenas; huge\n objects are managed separately in a single data structure that is shared by\n all threads. Huge objects are used by applications infrequently enough\n that this single data structure is not a scalability issue.\n\n Each chunk that is managed by an arena tracks its contents as runs of\n contiguous pages (unused, backing a set of small objects, or backing one\n large object). The combination of chunk alignment and chunk page maps\n makes it possible to determine all metadata regarding small and large\n allocations in constant time.\n\n Small objects are managed in groups by page runs. Each run maintains\n a frontier and free list to track which regions are in use. Allocation\n requests that are no more than half the quantum (8 or 16, depending on\n architecture) are rounded up to the nearest power of two that is at least\n sizeof(double). All other small\n object size classes are multiples of the quantum, spaced such that internal\n fragmentation is limited to approximately 25% for all but the smallest size\n classes. Allocation requests that are larger than the maximum small size\n class, but small enough to fit in an arena-managed chunk (see the opt.lg_chunk option), are\n rounded up to the nearest run size. Allocation requests that are too large\n to fit in an arena-managed chunk are rounded up to the nearest multiple of\n the chunk size.\n\n Allocations are packed tightly together, which can be an issue for\n multi-threaded applications. If you need to assure that allocations do not\n suffer from cacheline sharing, round your allocation requests up to the\n nearest multiple of the cacheline size, or specify cacheline alignment when\n allocating.\n\n Assuming 4 MiB chunks, 4 KiB pages, and a 16-byte quantum on a 64-bit\n system, the size classes in each category are as shown in .\n\n \n Size classes\n \n \n \n \n \n \n Category\n Spacing\n Size\n \n \n \n \n Small\n lg\n [8]\n \n \n 16\n [16, 32, 48, ..., 128]\n \n \n 32\n [160, 192, 224, 256]\n \n \n 64\n [320, 384, 448, 512]\n \n \n 128\n [640, 768, 896, 1024]\n \n \n 256\n [1280, 1536, 1792, 2048]\n \n \n 512\n [2560, 3072, 3584]\n \n \n Large\n 4 KiB\n [4 KiB, 8 KiB, 12 KiB, ..., 4072 KiB]\n \n \n Huge\n 4 MiB\n [4 MiB, 8 MiB, 12 MiB, ...]\n \n \n \n
\n \n \n MALLCTL NAMESPACE\n The following names are defined in the namespace accessible via the\n mallctl* functions. Value types are\n specified in parentheses, their readable/writable statuses are encoded as\n rw, r-, -w, or\n --, and required build configuration flags follow, if\n any. A name element encoded as <i> or\n <j> indicates an integer component, where the\n integer varies from 0 to some upper value that must be determined via\n introspection. In the case of stats.arenas.<i>.*,\n <i> equal to arenas.narenas can be\n used to access the summation of statistics from all arenas. Take special\n note of the epoch mallctl,\n which controls refreshing of cached dynamic statistics.\n\n \n \n \n version\n (const char *)\n r-\n \n Return the jemalloc version string.\n \n\n \n \n epoch\n (uint64_t)\n rw\n \n If a value is passed in, refresh the data from which\n the mallctl* functions report values,\n and increment the epoch. Return the current epoch. This is useful for\n detecting whether another thread caused a refresh.\n \n\n \n \n config.debug\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.dss\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.fill\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.lazy_lock\n (bool)\n r-\n \n was specified\n during build configuration.\n \n\n \n \n config.mremap\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.munmap\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.prof\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.prof_libgcc\n (bool)\n r-\n \n was not\n specified during build configuration.\n \n\n \n \n config.prof_libunwind\n (bool)\n r-\n \n was specified\n during build configuration.\n \n\n \n \n config.stats\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.tcache\n (bool)\n r-\n \n was not specified\n during build configuration.\n \n\n \n \n config.tls\n (bool)\n r-\n \n was not specified during\n build configuration.\n \n\n \n \n config.utrace\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.valgrind\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n config.xmalloc\n (bool)\n r-\n \n was specified during\n build configuration.\n \n\n \n \n opt.abort\n (bool)\n r-\n \n Abort-on-warning enabled/disabled. If true, most\n warnings are fatal. The process will call\n abort\n 3 in these cases. This option is\n disabled by default unless is\n specified during configuration, in which case it is enabled by default.\n \n \n\n \n \n opt.lg_chunk\n (size_t)\n r-\n \n Virtual memory chunk size (log base 2). The default\n chunk size is 4 MiB (2^22).\n \n\n \n \n opt.dss\n (const char *)\n r-\n \n dss (sbrk\n 2) allocation precedence as\n related to mmap\n 2 allocation. The following\n settings are supported: “disabled”, “primary”,\n and “secondary” (default).\n \n\n \n \n opt.narenas\n (size_t)\n r-\n \n Maximum number of arenas to use for automatic\n multiplexing of threads and arenas. The default is four times the\n number of CPUs, or one if there is a single CPU.\n \n\n \n \n opt.lg_dirty_mult\n (ssize_t)\n r-\n \n Per-arena minimum ratio (log base 2) of active to dirty\n pages. Some dirty unused pages may be allowed to accumulate, within\n the limit set by the ratio (or one chunk worth of dirty pages,\n whichever is greater), before informing the kernel about some of those\n pages via madvise\n 2 or a similar system call. This\n provides the kernel with sufficient information to recycle dirty pages\n if physical memory becomes scarce and the pages remain unused. The\n default minimum ratio is 8:1 (2^3:1); an option value of -1 will\n disable dirty page purging.\n \n\n \n \n opt.stats_print\n (bool)\n r-\n \n Enable/disable statistics printing at exit. If\n enabled, the malloc_stats_print\n function is called at program exit via an\n atexit\n 3 function. If\n is specified during configuration, this\n has the potential to cause deadlock for a multi-threaded process that\n exits while one or more threads are executing in the memory allocation\n functions. Therefore, this option should only be used with care; it is\n primarily intended as a performance tuning aid during application\n development. This option is disabled by default.\n \n\n \n \n opt.junk\n (bool)\n r-\n []\n \n Junk filling enabled/disabled. If enabled, each byte\n of uninitialized allocated memory will be initialized to\n 0xa5. All deallocated memory will be initialized to\n 0x5a. This is intended for debugging and will\n impact performance negatively. This option is disabled by default\n unless is specified during\n configuration, in which case it is enabled by default unless running\n inside Valgrind.\n \n\n \n \n opt.quarantine\n (size_t)\n r-\n []\n \n Per thread quarantine size in bytes. If non-zero, each\n thread maintains a FIFO object quarantine that stores up to the\n specified number of bytes of memory. The quarantined memory is not\n freed until it is released from quarantine, though it is immediately\n junk-filled if the opt.junk option is\n enabled. This feature is of particular use in combination with Valgrind, which can detect attempts\n to access quarantined objects. This is intended for debugging and will\n impact performance negatively. The default quarantine size is 0 unless\n running inside Valgrind, in which case the default is 16\n MiB.\n \n\n \n \n opt.redzone\n (bool)\n r-\n []\n \n Redzones enabled/disabled. If enabled, small\n allocations have redzones before and after them. Furthermore, if the\n opt.junk option is\n enabled, the redzones are checked for corruption during deallocation.\n However, the primary intended purpose of this feature is to be used in\n combination with Valgrind,\n which needs redzones in order to do effective buffer overflow/underflow\n detection. This option is intended for debugging and will impact\n performance negatively. This option is disabled by\n default unless running inside Valgrind.\n \n\n \n \n opt.zero\n (bool)\n r-\n []\n \n Zero filling enabled/disabled. If enabled, each byte\n of uninitialized allocated memory will be initialized to 0. Note that\n this initialization only happens once for each byte, so\n realloc and\n rallocm calls do not zero memory that\n was previously allocated. This is intended for debugging and will\n impact performance negatively. This option is disabled by default.\n \n \n\n \n \n opt.utrace\n (bool)\n r-\n []\n \n Allocation tracing based on\n utrace\n 2 enabled/disabled. This option\n is disabled by default.\n \n\n \n \n opt.valgrind\n (bool)\n r-\n []\n \n Valgrind\n support enabled/disabled. This option is vestigal because jemalloc\n auto-detects whether it is running inside Valgrind. This option is\n disabled by default, unless running inside Valgrind.\n \n\n \n \n opt.xmalloc\n (bool)\n r-\n []\n \n Abort-on-out-of-memory enabled/disabled. If enabled,\n rather than returning failure for any allocation function, display a\n diagnostic message on STDERR_FILENO and cause the\n program to drop core (using\n abort\n 3). If an application is\n designed to depend on this behavior, set the option at compile time by\n including the following in the source code:\n \n This option is disabled by default.\n \n\n \n \n opt.tcache\n (bool)\n r-\n []\n \n Thread-specific caching enabled/disabled. When there\n are multiple threads, each thread uses a thread-specific cache for\n objects up to a certain size. Thread-specific caching allows many\n allocations to be satisfied without performing any thread\n synchronization, at the cost of increased memory use. See the\n opt.lg_tcache_max\n option for related tuning information. This option is enabled by\n default unless running inside Valgrind.\n \n\n \n \n opt.lg_tcache_max\n (size_t)\n r-\n []\n \n Maximum size class (log base 2) to cache in the\n thread-specific cache. At a minimum, all small size classes are\n cached, and at a maximum all large size classes are cached. The\n default maximum is 32 KiB (2^15).\n \n\n \n \n opt.prof\n (bool)\n r-\n []\n \n Memory profiling enabled/disabled. If enabled, profile\n memory allocation activity. See the opt.prof_active\n option for on-the-fly activation/deactivation. See the opt.lg_prof_sample\n option for probabilistic sampling control. See the opt.prof_accum\n option for control of cumulative sample reporting. See the opt.lg_prof_interval\n option for information on interval-triggered profile dumping, the opt.prof_gdump\n option for information on high-water-triggered profile dumping, and the\n opt.prof_final\n option for final profile dumping. Profile output is compatible with\n the included pprof Perl script, which originates\n from the gperftools\n package.\n \n\n \n \n opt.prof_prefix\n (const char *)\n r-\n []\n \n Filename prefix for profile dumps. If the prefix is\n set to the empty string, no automatic dumps will occur; this is\n primarily useful for disabling the automatic final heap dump (which\n also disables leak reporting, if enabled). The default prefix is\n jeprof.\n \n\n \n \n opt.prof_active\n (bool)\n r-\n []\n \n Profiling activated/deactivated. This is a secondary\n control mechanism that makes it possible to start the application with\n profiling enabled (see the opt.prof option) but\n inactive, then toggle profiling at any time during program execution\n with the prof.active mallctl.\n This option is enabled by default.\n \n\n \n \n opt.lg_prof_sample\n (ssize_t)\n r-\n []\n \n Average interval (log base 2) between allocation\n samples, as measured in bytes of allocation activity. Increasing the\n sampling interval decreases profile fidelity, but also decreases the\n computational overhead. The default sample interval is 512 KiB (2^19\n B).\n \n\n \n \n opt.prof_accum\n (bool)\n r-\n []\n \n Reporting of cumulative object/byte counts in profile\n dumps enabled/disabled. If this option is enabled, every unique\n backtrace must be stored for the duration of execution. Depending on\n the application, this can impose a large memory overhead, and the\n cumulative counts are not always of interest. This option is disabled\n by default.\n \n\n \n \n opt.lg_prof_interval\n (ssize_t)\n r-\n []\n \n Average interval (log base 2) between memory profile\n dumps, as measured in bytes of allocation activity. The actual\n interval between dumps may be sporadic because decentralized allocation\n counters are used to avoid synchronization bottlenecks. Profiles are\n dumped to files named according to the pattern\n <prefix>.<pid>.<seq>.i<iseq>.heap,\n where <prefix> is controlled by the\n opt.prof_prefix\n option. By default, interval-triggered profile dumping is disabled\n (encoded as -1).\n \n \n\n \n \n opt.prof_gdump\n (bool)\n r-\n []\n \n Trigger a memory profile dump every time the total\n virtual memory exceeds the previous maximum. Profiles are dumped to\n files named according to the pattern\n <prefix>.<pid>.<seq>.u<useq>.heap,\n where <prefix> is controlled by the opt.prof_prefix\n option. This option is disabled by default.\n \n\n \n \n opt.prof_final\n (bool)\n r-\n []\n \n Use an\n atexit\n 3 function to dump final memory\n usage to a file named according to the pattern\n <prefix>.<pid>.<seq>.f.heap,\n where <prefix> is controlled by the opt.prof_prefix\n option. This option is enabled by default.\n \n\n \n \n opt.prof_leak\n (bool)\n r-\n []\n \n Leak reporting enabled/disabled. If enabled, use an\n atexit\n 3 function to report memory leaks\n detected by allocation sampling. See the\n opt.prof option for\n information on analyzing heap profile output. This option is disabled\n by default.\n \n\n \n \n thread.arena\n (unsigned)\n rw\n \n Get or set the arena associated with the calling\n thread. If the specified arena was not initialized beforehand (see the\n arenas.initialized\n mallctl), it will be automatically initialized as a side effect of\n calling this interface.\n \n\n \n \n thread.allocated\n (uint64_t)\n r-\n []\n \n Get the total number of bytes ever allocated by the\n calling thread. This counter has the potential to wrap around; it is\n up to the application to appropriately interpret the counter in such\n cases.\n \n\n \n \n thread.allocatedp\n (uint64_t *)\n r-\n []\n \n Get a pointer to the the value that is returned by the\n thread.allocated\n mallctl. This is useful for avoiding the overhead of repeated\n mallctl* calls.\n \n\n \n \n thread.deallocated\n (uint64_t)\n r-\n []\n \n Get the total number of bytes ever deallocated by the\n calling thread. This counter has the potential to wrap around; it is\n up to the application to appropriately interpret the counter in such\n cases.\n \n\n \n \n thread.deallocatedp\n (uint64_t *)\n r-\n []\n \n Get a pointer to the the value that is returned by the\n thread.deallocated\n mallctl. This is useful for avoiding the overhead of repeated\n mallctl* calls.\n \n\n \n \n thread.tcache.enabled\n (bool)\n rw\n []\n \n Enable/disable calling thread's tcache. The tcache is\n implicitly flushed as a side effect of becoming\n disabled (see thread.tcache.flush).\n \n \n\n \n \n thread.tcache.flush\n (void)\n --\n []\n \n Flush calling thread's tcache. This interface releases\n all cached objects and internal data structures associated with the\n calling thread's thread-specific cache. Ordinarily, this interface\n need not be called, since automatic periodic incremental garbage\n collection occurs, and the thread cache is automatically discarded when\n a thread exits. However, garbage collection is triggered by allocation\n activity, so it is possible for a thread that stops\n allocating/deallocating to retain its cache indefinitely, in which case\n the developer may find manual flushing useful.\n \n\n \n \n arena.<i>.purge\n (unsigned)\n --\n \n Purge unused dirty pages for arena <i>, or for\n all arenas if <i> equals arenas.narenas.\n \n \n\n \n \n arena.<i>.dss\n (const char *)\n rw\n \n Set the precedence of dss allocation as related to mmap\n allocation for arena <i>, or for all arenas if <i> equals\n arenas.narenas. See\n opt.dss for supported\n settings.\n \n \n\n \n \n arenas.narenas\n (unsigned)\n r-\n \n Current limit on number of arenas.\n \n\n \n \n arenas.initialized\n (bool *)\n r-\n \n An array of arenas.narenas\n booleans. Each boolean indicates whether the corresponding arena is\n initialized.\n \n\n \n \n arenas.quantum\n (size_t)\n r-\n \n Quantum size.\n \n\n \n \n arenas.page\n (size_t)\n r-\n \n Page size.\n \n\n \n \n arenas.tcache_max\n (size_t)\n r-\n []\n \n Maximum thread-cached size class.\n \n\n \n \n arenas.nbins\n (unsigned)\n r-\n \n Number of bin size classes.\n \n\n \n \n arenas.nhbins\n (unsigned)\n r-\n []\n \n Total number of thread cache bin size\n classes.\n \n\n \n \n arenas.bin.<i>.size\n (size_t)\n r-\n \n Maximum size supported by size class.\n \n\n \n \n arenas.bin.<i>.nregs\n (uint32_t)\n r-\n \n Number of regions per page run.\n \n\n \n \n arenas.bin.<i>.run_size\n (size_t)\n r-\n \n Number of bytes per page run.\n \n\n \n \n arenas.nlruns\n (size_t)\n r-\n \n Total number of large size classes.\n \n\n \n \n arenas.lrun.<i>.size\n (size_t)\n r-\n \n Maximum size supported by this large size\n class.\n \n\n \n \n arenas.purge\n (unsigned)\n -w\n \n Purge unused dirty pages for the specified arena, or\n for all arenas if none is specified.\n \n\n \n \n arenas.extend\n (unsigned)\n r-\n \n Extend the array of arenas by appending a new arena,\n and returning the new arena index.\n \n\n \n \n prof.active\n (bool)\n rw\n []\n \n Control whether sampling is currently active. See the\n opt.prof_active\n option for additional information.\n \n \n\n \n \n prof.dump\n (const char *)\n -w\n []\n \n Dump a memory profile to the specified file, or if NULL\n is specified, to a file according to the pattern\n <prefix>.<pid>.<seq>.m<mseq>.heap,\n where <prefix> is controlled by the\n opt.prof_prefix\n option.\n \n\n \n \n prof.interval\n (uint64_t)\n r-\n []\n \n Average number of bytes allocated between\n inverval-based profile dumps. See the\n opt.lg_prof_interval\n option for additional information.\n \n\n \n \n stats.cactive\n (size_t *)\n r-\n []\n \n Pointer to a counter that contains an approximate count\n of the current number of bytes in active pages. The estimate may be\n high, but never low, because each arena rounds up to the nearest\n multiple of the chunk size when computing its contribution to the\n counter. Note that the epoch mallctl has no bearing\n on this counter. Furthermore, counter consistency is maintained via\n atomic operations, so it is necessary to use an atomic operation in\n order to guarantee a consistent read when dereferencing the pointer.\n \n \n\n \n \n stats.allocated\n (size_t)\n r-\n []\n \n Total number of bytes allocated by the\n application.\n \n\n \n \n stats.active\n (size_t)\n r-\n []\n \n Total number of bytes in active pages allocated by the\n application. This is a multiple of the page size, and greater than or\n equal to stats.allocated.\n This does not include \n stats.arenas.<i>.pdirty and pages\n entirely devoted to allocator metadata.\n \n\n \n \n stats.mapped\n (size_t)\n r-\n []\n \n Total number of bytes in chunks mapped on behalf of the\n application. This is a multiple of the chunk size, and is at least as\n large as stats.active. This\n does not include inactive chunks.\n \n\n \n \n stats.chunks.current\n (size_t)\n r-\n []\n \n Total number of chunks actively mapped on behalf of the\n application. This does not include inactive chunks.\n \n \n\n \n \n stats.chunks.total\n (uint64_t)\n r-\n []\n \n Cumulative number of chunks allocated.\n \n\n \n \n stats.chunks.high\n (size_t)\n r-\n []\n \n Maximum number of active chunks at any time thus far.\n \n \n\n \n \n stats.huge.allocated\n (size_t)\n r-\n []\n \n Number of bytes currently allocated by huge objects.\n \n \n\n \n \n stats.huge.nmalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of huge allocation requests.\n \n \n\n \n \n stats.huge.ndalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of huge deallocation requests.\n \n \n\n \n \n stats.arenas.<i>.dss\n (const char *)\n r-\n \n dss (sbrk\n 2) allocation precedence as\n related to mmap\n 2 allocation. See opt.dss for details.\n \n \n\n \n \n stats.arenas.<i>.nthreads\n (unsigned)\n r-\n \n Number of threads currently assigned to\n arena.\n \n\n \n \n stats.arenas.<i>.pactive\n (size_t)\n r-\n \n Number of pages in active runs.\n \n\n \n \n stats.arenas.<i>.pdirty\n (size_t)\n r-\n \n Number of pages within unused runs that are potentially\n dirty, and for which madvise...\n MADV_DONTNEED or\n similar has not been called.\n \n\n \n \n stats.arenas.<i>.mapped\n (size_t)\n r-\n []\n \n Number of mapped bytes.\n \n\n \n \n stats.arenas.<i>.npurge\n (uint64_t)\n r-\n []\n \n Number of dirty page purge sweeps performed.\n \n \n\n \n \n stats.arenas.<i>.nmadvise\n (uint64_t)\n r-\n []\n \n Number of madvise...\n MADV_DONTNEED or\n similar calls made to purge dirty pages.\n \n\n \n \n stats.arenas.<i>.npurged\n (uint64_t)\n r-\n []\n \n Number of pages purged.\n \n\n \n \n stats.arenas.<i>.small.allocated\n (size_t)\n r-\n []\n \n Number of bytes currently allocated by small objects.\n \n \n\n \n \n stats.arenas.<i>.small.nmalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of allocation requests served by\n small bins.\n \n\n \n \n stats.arenas.<i>.small.ndalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of small objects returned to bins.\n \n \n\n \n \n stats.arenas.<i>.small.nrequests\n (uint64_t)\n r-\n []\n \n Cumulative number of small allocation requests.\n \n \n\n \n \n stats.arenas.<i>.large.allocated\n (size_t)\n r-\n []\n \n Number of bytes currently allocated by large objects.\n \n \n\n \n \n stats.arenas.<i>.large.nmalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of large allocation requests served\n directly by the arena.\n \n\n \n \n stats.arenas.<i>.large.ndalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of large deallocation requests served\n directly by the arena.\n \n\n \n \n stats.arenas.<i>.large.nrequests\n (uint64_t)\n r-\n []\n \n Cumulative number of large allocation requests.\n \n \n\n \n \n stats.arenas.<i>.bins.<j>.allocated\n (size_t)\n r-\n []\n \n Current number of bytes allocated by\n bin.\n \n\n \n \n stats.arenas.<i>.bins.<j>.nmalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of allocations served by bin.\n \n \n\n \n \n stats.arenas.<i>.bins.<j>.ndalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of allocations returned to bin.\n \n \n\n \n \n stats.arenas.<i>.bins.<j>.nrequests\n (uint64_t)\n r-\n []\n \n Cumulative number of allocation\n requests.\n \n\n \n \n stats.arenas.<i>.bins.<j>.nfills\n (uint64_t)\n r-\n [ ]\n \n Cumulative number of tcache fills.\n \n\n \n \n stats.arenas.<i>.bins.<j>.nflushes\n (uint64_t)\n r-\n [ ]\n \n Cumulative number of tcache flushes.\n \n\n \n \n stats.arenas.<i>.bins.<j>.nruns\n (uint64_t)\n r-\n []\n \n Cumulative number of runs created.\n \n\n \n \n stats.arenas.<i>.bins.<j>.nreruns\n (uint64_t)\n r-\n []\n \n Cumulative number of times the current run from which\n to allocate changed.\n \n\n \n \n stats.arenas.<i>.bins.<j>.curruns\n (size_t)\n r-\n []\n \n Current number of runs.\n \n\n \n \n stats.arenas.<i>.lruns.<j>.nmalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of allocation requests for this size\n class served directly by the arena.\n \n\n \n \n stats.arenas.<i>.lruns.<j>.ndalloc\n (uint64_t)\n r-\n []\n \n Cumulative number of deallocation requests for this\n size class served directly by the arena.\n \n\n \n \n stats.arenas.<i>.lruns.<j>.nrequests\n (uint64_t)\n r-\n []\n \n Cumulative number of allocation requests for this size\n class.\n \n\n \n \n stats.arenas.<i>.lruns.<j>.curruns\n (size_t)\n r-\n []\n \n Current number of runs for this size class.\n \n \n \n \n \n DEBUGGING MALLOC PROBLEMS\n When debugging, it is a good idea to configure/build jemalloc with\n the and \n options, and recompile the program with suitable options and symbols for\n debugger support. When so configured, jemalloc incorporates a wide variety\n of run-time assertions that catch application errors such as double-free,\n write-after-free, etc.\n\n Programs often accidentally depend on “uninitialized”\n memory actually being filled with zero bytes. Junk filling\n (see the opt.junk\n option) tends to expose such bugs in the form of obviously incorrect\n results and/or coredumps. Conversely, zero\n filling (see the opt.zero option) eliminates\n the symptoms of such bugs. Between these two options, it is usually\n possible to quickly detect, diagnose, and eliminate such bugs.\n\n This implementation does not provide much detail about the problems\n it detects, because the performance impact for storing such information\n would be prohibitive. However, jemalloc does integrate with the most\n excellent Valgrind tool if the\n configuration option is enabled.\n \n \n DIAGNOSTIC MESSAGES\n If any of the memory allocation/deallocation functions detect an\n error or warning condition, a message will be printed to file descriptor\n STDERR_FILENO. Errors will result in the process\n dumping core. If the opt.abort option is set, most\n warnings are treated as errors.\n\n The malloc_message variable allows the programmer\n to override the function which emits the text strings forming the errors\n and warnings if for some reason the STDERR_FILENO file\n descriptor is not suitable for this.\n malloc_message takes the\n cbopaque pointer argument that is\n NULL unless overridden by the arguments in a call to\n malloc_stats_print, followed by a string\n pointer. Please note that doing anything which tries to allocate memory in\n this function is likely to result in a crash or deadlock.\n\n All messages are prefixed by\n “<jemalloc>: ”.\n \n \n RETURN VALUES\n \n Standard API\n The malloc and\n calloc functions return a pointer to the\n allocated memory if successful; otherwise a NULL\n pointer is returned and errno is set to\n ENOMEM.\n\n The posix_memalign function\n returns the value 0 if successful; otherwise it returns an error value.\n The posix_memalign function will fail\n if:\n \n \n EINVAL\n\n The alignment parameter is\n not a power of 2 at least as large as\n sizeof(void *).\n \n \n \n ENOMEM\n\n Memory allocation error.\n \n \n \n\n The aligned_alloc function returns\n a pointer to the allocated memory if successful; otherwise a\n NULL pointer is returned and\n errno is set. The\n aligned_alloc function will fail if:\n \n \n EINVAL\n\n The alignment parameter is\n not a power of 2.\n \n \n \n ENOMEM\n\n Memory allocation error.\n \n \n \n\n The realloc function returns a\n pointer, possibly identical to ptr, to the\n allocated memory if successful; otherwise a NULL\n pointer is returned, and errno is set to\n ENOMEM if the error was the result of an\n allocation failure. The realloc\n function always leaves the original buffer intact when an error occurs.\n \n\n The free function returns no\n value.\n \n \n Non-standard API\n The malloc_usable_size function\n returns the usable size of the allocation pointed to by\n ptr. \n\n The mallctl,\n mallctlnametomib, and\n mallctlbymib functions return 0 on\n success; otherwise they return an error value. The functions will fail\n if:\n \n \n EINVAL\n\n newp is not\n NULL, and newlen is too\n large or too small. Alternatively, *oldlenp\n is too large or too small; in this case as much data as possible\n are read despite the error.\n \n \n ENOMEM\n\n *oldlenp is too short to\n hold the requested value.\n \n \n ENOENT\n\n name or\n mib specifies an unknown/invalid\n value.\n \n \n EPERM\n\n Attempt to read or write void value, or attempt to\n write read-only value.\n \n \n EAGAIN\n\n A memory allocation failure\n occurred.\n \n \n EFAULT\n\n An interface with side effects failed in some way\n not directly related to mallctl*\n read/write processing.\n \n \n \n \n \n Experimental API\n The allocm,\n rallocm,\n sallocm,\n dallocm, and\n nallocm functions return\n ALLOCM_SUCCESS on success; otherwise they return an\n error value. The allocm,\n rallocm, and\n nallocm functions will fail if:\n \n \n ALLOCM_ERR_OOM\n\n Out of memory. Insufficient contiguous memory was\n available to service the allocation request. The\n allocm function additionally sets\n *ptr to NULL, whereas\n the rallocm function leaves\n *ptr unmodified.\n \n \n The rallocm function will also\n fail if:\n \n \n ALLOCM_ERR_NOT_MOVED\n\n ALLOCM_NO_MOVE was specified,\n but the reallocation request could not be serviced without moving\n the object.\n \n \n \n \n \n \n ENVIRONMENT\n The following environment variable affects the execution of the\n allocation functions:\n \n \n MALLOC_CONF\n\n If the environment variable\n MALLOC_CONF is set, the characters it contains\n will be interpreted as options.\n \n \n \n \n \n EXAMPLES\n To dump core whenever a problem occurs:\n ln -s 'abort:true' /etc/malloc.conf\n \n To specify in the source a chunk size that is 16 MiB:\n \n \n \n SEE ALSO\n madvise\n 2,\n mmap\n 2,\n sbrk\n 2,\n utrace\n 2,\n alloca\n 3,\n atexit\n 3,\n getpagesize\n 3\n \n \n STANDARDS\n The malloc,\n calloc,\n realloc, and\n free functions conform to ISO/IEC\n 9899:1990 (“ISO C90”).\n\n The posix_memalign function conforms\n to IEEE Std 1003.1-2001 (“POSIX.1”).\n \n\n" }, { "path": "deps/jemalloc/doc/manpages.xsl.in", "content": "\n \n \n\n" }, { "path": "deps/jemalloc/doc/stylesheet.xsl", "content": "\n ansi\n \n \n \"\"\n \n\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/arena.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/*\n * RUN_MAX_OVRHD indicates maximum desired run header overhead. Runs are sized\n * as small as possible such that this setting is still honored, without\n * violating other constraints. The goal is to make runs as small as possible\n * without exceeding a per run external fragmentation threshold.\n *\n * We use binary fixed point math for overhead computations, where the binary\n * point is implicitly RUN_BFP bits to the left.\n *\n * Note that it is possible to set RUN_MAX_OVRHD low enough that it cannot be\n * honored for some/all object sizes, since when heap profiling is enabled\n * there is one pointer of header overhead per object (plus a constant). This\n * constraint is relaxed (ignored) for runs that are so small that the\n * per-region overhead is greater than:\n *\n * (RUN_MAX_OVRHD / (reg_interval << (3+RUN_BFP))\n */\n#define\tRUN_BFP\t\t\t12\n/* \\/ Implicit binary fixed point. */\n#define\tRUN_MAX_OVRHD\t\t0x0000003dU\n#define\tRUN_MAX_OVRHD_RELAX\t0x00001800U\n\n/* Maximum number of regions in one run. */\n#define\tLG_RUN_MAXREGS\t\t11\n#define\tRUN_MAXREGS\t\t(1U << LG_RUN_MAXREGS)\n\n/*\n * Minimum redzone size. Redzones may be larger than this if necessary to\n * preserve region alignment.\n */\n#define\tREDZONE_MINSIZE\t\t16\n\n/*\n * The minimum ratio of active:dirty pages per arena is computed as:\n *\n * (nactive >> opt_lg_dirty_mult) >= ndirty\n *\n * So, supposing that opt_lg_dirty_mult is 3, there can be no less than 8 times\n * as many active pages as dirty pages.\n */\n#define\tLG_DIRTY_MULT_DEFAULT\t3\n\ntypedef struct arena_chunk_map_s arena_chunk_map_t;\ntypedef struct arena_chunk_s arena_chunk_t;\ntypedef struct arena_run_s arena_run_t;\ntypedef struct arena_bin_info_s arena_bin_info_t;\ntypedef struct arena_bin_s arena_bin_t;\ntypedef struct arena_s arena_t;\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n/* Each element of the chunk map corresponds to one page within the chunk. */\nstruct arena_chunk_map_s {\n#ifndef JEMALLOC_PROF\n\t/*\n\t * Overlay prof_ctx in order to allow it to be referenced by dead code.\n\t * Such antics aren't warranted for per arena data structures, but\n\t * chunk map overhead accounts for a percentage of memory, rather than\n\t * being just a fixed cost.\n\t */\n\tunion {\n#endif\n\tunion {\n\t\t/*\n\t\t * Linkage for run trees. There are two disjoint uses:\n\t\t *\n\t\t * 1) arena_t's runs_avail tree.\n\t\t * 2) arena_run_t conceptually uses this linkage for in-use\n\t\t * non-full runs, rather than directly embedding linkage.\n\t\t */\n\t\trb_node(arena_chunk_map_t)\trb_link;\n\t\t/*\n\t\t * List of runs currently in purgatory. arena_chunk_purge()\n\t\t * temporarily allocates runs that contain dirty pages while\n\t\t * purging, so that other threads cannot use the runs while the\n\t\t * purging thread is operating without the arena lock held.\n\t\t */\n\t\tql_elm(arena_chunk_map_t)\tql_link;\n\t}\t\t\t\tu;\n\n\t/* Profile counters, used for large object runs. */\n\tprof_ctx_t\t\t\t*prof_ctx;\n#ifndef JEMALLOC_PROF\n\t}; /* union { ... }; */\n#endif\n\n\t/*\n\t * Run address (or size) and various flags are stored together. The bit\n\t * layout looks like (assuming 32-bit system):\n\t *\n\t * ???????? ???????? ????nnnn nnnndula\n\t *\n\t * ? : Unallocated: Run address for first/last pages, unset for internal\n\t * pages.\n\t * Small: Run page offset.\n\t * Large: Run size for first page, unset for trailing pages.\n\t * n : binind for small size class, BININD_INVALID for large size class.\n\t * d : dirty?\n\t * u : unzeroed?\n\t * l : large?\n\t * a : allocated?\n\t *\n\t * Following are example bit patterns for the three types of runs.\n\t *\n\t * p : run page offset\n\t * s : run size\n\t * n : binind for size class; large objects set these to BININD_INVALID\n\t * except for promoted allocations (see prof_promote)\n\t * x : don't care\n\t * - : 0\n\t * + : 1\n\t * [DULA] : bit set\n\t * [dula] : bit unset\n\t *\n\t * Unallocated (clean):\n\t * ssssssss ssssssss ssss++++ ++++du-a\n\t * xxxxxxxx xxxxxxxx xxxxxxxx xxxx-Uxx\n\t * ssssssss ssssssss ssss++++ ++++dU-a\n\t *\n\t * Unallocated (dirty):\n\t * ssssssss ssssssss ssss++++ ++++D--a\n\t * xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx\n\t * ssssssss ssssssss ssss++++ ++++D--a\n\t *\n\t * Small:\n\t * pppppppp pppppppp ppppnnnn nnnnd--A\n\t * pppppppp pppppppp ppppnnnn nnnn---A\n\t * pppppppp pppppppp ppppnnnn nnnnd--A\n\t *\n\t * Large:\n\t * ssssssss ssssssss ssss++++ ++++D-LA\n\t * xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx\n\t * -------- -------- ----++++ ++++D-LA\n\t *\n\t * Large (sampled, size <= PAGE):\n\t * ssssssss ssssssss ssssnnnn nnnnD-LA\n\t *\n\t * Large (not sampled, size == PAGE):\n\t * ssssssss ssssssss ssss++++ ++++D-LA\n\t */\n\tsize_t\t\t\t\tbits;\n#define\tCHUNK_MAP_BININD_SHIFT\t4\n#define\tBININD_INVALID\t\t((size_t)0xffU)\n/* CHUNK_MAP_BININD_MASK == (BININD_INVALID << CHUNK_MAP_BININD_SHIFT) */\n#define\tCHUNK_MAP_BININD_MASK\t((size_t)0xff0U)\n#define\tCHUNK_MAP_BININD_INVALID CHUNK_MAP_BININD_MASK\n#define\tCHUNK_MAP_FLAGS_MASK\t((size_t)0xcU)\n#define\tCHUNK_MAP_DIRTY\t\t((size_t)0x8U)\n#define\tCHUNK_MAP_UNZEROED\t((size_t)0x4U)\n#define\tCHUNK_MAP_LARGE\t\t((size_t)0x2U)\n#define\tCHUNK_MAP_ALLOCATED\t((size_t)0x1U)\n#define\tCHUNK_MAP_KEY\t\tCHUNK_MAP_ALLOCATED\n};\ntypedef rb_tree(arena_chunk_map_t) arena_avail_tree_t;\ntypedef rb_tree(arena_chunk_map_t) arena_run_tree_t;\n\n/* Arena chunk header. */\nstruct arena_chunk_s {\n\t/* Arena that owns the chunk. */\n\tarena_t\t\t\t*arena;\n\n\t/* Linkage for tree of arena chunks that contain dirty runs. */\n\trb_node(arena_chunk_t)\tdirty_link;\n\n\t/* Number of dirty pages. */\n\tsize_t\t\t\tndirty;\n\n\t/* Number of available runs. */\n\tsize_t\t\t\tnruns_avail;\n\n\t/*\n\t * Number of available run adjacencies. Clean and dirty available runs\n\t * are not coalesced, which causes virtual memory fragmentation. The\n\t * ratio of (nruns_avail-nruns_adjac):nruns_adjac is used for tracking\n\t * this fragmentation.\n\t * */\n\tsize_t\t\t\tnruns_adjac;\n\n\t/*\n\t * Map of pages within chunk that keeps track of free/large/small. The\n\t * first map_bias entries are omitted, since the chunk header does not\n\t * need to be tracked in the map. This omission saves a header page\n\t * for common chunk sizes (e.g. 4 MiB).\n\t */\n\tarena_chunk_map_t\tmap[1]; /* Dynamically sized. */\n};\ntypedef rb_tree(arena_chunk_t) arena_chunk_tree_t;\n\nstruct arena_run_s {\n\t/* Bin this run is associated with. */\n\tarena_bin_t\t*bin;\n\n\t/* Index of next region that has never been allocated, or nregs. */\n\tuint32_t\tnextind;\n\n\t/* Number of free regions in run. */\n\tunsigned\tnfree;\n};\n\n/*\n * Read-only information associated with each element of arena_t's bins array\n * is stored separately, partly to reduce memory usage (only one copy, rather\n * than one per arena), but mainly to avoid false cacheline sharing.\n *\n * Each run has the following layout:\n *\n * /--------------------\\\n * | arena_run_t header |\n * | ... |\n * bitmap_offset | bitmap |\n * | ... |\n * ctx0_offset | ctx map |\n * | ... |\n * |--------------------|\n * | redzone |\n * reg0_offset | region 0 |\n * | redzone |\n * |--------------------| \\\n * | redzone | |\n * | region 1 | > reg_interval\n * | redzone | /\n * |--------------------|\n * | ... |\n * | ... |\n * | ... |\n * |--------------------|\n * | redzone |\n * | region nregs-1 |\n * | redzone |\n * |--------------------|\n * | alignment pad? |\n * \\--------------------/\n *\n * reg_interval has at least the same minimum alignment as reg_size; this\n * preserves the alignment constraint that sa2u() depends on. Alignment pad is\n * either 0 or redzone_size; it is present only if needed to align reg0_offset.\n */\nstruct arena_bin_info_s {\n\t/* Size of regions in a run for this bin's size class. */\n\tsize_t\t\treg_size;\n\n\t/* Redzone size. */\n\tsize_t\t\tredzone_size;\n\n\t/* Interval between regions (reg_size + (redzone_size << 1)). */\n\tsize_t\t\treg_interval;\n\n\t/* Total size of a run for this bin's size class. */\n\tsize_t\t\trun_size;\n\n\t/* Total number of regions in a run for this bin's size class. */\n\tuint32_t\tnregs;\n\n\t/*\n\t * Offset of first bitmap_t element in a run header for this bin's size\n\t * class.\n\t */\n\tuint32_t\tbitmap_offset;\n\n\t/*\n\t * Metadata used to manipulate bitmaps for runs associated with this\n\t * bin.\n\t */\n\tbitmap_info_t\tbitmap_info;\n\n\t/*\n\t * Offset of first (prof_ctx_t *) in a run header for this bin's size\n\t * class, or 0 if (config_prof == false || opt_prof == false).\n\t */\n\tuint32_t\tctx0_offset;\n\n\t/* Offset of first region in a run for this bin's size class. */\n\tuint32_t\treg0_offset;\n};\n\nstruct arena_bin_s {\n\t/*\n\t * All operations on runcur, runs, and stats require that lock be\n\t * locked. Run allocation/deallocation are protected by the arena lock,\n\t * which may be acquired while holding one or more bin locks, but not\n\t * vise versa.\n\t */\n\tmalloc_mutex_t\tlock;\n\n\t/*\n\t * Current run being used to service allocations of this bin's size\n\t * class.\n\t */\n\tarena_run_t\t*runcur;\n\n\t/*\n\t * Tree of non-full runs. This tree is used when looking for an\n\t * existing run when runcur is no longer usable. We choose the\n\t * non-full run that is lowest in memory; this policy tends to keep\n\t * objects packed well, and it can also help reduce the number of\n\t * almost-empty chunks.\n\t */\n\tarena_run_tree_t runs;\n\n\t/* Bin statistics. */\n\tmalloc_bin_stats_t stats;\n};\n\nstruct arena_s {\n\t/* This arena's index within the arenas array. */\n\tunsigned\t\tind;\n\n\t/*\n\t * Number of threads currently assigned to this arena. This field is\n\t * protected by arenas_lock.\n\t */\n\tunsigned\t\tnthreads;\n\n\t/*\n\t * There are three classes of arena operations from a locking\n\t * perspective:\n\t * 1) Thread asssignment (modifies nthreads) is protected by\n\t * arenas_lock.\n\t * 2) Bin-related operations are protected by bin locks.\n\t * 3) Chunk- and run-related operations are protected by this mutex.\n\t */\n\tmalloc_mutex_t\t\tlock;\n\n\tarena_stats_t\t\tstats;\n\t/*\n\t * List of tcaches for extant threads associated with this arena.\n\t * Stats from these are merged incrementally, and at exit.\n\t */\n\tql_head(tcache_t)\ttcache_ql;\n\n\tuint64_t\t\tprof_accumbytes;\n\n\tdss_prec_t\t\tdss_prec;\n\n\t/* Tree of dirty-page-containing chunks this arena manages. */\n\tarena_chunk_tree_t\tchunks_dirty;\n\n\t/*\n\t * In order to avoid rapid chunk allocation/deallocation when an arena\n\t * oscillates right on the cusp of needing a new chunk, cache the most\n\t * recently freed chunk. The spare is left in the arena's chunk trees\n\t * until it is deleted.\n\t *\n\t * There is one spare chunk per arena, rather than one spare total, in\n\t * order to avoid interactions between multiple threads that could make\n\t * a single spare inadequate.\n\t */\n\tarena_chunk_t\t\t*spare;\n\n\t/* Number of pages in active runs. */\n\tsize_t\t\t\tnactive;\n\n\t/*\n\t * Current count of pages within unused runs that are potentially\n\t * dirty, and for which madvise(... MADV_DONTNEED) has not been called.\n\t * By tracking this, we can institute a limit on how much dirty unused\n\t * memory is mapped for each arena.\n\t */\n\tsize_t\t\t\tndirty;\n\n\t/*\n\t * Approximate number of pages being purged. It is possible for\n\t * multiple threads to purge dirty pages concurrently, and they use\n\t * npurgatory to indicate the total number of pages all threads are\n\t * attempting to purge.\n\t */\n\tsize_t\t\t\tnpurgatory;\n\n\t/*\n\t * Size/address-ordered trees of this arena's available runs. The trees\n\t * are used for first-best-fit run allocation.\n\t */\n\tarena_avail_tree_t\truns_avail;\n\n\t/* bins is used to store trees of free regions. */\n\tarena_bin_t\t\tbins[NBINS];\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nextern ssize_t\topt_lg_dirty_mult;\n/*\n * small_size2bin is a compact lookup table that rounds request sizes up to\n * size classes. In order to reduce cache footprint, the table is compressed,\n * and all accesses are via the SMALL_SIZE2BIN macro.\n */\nextern uint8_t const\tsmall_size2bin[];\n#define\tSMALL_SIZE2BIN(s)\t(small_size2bin[(s-1) >> LG_TINY_MIN])\n\nextern arena_bin_info_t\tarena_bin_info[NBINS];\n\n/* Number of large size classes. */\n#define\t\t\tnlclasses (chunk_npages - map_bias)\n\nvoid\tarena_purge_all(arena_t *arena);\nvoid\tarena_prof_accum(arena_t *arena, uint64_t accumbytes);\nvoid\tarena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin,\n size_t binind, uint64_t prof_accumbytes);\nvoid\tarena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,\n bool zero);\nvoid\tarena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info);\nvoid\t*arena_malloc_small(arena_t *arena, size_t size, bool zero);\nvoid\t*arena_malloc_large(arena_t *arena, size_t size, bool zero);\nvoid\t*arena_palloc(arena_t *arena, size_t size, size_t alignment, bool zero);\nvoid\tarena_prof_promoted(const void *ptr, size_t size);\nvoid\tarena_dalloc_bin_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n arena_chunk_map_t *mapelm);\nvoid\tarena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t pageind, arena_chunk_map_t *mapelm);\nvoid\tarena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t pageind);\nvoid\tarena_dalloc_large_locked(arena_t *arena, arena_chunk_t *chunk,\n void *ptr);\nvoid\tarena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr);\nvoid\t*arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size,\n size_t extra, bool zero);\nvoid\t*arena_ralloc(arena_t *arena, void *ptr, size_t oldsize, size_t size,\n size_t extra, size_t alignment, bool zero, bool try_tcache_alloc,\n bool try_tcache_dalloc);\ndss_prec_t\tarena_dss_prec_get(arena_t *arena);\nvoid\tarena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);\nvoid\tarena_stats_merge(arena_t *arena, const char **dss, size_t *nactive,\n size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats,\n malloc_large_stats_t *lstats);\nbool\tarena_new(arena_t *arena, unsigned ind);\nvoid\tarena_boot(void);\nvoid\tarena_prefork(arena_t *arena);\nvoid\tarena_postfork_parent(arena_t *arena);\nvoid\tarena_postfork_child(arena_t *arena);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\narena_chunk_map_t\t*arena_mapp_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\t*arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_unallocated_size_get(arena_chunk_t *chunk,\n size_t pageind);\nsize_t\tarena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind);\nsize_t\tarena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind);\nvoid\tarena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind,\n size_t size, size_t flags);\nvoid\tarena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,\n size_t size);\nvoid\tarena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind,\n size_t size, size_t flags);\nvoid\tarena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,\n size_t binind);\nvoid\tarena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind,\n size_t runind, size_t binind, size_t flags);\nvoid\tarena_mapbits_unzeroed_set(arena_chunk_t *chunk, size_t pageind,\n size_t unzeroed);\nsize_t\tarena_ptr_small_binind_get(const void *ptr, size_t mapbits);\nsize_t\tarena_bin_index(arena_t *arena, arena_bin_t *bin);\nunsigned\tarena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,\n const void *ptr);\nprof_ctx_t\t*arena_prof_ctx_get(const void *ptr);\nvoid\tarena_prof_ctx_set(const void *ptr, prof_ctx_t *ctx);\nvoid\t*arena_malloc(arena_t *arena, size_t size, bool zero, bool try_tcache);\nsize_t\tarena_salloc(const void *ptr, bool demote);\nvoid\tarena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n bool try_tcache);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_))\n# ifdef JEMALLOC_ARENA_INLINE_A\nJEMALLOC_INLINE arena_chunk_map_t *\narena_mapp_get(arena_chunk_t *chunk, size_t pageind)\n{\n\n\tassert(pageind >= map_bias);\n\tassert(pageind < chunk_npages);\n\n\treturn (&chunk->map[pageind-map_bias]);\n}\n\nJEMALLOC_INLINE size_t *\narena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind)\n{\n\n\treturn (&arena_mapp_get(chunk, pageind)->bits);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_get(arena_chunk_t *chunk, size_t pageind)\n{\n\n\treturn (*arena_mapbitsp_get(chunk, pageind));\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);\n\treturn (mapbits & ~PAGE_MASK);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) ==\n\t (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED));\n\treturn (mapbits & ~PAGE_MASK);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) ==\n\t CHUNK_MAP_ALLOCATED);\n\treturn (mapbits >> LG_PAGE);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\tsize_t binind;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tbinind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;\n\tassert(binind < NBINS || binind == BININD_INVALID);\n\treturn (binind);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\treturn (mapbits & CHUNK_MAP_DIRTY);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\treturn (mapbits & CHUNK_MAP_UNZEROED);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\treturn (mapbits & CHUNK_MAP_LARGE);\n}\n\nJEMALLOC_INLINE size_t\narena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind)\n{\n\tsize_t mapbits;\n\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\treturn (mapbits & CHUNK_MAP_ALLOCATED);\n}\n\nJEMALLOC_INLINE void\narena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,\n size_t flags)\n{\n\tsize_t *mapbitsp;\n\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\tassert((size & PAGE_MASK) == 0);\n\tassert((flags & ~CHUNK_MAP_FLAGS_MASK) == 0);\n\tassert((flags & (CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == flags);\n\t*mapbitsp = size | CHUNK_MAP_BININD_INVALID | flags;\n}\n\nJEMALLOC_INLINE void\narena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,\n size_t size)\n{\n\tsize_t *mapbitsp;\n\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\tassert((size & PAGE_MASK) == 0);\n\tassert((*mapbitsp & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);\n\t*mapbitsp = size | (*mapbitsp & PAGE_MASK);\n}\n\nJEMALLOC_INLINE void\narena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size,\n size_t flags)\n{\n\tsize_t *mapbitsp;\n\tsize_t unzeroed;\n\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\tassert((size & PAGE_MASK) == 0);\n\tassert((flags & CHUNK_MAP_DIRTY) == flags);\n\tunzeroed = *mapbitsp & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */\n\t*mapbitsp = size | CHUNK_MAP_BININD_INVALID | flags | unzeroed |\n\t CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED;\n}\n\nJEMALLOC_INLINE void\narena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,\n size_t binind)\n{\n\tsize_t *mapbitsp;\n\n\tassert(binind <= BININD_INVALID);\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\tassert(arena_mapbits_large_size_get(chunk, pageind) == PAGE);\n\t*mapbitsp = (*mapbitsp & ~CHUNK_MAP_BININD_MASK) | (binind <<\n\t CHUNK_MAP_BININD_SHIFT);\n}\n\nJEMALLOC_INLINE void\narena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,\n size_t binind, size_t flags)\n{\n\tsize_t *mapbitsp;\n\tsize_t unzeroed;\n\n\tassert(binind < BININD_INVALID);\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\tassert(pageind - runind >= map_bias);\n\tassert((flags & CHUNK_MAP_DIRTY) == flags);\n\tunzeroed = *mapbitsp & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */\n\t*mapbitsp = (runind << LG_PAGE) | (binind << CHUNK_MAP_BININD_SHIFT) |\n\t flags | unzeroed | CHUNK_MAP_ALLOCATED;\n}\n\nJEMALLOC_INLINE void\narena_mapbits_unzeroed_set(arena_chunk_t *chunk, size_t pageind,\n size_t unzeroed)\n{\n\tsize_t *mapbitsp;\n\n\tmapbitsp = arena_mapbitsp_get(chunk, pageind);\n\t*mapbitsp = (*mapbitsp & ~CHUNK_MAP_UNZEROED) | unzeroed;\n}\n\nJEMALLOC_INLINE size_t\narena_ptr_small_binind_get(const void *ptr, size_t mapbits)\n{\n\tsize_t binind;\n\n\tbinind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;\n\n\tif (config_debug) {\n\t\tarena_chunk_t *chunk;\n\t\tarena_t *arena;\n\t\tsize_t pageind;\n\t\tsize_t actual_mapbits;\n\t\tarena_run_t *run;\n\t\tarena_bin_t *bin;\n\t\tsize_t actual_binind;\n\t\tarena_bin_info_t *bin_info;\n\n\t\tassert(binind != BININD_INVALID);\n\t\tassert(binind < NBINS);\n\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\t\tarena = chunk->arena;\n\t\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\t\tactual_mapbits = arena_mapbits_get(chunk, pageind);\n\t\tassert(mapbits == actual_mapbits);\n\t\tassert(arena_mapbits_large_get(chunk, pageind) == 0);\n\t\tassert(arena_mapbits_allocated_get(chunk, pageind) != 0);\n\t\trun = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind -\n\t\t (actual_mapbits >> LG_PAGE)) << LG_PAGE));\n\t\tbin = run->bin;\n\t\tactual_binind = bin - arena->bins;\n\t\tassert(binind == actual_binind);\n\t\tbin_info = &arena_bin_info[actual_binind];\n\t\tassert(((uintptr_t)ptr - ((uintptr_t)run +\n\t\t (uintptr_t)bin_info->reg0_offset)) % bin_info->reg_interval\n\t\t == 0);\n\t}\n\n\treturn (binind);\n}\n# endif /* JEMALLOC_ARENA_INLINE_A */\n\n# ifdef JEMALLOC_ARENA_INLINE_B\nJEMALLOC_INLINE size_t\narena_bin_index(arena_t *arena, arena_bin_t *bin)\n{\n\tsize_t binind = bin - arena->bins;\n\tassert(binind < NBINS);\n\treturn (binind);\n}\n\nJEMALLOC_INLINE unsigned\narena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)\n{\n\tunsigned shift, diff, regind;\n\tsize_t interval;\n\n\t/*\n\t * Freeing a pointer lower than region zero can cause assertion\n\t * failure.\n\t */\n\tassert((uintptr_t)ptr >= (uintptr_t)run +\n\t (uintptr_t)bin_info->reg0_offset);\n\n\t/*\n\t * Avoid doing division with a variable divisor if possible. Using\n\t * actual division here can reduce allocator throughput by over 20%!\n\t */\n\tdiff = (unsigned)((uintptr_t)ptr - (uintptr_t)run -\n\t bin_info->reg0_offset);\n\n\t/* Rescale (factor powers of 2 out of the numerator and denominator). */\n\tinterval = bin_info->reg_interval;\n\tshift = ffs(interval) - 1;\n\tdiff >>= shift;\n\tinterval >>= shift;\n\n\tif (interval == 1) {\n\t\t/* The divisor was a power of 2. */\n\t\tregind = diff;\n\t} else {\n\t\t/*\n\t\t * To divide by a number D that is not a power of two we\n\t\t * multiply by (2^21 / D) and then right shift by 21 positions.\n\t\t *\n\t\t * X / D\n\t\t *\n\t\t * becomes\n\t\t *\n\t\t * (X * interval_invs[D - 3]) >> SIZE_INV_SHIFT\n\t\t *\n\t\t * We can omit the first three elements, because we never\n\t\t * divide by 0, and 1 and 2 are both powers of two, which are\n\t\t * handled above.\n\t\t */\n#define\tSIZE_INV_SHIFT\t((sizeof(unsigned) << 3) - LG_RUN_MAXREGS)\n#define\tSIZE_INV(s)\t(((1U << SIZE_INV_SHIFT) / (s)) + 1)\n\t\tstatic const unsigned interval_invs[] = {\n\t\t SIZE_INV(3),\n\t\t SIZE_INV(4), SIZE_INV(5), SIZE_INV(6), SIZE_INV(7),\n\t\t SIZE_INV(8), SIZE_INV(9), SIZE_INV(10), SIZE_INV(11),\n\t\t SIZE_INV(12), SIZE_INV(13), SIZE_INV(14), SIZE_INV(15),\n\t\t SIZE_INV(16), SIZE_INV(17), SIZE_INV(18), SIZE_INV(19),\n\t\t SIZE_INV(20), SIZE_INV(21), SIZE_INV(22), SIZE_INV(23),\n\t\t SIZE_INV(24), SIZE_INV(25), SIZE_INV(26), SIZE_INV(27),\n\t\t SIZE_INV(28), SIZE_INV(29), SIZE_INV(30), SIZE_INV(31)\n\t\t};\n\n\t\tif (interval <= ((sizeof(interval_invs) / sizeof(unsigned)) +\n\t\t 2)) {\n\t\t\tregind = (diff * interval_invs[interval - 3]) >>\n\t\t\t SIZE_INV_SHIFT;\n\t\t} else\n\t\t\tregind = diff / interval;\n#undef SIZE_INV\n#undef SIZE_INV_SHIFT\n\t}\n\tassert(diff == regind * interval);\n\tassert(regind < bin_info->nregs);\n\n\treturn (regind);\n}\n\nJEMALLOC_INLINE prof_ctx_t *\narena_prof_ctx_get(const void *ptr)\n{\n\tprof_ctx_t *ret;\n\tarena_chunk_t *chunk;\n\tsize_t pageind, mapbits;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\tassert(CHUNK_ADDR2BASE(ptr) != ptr);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert((mapbits & CHUNK_MAP_ALLOCATED) != 0);\n\tif ((mapbits & CHUNK_MAP_LARGE) == 0) {\n\t\tif (prof_promote)\n\t\t\tret = (prof_ctx_t *)(uintptr_t)1U;\n\t\telse {\n\t\t\tarena_run_t *run = (arena_run_t *)((uintptr_t)chunk +\n\t\t\t (uintptr_t)((pageind - (mapbits >> LG_PAGE)) <<\n\t\t\t LG_PAGE));\n\t\t\tsize_t binind = arena_ptr_small_binind_get(ptr,\n\t\t\t mapbits);\n\t\t\tarena_bin_info_t *bin_info = &arena_bin_info[binind];\n\t\t\tunsigned regind;\n\n\t\t\tregind = arena_run_regind(run, bin_info, ptr);\n\t\t\tret = *(prof_ctx_t **)((uintptr_t)run +\n\t\t\t bin_info->ctx0_offset + (regind *\n\t\t\t sizeof(prof_ctx_t *)));\n\t\t}\n\t} else\n\t\tret = arena_mapp_get(chunk, pageind)->prof_ctx;\n\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void\narena_prof_ctx_set(const void *ptr, prof_ctx_t *ctx)\n{\n\tarena_chunk_t *chunk;\n\tsize_t pageind, mapbits;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\tassert(CHUNK_ADDR2BASE(ptr) != ptr);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert((mapbits & CHUNK_MAP_ALLOCATED) != 0);\n\tif ((mapbits & CHUNK_MAP_LARGE) == 0) {\n\t\tif (prof_promote == false) {\n\t\t\tarena_run_t *run = (arena_run_t *)((uintptr_t)chunk +\n\t\t\t (uintptr_t)((pageind - (mapbits >> LG_PAGE)) <<\n\t\t\t LG_PAGE));\n\t\t\tsize_t binind;\n\t\t\tarena_bin_info_t *bin_info;\n\t\t\tunsigned regind;\n\n\t\t\tbinind = arena_ptr_small_binind_get(ptr, mapbits);\n\t\t\tbin_info = &arena_bin_info[binind];\n\t\t\tregind = arena_run_regind(run, bin_info, ptr);\n\n\t\t\t*((prof_ctx_t **)((uintptr_t)run + bin_info->ctx0_offset\n\t\t\t + (regind * sizeof(prof_ctx_t *)))) = ctx;\n\t\t} else\n\t\t\tassert((uintptr_t)ctx == (uintptr_t)1U);\n\t} else\n\t\tarena_mapp_get(chunk, pageind)->prof_ctx = ctx;\n}\n\nJEMALLOC_INLINE void *\narena_malloc(arena_t *arena, size_t size, bool zero, bool try_tcache)\n{\n\ttcache_t *tcache;\n\n\tassert(size != 0);\n\tassert(size <= arena_maxclass);\n\n\tif (size <= SMALL_MAXCLASS) {\n\t\tif (try_tcache && (tcache = tcache_get(true)) != NULL)\n\t\t\treturn (tcache_alloc_small(tcache, size, zero));\n\t\telse {\n\t\t\treturn (arena_malloc_small(choose_arena(arena), size,\n\t\t\t zero));\n\t\t}\n\t} else {\n\t\t/*\n\t\t * Initialize tcache after checking size in order to avoid\n\t\t * infinite recursion during tcache initialization.\n\t\t */\n\t\tif (try_tcache && size <= tcache_maxclass && (tcache =\n\t\t tcache_get(true)) != NULL)\n\t\t\treturn (tcache_alloc_large(tcache, size, zero));\n\t\telse {\n\t\t\treturn (arena_malloc_large(choose_arena(arena), size,\n\t\t\t zero));\n\t\t}\n\t}\n}\n\n/* Return the size of the allocation pointed to by ptr. */\nJEMALLOC_INLINE size_t\narena_salloc(const void *ptr, bool demote)\n{\n\tsize_t ret;\n\tarena_chunk_t *chunk;\n\tsize_t pageind, binind;\n\n\tassert(ptr != NULL);\n\tassert(CHUNK_ADDR2BASE(ptr) != ptr);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tassert(arena_mapbits_allocated_get(chunk, pageind) != 0);\n\tbinind = arena_mapbits_binind_get(chunk, pageind);\n\tif (binind == BININD_INVALID || (config_prof && demote == false &&\n\t prof_promote && arena_mapbits_large_get(chunk, pageind) != 0)) {\n\t\t/*\n\t\t * Large allocation. In the common case (demote == true), and\n\t\t * as this is an inline function, most callers will only end up\n\t\t * looking at binind to determine that ptr is a small\n\t\t * allocation.\n\t\t */\n\t\tassert(((uintptr_t)ptr & PAGE_MASK) == 0);\n\t\tret = arena_mapbits_large_size_get(chunk, pageind);\n\t\tassert(ret != 0);\n\t\tassert(pageind + (ret>>LG_PAGE) <= chunk_npages);\n\t\tassert(ret == PAGE || arena_mapbits_large_size_get(chunk,\n\t\t pageind+(ret>>LG_PAGE)-1) == 0);\n\t\tassert(binind == arena_mapbits_binind_get(chunk,\n\t\t pageind+(ret>>LG_PAGE)-1));\n\t\tassert(arena_mapbits_dirty_get(chunk, pageind) ==\n\t\t arena_mapbits_dirty_get(chunk, pageind+(ret>>LG_PAGE)-1));\n\t} else {\n\t\t/*\n\t\t * Small allocation (possibly promoted to a large object due to\n\t\t * prof_promote).\n\t\t */\n\t\tassert(arena_mapbits_large_get(chunk, pageind) != 0 ||\n\t\t arena_ptr_small_binind_get(ptr, arena_mapbits_get(chunk,\n\t\t pageind)) == binind);\n\t\tret = arena_bin_info[binind].reg_size;\n\t}\n\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void\narena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr, bool try_tcache)\n{\n\tsize_t pageind, mapbits;\n\ttcache_t *tcache;\n\n\tassert(arena != NULL);\n\tassert(chunk->arena == arena);\n\tassert(ptr != NULL);\n\tassert(CHUNK_ADDR2BASE(ptr) != ptr);\n\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tmapbits = arena_mapbits_get(chunk, pageind);\n\tassert(arena_mapbits_allocated_get(chunk, pageind) != 0);\n\tif ((mapbits & CHUNK_MAP_LARGE) == 0) {\n\t\t/* Small allocation. */\n\t\tif (try_tcache && (tcache = tcache_get(false)) != NULL) {\n\t\t\tsize_t binind;\n\n\t\t\tbinind = arena_ptr_small_binind_get(ptr, mapbits);\n\t\t\ttcache_dalloc_small(tcache, ptr, binind);\n\t\t} else\n\t\t\tarena_dalloc_small(arena, chunk, ptr, pageind);\n\t} else {\n\t\tsize_t size = arena_mapbits_large_size_get(chunk, pageind);\n\n\t\tassert(((uintptr_t)ptr & PAGE_MASK) == 0);\n\n\t\tif (try_tcache && size <= tcache_maxclass && (tcache =\n\t\t tcache_get(false)) != NULL) {\n\t\t\ttcache_dalloc_large(tcache, ptr, size);\n\t\t} else\n\t\t\tarena_dalloc_large(arena, chunk, ptr);\n\t}\n}\n# endif /* JEMALLOC_ARENA_INLINE_B */\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/atomic.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n#define\tatomic_read_uint64(p)\tatomic_add_uint64(p, 0)\n#define\tatomic_read_uint32(p)\tatomic_add_uint32(p, 0)\n#define\tatomic_read_z(p)\tatomic_add_z(p, 0)\n#define\tatomic_read_u(p)\tatomic_add_u(p, 0)\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nuint64_t\tatomic_add_uint64(uint64_t *p, uint64_t x);\nuint64_t\tatomic_sub_uint64(uint64_t *p, uint64_t x);\nuint32_t\tatomic_add_uint32(uint32_t *p, uint32_t x);\nuint32_t\tatomic_sub_uint32(uint32_t *p, uint32_t x);\nsize_t\tatomic_add_z(size_t *p, size_t x);\nsize_t\tatomic_sub_z(size_t *p, size_t x);\nunsigned\tatomic_add_u(unsigned *p, unsigned x);\nunsigned\tatomic_sub_u(unsigned *p, unsigned x);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ATOMIC_C_))\n/******************************************************************************/\n/* 64-bit operations. */\n#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)\n# ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (__sync_add_and_fetch(p, x));\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (__sync_sub_and_fetch(p, x));\n}\n#elif (defined(_MSC_VER))\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (InterlockedExchangeAdd64(p, x));\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (InterlockedExchangeAdd64(p, -((int64_t)x)));\n}\n#elif (defined(JEMALLOC_OSATOMIC))\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (OSAtomicAdd64((int64_t)x, (int64_t *)p));\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (OSAtomicAdd64(-((int64_t)x), (int64_t *)p));\n}\n# elif (defined(__amd64__) || defined(__x86_64__))\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\tasm volatile (\n\t \"lock; xaddq %0, %1;\"\n\t : \"+r\" (x), \"=m\" (*p) /* Outputs. */\n\t : \"m\" (*p) /* Inputs. */\n\t );\n\n\treturn (x);\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\tx = (uint64_t)(-(int64_t)x);\n\tasm volatile (\n\t \"lock; xaddq %0, %1;\"\n\t : \"+r\" (x), \"=m\" (*p) /* Outputs. */\n\t : \"m\" (*p) /* Inputs. */\n\t );\n\n\treturn (x);\n}\n# elif (defined(JEMALLOC_ATOMIC9))\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\t/*\n\t * atomic_fetchadd_64() doesn't exist, but we only ever use this\n\t * function on LP64 systems, so atomic_fetchadd_long() will do.\n\t */\n\tassert(sizeof(uint64_t) == sizeof(unsigned long));\n\n\treturn (atomic_fetchadd_long(p, (unsigned long)x) + x);\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\tassert(sizeof(uint64_t) == sizeof(unsigned long));\n\n\treturn (atomic_fetchadd_long(p, (unsigned long)(-(long)x)) - x);\n}\n# elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_8))\nJEMALLOC_INLINE uint64_t\natomic_add_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (__sync_add_and_fetch(p, x));\n}\n\nJEMALLOC_INLINE uint64_t\natomic_sub_uint64(uint64_t *p, uint64_t x)\n{\n\n\treturn (__sync_sub_and_fetch(p, x));\n}\n# else\n# error \"Missing implementation for 64-bit atomic operations\"\n# endif\n#endif\n\n/******************************************************************************/\n/* 32-bit operations. */\n#ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (__sync_add_and_fetch(p, x));\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (__sync_sub_and_fetch(p, x));\n}\n#elif (defined(_MSC_VER))\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (InterlockedExchangeAdd(p, x));\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (InterlockedExchangeAdd(p, -((int32_t)x)));\n}\n#elif (defined(JEMALLOC_OSATOMIC))\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (OSAtomicAdd32((int32_t)x, (int32_t *)p));\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (OSAtomicAdd32(-((int32_t)x), (int32_t *)p));\n}\n#elif (defined(__i386__) || defined(__amd64__) || defined(__x86_64__))\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\tasm volatile (\n\t \"lock; xaddl %0, %1;\"\n\t : \"+r\" (x), \"=m\" (*p) /* Outputs. */\n\t : \"m\" (*p) /* Inputs. */\n\t );\n\n\treturn (x);\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\tx = (uint32_t)(-(int32_t)x);\n\tasm volatile (\n\t \"lock; xaddl %0, %1;\"\n\t : \"+r\" (x), \"=m\" (*p) /* Outputs. */\n\t : \"m\" (*p) /* Inputs. */\n\t );\n\n\treturn (x);\n}\n#elif (defined(JEMALLOC_ATOMIC9))\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (atomic_fetchadd_32(p, x) + x);\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (atomic_fetchadd_32(p, (uint32_t)(-(int32_t)x)) - x);\n}\n#elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4))\nJEMALLOC_INLINE uint32_t\natomic_add_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (__sync_add_and_fetch(p, x));\n}\n\nJEMALLOC_INLINE uint32_t\natomic_sub_uint32(uint32_t *p, uint32_t x)\n{\n\n\treturn (__sync_sub_and_fetch(p, x));\n}\n#else\n# error \"Missing implementation for 32-bit atomic operations\"\n#endif\n\n/******************************************************************************/\n/* size_t operations. */\nJEMALLOC_INLINE size_t\natomic_add_z(size_t *p, size_t x)\n{\n\n#if (LG_SIZEOF_PTR == 3)\n\treturn ((size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)x));\n#elif (LG_SIZEOF_PTR == 2)\n\treturn ((size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)x));\n#endif\n}\n\nJEMALLOC_INLINE size_t\natomic_sub_z(size_t *p, size_t x)\n{\n\n#if (LG_SIZEOF_PTR == 3)\n\treturn ((size_t)atomic_add_uint64((uint64_t *)p,\n\t (uint64_t)-((int64_t)x)));\n#elif (LG_SIZEOF_PTR == 2)\n\treturn ((size_t)atomic_add_uint32((uint32_t *)p,\n\t (uint32_t)-((int32_t)x)));\n#endif\n}\n\n/******************************************************************************/\n/* unsigned operations. */\nJEMALLOC_INLINE unsigned\natomic_add_u(unsigned *p, unsigned x)\n{\n\n#if (LG_SIZEOF_INT == 3)\n\treturn ((unsigned)atomic_add_uint64((uint64_t *)p, (uint64_t)x));\n#elif (LG_SIZEOF_INT == 2)\n\treturn ((unsigned)atomic_add_uint32((uint32_t *)p, (uint32_t)x));\n#endif\n}\n\nJEMALLOC_INLINE unsigned\natomic_sub_u(unsigned *p, unsigned x)\n{\n\n#if (LG_SIZEOF_INT == 3)\n\treturn ((unsigned)atomic_add_uint64((uint64_t *)p,\n\t (uint64_t)-((int64_t)x)));\n#elif (LG_SIZEOF_INT == 2)\n\treturn ((unsigned)atomic_add_uint32((uint32_t *)p,\n\t (uint32_t)-((int32_t)x)));\n#endif\n}\n/******************************************************************************/\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/base.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nvoid\t*base_alloc(size_t size);\nvoid\t*base_calloc(size_t number, size_t size);\nextent_node_t *base_node_alloc(void);\nvoid\tbase_node_dealloc(extent_node_t *node);\nbool\tbase_boot(void);\nvoid\tbase_prefork(void);\nvoid\tbase_postfork_parent(void);\nvoid\tbase_postfork_child(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/bitmap.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/* Maximum bitmap bit count is 2^LG_BITMAP_MAXBITS. */\n#define\tLG_BITMAP_MAXBITS\tLG_RUN_MAXREGS\n\ntypedef struct bitmap_level_s bitmap_level_t;\ntypedef struct bitmap_info_s bitmap_info_t;\ntypedef unsigned long bitmap_t;\n#define\tLG_SIZEOF_BITMAP\tLG_SIZEOF_LONG\n\n/* Number of bits per group. */\n#define\tLG_BITMAP_GROUP_NBITS\t\t(LG_SIZEOF_BITMAP + 3)\n#define\tBITMAP_GROUP_NBITS\t\t(ZU(1) << LG_BITMAP_GROUP_NBITS)\n#define\tBITMAP_GROUP_NBITS_MASK\t\t(BITMAP_GROUP_NBITS-1)\n\n/* Maximum number of levels possible. */\n#define\tBITMAP_MAX_LEVELS\t\t\t\t\t\t\\\n (LG_BITMAP_MAXBITS / LG_SIZEOF_BITMAP)\t\t\t\t\\\n + !!(LG_BITMAP_MAXBITS % LG_SIZEOF_BITMAP)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct bitmap_level_s {\n\t/* Offset of this level's groups within the array of groups. */\n\tsize_t group_offset;\n};\n\nstruct bitmap_info_s {\n\t/* Logical number of bits in bitmap (stored at bottom level). */\n\tsize_t nbits;\n\n\t/* Number of levels necessary for nbits. */\n\tunsigned nlevels;\n\n\t/*\n\t * Only the first (nlevels+1) elements are used, and levels are ordered\n\t * bottom to top (e.g. the bottom level is stored in levels[0]).\n\t */\n\tbitmap_level_t levels[BITMAP_MAX_LEVELS+1];\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nvoid\tbitmap_info_init(bitmap_info_t *binfo, size_t nbits);\nsize_t\tbitmap_info_ngroups(const bitmap_info_t *binfo);\nsize_t\tbitmap_size(size_t nbits);\nvoid\tbitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nbool\tbitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo);\nbool\tbitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit);\nvoid\tbitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit);\nsize_t\tbitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo);\nvoid\tbitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_BITMAP_C_))\nJEMALLOC_INLINE bool\nbitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo)\n{\n\tunsigned rgoff = binfo->levels[binfo->nlevels].group_offset - 1;\n\tbitmap_t rg = bitmap[rgoff];\n\t/* The bitmap is full iff the root group is 0. */\n\treturn (rg == 0);\n}\n\nJEMALLOC_INLINE bool\nbitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)\n{\n\tsize_t goff;\n\tbitmap_t g;\n\n\tassert(bit < binfo->nbits);\n\tgoff = bit >> LG_BITMAP_GROUP_NBITS;\n\tg = bitmap[goff];\n\treturn (!(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK))));\n}\n\nJEMALLOC_INLINE void\nbitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)\n{\n\tsize_t goff;\n\tbitmap_t *gp;\n\tbitmap_t g;\n\n\tassert(bit < binfo->nbits);\n\tassert(bitmap_get(bitmap, binfo, bit) == false);\n\tgoff = bit >> LG_BITMAP_GROUP_NBITS;\n\tgp = &bitmap[goff];\n\tg = *gp;\n\tassert(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)));\n\tg ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);\n\t*gp = g;\n\tassert(bitmap_get(bitmap, binfo, bit));\n\t/* Propagate group state transitions up the tree. */\n\tif (g == 0) {\n\t\tunsigned i;\n\t\tfor (i = 1; i < binfo->nlevels; i++) {\n\t\t\tbit = goff;\n\t\t\tgoff = bit >> LG_BITMAP_GROUP_NBITS;\n\t\t\tgp = &bitmap[binfo->levels[i].group_offset + goff];\n\t\t\tg = *gp;\n\t\t\tassert(g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)));\n\t\t\tg ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);\n\t\t\t*gp = g;\n\t\t\tif (g != 0)\n\t\t\t\tbreak;\n\t\t}\n\t}\n}\n\n/* sfu: set first unset. */\nJEMALLOC_INLINE size_t\nbitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo)\n{\n\tsize_t bit;\n\tbitmap_t g;\n\tunsigned i;\n\n\tassert(bitmap_full(bitmap, binfo) == false);\n\n\ti = binfo->nlevels - 1;\n\tg = bitmap[binfo->levels[i].group_offset];\n\tbit = ffsl(g) - 1;\n\twhile (i > 0) {\n\t\ti--;\n\t\tg = bitmap[binfo->levels[i].group_offset + bit];\n\t\tbit = (bit << LG_BITMAP_GROUP_NBITS) + (ffsl(g) - 1);\n\t}\n\n\tbitmap_set(bitmap, binfo, bit);\n\treturn (bit);\n}\n\nJEMALLOC_INLINE void\nbitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit)\n{\n\tsize_t goff;\n\tbitmap_t *gp;\n\tbitmap_t g;\n\tbool propagate;\n\n\tassert(bit < binfo->nbits);\n\tassert(bitmap_get(bitmap, binfo, bit));\n\tgoff = bit >> LG_BITMAP_GROUP_NBITS;\n\tgp = &bitmap[goff];\n\tg = *gp;\n\tpropagate = (g == 0);\n\tassert((g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK))) == 0);\n\tg ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);\n\t*gp = g;\n\tassert(bitmap_get(bitmap, binfo, bit) == false);\n\t/* Propagate group state transitions up the tree. */\n\tif (propagate) {\n\t\tunsigned i;\n\t\tfor (i = 1; i < binfo->nlevels; i++) {\n\t\t\tbit = goff;\n\t\t\tgoff = bit >> LG_BITMAP_GROUP_NBITS;\n\t\t\tgp = &bitmap[binfo->levels[i].group_offset + goff];\n\t\t\tg = *gp;\n\t\t\tpropagate = (g == 0);\n\t\t\tassert((g & (1LU << (bit & BITMAP_GROUP_NBITS_MASK)))\n\t\t\t == 0);\n\t\t\tg ^= 1LU << (bit & BITMAP_GROUP_NBITS_MASK);\n\t\t\t*gp = g;\n\t\t\tif (propagate == false)\n\t\t\t\tbreak;\n\t\t}\n\t}\n}\n\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/chunk.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/*\n * Size and alignment of memory chunks that are allocated by the OS's virtual\n * memory system.\n */\n#define\tLG_CHUNK_DEFAULT\t22\n\n/* Return the chunk address for allocation address a. */\n#define\tCHUNK_ADDR2BASE(a)\t\t\t\t\t\t\\\n\t((void *)((uintptr_t)(a) & ~chunksize_mask))\n\n/* Return the chunk offset of address a. */\n#define\tCHUNK_ADDR2OFFSET(a)\t\t\t\t\t\t\\\n\t((size_t)((uintptr_t)(a) & chunksize_mask))\n\n/* Return the smallest chunk multiple that is >= s. */\n#define\tCHUNK_CEILING(s)\t\t\t\t\t\t\\\n\t(((s) + chunksize_mask) & ~chunksize_mask)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nextern size_t\t\topt_lg_chunk;\nextern const char\t*opt_dss;\n\n/* Protects stats_chunks; currently not used for any other purpose. */\nextern malloc_mutex_t\tchunks_mtx;\n/* Chunk statistics. */\nextern chunk_stats_t\tstats_chunks;\n\nextern rtree_t\t\t*chunks_rtree;\n\nextern size_t\t\tchunksize;\nextern size_t\t\tchunksize_mask; /* (chunksize - 1). */\nextern size_t\t\tchunk_npages;\nextern size_t\t\tmap_bias; /* Number of arena chunk header pages. */\nextern size_t\t\tarena_maxclass; /* Max size class for arenas. */\n\nvoid\t*chunk_alloc(size_t size, size_t alignment, bool base, bool *zero,\n dss_prec_t dss_prec);\nvoid\tchunk_unmap(void *chunk, size_t size);\nvoid\tchunk_dealloc(void *chunk, size_t size, bool unmap);\nbool\tchunk_boot(void);\nvoid\tchunk_prefork(void);\nvoid\tchunk_postfork_parent(void);\nvoid\tchunk_postfork_child(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n\n#include \"jemalloc/internal/chunk_dss.h\"\n#include \"jemalloc/internal/chunk_mmap.h\"\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/chunk_dss.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef enum {\n\tdss_prec_disabled = 0,\n\tdss_prec_primary = 1,\n\tdss_prec_secondary = 2,\n\n\tdss_prec_limit = 3\n} dss_prec_t ;\n#define\tDSS_PREC_DEFAULT\tdss_prec_secondary\n#define\tDSS_DEFAULT\t\t\"secondary\"\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nextern const char *dss_prec_names[];\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\ndss_prec_t\tchunk_dss_prec_get(void);\nbool\tchunk_dss_prec_set(dss_prec_t dss_prec);\nvoid\t*chunk_alloc_dss(size_t size, size_t alignment, bool *zero);\nbool\tchunk_in_dss(void *chunk);\nbool\tchunk_dss_boot(void);\nvoid\tchunk_dss_prefork(void);\nvoid\tchunk_dss_postfork_parent(void);\nvoid\tchunk_dss_postfork_child(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/chunk_mmap.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nbool\tpages_purge(void *addr, size_t length);\n\nvoid\t*chunk_alloc_mmap(size_t size, size_t alignment, bool *zero);\nbool\tchunk_dealloc_mmap(void *chunk, size_t size);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/ckh.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct ckh_s ckh_t;\ntypedef struct ckhc_s ckhc_t;\n\n/* Typedefs to allow easy function pointer passing. */\ntypedef void ckh_hash_t (const void *, unsigned, size_t *, size_t *);\ntypedef bool ckh_keycomp_t (const void *, const void *);\n\n/* Maintain counters used to get an idea of performance. */\n/* #define\tCKH_COUNT */\n/* Print counter values in ckh_delete() (requires CKH_COUNT). */\n/* #define\tCKH_VERBOSE */\n\n/*\n * There are 2^LG_CKH_BUCKET_CELLS cells in each hash table bucket. Try to fit\n * one bucket per L1 cache line.\n */\n#define LG_CKH_BUCKET_CELLS (LG_CACHELINE - LG_SIZEOF_PTR - 1)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n/* Hash table cell. */\nstruct ckhc_s {\n\tconst void\t*key;\n\tconst void\t*data;\n};\n\nstruct ckh_s {\n#ifdef CKH_COUNT\n\t/* Counters used to get an idea of performance. */\n\tuint64_t\tngrows;\n\tuint64_t\tnshrinks;\n\tuint64_t\tnshrinkfails;\n\tuint64_t\tninserts;\n\tuint64_t\tnrelocs;\n#endif\n\n\t/* Used for pseudo-random number generation. */\n#define\tCKH_A\t\t1103515241\n#define\tCKH_C\t\t12347\n\tuint32_t\tprng_state;\n\n\t/* Total number of items. */\n\tsize_t\t\tcount;\n\n\t/*\n\t * Minimum and current number of hash table buckets. There are\n\t * 2^LG_CKH_BUCKET_CELLS cells per bucket.\n\t */\n\tunsigned\tlg_minbuckets;\n\tunsigned\tlg_curbuckets;\n\n\t/* Hash and comparison functions. */\n\tckh_hash_t\t*hash;\n\tckh_keycomp_t\t*keycomp;\n\n\t/* Hash table with 2^lg_curbuckets buckets. */\n\tckhc_t\t\t*tab;\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nbool\tckh_new(ckh_t *ckh, size_t minitems, ckh_hash_t *hash,\n ckh_keycomp_t *keycomp);\nvoid\tckh_delete(ckh_t *ckh);\nsize_t\tckh_count(ckh_t *ckh);\nbool\tckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);\nbool\tckh_insert(ckh_t *ckh, const void *key, const void *data);\nbool\tckh_remove(ckh_t *ckh, const void *searchkey, void **key,\n void **data);\nbool\tckh_search(ckh_t *ckh, const void *seachkey, void **key, void **data);\nvoid\tckh_string_hash(const void *key, unsigned minbits, size_t *hash1,\n size_t *hash2);\nbool\tckh_string_keycomp(const void *k1, const void *k2);\nvoid\tckh_pointer_hash(const void *key, unsigned minbits, size_t *hash1,\n size_t *hash2);\nbool\tckh_pointer_keycomp(const void *k1, const void *k2);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/ctl.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct ctl_node_s ctl_node_t;\ntypedef struct ctl_named_node_s ctl_named_node_t;\ntypedef struct ctl_indexed_node_s ctl_indexed_node_t;\ntypedef struct ctl_arena_stats_s ctl_arena_stats_t;\ntypedef struct ctl_stats_s ctl_stats_t;\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct ctl_node_s {\n\tbool\t\t\tnamed;\n};\n\nstruct ctl_named_node_s {\n\tstruct ctl_node_s\tnode;\n\tconst char\t\t*name;\n\t/* If (nchildren == 0), this is a terminal node. */\n\tunsigned\t\tnchildren;\n\tconst\t\t\tctl_node_t *children;\n\tint\t\t\t(*ctl)(const size_t *, size_t, void *, size_t *,\n\t void *, size_t);\n};\n\nstruct ctl_indexed_node_s {\n\tstruct ctl_node_s\tnode;\n\tconst ctl_named_node_t\t*(*index)(const size_t *, size_t, size_t);\n};\n\nstruct ctl_arena_stats_s {\n\tbool\t\t\tinitialized;\n\tunsigned\t\tnthreads;\n\tconst char\t\t*dss;\n\tsize_t\t\t\tpactive;\n\tsize_t\t\t\tpdirty;\n\tarena_stats_t\t\tastats;\n\n\t/* Aggregate stats for small size classes, based on bin stats. */\n\tsize_t\t\t\tallocated_small;\n\tuint64_t\t\tnmalloc_small;\n\tuint64_t\t\tndalloc_small;\n\tuint64_t\t\tnrequests_small;\n\n\tmalloc_bin_stats_t\tbstats[NBINS];\n\tmalloc_large_stats_t\t*lstats;\t/* nlclasses elements. */\n};\n\nstruct ctl_stats_s {\n\tsize_t\t\t\tallocated;\n\tsize_t\t\t\tactive;\n\tsize_t\t\t\tmapped;\n\tstruct {\n\t\tsize_t\t\tcurrent;\t/* stats_chunks.curchunks */\n\t\tuint64_t\ttotal;\t\t/* stats_chunks.nchunks */\n\t\tsize_t\t\thigh;\t\t/* stats_chunks.highchunks */\n\t} chunks;\n\tstruct {\n\t\tsize_t\t\tallocated;\t/* huge_allocated */\n\t\tuint64_t\tnmalloc;\t/* huge_nmalloc */\n\t\tuint64_t\tndalloc;\t/* huge_ndalloc */\n\t} huge;\n\tunsigned\t\tnarenas;\n\tctl_arena_stats_t\t*arenas;\t/* (narenas + 1) elements. */\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nint\tctl_byname(const char *name, void *oldp, size_t *oldlenp, void *newp,\n size_t newlen);\nint\tctl_nametomib(const char *name, size_t *mibp, size_t *miblenp);\n\nint\tctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen);\nbool\tctl_boot(void);\nvoid\tctl_prefork(void);\nvoid\tctl_postfork_parent(void);\nvoid\tctl_postfork_child(void);\n\n#define\txmallctl(name, oldp, oldlenp, newp, newlen) do {\t\t\\\n\tif (je_mallctl(name, oldp, oldlenp, newp, newlen)\t\t\\\n\t != 0) {\t\t\t\t\t\t\t\\\n\t\tmalloc_printf(\t\t\t\t\t\t\\\n\t\t \": Failure in xmallctl(\\\"%s\\\", ...)\\n\",\t\\\n\t\t name);\t\t\t\t\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\txmallctlnametomib(name, mibp, miblenp) do {\t\t\t\\\n\tif (je_mallctlnametomib(name, mibp, miblenp) != 0) {\t\t\\\n\t\tmalloc_printf(\": Failure in \"\t\t\t\\\n\t\t \"xmallctlnametomib(\\\"%s\\\", ...)\\n\", name);\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\txmallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) do {\t\\\n\tif (je_mallctlbymib(mib, miblen, oldp, oldlenp, newp,\t\t\\\n\t newlen) != 0) {\t\t\t\t\t\t\\\n\t\tmalloc_write(\t\t\t\t\t\t\\\n\t\t \": Failure in xmallctlbymib()\\n\");\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/extent.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct extent_node_s extent_node_t;\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n/* Tree of extents. */\nstruct extent_node_s {\n\t/* Linkage for the size/address-ordered tree. */\n\trb_node(extent_node_t)\tlink_szad;\n\n\t/* Linkage for the address-ordered tree. */\n\trb_node(extent_node_t)\tlink_ad;\n\n\t/* Profile counters, used for huge objects. */\n\tprof_ctx_t\t\t*prof_ctx;\n\n\t/* Pointer to the extent that this tree node is responsible for. */\n\tvoid\t\t\t*addr;\n\n\t/* Total region size. */\n\tsize_t\t\t\tsize;\n\n\t/* True if zero-filled; used by chunk recycling code. */\n\tbool\t\t\tzeroed;\n};\ntypedef rb_tree(extent_node_t) extent_tree_t;\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nrb_proto(, extent_tree_szad_, extent_tree_t, extent_node_t)\n\nrb_proto(, extent_tree_ad_, extent_tree_t, extent_node_t)\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/hash.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nuint64_t\thash(const void *key, size_t len, uint64_t seed);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_HASH_C_))\n/*\n * The following hash function is based on MurmurHash64A(), placed into the\n * public domain by Austin Appleby. See http://murmurhash.googlepages.com/ for\n * details.\n */\nJEMALLOC_INLINE uint64_t\nhash(const void *key, size_t len, uint64_t seed)\n{\n\tconst uint64_t m = UINT64_C(0xc6a4a7935bd1e995);\n\tconst int r = 47;\n\tuint64_t h = seed ^ (len * m);\n\tconst uint64_t *data = (const uint64_t *)key;\n\tconst uint64_t *end = data + (len/8);\n\tconst unsigned char *data2;\n\n\tassert(((uintptr_t)key & 0x7) == 0);\n\n\twhile(data != end) {\n\t\tuint64_t k = *data++;\n\n\t\tk *= m;\n\t\tk ^= k >> r;\n\t\tk *= m;\n\n\t\th ^= k;\n\t\th *= m;\n\t}\n\n\tdata2 = (const unsigned char *)data;\n\tswitch(len & 7) {\n\tcase 7: h ^= ((uint64_t)(data2[6])) << 48;\n\tcase 6: h ^= ((uint64_t)(data2[5])) << 40;\n\tcase 5: h ^= ((uint64_t)(data2[4])) << 32;\n\tcase 4: h ^= ((uint64_t)(data2[3])) << 24;\n\tcase 3: h ^= ((uint64_t)(data2[2])) << 16;\n\tcase 2: h ^= ((uint64_t)(data2[1])) << 8;\n\tcase 1: h ^= ((uint64_t)(data2[0]));\n\t\th *= m;\n\t}\n\n\th ^= h >> r;\n\th *= m;\n\th ^= h >> r;\n\n\treturn (h);\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/huge.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n/* Huge allocation statistics. */\nextern uint64_t\t\thuge_nmalloc;\nextern uint64_t\t\thuge_ndalloc;\nextern size_t\t\thuge_allocated;\n\n/* Protects chunk-related data structures. */\nextern malloc_mutex_t\thuge_mtx;\n\nvoid\t*huge_malloc(size_t size, bool zero);\nvoid\t*huge_palloc(size_t size, size_t alignment, bool zero);\nvoid\t*huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size,\n size_t extra);\nvoid\t*huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra,\n size_t alignment, bool zero, bool try_tcache_dalloc);\nvoid\thuge_dalloc(void *ptr, bool unmap);\nsize_t\thuge_salloc(const void *ptr);\nprof_ctx_t\t*huge_prof_ctx_get(const void *ptr);\nvoid\thuge_prof_ctx_set(const void *ptr, prof_ctx_t *ctx);\nbool\thuge_boot(void);\nvoid\thuge_prefork(void);\nvoid\thuge_postfork_parent(void);\nvoid\thuge_postfork_child(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/jemalloc_internal.h.in", "content": "#ifndef JEMALLOC_INTERNAL_H\n#define JEMALLOC_INTERNAL_H\n#include \n#ifdef _WIN32\n# include \n# define ENOENT ERROR_PATH_NOT_FOUND\n# define EINVAL ERROR_BAD_ARGUMENTS\n# define EAGAIN ERROR_OUTOFMEMORY\n# define EPERM ERROR_WRITE_FAULT\n# define EFAULT ERROR_INVALID_ADDRESS\n# define ENOMEM ERROR_NOT_ENOUGH_MEMORY\n# undef ERANGE\n# define ERANGE ERROR_INVALID_DATA\n#else\n# include \n# include \n# include \n# if !defined(SYS_write) && defined(__NR_write)\n# define SYS_write __NR_write\n# endif\n# include \n# include \n# include \n#endif\n#include \n\n#include \n#ifndef SIZE_T_MAX\n# define SIZE_T_MAX\tSIZE_MAX\n#endif\n#include \n#include \n#include \n#include \n#include \n#include \n#ifndef offsetof\n# define offsetof(type, member)\t((size_t)&(((type *)NULL)->member))\n#endif\n#include \n#include \n#include \n#include \n#ifdef _MSC_VER\n# include \ntypedef intptr_t ssize_t;\n# define PATH_MAX 1024\n# define STDERR_FILENO 2\n# define __func__ __FUNCTION__\n/* Disable warnings about deprecated system functions */\n# pragma warning(disable: 4996)\n#else\n# include \n#endif\n#include \n\n#define\tJEMALLOC_NO_DEMANGLE\n#include \"../jemalloc@install_suffix@.h\"\n\n#ifdef JEMALLOC_UTRACE\n#include \n#endif\n\n#ifdef JEMALLOC_VALGRIND\n#include \n#include \n#endif\n\n#include \"jemalloc/internal/private_namespace.h\"\n\n#ifdef JEMALLOC_CC_SILENCE\n#define\tUNUSED JEMALLOC_ATTR(unused)\n#else\n#define\tUNUSED\n#endif\n\nstatic const bool config_debug =\n#ifdef JEMALLOC_DEBUG\n true\n#else\n false\n#endif\n ;\nstatic const bool config_dss =\n#ifdef JEMALLOC_DSS\n true\n#else\n false\n#endif\n ;\nstatic const bool config_fill =\n#ifdef JEMALLOC_FILL\n true\n#else\n false\n#endif\n ;\nstatic const bool config_lazy_lock =\n#ifdef JEMALLOC_LAZY_LOCK\n true\n#else\n false\n#endif\n ;\nstatic const bool config_prof =\n#ifdef JEMALLOC_PROF\n true\n#else\n false\n#endif\n ;\nstatic const bool config_prof_libgcc =\n#ifdef JEMALLOC_PROF_LIBGCC\n true\n#else\n false\n#endif\n ;\nstatic const bool config_prof_libunwind =\n#ifdef JEMALLOC_PROF_LIBUNWIND\n true\n#else\n false\n#endif\n ;\nstatic const bool config_mremap =\n#ifdef JEMALLOC_MREMAP\n true\n#else\n false\n#endif\n ;\nstatic const bool config_munmap =\n#ifdef JEMALLOC_MUNMAP\n true\n#else\n false\n#endif\n ;\nstatic const bool config_stats =\n#ifdef JEMALLOC_STATS\n true\n#else\n false\n#endif\n ;\nstatic const bool config_tcache =\n#ifdef JEMALLOC_TCACHE\n true\n#else\n false\n#endif\n ;\nstatic const bool config_tls =\n#ifdef JEMALLOC_TLS\n true\n#else\n false\n#endif\n ;\nstatic const bool config_utrace =\n#ifdef JEMALLOC_UTRACE\n true\n#else\n false\n#endif\n ;\nstatic const bool config_valgrind =\n#ifdef JEMALLOC_VALGRIND\n true\n#else\n false\n#endif\n ;\nstatic const bool config_xmalloc =\n#ifdef JEMALLOC_XMALLOC\n true\n#else\n false\n#endif\n ;\nstatic const bool config_ivsalloc =\n#ifdef JEMALLOC_IVSALLOC\n true\n#else\n false\n#endif\n ;\n\n#ifdef JEMALLOC_ATOMIC9\n#include \n#endif\n\n#if (defined(JEMALLOC_OSATOMIC) || defined(JEMALLOC_OSSPIN))\n#include \n#endif\n\n#ifdef JEMALLOC_ZONE\n#include \n#include \n#include \n#include \n#endif\n\n#define\tRB_COMPACT\n#include \"jemalloc/internal/rb.h\"\n#include \"jemalloc/internal/qr.h\"\n#include \"jemalloc/internal/ql.h\"\n\n/*\n * jemalloc can conceptually be broken into components (arena, tcache, etc.),\n * but there are circular dependencies that cannot be broken without\n * substantial performance degradation. In order to reduce the effect on\n * visual code flow, read the header files in multiple passes, with one of the\n * following cpp variables defined during each pass:\n *\n * JEMALLOC_H_TYPES : Preprocessor-defined constants and psuedo-opaque data\n * types.\n * JEMALLOC_H_STRUCTS : Data structures.\n * JEMALLOC_H_EXTERNS : Extern data declarations and function prototypes.\n * JEMALLOC_H_INLINES : Inline functions.\n */\n/******************************************************************************/\n#define JEMALLOC_H_TYPES\n\n#define\tALLOCM_LG_ALIGN_MASK\t((int)0x3f)\n\n#define\tZU(z)\t((size_t)z)\n\n#ifndef __DECONST\n# define\t__DECONST(type, var)\t((type)(uintptr_t)(const void *)(var))\n#endif\n\n#ifdef JEMALLOC_DEBUG\n /* Disable inlining to make debugging easier. */\n# define JEMALLOC_INLINE\n# define inline\n#else\n# define JEMALLOC_ENABLE_INLINE\n# define JEMALLOC_INLINE static inline\n# ifdef _MSC_VER\n# define inline _inline\n# endif\n#endif\n\n/* Smallest size class to support. */\n#define\tLG_TINY_MIN\t\t3\n#define\tTINY_MIN\t\t(1U << LG_TINY_MIN)\n\n/*\n * Minimum alignment of allocations is 2^LG_QUANTUM bytes (ignoring tiny size\n * classes).\n */\n#ifndef LG_QUANTUM\n# if (defined(__i386__) || defined(_M_IX86))\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __ia64__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __alpha__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __sparc64__\n# define LG_QUANTUM\t\t4\n# endif\n# if (defined(__amd64__) || defined(__x86_64__) || defined(_M_X64))\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __arm__\n# define LG_QUANTUM\t\t3\n# endif\n# ifdef __hppa__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __mips__\n# define LG_QUANTUM\t\t3\n# endif\n# ifdef __powerpc__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __s390x__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __SH4__\n# define LG_QUANTUM\t\t4\n# endif\n# ifdef __tile__\n# define LG_QUANTUM\t\t4\n# endif\n# ifndef LG_QUANTUM\n# error \"No LG_QUANTUM definition for architecture; specify via CPPFLAGS\"\n# endif\n#endif\n\n#define\tQUANTUM\t\t\t((size_t)(1U << LG_QUANTUM))\n#define\tQUANTUM_MASK\t\t(QUANTUM - 1)\n\n/* Return the smallest quantum multiple that is >= a. */\n#define\tQUANTUM_CEILING(a)\t\t\t\t\t\t\\\n\t(((a) + QUANTUM_MASK) & ~QUANTUM_MASK)\n\n#define\tLONG\t\t\t((size_t)(1U << LG_SIZEOF_LONG))\n#define\tLONG_MASK\t\t(LONG - 1)\n\n/* Return the smallest long multiple that is >= a. */\n#define\tLONG_CEILING(a)\t\t\t\t\t\t\t\\\n\t(((a) + LONG_MASK) & ~LONG_MASK)\n\n#define\tSIZEOF_PTR\t\t(1U << LG_SIZEOF_PTR)\n#define\tPTR_MASK\t\t(SIZEOF_PTR - 1)\n\n/* Return the smallest (void *) multiple that is >= a. */\n#define\tPTR_CEILING(a)\t\t\t\t\t\t\t\\\n\t(((a) + PTR_MASK) & ~PTR_MASK)\n\n/*\n * Maximum size of L1 cache line. This is used to avoid cache line aliasing.\n * In addition, this controls the spacing of cacheline-spaced size classes.\n *\n * CACHELINE cannot be based on LG_CACHELINE because __declspec(align()) can\n * only handle raw constants.\n */\n#define\tLG_CACHELINE\t\t6\n#define\tCACHELINE\t\t64\n#define\tCACHELINE_MASK\t\t(CACHELINE - 1)\n\n/* Return the smallest cacheline multiple that is >= s. */\n#define\tCACHELINE_CEILING(s)\t\t\t\t\t\t\\\n\t(((s) + CACHELINE_MASK) & ~CACHELINE_MASK)\n\n/* Page size. STATIC_PAGE_SHIFT is determined by the configure script. */\n#ifdef PAGE_MASK\n# undef PAGE_MASK\n#endif\n#define\tLG_PAGE\t\tSTATIC_PAGE_SHIFT\n#define\tPAGE\t\t((size_t)(1U << STATIC_PAGE_SHIFT))\n#define\tPAGE_MASK\t((size_t)(PAGE - 1))\n\n/* Return the smallest pagesize multiple that is >= s. */\n#define\tPAGE_CEILING(s)\t\t\t\t\t\t\t\\\n\t(((s) + PAGE_MASK) & ~PAGE_MASK)\n\n/* Return the nearest aligned address at or below a. */\n#define\tALIGNMENT_ADDR2BASE(a, alignment)\t\t\t\t\\\n\t((void *)((uintptr_t)(a) & (-(alignment))))\n\n/* Return the offset between a and the nearest aligned address at or below a. */\n#define\tALIGNMENT_ADDR2OFFSET(a, alignment)\t\t\t\t\\\n\t((size_t)((uintptr_t)(a) & (alignment - 1)))\n\n/* Return the smallest alignment multiple that is >= s. */\n#define\tALIGNMENT_CEILING(s, alignment)\t\t\t\t\t\\\n\t(((s) + (alignment - 1)) & (-(alignment)))\n\n/* Declare a variable length array */\n#if __STDC_VERSION__ < 199901L\n# ifdef _MSC_VER\n# include \n# define alloca _alloca\n# else\n# include \n# endif\n# define VARIABLE_ARRAY(type, name, count) \\\n\ttype *name = alloca(sizeof(type) * count)\n#else\n# define VARIABLE_ARRAY(type, name, count) type name[count]\n#endif\n\n#ifdef JEMALLOC_VALGRIND\n/*\n * The JEMALLOC_VALGRIND_*() macros must be macros rather than functions\n * so that when Valgrind reports errors, there are no extra stack frames\n * in the backtraces.\n *\n * The size that is reported to valgrind must be consistent through a chain of\n * malloc..realloc..realloc calls. Request size isn't recorded anywhere in\n * jemalloc, so it is critical that all callers of these macros provide usize\n * rather than request size. As a result, buffer overflow detection is\n * technically weakened for the standard API, though it is generally accepted\n * practice to consider any extra bytes reported by malloc_usable_size() as\n * usable space.\n */\n#define\tJEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do {\t\t\\\n\tif (config_valgrind && opt_valgrind && cond)\t\t\t\\\n\t\tVALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(ptr), zero);\t\\\n} while (0)\n#define\tJEMALLOC_VALGRIND_REALLOC(ptr, usize, old_ptr, old_usize,\t\\\n old_rzsize, zero) do {\t\t\t\t\t\t\\\n\tif (config_valgrind && opt_valgrind) {\t\t\t\t\\\n\t\tsize_t rzsize = p2rz(ptr);\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\t\tif (ptr == old_ptr) {\t\t\t\t\t\\\n\t\t\tVALGRIND_RESIZEINPLACE_BLOCK(ptr, old_usize,\t\\\n\t\t\t usize, rzsize);\t\t\t\t\\\n\t\t\tif (zero && old_usize < usize) {\t\t\\\n\t\t\t\tVALGRIND_MAKE_MEM_DEFINED(\t\t\\\n\t\t\t\t (void *)((uintptr_t)ptr +\t\t\\\n\t\t\t\t old_usize), usize - old_usize);\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t\tif (old_ptr != NULL) {\t\t\t\t\\\n\t\t\t\tVALGRIND_FREELIKE_BLOCK(old_ptr,\t\\\n\t\t\t\t old_rzsize);\t\t\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t\tif (ptr != NULL) {\t\t\t\t\\\n\t\t\t\tsize_t copy_size = (old_usize < usize)\t\\\n\t\t\t\t ? old_usize : usize;\t\t\\\n\t\t\t\tsize_t tail_size = usize - copy_size;\t\\\n\t\t\t\tVALGRIND_MALLOCLIKE_BLOCK(ptr, usize,\t\\\n\t\t\t\t rzsize, false);\t\t\t\\\n\t\t\t\tif (copy_size > 0) {\t\t\t\\\n\t\t\t\t\tVALGRIND_MAKE_MEM_DEFINED(ptr,\t\\\n\t\t\t\t\t copy_size);\t\t\t\\\n\t\t\t\t}\t\t\t\t\t\\\n\t\t\t\tif (zero && tail_size > 0) {\t\t\\\n\t\t\t\t\tVALGRIND_MAKE_MEM_DEFINED(\t\\\n\t\t\t\t\t (void *)((uintptr_t)ptr +\t\\\n\t\t\t\t\t copy_size), tail_size);\t\\\n\t\t\t\t}\t\t\t\t\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#define\tJEMALLOC_VALGRIND_FREE(ptr, rzsize) do {\t\t\t\\\n\tif (config_valgrind && opt_valgrind)\t\t\t\t\\\n\t\tVALGRIND_FREELIKE_BLOCK(ptr, rzsize);\t\t\t\\\n} while (0)\n#else\n#define\tRUNNING_ON_VALGRIND\t((unsigned)0)\n#define\tVALGRIND_MALLOCLIKE_BLOCK(addr, sizeB, rzB, is_zeroed)\n#define\tVALGRIND_RESIZEINPLACE_BLOCK(addr, oldSizeB, newSizeB, rzB)\n#define\tVALGRIND_FREELIKE_BLOCK(addr, rzB)\n#define\tVALGRIND_MAKE_MEM_UNDEFINED(_qzz_addr, _qzz_len)\n#define\tVALGRIND_MAKE_MEM_DEFINED(_qzz_addr, _qzz_len)\n#define\tJEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero)\n#define\tJEMALLOC_VALGRIND_REALLOC(ptr, usize, old_ptr, old_usize,\t\\\n old_rzsize, zero)\n#define\tJEMALLOC_VALGRIND_FREE(ptr, rzsize)\n#endif\n\n#include \"jemalloc/internal/util.h\"\n#include \"jemalloc/internal/atomic.h\"\n#include \"jemalloc/internal/prng.h\"\n#include \"jemalloc/internal/ckh.h\"\n#include \"jemalloc/internal/size_classes.h\"\n#include \"jemalloc/internal/stats.h\"\n#include \"jemalloc/internal/ctl.h\"\n#include \"jemalloc/internal/mutex.h\"\n#include \"jemalloc/internal/tsd.h\"\n#include \"jemalloc/internal/mb.h\"\n#include \"jemalloc/internal/extent.h\"\n#include \"jemalloc/internal/arena.h\"\n#include \"jemalloc/internal/bitmap.h\"\n#include \"jemalloc/internal/base.h\"\n#include \"jemalloc/internal/chunk.h\"\n#include \"jemalloc/internal/huge.h\"\n#include \"jemalloc/internal/rtree.h\"\n#include \"jemalloc/internal/tcache.h\"\n#include \"jemalloc/internal/hash.h\"\n#include \"jemalloc/internal/quarantine.h\"\n#include \"jemalloc/internal/prof.h\"\n\n#undef JEMALLOC_H_TYPES\n/******************************************************************************/\n#define JEMALLOC_H_STRUCTS\n\n#include \"jemalloc/internal/util.h\"\n#include \"jemalloc/internal/atomic.h\"\n#include \"jemalloc/internal/prng.h\"\n#include \"jemalloc/internal/ckh.h\"\n#include \"jemalloc/internal/size_classes.h\"\n#include \"jemalloc/internal/stats.h\"\n#include \"jemalloc/internal/ctl.h\"\n#include \"jemalloc/internal/mutex.h\"\n#include \"jemalloc/internal/tsd.h\"\n#include \"jemalloc/internal/mb.h\"\n#include \"jemalloc/internal/bitmap.h\"\n#include \"jemalloc/internal/extent.h\"\n#include \"jemalloc/internal/arena.h\"\n#include \"jemalloc/internal/base.h\"\n#include \"jemalloc/internal/chunk.h\"\n#include \"jemalloc/internal/huge.h\"\n#include \"jemalloc/internal/rtree.h\"\n#include \"jemalloc/internal/tcache.h\"\n#include \"jemalloc/internal/hash.h\"\n#include \"jemalloc/internal/quarantine.h\"\n#include \"jemalloc/internal/prof.h\"\n\ntypedef struct {\n\tuint64_t\tallocated;\n\tuint64_t\tdeallocated;\n} thread_allocated_t;\n/*\n * The JEMALLOC_CONCAT() wrapper is necessary to pass {0, 0} via a cpp macro\n * argument.\n */\n#define\tTHREAD_ALLOCATED_INITIALIZER\tJEMALLOC_CONCAT({0, 0})\n\n#undef JEMALLOC_H_STRUCTS\n/******************************************************************************/\n#define JEMALLOC_H_EXTERNS\n\nextern bool\topt_abort;\nextern bool\topt_junk;\nextern size_t\topt_quarantine;\nextern bool\topt_redzone;\nextern bool\topt_utrace;\nextern bool\topt_valgrind;\nextern bool\topt_xmalloc;\nextern bool\topt_zero;\nextern size_t\topt_narenas;\n\n/* Number of CPUs. */\nextern unsigned\t\tncpus;\n\n/* Protects arenas initialization (arenas, arenas_total). */\nextern malloc_mutex_t\tarenas_lock;\n/*\n * Arenas that are used to service external requests. Not all elements of the\n * arenas array are necessarily used; arenas are created lazily as needed.\n *\n * arenas[0..narenas_auto) are used for automatic multiplexing of threads and\n * arenas. arenas[narenas_auto..narenas_total) are only used if the application\n * takes some action to create them and allocate from them.\n */\nextern arena_t\t\t**arenas;\nextern unsigned\t\tnarenas_total;\nextern unsigned\t\tnarenas_auto; /* Read-only after initialization. */\n\narena_t\t*arenas_extend(unsigned ind);\nvoid\tarenas_cleanup(void *arg);\narena_t\t*choose_arena_hard(void);\nvoid\tjemalloc_prefork(void);\nvoid\tjemalloc_postfork_parent(void);\nvoid\tjemalloc_postfork_child(void);\n\n#include \"jemalloc/internal/util.h\"\n#include \"jemalloc/internal/atomic.h\"\n#include \"jemalloc/internal/prng.h\"\n#include \"jemalloc/internal/ckh.h\"\n#include \"jemalloc/internal/size_classes.h\"\n#include \"jemalloc/internal/stats.h\"\n#include \"jemalloc/internal/ctl.h\"\n#include \"jemalloc/internal/mutex.h\"\n#include \"jemalloc/internal/tsd.h\"\n#include \"jemalloc/internal/mb.h\"\n#include \"jemalloc/internal/bitmap.h\"\n#include \"jemalloc/internal/extent.h\"\n#include \"jemalloc/internal/arena.h\"\n#include \"jemalloc/internal/base.h\"\n#include \"jemalloc/internal/chunk.h\"\n#include \"jemalloc/internal/huge.h\"\n#include \"jemalloc/internal/rtree.h\"\n#include \"jemalloc/internal/tcache.h\"\n#include \"jemalloc/internal/hash.h\"\n#include \"jemalloc/internal/quarantine.h\"\n#include \"jemalloc/internal/prof.h\"\n\n#undef JEMALLOC_H_EXTERNS\n/******************************************************************************/\n#define JEMALLOC_H_INLINES\n\n#include \"jemalloc/internal/util.h\"\n#include \"jemalloc/internal/atomic.h\"\n#include \"jemalloc/internal/prng.h\"\n#include \"jemalloc/internal/ckh.h\"\n#include \"jemalloc/internal/size_classes.h\"\n#include \"jemalloc/internal/stats.h\"\n#include \"jemalloc/internal/ctl.h\"\n#include \"jemalloc/internal/mutex.h\"\n#include \"jemalloc/internal/tsd.h\"\n#include \"jemalloc/internal/mb.h\"\n#include \"jemalloc/internal/extent.h\"\n#include \"jemalloc/internal/base.h\"\n#include \"jemalloc/internal/chunk.h\"\n#include \"jemalloc/internal/huge.h\"\n\n#ifndef JEMALLOC_ENABLE_INLINE\nmalloc_tsd_protos(JEMALLOC_ATTR(unused), arenas, arena_t *)\n\nsize_t\ts2u(size_t size);\nsize_t\tsa2u(size_t size, size_t alignment);\nunsigned\tnarenas_total_get(void);\narena_t\t*choose_arena(arena_t *arena);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))\n/*\n * Map of pthread_self() --> arenas[???], used for selecting an arena to use\n * for allocations.\n */\nmalloc_tsd_externs(arenas, arena_t *)\nmalloc_tsd_funcs(JEMALLOC_INLINE, arenas, arena_t *, NULL, arenas_cleanup)\n\n/*\n * Compute usable size that would result from allocating an object with the\n * specified size.\n */\nJEMALLOC_INLINE size_t\ns2u(size_t size)\n{\n\n\tif (size <= SMALL_MAXCLASS)\n\t\treturn (arena_bin_info[SMALL_SIZE2BIN(size)].reg_size);\n\tif (size <= arena_maxclass)\n\t\treturn (PAGE_CEILING(size));\n\treturn (CHUNK_CEILING(size));\n}\n\n/*\n * Compute usable size that would result from allocating an object with the\n * specified size and alignment.\n */\nJEMALLOC_INLINE size_t\nsa2u(size_t size, size_t alignment)\n{\n\tsize_t usize;\n\n\tassert(alignment != 0 && ((alignment - 1) & alignment) == 0);\n\n\t/*\n\t * Round size up to the nearest multiple of alignment.\n\t *\n\t * This done, we can take advantage of the fact that for each small\n\t * size class, every object is aligned at the smallest power of two\n\t * that is non-zero in the base two representation of the size. For\n\t * example:\n\t *\n\t * Size | Base 2 | Minimum alignment\n\t * -----+----------+------------------\n\t * 96 | 1100000 | 32\n\t * 144 | 10100000 | 32\n\t * 192 | 11000000 | 64\n\t */\n\tusize = ALIGNMENT_CEILING(size, alignment);\n\t/*\n\t * (usize < size) protects against the combination of maximal\n\t * alignment and size greater than maximal alignment.\n\t */\n\tif (usize < size) {\n\t\t/* size_t overflow. */\n\t\treturn (0);\n\t}\n\n\tif (usize <= arena_maxclass && alignment <= PAGE) {\n\t\tif (usize <= SMALL_MAXCLASS)\n\t\t\treturn (arena_bin_info[SMALL_SIZE2BIN(usize)].reg_size);\n\t\treturn (PAGE_CEILING(usize));\n\t} else {\n\t\tsize_t run_size;\n\n\t\t/*\n\t\t * We can't achieve subpage alignment, so round up alignment\n\t\t * permanently; it makes later calculations simpler.\n\t\t */\n\t\talignment = PAGE_CEILING(alignment);\n\t\tusize = PAGE_CEILING(size);\n\t\t/*\n\t\t * (usize < size) protects against very large sizes within\n\t\t * PAGE of SIZE_T_MAX.\n\t\t *\n\t\t * (usize + alignment < usize) protects against the\n\t\t * combination of maximal alignment and usize large enough\n\t\t * to cause overflow. This is similar to the first overflow\n\t\t * check above, but it needs to be repeated due to the new\n\t\t * usize value, which may now be *equal* to maximal\n\t\t * alignment, whereas before we only detected overflow if the\n\t\t * original size was *greater* than maximal alignment.\n\t\t */\n\t\tif (usize < size || usize + alignment < usize) {\n\t\t\t/* size_t overflow. */\n\t\t\treturn (0);\n\t\t}\n\n\t\t/*\n\t\t * Calculate the size of the over-size run that arena_palloc()\n\t\t * would need to allocate in order to guarantee the alignment.\n\t\t * If the run wouldn't fit within a chunk, round up to a huge\n\t\t * allocation size.\n\t\t */\n\t\trun_size = usize + alignment - PAGE;\n\t\tif (run_size <= arena_maxclass)\n\t\t\treturn (PAGE_CEILING(usize));\n\t\treturn (CHUNK_CEILING(usize));\n\t}\n}\n\nJEMALLOC_INLINE unsigned\nnarenas_total_get(void)\n{\n\tunsigned narenas;\n\n\tmalloc_mutex_lock(&arenas_lock);\n\tnarenas = narenas_total;\n\tmalloc_mutex_unlock(&arenas_lock);\n\n\treturn (narenas);\n}\n\n/* Choose an arena based on a per-thread value. */\nJEMALLOC_INLINE arena_t *\nchoose_arena(arena_t *arena)\n{\n\tarena_t *ret;\n\n\tif (arena != NULL)\n\t\treturn (arena);\n\n\tif ((ret = *arenas_tsd_get()) == NULL) {\n\t\tret = choose_arena_hard();\n\t\tassert(ret != NULL);\n\t}\n\n\treturn (ret);\n}\n#endif\n\n#include \"jemalloc/internal/bitmap.h\"\n#include \"jemalloc/internal/rtree.h\"\n/*\n * Include arena.h twice in order to resolve circular dependencies with\n * tcache.h.\n */\n#define\tJEMALLOC_ARENA_INLINE_A\n#include \"jemalloc/internal/arena.h\"\n#undef JEMALLOC_ARENA_INLINE_A\n#include \"jemalloc/internal/tcache.h\"\n#define\tJEMALLOC_ARENA_INLINE_B\n#include \"jemalloc/internal/arena.h\"\n#undef JEMALLOC_ARENA_INLINE_B\n#include \"jemalloc/internal/hash.h\"\n#include \"jemalloc/internal/quarantine.h\"\n\n#ifndef JEMALLOC_ENABLE_INLINE\nvoid\t*imallocx(size_t size, bool try_tcache, arena_t *arena);\nvoid\t*imalloc(size_t size);\nvoid\t*icallocx(size_t size, bool try_tcache, arena_t *arena);\nvoid\t*icalloc(size_t size);\nvoid\t*ipallocx(size_t usize, size_t alignment, bool zero, bool try_tcache,\n arena_t *arena);\nvoid\t*ipalloc(size_t usize, size_t alignment, bool zero);\nsize_t\tisalloc(const void *ptr, bool demote);\nsize_t\tivsalloc(const void *ptr, bool demote);\nsize_t\tu2rz(size_t usize);\nsize_t\tp2rz(const void *ptr);\nvoid\tidallocx(void *ptr, bool try_tcache);\nvoid\tidalloc(void *ptr);\nvoid\tiqallocx(void *ptr, bool try_tcache);\nvoid\tiqalloc(void *ptr);\nvoid\t*irallocx(void *ptr, size_t size, size_t extra, size_t alignment,\n bool zero, bool no_move, bool try_tcache_alloc, bool try_tcache_dalloc,\n arena_t *arena);\nvoid\t*iralloc(void *ptr, size_t size, size_t extra, size_t alignment,\n bool zero, bool no_move);\nmalloc_tsd_protos(JEMALLOC_ATTR(unused), thread_allocated, thread_allocated_t)\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))\nJEMALLOC_INLINE void *\nimallocx(size_t size, bool try_tcache, arena_t *arena)\n{\n\n\tassert(size != 0);\n\n\tif (size <= arena_maxclass)\n\t\treturn (arena_malloc(arena, size, false, try_tcache));\n\telse\n\t\treturn (huge_malloc(size, false));\n}\n\nJEMALLOC_INLINE void *\nimalloc(size_t size)\n{\n\n\treturn (imallocx(size, true, NULL));\n}\n\nJEMALLOC_INLINE void *\nicallocx(size_t size, bool try_tcache, arena_t *arena)\n{\n\n\tif (size <= arena_maxclass)\n\t\treturn (arena_malloc(arena, size, true, try_tcache));\n\telse\n\t\treturn (huge_malloc(size, true));\n}\n\nJEMALLOC_INLINE void *\nicalloc(size_t size)\n{\n\n\treturn (icallocx(size, true, NULL));\n}\n\nJEMALLOC_INLINE void *\nipallocx(size_t usize, size_t alignment, bool zero, bool try_tcache,\n arena_t *arena)\n{\n\tvoid *ret;\n\n\tassert(usize != 0);\n\tassert(usize == sa2u(usize, alignment));\n\n\tif (usize <= arena_maxclass && alignment <= PAGE)\n\t\tret = arena_malloc(arena, usize, zero, try_tcache);\n\telse {\n\t\tif (usize <= arena_maxclass) {\n\t\t\tret = arena_palloc(choose_arena(arena), usize,\n\t\t\t alignment, zero);\n\t\t} else if (alignment <= chunksize)\n\t\t\tret = huge_malloc(usize, zero);\n\t\telse\n\t\t\tret = huge_palloc(usize, alignment, zero);\n\t}\n\n\tassert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void *\nipalloc(size_t usize, size_t alignment, bool zero)\n{\n\n\treturn (ipallocx(usize, alignment, zero, true, NULL));\n}\n\n/*\n * Typical usage:\n * void *ptr = [...]\n * size_t sz = isalloc(ptr, config_prof);\n */\nJEMALLOC_INLINE size_t\nisalloc(const void *ptr, bool demote)\n{\n\tsize_t ret;\n\tarena_chunk_t *chunk;\n\n\tassert(ptr != NULL);\n\t/* Demotion only makes sense if config_prof is true. */\n\tassert(config_prof || demote == false);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tif (chunk != ptr)\n\t\tret = arena_salloc(ptr, demote);\n\telse\n\t\tret = huge_salloc(ptr);\n\n\treturn (ret);\n}\n\nJEMALLOC_INLINE size_t\nivsalloc(const void *ptr, bool demote)\n{\n\n\t/* Return 0 if ptr is not within a chunk managed by jemalloc. */\n\tif (rtree_get(chunks_rtree, (uintptr_t)CHUNK_ADDR2BASE(ptr)) == NULL)\n\t\treturn (0);\n\n\treturn (isalloc(ptr, demote));\n}\n\nJEMALLOC_INLINE size_t\nu2rz(size_t usize)\n{\n\tsize_t ret;\n\n\tif (usize <= SMALL_MAXCLASS) {\n\t\tsize_t binind = SMALL_SIZE2BIN(usize);\n\t\tret = arena_bin_info[binind].redzone_size;\n\t} else\n\t\tret = 0;\n\n\treturn (ret);\n}\n\nJEMALLOC_INLINE size_t\np2rz(const void *ptr)\n{\n\tsize_t usize = isalloc(ptr, false);\n\n\treturn (u2rz(usize));\n}\n\nJEMALLOC_INLINE void\nidallocx(void *ptr, bool try_tcache)\n{\n\tarena_chunk_t *chunk;\n\n\tassert(ptr != NULL);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tif (chunk != ptr)\n\t\tarena_dalloc(chunk->arena, chunk, ptr, try_tcache);\n\telse\n\t\thuge_dalloc(ptr, true);\n}\n\nJEMALLOC_INLINE void\nidalloc(void *ptr)\n{\n\n\tidallocx(ptr, true);\n}\n\nJEMALLOC_INLINE void\niqallocx(void *ptr, bool try_tcache)\n{\n\n\tif (config_fill && opt_quarantine)\n\t\tquarantine(ptr);\n\telse\n\t\tidallocx(ptr, try_tcache);\n}\n\nJEMALLOC_INLINE void\niqalloc(void *ptr)\n{\n\n\tiqallocx(ptr, true);\n}\n\nJEMALLOC_INLINE void *\nirallocx(void *ptr, size_t size, size_t extra, size_t alignment, bool zero,\n bool no_move, bool try_tcache_alloc, bool try_tcache_dalloc, arena_t *arena)\n{\n\tvoid *ret;\n\tsize_t oldsize;\n\n\tassert(ptr != NULL);\n\tassert(size != 0);\n\n\toldsize = isalloc(ptr, config_prof);\n\n\tif (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))\n\t != 0) {\n\t\tsize_t usize, copysize;\n\n\t\t/*\n\t\t * Existing object alignment is inadequate; allocate new space\n\t\t * and copy.\n\t\t */\n\t\tif (no_move)\n\t\t\treturn (NULL);\n\t\tusize = sa2u(size + extra, alignment);\n\t\tif (usize == 0)\n\t\t\treturn (NULL);\n\t\tret = ipallocx(usize, alignment, zero, try_tcache_alloc, arena);\n\t\tif (ret == NULL) {\n\t\t\tif (extra == 0)\n\t\t\t\treturn (NULL);\n\t\t\t/* Try again, without extra this time. */\n\t\t\tusize = sa2u(size, alignment);\n\t\t\tif (usize == 0)\n\t\t\t\treturn (NULL);\n\t\t\tret = ipallocx(usize, alignment, zero, try_tcache_alloc,\n\t\t\t arena);\n\t\t\tif (ret == NULL)\n\t\t\t\treturn (NULL);\n\t\t}\n\t\t/*\n\t\t * Copy at most size bytes (not size+extra), since the caller\n\t\t * has no expectation that the extra bytes will be reliably\n\t\t * preserved.\n\t\t */\n\t\tcopysize = (size < oldsize) ? size : oldsize;\n\t\tmemcpy(ret, ptr, copysize);\n\t\tiqallocx(ptr, try_tcache_dalloc);\n\t\treturn (ret);\n\t}\n\n\tif (no_move) {\n\t\tif (size <= arena_maxclass) {\n\t\t\treturn (arena_ralloc_no_move(ptr, oldsize, size,\n\t\t\t extra, zero));\n\t\t} else {\n\t\t\treturn (huge_ralloc_no_move(ptr, oldsize, size,\n\t\t\t extra));\n\t\t}\n\t} else {\n\t\tif (size + extra <= arena_maxclass) {\n\t\t\treturn (arena_ralloc(arena, ptr, oldsize, size, extra,\n\t\t\t alignment, zero, try_tcache_alloc,\n\t\t\t try_tcache_dalloc));\n\t\t} else {\n\t\t\treturn (huge_ralloc(ptr, oldsize, size, extra,\n\t\t\t alignment, zero, try_tcache_dalloc));\n\t\t}\n\t}\n}\n\nJEMALLOC_INLINE void *\niralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero,\n bool no_move)\n{\n\n\treturn (irallocx(ptr, size, extra, alignment, zero, no_move, true, true,\n\t NULL));\n}\n\nmalloc_tsd_externs(thread_allocated, thread_allocated_t)\nmalloc_tsd_funcs(JEMALLOC_INLINE, thread_allocated, thread_allocated_t,\n THREAD_ALLOCATED_INITIALIZER, malloc_tsd_no_cleanup)\n#endif\n\n#include \"jemalloc/internal/prof.h\"\n\n#undef JEMALLOC_H_INLINES\n/******************************************************************************/\n#endif /* JEMALLOC_INTERNAL_H */\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/mb.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nvoid\tmb_write(void);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MB_C_))\n#ifdef __i386__\n/*\n * According to the Intel Architecture Software Developer's Manual, current\n * processors execute instructions in order from the perspective of other\n * processors in a multiprocessor system, but 1) Intel reserves the right to\n * change that, and 2) the compiler's optimizer could re-order instructions if\n * there weren't some form of barrier. Therefore, even if running on an\n * architecture that does not need memory barriers (everything through at least\n * i686), an \"optimizer barrier\" is necessary.\n */\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\n# if 0\n\t/* This is a true memory barrier. */\n\tasm volatile (\"pusha;\"\n\t \"xor %%eax,%%eax;\"\n\t \"cpuid;\"\n\t \"popa;\"\n\t : /* Outputs. */\n\t : /* Inputs. */\n\t : \"memory\" /* Clobbers. */\n\t );\n#else\n\t/*\n\t * This is hopefully enough to keep the compiler from reordering\n\t * instructions around this one.\n\t */\n\tasm volatile (\"nop;\"\n\t : /* Outputs. */\n\t : /* Inputs. */\n\t : \"memory\" /* Clobbers. */\n\t );\n#endif\n}\n#elif (defined(__amd64__) || defined(__x86_64__))\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\n\tasm volatile (\"sfence\"\n\t : /* Outputs. */\n\t : /* Inputs. */\n\t : \"memory\" /* Clobbers. */\n\t );\n}\n#elif defined(__powerpc__)\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\n\tasm volatile (\"eieio\"\n\t : /* Outputs. */\n\t : /* Inputs. */\n\t : \"memory\" /* Clobbers. */\n\t );\n}\n#elif defined(__sparc64__)\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\n\tasm volatile (\"membar #StoreStore\"\n\t : /* Outputs. */\n\t : /* Inputs. */\n\t : \"memory\" /* Clobbers. */\n\t );\n}\n#elif defined(__tile__)\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\n\t__sync_synchronize();\n}\n#else\n/*\n * This is much slower than a simple memory barrier, but the semantics of mutex\n * unlock make this work.\n */\nJEMALLOC_INLINE void\nmb_write(void)\n{\n\tmalloc_mutex_t mtx;\n\n\tmalloc_mutex_init(&mtx);\n\tmalloc_mutex_lock(&mtx);\n\tmalloc_mutex_unlock(&mtx);\n}\n#endif\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/mutex.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct malloc_mutex_s malloc_mutex_t;\n\n#ifdef _WIN32\n# define MALLOC_MUTEX_INITIALIZER\n#elif (defined(JEMALLOC_OSSPIN))\n# define MALLOC_MUTEX_INITIALIZER {0}\n#elif (defined(JEMALLOC_MUTEX_INIT_CB))\n# define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL}\n#else\n# if (defined(PTHREAD_MUTEX_ADAPTIVE_NP) &&\t\t\t\t\\\n defined(PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP))\n# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_ADAPTIVE_NP\n# define MALLOC_MUTEX_INITIALIZER {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP}\n# else\n# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT\n# define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER}\n# endif\n#endif\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct malloc_mutex_s {\n#ifdef _WIN32\n\tCRITICAL_SECTION\tlock;\n#elif (defined(JEMALLOC_OSSPIN))\n\tOSSpinLock\t\tlock;\n#elif (defined(JEMALLOC_MUTEX_INIT_CB))\n\tpthread_mutex_t\t\tlock;\n\tmalloc_mutex_t\t\t*postponed_next;\n#else\n\tpthread_mutex_t\t\tlock;\n#endif\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n#ifdef JEMALLOC_LAZY_LOCK\nextern bool isthreaded;\n#else\n# undef isthreaded /* Undo private_namespace.h definition. */\n# define isthreaded true\n#endif\n\nbool\tmalloc_mutex_init(malloc_mutex_t *mutex);\nvoid\tmalloc_mutex_prefork(malloc_mutex_t *mutex);\nvoid\tmalloc_mutex_postfork_parent(malloc_mutex_t *mutex);\nvoid\tmalloc_mutex_postfork_child(malloc_mutex_t *mutex);\nbool\tmutex_boot(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nvoid\tmalloc_mutex_lock(malloc_mutex_t *mutex);\nvoid\tmalloc_mutex_unlock(malloc_mutex_t *mutex);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MUTEX_C_))\nJEMALLOC_INLINE void\nmalloc_mutex_lock(malloc_mutex_t *mutex)\n{\n\n\tif (isthreaded) {\n#ifdef _WIN32\n\t\tEnterCriticalSection(&mutex->lock);\n#elif (defined(JEMALLOC_OSSPIN))\n\t\tOSSpinLockLock(&mutex->lock);\n#else\n\t\tpthread_mutex_lock(&mutex->lock);\n#endif\n\t}\n}\n\nJEMALLOC_INLINE void\nmalloc_mutex_unlock(malloc_mutex_t *mutex)\n{\n\n\tif (isthreaded) {\n#ifdef _WIN32\n\t\tLeaveCriticalSection(&mutex->lock);\n#elif (defined(JEMALLOC_OSSPIN))\n\t\tOSSpinLockUnlock(&mutex->lock);\n#else\n\t\tpthread_mutex_unlock(&mutex->lock);\n#endif\n\t}\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/private_namespace.h", "content": "#define\ta0calloc JEMALLOC_N(a0calloc)\n#define\ta0free JEMALLOC_N(a0free)\n#define\ta0malloc JEMALLOC_N(a0malloc)\n#define\tarena_alloc_junk_small JEMALLOC_N(arena_alloc_junk_small)\n#define\tarena_bin_index JEMALLOC_N(arena_bin_index)\n#define\tarena_bin_info JEMALLOC_N(arena_bin_info)\n#define\tarena_boot JEMALLOC_N(arena_boot)\n#define\tarena_dalloc JEMALLOC_N(arena_dalloc)\n#define\tarena_dalloc_bin JEMALLOC_N(arena_dalloc_bin)\n#define\tarena_dalloc_bin_locked JEMALLOC_N(arena_dalloc_bin_locked)\n#define\tarena_dalloc_junk_small JEMALLOC_N(arena_dalloc_junk_small)\n#define\tarena_dalloc_large JEMALLOC_N(arena_dalloc_large)\n#define\tarena_dalloc_large_locked JEMALLOC_N(arena_dalloc_large_locked)\n#define\tarena_dalloc_small JEMALLOC_N(arena_dalloc_small)\n#define\tarena_dss_prec_get JEMALLOC_N(arena_dss_prec_get)\n#define\tarena_dss_prec_set JEMALLOC_N(arena_dss_prec_set)\n#define\tarena_malloc JEMALLOC_N(arena_malloc)\n#define\tarena_malloc_large JEMALLOC_N(arena_malloc_large)\n#define\tarena_malloc_small JEMALLOC_N(arena_malloc_small)\n#define\tarena_mapbits_allocated_get JEMALLOC_N(arena_mapbits_allocated_get)\n#define\tarena_mapbits_binind_get JEMALLOC_N(arena_mapbits_binind_get)\n#define\tarena_mapbits_dirty_get JEMALLOC_N(arena_mapbits_dirty_get)\n#define\tarena_mapbits_get JEMALLOC_N(arena_mapbits_get)\n#define\tarena_mapbits_large_binind_set JEMALLOC_N(arena_mapbits_large_binind_set)\n#define\tarena_mapbits_large_get JEMALLOC_N(arena_mapbits_large_get)\n#define\tarena_mapbits_large_set JEMALLOC_N(arena_mapbits_large_set)\n#define\tarena_mapbits_large_size_get JEMALLOC_N(arena_mapbits_large_size_get)\n#define\tarena_mapbits_small_runind_get JEMALLOC_N(arena_mapbits_small_runind_get)\n#define\tarena_mapbits_small_set JEMALLOC_N(arena_mapbits_small_set)\n#define\tarena_mapbits_unallocated_set JEMALLOC_N(arena_mapbits_unallocated_set)\n#define\tarena_mapbits_unallocated_size_get JEMALLOC_N(arena_mapbits_unallocated_size_get)\n#define\tarena_mapbits_unallocated_size_set JEMALLOC_N(arena_mapbits_unallocated_size_set)\n#define\tarena_mapbits_unzeroed_get JEMALLOC_N(arena_mapbits_unzeroed_get)\n#define\tarena_mapbits_unzeroed_set JEMALLOC_N(arena_mapbits_unzeroed_set)\n#define\tarena_mapbitsp_get JEMALLOC_N(arena_mapbitsp_get)\n#define\tarena_mapp_get JEMALLOC_N(arena_mapp_get)\n#define\tarena_maxclass JEMALLOC_N(arena_maxclass)\n#define\tarena_new JEMALLOC_N(arena_new)\n#define\tarena_palloc JEMALLOC_N(arena_palloc)\n#define\tarena_postfork_child JEMALLOC_N(arena_postfork_child)\n#define\tarena_postfork_parent JEMALLOC_N(arena_postfork_parent)\n#define\tarena_prefork JEMALLOC_N(arena_prefork)\n#define\tarena_prof_accum JEMALLOC_N(arena_prof_accum)\n#define\tarena_prof_ctx_get JEMALLOC_N(arena_prof_ctx_get)\n#define\tarena_prof_ctx_set JEMALLOC_N(arena_prof_ctx_set)\n#define\tarena_prof_promoted JEMALLOC_N(arena_prof_promoted)\n#define\tarena_ptr_small_binind_get JEMALLOC_N(arena_ptr_small_binind_get)\n#define\tarena_purge_all JEMALLOC_N(arena_purge_all)\n#define\tarena_ralloc JEMALLOC_N(arena_ralloc)\n#define\tarena_ralloc_no_move JEMALLOC_N(arena_ralloc_no_move)\n#define\tarena_run_regind JEMALLOC_N(arena_run_regind)\n#define\tarena_salloc JEMALLOC_N(arena_salloc)\n#define\tarena_stats_merge JEMALLOC_N(arena_stats_merge)\n#define\tarena_tcache_fill_small JEMALLOC_N(arena_tcache_fill_small)\n#define\tarenas JEMALLOC_N(arenas)\n#define\tarenas_booted JEMALLOC_N(arenas_booted)\n#define\tarenas_cleanup JEMALLOC_N(arenas_cleanup)\n#define\tarenas_extend JEMALLOC_N(arenas_extend)\n#define\tarenas_initialized JEMALLOC_N(arenas_initialized)\n#define\tarenas_lock JEMALLOC_N(arenas_lock)\n#define\tarenas_tls JEMALLOC_N(arenas_tls)\n#define\tarenas_tsd JEMALLOC_N(arenas_tsd)\n#define\tarenas_tsd_boot JEMALLOC_N(arenas_tsd_boot)\n#define\tarenas_tsd_cleanup_wrapper JEMALLOC_N(arenas_tsd_cleanup_wrapper)\n#define\tarenas_tsd_get JEMALLOC_N(arenas_tsd_get)\n#define\tarenas_tsd_set JEMALLOC_N(arenas_tsd_set)\n#define\tatomic_add_u JEMALLOC_N(atomic_add_u)\n#define\tatomic_add_uint32 JEMALLOC_N(atomic_add_uint32)\n#define\tatomic_add_uint64 JEMALLOC_N(atomic_add_uint64)\n#define\tatomic_add_z JEMALLOC_N(atomic_add_z)\n#define\tatomic_sub_u JEMALLOC_N(atomic_sub_u)\n#define\tatomic_sub_uint32 JEMALLOC_N(atomic_sub_uint32)\n#define\tatomic_sub_uint64 JEMALLOC_N(atomic_sub_uint64)\n#define\tatomic_sub_z JEMALLOC_N(atomic_sub_z)\n#define\tbase_alloc JEMALLOC_N(base_alloc)\n#define\tbase_boot JEMALLOC_N(base_boot)\n#define\tbase_calloc JEMALLOC_N(base_calloc)\n#define\tbase_node_alloc JEMALLOC_N(base_node_alloc)\n#define\tbase_node_dealloc JEMALLOC_N(base_node_dealloc)\n#define\tbase_postfork_child JEMALLOC_N(base_postfork_child)\n#define\tbase_postfork_parent JEMALLOC_N(base_postfork_parent)\n#define\tbase_prefork JEMALLOC_N(base_prefork)\n#define\tbitmap_full JEMALLOC_N(bitmap_full)\n#define\tbitmap_get JEMALLOC_N(bitmap_get)\n#define\tbitmap_info_init JEMALLOC_N(bitmap_info_init)\n#define\tbitmap_info_ngroups JEMALLOC_N(bitmap_info_ngroups)\n#define\tbitmap_init JEMALLOC_N(bitmap_init)\n#define\tbitmap_set JEMALLOC_N(bitmap_set)\n#define\tbitmap_sfu JEMALLOC_N(bitmap_sfu)\n#define\tbitmap_size JEMALLOC_N(bitmap_size)\n#define\tbitmap_unset JEMALLOC_N(bitmap_unset)\n#define\tbt_init JEMALLOC_N(bt_init)\n#define\tbuferror JEMALLOC_N(buferror)\n#define\tchoose_arena JEMALLOC_N(choose_arena)\n#define\tchoose_arena_hard JEMALLOC_N(choose_arena_hard)\n#define\tchunk_alloc JEMALLOC_N(chunk_alloc)\n#define\tchunk_alloc_dss JEMALLOC_N(chunk_alloc_dss)\n#define\tchunk_alloc_mmap JEMALLOC_N(chunk_alloc_mmap)\n#define\tchunk_boot JEMALLOC_N(chunk_boot)\n#define\tchunk_dealloc JEMALLOC_N(chunk_dealloc)\n#define\tchunk_dealloc_mmap JEMALLOC_N(chunk_dealloc_mmap)\n#define\tchunk_dss_boot JEMALLOC_N(chunk_dss_boot)\n#define\tchunk_dss_postfork_child JEMALLOC_N(chunk_dss_postfork_child)\n#define\tchunk_dss_postfork_parent JEMALLOC_N(chunk_dss_postfork_parent)\n#define\tchunk_dss_prec_get JEMALLOC_N(chunk_dss_prec_get)\n#define\tchunk_dss_prec_set JEMALLOC_N(chunk_dss_prec_set)\n#define\tchunk_dss_prefork JEMALLOC_N(chunk_dss_prefork)\n#define\tchunk_in_dss JEMALLOC_N(chunk_in_dss)\n#define\tchunk_npages JEMALLOC_N(chunk_npages)\n#define\tchunk_postfork_child JEMALLOC_N(chunk_postfork_child)\n#define\tchunk_postfork_parent JEMALLOC_N(chunk_postfork_parent)\n#define\tchunk_prefork JEMALLOC_N(chunk_prefork)\n#define\tchunk_unmap JEMALLOC_N(chunk_unmap)\n#define\tchunks_mtx JEMALLOC_N(chunks_mtx)\n#define\tchunks_rtree JEMALLOC_N(chunks_rtree)\n#define\tchunksize JEMALLOC_N(chunksize)\n#define\tchunksize_mask JEMALLOC_N(chunksize_mask)\n#define\tckh_bucket_search JEMALLOC_N(ckh_bucket_search)\n#define\tckh_count JEMALLOC_N(ckh_count)\n#define\tckh_delete JEMALLOC_N(ckh_delete)\n#define\tckh_evict_reloc_insert JEMALLOC_N(ckh_evict_reloc_insert)\n#define\tckh_insert JEMALLOC_N(ckh_insert)\n#define\tckh_isearch JEMALLOC_N(ckh_isearch)\n#define\tckh_iter JEMALLOC_N(ckh_iter)\n#define\tckh_new JEMALLOC_N(ckh_new)\n#define\tckh_pointer_hash JEMALLOC_N(ckh_pointer_hash)\n#define\tckh_pointer_keycomp JEMALLOC_N(ckh_pointer_keycomp)\n#define\tckh_rebuild JEMALLOC_N(ckh_rebuild)\n#define\tckh_remove JEMALLOC_N(ckh_remove)\n#define\tckh_search JEMALLOC_N(ckh_search)\n#define\tckh_string_hash JEMALLOC_N(ckh_string_hash)\n#define\tckh_string_keycomp JEMALLOC_N(ckh_string_keycomp)\n#define\tckh_try_bucket_insert JEMALLOC_N(ckh_try_bucket_insert)\n#define\tckh_try_insert JEMALLOC_N(ckh_try_insert)\n#define\tctl_boot JEMALLOC_N(ctl_boot)\n#define\tctl_bymib JEMALLOC_N(ctl_bymib)\n#define\tctl_byname JEMALLOC_N(ctl_byname)\n#define\tctl_nametomib JEMALLOC_N(ctl_nametomib)\n#define\tctl_postfork_child JEMALLOC_N(ctl_postfork_child)\n#define\tctl_postfork_parent JEMALLOC_N(ctl_postfork_parent)\n#define\tctl_prefork JEMALLOC_N(ctl_prefork)\n#define\tdss_prec_names JEMALLOC_N(dss_prec_names)\n#define\textent_tree_ad_first JEMALLOC_N(extent_tree_ad_first)\n#define\textent_tree_ad_insert JEMALLOC_N(extent_tree_ad_insert)\n#define\textent_tree_ad_iter JEMALLOC_N(extent_tree_ad_iter)\n#define\textent_tree_ad_iter_recurse JEMALLOC_N(extent_tree_ad_iter_recurse)\n#define\textent_tree_ad_iter_start JEMALLOC_N(extent_tree_ad_iter_start)\n#define\textent_tree_ad_last JEMALLOC_N(extent_tree_ad_last)\n#define\textent_tree_ad_new JEMALLOC_N(extent_tree_ad_new)\n#define\textent_tree_ad_next JEMALLOC_N(extent_tree_ad_next)\n#define\textent_tree_ad_nsearch JEMALLOC_N(extent_tree_ad_nsearch)\n#define\textent_tree_ad_prev JEMALLOC_N(extent_tree_ad_prev)\n#define\textent_tree_ad_psearch JEMALLOC_N(extent_tree_ad_psearch)\n#define\textent_tree_ad_remove JEMALLOC_N(extent_tree_ad_remove)\n#define\textent_tree_ad_reverse_iter JEMALLOC_N(extent_tree_ad_reverse_iter)\n#define\textent_tree_ad_reverse_iter_recurse JEMALLOC_N(extent_tree_ad_reverse_iter_recurse)\n#define\textent_tree_ad_reverse_iter_start JEMALLOC_N(extent_tree_ad_reverse_iter_start)\n#define\textent_tree_ad_search JEMALLOC_N(extent_tree_ad_search)\n#define\textent_tree_szad_first JEMALLOC_N(extent_tree_szad_first)\n#define\textent_tree_szad_insert JEMALLOC_N(extent_tree_szad_insert)\n#define\textent_tree_szad_iter JEMALLOC_N(extent_tree_szad_iter)\n#define\textent_tree_szad_iter_recurse JEMALLOC_N(extent_tree_szad_iter_recurse)\n#define\textent_tree_szad_iter_start JEMALLOC_N(extent_tree_szad_iter_start)\n#define\textent_tree_szad_last JEMALLOC_N(extent_tree_szad_last)\n#define\textent_tree_szad_new JEMALLOC_N(extent_tree_szad_new)\n#define\textent_tree_szad_next JEMALLOC_N(extent_tree_szad_next)\n#define\textent_tree_szad_nsearch JEMALLOC_N(extent_tree_szad_nsearch)\n#define\textent_tree_szad_prev JEMALLOC_N(extent_tree_szad_prev)\n#define\textent_tree_szad_psearch JEMALLOC_N(extent_tree_szad_psearch)\n#define\textent_tree_szad_remove JEMALLOC_N(extent_tree_szad_remove)\n#define\textent_tree_szad_reverse_iter JEMALLOC_N(extent_tree_szad_reverse_iter)\n#define\textent_tree_szad_reverse_iter_recurse JEMALLOC_N(extent_tree_szad_reverse_iter_recurse)\n#define\textent_tree_szad_reverse_iter_start JEMALLOC_N(extent_tree_szad_reverse_iter_start)\n#define\textent_tree_szad_search JEMALLOC_N(extent_tree_szad_search)\n#define\tget_errno JEMALLOC_N(get_errno)\n#define\thash JEMALLOC_N(hash)\n#define\thuge_allocated JEMALLOC_N(huge_allocated)\n#define\thuge_boot JEMALLOC_N(huge_boot)\n#define\thuge_dalloc JEMALLOC_N(huge_dalloc)\n#define\thuge_malloc JEMALLOC_N(huge_malloc)\n#define\thuge_mtx JEMALLOC_N(huge_mtx)\n#define\thuge_ndalloc JEMALLOC_N(huge_ndalloc)\n#define\thuge_nmalloc JEMALLOC_N(huge_nmalloc)\n#define\thuge_palloc JEMALLOC_N(huge_palloc)\n#define\thuge_postfork_child JEMALLOC_N(huge_postfork_child)\n#define\thuge_postfork_parent JEMALLOC_N(huge_postfork_parent)\n#define\thuge_prefork JEMALLOC_N(huge_prefork)\n#define\thuge_prof_ctx_get JEMALLOC_N(huge_prof_ctx_get)\n#define\thuge_prof_ctx_set JEMALLOC_N(huge_prof_ctx_set)\n#define\thuge_ralloc JEMALLOC_N(huge_ralloc)\n#define\thuge_ralloc_no_move JEMALLOC_N(huge_ralloc_no_move)\n#define\thuge_salloc JEMALLOC_N(huge_salloc)\n#define\tiallocm JEMALLOC_N(iallocm)\n#define\ticalloc JEMALLOC_N(icalloc)\n#define\ticallocx JEMALLOC_N(icallocx)\n#define\tidalloc JEMALLOC_N(idalloc)\n#define\tidallocx JEMALLOC_N(idallocx)\n#define\timalloc JEMALLOC_N(imalloc)\n#define\timallocx JEMALLOC_N(imallocx)\n#define\tipalloc JEMALLOC_N(ipalloc)\n#define\tipallocx JEMALLOC_N(ipallocx)\n#define\tiqalloc JEMALLOC_N(iqalloc)\n#define\tiqallocx JEMALLOC_N(iqallocx)\n#define\tiralloc JEMALLOC_N(iralloc)\n#define\tirallocx JEMALLOC_N(irallocx)\n#define\tisalloc JEMALLOC_N(isalloc)\n#define\tisthreaded JEMALLOC_N(isthreaded)\n#define\tivsalloc JEMALLOC_N(ivsalloc)\n#define\tjemalloc_postfork_child JEMALLOC_N(jemalloc_postfork_child)\n#define\tjemalloc_postfork_parent JEMALLOC_N(jemalloc_postfork_parent)\n#define\tjemalloc_prefork JEMALLOC_N(jemalloc_prefork)\n#define\tmalloc_cprintf JEMALLOC_N(malloc_cprintf)\n#define\tmalloc_mutex_init JEMALLOC_N(malloc_mutex_init)\n#define\tmalloc_mutex_lock JEMALLOC_N(malloc_mutex_lock)\n#define\tmalloc_mutex_postfork_child JEMALLOC_N(malloc_mutex_postfork_child)\n#define\tmalloc_mutex_postfork_parent JEMALLOC_N(malloc_mutex_postfork_parent)\n#define\tmalloc_mutex_prefork JEMALLOC_N(malloc_mutex_prefork)\n#define\tmalloc_mutex_unlock JEMALLOC_N(malloc_mutex_unlock)\n#define\tmalloc_printf JEMALLOC_N(malloc_printf)\n#define\tmalloc_snprintf JEMALLOC_N(malloc_snprintf)\n#define\tmalloc_strtoumax JEMALLOC_N(malloc_strtoumax)\n#define\tmalloc_tsd_boot JEMALLOC_N(malloc_tsd_boot)\n#define\tmalloc_tsd_cleanup_register JEMALLOC_N(malloc_tsd_cleanup_register)\n#define\tmalloc_tsd_dalloc JEMALLOC_N(malloc_tsd_dalloc)\n#define\tmalloc_tsd_malloc JEMALLOC_N(malloc_tsd_malloc)\n#define\tmalloc_tsd_no_cleanup JEMALLOC_N(malloc_tsd_no_cleanup)\n#define\tmalloc_vcprintf JEMALLOC_N(malloc_vcprintf)\n#define\tmalloc_vsnprintf JEMALLOC_N(malloc_vsnprintf)\n#define\tmalloc_write JEMALLOC_N(malloc_write)\n#define\tmap_bias JEMALLOC_N(map_bias)\n#define\tmb_write JEMALLOC_N(mb_write)\n#define\tmutex_boot JEMALLOC_N(mutex_boot)\n#define\tnarenas_auto JEMALLOC_N(narenas_auto)\n#define\tnarenas_total JEMALLOC_N(narenas_total)\n#define\tnarenas_total_get JEMALLOC_N(narenas_total_get)\n#define\tncpus JEMALLOC_N(ncpus)\n#define\tnhbins JEMALLOC_N(nhbins)\n#define\topt_abort JEMALLOC_N(opt_abort)\n#define\topt_junk JEMALLOC_N(opt_junk)\n#define\topt_lg_chunk JEMALLOC_N(opt_lg_chunk)\n#define\topt_lg_dirty_mult JEMALLOC_N(opt_lg_dirty_mult)\n#define\topt_lg_prof_interval JEMALLOC_N(opt_lg_prof_interval)\n#define\topt_lg_prof_sample JEMALLOC_N(opt_lg_prof_sample)\n#define\topt_lg_tcache_max JEMALLOC_N(opt_lg_tcache_max)\n#define\topt_narenas JEMALLOC_N(opt_narenas)\n#define\topt_prof JEMALLOC_N(opt_prof)\n#define\topt_prof_accum JEMALLOC_N(opt_prof_accum)\n#define\topt_prof_active JEMALLOC_N(opt_prof_active)\n#define\topt_prof_final JEMALLOC_N(opt_prof_final)\n#define\topt_prof_gdump JEMALLOC_N(opt_prof_gdump)\n#define\topt_prof_leak JEMALLOC_N(opt_prof_leak)\n#define\topt_prof_prefix JEMALLOC_N(opt_prof_prefix)\n#define\topt_quarantine JEMALLOC_N(opt_quarantine)\n#define\topt_redzone JEMALLOC_N(opt_redzone)\n#define\topt_stats_print JEMALLOC_N(opt_stats_print)\n#define\topt_tcache JEMALLOC_N(opt_tcache)\n#define\topt_utrace JEMALLOC_N(opt_utrace)\n#define\topt_valgrind JEMALLOC_N(opt_valgrind)\n#define\topt_xmalloc JEMALLOC_N(opt_xmalloc)\n#define\topt_zero JEMALLOC_N(opt_zero)\n#define\tp2rz JEMALLOC_N(p2rz)\n#define\tpages_purge JEMALLOC_N(pages_purge)\n#define\tpow2_ceil JEMALLOC_N(pow2_ceil)\n#define\tprof_backtrace JEMALLOC_N(prof_backtrace)\n#define\tprof_boot0 JEMALLOC_N(prof_boot0)\n#define\tprof_boot1 JEMALLOC_N(prof_boot1)\n#define\tprof_boot2 JEMALLOC_N(prof_boot2)\n#define\tprof_ctx_get JEMALLOC_N(prof_ctx_get)\n#define\tprof_ctx_set JEMALLOC_N(prof_ctx_set)\n#define\tprof_free JEMALLOC_N(prof_free)\n#define\tprof_gdump JEMALLOC_N(prof_gdump)\n#define\tprof_idump JEMALLOC_N(prof_idump)\n#define\tprof_interval JEMALLOC_N(prof_interval)\n#define\tprof_lookup JEMALLOC_N(prof_lookup)\n#define\tprof_malloc JEMALLOC_N(prof_malloc)\n#define\tprof_mdump JEMALLOC_N(prof_mdump)\n#define\tprof_postfork_child JEMALLOC_N(prof_postfork_child)\n#define\tprof_postfork_parent JEMALLOC_N(prof_postfork_parent)\n#define\tprof_prefork JEMALLOC_N(prof_prefork)\n#define\tprof_promote JEMALLOC_N(prof_promote)\n#define\tprof_realloc JEMALLOC_N(prof_realloc)\n#define\tprof_sample_accum_update JEMALLOC_N(prof_sample_accum_update)\n#define\tprof_sample_threshold_update JEMALLOC_N(prof_sample_threshold_update)\n#define\tprof_tdata_booted JEMALLOC_N(prof_tdata_booted)\n#define\tprof_tdata_cleanup JEMALLOC_N(prof_tdata_cleanup)\n#define\tprof_tdata_get JEMALLOC_N(prof_tdata_get)\n#define\tprof_tdata_init JEMALLOC_N(prof_tdata_init)\n#define\tprof_tdata_initialized JEMALLOC_N(prof_tdata_initialized)\n#define\tprof_tdata_tls JEMALLOC_N(prof_tdata_tls)\n#define\tprof_tdata_tsd JEMALLOC_N(prof_tdata_tsd)\n#define\tprof_tdata_tsd_boot JEMALLOC_N(prof_tdata_tsd_boot)\n#define\tprof_tdata_tsd_cleanup_wrapper JEMALLOC_N(prof_tdata_tsd_cleanup_wrapper)\n#define\tprof_tdata_tsd_get JEMALLOC_N(prof_tdata_tsd_get)\n#define\tprof_tdata_tsd_set JEMALLOC_N(prof_tdata_tsd_set)\n#define\tquarantine JEMALLOC_N(quarantine)\n#define\tquarantine_boot JEMALLOC_N(quarantine_boot)\n#define\tquarantine_tsd_boot JEMALLOC_N(quarantine_tsd_boot)\n#define\tquarantine_tsd_cleanup_wrapper JEMALLOC_N(quarantine_tsd_cleanup_wrapper)\n#define\tquarantine_tsd_get JEMALLOC_N(quarantine_tsd_get)\n#define\tquarantine_tsd_set JEMALLOC_N(quarantine_tsd_set)\n#define\tregister_zone JEMALLOC_N(register_zone)\n#define\trtree_get JEMALLOC_N(rtree_get)\n#define\trtree_get_locked JEMALLOC_N(rtree_get_locked)\n#define\trtree_new JEMALLOC_N(rtree_new)\n#define\trtree_postfork_child JEMALLOC_N(rtree_postfork_child)\n#define\trtree_postfork_parent JEMALLOC_N(rtree_postfork_parent)\n#define\trtree_prefork JEMALLOC_N(rtree_prefork)\n#define\trtree_set JEMALLOC_N(rtree_set)\n#define\ts2u JEMALLOC_N(s2u)\n#define\tsa2u JEMALLOC_N(sa2u)\n#define\tset_errno JEMALLOC_N(set_errno)\n#define\tstats_cactive JEMALLOC_N(stats_cactive)\n#define\tstats_cactive_add JEMALLOC_N(stats_cactive_add)\n#define\tstats_cactive_get JEMALLOC_N(stats_cactive_get)\n#define\tstats_cactive_sub JEMALLOC_N(stats_cactive_sub)\n#define\tstats_chunks JEMALLOC_N(stats_chunks)\n#define\tstats_print JEMALLOC_N(stats_print)\n#define\ttcache_alloc_easy JEMALLOC_N(tcache_alloc_easy)\n#define\ttcache_alloc_large JEMALLOC_N(tcache_alloc_large)\n#define\ttcache_alloc_small JEMALLOC_N(tcache_alloc_small)\n#define\ttcache_alloc_small_hard JEMALLOC_N(tcache_alloc_small_hard)\n#define\ttcache_arena_associate JEMALLOC_N(tcache_arena_associate)\n#define\ttcache_arena_dissociate JEMALLOC_N(tcache_arena_dissociate)\n#define\ttcache_bin_flush_large JEMALLOC_N(tcache_bin_flush_large)\n#define\ttcache_bin_flush_small JEMALLOC_N(tcache_bin_flush_small)\n#define\ttcache_bin_info JEMALLOC_N(tcache_bin_info)\n#define\ttcache_boot0 JEMALLOC_N(tcache_boot0)\n#define\ttcache_boot1 JEMALLOC_N(tcache_boot1)\n#define\ttcache_booted JEMALLOC_N(tcache_booted)\n#define\ttcache_create JEMALLOC_N(tcache_create)\n#define\ttcache_dalloc_large JEMALLOC_N(tcache_dalloc_large)\n#define\ttcache_dalloc_small JEMALLOC_N(tcache_dalloc_small)\n#define\ttcache_destroy JEMALLOC_N(tcache_destroy)\n#define\ttcache_enabled_booted JEMALLOC_N(tcache_enabled_booted)\n#define\ttcache_enabled_get JEMALLOC_N(tcache_enabled_get)\n#define\ttcache_enabled_initialized JEMALLOC_N(tcache_enabled_initialized)\n#define\ttcache_enabled_set JEMALLOC_N(tcache_enabled_set)\n#define\ttcache_enabled_tls JEMALLOC_N(tcache_enabled_tls)\n#define\ttcache_enabled_tsd JEMALLOC_N(tcache_enabled_tsd)\n#define\ttcache_enabled_tsd_boot JEMALLOC_N(tcache_enabled_tsd_boot)\n#define\ttcache_enabled_tsd_cleanup_wrapper JEMALLOC_N(tcache_enabled_tsd_cleanup_wrapper)\n#define\ttcache_enabled_tsd_get JEMALLOC_N(tcache_enabled_tsd_get)\n#define\ttcache_enabled_tsd_set JEMALLOC_N(tcache_enabled_tsd_set)\n#define\ttcache_event JEMALLOC_N(tcache_event)\n#define\ttcache_event_hard JEMALLOC_N(tcache_event_hard)\n#define\ttcache_flush JEMALLOC_N(tcache_flush)\n#define\ttcache_get JEMALLOC_N(tcache_get)\n#define\ttcache_initialized JEMALLOC_N(tcache_initialized)\n#define\ttcache_maxclass JEMALLOC_N(tcache_maxclass)\n#define\ttcache_salloc JEMALLOC_N(tcache_salloc)\n#define\ttcache_stats_merge JEMALLOC_N(tcache_stats_merge)\n#define\ttcache_thread_cleanup JEMALLOC_N(tcache_thread_cleanup)\n#define\ttcache_tls JEMALLOC_N(tcache_tls)\n#define\ttcache_tsd JEMALLOC_N(tcache_tsd)\n#define\ttcache_tsd_boot JEMALLOC_N(tcache_tsd_boot)\n#define\ttcache_tsd_cleanup_wrapper JEMALLOC_N(tcache_tsd_cleanup_wrapper)\n#define\ttcache_tsd_get JEMALLOC_N(tcache_tsd_get)\n#define\ttcache_tsd_set JEMALLOC_N(tcache_tsd_set)\n#define\tthread_allocated_booted JEMALLOC_N(thread_allocated_booted)\n#define\tthread_allocated_initialized JEMALLOC_N(thread_allocated_initialized)\n#define\tthread_allocated_tls JEMALLOC_N(thread_allocated_tls)\n#define\tthread_allocated_tsd JEMALLOC_N(thread_allocated_tsd)\n#define\tthread_allocated_tsd_boot JEMALLOC_N(thread_allocated_tsd_boot)\n#define\tthread_allocated_tsd_cleanup_wrapper JEMALLOC_N(thread_allocated_tsd_cleanup_wrapper)\n#define\tthread_allocated_tsd_get JEMALLOC_N(thread_allocated_tsd_get)\n#define\tthread_allocated_tsd_set JEMALLOC_N(thread_allocated_tsd_set)\n#define\tu2rz JEMALLOC_N(u2rz)\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/prng.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/*\n * Simple linear congruential pseudo-random number generator:\n *\n * prng(y) = (a*x + c) % m\n *\n * where the following constants ensure maximal period:\n *\n * a == Odd number (relatively prime to 2^n), and (a-1) is a multiple of 4.\n * c == Odd number (relatively prime to 2^n).\n * m == 2^32\n *\n * See Knuth's TAOCP 3rd Ed., Vol. 2, pg. 17 for details on these constraints.\n *\n * This choice of m has the disadvantage that the quality of the bits is\n * proportional to bit position. For example. the lowest bit has a cycle of 2,\n * the next has a cycle of 4, etc. For this reason, we prefer to use the upper\n * bits.\n *\n * Macro parameters:\n * uint32_t r : Result.\n * unsigned lg_range : (0..32], number of least significant bits to return.\n * uint32_t state : Seed value.\n * const uint32_t a, c : See above discussion.\n */\n#define prng32(r, lg_range, state, a, c) do {\t\t\t\t\\\n\tassert(lg_range > 0);\t\t\t\t\t\t\\\n\tassert(lg_range <= 32);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tr = (state * (a)) + (c);\t\t\t\t\t\\\n\tstate = r;\t\t\t\t\t\t\t\\\n\tr >>= (32 - lg_range);\t\t\t\t\t\t\\\n} while (false)\n\n/* Same as prng32(), but 64 bits of pseudo-randomness, using uint64_t. */\n#define prng64(r, lg_range, state, a, c) do {\t\t\t\t\\\n\tassert(lg_range > 0);\t\t\t\t\t\t\\\n\tassert(lg_range <= 64);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tr = (state * (a)) + (c);\t\t\t\t\t\\\n\tstate = r;\t\t\t\t\t\t\t\\\n\tr >>= (64 - lg_range);\t\t\t\t\t\t\\\n} while (false)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/prof.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct prof_bt_s prof_bt_t;\ntypedef struct prof_cnt_s prof_cnt_t;\ntypedef struct prof_thr_cnt_s prof_thr_cnt_t;\ntypedef struct prof_ctx_s prof_ctx_t;\ntypedef struct prof_tdata_s prof_tdata_t;\n\n/* Option defaults. */\n#define\tPROF_PREFIX_DEFAULT\t\t\"jeprof\"\n#define\tLG_PROF_SAMPLE_DEFAULT\t\t19\n#define\tLG_PROF_INTERVAL_DEFAULT\t-1\n\n/*\n * Hard limit on stack backtrace depth. The version of prof_backtrace() that\n * is based on __builtin_return_address() necessarily has a hard-coded number\n * of backtrace frame handlers, and should be kept in sync with this setting.\n */\n#define\tPROF_BT_MAX\t\t\t128\n\n/* Maximum number of backtraces to store in each per thread LRU cache. */\n#define\tPROF_TCMAX\t\t\t1024\n\n/* Initial hash table size. */\n#define\tPROF_CKH_MINITEMS\t\t64\n\n/* Size of memory buffer to use when writing dump files. */\n#define\tPROF_DUMP_BUFSIZE\t\t65536\n\n/* Size of stack-allocated buffer used by prof_printf(). */\n#define\tPROF_PRINTF_BUFSIZE\t\t128\n\n/*\n * Number of mutexes shared among all ctx's. No space is allocated for these\n * unless profiling is enabled, so it's okay to over-provision.\n */\n#define\tPROF_NCTX_LOCKS\t\t\t1024\n\n/*\n * prof_tdata pointers close to NULL are used to encode state information that\n * is used for cleaning up during thread shutdown.\n */\n#define\tPROF_TDATA_STATE_REINCARNATED\t((prof_tdata_t *)(uintptr_t)1)\n#define\tPROF_TDATA_STATE_PURGATORY\t((prof_tdata_t *)(uintptr_t)2)\n#define\tPROF_TDATA_STATE_MAX\t\tPROF_TDATA_STATE_PURGATORY\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct prof_bt_s {\n\t/* Backtrace, stored as len program counters. */\n\tvoid\t\t**vec;\n\tunsigned\tlen;\n};\n\n#ifdef JEMALLOC_PROF_LIBGCC\n/* Data structure passed to libgcc _Unwind_Backtrace() callback functions. */\ntypedef struct {\n\tprof_bt_t\t*bt;\n\tunsigned\tnignore;\n\tunsigned\tmax;\n} prof_unwind_data_t;\n#endif\n\nstruct prof_cnt_s {\n\t/*\n\t * Profiling counters. An allocation/deallocation pair can operate on\n\t * different prof_thr_cnt_t objects that are linked into the same\n\t * prof_ctx_t cnts_ql, so it is possible for the cur* counters to go\n\t * negative. In principle it is possible for the *bytes counters to\n\t * overflow/underflow, but a general solution would require something\n\t * like 128-bit counters; this implementation doesn't bother to solve\n\t * that problem.\n\t */\n\tint64_t\t\tcurobjs;\n\tint64_t\t\tcurbytes;\n\tuint64_t\taccumobjs;\n\tuint64_t\taccumbytes;\n};\n\nstruct prof_thr_cnt_s {\n\t/* Linkage into prof_ctx_t's cnts_ql. */\n\tql_elm(prof_thr_cnt_t)\tcnts_link;\n\n\t/* Linkage into thread's LRU. */\n\tql_elm(prof_thr_cnt_t)\tlru_link;\n\n\t/*\n\t * Associated context. If a thread frees an object that it did not\n\t * allocate, it is possible that the context is not cached in the\n\t * thread's hash table, in which case it must be able to look up the\n\t * context, insert a new prof_thr_cnt_t into the thread's hash table,\n\t * and link it into the prof_ctx_t's cnts_ql.\n\t */\n\tprof_ctx_t\t\t*ctx;\n\n\t/*\n\t * Threads use memory barriers to update the counters. Since there is\n\t * only ever one writer, the only challenge is for the reader to get a\n\t * consistent read of the counters.\n\t *\n\t * The writer uses this series of operations:\n\t *\n\t * 1) Increment epoch to an odd number.\n\t * 2) Update counters.\n\t * 3) Increment epoch to an even number.\n\t *\n\t * The reader must assure 1) that the epoch is even while it reads the\n\t * counters, and 2) that the epoch doesn't change between the time it\n\t * starts and finishes reading the counters.\n\t */\n\tunsigned\t\tepoch;\n\n\t/* Profiling counters. */\n\tprof_cnt_t\t\tcnts;\n};\n\nstruct prof_ctx_s {\n\t/* Associated backtrace. */\n\tprof_bt_t\t\t*bt;\n\n\t/* Protects nlimbo, cnt_merged, and cnts_ql. */\n\tmalloc_mutex_t\t\t*lock;\n\n\t/*\n\t * Number of threads that currently cause this ctx to be in a state of\n\t * limbo due to one of:\n\t * - Initializing per thread counters associated with this ctx.\n\t * - Preparing to destroy this ctx.\n\t * nlimbo must be 1 (single destroyer) in order to safely destroy the\n\t * ctx.\n\t */\n\tunsigned\t\tnlimbo;\n\n\t/* Temporary storage for summation during dump. */\n\tprof_cnt_t\t\tcnt_summed;\n\n\t/* When threads exit, they merge their stats into cnt_merged. */\n\tprof_cnt_t\t\tcnt_merged;\n\n\t/*\n\t * List of profile counters, one for each thread that has allocated in\n\t * this context.\n\t */\n\tql_head(prof_thr_cnt_t)\tcnts_ql;\n};\n\nstruct prof_tdata_s {\n\t/*\n\t * Hash of (prof_bt_t *)-->(prof_thr_cnt_t *). Each thread keeps a\n\t * cache of backtraces, with associated thread-specific prof_thr_cnt_t\n\t * objects. Other threads may read the prof_thr_cnt_t contents, but no\n\t * others will ever write them.\n\t *\n\t * Upon thread exit, the thread must merge all the prof_thr_cnt_t\n\t * counter data into the associated prof_ctx_t objects, and unlink/free\n\t * the prof_thr_cnt_t objects.\n\t */\n\tckh_t\t\t\tbt2cnt;\n\n\t/* LRU for contents of bt2cnt. */\n\tql_head(prof_thr_cnt_t)\tlru_ql;\n\n\t/* Backtrace vector, used for calls to prof_backtrace(). */\n\tvoid\t\t\t**vec;\n\n\t/* Sampling state. */\n\tuint64_t\t\tprng_state;\n\tuint64_t\t\tthreshold;\n\tuint64_t\t\taccum;\n\n\t/* State used to avoid dumping while operating on prof internals. */\n\tbool\t\t\tenq;\n\tbool\t\t\tenq_idump;\n\tbool\t\t\tenq_gdump;\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nextern bool\topt_prof;\n/*\n * Even if opt_prof is true, sampling can be temporarily disabled by setting\n * opt_prof_active to false. No locking is used when updating opt_prof_active,\n * so there are no guarantees regarding how long it will take for all threads\n * to notice state changes.\n */\nextern bool\topt_prof_active;\nextern size_t\topt_lg_prof_sample; /* Mean bytes between samples. */\nextern ssize_t\topt_lg_prof_interval; /* lg(prof_interval). */\nextern bool\topt_prof_gdump; /* High-water memory dumping. */\nextern bool\topt_prof_final; /* Final profile dumping. */\nextern bool\topt_prof_leak; /* Dump leak summary at exit. */\nextern bool\topt_prof_accum; /* Report cumulative bytes. */\nextern char\topt_prof_prefix[PATH_MAX + 1];\n\n/*\n * Profile dump interval, measured in bytes allocated. Each arena triggers a\n * profile dump when it reaches this threshold. The effect is that the\n * interval between profile dumps averages prof_interval, though the actual\n * interval between dumps will tend to be sporadic, and the interval will be a\n * maximum of approximately (prof_interval * narenas).\n */\nextern uint64_t\tprof_interval;\n\n/*\n * If true, promote small sampled objects to large objects, since small run\n * headers do not have embedded profile context pointers.\n */\nextern bool\tprof_promote;\n\nvoid\tbt_init(prof_bt_t *bt, void **vec);\nvoid\tprof_backtrace(prof_bt_t *bt, unsigned nignore);\nprof_thr_cnt_t\t*prof_lookup(prof_bt_t *bt);\nvoid\tprof_idump(void);\nbool\tprof_mdump(const char *filename);\nvoid\tprof_gdump(void);\nprof_tdata_t\t*prof_tdata_init(void);\nvoid\tprof_tdata_cleanup(void *arg);\nvoid\tprof_boot0(void);\nvoid\tprof_boot1(void);\nbool\tprof_boot2(void);\nvoid\tprof_prefork(void);\nvoid\tprof_postfork_parent(void);\nvoid\tprof_postfork_child(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#define\tPROF_ALLOC_PREP(nignore, size, ret) do {\t\t\t\\\n\tprof_tdata_t *prof_tdata;\t\t\t\t\t\\\n\tprof_bt_t bt;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(size == s2u(size));\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tprof_tdata = prof_tdata_get();\t\t\t\t\t\\\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) {\t\\\n\t\tif (prof_tdata != NULL)\t\t\t\t\t\\\n\t\t\tret = (prof_thr_cnt_t *)(uintptr_t)1U;\t\t\\\n\t\telse\t\t\t\t\t\t\t\\\n\t\t\tret = NULL;\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (opt_prof_active == false) {\t\t\t\t\t\\\n\t\t/* Sampling is currently inactive, so avoid sampling. */\\\n\t\tret = (prof_thr_cnt_t *)(uintptr_t)1U;\t\t\t\\\n\t} else if (opt_lg_prof_sample == 0) {\t\t\t\t\\\n\t\t/* Don't bother with sampling logic, since sampling */\\\n\t\t/* interval is 1. */\\\n\t\tbt_init(&bt, prof_tdata->vec);\t\t\t\t\\\n\t\tprof_backtrace(&bt, nignore);\t\t\t\t\\\n\t\tret = prof_lookup(&bt);\t\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t\tif (prof_tdata->threshold == 0) {\t\t\t\\\n\t\t\t/* Initialize. Seed the prng differently for */\\\n\t\t\t/* each thread. */\\\n\t\t\tprof_tdata->prng_state =\t\t\t\\\n\t\t\t (uint64_t)(uintptr_t)&size;\t\t\t\\\n\t\t\tprof_sample_threshold_update(prof_tdata);\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\t\t/* Determine whether to capture a backtrace based on */\\\n\t\t/* whether size is enough for prof_accum to reach */\\\n\t\t/* prof_tdata->threshold. However, delay updating */\\\n\t\t/* these variables until prof_{m,re}alloc(), because */\\\n\t\t/* we don't know for sure that the allocation will */\\\n\t\t/* succeed. */\\\n\t\t/* */\\\n\t\t/* Use subtraction rather than addition to avoid */\\\n\t\t/* potential integer overflow. */\\\n\t\tif (size >= prof_tdata->threshold -\t\t\t\\\n\t\t prof_tdata->accum) {\t\t\t\t\\\n\t\t\tbt_init(&bt, prof_tdata->vec);\t\t\t\\\n\t\t\tprof_backtrace(&bt, nignore);\t\t\t\\\n\t\t\tret = prof_lookup(&bt);\t\t\t\t\\\n\t\t} else\t\t\t\t\t\t\t\\\n\t\t\tret = (prof_thr_cnt_t *)(uintptr_t)1U;\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#ifndef JEMALLOC_ENABLE_INLINE\nmalloc_tsd_protos(JEMALLOC_ATTR(unused), prof_tdata, prof_tdata_t *)\n\nprof_tdata_t\t*prof_tdata_get(void);\nvoid\tprof_sample_threshold_update(prof_tdata_t *prof_tdata);\nprof_ctx_t\t*prof_ctx_get(const void *ptr);\nvoid\tprof_ctx_set(const void *ptr, prof_ctx_t *ctx);\nbool\tprof_sample_accum_update(size_t size);\nvoid\tprof_malloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt);\nvoid\tprof_realloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt,\n size_t old_size, prof_ctx_t *old_ctx);\nvoid\tprof_free(const void *ptr, size_t size);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_PROF_C_))\n/* Thread-specific backtrace cache, used to reduce bt2ctx contention. */\nmalloc_tsd_externs(prof_tdata, prof_tdata_t *)\nmalloc_tsd_funcs(JEMALLOC_INLINE, prof_tdata, prof_tdata_t *, NULL,\n prof_tdata_cleanup)\n\nJEMALLOC_INLINE prof_tdata_t *\nprof_tdata_get(void)\n{\n\tprof_tdata_t *prof_tdata;\n\n\tcassert(config_prof);\n\n\tprof_tdata = *prof_tdata_tsd_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) {\n\t\tif (prof_tdata == NULL)\n\t\t\tprof_tdata = prof_tdata_init();\n\t}\n\n\treturn (prof_tdata);\n}\n\nJEMALLOC_INLINE void\nprof_sample_threshold_update(prof_tdata_t *prof_tdata)\n{\n\tuint64_t r;\n\tdouble u;\n\n\tcassert(config_prof);\n\n\t/*\n\t * Compute sample threshold as a geometrically distributed random\n\t * variable with mean (2^opt_lg_prof_sample).\n\t *\n\t * __ __\n\t * | log(u) | 1\n\t * prof_tdata->threshold = | -------- |, where p = -------------------\n\t * | log(1-p) | opt_lg_prof_sample\n\t * 2\n\t *\n\t * For more information on the math, see:\n\t *\n\t * Non-Uniform Random Variate Generation\n\t * Luc Devroye\n\t * Springer-Verlag, New York, 1986\n\t * pp 500\n\t * (http://cg.scs.carleton.ca/~luc/rnbookindex.html)\n\t */\n\tprng64(r, 53, prof_tdata->prng_state,\n\t UINT64_C(6364136223846793005), UINT64_C(1442695040888963407));\n\tu = (double)r * (1.0/9007199254740992.0L);\n\tprof_tdata->threshold = (uint64_t)(log(u) /\n\t log(1.0 - (1.0 / (double)((uint64_t)1U << opt_lg_prof_sample))))\n\t + (uint64_t)1U;\n}\n\nJEMALLOC_INLINE prof_ctx_t *\nprof_ctx_get(const void *ptr)\n{\n\tprof_ctx_t *ret;\n\tarena_chunk_t *chunk;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tif (chunk != ptr) {\n\t\t/* Region. */\n\t\tret = arena_prof_ctx_get(ptr);\n\t} else\n\t\tret = huge_prof_ctx_get(ptr);\n\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void\nprof_ctx_set(const void *ptr, prof_ctx_t *ctx)\n{\n\tarena_chunk_t *chunk;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tif (chunk != ptr) {\n\t\t/* Region. */\n\t\tarena_prof_ctx_set(ptr, ctx);\n\t} else\n\t\thuge_prof_ctx_set(ptr, ctx);\n}\n\nJEMALLOC_INLINE bool\nprof_sample_accum_update(size_t size)\n{\n\tprof_tdata_t *prof_tdata;\n\n\tcassert(config_prof);\n\t/* Sampling logic is unnecessary if the interval is 1. */\n\tassert(opt_lg_prof_sample != 0);\n\n\tprof_tdata = *prof_tdata_tsd_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)\n\t\treturn (true);\n\n\t/* Take care to avoid integer overflow. */\n\tif (size >= prof_tdata->threshold - prof_tdata->accum) {\n\t\tprof_tdata->accum -= (prof_tdata->threshold - size);\n\t\t/* Compute new sample threshold. */\n\t\tprof_sample_threshold_update(prof_tdata);\n\t\twhile (prof_tdata->accum >= prof_tdata->threshold) {\n\t\t\tprof_tdata->accum -= prof_tdata->threshold;\n\t\t\tprof_sample_threshold_update(prof_tdata);\n\t\t}\n\t\treturn (false);\n\t} else {\n\t\tprof_tdata->accum += size;\n\t\treturn (true);\n\t}\n}\n\nJEMALLOC_INLINE void\nprof_malloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt)\n{\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\tassert(size == isalloc(ptr, true));\n\n\tif (opt_lg_prof_sample != 0) {\n\t\tif (prof_sample_accum_update(size)) {\n\t\t\t/*\n\t\t\t * Don't sample. For malloc()-like allocation, it is\n\t\t\t * always possible to tell in advance how large an\n\t\t\t * object's usable size will be, so there should never\n\t\t\t * be a difference between the size passed to\n\t\t\t * PROF_ALLOC_PREP() and prof_malloc().\n\t\t\t */\n\t\t\tassert((uintptr_t)cnt == (uintptr_t)1U);\n\t\t}\n\t}\n\n\tif ((uintptr_t)cnt > (uintptr_t)1U) {\n\t\tprof_ctx_set(ptr, cnt->ctx);\n\n\t\tcnt->epoch++;\n\t\t/*********/\n\t\tmb_write();\n\t\t/*********/\n\t\tcnt->cnts.curobjs++;\n\t\tcnt->cnts.curbytes += size;\n\t\tif (opt_prof_accum) {\n\t\t\tcnt->cnts.accumobjs++;\n\t\t\tcnt->cnts.accumbytes += size;\n\t\t}\n\t\t/*********/\n\t\tmb_write();\n\t\t/*********/\n\t\tcnt->epoch++;\n\t\t/*********/\n\t\tmb_write();\n\t\t/*********/\n\t} else\n\t\tprof_ctx_set(ptr, (prof_ctx_t *)(uintptr_t)1U);\n}\n\nJEMALLOC_INLINE void\nprof_realloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt,\n size_t old_size, prof_ctx_t *old_ctx)\n{\n\tprof_thr_cnt_t *told_cnt;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL || (uintptr_t)cnt <= (uintptr_t)1U);\n\n\tif (ptr != NULL) {\n\t\tassert(size == isalloc(ptr, true));\n\t\tif (opt_lg_prof_sample != 0) {\n\t\t\tif (prof_sample_accum_update(size)) {\n\t\t\t\t/*\n\t\t\t\t * Don't sample. The size passed to\n\t\t\t\t * PROF_ALLOC_PREP() was larger than what\n\t\t\t\t * actually got allocated, so a backtrace was\n\t\t\t\t * captured for this allocation, even though\n\t\t\t\t * its actual size was insufficient to cross\n\t\t\t\t * the sample threshold.\n\t\t\t\t */\n\t\t\t\tcnt = (prof_thr_cnt_t *)(uintptr_t)1U;\n\t\t\t}\n\t\t}\n\t}\n\n\tif ((uintptr_t)old_ctx > (uintptr_t)1U) {\n\t\ttold_cnt = prof_lookup(old_ctx->bt);\n\t\tif (told_cnt == NULL) {\n\t\t\t/*\n\t\t\t * It's too late to propagate OOM for this realloc(),\n\t\t\t * so operate directly on old_cnt->ctx->cnt_merged.\n\t\t\t */\n\t\t\tmalloc_mutex_lock(old_ctx->lock);\n\t\t\told_ctx->cnt_merged.curobjs--;\n\t\t\told_ctx->cnt_merged.curbytes -= old_size;\n\t\t\tmalloc_mutex_unlock(old_ctx->lock);\n\t\t\ttold_cnt = (prof_thr_cnt_t *)(uintptr_t)1U;\n\t\t}\n\t} else\n\t\ttold_cnt = (prof_thr_cnt_t *)(uintptr_t)1U;\n\n\tif ((uintptr_t)told_cnt > (uintptr_t)1U)\n\t\ttold_cnt->epoch++;\n\tif ((uintptr_t)cnt > (uintptr_t)1U) {\n\t\tprof_ctx_set(ptr, cnt->ctx);\n\t\tcnt->epoch++;\n\t} else if (ptr != NULL)\n\t\tprof_ctx_set(ptr, (prof_ctx_t *)(uintptr_t)1U);\n\t/*********/\n\tmb_write();\n\t/*********/\n\tif ((uintptr_t)told_cnt > (uintptr_t)1U) {\n\t\ttold_cnt->cnts.curobjs--;\n\t\ttold_cnt->cnts.curbytes -= old_size;\n\t}\n\tif ((uintptr_t)cnt > (uintptr_t)1U) {\n\t\tcnt->cnts.curobjs++;\n\t\tcnt->cnts.curbytes += size;\n\t\tif (opt_prof_accum) {\n\t\t\tcnt->cnts.accumobjs++;\n\t\t\tcnt->cnts.accumbytes += size;\n\t\t}\n\t}\n\t/*********/\n\tmb_write();\n\t/*********/\n\tif ((uintptr_t)told_cnt > (uintptr_t)1U)\n\t\ttold_cnt->epoch++;\n\tif ((uintptr_t)cnt > (uintptr_t)1U)\n\t\tcnt->epoch++;\n\t/*********/\n\tmb_write(); /* Not strictly necessary. */\n}\n\nJEMALLOC_INLINE void\nprof_free(const void *ptr, size_t size)\n{\n\tprof_ctx_t *ctx = prof_ctx_get(ptr);\n\n\tcassert(config_prof);\n\n\tif ((uintptr_t)ctx > (uintptr_t)1) {\n\t\tprof_thr_cnt_t *tcnt;\n\t\tassert(size == isalloc(ptr, true));\n\t\ttcnt = prof_lookup(ctx->bt);\n\n\t\tif (tcnt != NULL) {\n\t\t\ttcnt->epoch++;\n\t\t\t/*********/\n\t\t\tmb_write();\n\t\t\t/*********/\n\t\t\ttcnt->cnts.curobjs--;\n\t\t\ttcnt->cnts.curbytes -= size;\n\t\t\t/*********/\n\t\t\tmb_write();\n\t\t\t/*********/\n\t\t\ttcnt->epoch++;\n\t\t\t/*********/\n\t\t\tmb_write();\n\t\t\t/*********/\n\t\t} else {\n\t\t\t/*\n\t\t\t * OOM during free() cannot be propagated, so operate\n\t\t\t * directly on cnt->ctx->cnt_merged.\n\t\t\t */\n\t\t\tmalloc_mutex_lock(ctx->lock);\n\t\t\tctx->cnt_merged.curobjs--;\n\t\t\tctx->cnt_merged.curbytes -= size;\n\t\t\tmalloc_mutex_unlock(ctx->lock);\n\t\t}\n\t}\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/ql.h", "content": "/*\n * List definitions.\n */\n#define ql_head(a_type)\t\t\t\t\t\t\t\\\nstruct {\t\t\t\t\t\t\t\t\\\n\ta_type *qlh_first;\t\t\t\t\t\t\\\n}\n\n#define ql_head_initializer(a_head) {NULL}\n\n#define ql_elm(a_type)\tqr(a_type)\n\n/* List functions. */\n#define ql_new(a_head) do {\t\t\t\t\t\t\\\n\t(a_head)->qlh_first = NULL;\t\t\t\t\t\\\n} while (0)\n\n#define ql_elm_new(a_elm, a_field) qr_new((a_elm), a_field)\n\n#define ql_first(a_head) ((a_head)->qlh_first)\n\n#define ql_last(a_head, a_field)\t\t\t\t\t\\\n\t((ql_first(a_head) != NULL)\t\t\t\t\t\\\n\t ? qr_prev(ql_first(a_head), a_field) : NULL)\n\n#define ql_next(a_head, a_elm, a_field)\t\t\t\t\t\\\n\t((ql_last(a_head, a_field) != (a_elm))\t\t\t\t\\\n\t ? qr_next((a_elm), a_field)\t: NULL)\n\n#define ql_prev(a_head, a_elm, a_field)\t\t\t\t\t\\\n\t((ql_first(a_head) != (a_elm)) ? qr_prev((a_elm), a_field)\t\\\n\t\t\t\t : NULL)\n\n#define ql_before_insert(a_head, a_qlelm, a_elm, a_field) do {\t\t\\\n\tqr_before_insert((a_qlelm), (a_elm), a_field);\t\t\t\\\n\tif (ql_first(a_head) == (a_qlelm)) {\t\t\t\t\\\n\t\tql_first(a_head) = (a_elm);\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define ql_after_insert(a_qlelm, a_elm, a_field)\t\t\t\\\n\tqr_after_insert((a_qlelm), (a_elm), a_field)\n\n#define ql_head_insert(a_head, a_elm, a_field) do {\t\t\t\\\n\tif (ql_first(a_head) != NULL) {\t\t\t\t\t\\\n\t\tqr_before_insert(ql_first(a_head), (a_elm), a_field);\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tql_first(a_head) = (a_elm);\t\t\t\t\t\\\n} while (0)\n\n#define ql_tail_insert(a_head, a_elm, a_field) do {\t\t\t\\\n\tif (ql_first(a_head) != NULL) {\t\t\t\t\t\\\n\t\tqr_before_insert(ql_first(a_head), (a_elm), a_field);\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tql_first(a_head) = qr_next((a_elm), a_field);\t\t\t\\\n} while (0)\n\n#define ql_remove(a_head, a_elm, a_field) do {\t\t\t\t\\\n\tif (ql_first(a_head) == (a_elm)) {\t\t\t\t\\\n\t\tql_first(a_head) = qr_next(ql_first(a_head), a_field);\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tif (ql_first(a_head) != (a_elm)) {\t\t\t\t\\\n\t\tqr_remove((a_elm), a_field);\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t\tql_first(a_head) = NULL;\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define ql_head_remove(a_head, a_type, a_field) do {\t\t\t\\\n\ta_type *t = ql_first(a_head);\t\t\t\t\t\\\n\tql_remove((a_head), t, a_field);\t\t\t\t\\\n} while (0)\n\n#define ql_tail_remove(a_head, a_type, a_field) do {\t\t\t\\\n\ta_type *t = ql_last(a_head, a_field);\t\t\t\t\\\n\tql_remove((a_head), t, a_field);\t\t\t\t\\\n} while (0)\n\n#define ql_foreach(a_var, a_head, a_field)\t\t\t\t\\\n\tqr_foreach((a_var), ql_first(a_head), a_field)\n\n#define ql_reverse_foreach(a_var, a_head, a_field)\t\t\t\\\n\tqr_reverse_foreach((a_var), ql_first(a_head), a_field)\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/qr.h", "content": "/* Ring definitions. */\n#define qr(a_type)\t\t\t\t\t\t\t\\\nstruct {\t\t\t\t\t\t\t\t\\\n\ta_type\t*qre_next;\t\t\t\t\t\t\\\n\ta_type\t*qre_prev;\t\t\t\t\t\t\\\n}\n\n/* Ring functions. */\n#define qr_new(a_qr, a_field) do {\t\t\t\t\t\\\n\t(a_qr)->a_field.qre_next = (a_qr);\t\t\t\t\\\n\t(a_qr)->a_field.qre_prev = (a_qr);\t\t\t\t\\\n} while (0)\n\n#define qr_next(a_qr, a_field) ((a_qr)->a_field.qre_next)\n\n#define qr_prev(a_qr, a_field) ((a_qr)->a_field.qre_prev)\n\n#define qr_before_insert(a_qrelm, a_qr, a_field) do {\t\t\t\\\n\t(a_qr)->a_field.qre_prev = (a_qrelm)->a_field.qre_prev;\t\t\\\n\t(a_qr)->a_field.qre_next = (a_qrelm);\t\t\t\t\\\n\t(a_qr)->a_field.qre_prev->a_field.qre_next = (a_qr);\t\t\\\n\t(a_qrelm)->a_field.qre_prev = (a_qr);\t\t\t\t\\\n} while (0)\n\n#define qr_after_insert(a_qrelm, a_qr, a_field)\t\t\t\t\\\n do\t\t\t\t\t\t\t\t\t\\\n {\t\t\t\t\t\t\t\t\t\\\n\t(a_qr)->a_field.qre_next = (a_qrelm)->a_field.qre_next;\t\t\\\n\t(a_qr)->a_field.qre_prev = (a_qrelm);\t\t\t\t\\\n\t(a_qr)->a_field.qre_next->a_field.qre_prev = (a_qr);\t\t\\\n\t(a_qrelm)->a_field.qre_next = (a_qr);\t\t\t\t\\\n } while (0)\n\n#define qr_meld(a_qr_a, a_qr_b, a_field) do {\t\t\t\t\\\n\tvoid *t;\t\t\t\t\t\t\t\\\n\t(a_qr_a)->a_field.qre_prev->a_field.qre_next = (a_qr_b);\t\\\n\t(a_qr_b)->a_field.qre_prev->a_field.qre_next = (a_qr_a);\t\\\n\tt = (a_qr_a)->a_field.qre_prev;\t\t\t\t\t\\\n\t(a_qr_a)->a_field.qre_prev = (a_qr_b)->a_field.qre_prev;\t\\\n\t(a_qr_b)->a_field.qre_prev = t;\t\t\t\t\t\\\n} while (0)\n\n/* qr_meld() and qr_split() are functionally equivalent, so there's no need to\n * have two copies of the code. */\n#define qr_split(a_qr_a, a_qr_b, a_field)\t\t\t\t\\\n\tqr_meld((a_qr_a), (a_qr_b), a_field)\n\n#define qr_remove(a_qr, a_field) do {\t\t\t\t\t\\\n\t(a_qr)->a_field.qre_prev->a_field.qre_next\t\t\t\\\n\t = (a_qr)->a_field.qre_next;\t\t\t\t\t\\\n\t(a_qr)->a_field.qre_next->a_field.qre_prev\t\t\t\\\n\t = (a_qr)->a_field.qre_prev;\t\t\t\t\t\\\n\t(a_qr)->a_field.qre_next = (a_qr);\t\t\t\t\\\n\t(a_qr)->a_field.qre_prev = (a_qr);\t\t\t\t\\\n} while (0)\n\n#define qr_foreach(var, a_qr, a_field)\t\t\t\t\t\\\n\tfor ((var) = (a_qr);\t\t\t\t\t\t\\\n\t (var) != NULL;\t\t\t\t\t\t\\\n\t (var) = (((var)->a_field.qre_next != (a_qr))\t\t\\\n\t ? (var)->a_field.qre_next : NULL))\n\n#define qr_reverse_foreach(var, a_qr, a_field)\t\t\t\t\\\n\tfor ((var) = ((a_qr) != NULL) ? qr_prev(a_qr, a_field) : NULL;\t\\\n\t (var) != NULL;\t\t\t\t\t\t\\\n\t (var) = (((var) != (a_qr))\t\t\t\t\t\\\n\t ? (var)->a_field.qre_prev : NULL))\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/quarantine.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/* Default per thread quarantine size if valgrind is enabled. */\n#define\tJEMALLOC_VALGRIND_QUARANTINE_DEFAULT\t(ZU(1) << 24)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nvoid\tquarantine(void *ptr);\nbool\tquarantine_boot(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/rb.h", "content": "/*-\n *******************************************************************************\n *\n * cpp macro implementation of left-leaning 2-3 red-black trees. Parent\n * pointers are not used, and color bits are stored in the least significant\n * bit of right-child pointers (if RB_COMPACT is defined), thus making node\n * linkage as compact as is possible for red-black trees.\n *\n * Usage:\n *\n * #include \n * #include \n * #define NDEBUG // (Optional, see assert(3).)\n * #include \n * #define RB_COMPACT // (Optional, embed color bits in right-child pointers.)\n * #include \n * ...\n *\n *******************************************************************************\n */\n\n#ifndef RB_H_\n#define\tRB_H_\n\n#if 0\n__FBSDID(\"$FreeBSD: head/lib/libc/stdlib/rb.h 204493 2010-02-28 22:57:13Z jasone $\");\n#endif\n\n#ifdef RB_COMPACT\n/* Node structure. */\n#define\trb_node(a_type)\t\t\t\t\t\t\t\\\nstruct {\t\t\t\t\t\t\t\t\\\n a_type *rbn_left;\t\t\t\t\t\t\t\\\n a_type *rbn_right_red;\t\t\t\t\t\t\\\n}\n#else\n#define\trb_node(a_type)\t\t\t\t\t\t\t\\\nstruct {\t\t\t\t\t\t\t\t\\\n a_type *rbn_left;\t\t\t\t\t\t\t\\\n a_type *rbn_right;\t\t\t\t\t\t\t\\\n bool rbn_red;\t\t\t\t\t\t\t\\\n}\n#endif\n\n/* Root structure. */\n#define\trb_tree(a_type)\t\t\t\t\t\t\t\\\nstruct {\t\t\t\t\t\t\t\t\\\n a_type *rbt_root;\t\t\t\t\t\t\t\\\n a_type rbt_nil;\t\t\t\t\t\t\t\\\n}\n\n/* Left accessors. */\n#define\trbtn_left_get(a_type, a_field, a_node)\t\t\t\t\\\n ((a_node)->a_field.rbn_left)\n#define\trbtn_left_set(a_type, a_field, a_node, a_left) do {\t\t\\\n (a_node)->a_field.rbn_left = a_left;\t\t\t\t\\\n} while (0)\n\n#ifdef RB_COMPACT\n/* Right accessors. */\n#define\trbtn_right_get(a_type, a_field, a_node)\t\t\t\t\\\n ((a_type *) (((intptr_t) (a_node)->a_field.rbn_right_red)\t\t\\\n & ((ssize_t)-2)))\n#define\trbtn_right_set(a_type, a_field, a_node, a_right) do {\t\t\\\n (a_node)->a_field.rbn_right_red = (a_type *) (((uintptr_t) a_right)\t\\\n | (((uintptr_t) (a_node)->a_field.rbn_right_red) & ((size_t)1)));\t\\\n} while (0)\n\n/* Color accessors. */\n#define\trbtn_red_get(a_type, a_field, a_node)\t\t\t\t\\\n ((bool) (((uintptr_t) (a_node)->a_field.rbn_right_red)\t\t\\\n & ((size_t)1)))\n#define\trbtn_color_set(a_type, a_field, a_node, a_red) do {\t\t\\\n (a_node)->a_field.rbn_right_red = (a_type *) ((((intptr_t)\t\t\\\n (a_node)->a_field.rbn_right_red) & ((ssize_t)-2))\t\t\t\\\n | ((ssize_t)a_red));\t\t\t\t\t\t\\\n} while (0)\n#define\trbtn_red_set(a_type, a_field, a_node) do {\t\t\t\\\n (a_node)->a_field.rbn_right_red = (a_type *) (((uintptr_t)\t\t\\\n (a_node)->a_field.rbn_right_red) | ((size_t)1));\t\t\t\\\n} while (0)\n#define\trbtn_black_set(a_type, a_field, a_node) do {\t\t\t\\\n (a_node)->a_field.rbn_right_red = (a_type *) (((intptr_t)\t\t\\\n (a_node)->a_field.rbn_right_red) & ((ssize_t)-2));\t\t\\\n} while (0)\n#else\n/* Right accessors. */\n#define\trbtn_right_get(a_type, a_field, a_node)\t\t\t\t\\\n ((a_node)->a_field.rbn_right)\n#define\trbtn_right_set(a_type, a_field, a_node, a_right) do {\t\t\\\n (a_node)->a_field.rbn_right = a_right;\t\t\t\t\\\n} while (0)\n\n/* Color accessors. */\n#define\trbtn_red_get(a_type, a_field, a_node)\t\t\t\t\\\n ((a_node)->a_field.rbn_red)\n#define\trbtn_color_set(a_type, a_field, a_node, a_red) do {\t\t\\\n (a_node)->a_field.rbn_red = (a_red);\t\t\t\t\\\n} while (0)\n#define\trbtn_red_set(a_type, a_field, a_node) do {\t\t\t\\\n (a_node)->a_field.rbn_red = true;\t\t\t\t\t\\\n} while (0)\n#define\trbtn_black_set(a_type, a_field, a_node) do {\t\t\t\\\n (a_node)->a_field.rbn_red = false;\t\t\t\t\t\\\n} while (0)\n#endif\n\n/* Node initializer. */\n#define\trbt_node_new(a_type, a_field, a_rbt, a_node) do {\t\t\\\n rbtn_left_set(a_type, a_field, (a_node), &(a_rbt)->rbt_nil);\t\\\n rbtn_right_set(a_type, a_field, (a_node), &(a_rbt)->rbt_nil);\t\\\n rbtn_red_set(a_type, a_field, (a_node));\t\t\t\t\\\n} while (0)\n\n/* Tree initializer. */\n#define\trb_new(a_type, a_field, a_rbt) do {\t\t\t\t\\\n (a_rbt)->rbt_root = &(a_rbt)->rbt_nil;\t\t\t\t\\\n rbt_node_new(a_type, a_field, a_rbt, &(a_rbt)->rbt_nil);\t\t\\\n rbtn_black_set(a_type, a_field, &(a_rbt)->rbt_nil);\t\t\t\\\n} while (0)\n\n/* Internal utility macros. */\n#define\trbtn_first(a_type, a_field, a_rbt, a_root, r_node) do {\t\t\\\n (r_node) = (a_root);\t\t\t\t\t\t\\\n if ((r_node) != &(a_rbt)->rbt_nil) {\t\t\t\t\\\n\tfor (;\t\t\t\t\t\t\t\t\\\n\t rbtn_left_get(a_type, a_field, (r_node)) != &(a_rbt)->rbt_nil;\\\n\t (r_node) = rbtn_left_get(a_type, a_field, (r_node))) {\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\trbtn_last(a_type, a_field, a_rbt, a_root, r_node) do {\t\t\\\n (r_node) = (a_root);\t\t\t\t\t\t\\\n if ((r_node) != &(a_rbt)->rbt_nil) {\t\t\t\t\\\n\tfor (; rbtn_right_get(a_type, a_field, (r_node)) !=\t\t\\\n\t &(a_rbt)->rbt_nil; (r_node) = rbtn_right_get(a_type, a_field,\t\\\n\t (r_node))) {\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\trbtn_rotate_left(a_type, a_field, a_node, r_node) do {\t\t\\\n (r_node) = rbtn_right_get(a_type, a_field, (a_node));\t\t\\\n rbtn_right_set(a_type, a_field, (a_node),\t\t\t\t\\\n rbtn_left_get(a_type, a_field, (r_node)));\t\t\t\\\n rbtn_left_set(a_type, a_field, (r_node), (a_node));\t\t\t\\\n} while (0)\n\n#define\trbtn_rotate_right(a_type, a_field, a_node, r_node) do {\t\t\\\n (r_node) = rbtn_left_get(a_type, a_field, (a_node));\t\t\\\n rbtn_left_set(a_type, a_field, (a_node),\t\t\t\t\\\n rbtn_right_get(a_type, a_field, (r_node)));\t\t\t\\\n rbtn_right_set(a_type, a_field, (r_node), (a_node));\t\t\\\n} while (0)\n\n/*\n * The rb_proto() macro generates function prototypes that correspond to the\n * functions generated by an equivalently parameterized call to rb_gen().\n */\n\n#define\trb_proto(a_attr, a_prefix, a_rbt_type, a_type)\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##new(a_rbt_type *rbtree);\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##first(a_rbt_type *rbtree);\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##last(a_rbt_type *rbtree);\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##next(a_rbt_type *rbtree, a_type *node);\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##prev(a_rbt_type *rbtree, a_type *node);\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##search(a_rbt_type *rbtree, a_type *key);\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##nsearch(a_rbt_type *rbtree, a_type *key);\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##psearch(a_rbt_type *rbtree, a_type *key);\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##insert(a_rbt_type *rbtree, a_type *node);\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##remove(a_rbt_type *rbtree, a_type *node);\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##iter(a_rbt_type *rbtree, a_type *start, a_type *(*cb)(\t\\\n a_rbt_type *, a_type *, void *), void *arg);\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,\t\t\\\n a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg);\n\n/*\n * The rb_gen() macro generates a type-specific red-black tree implementation,\n * based on the above cpp macros.\n *\n * Arguments:\n *\n * a_attr : Function attribute for generated functions (ex: static).\n * a_prefix : Prefix for generated functions (ex: ex_).\n * a_rb_type : Type for red-black tree data structure (ex: ex_t).\n * a_type : Type for red-black tree node data structure (ex: ex_node_t).\n * a_field : Name of red-black tree node linkage (ex: ex_link).\n * a_cmp : Node comparison function name, with the following prototype:\n * int (a_cmp *)(a_type *a_node, a_type *a_other);\n * ^^^^^^\n * or a_key\n * Interpretation of comparision function return values:\n * -1 : a_node < a_other\n * 0 : a_node == a_other\n * 1 : a_node > a_other\n * In all cases, the a_node or a_key macro argument is the first\n * argument to the comparison function, which makes it possible\n * to write comparison functions that treat the first argument\n * specially.\n *\n * Assuming the following setup:\n *\n * typedef struct ex_node_s ex_node_t;\n * struct ex_node_s {\n * rb_node(ex_node_t) ex_link;\n * };\n * typedef rb_tree(ex_node_t) ex_t;\n * rb_gen(static, ex_, ex_t, ex_node_t, ex_link, ex_cmp)\n *\n * The following API is generated:\n *\n * static void\n * ex_new(ex_t *tree);\n * Description: Initialize a red-black tree structure.\n * Args:\n * tree: Pointer to an uninitialized red-black tree object.\n *\n * static ex_node_t *\n * ex_first(ex_t *tree);\n * static ex_node_t *\n * ex_last(ex_t *tree);\n * Description: Get the first/last node in tree.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * Ret: First/last node in tree, or NULL if tree is empty.\n *\n * static ex_node_t *\n * ex_next(ex_t *tree, ex_node_t *node);\n * static ex_node_t *\n * ex_prev(ex_t *tree, ex_node_t *node);\n * Description: Get node's successor/predecessor.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * node: A node in tree.\n * Ret: node's successor/predecessor in tree, or NULL if node is\n * last/first.\n *\n * static ex_node_t *\n * ex_search(ex_t *tree, ex_node_t *key);\n * Description: Search for node that matches key.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * key : Search key.\n * Ret: Node in tree that matches key, or NULL if no match.\n *\n * static ex_node_t *\n * ex_nsearch(ex_t *tree, ex_node_t *key);\n * static ex_node_t *\n * ex_psearch(ex_t *tree, ex_node_t *key);\n * Description: Search for node that matches key. If no match is found,\n * return what would be key's successor/predecessor, were\n * key in tree.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * key : Search key.\n * Ret: Node in tree that matches key, or if no match, hypothetical node's\n * successor/predecessor (NULL if no successor/predecessor).\n *\n * static void\n * ex_insert(ex_t *tree, ex_node_t *node);\n * Description: Insert node into tree.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * node: Node to be inserted into tree.\n *\n * static void\n * ex_remove(ex_t *tree, ex_node_t *node);\n * Description: Remove node from tree.\n * Args:\n * tree: Pointer to an initialized red-black tree object.\n * node: Node in tree to be removed.\n *\n * static ex_node_t *\n * ex_iter(ex_t *tree, ex_node_t *start, ex_node_t *(*cb)(ex_t *,\n * ex_node_t *, void *), void *arg);\n * static ex_node_t *\n * ex_reverse_iter(ex_t *tree, ex_node_t *start, ex_node *(*cb)(ex_t *,\n * ex_node_t *, void *), void *arg);\n * Description: Iterate forward/backward over tree, starting at node. If\n * tree is modified, iteration must be immediately\n * terminated by the callback function that causes the\n * modification.\n * Args:\n * tree : Pointer to an initialized red-black tree object.\n * start: Node at which to start iteration, or NULL to start at\n * first/last node.\n * cb : Callback function, which is called for each node during\n * iteration. Under normal circumstances the callback function\n * should return NULL, which causes iteration to continue. If a\n * callback function returns non-NULL, iteration is immediately\n * terminated and the non-NULL return value is returned by the\n * iterator. This is useful for re-starting iteration after\n * modifying tree.\n * arg : Opaque pointer passed to cb().\n * Ret: NULL if iteration completed, or the non-NULL callback return value\n * that caused termination of the iteration.\n */\n#define\trb_gen(a_attr, a_prefix, a_rbt_type, a_type, a_field, a_cmp)\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##new(a_rbt_type *rbtree) {\t\t\t\t\t\\\n rb_new(a_type, a_field, rbtree);\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##first(a_rbt_type *rbtree) {\t\t\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n rbtn_first(a_type, a_field, rbtree, rbtree->rbt_root, ret);\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = NULL;\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##last(a_rbt_type *rbtree) {\t\t\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n rbtn_last(a_type, a_field, rbtree, rbtree->rbt_root, ret);\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = NULL;\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##next(a_rbt_type *rbtree, a_type *node) {\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n if (rbtn_right_get(a_type, a_field, node) != &rbtree->rbt_nil) {\t\\\n\trbtn_first(a_type, a_field, rbtree, rbtn_right_get(a_type,\t\\\n\t a_field, node), ret);\t\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *tnode = rbtree->rbt_root;\t\t\t\t\\\n\tassert(tnode != &rbtree->rbt_nil);\t\t\t\t\\\n\tret = &rbtree->rbt_nil;\t\t\t\t\t\t\\\n\twhile (true) {\t\t\t\t\t\t\t\\\n\t int cmp = (a_cmp)(node, tnode);\t\t\t\t\\\n\t if (cmp < 0) {\t\t\t\t\t\t\\\n\t\tret = tnode;\t\t\t\t\t\t\\\n\t\ttnode = rbtn_left_get(a_type, a_field, tnode);\t\t\\\n\t } else if (cmp > 0) {\t\t\t\t\t\\\n\t\ttnode = rbtn_right_get(a_type, a_field, tnode);\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t assert(tnode != &rbtree->rbt_nil);\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = (NULL);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##prev(a_rbt_type *rbtree, a_type *node) {\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n if (rbtn_left_get(a_type, a_field, node) != &rbtree->rbt_nil) {\t\\\n\trbtn_last(a_type, a_field, rbtree, rbtn_left_get(a_type,\t\\\n\t a_field, node), ret);\t\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *tnode = rbtree->rbt_root;\t\t\t\t\\\n\tassert(tnode != &rbtree->rbt_nil);\t\t\t\t\\\n\tret = &rbtree->rbt_nil;\t\t\t\t\t\t\\\n\twhile (true) {\t\t\t\t\t\t\t\\\n\t int cmp = (a_cmp)(node, tnode);\t\t\t\t\\\n\t if (cmp < 0) {\t\t\t\t\t\t\\\n\t\ttnode = rbtn_left_get(a_type, a_field, tnode);\t\t\\\n\t } else if (cmp > 0) {\t\t\t\t\t\\\n\t\tret = tnode;\t\t\t\t\t\t\\\n\t\ttnode = rbtn_right_get(a_type, a_field, tnode);\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t assert(tnode != &rbtree->rbt_nil);\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = (NULL);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##search(a_rbt_type *rbtree, a_type *key) {\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n int cmp;\t\t\t\t\t\t\t\t\\\n ret = rbtree->rbt_root;\t\t\t\t\t\t\\\n while (ret != &rbtree->rbt_nil\t\t\t\t\t\\\n && (cmp = (a_cmp)(key, ret)) != 0) {\t\t\t\t\\\n\tif (cmp < 0) {\t\t\t\t\t\t\t\\\n\t ret = rbtn_left_get(a_type, a_field, ret);\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t ret = rbtn_right_get(a_type, a_field, ret);\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = (NULL);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##nsearch(a_rbt_type *rbtree, a_type *key) {\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n a_type *tnode = rbtree->rbt_root;\t\t\t\t\t\\\n ret = &rbtree->rbt_nil;\t\t\t\t\t\t\\\n while (tnode != &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tint cmp = (a_cmp)(key, tnode);\t\t\t\t\t\\\n\tif (cmp < 0) {\t\t\t\t\t\t\t\\\n\t ret = tnode;\t\t\t\t\t\t\\\n\t tnode = rbtn_left_get(a_type, a_field, tnode);\t\t\\\n\t} else if (cmp > 0) {\t\t\t\t\t\t\\\n\t tnode = rbtn_right_get(a_type, a_field, tnode);\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t ret = tnode;\t\t\t\t\t\t\\\n\t break;\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = (NULL);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##psearch(a_rbt_type *rbtree, a_type *key) {\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n a_type *tnode = rbtree->rbt_root;\t\t\t\t\t\\\n ret = &rbtree->rbt_nil;\t\t\t\t\t\t\\\n while (tnode != &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tint cmp = (a_cmp)(key, tnode);\t\t\t\t\t\\\n\tif (cmp < 0) {\t\t\t\t\t\t\t\\\n\t tnode = rbtn_left_get(a_type, a_field, tnode);\t\t\\\n\t} else if (cmp > 0) {\t\t\t\t\t\t\\\n\t ret = tnode;\t\t\t\t\t\t\\\n\t tnode = rbtn_right_get(a_type, a_field, tnode);\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t ret = tnode;\t\t\t\t\t\t\\\n\t break;\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = (NULL);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##insert(a_rbt_type *rbtree, a_type *node) {\t\t\t\\\n struct {\t\t\t\t\t\t\t\t\\\n\ta_type *node;\t\t\t\t\t\t\t\\\n\tint cmp;\t\t\t\t\t\t\t\\\n } path[sizeof(void *) << 4], *pathp;\t\t\t\t\\\n rbt_node_new(a_type, a_field, rbtree, node);\t\t\t\\\n /* Wind. */\t\t\t\t\t\t\t\t\\\n path->node = rbtree->rbt_root;\t\t\t\t\t\\\n for (pathp = path; pathp->node != &rbtree->rbt_nil; pathp++) {\t\\\n\tint cmp = pathp->cmp = a_cmp(node, pathp->node);\t\t\\\n\tassert(cmp != 0);\t\t\t\t\t\t\\\n\tif (cmp < 0) {\t\t\t\t\t\t\t\\\n\t pathp[1].node = rbtn_left_get(a_type, a_field,\t\t\\\n\t pathp->node);\t\t\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t pathp[1].node = rbtn_right_get(a_type, a_field,\t\t\\\n\t pathp->node);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n pathp->node = node;\t\t\t\t\t\t\t\\\n /* Unwind. */\t\t\t\t\t\t\t\\\n for (pathp--; (uintptr_t)pathp >= (uintptr_t)path; pathp--) {\t\\\n\ta_type *cnode = pathp->node;\t\t\t\t\t\\\n\tif (pathp->cmp < 0) {\t\t\t\t\t\t\\\n\t a_type *left = pathp[1].node;\t\t\t\t\\\n\t rbtn_left_set(a_type, a_field, cnode, left);\t\t\\\n\t if (rbtn_red_get(a_type, a_field, left)) {\t\t\t\\\n\t\ta_type *leftleft = rbtn_left_get(a_type, a_field, left);\\\n\t\tif (rbtn_red_get(a_type, a_field, leftleft)) {\t\t\\\n\t\t /* Fix up 4-node. */\t\t\t\t\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, leftleft);\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, cnode, tnode);\t\\\n\t\t cnode = tnode;\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\treturn;\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t a_type *right = pathp[1].node;\t\t\t\t\\\n\t rbtn_right_set(a_type, a_field, cnode, right);\t\t\\\n\t if (rbtn_red_get(a_type, a_field, right)) {\t\t\t\\\n\t\ta_type *left = rbtn_left_get(a_type, a_field, cnode);\t\\\n\t\tif (rbtn_red_get(a_type, a_field, left)) {\t\t\\\n\t\t /* Split 4-node. */\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, left);\t\t\\\n\t\t rbtn_black_set(a_type, a_field, right);\t\t\\\n\t\t rbtn_red_set(a_type, a_field, cnode);\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* Lean left. */\t\t\t\t\t\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t bool tred = rbtn_red_get(a_type, a_field, cnode);\t\\\n\t\t rbtn_rotate_left(a_type, a_field, cnode, tnode);\t\\\n\t\t rbtn_color_set(a_type, a_field, tnode, tred);\t\\\n\t\t rbtn_red_set(a_type, a_field, cnode);\t\t\\\n\t\t cnode = tnode;\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\treturn;\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tpathp->node = cnode;\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n /* Set root, and make it black. */\t\t\t\t\t\\\n rbtree->rbt_root = path->node;\t\t\t\t\t\\\n rbtn_black_set(a_type, a_field, rbtree->rbt_root);\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_prefix##remove(a_rbt_type *rbtree, a_type *node) {\t\t\t\\\n struct {\t\t\t\t\t\t\t\t\\\n\ta_type *node;\t\t\t\t\t\t\t\\\n\tint cmp;\t\t\t\t\t\t\t\\\n } *pathp, *nodep, path[sizeof(void *) << 4];\t\t\t\\\n /* Wind. */\t\t\t\t\t\t\t\t\\\n nodep = NULL; /* Silence compiler warning. */\t\t\t\\\n path->node = rbtree->rbt_root;\t\t\t\t\t\\\n for (pathp = path; pathp->node != &rbtree->rbt_nil; pathp++) {\t\\\n\tint cmp = pathp->cmp = a_cmp(node, pathp->node);\t\t\\\n\tif (cmp < 0) {\t\t\t\t\t\t\t\\\n\t pathp[1].node = rbtn_left_get(a_type, a_field,\t\t\\\n\t pathp->node);\t\t\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t pathp[1].node = rbtn_right_get(a_type, a_field,\t\t\\\n\t pathp->node);\t\t\t\t\t\t\\\n\t if (cmp == 0) {\t\t\t\t\t\t\\\n\t /* Find node's successor, in preparation for swap. */\t\\\n\t\tpathp->cmp = 1;\t\t\t\t\t\t\\\n\t\tnodep = pathp;\t\t\t\t\t\t\\\n\t\tfor (pathp++; pathp->node != &rbtree->rbt_nil;\t\t\\\n\t\t pathp++) {\t\t\t\t\t\t\\\n\t\t pathp->cmp = -1;\t\t\t\t\t\\\n\t\t pathp[1].node = rbtn_left_get(a_type, a_field,\t\\\n\t\t pathp->node);\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n assert(nodep->node == node);\t\t\t\t\t\\\n pathp--;\t\t\t\t\t\t\t\t\\\n if (pathp->node != node) {\t\t\t\t\t\t\\\n\t/* Swap node with its successor. */\t\t\t\t\\\n\tbool tred = rbtn_red_get(a_type, a_field, pathp->node);\t\t\\\n\trbtn_color_set(a_type, a_field, pathp->node,\t\t\t\\\n\t rbtn_red_get(a_type, a_field, node));\t\t\t\t\\\n\trbtn_left_set(a_type, a_field, pathp->node,\t\t\t\\\n\t rbtn_left_get(a_type, a_field, node));\t\t\t\\\n\t/* If node's successor is its right child, the following code */\\\n\t/* will do the wrong thing for the right child pointer. */\\\n\t/* However, it doesn't matter, because the pointer will be */\\\n\t/* properly set when the successor is pruned. */\\\n\trbtn_right_set(a_type, a_field, pathp->node,\t\t\t\\\n\t rbtn_right_get(a_type, a_field, node));\t\t\t\\\n\trbtn_color_set(a_type, a_field, node, tred);\t\t\t\\\n\t/* The pruned leaf node's child pointers are never accessed */\\\n\t/* again, so don't bother setting them to nil. */\\\n\tnodep->node = pathp->node;\t\t\t\t\t\\\n\tpathp->node = node;\t\t\t\t\t\t\\\n\tif (nodep == path) {\t\t\t\t\t\t\\\n\t rbtree->rbt_root = nodep->node;\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t if (nodep[-1].cmp < 0) {\t\t\t\t\t\\\n\t\trbtn_left_set(a_type, a_field, nodep[-1].node,\t\t\\\n\t\t nodep->node);\t\t\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\trbtn_right_set(a_type, a_field, nodep[-1].node,\t\t\\\n\t\t nodep->node);\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *left = rbtn_left_get(a_type, a_field, node);\t\t\\\n\tif (left != &rbtree->rbt_nil) {\t\t\t\t\t\\\n\t /* node has no successor, but it has a left child. */\\\n\t /* Splice node out, without losing the left child. */\\\n\t assert(rbtn_red_get(a_type, a_field, node) == false);\t\\\n\t assert(rbtn_red_get(a_type, a_field, left));\t\t\\\n\t rbtn_black_set(a_type, a_field, left);\t\t\t\\\n\t if (pathp == path) {\t\t\t\t\t\\\n\t\trbtree->rbt_root = left;\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\tif (pathp[-1].cmp < 0) {\t\t\t\t\\\n\t\t rbtn_left_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t left);\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t rbtn_right_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t left);\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t return;\t\t\t\t\t\t\t\\\n\t} else if (pathp == path) {\t\t\t\t\t\\\n\t /* The tree only contained one node. */\t\t\t\\\n\t rbtree->rbt_root = &rbtree->rbt_nil;\t\t\t\\\n\t return;\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (rbtn_red_get(a_type, a_field, pathp->node)) {\t\t\t\\\n\t/* Prune red node, which requires no fixup. */\t\t\t\\\n\tassert(pathp[-1].cmp < 0);\t\t\t\t\t\\\n\trbtn_left_set(a_type, a_field, pathp[-1].node,\t\t\t\\\n\t &rbtree->rbt_nil);\t\t\t\t\t\t\\\n\treturn;\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n /* The node to be pruned is black, so unwind until balance is */\\\n /* restored. */\\\n pathp->node = &rbtree->rbt_nil;\t\t\t\t\t\\\n for (pathp--; (uintptr_t)pathp >= (uintptr_t)path; pathp--) {\t\\\n\tassert(pathp->cmp != 0);\t\t\t\t\t\\\n\tif (pathp->cmp < 0) {\t\t\t\t\t\t\\\n\t rbtn_left_set(a_type, a_field, pathp->node,\t\t\t\\\n\t pathp[1].node);\t\t\t\t\t\t\\\n\t assert(rbtn_red_get(a_type, a_field, pathp[1].node)\t\t\\\n\t == false);\t\t\t\t\t\t\\\n\t if (rbtn_red_get(a_type, a_field, pathp->node)) {\t\t\\\n\t\ta_type *right = rbtn_right_get(a_type, a_field,\t\t\\\n\t\t pathp->node);\t\t\t\t\t\t\\\n\t\ta_type *rightleft = rbtn_left_get(a_type, a_field,\t\\\n\t\t right);\t\t\t\t\t\t\\\n\t\ta_type *tnode;\t\t\t\t\t\t\\\n\t\tif (rbtn_red_get(a_type, a_field, rightleft)) {\t\t\\\n\t\t /* In the following diagrams, ||, //, and \\\\ */\\\n\t\t /* indicate the path to the removed node. */\\\n\t\t /* */\\\n\t\t /* || */\\\n\t\t /* pathp(r) */\\\n\t\t /* // \\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (r) */\\\n\t\t /* */\\\n\t\t rbtn_black_set(a_type, a_field, pathp->node);\t\\\n\t\t rbtn_rotate_right(a_type, a_field, right, tnode);\t\\\n\t\t rbtn_right_set(a_type, a_field, pathp->node, tnode);\\\n\t\t rbtn_rotate_left(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(r) */\\\n\t\t /* // \\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (b) */\\\n\t\t /* */\\\n\t\t rbtn_rotate_left(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\t/* Balance restored, but rotation modified subtree */\\\n\t\t/* root. */\\\n\t\tassert((uintptr_t)pathp > (uintptr_t)path);\t\t\\\n\t\tif (pathp[-1].cmp < 0) {\t\t\t\t\\\n\t\t rbtn_left_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t rbtn_right_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\treturn;\t\t\t\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\ta_type *right = rbtn_right_get(a_type, a_field,\t\t\\\n\t\t pathp->node);\t\t\t\t\t\t\\\n\t\ta_type *rightleft = rbtn_left_get(a_type, a_field,\t\\\n\t\t right);\t\t\t\t\t\t\\\n\t\tif (rbtn_red_get(a_type, a_field, rightleft)) {\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* // \\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (r) */\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, rightleft);\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, right, tnode);\t\\\n\t\t rbtn_right_set(a_type, a_field, pathp->node, tnode);\\\n\t\t rbtn_rotate_left(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t /* Balance restored, but rotation modified */\\\n\t\t /* subree root, which may actually be the tree */\\\n\t\t /* root. */\\\n\t\t if (pathp == path) {\t\t\t\t\\\n\t\t\t/* Set root. */\t\t\t\t\t\\\n\t\t\trbtree->rbt_root = tnode;\t\t\t\\\n\t\t } else {\t\t\t\t\t\t\\\n\t\t\tif (pathp[-1].cmp < 0) {\t\t\t\\\n\t\t\t rbtn_left_set(a_type, a_field,\t\t\\\n\t\t\t pathp[-1].node, tnode);\t\t\t\\\n\t\t\t} else {\t\t\t\t\t\\\n\t\t\t rbtn_right_set(a_type, a_field,\t\t\\\n\t\t\t pathp[-1].node, tnode);\t\t\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t }\t\t\t\t\t\t\t\\\n\t\t return;\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* // \\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (b) */\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t rbtn_red_set(a_type, a_field, pathp->node);\t\t\\\n\t\t rbtn_rotate_left(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t pathp->node = tnode;\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t} else {\t\t\t\t\t\t\t\\\n\t a_type *left;\t\t\t\t\t\t\\\n\t rbtn_right_set(a_type, a_field, pathp->node,\t\t\\\n\t pathp[1].node);\t\t\t\t\t\t\\\n\t left = rbtn_left_get(a_type, a_field, pathp->node);\t\t\\\n\t if (rbtn_red_get(a_type, a_field, left)) {\t\t\t\\\n\t\ta_type *tnode;\t\t\t\t\t\t\\\n\t\ta_type *leftright = rbtn_right_get(a_type, a_field,\t\\\n\t\t left);\t\t\t\t\t\t\\\n\t\ta_type *leftrightleft = rbtn_left_get(a_type, a_field,\t\\\n\t\t leftright);\t\t\t\t\t\t\\\n\t\tif (rbtn_red_get(a_type, a_field, leftrightleft)) {\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (r) (b) */\\\n\t\t /* \\ */\\\n\t\t /* (b) */\\\n\t\t /* / */\\\n\t\t /* (r) */\\\n\t\t a_type *unode;\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, leftrightleft);\t\\\n\t\t rbtn_rotate_right(a_type, a_field, pathp->node,\t\\\n\t\t unode);\t\t\t\t\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t rbtn_right_set(a_type, a_field, unode, tnode);\t\\\n\t\t rbtn_rotate_left(a_type, a_field, unode, tnode);\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (r) (b) */\\\n\t\t /* \\ */\\\n\t\t /* (b) */\\\n\t\t /* / */\\\n\t\t /* (b) */\\\n\t\t assert(leftright != &rbtree->rbt_nil);\t\t\\\n\t\t rbtn_red_set(a_type, a_field, leftright);\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, tnode);\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\t/* Balance restored, but rotation modified subtree */\\\n\t\t/* root, which may actually be the tree root. */\\\n\t\tif (pathp == path) {\t\t\t\t\t\\\n\t\t /* Set root. */\t\t\t\t\t\\\n\t\t rbtree->rbt_root = tnode;\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t if (pathp[-1].cmp < 0) {\t\t\t\t\\\n\t\t\trbtn_left_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t\t tnode);\t\t\t\t\t\\\n\t\t } else {\t\t\t\t\t\t\\\n\t\t\trbtn_right_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t\t tnode);\t\t\t\t\t\\\n\t\t }\t\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\treturn;\t\t\t\t\t\t\t\\\n\t } else if (rbtn_red_get(a_type, a_field, pathp->node)) {\t\\\n\t\ta_type *leftleft = rbtn_left_get(a_type, a_field, left);\\\n\t\tif (rbtn_red_get(a_type, a_field, leftleft)) {\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(r) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (r) */\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, pathp->node);\t\\\n\t\t rbtn_red_set(a_type, a_field, left);\t\t\\\n\t\t rbtn_black_set(a_type, a_field, leftleft);\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t /* Balance restored, but rotation modified */\\\n\t\t /* subtree root. */\\\n\t\t assert((uintptr_t)pathp > (uintptr_t)path);\t\t\\\n\t\t if (pathp[-1].cmp < 0) {\t\t\t\t\\\n\t\t\trbtn_left_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t\t tnode);\t\t\t\t\t\\\n\t\t } else {\t\t\t\t\t\t\\\n\t\t\trbtn_right_set(a_type, a_field, pathp[-1].node,\t\\\n\t\t\t tnode);\t\t\t\t\t\\\n\t\t }\t\t\t\t\t\t\t\\\n\t\t return;\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(r) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (b) */\\\n\t\t rbtn_red_set(a_type, a_field, left);\t\t\\\n\t\t rbtn_black_set(a_type, a_field, pathp->node);\t\\\n\t\t /* Balance restored. */\t\t\t\t\\\n\t\t return;\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t } else {\t\t\t\t\t\t\t\\\n\t\ta_type *leftleft = rbtn_left_get(a_type, a_field, left);\\\n\t\tif (rbtn_red_get(a_type, a_field, leftleft)) {\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (r) */\\\n\t\t a_type *tnode;\t\t\t\t\t\\\n\t\t rbtn_black_set(a_type, a_field, leftleft);\t\t\\\n\t\t rbtn_rotate_right(a_type, a_field, pathp->node,\t\\\n\t\t tnode);\t\t\t\t\t\t\\\n\t\t /* Balance restored, but rotation modified */\\\n\t\t /* subtree root, which may actually be the tree */\\\n\t\t /* root. */\\\n\t\t if (pathp == path) {\t\t\t\t\\\n\t\t\t/* Set root. */\t\t\t\t\t\\\n\t\t\trbtree->rbt_root = tnode;\t\t\t\\\n\t\t } else {\t\t\t\t\t\t\\\n\t\t\tif (pathp[-1].cmp < 0) {\t\t\t\\\n\t\t\t rbtn_left_set(a_type, a_field,\t\t\\\n\t\t\t pathp[-1].node, tnode);\t\t\t\\\n\t\t\t} else {\t\t\t\t\t\\\n\t\t\t rbtn_right_set(a_type, a_field,\t\t\\\n\t\t\t pathp[-1].node, tnode);\t\t\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t }\t\t\t\t\t\t\t\\\n\t\t return;\t\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t /* || */\\\n\t\t /* pathp(b) */\\\n\t\t /* / \\\\ */\\\n\t\t /* (b) (b) */\\\n\t\t /* / */\\\n\t\t /* (b) */\\\n\t\t rbtn_red_set(a_type, a_field, left);\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t }\t\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n /* Set root. */\t\t\t\t\t\t\t\\\n rbtree->rbt_root = path->node;\t\t\t\t\t\\\n assert(rbtn_red_get(a_type, a_field, rbtree->rbt_root) == false);\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##iter_recurse(a_rbt_type *rbtree, a_type *node,\t\t\\\n a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {\t\t\\\n if (node == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\treturn (&rbtree->rbt_nil);\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = a_prefix##iter_recurse(rbtree, rbtn_left_get(a_type,\t\\\n\t a_field, node), cb, arg)) != &rbtree->rbt_nil\t\t\t\\\n\t || (ret = cb(rbtree, node, arg)) != NULL) {\t\t\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##iter_recurse(rbtree, rbtn_right_get(a_type,\t\\\n\t a_field, node), cb, arg));\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##iter_start(a_rbt_type *rbtree, a_type *start, a_type *node,\t\\\n a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {\t\t\\\n int cmp = a_cmp(start, node);\t\t\t\t\t\\\n if (cmp < 0) {\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = a_prefix##iter_start(rbtree, start,\t\t\t\\\n\t rbtn_left_get(a_type, a_field, node), cb, arg)) !=\t\t\\\n\t &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##iter_recurse(rbtree, rbtn_right_get(a_type,\t\\\n\t a_field, node), cb, arg));\t\t\t\t\t\\\n } else if (cmp > 0) {\t\t\t\t\t\t\\\n\treturn (a_prefix##iter_start(rbtree, start,\t\t\t\\\n\t rbtn_right_get(a_type, a_field, node), cb, arg));\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = cb(rbtree, node, arg)) != NULL) {\t\t\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##iter_recurse(rbtree, rbtn_right_get(a_type,\t\\\n\t a_field, node), cb, arg));\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##iter(a_rbt_type *rbtree, a_type *start, a_type *(*cb)(\t\\\n a_rbt_type *, a_type *, void *), void *arg) {\t\t\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n if (start != NULL) {\t\t\t\t\t\t\\\n\tret = a_prefix##iter_start(rbtree, start, rbtree->rbt_root,\t\\\n\t cb, arg);\t\t\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\tret = a_prefix##iter_recurse(rbtree, rbtree->rbt_root, cb, arg);\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = NULL;\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##reverse_iter_recurse(a_rbt_type *rbtree, a_type *node,\t\\\n a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {\t\t\\\n if (node == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\treturn (&rbtree->rbt_nil);\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = a_prefix##reverse_iter_recurse(rbtree,\t\t\\\n\t rbtn_right_get(a_type, a_field, node), cb, arg)) !=\t\t\\\n\t &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##reverse_iter_recurse(rbtree,\t\t\t\\\n\t rbtn_left_get(a_type, a_field, node), cb, arg));\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##reverse_iter_start(a_rbt_type *rbtree, a_type *start,\t\t\\\n a_type *node, a_type *(*cb)(a_rbt_type *, a_type *, void *),\t\t\\\n void *arg) {\t\t\t\t\t\t\t\t\\\n int cmp = a_cmp(start, node);\t\t\t\t\t\\\n if (cmp > 0) {\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = a_prefix##reverse_iter_start(rbtree, start,\t\t\\\n\t rbtn_right_get(a_type, a_field, node), cb, arg)) !=\t\t\\\n\t &rbtree->rbt_nil || (ret = cb(rbtree, node, arg)) != NULL) {\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##reverse_iter_recurse(rbtree,\t\t\t\\\n\t rbtn_left_get(a_type, a_field, node), cb, arg));\t\t\\\n } else if (cmp < 0) {\t\t\t\t\t\t\\\n\treturn (a_prefix##reverse_iter_start(rbtree, start,\t\t\\\n\t rbtn_left_get(a_type, a_field, node), cb, arg));\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\ta_type *ret;\t\t\t\t\t\t\t\\\n\tif ((ret = cb(rbtree, node, arg)) != NULL) {\t\t\t\\\n\t return (ret);\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_prefix##reverse_iter_recurse(rbtree,\t\t\t\\\n\t rbtn_left_get(a_type, a_field, node), cb, arg));\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start,\t\t\\\n a_type *(*cb)(a_rbt_type *, a_type *, void *), void *arg) {\t\t\\\n a_type *ret;\t\t\t\t\t\t\t\\\n if (start != NULL) {\t\t\t\t\t\t\\\n\tret = a_prefix##reverse_iter_start(rbtree, start,\t\t\\\n\t rbtree->rbt_root, cb, arg);\t\t\t\t\t\\\n } else {\t\t\t\t\t\t\t\t\\\n\tret = a_prefix##reverse_iter_recurse(rbtree, rbtree->rbt_root,\t\\\n\t cb, arg);\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n if (ret == &rbtree->rbt_nil) {\t\t\t\t\t\\\n\tret = NULL;\t\t\t\t\t\t\t\\\n }\t\t\t\t\t\t\t\t\t\\\n return (ret);\t\t\t\t\t\t\t\\\n}\n\n#endif /* RB_H_ */\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/rtree.h", "content": "/*\n * This radix tree implementation is tailored to the singular purpose of\n * tracking which chunks are currently owned by jemalloc. This functionality\n * is mandatory for OS X, where jemalloc must be able to respond to object\n * ownership queries.\n *\n *******************************************************************************\n */\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct rtree_s rtree_t;\n\n/*\n * Size of each radix tree node (must be a power of 2). This impacts tree\n * depth.\n */\n#if (LG_SIZEOF_PTR == 2)\n# define RTREE_NODESIZE (1U << 14)\n#else\n# define RTREE_NODESIZE CACHELINE\n#endif\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct rtree_s {\n\tmalloc_mutex_t\tmutex;\n\tvoid\t\t**root;\n\tunsigned\theight;\n\tunsigned\tlevel2bits[1]; /* Dynamically sized. */\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nrtree_t\t*rtree_new(unsigned bits);\nvoid\trtree_prefork(rtree_t *rtree);\nvoid\trtree_postfork_parent(rtree_t *rtree);\nvoid\trtree_postfork_child(rtree_t *rtree);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\n#ifndef JEMALLOC_DEBUG\nvoid\t*rtree_get_locked(rtree_t *rtree, uintptr_t key);\n#endif\nvoid\t*rtree_get(rtree_t *rtree, uintptr_t key);\nbool\trtree_set(rtree_t *rtree, uintptr_t key, void *val);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_RTREE_C_))\n#define\tRTREE_GET_GENERATE(f)\t\t\t\t\t\t\\\n/* The least significant bits of the key are ignored. */\t\t\\\nJEMALLOC_INLINE void *\t\t\t\t\t\t\t\\\nf(rtree_t *rtree, uintptr_t key)\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tvoid *ret;\t\t\t\t\t\t\t\\\n\tuintptr_t subkey;\t\t\t\t\t\t\\\n\tunsigned i, lshift, height, bits;\t\t\t\t\\\n\tvoid **node, **child;\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tRTREE_LOCK(&rtree->mutex);\t\t\t\t\t\\\n\tfor (i = lshift = 0, height = rtree->height, node = rtree->root;\\\n\t i < height - 1;\t\t\t\t\t\t\\\n\t i++, lshift += bits, node = child) {\t\t\t\\\n\t\tbits = rtree->level2bits[i];\t\t\t\t\\\n\t\tsubkey = (key << lshift) >> ((ZU(1) << (LG_SIZEOF_PTR + \\\n\t\t 3)) - bits);\t\t\t\t\t\\\n\t\tchild = (void**)node[subkey];\t\t\t\t\\\n\t\tif (child == NULL) {\t\t\t\t\t\\\n\t\t\tRTREE_UNLOCK(&rtree->mutex);\t\t\t\\\n\t\t\treturn (NULL);\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\t/*\t\t\t\t\t\t\t\t\\\n\t * node is a leaf, so it contains values rather than node\t\\\n\t * pointers.\t\t\t\t\t\t\t\\\n\t */\t\t\t\t\t\t\t\t\\\n\tbits = rtree->level2bits[i];\t\t\t\t\t\\\n\tsubkey = (key << lshift) >> ((ZU(1) << (LG_SIZEOF_PTR+3)) -\t\\\n\t bits);\t\t\t\t\t\t\t\\\n\tret = node[subkey];\t\t\t\t\t\t\\\n\tRTREE_UNLOCK(&rtree->mutex);\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tRTREE_GET_VALIDATE\t\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n#ifdef JEMALLOC_DEBUG\n# define RTREE_LOCK(l)\t\tmalloc_mutex_lock(l)\n# define RTREE_UNLOCK(l)\tmalloc_mutex_unlock(l)\n# define RTREE_GET_VALIDATE\nRTREE_GET_GENERATE(rtree_get_locked)\n# undef RTREE_LOCK\n# undef RTREE_UNLOCK\n# undef RTREE_GET_VALIDATE\n#endif\n\n#define\tRTREE_LOCK(l)\n#define\tRTREE_UNLOCK(l)\n#ifdef JEMALLOC_DEBUG\n /*\n * Suppose that it were possible for a jemalloc-allocated chunk to be\n * munmap()ped, followed by a different allocator in another thread re-using\n * overlapping virtual memory, all without invalidating the cached rtree\n * value. The result would be a false positive (the rtree would claim that\n * jemalloc owns memory that it had actually discarded). This scenario\n * seems impossible, but the following assertion is a prudent sanity check.\n */\n# define RTREE_GET_VALIDATE\t\t\t\t\t\t\\\n\tassert(rtree_get_locked(rtree, key) == ret);\n#else\n# define RTREE_GET_VALIDATE\n#endif\nRTREE_GET_GENERATE(rtree_get)\n#undef RTREE_LOCK\n#undef RTREE_UNLOCK\n#undef RTREE_GET_VALIDATE\n\nJEMALLOC_INLINE bool\nrtree_set(rtree_t *rtree, uintptr_t key, void *val)\n{\n\tuintptr_t subkey;\n\tunsigned i, lshift, height, bits;\n\tvoid **node, **child;\n\n\tmalloc_mutex_lock(&rtree->mutex);\n\tfor (i = lshift = 0, height = rtree->height, node = rtree->root;\n\t i < height - 1;\n\t i++, lshift += bits, node = child) {\n\t\tbits = rtree->level2bits[i];\n\t\tsubkey = (key << lshift) >> ((ZU(1) << (LG_SIZEOF_PTR+3)) -\n\t\t bits);\n\t\tchild = (void**)node[subkey];\n\t\tif (child == NULL) {\n\t\t\tchild = (void**)base_alloc(sizeof(void *) <<\n\t\t\t rtree->level2bits[i+1]);\n\t\t\tif (child == NULL) {\n\t\t\t\tmalloc_mutex_unlock(&rtree->mutex);\n\t\t\t\treturn (true);\n\t\t\t}\n\t\t\tmemset(child, 0, sizeof(void *) <<\n\t\t\t rtree->level2bits[i+1]);\n\t\t\tnode[subkey] = child;\n\t\t}\n\t}\n\n\t/* node is a leaf, so it contains values rather than node pointers. */\n\tbits = rtree->level2bits[i];\n\tsubkey = (key << lshift) >> ((ZU(1) << (LG_SIZEOF_PTR+3)) - bits);\n\tnode[subkey] = val;\n\tmalloc_mutex_unlock(&rtree->mutex);\n\n\treturn (false);\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/size_classes.sh", "content": "#!/bin/sh\n\n# The following limits are chosen such that they cover all supported platforms.\n\n# Range of quanta.\nlg_qmin=3\nlg_qmax=4\n\n# The range of tiny size classes is [2^lg_tmin..2^(lg_q-1)].\nlg_tmin=3\n\n# Range of page sizes.\nlg_pmin=12\nlg_pmax=16\n\npow2() {\n e=$1\n pow2_result=1\n while [ ${e} -gt 0 ] ; do\n pow2_result=$((${pow2_result} + ${pow2_result}))\n e=$((${e} - 1))\n done\n}\n\ncat < 255)\n# error \"Too many small size classes\"\n#endif\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\nEOF\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/stats.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct tcache_bin_stats_s tcache_bin_stats_t;\ntypedef struct malloc_bin_stats_s malloc_bin_stats_t;\ntypedef struct malloc_large_stats_s malloc_large_stats_t;\ntypedef struct arena_stats_s arena_stats_t;\ntypedef struct chunk_stats_s chunk_stats_t;\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\nstruct tcache_bin_stats_s {\n\t/*\n\t * Number of allocation requests that corresponded to the size of this\n\t * bin.\n\t */\n\tuint64_t\tnrequests;\n};\n\nstruct malloc_bin_stats_s {\n\t/*\n\t * Current number of bytes allocated, including objects currently\n\t * cached by tcache.\n\t */\n\tsize_t\t\tallocated;\n\n\t/*\n\t * Total number of allocation/deallocation requests served directly by\n\t * the bin. Note that tcache may allocate an object, then recycle it\n\t * many times, resulting many increments to nrequests, but only one\n\t * each to nmalloc and ndalloc.\n\t */\n\tuint64_t\tnmalloc;\n\tuint64_t\tndalloc;\n\n\t/*\n\t * Number of allocation requests that correspond to the size of this\n\t * bin. This includes requests served by tcache, though tcache only\n\t * periodically merges into this counter.\n\t */\n\tuint64_t\tnrequests;\n\n\t/* Number of tcache fills from this bin. */\n\tuint64_t\tnfills;\n\n\t/* Number of tcache flushes to this bin. */\n\tuint64_t\tnflushes;\n\n\t/* Total number of runs created for this bin's size class. */\n\tuint64_t\tnruns;\n\n\t/*\n\t * Total number of runs reused by extracting them from the runs tree for\n\t * this bin's size class.\n\t */\n\tuint64_t\treruns;\n\n\t/* Current number of runs in this bin. */\n\tsize_t\t\tcurruns;\n};\n\nstruct malloc_large_stats_s {\n\t/*\n\t * Total number of allocation/deallocation requests served directly by\n\t * the arena. Note that tcache may allocate an object, then recycle it\n\t * many times, resulting many increments to nrequests, but only one\n\t * each to nmalloc and ndalloc.\n\t */\n\tuint64_t\tnmalloc;\n\tuint64_t\tndalloc;\n\n\t/*\n\t * Number of allocation requests that correspond to this size class.\n\t * This includes requests served by tcache, though tcache only\n\t * periodically merges into this counter.\n\t */\n\tuint64_t\tnrequests;\n\n\t/* Current number of runs of this size class. */\n\tsize_t\t\tcurruns;\n};\n\nstruct arena_stats_s {\n\t/* Number of bytes currently mapped. */\n\tsize_t\t\tmapped;\n\n\t/*\n\t * Total number of purge sweeps, total number of madvise calls made,\n\t * and total pages purged in order to keep dirty unused memory under\n\t * control.\n\t */\n\tuint64_t\tnpurge;\n\tuint64_t\tnmadvise;\n\tuint64_t\tpurged;\n\n\t/* Per-size-category statistics. */\n\tsize_t\t\tallocated_large;\n\tuint64_t\tnmalloc_large;\n\tuint64_t\tndalloc_large;\n\tuint64_t\tnrequests_large;\n\n\t/*\n\t * One element for each possible size class, including sizes that\n\t * overlap with bin size classes. This is necessary because ipalloc()\n\t * sometimes has to use such large objects in order to assure proper\n\t * alignment.\n\t */\n\tmalloc_large_stats_t\t*lstats;\n};\n\nstruct chunk_stats_s {\n\t/* Number of chunks that were allocated. */\n\tuint64_t\tnchunks;\n\n\t/* High-water mark for number of chunks allocated. */\n\tsize_t\t\thighchunks;\n\n\t/*\n\t * Current number of chunks allocated. This value isn't maintained for\n\t * any other purpose, so keep track of it in order to be able to set\n\t * highchunks.\n\t */\n\tsize_t\t\tcurchunks;\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nextern bool\topt_stats_print;\n\nextern size_t\tstats_cactive;\n\nvoid\tstats_print(void (*write)(void *, const char *), void *cbopaque,\n const char *opts);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nsize_t\tstats_cactive_get(void);\nvoid\tstats_cactive_add(size_t size);\nvoid\tstats_cactive_sub(size_t size);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_STATS_C_))\nJEMALLOC_INLINE size_t\nstats_cactive_get(void)\n{\n\n\treturn (atomic_read_z(&stats_cactive));\n}\n\nJEMALLOC_INLINE void\nstats_cactive_add(size_t size)\n{\n\n\tatomic_add_z(&stats_cactive, size);\n}\n\nJEMALLOC_INLINE void\nstats_cactive_sub(size_t size)\n{\n\n\tatomic_sub_z(&stats_cactive, size);\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/tcache.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\ntypedef struct tcache_bin_info_s tcache_bin_info_t;\ntypedef struct tcache_bin_s tcache_bin_t;\ntypedef struct tcache_s tcache_t;\n\n/*\n * tcache pointers close to NULL are used to encode state information that is\n * used for two purposes: preventing thread caching on a per thread basis and\n * cleaning up during thread shutdown.\n */\n#define\tTCACHE_STATE_DISABLED\t\t((tcache_t *)(uintptr_t)1)\n#define\tTCACHE_STATE_REINCARNATED\t((tcache_t *)(uintptr_t)2)\n#define\tTCACHE_STATE_PURGATORY\t\t((tcache_t *)(uintptr_t)3)\n#define\tTCACHE_STATE_MAX\t\tTCACHE_STATE_PURGATORY\n\n/*\n * Absolute maximum number of cache slots for each small bin in the thread\n * cache. This is an additional constraint beyond that imposed as: twice the\n * number of regions per run for this size class.\n *\n * This constant must be an even number.\n */\n#define\tTCACHE_NSLOTS_SMALL_MAX\t\t200\n\n/* Number of cache slots for large size classes. */\n#define\tTCACHE_NSLOTS_LARGE\t\t20\n\n/* (1U << opt_lg_tcache_max) is used to compute tcache_maxclass. */\n#define\tLG_TCACHE_MAXCLASS_DEFAULT\t15\n\n/*\n * TCACHE_GC_SWEEP is the approximate number of allocation events between\n * full GC sweeps. Integer rounding may cause the actual number to be\n * slightly higher, since GC is performed incrementally.\n */\n#define\tTCACHE_GC_SWEEP\t\t\t8192\n\n/* Number of tcache allocation/deallocation events between incremental GCs. */\n#define\tTCACHE_GC_INCR\t\t\t\t\t\t\t\\\n ((TCACHE_GC_SWEEP / NBINS) + ((TCACHE_GC_SWEEP / NBINS == 0) ? 0 : 1))\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\ntypedef enum {\n\ttcache_enabled_false = 0, /* Enable cast to/from bool. */\n\ttcache_enabled_true = 1,\n\ttcache_enabled_default = 2\n} tcache_enabled_t;\n\n/*\n * Read-only information associated with each element of tcache_t's tbins array\n * is stored separately, mainly to reduce memory usage.\n */\nstruct tcache_bin_info_s {\n\tunsigned\tncached_max;\t/* Upper limit on ncached. */\n};\n\nstruct tcache_bin_s {\n\ttcache_bin_stats_t tstats;\n\tint\t\tlow_water;\t/* Min # cached since last GC. */\n\tunsigned\tlg_fill_div;\t/* Fill (ncached_max >> lg_fill_div). */\n\tunsigned\tncached;\t/* # of cached objects. */\n\tvoid\t\t**avail;\t/* Stack of available objects. */\n};\n\nstruct tcache_s {\n\tql_elm(tcache_t) link;\t\t/* Used for aggregating stats. */\n\tuint64_t\tprof_accumbytes;/* Cleared after arena_prof_accum() */\n\tarena_t\t\t*arena;\t\t/* This thread's arena. */\n\tunsigned\tev_cnt;\t\t/* Event count since incremental GC. */\n\tunsigned\tnext_gc_bin;\t/* Next bin to GC. */\n\ttcache_bin_t\ttbins[1];\t/* Dynamically sized. */\n\t/*\n\t * The pointer stacks associated with tbins follow as a contiguous\n\t * array. During tcache initialization, the avail pointer in each\n\t * element of tbins is initialized to point to the proper offset within\n\t * this array.\n\t */\n};\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nextern bool\topt_tcache;\nextern ssize_t\topt_lg_tcache_max;\n\nextern tcache_bin_info_t\t*tcache_bin_info;\n\n/*\n * Number of tcache bins. There are NBINS small-object bins, plus 0 or more\n * large-object bins.\n */\nextern size_t\t\t\tnhbins;\n\n/* Maximum cached size class. */\nextern size_t\t\t\ttcache_maxclass;\n\nsize_t\ttcache_salloc(const void *ptr);\nvoid\ttcache_event_hard(tcache_t *tcache);\nvoid\t*tcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin,\n size_t binind);\nvoid\ttcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem,\n tcache_t *tcache);\nvoid\ttcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem,\n tcache_t *tcache);\nvoid\ttcache_arena_associate(tcache_t *tcache, arena_t *arena);\nvoid\ttcache_arena_dissociate(tcache_t *tcache);\ntcache_t *tcache_create(arena_t *arena);\nvoid\ttcache_destroy(tcache_t *tcache);\nvoid\ttcache_thread_cleanup(void *arg);\nvoid\ttcache_stats_merge(tcache_t *tcache, arena_t *arena);\nbool\ttcache_boot0(void);\nbool\ttcache_boot1(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nmalloc_tsd_protos(JEMALLOC_ATTR(unused), tcache, tcache_t *)\nmalloc_tsd_protos(JEMALLOC_ATTR(unused), tcache_enabled, tcache_enabled_t)\n\nvoid\ttcache_event(tcache_t *tcache);\nvoid\ttcache_flush(void);\nbool\ttcache_enabled_get(void);\ntcache_t *tcache_get(bool create);\nvoid\ttcache_enabled_set(bool enabled);\nvoid\t*tcache_alloc_easy(tcache_bin_t *tbin);\nvoid\t*tcache_alloc_small(tcache_t *tcache, size_t size, bool zero);\nvoid\t*tcache_alloc_large(tcache_t *tcache, size_t size, bool zero);\nvoid\ttcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind);\nvoid\ttcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TCACHE_C_))\n/* Map of thread-specific caches. */\nmalloc_tsd_externs(tcache, tcache_t *)\nmalloc_tsd_funcs(JEMALLOC_INLINE, tcache, tcache_t *, NULL,\n tcache_thread_cleanup)\n/* Per thread flag that allows thread caches to be disabled. */\nmalloc_tsd_externs(tcache_enabled, tcache_enabled_t)\nmalloc_tsd_funcs(JEMALLOC_INLINE, tcache_enabled, tcache_enabled_t,\n tcache_enabled_default, malloc_tsd_no_cleanup)\n\nJEMALLOC_INLINE void\ntcache_flush(void)\n{\n\ttcache_t *tcache;\n\n\tcassert(config_tcache);\n\n\ttcache = *tcache_tsd_get();\n\tif ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX)\n\t\treturn;\n\ttcache_destroy(tcache);\n\ttcache = NULL;\n\ttcache_tsd_set(&tcache);\n}\n\nJEMALLOC_INLINE bool\ntcache_enabled_get(void)\n{\n\ttcache_enabled_t tcache_enabled;\n\n\tcassert(config_tcache);\n\n\ttcache_enabled = *tcache_enabled_tsd_get();\n\tif (tcache_enabled == tcache_enabled_default) {\n\t\ttcache_enabled = (tcache_enabled_t)opt_tcache;\n\t\ttcache_enabled_tsd_set(&tcache_enabled);\n\t}\n\n\treturn ((bool)tcache_enabled);\n}\n\nJEMALLOC_INLINE void\ntcache_enabled_set(bool enabled)\n{\n\ttcache_enabled_t tcache_enabled;\n\ttcache_t *tcache;\n\n\tcassert(config_tcache);\n\n\ttcache_enabled = (tcache_enabled_t)enabled;\n\ttcache_enabled_tsd_set(&tcache_enabled);\n\ttcache = *tcache_tsd_get();\n\tif (enabled) {\n\t\tif (tcache == TCACHE_STATE_DISABLED) {\n\t\t\ttcache = NULL;\n\t\t\ttcache_tsd_set(&tcache);\n\t\t}\n\t} else /* disabled */ {\n\t\tif (tcache > TCACHE_STATE_MAX) {\n\t\t\ttcache_destroy(tcache);\n\t\t\ttcache = NULL;\n\t\t}\n\t\tif (tcache == NULL) {\n\t\t\ttcache = TCACHE_STATE_DISABLED;\n\t\t\ttcache_tsd_set(&tcache);\n\t\t}\n\t}\n}\n\nJEMALLOC_INLINE tcache_t *\ntcache_get(bool create)\n{\n\ttcache_t *tcache;\n\n\tif (config_tcache == false)\n\t\treturn (NULL);\n\tif (config_lazy_lock && isthreaded == false)\n\t\treturn (NULL);\n\n\ttcache = *tcache_tsd_get();\n\tif ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX) {\n\t\tif (tcache == TCACHE_STATE_DISABLED)\n\t\t\treturn (NULL);\n\t\tif (tcache == NULL) {\n\t\t\tif (create == false) {\n\t\t\t\t/*\n\t\t\t\t * Creating a tcache here would cause\n\t\t\t\t * allocation as a side effect of free().\n\t\t\t\t * Ordinarily that would be okay since\n\t\t\t\t * tcache_create() failure is a soft failure\n\t\t\t\t * that doesn't propagate. However, if TLS\n\t\t\t\t * data are freed via free() as in glibc,\n\t\t\t\t * subtle corruption could result from setting\n\t\t\t\t * a TLS variable after its backing memory is\n\t\t\t\t * freed.\n\t\t\t\t */\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tif (tcache_enabled_get() == false) {\n\t\t\t\ttcache_enabled_set(false); /* Memoize. */\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\treturn (tcache_create(choose_arena(NULL)));\n\t\t}\n\t\tif (tcache == TCACHE_STATE_PURGATORY) {\n\t\t\t/*\n\t\t\t * Make a note that an allocator function was called\n\t\t\t * after tcache_thread_cleanup() was called.\n\t\t\t */\n\t\t\ttcache = TCACHE_STATE_REINCARNATED;\n\t\t\ttcache_tsd_set(&tcache);\n\t\t\treturn (NULL);\n\t\t}\n\t\tif (tcache == TCACHE_STATE_REINCARNATED)\n\t\t\treturn (NULL);\n\t\tnot_reached();\n\t}\n\n\treturn (tcache);\n}\n\nJEMALLOC_INLINE void\ntcache_event(tcache_t *tcache)\n{\n\n\tif (TCACHE_GC_INCR == 0)\n\t\treturn;\n\n\ttcache->ev_cnt++;\n\tassert(tcache->ev_cnt <= TCACHE_GC_INCR);\n\tif (tcache->ev_cnt == TCACHE_GC_INCR)\n\t\ttcache_event_hard(tcache);\n}\n\nJEMALLOC_INLINE void *\ntcache_alloc_easy(tcache_bin_t *tbin)\n{\n\tvoid *ret;\n\n\tif (tbin->ncached == 0) {\n\t\ttbin->low_water = -1;\n\t\treturn (NULL);\n\t}\n\ttbin->ncached--;\n\tif ((int)tbin->ncached < tbin->low_water)\n\t\ttbin->low_water = tbin->ncached;\n\tret = tbin->avail[tbin->ncached];\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void *\ntcache_alloc_small(tcache_t *tcache, size_t size, bool zero)\n{\n\tvoid *ret;\n\tsize_t binind;\n\ttcache_bin_t *tbin;\n\n\tbinind = SMALL_SIZE2BIN(size);\n\tassert(binind < NBINS);\n\ttbin = &tcache->tbins[binind];\n\tret = tcache_alloc_easy(tbin);\n\tif (ret == NULL) {\n\t\tret = tcache_alloc_small_hard(tcache, tbin, binind);\n\t\tif (ret == NULL)\n\t\t\treturn (NULL);\n\t}\n\tassert(tcache_salloc(ret) == arena_bin_info[binind].reg_size);\n\n\tif (zero == false) {\n\t\tif (config_fill) {\n\t\t\tif (opt_junk) {\n\t\t\t\tarena_alloc_junk_small(ret,\n\t\t\t\t &arena_bin_info[binind], false);\n\t\t\t} else if (opt_zero)\n\t\t\t\tmemset(ret, 0, size);\n\t\t}\n\t} else {\n\t\tif (config_fill && opt_junk) {\n\t\t\tarena_alloc_junk_small(ret, &arena_bin_info[binind],\n\t\t\t true);\n\t\t}\n\t\tVALGRIND_MAKE_MEM_UNDEFINED(ret, size);\n\t\tmemset(ret, 0, size);\n\t}\n\n\tif (config_stats)\n\t\ttbin->tstats.nrequests++;\n\tif (config_prof)\n\t\ttcache->prof_accumbytes += arena_bin_info[binind].reg_size;\n\ttcache_event(tcache);\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void *\ntcache_alloc_large(tcache_t *tcache, size_t size, bool zero)\n{\n\tvoid *ret;\n\tsize_t binind;\n\ttcache_bin_t *tbin;\n\n\tsize = PAGE_CEILING(size);\n\tassert(size <= tcache_maxclass);\n\tbinind = NBINS + (size >> LG_PAGE) - 1;\n\tassert(binind < nhbins);\n\ttbin = &tcache->tbins[binind];\n\tret = tcache_alloc_easy(tbin);\n\tif (ret == NULL) {\n\t\t/*\n\t\t * Only allocate one large object at a time, because it's quite\n\t\t * expensive to create one and not use it.\n\t\t */\n\t\tret = arena_malloc_large(tcache->arena, size, zero);\n\t\tif (ret == NULL)\n\t\t\treturn (NULL);\n\t} else {\n\t\tif (config_prof && prof_promote && size == PAGE) {\n\t\t\tarena_chunk_t *chunk =\n\t\t\t (arena_chunk_t *)CHUNK_ADDR2BASE(ret);\n\t\t\tsize_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >>\n\t\t\t LG_PAGE);\n\t\t\tarena_mapbits_large_binind_set(chunk, pageind,\n\t\t\t BININD_INVALID);\n\t\t}\n\t\tif (zero == false) {\n\t\t\tif (config_fill) {\n\t\t\t\tif (opt_junk)\n\t\t\t\t\tmemset(ret, 0xa5, size);\n\t\t\t\telse if (opt_zero)\n\t\t\t\t\tmemset(ret, 0, size);\n\t\t\t}\n\t\t} else {\n\t\t\tVALGRIND_MAKE_MEM_UNDEFINED(ret, size);\n\t\t\tmemset(ret, 0, size);\n\t\t}\n\n\t\tif (config_stats)\n\t\t\ttbin->tstats.nrequests++;\n\t\tif (config_prof)\n\t\t\ttcache->prof_accumbytes += size;\n\t}\n\n\ttcache_event(tcache);\n\treturn (ret);\n}\n\nJEMALLOC_INLINE void\ntcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind)\n{\n\ttcache_bin_t *tbin;\n\ttcache_bin_info_t *tbin_info;\n\n\tassert(tcache_salloc(ptr) <= SMALL_MAXCLASS);\n\n\tif (config_fill && opt_junk)\n\t\tarena_dalloc_junk_small(ptr, &arena_bin_info[binind]);\n\n\ttbin = &tcache->tbins[binind];\n\ttbin_info = &tcache_bin_info[binind];\n\tif (tbin->ncached == tbin_info->ncached_max) {\n\t\ttcache_bin_flush_small(tbin, binind, (tbin_info->ncached_max >>\n\t\t 1), tcache);\n\t}\n\tassert(tbin->ncached < tbin_info->ncached_max);\n\ttbin->avail[tbin->ncached] = ptr;\n\ttbin->ncached++;\n\n\ttcache_event(tcache);\n}\n\nJEMALLOC_INLINE void\ntcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size)\n{\n\tsize_t binind;\n\ttcache_bin_t *tbin;\n\ttcache_bin_info_t *tbin_info;\n\n\tassert((size & PAGE_MASK) == 0);\n\tassert(tcache_salloc(ptr) > SMALL_MAXCLASS);\n\tassert(tcache_salloc(ptr) <= tcache_maxclass);\n\n\tbinind = NBINS + (size >> LG_PAGE) - 1;\n\n\tif (config_fill && opt_junk)\n\t\tmemset(ptr, 0x5a, size);\n\n\ttbin = &tcache->tbins[binind];\n\ttbin_info = &tcache_bin_info[binind];\n\tif (tbin->ncached == tbin_info->ncached_max) {\n\t\ttcache_bin_flush_large(tbin, binind, (tbin_info->ncached_max >>\n\t\t 1), tcache);\n\t}\n\tassert(tbin->ncached < tbin_info->ncached_max);\n\ttbin->avail[tbin->ncached] = ptr;\n\ttbin->ncached++;\n\n\ttcache_event(tcache);\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/tsd.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/* Maximum number of malloc_tsd users with cleanup functions. */\n#define\tMALLOC_TSD_CLEANUPS_MAX\t8\n\ntypedef bool (*malloc_tsd_cleanup_t)(void);\n\n/*\n * TLS/TSD-agnostic macro-based implementation of thread-specific data. There\n * are four macros that support (at least) three use cases: file-private,\n * library-private, and library-private inlined. Following is an example\n * library-private tsd variable:\n *\n * In example.h:\n * typedef struct {\n * int x;\n * int y;\n * } example_t;\n * #define EX_INITIALIZER JEMALLOC_CONCAT({0, 0})\n * malloc_tsd_protos(, example, example_t *)\n * malloc_tsd_externs(example, example_t *)\n * In example.c:\n * malloc_tsd_data(, example, example_t *, EX_INITIALIZER)\n * malloc_tsd_funcs(, example, example_t *, EX_INITIALIZER,\n * example_tsd_cleanup)\n *\n * The result is a set of generated functions, e.g.:\n *\n * bool example_tsd_boot(void) {...}\n * example_t **example_tsd_get() {...}\n * void example_tsd_set(example_t **val) {...}\n *\n * Note that all of the functions deal in terms of (a_type *) rather than\n * (a_type) so that it is possible to support non-pointer types (unlike\n * pthreads TSD). example_tsd_cleanup() is passed an (a_type *) pointer that is\n * cast to (void *). This means that the cleanup function needs to cast *and*\n * dereference the function argument, e.g.:\n *\n * void\n * example_tsd_cleanup(void *arg)\n * {\n * example_t *example = *(example_t **)arg;\n *\n * [...]\n * if ([want the cleanup function to be called again]) {\n * example_tsd_set(&example);\n * }\n * }\n *\n * If example_tsd_set() is called within example_tsd_cleanup(), it will be\n * called again. This is similar to how pthreads TSD destruction works, except\n * that pthreads only calls the cleanup function again if the value was set to\n * non-NULL.\n */\n\n/* malloc_tsd_protos(). */\n#define\tmalloc_tsd_protos(a_attr, a_name, a_type)\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_boot(void);\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_name##_tsd_get(void);\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_set(a_type *val);\n\n/* malloc_tsd_externs(). */\n#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP\n#define\tmalloc_tsd_externs(a_name, a_type)\t\t\t\t\\\nextern __thread a_type\ta_name##_tls;\t\t\t\t\t\\\nextern __thread bool\ta_name##_initialized;\t\t\t\t\\\nextern bool\t\ta_name##_booted;\n#elif (defined(JEMALLOC_TLS))\n#define\tmalloc_tsd_externs(a_name, a_type)\t\t\t\t\\\nextern __thread a_type\ta_name##_tls;\t\t\t\t\t\\\nextern pthread_key_t\ta_name##_tsd;\t\t\t\t\t\\\nextern bool\t\ta_name##_booted;\n#elif (defined(_WIN32))\n#define malloc_tsd_externs(a_name, a_type)\t\t\t\t\\\nextern DWORD\t\ta_name##_tsd;\t\t\t\t\t\\\nextern bool\t\ta_name##_booted;\n#else\n#define\tmalloc_tsd_externs(a_name, a_type)\t\t\t\t\\\nextern pthread_key_t\ta_name##_tsd;\t\t\t\t\t\\\nextern bool\t\ta_name##_booted;\n#endif\n\n/* malloc_tsd_data(). */\n#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP\n#define\tmalloc_tsd_data(a_attr, a_name, a_type, a_initializer)\t\t\\\na_attr __thread a_type JEMALLOC_TLS_MODEL\t\t\t\t\\\n a_name##_tls = a_initializer;\t\t\t\t\t\\\na_attr __thread bool JEMALLOC_TLS_MODEL\t\t\t\t\t\\\n a_name##_initialized = false;\t\t\t\t\t\\\na_attr bool\t\ta_name##_booted = false;\n#elif (defined(JEMALLOC_TLS))\n#define\tmalloc_tsd_data(a_attr, a_name, a_type, a_initializer)\t\t\\\na_attr __thread a_type JEMALLOC_TLS_MODEL\t\t\t\t\\\n a_name##_tls = a_initializer;\t\t\t\t\t\\\na_attr pthread_key_t\ta_name##_tsd;\t\t\t\t\t\\\na_attr bool\t\ta_name##_booted = false;\n#elif (defined(_WIN32))\n#define\tmalloc_tsd_data(a_attr, a_name, a_type, a_initializer)\t\t\\\na_attr DWORD\t\ta_name##_tsd;\t\t\t\t\t\\\na_attr bool\t\ta_name##_booted = false;\n#else\n#define\tmalloc_tsd_data(a_attr, a_name, a_type, a_initializer)\t\t\\\na_attr pthread_key_t\ta_name##_tsd;\t\t\t\t\t\\\na_attr bool\t\ta_name##_booted = false;\n#endif\n\n/* malloc_tsd_funcs(). */\n#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP\n#define\tmalloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,\t\t\\\n a_cleanup)\t\t\t\t\t\t\t\t\\\n/* Initialization/cleanup. */\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_cleanup_wrapper(void)\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (a_name##_initialized) {\t\t\t\t\t\\\n\t\ta_name##_initialized = false;\t\t\t\t\\\n\t\ta_cleanup(&a_name##_tls);\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (a_name##_initialized);\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_boot(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup) {\t\t\t\\\n\t\tmalloc_tsd_cleanup_register(\t\t\t\t\\\n\t\t &a_name##_tsd_cleanup_wrapper);\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\ta_name##_booted = true;\t\t\t\t\t\t\\\n\treturn (false);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\n/* Get/set. */\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_name##_tsd_get(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\treturn (&a_name##_tls);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_set(a_type *val)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\ta_name##_tls = (*val);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup)\t\t\t\t\\\n\t\ta_name##_initialized = true;\t\t\t\t\\\n}\n#elif (defined(JEMALLOC_TLS))\n#define\tmalloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,\t\t\\\n a_cleanup)\t\t\t\t\t\t\t\t\\\n/* Initialization/cleanup. */\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_boot(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup) {\t\t\t\\\n\t\tif (pthread_key_create(&a_name##_tsd, a_cleanup) != 0)\t\\\n\t\t\treturn (true);\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\ta_name##_booted = true;\t\t\t\t\t\t\\\n\treturn (false);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\n/* Get/set. */\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_name##_tsd_get(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\treturn (&a_name##_tls);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_set(a_type *val)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\ta_name##_tls = (*val);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup) {\t\t\t\\\n\t\tif (pthread_setspecific(a_name##_tsd,\t\t\t\\\n\t\t (void *)(&a_name##_tls))) {\t\t\t\t\\\n\t\t\tmalloc_write(\": Error\"\t\t\\\n\t\t\t \" setting TSD for \"#a_name\"\\n\");\t\t\\\n\t\t\tif (opt_abort)\t\t\t\t\t\\\n\t\t\t\tabort();\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n}\n#elif (defined(_WIN32))\n#define\tmalloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,\t\t\\\n a_cleanup)\t\t\t\t\t\t\t\t\\\n/* Data structure. */\t\t\t\t\t\t\t\\\ntypedef struct {\t\t\t\t\t\t\t\\\n\tbool\tinitialized;\t\t\t\t\t\t\\\n\ta_type\tval;\t\t\t\t\t\t\t\\\n} a_name##_tsd_wrapper_t;\t\t\t\t\t\t\\\n/* Initialization/cleanup. */\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_cleanup_wrapper(void)\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\twrapper = (a_name##_tsd_wrapper_t *) TlsGetValue(a_name##_tsd);\t\\\n\tif (wrapper == NULL)\t\t\t\t\t\t\\\n\t\treturn (false);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup &&\t\t\t\\\n\t wrapper->initialized) {\t\t\t\t\t\\\n\t\ta_type val = wrapper->val;\t\t\t\t\\\n\t\ta_type tsd_static_data = a_initializer;\t\t\t\\\n\t\twrapper->initialized = false;\t\t\t\t\\\n\t\twrapper->val = tsd_static_data;\t\t\t\t\\\n\t\ta_cleanup(&val);\t\t\t\t\t\\\n\t\tif (wrapper->initialized) {\t\t\t\t\\\n\t\t\t/* Trigger another cleanup round. */\t\t\\\n\t\t\treturn (true);\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tmalloc_tsd_dalloc(wrapper);\t\t\t\t\t\\\n\treturn (false);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_boot(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd = TlsAlloc();\t\t\t\t\t\\\n\tif (a_name##_tsd == TLS_OUT_OF_INDEXES)\t\t\t\t\\\n\t\treturn (true);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup) {\t\t\t\\\n\t\tmalloc_tsd_cleanup_register(\t\t\t\t\\\n\t\t &a_name##_tsd_cleanup_wrapper);\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\ta_name##_booted = true;\t\t\t\t\t\t\\\n\treturn (false);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\n/* Get/set. */\t\t\t\t\t\t\t\t\\\na_attr a_name##_tsd_wrapper_t *\t\t\t\t\t\t\\\na_name##_tsd_get_wrapper(void)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *)\t\\\n\t TlsGetValue(a_name##_tsd);\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (wrapper == NULL) {\t\t\t\t\t\t\\\n\t\twrapper = (a_name##_tsd_wrapper_t *)\t\t\t\\\n\t\t malloc_tsd_malloc(sizeof(a_name##_tsd_wrapper_t));\t\\\n\t\tif (wrapper == NULL) {\t\t\t\t\t\\\n\t\t\tmalloc_write(\": Error allocating\"\t\\\n\t\t\t \" TSD for \"#a_name\"\\n\");\t\t\t\\\n\t\t\tabort();\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t\tstatic a_type tsd_static_data = a_initializer;\t\\\n\t\t\twrapper->initialized = false;\t\t\t\\\n\t\t\twrapper->val = tsd_static_data;\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\tif (!TlsSetValue(a_name##_tsd, (void *)wrapper)) {\t\\\n\t\t\tmalloc_write(\": Error setting\"\t\\\n\t\t\t \" TSD for \"#a_name\"\\n\");\t\t\t\\\n\t\t\tabort();\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (wrapper);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_name##_tsd_get(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\twrapper = a_name##_tsd_get_wrapper();\t\t\t\t\\\n\treturn (&wrapper->val);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_set(a_type *val)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\twrapper = a_name##_tsd_get_wrapper();\t\t\t\t\\\n\twrapper->val = *(val);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup)\t\t\t\t\\\n\t\twrapper->initialized = true;\t\t\t\t\\\n}\n#else\n#define\tmalloc_tsd_funcs(a_attr, a_name, a_type, a_initializer,\t\t\\\n a_cleanup)\t\t\t\t\t\t\t\t\\\n/* Data structure. */\t\t\t\t\t\t\t\\\ntypedef struct {\t\t\t\t\t\t\t\\\n\tbool\tinitialized;\t\t\t\t\t\t\\\n\ta_type\tval;\t\t\t\t\t\t\t\\\n} a_name##_tsd_wrapper_t;\t\t\t\t\t\t\\\n/* Initialization/cleanup. */\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_cleanup_wrapper(void *arg)\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *)arg;\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup &&\t\t\t\\\n\t wrapper->initialized) {\t\t\t\t\t\\\n\t\twrapper->initialized = false;\t\t\t\t\\\n\t\ta_cleanup(&wrapper->val);\t\t\t\t\\\n\t\tif (wrapper->initialized) {\t\t\t\t\\\n\t\t\t/* Trigger another cleanup round. */\t\t\\\n\t\t\tif (pthread_setspecific(a_name##_tsd,\t\t\\\n\t\t\t (void *)wrapper)) {\t\t\t\t\\\n\t\t\t\tmalloc_write(\": Error\"\t\\\n\t\t\t\t \" setting TSD for \"#a_name\"\\n\");\t\\\n\t\t\t\tif (opt_abort)\t\t\t\t\\\n\t\t\t\t\tabort();\t\t\t\\\n\t\t\t}\t\t\t\t\t\t\\\n\t\t\treturn;\t\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\tmalloc_tsd_dalloc(wrapper);\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr bool\t\t\t\t\t\t\t\t\\\na_name##_tsd_boot(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (pthread_key_create(&a_name##_tsd,\t\t\t\t\\\n\t a_name##_tsd_cleanup_wrapper) != 0)\t\t\t\t\\\n\t\treturn (true);\t\t\t\t\t\t\\\n\ta_name##_booted = true;\t\t\t\t\t\t\\\n\treturn (false);\t\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\n/* Get/set. */\t\t\t\t\t\t\t\t\\\na_attr a_name##_tsd_wrapper_t *\t\t\t\t\t\t\\\na_name##_tsd_get_wrapper(void)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *)\t\\\n\t pthread_getspecific(a_name##_tsd);\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif (wrapper == NULL) {\t\t\t\t\t\t\\\n\t\twrapper = (a_name##_tsd_wrapper_t *)\t\t\t\\\n\t\t malloc_tsd_malloc(sizeof(a_name##_tsd_wrapper_t));\t\\\n\t\tif (wrapper == NULL) {\t\t\t\t\t\\\n\t\t\tmalloc_write(\": Error allocating\"\t\\\n\t\t\t \" TSD for \"#a_name\"\\n\");\t\t\t\\\n\t\t\tabort();\t\t\t\t\t\\\n\t\t} else {\t\t\t\t\t\t\\\n\t\t\tstatic a_type tsd_static_data = a_initializer;\t\\\n\t\t\twrapper->initialized = false;\t\t\t\\\n\t\t\twrapper->val = tsd_static_data;\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\tif (pthread_setspecific(a_name##_tsd,\t\t\t\\\n\t\t (void *)wrapper)) {\t\t\t\t\t\\\n\t\t\tmalloc_write(\": Error setting\"\t\\\n\t\t\t \" TSD for \"#a_name\"\\n\");\t\t\t\\\n\t\t\tabort();\t\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\treturn (wrapper);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr a_type *\t\t\t\t\t\t\t\t\\\na_name##_tsd_get(void)\t\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\twrapper = a_name##_tsd_get_wrapper();\t\t\t\t\\\n\treturn (&wrapper->val);\t\t\t\t\t\t\\\n}\t\t\t\t\t\t\t\t\t\\\na_attr void\t\t\t\t\t\t\t\t\\\na_name##_tsd_set(a_type *val)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\ta_name##_tsd_wrapper_t *wrapper;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tassert(a_name##_booted);\t\t\t\t\t\\\n\twrapper = a_name##_tsd_get_wrapper();\t\t\t\t\\\n\twrapper->val = *(val);\t\t\t\t\t\t\\\n\tif (a_cleanup != malloc_tsd_no_cleanup)\t\t\t\t\\\n\t\twrapper->initialized = true;\t\t\t\t\\\n}\n#endif\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nvoid\t*malloc_tsd_malloc(size_t size);\nvoid\tmalloc_tsd_dalloc(void *wrapper);\nvoid\tmalloc_tsd_no_cleanup(void *);\nvoid\tmalloc_tsd_cleanup_register(bool (*f)(void));\nvoid\tmalloc_tsd_boot(void);\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/internal/util.h", "content": "/******************************************************************************/\n#ifdef JEMALLOC_H_TYPES\n\n/* Size of stack-allocated buffer passed to buferror(). */\n#define\tBUFERROR_BUF\t\t64\n\n/*\n * Size of stack-allocated buffer used by malloc_{,v,vc}printf(). This must be\n * large enough for all possible uses within jemalloc.\n */\n#define\tMALLOC_PRINTF_BUFSIZE\t4096\n\n/*\n * Wrap a cpp argument that contains commas such that it isn't broken up into\n * multiple arguments.\n */\n#define JEMALLOC_CONCAT(...) __VA_ARGS__\n\n/*\n * Silence compiler warnings due to uninitialized values. This is used\n * wherever the compiler fails to recognize that the variable is never used\n * uninitialized.\n */\n#ifdef JEMALLOC_CC_SILENCE\n# define JEMALLOC_CC_SILENCE_INIT(v) = v\n#else\n# define JEMALLOC_CC_SILENCE_INIT(v)\n#endif\n\n/*\n * Define a custom assert() in order to reduce the chances of deadlock during\n * assertion failure.\n */\n#ifndef assert\n#define\tassert(e) do {\t\t\t\t\t\t\t\\\n\tif (config_debug && !(e)) {\t\t\t\t\t\\\n\t\tmalloc_printf(\t\t\t\t\t\t\\\n\t\t \": %s:%d: Failed assertion: \\\"%s\\\"\\n\",\t\\\n\t\t __FILE__, __LINE__, #e);\t\t\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#endif\n\n/* Use to assert a particular configuration, e.g., cassert(config_debug). */\n#define\tcassert(c) do {\t\t\t\t\t\t\t\\\n\tif ((c) == false)\t\t\t\t\t\t\\\n\t\tassert(false);\t\t\t\t\t\t\\\n} while (0)\n\n#ifndef not_reached\n#define\tnot_reached() do {\t\t\t\t\t\t\\\n\tif (config_debug) {\t\t\t\t\t\t\\\n\t\tmalloc_printf(\t\t\t\t\t\t\\\n\t\t \": %s:%d: Unreachable code reached\\n\",\t\\\n\t\t __FILE__, __LINE__);\t\t\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#endif\n\n#ifndef not_implemented\n#define\tnot_implemented() do {\t\t\t\t\t\t\\\n\tif (config_debug) {\t\t\t\t\t\t\\\n\t\tmalloc_printf(\": %s:%d: Not implemented\\n\",\t\\\n\t\t __FILE__, __LINE__);\t\t\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#endif\n\n#define\tassert_not_implemented(e) do {\t\t\t\t\t\\\n\tif (config_debug && !(e))\t\t\t\t\t\\\n\t\tnot_implemented();\t\t\t\t\t\\\n} while (0)\n\n#endif /* JEMALLOC_H_TYPES */\n/******************************************************************************/\n#ifdef JEMALLOC_H_STRUCTS\n\n#endif /* JEMALLOC_H_STRUCTS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_EXTERNS\n\nint\tbuferror(char *buf, size_t buflen);\nuintmax_t\tmalloc_strtoumax(const char *nptr, char **endptr, int base);\nvoid\tmalloc_write(const char *s);\n\n/*\n * malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating\n * point math.\n */\nint\tmalloc_vsnprintf(char *str, size_t size, const char *format,\n va_list ap);\nint\tmalloc_snprintf(char *str, size_t size, const char *format, ...)\n JEMALLOC_ATTR(format(printf, 3, 4));\nvoid\tmalloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque,\n const char *format, va_list ap);\nvoid malloc_cprintf(void (*write)(void *, const char *), void *cbopaque,\n const char *format, ...) JEMALLOC_ATTR(format(printf, 3, 4));\nvoid\tmalloc_printf(const char *format, ...)\n JEMALLOC_ATTR(format(printf, 1, 2));\n\n#endif /* JEMALLOC_H_EXTERNS */\n/******************************************************************************/\n#ifdef JEMALLOC_H_INLINES\n\n#ifndef JEMALLOC_ENABLE_INLINE\nsize_t\tpow2_ceil(size_t x);\nvoid\tmalloc_write(const char *s);\nvoid\tset_errno(int errnum);\nint\tget_errno(void);\n#endif\n\n#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_UTIL_C_))\n/* Compute the smallest power of 2 that is >= x. */\nJEMALLOC_INLINE size_t\npow2_ceil(size_t x)\n{\n\n\tx--;\n\tx |= x >> 1;\n\tx |= x >> 2;\n\tx |= x >> 4;\n\tx |= x >> 8;\n\tx |= x >> 16;\n#if (LG_SIZEOF_PTR == 3)\n\tx |= x >> 32;\n#endif\n\tx++;\n\treturn (x);\n}\n\n/* Sets error code */\nJEMALLOC_INLINE void\nset_errno(int errnum)\n{\n\n#ifdef _WIN32\n\tSetLastError(errnum);\n#else\n\terrno = errnum;\n#endif\n}\n\n/* Get last error code */\nJEMALLOC_INLINE int\nget_errno(void)\n{\n\n#ifdef _WIN32\n\treturn (GetLastError());\n#else\n\treturn (errno);\n#endif\n}\n#endif\n\n#endif /* JEMALLOC_H_INLINES */\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/include/jemalloc/jemalloc.h.in", "content": "#ifndef JEMALLOC_H_\n#define\tJEMALLOC_H_\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n#include \n#include \n\n#define\tJEMALLOC_VERSION \"@jemalloc_version@\"\n#define\tJEMALLOC_VERSION_MAJOR @jemalloc_version_major@\n#define\tJEMALLOC_VERSION_MINOR @jemalloc_version_minor@\n#define\tJEMALLOC_VERSION_BUGFIX @jemalloc_version_bugfix@\n#define\tJEMALLOC_VERSION_NREV @jemalloc_version_nrev@\n#define\tJEMALLOC_VERSION_GID \"@jemalloc_version_gid@\"\n\n#include \"jemalloc_defs@install_suffix@.h\"\n\n#ifdef JEMALLOC_EXPERIMENTAL\n#define\tALLOCM_LG_ALIGN(la)\t(la)\n#if LG_SIZEOF_PTR == 2\n#define\tALLOCM_ALIGN(a)\t(ffs(a)-1)\n#else\n#define\tALLOCM_ALIGN(a)\t((a < (size_t)INT_MAX) ? ffs(a)-1 : ffs(a>>32)+31)\n#endif\n#define\tALLOCM_ZERO\t((int)0x40)\n#define\tALLOCM_NO_MOVE\t((int)0x80)\n/* Bias arena index bits so that 0 encodes \"ALLOCM_ARENA() unspecified\". */\n#define\tALLOCM_ARENA(a)\t((int)(((a)+1) << 8))\n\n#define\tALLOCM_SUCCESS\t\t0\n#define\tALLOCM_ERR_OOM\t\t1\n#define\tALLOCM_ERR_NOT_MOVED\t2\n#endif\n\n/*\n * The je_ prefix on the following public symbol declarations is an artifact of\n * namespace management, and should be omitted in application code unless\n * JEMALLOC_NO_DEMANGLE is defined (see below).\n */\nextern JEMALLOC_EXPORT const char\t*je_malloc_conf;\nextern JEMALLOC_EXPORT void\t\t(*je_malloc_message)(void *cbopaque,\n const char *s);\n\nJEMALLOC_EXPORT void\t*je_malloc(size_t size) JEMALLOC_ATTR(malloc);\nJEMALLOC_EXPORT void\t*je_calloc(size_t num, size_t size)\n JEMALLOC_ATTR(malloc);\nJEMALLOC_EXPORT int\tje_posix_memalign(void **memptr, size_t alignment,\n size_t size) JEMALLOC_ATTR(nonnull(1));\nJEMALLOC_EXPORT void\t*je_aligned_alloc(size_t alignment, size_t size)\n JEMALLOC_ATTR(malloc);\nJEMALLOC_EXPORT void\t*je_realloc(void *ptr, size_t size);\nJEMALLOC_EXPORT void\tje_free(void *ptr);\n\n#ifdef JEMALLOC_OVERRIDE_MEMALIGN\nJEMALLOC_EXPORT void *\tje_memalign(size_t alignment, size_t size)\n JEMALLOC_ATTR(malloc);\n#endif\n\n#ifdef JEMALLOC_OVERRIDE_VALLOC\nJEMALLOC_EXPORT void *\tje_valloc(size_t size) JEMALLOC_ATTR(malloc);\n#endif\n\nJEMALLOC_EXPORT size_t\tje_malloc_usable_size(\n JEMALLOC_USABLE_SIZE_CONST void *ptr);\nJEMALLOC_EXPORT void\tje_malloc_stats_print(void (*write_cb)(void *,\n const char *), void *je_cbopaque, const char *opts);\nJEMALLOC_EXPORT int\tje_mallctl(const char *name, void *oldp,\n size_t *oldlenp, void *newp, size_t newlen);\nJEMALLOC_EXPORT int\tje_mallctlnametomib(const char *name, size_t *mibp,\n size_t *miblenp);\nJEMALLOC_EXPORT int\tje_mallctlbymib(const size_t *mib, size_t miblen,\n void *oldp, size_t *oldlenp, void *newp, size_t newlen);\n\n#ifdef JEMALLOC_EXPERIMENTAL\nJEMALLOC_EXPORT int\tje_allocm(void **ptr, size_t *rsize, size_t size,\n int flags) JEMALLOC_ATTR(nonnull(1));\nJEMALLOC_EXPORT int\tje_rallocm(void **ptr, size_t *rsize, size_t size,\n size_t extra, int flags) JEMALLOC_ATTR(nonnull(1));\nJEMALLOC_EXPORT int\tje_sallocm(const void *ptr, size_t *rsize, int flags)\n JEMALLOC_ATTR(nonnull(1));\nJEMALLOC_EXPORT int\tje_dallocm(void *ptr, int flags)\n JEMALLOC_ATTR(nonnull(1));\nJEMALLOC_EXPORT int\tje_nallocm(size_t *rsize, size_t size, int flags);\n#endif\n\n/*\n * By default application code must explicitly refer to mangled symbol names,\n * so that it is possible to use jemalloc in conjunction with another allocator\n * in the same application. Define JEMALLOC_MANGLE in order to cause automatic\n * name mangling that matches the API prefixing that happened as a result of\n * --with-mangling and/or --with-jemalloc-prefix configuration settings.\n */\n#ifdef JEMALLOC_MANGLE\n#ifndef JEMALLOC_NO_DEMANGLE\n#define\tJEMALLOC_NO_DEMANGLE\n#endif\n#define\tmalloc_conf je_malloc_conf\n#define\tmalloc_message je_malloc_message\n#define\tmalloc je_malloc\n#define\tcalloc je_calloc\n#define\tposix_memalign je_posix_memalign\n#define\taligned_alloc je_aligned_alloc\n#define\trealloc je_realloc\n#define\tfree je_free\n#define\tmalloc_usable_size je_malloc_usable_size\n#define\tmalloc_stats_print je_malloc_stats_print\n#define\tmallctl je_mallctl\n#define\tmallctlnametomib je_mallctlnametomib\n#define\tmallctlbymib je_mallctlbymib\n#define\tmemalign je_memalign\n#define\tvalloc je_valloc\n#ifdef JEMALLOC_EXPERIMENTAL\n#define\tallocm je_allocm\n#define\trallocm je_rallocm\n#define\tsallocm je_sallocm\n#define\tdallocm je_dallocm\n#define\tnallocm je_nallocm\n#endif\n#endif\n\n/*\n * The je_* macros can be used as stable alternative names for the public\n * jemalloc API if JEMALLOC_NO_DEMANGLE is defined. This is primarily meant\n * for use in jemalloc itself, but it can be used by application code to\n * provide isolation from the name mangling specified via --with-mangling\n * and/or --with-jemalloc-prefix.\n */\n#ifndef JEMALLOC_NO_DEMANGLE\n#undef je_malloc_conf\n#undef je_malloc_message\n#undef je_malloc\n#undef je_calloc\n#undef je_posix_memalign\n#undef je_aligned_alloc\n#undef je_realloc\n#undef je_free\n#undef je_malloc_usable_size\n#undef je_malloc_stats_print\n#undef je_mallctl\n#undef je_mallctlnametomib\n#undef je_mallctlbymib\n#undef je_memalign\n#undef je_valloc\n#ifdef JEMALLOC_EXPERIMENTAL\n#undef je_allocm\n#undef je_rallocm\n#undef je_sallocm\n#undef je_dallocm\n#undef je_nallocm\n#endif\n#endif\n\n#ifdef __cplusplus\n};\n#endif\n#endif /* JEMALLOC_H_ */\n" }, { "path": "deps/jemalloc/include/jemalloc/jemalloc_defs.h.in", "content": "/*\n * If JEMALLOC_PREFIX is defined via --with-jemalloc-prefix, it will cause all\n * public APIs to be prefixed. This makes it possible, with some care, to use\n * multiple allocators simultaneously.\n */\n#undef JEMALLOC_PREFIX\n#undef JEMALLOC_CPREFIX\n\n/*\n * Name mangling for public symbols is controlled by --with-mangling and\n * --with-jemalloc-prefix. With default settings the je_ prefix is stripped by\n * these macro definitions.\n */\n#undef je_malloc_conf\n#undef je_malloc_message\n#undef je_malloc\n#undef je_calloc\n#undef je_posix_memalign\n#undef je_aligned_alloc\n#undef je_realloc\n#undef je_free\n#undef je_malloc_usable_size\n#undef je_malloc_stats_print\n#undef je_mallctl\n#undef je_mallctlnametomib\n#undef je_mallctlbymib\n#undef je_memalign\n#undef je_valloc\n#undef je_allocm\n#undef je_rallocm\n#undef je_sallocm\n#undef je_dallocm\n#undef je_nallocm\n\n/*\n * JEMALLOC_PRIVATE_NAMESPACE is used as a prefix for all library-private APIs.\n * For shared libraries, symbol visibility mechanisms prevent these symbols\n * from being exported, but for static libraries, naming collisions are a real\n * possibility.\n */\n#undef JEMALLOC_PRIVATE_NAMESPACE\n#undef JEMALLOC_N\n\n/*\n * Hyper-threaded CPUs may need a special instruction inside spin loops in\n * order to yield to another virtual CPU.\n */\n#undef CPU_SPINWAIT\n\n/* Defined if the equivalent of FreeBSD's atomic(9) functions are available. */\n#undef JEMALLOC_ATOMIC9\n\n/*\n * Defined if OSAtomic*() functions are available, as provided by Darwin, and\n * documented in the atomic(3) manual page.\n */\n#undef JEMALLOC_OSATOMIC\n\n/*\n * Defined if __sync_add_and_fetch(uint32_t *, uint32_t) and\n * __sync_sub_and_fetch(uint32_t *, uint32_t) are available, despite\n * __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4 not being defined (which means the\n * functions are defined in libgcc instead of being inlines)\n */\n#undef JE_FORCE_SYNC_COMPARE_AND_SWAP_4\n\n/*\n * Defined if __sync_add_and_fetch(uint64_t *, uint64_t) and\n * __sync_sub_and_fetch(uint64_t *, uint64_t) are available, despite\n * __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 not being defined (which means the\n * functions are defined in libgcc instead of being inlines)\n */\n#undef JE_FORCE_SYNC_COMPARE_AND_SWAP_8\n\n/*\n * Defined if OSSpin*() functions are available, as provided by Darwin, and\n * documented in the spinlock(3) manual page.\n */\n#undef JEMALLOC_OSSPIN\n\n/*\n * Defined if _malloc_thread_cleanup() exists. At least in the case of\n * FreeBSD, pthread_key_create() allocates, which if used during malloc\n * bootstrapping will cause recursion into the pthreads library. Therefore, if\n * _malloc_thread_cleanup() exists, use it as the basis for thread cleanup in\n * malloc_tsd.\n */\n#undef JEMALLOC_MALLOC_THREAD_CLEANUP\n\n/*\n * Defined if threaded initialization is known to be safe on this platform.\n * Among other things, it must be possible to initialize a mutex without\n * triggering allocation in order for threaded allocation to be safe.\n */\n#undef JEMALLOC_THREADED_INIT\n\n/*\n * Defined if the pthreads implementation defines\n * _pthread_mutex_init_calloc_cb(), in which case the function is used in order\n * to avoid recursive allocation during mutex initialization.\n */\n#undef JEMALLOC_MUTEX_INIT_CB\n\n/* Defined if __attribute__((...)) syntax is supported. */\n#undef JEMALLOC_HAVE_ATTR\n#ifdef JEMALLOC_HAVE_ATTR\n# define JEMALLOC_ATTR(s) __attribute__((s))\n# define JEMALLOC_EXPORT JEMALLOC_ATTR(visibility(\"default\"))\n# define JEMALLOC_ALIGNED(s) JEMALLOC_ATTR(aligned(s))\n# define JEMALLOC_SECTION(s) JEMALLOC_ATTR(section(s))\n# define JEMALLOC_NOINLINE JEMALLOC_ATTR(noinline)\n#elif _MSC_VER\n# define JEMALLOC_ATTR(s)\n# ifdef DLLEXPORT\n# define JEMALLOC_EXPORT __declspec(dllexport)\n# else\n# define JEMALLOC_EXPORT __declspec(dllimport)\n# endif\n# define JEMALLOC_ALIGNED(s) __declspec(align(s))\n# define JEMALLOC_SECTION(s) __declspec(allocate(s))\n# define JEMALLOC_NOINLINE __declspec(noinline)\n#else\n# define JEMALLOC_ATTR(s)\n# define JEMALLOC_EXPORT\n# define JEMALLOC_ALIGNED(s)\n# define JEMALLOC_SECTION(s)\n# define JEMALLOC_NOINLINE\n#endif\n\n/* Defined if sbrk() is supported. */\n#undef JEMALLOC_HAVE_SBRK\n\n/* Non-empty if the tls_model attribute is supported. */\n#undef JEMALLOC_TLS_MODEL\n\n/* JEMALLOC_CC_SILENCE enables code that silences unuseful compiler warnings. */\n#undef JEMALLOC_CC_SILENCE\n\n/*\n * JEMALLOC_DEBUG enables assertions and other sanity checks, and disables\n * inline functions.\n */\n#undef JEMALLOC_DEBUG\n\n/* JEMALLOC_STATS enables statistics calculation. */\n#undef JEMALLOC_STATS\n\n/* JEMALLOC_PROF enables allocation profiling. */\n#undef JEMALLOC_PROF\n\n/* Use libunwind for profile backtracing if defined. */\n#undef JEMALLOC_PROF_LIBUNWIND\n\n/* Use libgcc for profile backtracing if defined. */\n#undef JEMALLOC_PROF_LIBGCC\n\n/* Use gcc intrinsics for profile backtracing if defined. */\n#undef JEMALLOC_PROF_GCC\n\n/*\n * JEMALLOC_TCACHE enables a thread-specific caching layer for small objects.\n * This makes it possible to allocate/deallocate objects without any locking\n * when the cache is in the steady state.\n */\n#undef JEMALLOC_TCACHE\n\n/*\n * JEMALLOC_DSS enables use of sbrk(2) to allocate chunks from the data storage\n * segment (DSS).\n */\n#undef JEMALLOC_DSS\n\n/* Support memory filling (junk/zero/quarantine/redzone). */\n#undef JEMALLOC_FILL\n\n/* Support the experimental API. */\n#undef JEMALLOC_EXPERIMENTAL\n\n/* Support utrace(2)-based tracing. */\n#undef JEMALLOC_UTRACE\n\n/* Support Valgrind. */\n#undef JEMALLOC_VALGRIND\n\n/* Support optional abort() on OOM. */\n#undef JEMALLOC_XMALLOC\n\n/* Support lazy locking (avoid locking unless a second thread is launched). */\n#undef JEMALLOC_LAZY_LOCK\n\n/* One page is 2^STATIC_PAGE_SHIFT bytes. */\n#undef STATIC_PAGE_SHIFT\n\n/*\n * If defined, use munmap() to unmap freed chunks, rather than storing them for\n * later reuse. This is disabled by default on Linux because common sequences\n * of mmap()/munmap() calls will cause virtual memory map holes.\n */\n#undef JEMALLOC_MUNMAP\n\n/*\n * If defined, use mremap(...MREMAP_FIXED...) for huge realloc(). This is\n * disabled by default because it is Linux-specific and it will cause virtual\n * memory map holes, much like munmap(2) does.\n */\n#undef JEMALLOC_MREMAP\n\n/* TLS is used to map arenas and magazine caches to threads. */\n#undef JEMALLOC_TLS\n\n/*\n * JEMALLOC_IVSALLOC enables ivsalloc(), which verifies that pointers reside\n * within jemalloc-owned chunks before dereferencing them.\n */\n#undef JEMALLOC_IVSALLOC\n\n/*\n * Define overrides for non-standard allocator-related functions if they\n * are present on the system.\n */\n#undef JEMALLOC_OVERRIDE_MEMALIGN\n#undef JEMALLOC_OVERRIDE_VALLOC\n\n/*\n * At least Linux omits the \"const\" in:\n *\n * size_t malloc_usable_size(const void *ptr);\n *\n * Match the operating system's prototype.\n */\n#undef JEMALLOC_USABLE_SIZE_CONST\n\n/*\n * Darwin (OS X) uses zones to work around Mach-O symbol override shortcomings.\n */\n#undef JEMALLOC_ZONE\n#undef JEMALLOC_ZONE_VERSION\n\n/*\n * Methods for purging unused pages differ between operating systems.\n *\n * madvise(..., MADV_DONTNEED) : On Linux, this immediately discards pages,\n * such that new pages will be demand-zeroed if\n * the address region is later touched.\n * madvise(..., MADV_FREE) : On FreeBSD and Darwin, this marks pages as being\n * unused, such that they will be discarded rather\n * than swapped out.\n */\n#undef JEMALLOC_PURGE_MADVISE_DONTNEED\n#undef JEMALLOC_PURGE_MADVISE_FREE\n\n/* sizeof(void *) == 2^LG_SIZEOF_PTR. */\n#undef LG_SIZEOF_PTR\n\n/* sizeof(int) == 2^LG_SIZEOF_INT. */\n#undef LG_SIZEOF_INT\n\n/* sizeof(long) == 2^LG_SIZEOF_LONG. */\n#undef LG_SIZEOF_LONG\n\n/* sizeof(intmax_t) == 2^LG_SIZEOF_INTMAX_T. */\n#undef LG_SIZEOF_INTMAX_T\n" }, { "path": "deps/jemalloc/include/msvc_compat/inttypes.h", "content": "// ISO C9x compliant inttypes.h for Microsoft Visual Studio\n// Based on ISO/IEC 9899:TC2 Committee draft (May 6, 2005) WG14/N1124 \n// \n// Copyright (c) 2006 Alexander Chemeris\n// \n// Redistribution and use in source and binary forms, with or without\n// modification, are permitted provided that the following conditions are met:\n// \n// 1. Redistributions of source code must retain the above copyright notice,\n// this list of conditions and the following disclaimer.\n// \n// 2. Redistributions in binary form must reproduce the above copyright\n// notice, this list of conditions and the following disclaimer in the\n// documentation and/or other materials provided with the distribution.\n// \n// 3. The name of the author may be used to endorse or promote products\n// derived from this software without specific prior written permission.\n// \n// THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED\n// WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF\n// MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO\n// EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\n// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,\n// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;\n// OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, \n// WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR\n// OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF\n// ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n// \n///////////////////////////////////////////////////////////////////////////////\n\n#ifndef _MSC_VER // [\n#error \"Use this header only with Microsoft Visual C++ compilers!\"\n#endif // _MSC_VER ]\n\n#ifndef _MSC_INTTYPES_H_ // [\n#define _MSC_INTTYPES_H_\n\n#if _MSC_VER > 1000\n#pragma once\n#endif\n\n#include \"stdint.h\"\n\n// 7.8 Format conversion of integer types\n\ntypedef struct {\n intmax_t quot;\n intmax_t rem;\n} imaxdiv_t;\n\n// 7.8.1 Macros for format specifiers\n\n#if !defined(__cplusplus) || defined(__STDC_FORMAT_MACROS) // [ See footnote 185 at page 198\n\n#ifdef _WIN64\n# define __PRI64_PREFIX \"l\"\n# define __PRIPTR_PREFIX \"l\"\n#else\n# define __PRI64_PREFIX \"ll\"\n# define __PRIPTR_PREFIX\n#endif\n\n// The fprintf macros for signed integers are:\n#define PRId8 \"d\"\n#define PRIi8 \"i\"\n#define PRIdLEAST8 \"d\"\n#define PRIiLEAST8 \"i\"\n#define PRIdFAST8 \"d\"\n#define PRIiFAST8 \"i\"\n\n#define PRId16 \"hd\"\n#define PRIi16 \"hi\"\n#define PRIdLEAST16 \"hd\"\n#define PRIiLEAST16 \"hi\"\n#define PRIdFAST16 \"hd\"\n#define PRIiFAST16 \"hi\"\n\n#define PRId32 \"d\"\n#define PRIi32 \"i\"\n#define PRIdLEAST32 \"d\"\n#define PRIiLEAST32 \"i\"\n#define PRIdFAST32 \"d\"\n#define PRIiFAST32 \"i\"\n\n#define PRId64 __PRI64_PREFIX \"d\"\n#define PRIi64 __PRI64_PREFIX \"i\"\n#define PRIdLEAST64 __PRI64_PREFIX \"d\"\n#define PRIiLEAST64 __PRI64_PREFIX \"i\"\n#define PRIdFAST64 __PRI64_PREFIX \"d\"\n#define PRIiFAST64 __PRI64_PREFIX \"i\"\n\n#define PRIdMAX __PRI64_PREFIX \"d\"\n#define PRIiMAX __PRI64_PREFIX \"i\"\n\n#define PRIdPTR __PRIPTR_PREFIX \"d\"\n#define PRIiPTR __PRIPTR_PREFIX \"i\"\n\n// The fprintf macros for unsigned integers are:\n#define PRIo8 \"o\"\n#define PRIu8 \"u\"\n#define PRIx8 \"x\"\n#define PRIX8 \"X\"\n#define PRIoLEAST8 \"o\"\n#define PRIuLEAST8 \"u\"\n#define PRIxLEAST8 \"x\"\n#define PRIXLEAST8 \"X\"\n#define PRIoFAST8 \"o\"\n#define PRIuFAST8 \"u\"\n#define PRIxFAST8 \"x\"\n#define PRIXFAST8 \"X\"\n\n#define PRIo16 \"ho\"\n#define PRIu16 \"hu\"\n#define PRIx16 \"hx\"\n#define PRIX16 \"hX\"\n#define PRIoLEAST16 \"ho\"\n#define PRIuLEAST16 \"hu\"\n#define PRIxLEAST16 \"hx\"\n#define PRIXLEAST16 \"hX\"\n#define PRIoFAST16 \"ho\"\n#define PRIuFAST16 \"hu\"\n#define PRIxFAST16 \"hx\"\n#define PRIXFAST16 \"hX\"\n\n#define PRIo32 \"o\"\n#define PRIu32 \"u\"\n#define PRIx32 \"x\"\n#define PRIX32 \"X\"\n#define PRIoLEAST32 \"o\"\n#define PRIuLEAST32 \"u\"\n#define PRIxLEAST32 \"x\"\n#define PRIXLEAST32 \"X\"\n#define PRIoFAST32 \"o\"\n#define PRIuFAST32 \"u\"\n#define PRIxFAST32 \"x\"\n#define PRIXFAST32 \"X\"\n\n#define PRIo64 __PRI64_PREFIX \"o\"\n#define PRIu64 __PRI64_PREFIX \"u\"\n#define PRIx64 __PRI64_PREFIX \"x\"\n#define PRIX64 __PRI64_PREFIX \"X\"\n#define PRIoLEAST64 __PRI64_PREFIX \"o\"\n#define PRIuLEAST64 __PRI64_PREFIX \"u\"\n#define PRIxLEAST64 __PRI64_PREFIX \"x\"\n#define PRIXLEAST64 __PRI64_PREFIX \"X\"\n#define PRIoFAST64 __PRI64_PREFIX \"o\"\n#define PRIuFAST64 __PRI64_PREFIX \"u\"\n#define PRIxFAST64 __PRI64_PREFIX \"x\"\n#define PRIXFAST64 __PRI64_PREFIX \"X\"\n\n#define PRIoMAX __PRI64_PREFIX \"o\"\n#define PRIuMAX __PRI64_PREFIX \"u\"\n#define PRIxMAX __PRI64_PREFIX \"x\"\n#define PRIXMAX __PRI64_PREFIX \"X\"\n\n#define PRIoPTR __PRIPTR_PREFIX \"o\"\n#define PRIuPTR __PRIPTR_PREFIX \"u\"\n#define PRIxPTR __PRIPTR_PREFIX \"x\"\n#define PRIXPTR __PRIPTR_PREFIX \"X\"\n\n// The fscanf macros for signed integers are:\n#define SCNd8 \"d\"\n#define SCNi8 \"i\"\n#define SCNdLEAST8 \"d\"\n#define SCNiLEAST8 \"i\"\n#define SCNdFAST8 \"d\"\n#define SCNiFAST8 \"i\"\n\n#define SCNd16 \"hd\"\n#define SCNi16 \"hi\"\n#define SCNdLEAST16 \"hd\"\n#define SCNiLEAST16 \"hi\"\n#define SCNdFAST16 \"hd\"\n#define SCNiFAST16 \"hi\"\n\n#define SCNd32 \"ld\"\n#define SCNi32 \"li\"\n#define SCNdLEAST32 \"ld\"\n#define SCNiLEAST32 \"li\"\n#define SCNdFAST32 \"ld\"\n#define SCNiFAST32 \"li\"\n\n#define SCNd64 \"I64d\"\n#define SCNi64 \"I64i\"\n#define SCNdLEAST64 \"I64d\"\n#define SCNiLEAST64 \"I64i\"\n#define SCNdFAST64 \"I64d\"\n#define SCNiFAST64 \"I64i\"\n\n#define SCNdMAX \"I64d\"\n#define SCNiMAX \"I64i\"\n\n#ifdef _WIN64 // [\n# define SCNdPTR \"I64d\"\n# define SCNiPTR \"I64i\"\n#else // _WIN64 ][\n# define SCNdPTR \"ld\"\n# define SCNiPTR \"li\"\n#endif // _WIN64 ]\n\n// The fscanf macros for unsigned integers are:\n#define SCNo8 \"o\"\n#define SCNu8 \"u\"\n#define SCNx8 \"x\"\n#define SCNX8 \"X\"\n#define SCNoLEAST8 \"o\"\n#define SCNuLEAST8 \"u\"\n#define SCNxLEAST8 \"x\"\n#define SCNXLEAST8 \"X\"\n#define SCNoFAST8 \"o\"\n#define SCNuFAST8 \"u\"\n#define SCNxFAST8 \"x\"\n#define SCNXFAST8 \"X\"\n\n#define SCNo16 \"ho\"\n#define SCNu16 \"hu\"\n#define SCNx16 \"hx\"\n#define SCNX16 \"hX\"\n#define SCNoLEAST16 \"ho\"\n#define SCNuLEAST16 \"hu\"\n#define SCNxLEAST16 \"hx\"\n#define SCNXLEAST16 \"hX\"\n#define SCNoFAST16 \"ho\"\n#define SCNuFAST16 \"hu\"\n#define SCNxFAST16 \"hx\"\n#define SCNXFAST16 \"hX\"\n\n#define SCNo32 \"lo\"\n#define SCNu32 \"lu\"\n#define SCNx32 \"lx\"\n#define SCNX32 \"lX\"\n#define SCNoLEAST32 \"lo\"\n#define SCNuLEAST32 \"lu\"\n#define SCNxLEAST32 \"lx\"\n#define SCNXLEAST32 \"lX\"\n#define SCNoFAST32 \"lo\"\n#define SCNuFAST32 \"lu\"\n#define SCNxFAST32 \"lx\"\n#define SCNXFAST32 \"lX\"\n\n#define SCNo64 \"I64o\"\n#define SCNu64 \"I64u\"\n#define SCNx64 \"I64x\"\n#define SCNX64 \"I64X\"\n#define SCNoLEAST64 \"I64o\"\n#define SCNuLEAST64 \"I64u\"\n#define SCNxLEAST64 \"I64x\"\n#define SCNXLEAST64 \"I64X\"\n#define SCNoFAST64 \"I64o\"\n#define SCNuFAST64 \"I64u\"\n#define SCNxFAST64 \"I64x\"\n#define SCNXFAST64 \"I64X\"\n\n#define SCNoMAX \"I64o\"\n#define SCNuMAX \"I64u\"\n#define SCNxMAX \"I64x\"\n#define SCNXMAX \"I64X\"\n\n#ifdef _WIN64 // [\n# define SCNoPTR \"I64o\"\n# define SCNuPTR \"I64u\"\n# define SCNxPTR \"I64x\"\n# define SCNXPTR \"I64X\"\n#else // _WIN64 ][\n# define SCNoPTR \"lo\"\n# define SCNuPTR \"lu\"\n# define SCNxPTR \"lx\"\n# define SCNXPTR \"lX\"\n#endif // _WIN64 ]\n\n#endif // __STDC_FORMAT_MACROS ]\n\n// 7.8.2 Functions for greatest-width integer types\n\n// 7.8.2.1 The imaxabs function\n#define imaxabs _abs64\n\n// 7.8.2.2 The imaxdiv function\n\n// This is modified version of div() function from Microsoft's div.c found\n// in %MSVC.NET%\\crt\\src\\div.c\n#ifdef STATIC_IMAXDIV // [\nstatic\n#else // STATIC_IMAXDIV ][\n_inline\n#endif // STATIC_IMAXDIV ]\nimaxdiv_t __cdecl imaxdiv(intmax_t numer, intmax_t denom)\n{\n imaxdiv_t result;\n\n result.quot = numer / denom;\n result.rem = numer % denom;\n\n if (numer < 0 && result.rem > 0) {\n // did division wrong; must fix up\n ++result.quot;\n result.rem -= denom;\n }\n\n return result;\n}\n\n// 7.8.2.3 The strtoimax and strtoumax functions\n#define strtoimax _strtoi64\n#define strtoumax _strtoui64\n\n// 7.8.2.4 The wcstoimax and wcstoumax functions\n#define wcstoimax _wcstoi64\n#define wcstoumax _wcstoui64\n\n\n#endif // _MSC_INTTYPES_H_ ]\n" }, { "path": "deps/jemalloc/include/msvc_compat/stdbool.h", "content": "#ifndef stdbool_h\n#define stdbool_h\n\n#include \n\n/* MSVC doesn't define _Bool or bool in C, but does have BOOL */\n/* Note this doesn't pass autoconf's test because (bool) 0.5 != true */\ntypedef BOOL _Bool;\n\n#define bool _Bool\n#define true 1\n#define false 0\n\n#define __bool_true_false_are_defined 1\n\n#endif /* stdbool_h */\n" }, { "path": "deps/jemalloc/include/msvc_compat/stdint.h", "content": "// ISO C9x compliant stdint.h for Microsoft Visual Studio\n// Based on ISO/IEC 9899:TC2 Committee draft (May 6, 2005) WG14/N1124 \n// \n// Copyright (c) 2006-2008 Alexander Chemeris\n// \n// Redistribution and use in source and binary forms, with or without\n// modification, are permitted provided that the following conditions are met:\n// \n// 1. Redistributions of source code must retain the above copyright notice,\n// this list of conditions and the following disclaimer.\n// \n// 2. Redistributions in binary form must reproduce the above copyright\n// notice, this list of conditions and the following disclaimer in the\n// documentation and/or other materials provided with the distribution.\n// \n// 3. The name of the author may be used to endorse or promote products\n// derived from this software without specific prior written permission.\n// \n// THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED\n// WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF\n// MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO\n// EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\n// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,\n// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;\n// OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, \n// WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR\n// OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF\n// ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n// \n///////////////////////////////////////////////////////////////////////////////\n\n#ifndef _MSC_VER // [\n#error \"Use this header only with Microsoft Visual C++ compilers!\"\n#endif // _MSC_VER ]\n\n#ifndef _MSC_STDINT_H_ // [\n#define _MSC_STDINT_H_\n\n#if _MSC_VER > 1000\n#pragma once\n#endif\n\n#include \n\n// For Visual Studio 6 in C++ mode and for many Visual Studio versions when\n// compiling for ARM we should wrap include with 'extern \"C++\" {}'\n// or compiler give many errors like this:\n// error C2733: second C linkage of overloaded function 'wmemchr' not allowed\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n# include \n#ifdef __cplusplus\n}\n#endif\n\n// Define _W64 macros to mark types changing their size, like intptr_t.\n#ifndef _W64\n# if !defined(__midl) && (defined(_X86_) || defined(_M_IX86)) && _MSC_VER >= 1300\n# define _W64 __w64\n# else\n# define _W64\n# endif\n#endif\n\n\n// 7.18.1 Integer types\n\n// 7.18.1.1 Exact-width integer types\n\n// Visual Studio 6 and Embedded Visual C++ 4 doesn't\n// realize that, e.g. char has the same size as __int8\n// so we give up on __intX for them.\n#if (_MSC_VER < 1300)\n typedef signed char int8_t;\n typedef signed short int16_t;\n typedef signed int int32_t;\n typedef unsigned char uint8_t;\n typedef unsigned short uint16_t;\n typedef unsigned int uint32_t;\n#else\n typedef signed __int8 int8_t;\n typedef signed __int16 int16_t;\n typedef signed __int32 int32_t;\n typedef unsigned __int8 uint8_t;\n typedef unsigned __int16 uint16_t;\n typedef unsigned __int32 uint32_t;\n#endif\ntypedef signed __int64 int64_t;\ntypedef unsigned __int64 uint64_t;\n\n\n// 7.18.1.2 Minimum-width integer types\ntypedef int8_t int_least8_t;\ntypedef int16_t int_least16_t;\ntypedef int32_t int_least32_t;\ntypedef int64_t int_least64_t;\ntypedef uint8_t uint_least8_t;\ntypedef uint16_t uint_least16_t;\ntypedef uint32_t uint_least32_t;\ntypedef uint64_t uint_least64_t;\n\n// 7.18.1.3 Fastest minimum-width integer types\ntypedef int8_t int_fast8_t;\ntypedef int16_t int_fast16_t;\ntypedef int32_t int_fast32_t;\ntypedef int64_t int_fast64_t;\ntypedef uint8_t uint_fast8_t;\ntypedef uint16_t uint_fast16_t;\ntypedef uint32_t uint_fast32_t;\ntypedef uint64_t uint_fast64_t;\n\n// 7.18.1.4 Integer types capable of holding object pointers\n#ifdef _WIN64 // [\n typedef signed __int64 intptr_t;\n typedef unsigned __int64 uintptr_t;\n#else // _WIN64 ][\n typedef _W64 signed int intptr_t;\n typedef _W64 unsigned int uintptr_t;\n#endif // _WIN64 ]\n\n// 7.18.1.5 Greatest-width integer types\ntypedef int64_t intmax_t;\ntypedef uint64_t uintmax_t;\n\n\n// 7.18.2 Limits of specified-width integer types\n\n#if !defined(__cplusplus) || defined(__STDC_LIMIT_MACROS) // [ See footnote 220 at page 257 and footnote 221 at page 259\n\n// 7.18.2.1 Limits of exact-width integer types\n#define INT8_MIN ((int8_t)_I8_MIN)\n#define INT8_MAX _I8_MAX\n#define INT16_MIN ((int16_t)_I16_MIN)\n#define INT16_MAX _I16_MAX\n#define INT32_MIN ((int32_t)_I32_MIN)\n#define INT32_MAX _I32_MAX\n#define INT64_MIN ((int64_t)_I64_MIN)\n#define INT64_MAX _I64_MAX\n#define UINT8_MAX _UI8_MAX\n#define UINT16_MAX _UI16_MAX\n#define UINT32_MAX _UI32_MAX\n#define UINT64_MAX _UI64_MAX\n\n// 7.18.2.2 Limits of minimum-width integer types\n#define INT_LEAST8_MIN INT8_MIN\n#define INT_LEAST8_MAX INT8_MAX\n#define INT_LEAST16_MIN INT16_MIN\n#define INT_LEAST16_MAX INT16_MAX\n#define INT_LEAST32_MIN INT32_MIN\n#define INT_LEAST32_MAX INT32_MAX\n#define INT_LEAST64_MIN INT64_MIN\n#define INT_LEAST64_MAX INT64_MAX\n#define UINT_LEAST8_MAX UINT8_MAX\n#define UINT_LEAST16_MAX UINT16_MAX\n#define UINT_LEAST32_MAX UINT32_MAX\n#define UINT_LEAST64_MAX UINT64_MAX\n\n// 7.18.2.3 Limits of fastest minimum-width integer types\n#define INT_FAST8_MIN INT8_MIN\n#define INT_FAST8_MAX INT8_MAX\n#define INT_FAST16_MIN INT16_MIN\n#define INT_FAST16_MAX INT16_MAX\n#define INT_FAST32_MIN INT32_MIN\n#define INT_FAST32_MAX INT32_MAX\n#define INT_FAST64_MIN INT64_MIN\n#define INT_FAST64_MAX INT64_MAX\n#define UINT_FAST8_MAX UINT8_MAX\n#define UINT_FAST16_MAX UINT16_MAX\n#define UINT_FAST32_MAX UINT32_MAX\n#define UINT_FAST64_MAX UINT64_MAX\n\n// 7.18.2.4 Limits of integer types capable of holding object pointers\n#ifdef _WIN64 // [\n# define INTPTR_MIN INT64_MIN\n# define INTPTR_MAX INT64_MAX\n# define UINTPTR_MAX UINT64_MAX\n#else // _WIN64 ][\n# define INTPTR_MIN INT32_MIN\n# define INTPTR_MAX INT32_MAX\n# define UINTPTR_MAX UINT32_MAX\n#endif // _WIN64 ]\n\n// 7.18.2.5 Limits of greatest-width integer types\n#define INTMAX_MIN INT64_MIN\n#define INTMAX_MAX INT64_MAX\n#define UINTMAX_MAX UINT64_MAX\n\n// 7.18.3 Limits of other integer types\n\n#ifdef _WIN64 // [\n# define PTRDIFF_MIN _I64_MIN\n# define PTRDIFF_MAX _I64_MAX\n#else // _WIN64 ][\n# define PTRDIFF_MIN _I32_MIN\n# define PTRDIFF_MAX _I32_MAX\n#endif // _WIN64 ]\n\n#define SIG_ATOMIC_MIN INT_MIN\n#define SIG_ATOMIC_MAX INT_MAX\n\n#ifndef SIZE_MAX // [\n# ifdef _WIN64 // [\n# define SIZE_MAX _UI64_MAX\n# else // _WIN64 ][\n# define SIZE_MAX _UI32_MAX\n# endif // _WIN64 ]\n#endif // SIZE_MAX ]\n\n// WCHAR_MIN and WCHAR_MAX are also defined in \n#ifndef WCHAR_MIN // [\n# define WCHAR_MIN 0\n#endif // WCHAR_MIN ]\n#ifndef WCHAR_MAX // [\n# define WCHAR_MAX _UI16_MAX\n#endif // WCHAR_MAX ]\n\n#define WINT_MIN 0\n#define WINT_MAX _UI16_MAX\n\n#endif // __STDC_LIMIT_MACROS ]\n\n\n// 7.18.4 Limits of other integer types\n\n#if !defined(__cplusplus) || defined(__STDC_CONSTANT_MACROS) // [ See footnote 224 at page 260\n\n// 7.18.4.1 Macros for minimum-width integer constants\n\n#define INT8_C(val) val##i8\n#define INT16_C(val) val##i16\n#define INT32_C(val) val##i32\n#define INT64_C(val) val##i64\n\n#define UINT8_C(val) val##ui8\n#define UINT16_C(val) val##ui16\n#define UINT32_C(val) val##ui32\n#define UINT64_C(val) val##ui64\n\n// 7.18.4.2 Macros for greatest-width integer constants\n#define INTMAX_C INT64_C\n#define UINTMAX_C UINT64_C\n\n#endif // __STDC_CONSTANT_MACROS ]\n\n\n#endif // _MSC_STDINT_H_ ]\n" }, { "path": "deps/jemalloc/include/msvc_compat/strings.h", "content": "#ifndef strings_h\n#define strings_h\n\n/* MSVC doesn't define ffs/ffsl. This dummy strings.h header is provided\n * for both */\n#include \n#pragma intrinsic(_BitScanForward)\nstatic __forceinline int ffsl(long x)\n{\n\tunsigned long i;\n\n\tif (_BitScanForward(&i, x))\n\t\treturn (i + 1);\n\treturn (0);\n}\n\nstatic __forceinline int ffs(int x)\n{\n\n\treturn (ffsl(x));\n}\n\n#endif\n" }, { "path": "deps/jemalloc/install-sh", "content": "#! /bin/sh\n#\n# install - install a program, script, or datafile\n# This comes from X11R5 (mit/util/scripts/install.sh).\n#\n# Copyright 1991 by the Massachusetts Institute of Technology\n#\n# Permission to use, copy, modify, distribute, and sell this software and its\n# documentation for any purpose is hereby granted without fee, provided that\n# the above copyright notice appear in all copies and that both that\n# copyright notice and this permission notice appear in supporting\n# documentation, and that the name of M.I.T. not be used in advertising or\n# publicity pertaining to distribution of the software without specific,\n# written prior permission. M.I.T. makes no representations about the\n# suitability of this software for any purpose. It is provided \"as is\"\n# without express or implied warranty.\n#\n# Calling this script install-sh is preferred over install.sh, to prevent\n# `make' implicit rules from creating a file called install from it\n# when there is no Makefile.\n#\n# This script is compatible with the BSD install script, but was written\n# from scratch. It can only install one file at a time, a restriction\n# shared with many OS's install programs.\n\n\n# set DOITPROG to echo to test this script\n\n# Don't use :- since 4.3BSD and earlier shells don't like it.\ndoit=\"${DOITPROG-}\"\n\n\n# put in absolute paths if you don't have them in your path; or use env. vars.\n\nmvprog=\"${MVPROG-mv}\"\ncpprog=\"${CPPROG-cp}\"\nchmodprog=\"${CHMODPROG-chmod}\"\nchownprog=\"${CHOWNPROG-chown}\"\nchgrpprog=\"${CHGRPPROG-chgrp}\"\nstripprog=\"${STRIPPROG-strip}\"\nrmprog=\"${RMPROG-rm}\"\nmkdirprog=\"${MKDIRPROG-mkdir}\"\n\ntransformbasename=\"\"\ntransform_arg=\"\"\ninstcmd=\"$mvprog\"\nchmodcmd=\"$chmodprog 0755\"\nchowncmd=\"\"\nchgrpcmd=\"\"\nstripcmd=\"\"\nrmcmd=\"$rmprog -f\"\nmvcmd=\"$mvprog\"\nsrc=\"\"\ndst=\"\"\ndir_arg=\"\"\n\nwhile [ x\"$1\" != x ]; do\n case $1 in\n\t-c) instcmd=\"$cpprog\"\n\t shift\n\t continue;;\n\n\t-d) dir_arg=true\n\t shift\n\t continue;;\n\n\t-m) chmodcmd=\"$chmodprog $2\"\n\t shift\n\t shift\n\t continue;;\n\n\t-o) chowncmd=\"$chownprog $2\"\n\t shift\n\t shift\n\t continue;;\n\n\t-g) chgrpcmd=\"$chgrpprog $2\"\n\t shift\n\t shift\n\t continue;;\n\n\t-s) stripcmd=\"$stripprog\"\n\t shift\n\t continue;;\n\n\t-t=*) transformarg=`echo $1 | sed 's/-t=//'`\n\t shift\n\t continue;;\n\n\t-b=*) transformbasename=`echo $1 | sed 's/-b=//'`\n\t shift\n\t continue;;\n\n\t*) if [ x\"$src\" = x ]\n\t then\n\t\tsrc=$1\n\t else\n\t\t# this colon is to work around a 386BSD /bin/sh bug\n\t\t:\n\t\tdst=$1\n\t fi\n\t shift\n\t continue;;\n esac\ndone\n\nif [ x\"$src\" = x ]\nthen\n\techo \"install:\tno input file specified\"\n\texit 1\nelse\n\ttrue\nfi\n\nif [ x\"$dir_arg\" != x ]; then\n\tdst=$src\n\tsrc=\"\"\n\t\n\tif [ -d $dst ]; then\n\t\tinstcmd=:\n\telse\n\t\tinstcmd=mkdir\n\tfi\nelse\n\n# Waiting for this to be detected by the \"$instcmd $src $dsttmp\" command\n# might cause directories to be created, which would be especially bad \n# if $src (and thus $dsttmp) contains '*'.\n\n\tif [ -f $src -o -d $src ]\n\tthen\n\t\ttrue\n\telse\n\t\techo \"install: $src does not exist\"\n\t\texit 1\n\tfi\n\t\n\tif [ x\"$dst\" = x ]\n\tthen\n\t\techo \"install:\tno destination specified\"\n\t\texit 1\n\telse\n\t\ttrue\n\tfi\n\n# If destination is a directory, append the input filename; if your system\n# does not like double slashes in filenames, you may need to add some logic\n\n\tif [ -d $dst ]\n\tthen\n\t\tdst=\"$dst\"/`basename $src`\n\telse\n\t\ttrue\n\tfi\nfi\n\n## this sed command emulates the dirname command\ndstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`\n\n# Make sure that the destination directory exists.\n# this part is taken from Noah Friedman's mkinstalldirs script\n\n# Skip lots of stat calls in the usual case.\nif [ ! -d \"$dstdir\" ]; then\ndefaultIFS='\t\n'\nIFS=\"${IFS-${defaultIFS}}\"\n\noIFS=\"${IFS}\"\n# Some sh's can't handle IFS=/ for some reason.\nIFS='%'\nset - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`\nIFS=\"${oIFS}\"\n\npathcomp=''\n\nwhile [ $# -ne 0 ] ; do\n\tpathcomp=\"${pathcomp}${1}\"\n\tshift\n\n\tif [ ! -d \"${pathcomp}\" ] ;\n then\n\t\t$mkdirprog \"${pathcomp}\"\n\telse\n\t\ttrue\n\tfi\n\n\tpathcomp=\"${pathcomp}/\"\ndone\nfi\n\nif [ x\"$dir_arg\" != x ]\nthen\n\t$doit $instcmd $dst &&\n\n\tif [ x\"$chowncmd\" != x ]; then $doit $chowncmd $dst; else true ; fi &&\n\tif [ x\"$chgrpcmd\" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&\n\tif [ x\"$stripcmd\" != x ]; then $doit $stripcmd $dst; else true ; fi &&\n\tif [ x\"$chmodcmd\" != x ]; then $doit $chmodcmd $dst; else true ; fi\nelse\n\n# If we're going to rename the final executable, determine the name now.\n\n\tif [ x\"$transformarg\" = x ] \n\tthen\n\t\tdstfile=`basename $dst`\n\telse\n\t\tdstfile=`basename $dst $transformbasename | \n\t\t\tsed $transformarg`$transformbasename\n\tfi\n\n# don't allow the sed command to completely eliminate the filename\n\n\tif [ x\"$dstfile\" = x ] \n\tthen\n\t\tdstfile=`basename $dst`\n\telse\n\t\ttrue\n\tfi\n\n# Make a temp file name in the proper directory.\n\n\tdsttmp=$dstdir/#inst.$$#\n\n# Move or copy the file name to the temp name\n\n\t$doit $instcmd $src $dsttmp &&\n\n\ttrap \"rm -f ${dsttmp}\" 0 &&\n\n# and set any options; do chmod last to preserve setuid bits\n\n# If any of these fail, we abort the whole thing. If we want to\n# ignore errors from any of these, just make sure not to ignore\n# errors from the above \"$doit $instcmd $src $dsttmp\" command.\n\n\tif [ x\"$chowncmd\" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&\n\tif [ x\"$chgrpcmd\" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&\n\tif [ x\"$stripcmd\" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&\n\tif [ x\"$chmodcmd\" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&\n\n# Now rename the file to the real destination.\n\n\t$doit $rmcmd -f $dstdir/$dstfile &&\n\t$doit $mvcmd $dsttmp $dstdir/$dstfile \n\nfi &&\n\n\nexit 0\n" }, { "path": "deps/jemalloc/src/arena.c", "content": "#define\tJEMALLOC_ARENA_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nssize_t\t\topt_lg_dirty_mult = LG_DIRTY_MULT_DEFAULT;\narena_bin_info_t\tarena_bin_info[NBINS];\n\nJEMALLOC_ALIGNED(CACHELINE)\nconst uint8_t\tsmall_size2bin[] = {\n#define\tS2B_8(i)\ti,\n#define\tS2B_16(i)\tS2B_8(i) S2B_8(i)\n#define\tS2B_32(i)\tS2B_16(i) S2B_16(i)\n#define\tS2B_64(i)\tS2B_32(i) S2B_32(i)\n#define\tS2B_128(i)\tS2B_64(i) S2B_64(i)\n#define\tS2B_256(i)\tS2B_128(i) S2B_128(i)\n#define\tS2B_512(i)\tS2B_256(i) S2B_256(i)\n#define\tS2B_1024(i)\tS2B_512(i) S2B_512(i)\n#define\tS2B_2048(i)\tS2B_1024(i) S2B_1024(i)\n#define\tS2B_4096(i)\tS2B_2048(i) S2B_2048(i)\n#define\tS2B_8192(i)\tS2B_4096(i) S2B_4096(i)\n#define\tSIZE_CLASS(bin, delta, size)\t\t\t\t\t\\\n\tS2B_##delta(bin)\n\tSIZE_CLASSES\n#undef S2B_8\n#undef S2B_16\n#undef S2B_32\n#undef S2B_64\n#undef S2B_128\n#undef S2B_256\n#undef S2B_512\n#undef S2B_1024\n#undef S2B_2048\n#undef S2B_4096\n#undef S2B_8192\n#undef SIZE_CLASS\n};\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\tarena_avail_insert(arena_t *arena, arena_chunk_t *chunk,\n size_t pageind, size_t npages, bool maybe_adjac_pred,\n bool maybe_adjac_succ);\nstatic void\tarena_avail_remove(arena_t *arena, arena_chunk_t *chunk,\n size_t pageind, size_t npages, bool maybe_adjac_pred,\n bool maybe_adjac_succ);\nstatic void\tarena_run_split(arena_t *arena, arena_run_t *run, size_t size,\n bool large, size_t binind, bool zero);\nstatic arena_chunk_t *arena_chunk_alloc(arena_t *arena);\nstatic void\tarena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk);\nstatic arena_run_t\t*arena_run_alloc_helper(arena_t *arena, size_t size,\n bool large, size_t binind, bool zero);\nstatic arena_run_t *arena_run_alloc(arena_t *arena, size_t size, bool large,\n size_t binind, bool zero);\nstatic arena_chunk_t\t*chunks_dirty_iter_cb(arena_chunk_tree_t *tree,\n arena_chunk_t *chunk, void *arg);\nstatic void\tarena_purge(arena_t *arena, bool all);\nstatic void\tarena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty,\n bool cleaned);\nstatic void\tarena_run_trim_head(arena_t *arena, arena_chunk_t *chunk,\n arena_run_t *run, size_t oldsize, size_t newsize);\nstatic void\tarena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk,\n arena_run_t *run, size_t oldsize, size_t newsize, bool dirty);\nstatic arena_run_t\t*arena_bin_runs_first(arena_bin_t *bin);\nstatic void\tarena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run);\nstatic void\tarena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run);\nstatic arena_run_t *arena_bin_nonfull_run_tryget(arena_bin_t *bin);\nstatic arena_run_t *arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin);\nstatic void\t*arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin);\nstatic void\tarena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run,\n arena_bin_t *bin);\nstatic void\tarena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk,\n arena_run_t *run, arena_bin_t *bin);\nstatic void\tarena_bin_lower_run(arena_t *arena, arena_chunk_t *chunk,\n arena_run_t *run, arena_bin_t *bin);\nstatic void\tarena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk,\n void *ptr, size_t oldsize, size_t size);\nstatic bool\tarena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk,\n void *ptr, size_t oldsize, size_t size, size_t extra, bool zero);\nstatic bool\tarena_ralloc_large(void *ptr, size_t oldsize, size_t size,\n size_t extra, bool zero);\nstatic size_t\tbin_info_run_size_calc(arena_bin_info_t *bin_info,\n size_t min_run_size);\nstatic void\tbin_info_init(void);\n\n/******************************************************************************/\n\nstatic inline int\narena_run_comp(arena_chunk_map_t *a, arena_chunk_map_t *b)\n{\n\tuintptr_t a_mapelm = (uintptr_t)a;\n\tuintptr_t b_mapelm = (uintptr_t)b;\n\n\tassert(a != NULL);\n\tassert(b != NULL);\n\n\treturn ((a_mapelm > b_mapelm) - (a_mapelm < b_mapelm));\n}\n\n/* Generate red-black tree functions. */\nrb_gen(static UNUSED, arena_run_tree_, arena_run_tree_t, arena_chunk_map_t,\n u.rb_link, arena_run_comp)\n\nstatic inline int\narena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b)\n{\n\tint ret;\n\tsize_t a_size = a->bits & ~PAGE_MASK;\n\tsize_t b_size = b->bits & ~PAGE_MASK;\n\n\tret = (a_size > b_size) - (a_size < b_size);\n\tif (ret == 0) {\n\t\tuintptr_t a_mapelm, b_mapelm;\n\n\t\tif ((a->bits & CHUNK_MAP_KEY) != CHUNK_MAP_KEY)\n\t\t\ta_mapelm = (uintptr_t)a;\n\t\telse {\n\t\t\t/*\n\t\t\t * Treat keys as though they are lower than anything\n\t\t\t * else.\n\t\t\t */\n\t\t\ta_mapelm = 0;\n\t\t}\n\t\tb_mapelm = (uintptr_t)b;\n\n\t\tret = (a_mapelm > b_mapelm) - (a_mapelm < b_mapelm);\n\t}\n\n\treturn (ret);\n}\n\n/* Generate red-black tree functions. */\nrb_gen(static UNUSED, arena_avail_tree_, arena_avail_tree_t, arena_chunk_map_t,\n u.rb_link, arena_avail_comp)\n\nstatic inline int\narena_chunk_dirty_comp(arena_chunk_t *a, arena_chunk_t *b)\n{\n\n\tassert(a != NULL);\n\tassert(b != NULL);\n\n\t/*\n\t * Short-circuit for self comparison. The following comparison code\n\t * would come to the same result, but at the cost of executing the slow\n\t * path.\n\t */\n\tif (a == b)\n\t\treturn (0);\n\n\t/*\n\t * Order such that chunks with higher fragmentation are \"less than\"\n\t * those with lower fragmentation -- purging order is from \"least\" to\n\t * \"greatest\". Fragmentation is measured as:\n\t *\n\t * mean current avail run size\n\t * --------------------------------\n\t * mean defragmented avail run size\n\t *\n\t * navail\n\t * -----------\n\t * nruns_avail nruns_avail-nruns_adjac\n\t * = ========================= = -----------------------\n\t * navail nruns_avail\n\t * -----------------------\n\t * nruns_avail-nruns_adjac\n\t *\n\t * The following code multiplies away the denominator prior to\n\t * comparison, in order to avoid division.\n\t *\n\t */\n\t{\n\t\tsize_t a_val = (a->nruns_avail - a->nruns_adjac) *\n\t\t b->nruns_avail;\n\t\tsize_t b_val = (b->nruns_avail - b->nruns_adjac) *\n\t\t a->nruns_avail;\n\n\t\tif (a_val < b_val)\n\t\t\treturn (1);\n\t\tif (a_val > b_val)\n\t\t\treturn (-1);\n\t}\n\t/*\n\t * Break ties by chunk address. For fragmented chunks, report lower\n\t * addresses as \"lower\", so that fragmentation reduction happens first\n\t * at lower addresses. However, use the opposite ordering for\n\t * unfragmented chunks, in order to increase the chances of\n\t * re-allocating dirty runs.\n\t */\n\t{\n\t\tuintptr_t a_chunk = (uintptr_t)a;\n\t\tuintptr_t b_chunk = (uintptr_t)b;\n\t\tint ret = ((a_chunk > b_chunk) - (a_chunk < b_chunk));\n\t\tif (a->nruns_adjac == 0) {\n\t\t\tassert(b->nruns_adjac == 0);\n\t\t\tret = -ret;\n\t\t}\n\t\treturn (ret);\n\t}\n}\n\n/* Generate red-black tree functions. */\nrb_gen(static UNUSED, arena_chunk_dirty_, arena_chunk_tree_t, arena_chunk_t,\n dirty_link, arena_chunk_dirty_comp)\n\nstatic inline bool\narena_avail_adjac_pred(arena_chunk_t *chunk, size_t pageind)\n{\n\tbool ret;\n\n\tif (pageind-1 < map_bias)\n\t\tret = false;\n\telse {\n\t\tret = (arena_mapbits_allocated_get(chunk, pageind-1) == 0);\n\t\tassert(ret == false || arena_mapbits_dirty_get(chunk,\n\t\t pageind-1) != arena_mapbits_dirty_get(chunk, pageind));\n\t}\n\treturn (ret);\n}\n\nstatic inline bool\narena_avail_adjac_succ(arena_chunk_t *chunk, size_t pageind, size_t npages)\n{\n\tbool ret;\n\n\tif (pageind+npages == chunk_npages)\n\t\tret = false;\n\telse {\n\t\tassert(pageind+npages < chunk_npages);\n\t\tret = (arena_mapbits_allocated_get(chunk, pageind+npages) == 0);\n\t\tassert(ret == false || arena_mapbits_dirty_get(chunk, pageind)\n\t\t != arena_mapbits_dirty_get(chunk, pageind+npages));\n\t}\n\treturn (ret);\n}\n\nstatic inline bool\narena_avail_adjac(arena_chunk_t *chunk, size_t pageind, size_t npages)\n{\n\n\treturn (arena_avail_adjac_pred(chunk, pageind) ||\n\t arena_avail_adjac_succ(chunk, pageind, npages));\n}\n\nstatic void\narena_avail_insert(arena_t *arena, arena_chunk_t *chunk, size_t pageind,\n size_t npages, bool maybe_adjac_pred, bool maybe_adjac_succ)\n{\n\n\tassert(npages == (arena_mapbits_unallocated_size_get(chunk, pageind) >>\n\t LG_PAGE));\n\n\t/*\n\t * chunks_dirty is keyed by nruns_{avail,adjac}, so the chunk must be\n\t * removed and reinserted even if the run to be inserted is clean.\n\t */\n\tif (chunk->ndirty != 0)\n\t\tarena_chunk_dirty_remove(&arena->chunks_dirty, chunk);\n\n\tif (maybe_adjac_pred && arena_avail_adjac_pred(chunk, pageind))\n\t\tchunk->nruns_adjac++;\n\tif (maybe_adjac_succ && arena_avail_adjac_succ(chunk, pageind, npages))\n\t\tchunk->nruns_adjac++;\n\tchunk->nruns_avail++;\n\tassert(chunk->nruns_avail > chunk->nruns_adjac);\n\n\tif (arena_mapbits_dirty_get(chunk, pageind) != 0) {\n\t\tarena->ndirty += npages;\n\t\tchunk->ndirty += npages;\n\t}\n\tif (chunk->ndirty != 0)\n\t\tarena_chunk_dirty_insert(&arena->chunks_dirty, chunk);\n\n\tarena_avail_tree_insert(&arena->runs_avail, arena_mapp_get(chunk,\n\t pageind));\n}\n\nstatic void\narena_avail_remove(arena_t *arena, arena_chunk_t *chunk, size_t pageind,\n size_t npages, bool maybe_adjac_pred, bool maybe_adjac_succ)\n{\n\n\tassert(npages == (arena_mapbits_unallocated_size_get(chunk, pageind) >>\n\t LG_PAGE));\n\n\t/*\n\t * chunks_dirty is keyed by nruns_{avail,adjac}, so the chunk must be\n\t * removed and reinserted even if the run to be removed is clean.\n\t */\n\tif (chunk->ndirty != 0)\n\t\tarena_chunk_dirty_remove(&arena->chunks_dirty, chunk);\n\n\tif (maybe_adjac_pred && arena_avail_adjac_pred(chunk, pageind))\n\t\tchunk->nruns_adjac--;\n\tif (maybe_adjac_succ && arena_avail_adjac_succ(chunk, pageind, npages))\n\t\tchunk->nruns_adjac--;\n\tchunk->nruns_avail--;\n\tassert(chunk->nruns_avail > chunk->nruns_adjac || (chunk->nruns_avail\n\t == 0 && chunk->nruns_adjac == 0));\n\n\tif (arena_mapbits_dirty_get(chunk, pageind) != 0) {\n\t\tarena->ndirty -= npages;\n\t\tchunk->ndirty -= npages;\n\t}\n\tif (chunk->ndirty != 0)\n\t\tarena_chunk_dirty_insert(&arena->chunks_dirty, chunk);\n\n\tarena_avail_tree_remove(&arena->runs_avail, arena_mapp_get(chunk,\n\t pageind));\n}\n\nstatic inline void *\narena_run_reg_alloc(arena_run_t *run, arena_bin_info_t *bin_info)\n{\n\tvoid *ret;\n\tunsigned regind;\n\tbitmap_t *bitmap = (bitmap_t *)((uintptr_t)run +\n\t (uintptr_t)bin_info->bitmap_offset);\n\n\tassert(run->nfree > 0);\n\tassert(bitmap_full(bitmap, &bin_info->bitmap_info) == false);\n\n\tregind = bitmap_sfu(bitmap, &bin_info->bitmap_info);\n\tret = (void *)((uintptr_t)run + (uintptr_t)bin_info->reg0_offset +\n\t (uintptr_t)(bin_info->reg_interval * regind));\n\trun->nfree--;\n\tif (regind == run->nextind)\n\t\trun->nextind++;\n\tassert(regind < run->nextind);\n\treturn (ret);\n}\n\nstatic inline void\narena_run_reg_dalloc(arena_run_t *run, void *ptr)\n{\n\tarena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\tsize_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tsize_t mapbits = arena_mapbits_get(chunk, pageind);\n\tsize_t binind = arena_ptr_small_binind_get(ptr, mapbits);\n\tarena_bin_info_t *bin_info = &arena_bin_info[binind];\n\tunsigned regind = arena_run_regind(run, bin_info, ptr);\n\tbitmap_t *bitmap = (bitmap_t *)((uintptr_t)run +\n\t (uintptr_t)bin_info->bitmap_offset);\n\n\tassert(run->nfree < bin_info->nregs);\n\t/* Freeing an interior pointer can cause assertion failure. */\n\tassert(((uintptr_t)ptr - ((uintptr_t)run +\n\t (uintptr_t)bin_info->reg0_offset)) %\n\t (uintptr_t)bin_info->reg_interval == 0);\n\tassert((uintptr_t)ptr >= (uintptr_t)run +\n\t (uintptr_t)bin_info->reg0_offset);\n\t/* Freeing an unallocated pointer can cause assertion failure. */\n\tassert(bitmap_get(bitmap, &bin_info->bitmap_info, regind));\n\n\tbitmap_unset(bitmap, &bin_info->bitmap_info, regind);\n\trun->nfree++;\n}\n\nstatic inline void\narena_chunk_validate_zeroed(arena_chunk_t *chunk, size_t run_ind)\n{\n\tsize_t i;\n\tUNUSED size_t *p = (size_t *)((uintptr_t)chunk + (run_ind << LG_PAGE));\n\n\tfor (i = 0; i < PAGE / sizeof(size_t); i++)\n\t\tassert(p[i] == 0);\n}\n\nstatic void\narena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large,\n size_t binind, bool zero)\n{\n\tarena_chunk_t *chunk;\n\tsize_t run_ind, total_pages, need_pages, rem_pages, i;\n\tsize_t flag_dirty;\n\n\tassert((large && binind == BININD_INVALID) || (large == false && binind\n\t != BININD_INVALID));\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\trun_ind = (unsigned)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE);\n\tflag_dirty = arena_mapbits_dirty_get(chunk, run_ind);\n\ttotal_pages = arena_mapbits_unallocated_size_get(chunk, run_ind) >>\n\t LG_PAGE;\n\tassert(arena_mapbits_dirty_get(chunk, run_ind+total_pages-1) ==\n\t flag_dirty);\n\tneed_pages = (size >> LG_PAGE);\n\tassert(need_pages > 0);\n\tassert(need_pages <= total_pages);\n\trem_pages = total_pages - need_pages;\n\n\tarena_avail_remove(arena, chunk, run_ind, total_pages, true, true);\n\tif (config_stats) {\n\t\t/*\n\t\t * Update stats_cactive if nactive is crossing a chunk\n\t\t * multiple.\n\t\t */\n\t\tsize_t cactive_diff = CHUNK_CEILING((arena->nactive +\n\t\t need_pages) << LG_PAGE) - CHUNK_CEILING(arena->nactive <<\n\t\t LG_PAGE);\n\t\tif (cactive_diff != 0)\n\t\t\tstats_cactive_add(cactive_diff);\n\t}\n\tarena->nactive += need_pages;\n\n\t/* Keep track of trailing unused pages for later use. */\n\tif (rem_pages > 0) {\n\t\tif (flag_dirty != 0) {\n\t\t\tarena_mapbits_unallocated_set(chunk, run_ind+need_pages,\n\t\t\t (rem_pages << LG_PAGE), CHUNK_MAP_DIRTY);\n\t\t\tarena_mapbits_unallocated_set(chunk,\n\t\t\t run_ind+total_pages-1, (rem_pages << LG_PAGE),\n\t\t\t CHUNK_MAP_DIRTY);\n\t\t} else {\n\t\t\tarena_mapbits_unallocated_set(chunk, run_ind+need_pages,\n\t\t\t (rem_pages << LG_PAGE),\n\t\t\t arena_mapbits_unzeroed_get(chunk,\n\t\t\t run_ind+need_pages));\n\t\t\tarena_mapbits_unallocated_set(chunk,\n\t\t\t run_ind+total_pages-1, (rem_pages << LG_PAGE),\n\t\t\t arena_mapbits_unzeroed_get(chunk,\n\t\t\t run_ind+total_pages-1));\n\t\t}\n\t\tarena_avail_insert(arena, chunk, run_ind+need_pages, rem_pages,\n\t\t false, true);\n\t}\n\n\t/*\n\t * Update the page map separately for large vs. small runs, since it is\n\t * possible to avoid iteration for large mallocs.\n\t */\n\tif (large) {\n\t\tif (zero) {\n\t\t\tif (flag_dirty == 0) {\n\t\t\t\t/*\n\t\t\t\t * The run is clean, so some pages may be\n\t\t\t\t * zeroed (i.e. never before touched).\n\t\t\t\t */\n\t\t\t\tfor (i = 0; i < need_pages; i++) {\n\t\t\t\t\tif (arena_mapbits_unzeroed_get(chunk,\n\t\t\t\t\t run_ind+i) != 0) {\n\t\t\t\t\t\tVALGRIND_MAKE_MEM_UNDEFINED(\n\t\t\t\t\t\t (void *)((uintptr_t)\n\t\t\t\t\t\t chunk + ((run_ind+i) <<\n\t\t\t\t\t\t LG_PAGE)), PAGE);\n\t\t\t\t\t\tmemset((void *)((uintptr_t)\n\t\t\t\t\t\t chunk + ((run_ind+i) <<\n\t\t\t\t\t\t LG_PAGE)), 0, PAGE);\n\t\t\t\t\t} else if (config_debug) {\n\t\t\t\t\t\tVALGRIND_MAKE_MEM_DEFINED(\n\t\t\t\t\t\t (void *)((uintptr_t)\n\t\t\t\t\t\t chunk + ((run_ind+i) <<\n\t\t\t\t\t\t LG_PAGE)), PAGE);\n\t\t\t\t\t\tarena_chunk_validate_zeroed(\n\t\t\t\t\t\t chunk, run_ind+i);\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\t/*\n\t\t\t\t * The run is dirty, so all pages must be\n\t\t\t\t * zeroed.\n\t\t\t\t */\n\t\t\t\tVALGRIND_MAKE_MEM_UNDEFINED((void\n\t\t\t\t *)((uintptr_t)chunk + (run_ind <<\n\t\t\t\t LG_PAGE)), (need_pages << LG_PAGE));\n\t\t\t\tmemset((void *)((uintptr_t)chunk + (run_ind <<\n\t\t\t\t LG_PAGE)), 0, (need_pages << LG_PAGE));\n\t\t\t}\n\t\t}\n\n\t\t/*\n\t\t * Set the last element first, in case the run only contains one\n\t\t * page (i.e. both statements set the same element).\n\t\t */\n\t\tarena_mapbits_large_set(chunk, run_ind+need_pages-1, 0,\n\t\t flag_dirty);\n\t\tarena_mapbits_large_set(chunk, run_ind, size, flag_dirty);\n\t} else {\n\t\tassert(zero == false);\n\t\t/*\n\t\t * Propagate the dirty and unzeroed flags to the allocated\n\t\t * small run, so that arena_dalloc_bin_run() has the ability to\n\t\t * conditionally trim clean pages.\n\t\t */\n\t\tarena_mapbits_small_set(chunk, run_ind, 0, binind, flag_dirty);\n\t\t/*\n\t\t * The first page will always be dirtied during small run\n\t\t * initialization, so a validation failure here would not\n\t\t * actually cause an observable failure.\n\t\t */\n\t\tif (config_debug && flag_dirty == 0 &&\n\t\t arena_mapbits_unzeroed_get(chunk, run_ind) == 0)\n\t\t\tarena_chunk_validate_zeroed(chunk, run_ind);\n\t\tfor (i = 1; i < need_pages - 1; i++) {\n\t\t\tarena_mapbits_small_set(chunk, run_ind+i, i, binind, 0);\n\t\t\tif (config_debug && flag_dirty == 0 &&\n\t\t\t arena_mapbits_unzeroed_get(chunk, run_ind+i) == 0)\n\t\t\t\tarena_chunk_validate_zeroed(chunk, run_ind+i);\n\t\t}\n\t\tarena_mapbits_small_set(chunk, run_ind+need_pages-1,\n\t\t need_pages-1, binind, flag_dirty);\n\t\tif (config_debug && flag_dirty == 0 &&\n\t\t arena_mapbits_unzeroed_get(chunk, run_ind+need_pages-1) ==\n\t\t 0) {\n\t\t\tarena_chunk_validate_zeroed(chunk,\n\t\t\t run_ind+need_pages-1);\n\t\t}\n\t}\n}\n\nstatic arena_chunk_t *\narena_chunk_alloc(arena_t *arena)\n{\n\tarena_chunk_t *chunk;\n\tsize_t i;\n\n\tif (arena->spare != NULL) {\n\t\tchunk = arena->spare;\n\t\tarena->spare = NULL;\n\n\t\tassert(arena_mapbits_allocated_get(chunk, map_bias) == 0);\n\t\tassert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0);\n\t\tassert(arena_mapbits_unallocated_size_get(chunk, map_bias) ==\n\t\t arena_maxclass);\n\t\tassert(arena_mapbits_unallocated_size_get(chunk,\n\t\t chunk_npages-1) == arena_maxclass);\n\t\tassert(arena_mapbits_dirty_get(chunk, map_bias) ==\n\t\t arena_mapbits_dirty_get(chunk, chunk_npages-1));\n\t} else {\n\t\tbool zero;\n\t\tsize_t unzeroed;\n\n\t\tzero = false;\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t\tchunk = (arena_chunk_t *)chunk_alloc(chunksize, chunksize,\n\t\t false, &zero, arena->dss_prec);\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tif (chunk == NULL)\n\t\t\treturn (NULL);\n\t\tif (config_stats)\n\t\t\tarena->stats.mapped += chunksize;\n\n\t\tchunk->arena = arena;\n\n\t\t/*\n\t\t * Claim that no pages are in use, since the header is merely\n\t\t * overhead.\n\t\t */\n\t\tchunk->ndirty = 0;\n\n\t\tchunk->nruns_avail = 0;\n\t\tchunk->nruns_adjac = 0;\n\n\t\t/*\n\t\t * Initialize the map to contain one maximal free untouched run.\n\t\t * Mark the pages as zeroed iff chunk_alloc() returned a zeroed\n\t\t * chunk.\n\t\t */\n\t\tunzeroed = zero ? 0 : CHUNK_MAP_UNZEROED;\n\t\tarena_mapbits_unallocated_set(chunk, map_bias, arena_maxclass,\n\t\t unzeroed);\n\t\t/*\n\t\t * There is no need to initialize the internal page map entries\n\t\t * unless the chunk is not zeroed.\n\t\t */\n\t\tif (zero == false) {\n\t\t\tfor (i = map_bias+1; i < chunk_npages-1; i++)\n\t\t\t\tarena_mapbits_unzeroed_set(chunk, i, unzeroed);\n\t\t} else if (config_debug) {\n\t\t\tfor (i = map_bias+1; i < chunk_npages-1; i++) {\n\t\t\t\tassert(arena_mapbits_unzeroed_get(chunk, i) ==\n\t\t\t\t unzeroed);\n\t\t\t}\n\t\t}\n\t\tarena_mapbits_unallocated_set(chunk, chunk_npages-1,\n\t\t arena_maxclass, unzeroed);\n\t}\n\n\t/* Insert the run into the runs_avail tree. */\n\tarena_avail_insert(arena, chunk, map_bias, chunk_npages-map_bias,\n\t false, false);\n\n\treturn (chunk);\n}\n\nstatic void\narena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk)\n{\n\tassert(arena_mapbits_allocated_get(chunk, map_bias) == 0);\n\tassert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0);\n\tassert(arena_mapbits_unallocated_size_get(chunk, map_bias) ==\n\t arena_maxclass);\n\tassert(arena_mapbits_unallocated_size_get(chunk, chunk_npages-1) ==\n\t arena_maxclass);\n\tassert(arena_mapbits_dirty_get(chunk, map_bias) ==\n\t arena_mapbits_dirty_get(chunk, chunk_npages-1));\n\n\t/*\n\t * Remove run from the runs_avail tree, so that the arena does not use\n\t * it.\n\t */\n\tarena_avail_remove(arena, chunk, map_bias, chunk_npages-map_bias,\n\t false, false);\n\n\tif (arena->spare != NULL) {\n\t\tarena_chunk_t *spare = arena->spare;\n\n\t\tarena->spare = chunk;\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t\tchunk_dealloc((void *)spare, chunksize, true);\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tif (config_stats)\n\t\t\tarena->stats.mapped -= chunksize;\n\t} else\n\t\tarena->spare = chunk;\n}\n\nstatic arena_run_t *\narena_run_alloc_helper(arena_t *arena, size_t size, bool large, size_t binind,\n bool zero)\n{\n\tarena_run_t *run;\n\tarena_chunk_map_t *mapelm, key;\n\n\tkey.bits = size | CHUNK_MAP_KEY;\n\tmapelm = arena_avail_tree_nsearch(&arena->runs_avail, &key);\n\tif (mapelm != NULL) {\n\t\tarena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm);\n\t\tsize_t pageind = (((uintptr_t)mapelm -\n\t\t (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t))\n\t\t + map_bias;\n\n\t\trun = (arena_run_t *)((uintptr_t)run_chunk + (pageind <<\n\t\t LG_PAGE));\n\t\tarena_run_split(arena, run, size, large, binind, zero);\n\t\treturn (run);\n\t}\n\n\treturn (NULL);\n}\n\nstatic arena_run_t *\narena_run_alloc(arena_t *arena, size_t size, bool large, size_t binind,\n bool zero)\n{\n\tarena_chunk_t *chunk;\n\tarena_run_t *run;\n\n\tassert(size <= arena_maxclass);\n\tassert((size & PAGE_MASK) == 0);\n\tassert((large && binind == BININD_INVALID) || (large == false && binind\n\t != BININD_INVALID));\n\n\t/* Search the arena's chunks for the lowest best fit. */\n\trun = arena_run_alloc_helper(arena, size, large, binind, zero);\n\tif (run != NULL)\n\t\treturn (run);\n\n\t/*\n\t * No usable runs. Create a new chunk from which to allocate the run.\n\t */\n\tchunk = arena_chunk_alloc(arena);\n\tif (chunk != NULL) {\n\t\trun = (arena_run_t *)((uintptr_t)chunk + (map_bias << LG_PAGE));\n\t\tarena_run_split(arena, run, size, large, binind, zero);\n\t\treturn (run);\n\t}\n\n\t/*\n\t * arena_chunk_alloc() failed, but another thread may have made\n\t * sufficient memory available while this one dropped arena->lock in\n\t * arena_chunk_alloc(), so search one more time.\n\t */\n\treturn (arena_run_alloc_helper(arena, size, large, binind, zero));\n}\n\nstatic inline void\narena_maybe_purge(arena_t *arena)\n{\n\tsize_t npurgeable, threshold;\n\n\t/* Don't purge if the option is disabled. */\n\tif (opt_lg_dirty_mult < 0)\n\t\treturn;\n\t/* Don't purge if all dirty pages are already being purged. */\n\tif (arena->ndirty <= arena->npurgatory)\n\t\treturn;\n\tnpurgeable = arena->ndirty - arena->npurgatory;\n\tthreshold = (arena->nactive >> opt_lg_dirty_mult);\n\t/*\n\t * Don't purge unless the number of purgeable pages exceeds the\n\t * threshold.\n\t */\n\tif (npurgeable <= threshold)\n\t\treturn;\n\n\tarena_purge(arena, false);\n}\n\nstatic inline size_t\narena_chunk_purge(arena_t *arena, arena_chunk_t *chunk, bool all)\n{\n\tsize_t npurged;\n\tql_head(arena_chunk_map_t) mapelms;\n\tarena_chunk_map_t *mapelm;\n\tsize_t pageind, npages;\n\tsize_t nmadvise;\n\n\tql_new(&mapelms);\n\n\t/*\n\t * If chunk is the spare, temporarily re-allocate it, 1) so that its\n\t * run is reinserted into runs_avail, and 2) so that it cannot be\n\t * completely discarded by another thread while arena->lock is dropped\n\t * by this thread. Note that the arena_run_dalloc() call will\n\t * implicitly deallocate the chunk, so no explicit action is required\n\t * in this function to deallocate the chunk.\n\t *\n\t * Note that once a chunk contains dirty pages, it cannot again contain\n\t * a single run unless 1) it is a dirty run, or 2) this function purges\n\t * dirty pages and causes the transition to a single clean run. Thus\n\t * (chunk == arena->spare) is possible, but it is not possible for\n\t * this function to be called on the spare unless it contains a dirty\n\t * run.\n\t */\n\tif (chunk == arena->spare) {\n\t\tassert(arena_mapbits_dirty_get(chunk, map_bias) != 0);\n\t\tassert(arena_mapbits_dirty_get(chunk, chunk_npages-1) != 0);\n\n\t\tarena_chunk_alloc(arena);\n\t}\n\n\tif (config_stats)\n\t\tarena->stats.purged += chunk->ndirty;\n\n\t/*\n\t * Operate on all dirty runs if there is no clean/dirty run\n\t * fragmentation.\n\t */\n\tif (chunk->nruns_adjac == 0)\n\t\tall = true;\n\n\t/*\n\t * Temporarily allocate free dirty runs within chunk. If all is false,\n\t * only operate on dirty runs that are fragments; otherwise operate on\n\t * all dirty runs.\n\t */\n\tfor (pageind = map_bias; pageind < chunk_npages; pageind += npages) {\n\t\tmapelm = arena_mapp_get(chunk, pageind);\n\t\tif (arena_mapbits_allocated_get(chunk, pageind) == 0) {\n\t\t\tsize_t run_size =\n\t\t\t arena_mapbits_unallocated_size_get(chunk, pageind);\n\n\t\t\tnpages = run_size >> LG_PAGE;\n\t\t\tassert(pageind + npages <= chunk_npages);\n\t\t\tassert(arena_mapbits_dirty_get(chunk, pageind) ==\n\t\t\t arena_mapbits_dirty_get(chunk, pageind+npages-1));\n\n\t\t\tif (arena_mapbits_dirty_get(chunk, pageind) != 0 &&\n\t\t\t (all || arena_avail_adjac(chunk, pageind,\n\t\t\t npages))) {\n\t\t\t\tarena_run_t *run = (arena_run_t *)((uintptr_t)\n\t\t\t\t chunk + (uintptr_t)(pageind << LG_PAGE));\n\n\t\t\t\tarena_run_split(arena, run, run_size, true,\n\t\t\t\t BININD_INVALID, false);\n\t\t\t\t/* Append to list for later processing. */\n\t\t\t\tql_elm_new(mapelm, u.ql_link);\n\t\t\t\tql_tail_insert(&mapelms, mapelm, u.ql_link);\n\t\t\t}\n\t\t} else {\n\t\t\t/* Skip run. */\n\t\t\tif (arena_mapbits_large_get(chunk, pageind) != 0) {\n\t\t\t\tnpages = arena_mapbits_large_size_get(chunk,\n\t\t\t\t pageind) >> LG_PAGE;\n\t\t\t} else {\n\t\t\t\tsize_t binind;\n\t\t\t\tarena_bin_info_t *bin_info;\n\t\t\t\tarena_run_t *run = (arena_run_t *)((uintptr_t)\n\t\t\t\t chunk + (uintptr_t)(pageind << LG_PAGE));\n\n\t\t\t\tassert(arena_mapbits_small_runind_get(chunk,\n\t\t\t\t pageind) == 0);\n\t\t\t\tbinind = arena_bin_index(arena, run->bin);\n\t\t\t\tbin_info = &arena_bin_info[binind];\n\t\t\t\tnpages = bin_info->run_size >> LG_PAGE;\n\t\t\t}\n\t\t}\n\t}\n\tassert(pageind == chunk_npages);\n\tassert(chunk->ndirty == 0 || all == false);\n\tassert(chunk->nruns_adjac == 0);\n\n\tmalloc_mutex_unlock(&arena->lock);\n\tif (config_stats)\n\t\tnmadvise = 0;\n\tnpurged = 0;\n\tql_foreach(mapelm, &mapelms, u.ql_link) {\n\t\tbool unzeroed;\n\t\tsize_t flag_unzeroed, i;\n\n\t\tpageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) /\n\t\t sizeof(arena_chunk_map_t)) + map_bias;\n\t\tnpages = arena_mapbits_large_size_get(chunk, pageind) >>\n\t\t LG_PAGE;\n\t\tassert(pageind + npages <= chunk_npages);\n\t\tunzeroed = pages_purge((void *)((uintptr_t)chunk + (pageind <<\n\t\t LG_PAGE)), (npages << LG_PAGE));\n\t\tflag_unzeroed = unzeroed ? CHUNK_MAP_UNZEROED : 0;\n\t\t/*\n\t\t * Set the unzeroed flag for all pages, now that pages_purge()\n\t\t * has returned whether the pages were zeroed as a side effect\n\t\t * of purging. This chunk map modification is safe even though\n\t\t * the arena mutex isn't currently owned by this thread,\n\t\t * because the run is marked as allocated, thus protecting it\n\t\t * from being modified by any other thread. As long as these\n\t\t * writes don't perturb the first and last elements'\n\t\t * CHUNK_MAP_ALLOCATED bits, behavior is well defined.\n\t\t */\n\t\tfor (i = 0; i < npages; i++) {\n\t\t\tarena_mapbits_unzeroed_set(chunk, pageind+i,\n\t\t\t flag_unzeroed);\n\t\t}\n\t\tnpurged += npages;\n\t\tif (config_stats)\n\t\t\tnmadvise++;\n\t}\n\tmalloc_mutex_lock(&arena->lock);\n\tif (config_stats)\n\t\tarena->stats.nmadvise += nmadvise;\n\n\t/* Deallocate runs. */\n\tfor (mapelm = ql_first(&mapelms); mapelm != NULL;\n\t mapelm = ql_first(&mapelms)) {\n\t\tarena_run_t *run;\n\n\t\tpageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) /\n\t\t sizeof(arena_chunk_map_t)) + map_bias;\n\t\trun = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)(pageind <<\n\t\t LG_PAGE));\n\t\tql_remove(&mapelms, mapelm, u.ql_link);\n\t\tarena_run_dalloc(arena, run, false, true);\n\t}\n\n\treturn (npurged);\n}\n\nstatic arena_chunk_t *\nchunks_dirty_iter_cb(arena_chunk_tree_t *tree, arena_chunk_t *chunk, void *arg)\n{\n size_t *ndirty = (size_t *)arg;\n\n assert(chunk->ndirty != 0);\n *ndirty += chunk->ndirty;\n return (NULL);\n}\n\nstatic void\narena_purge(arena_t *arena, bool all)\n{\n\tarena_chunk_t *chunk;\n\tsize_t npurgatory;\n\tif (config_debug) {\n\t\tsize_t ndirty = 0;\n\n\t\tarena_chunk_dirty_iter(&arena->chunks_dirty, NULL,\n\t\t chunks_dirty_iter_cb, (void *)&ndirty);\n\t\tassert(ndirty == arena->ndirty);\n\t}\n\tassert(arena->ndirty > arena->npurgatory || all);\n\tassert((arena->nactive >> opt_lg_dirty_mult) < (arena->ndirty -\n\t arena->npurgatory) || all);\n\n\tif (config_stats)\n\t\tarena->stats.npurge++;\n\n\t/*\n\t * Compute the minimum number of pages that this thread should try to\n\t * purge, and add the result to arena->npurgatory. This will keep\n\t * multiple threads from racing to reduce ndirty below the threshold.\n\t */\n\t{\n\t\tsize_t npurgeable = arena->ndirty - arena->npurgatory;\n\n\t\tif (all == false) {\n\t\t\tsize_t threshold = (arena->nactive >>\n\t\t\t opt_lg_dirty_mult);\n\n\t\t\tnpurgatory = npurgeable - threshold;\n\t\t} else\n\t\t\tnpurgatory = npurgeable;\n\t}\n\tarena->npurgatory += npurgatory;\n\n\twhile (npurgatory > 0) {\n\t\tsize_t npurgeable, npurged, nunpurged;\n\n\t\t/* Get next chunk with dirty pages. */\n\t\tchunk = arena_chunk_dirty_first(&arena->chunks_dirty);\n\t\tif (chunk == NULL) {\n\t\t\t/*\n\t\t\t * This thread was unable to purge as many pages as\n\t\t\t * originally intended, due to races with other threads\n\t\t\t * that either did some of the purging work, or re-used\n\t\t\t * dirty pages.\n\t\t\t */\n\t\t\tarena->npurgatory -= npurgatory;\n\t\t\treturn;\n\t\t}\n\t\tnpurgeable = chunk->ndirty;\n\t\tassert(npurgeable != 0);\n\n\t\tif (npurgeable > npurgatory && chunk->nruns_adjac == 0) {\n\t\t\t/*\n\t\t\t * This thread will purge all the dirty pages in chunk,\n\t\t\t * so set npurgatory to reflect this thread's intent to\n\t\t\t * purge the pages. This tends to reduce the chances\n\t\t\t * of the following scenario:\n\t\t\t *\n\t\t\t * 1) This thread sets arena->npurgatory such that\n\t\t\t * (arena->ndirty - arena->npurgatory) is at the\n\t\t\t * threshold.\n\t\t\t * 2) This thread drops arena->lock.\n\t\t\t * 3) Another thread causes one or more pages to be\n\t\t\t * dirtied, and immediately determines that it must\n\t\t\t * purge dirty pages.\n\t\t\t *\n\t\t\t * If this scenario *does* play out, that's okay,\n\t\t\t * because all of the purging work being done really\n\t\t\t * needs to happen.\n\t\t\t */\n\t\t\tarena->npurgatory += npurgeable - npurgatory;\n\t\t\tnpurgatory = npurgeable;\n\t\t}\n\n\t\t/*\n\t\t * Keep track of how many pages are purgeable, versus how many\n\t\t * actually get purged, and adjust counters accordingly.\n\t\t */\n\t\tarena->npurgatory -= npurgeable;\n\t\tnpurgatory -= npurgeable;\n\t\tnpurged = arena_chunk_purge(arena, chunk, all);\n\t\tnunpurged = npurgeable - npurged;\n\t\tarena->npurgatory += nunpurged;\n\t\tnpurgatory += nunpurged;\n\t}\n}\n\nvoid\narena_purge_all(arena_t *arena)\n{\n\n\tmalloc_mutex_lock(&arena->lock);\n\tarena_purge(arena, true);\n\tmalloc_mutex_unlock(&arena->lock);\n}\n\nstatic void\narena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty, bool cleaned)\n{\n\tarena_chunk_t *chunk;\n\tsize_t size, run_ind, run_pages, flag_dirty;\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\trun_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE);\n\tassert(run_ind >= map_bias);\n\tassert(run_ind < chunk_npages);\n\tif (arena_mapbits_large_get(chunk, run_ind) != 0) {\n\t\tsize = arena_mapbits_large_size_get(chunk, run_ind);\n\t\tassert(size == PAGE ||\n\t\t arena_mapbits_large_size_get(chunk,\n\t\t run_ind+(size>>LG_PAGE)-1) == 0);\n\t} else {\n\t\tsize_t binind = arena_bin_index(arena, run->bin);\n\t\tarena_bin_info_t *bin_info = &arena_bin_info[binind];\n\t\tsize = bin_info->run_size;\n\t}\n\trun_pages = (size >> LG_PAGE);\n\tif (config_stats) {\n\t\t/*\n\t\t * Update stats_cactive if nactive is crossing a chunk\n\t\t * multiple.\n\t\t */\n\t\tsize_t cactive_diff = CHUNK_CEILING(arena->nactive << LG_PAGE) -\n\t\t CHUNK_CEILING((arena->nactive - run_pages) << LG_PAGE);\n\t\tif (cactive_diff != 0)\n\t\t\tstats_cactive_sub(cactive_diff);\n\t}\n\tarena->nactive -= run_pages;\n\n\t/*\n\t * The run is dirty if the caller claims to have dirtied it, as well as\n\t * if it was already dirty before being allocated and the caller\n\t * doesn't claim to have cleaned it.\n\t */\n\tassert(arena_mapbits_dirty_get(chunk, run_ind) ==\n\t arena_mapbits_dirty_get(chunk, run_ind+run_pages-1));\n\tif (cleaned == false && arena_mapbits_dirty_get(chunk, run_ind) != 0)\n\t\tdirty = true;\n\tflag_dirty = dirty ? CHUNK_MAP_DIRTY : 0;\n\n\t/* Mark pages as unallocated in the chunk map. */\n\tif (dirty) {\n\t\tarena_mapbits_unallocated_set(chunk, run_ind, size,\n\t\t CHUNK_MAP_DIRTY);\n\t\tarena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size,\n\t\t CHUNK_MAP_DIRTY);\n\t} else {\n\t\tarena_mapbits_unallocated_set(chunk, run_ind, size,\n\t\t arena_mapbits_unzeroed_get(chunk, run_ind));\n\t\tarena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size,\n\t\t arena_mapbits_unzeroed_get(chunk, run_ind+run_pages-1));\n\t}\n\n\t/* Try to coalesce forward. */\n\tif (run_ind + run_pages < chunk_npages &&\n\t arena_mapbits_allocated_get(chunk, run_ind+run_pages) == 0 &&\n\t arena_mapbits_dirty_get(chunk, run_ind+run_pages) == flag_dirty) {\n\t\tsize_t nrun_size = arena_mapbits_unallocated_size_get(chunk,\n\t\t run_ind+run_pages);\n\t\tsize_t nrun_pages = nrun_size >> LG_PAGE;\n\n\t\t/*\n\t\t * Remove successor from runs_avail; the coalesced run is\n\t\t * inserted later.\n\t\t */\n\t\tassert(arena_mapbits_unallocated_size_get(chunk,\n\t\t run_ind+run_pages+nrun_pages-1) == nrun_size);\n\t\tassert(arena_mapbits_dirty_get(chunk,\n\t\t run_ind+run_pages+nrun_pages-1) == flag_dirty);\n\t\tarena_avail_remove(arena, chunk, run_ind+run_pages, nrun_pages,\n\t\t false, true);\n\n\t\tsize += nrun_size;\n\t\trun_pages += nrun_pages;\n\n\t\tarena_mapbits_unallocated_size_set(chunk, run_ind, size);\n\t\tarena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1,\n\t\t size);\n\t}\n\n\t/* Try to coalesce backward. */\n\tif (run_ind > map_bias && arena_mapbits_allocated_get(chunk, run_ind-1)\n\t == 0 && arena_mapbits_dirty_get(chunk, run_ind-1) == flag_dirty) {\n\t\tsize_t prun_size = arena_mapbits_unallocated_size_get(chunk,\n\t\t run_ind-1);\n\t\tsize_t prun_pages = prun_size >> LG_PAGE;\n\n\t\trun_ind -= prun_pages;\n\n\t\t/*\n\t\t * Remove predecessor from runs_avail; the coalesced run is\n\t\t * inserted later.\n\t\t */\n\t\tassert(arena_mapbits_unallocated_size_get(chunk, run_ind) ==\n\t\t prun_size);\n\t\tassert(arena_mapbits_dirty_get(chunk, run_ind) == flag_dirty);\n\t\tarena_avail_remove(arena, chunk, run_ind, prun_pages, true,\n\t\t false);\n\n\t\tsize += prun_size;\n\t\trun_pages += prun_pages;\n\n\t\tarena_mapbits_unallocated_size_set(chunk, run_ind, size);\n\t\tarena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1,\n\t\t size);\n\t}\n\n\t/* Insert into runs_avail, now that coalescing is complete. */\n\tassert(arena_mapbits_unallocated_size_get(chunk, run_ind) ==\n\t arena_mapbits_unallocated_size_get(chunk, run_ind+run_pages-1));\n\tassert(arena_mapbits_dirty_get(chunk, run_ind) ==\n\t arena_mapbits_dirty_get(chunk, run_ind+run_pages-1));\n\tarena_avail_insert(arena, chunk, run_ind, run_pages, true, true);\n\n\t/* Deallocate chunk if it is now completely unused. */\n\tif (size == arena_maxclass) {\n\t\tassert(run_ind == map_bias);\n\t\tassert(run_pages == (arena_maxclass >> LG_PAGE));\n\t\tarena_chunk_dealloc(arena, chunk);\n\t}\n\n\t/*\n\t * It is okay to do dirty page processing here even if the chunk was\n\t * deallocated above, since in that case it is the spare. Waiting\n\t * until after possible chunk deallocation to do dirty processing\n\t * allows for an old spare to be fully deallocated, thus decreasing the\n\t * chances of spuriously crossing the dirty page purging threshold.\n\t */\n\tif (dirty)\n\t\tarena_maybe_purge(arena);\n}\n\nstatic void\narena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run,\n size_t oldsize, size_t newsize)\n{\n\tsize_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE;\n\tsize_t head_npages = (oldsize - newsize) >> LG_PAGE;\n\tsize_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind);\n\n\tassert(oldsize > newsize);\n\n\t/*\n\t * Update the chunk map so that arena_run_dalloc() can treat the\n\t * leading run as separately allocated. Set the last element of each\n\t * run first, in case of single-page runs.\n\t */\n\tassert(arena_mapbits_large_size_get(chunk, pageind) == oldsize);\n\tarena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty);\n\tarena_mapbits_large_set(chunk, pageind, oldsize-newsize, flag_dirty);\n\n\tif (config_debug) {\n\t\tUNUSED size_t tail_npages = newsize >> LG_PAGE;\n\t\tassert(arena_mapbits_large_size_get(chunk,\n\t\t pageind+head_npages+tail_npages-1) == 0);\n\t\tassert(arena_mapbits_dirty_get(chunk,\n\t\t pageind+head_npages+tail_npages-1) == flag_dirty);\n\t}\n\tarena_mapbits_large_set(chunk, pageind+head_npages, newsize,\n\t flag_dirty);\n\n\tarena_run_dalloc(arena, run, false, false);\n}\n\nstatic void\narena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run,\n size_t oldsize, size_t newsize, bool dirty)\n{\n\tsize_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE;\n\tsize_t head_npages = newsize >> LG_PAGE;\n\tsize_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind);\n\n\tassert(oldsize > newsize);\n\n\t/*\n\t * Update the chunk map so that arena_run_dalloc() can treat the\n\t * trailing run as separately allocated. Set the last element of each\n\t * run first, in case of single-page runs.\n\t */\n\tassert(arena_mapbits_large_size_get(chunk, pageind) == oldsize);\n\tarena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty);\n\tarena_mapbits_large_set(chunk, pageind, newsize, flag_dirty);\n\n\tif (config_debug) {\n\t\tUNUSED size_t tail_npages = (oldsize - newsize) >> LG_PAGE;\n\t\tassert(arena_mapbits_large_size_get(chunk,\n\t\t pageind+head_npages+tail_npages-1) == 0);\n\t\tassert(arena_mapbits_dirty_get(chunk,\n\t\t pageind+head_npages+tail_npages-1) == flag_dirty);\n\t}\n\tarena_mapbits_large_set(chunk, pageind+head_npages, oldsize-newsize,\n\t flag_dirty);\n\n\tarena_run_dalloc(arena, (arena_run_t *)((uintptr_t)run + newsize),\n\t dirty, false);\n}\n\nstatic arena_run_t *\narena_bin_runs_first(arena_bin_t *bin)\n{\n\tarena_chunk_map_t *mapelm = arena_run_tree_first(&bin->runs);\n\tif (mapelm != NULL) {\n\t\tarena_chunk_t *chunk;\n\t\tsize_t pageind;\n\t\tarena_run_t *run;\n\n\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(mapelm);\n\t\tpageind = ((((uintptr_t)mapelm - (uintptr_t)chunk->map) /\n\t\t sizeof(arena_chunk_map_t))) + map_bias;\n\t\trun = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind -\n\t\t arena_mapbits_small_runind_get(chunk, pageind)) <<\n\t\t LG_PAGE));\n\t\treturn (run);\n\t}\n\n\treturn (NULL);\n}\n\nstatic void\narena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run)\n{\n\tarena_chunk_t *chunk = CHUNK_ADDR2BASE(run);\n\tsize_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE;\n\tarena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind);\n\n\tassert(arena_run_tree_search(&bin->runs, mapelm) == NULL);\n\n\tarena_run_tree_insert(&bin->runs, mapelm);\n}\n\nstatic void\narena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run)\n{\n\tarena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\tsize_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE;\n\tarena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind);\n\n\tassert(arena_run_tree_search(&bin->runs, mapelm) != NULL);\n\n\tarena_run_tree_remove(&bin->runs, mapelm);\n}\n\nstatic arena_run_t *\narena_bin_nonfull_run_tryget(arena_bin_t *bin)\n{\n\tarena_run_t *run = arena_bin_runs_first(bin);\n\tif (run != NULL) {\n\t\tarena_bin_runs_remove(bin, run);\n\t\tif (config_stats)\n\t\t\tbin->stats.reruns++;\n\t}\n\treturn (run);\n}\n\nstatic arena_run_t *\narena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin)\n{\n\tarena_run_t *run;\n\tsize_t binind;\n\tarena_bin_info_t *bin_info;\n\n\t/* Look for a usable run. */\n\trun = arena_bin_nonfull_run_tryget(bin);\n\tif (run != NULL)\n\t\treturn (run);\n\t/* No existing runs have any space available. */\n\n\tbinind = arena_bin_index(arena, bin);\n\tbin_info = &arena_bin_info[binind];\n\n\t/* Allocate a new run. */\n\tmalloc_mutex_unlock(&bin->lock);\n\t/******************************/\n\tmalloc_mutex_lock(&arena->lock);\n\trun = arena_run_alloc(arena, bin_info->run_size, false, binind, false);\n\tif (run != NULL) {\n\t\tbitmap_t *bitmap = (bitmap_t *)((uintptr_t)run +\n\t\t (uintptr_t)bin_info->bitmap_offset);\n\n\t\t/* Initialize run internals. */\n\t\tVALGRIND_MAKE_MEM_UNDEFINED(run, bin_info->reg0_offset -\n\t\t bin_info->redzone_size);\n\t\trun->bin = bin;\n\t\trun->nextind = 0;\n\t\trun->nfree = bin_info->nregs;\n\t\tbitmap_init(bitmap, &bin_info->bitmap_info);\n\t}\n\tmalloc_mutex_unlock(&arena->lock);\n\t/********************************/\n\tmalloc_mutex_lock(&bin->lock);\n\tif (run != NULL) {\n\t\tif (config_stats) {\n\t\t\tbin->stats.nruns++;\n\t\t\tbin->stats.curruns++;\n\t\t}\n\t\treturn (run);\n\t}\n\n\t/*\n\t * arena_run_alloc() failed, but another thread may have made\n\t * sufficient memory available while this one dropped bin->lock above,\n\t * so search one more time.\n\t */\n\trun = arena_bin_nonfull_run_tryget(bin);\n\tif (run != NULL)\n\t\treturn (run);\n\n\treturn (NULL);\n}\n\n/* Re-fill bin->runcur, then call arena_run_reg_alloc(). */\nstatic void *\narena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin)\n{\n\tvoid *ret;\n\tsize_t binind;\n\tarena_bin_info_t *bin_info;\n\tarena_run_t *run;\n\n\tbinind = arena_bin_index(arena, bin);\n\tbin_info = &arena_bin_info[binind];\n\tbin->runcur = NULL;\n\trun = arena_bin_nonfull_run_get(arena, bin);\n\tif (bin->runcur != NULL && bin->runcur->nfree > 0) {\n\t\t/*\n\t\t * Another thread updated runcur while this one ran without the\n\t\t * bin lock in arena_bin_nonfull_run_get().\n\t\t */\n\t\tassert(bin->runcur->nfree > 0);\n\t\tret = arena_run_reg_alloc(bin->runcur, bin_info);\n\t\tif (run != NULL) {\n\t\t\tarena_chunk_t *chunk;\n\n\t\t\t/*\n\t\t\t * arena_run_alloc() may have allocated run, or it may\n\t\t\t * have pulled run from the bin's run tree. Therefore\n\t\t\t * it is unsafe to make any assumptions about how run\n\t\t\t * has previously been used, and arena_bin_lower_run()\n\t\t\t * must be called, as if a region were just deallocated\n\t\t\t * from the run.\n\t\t\t */\n\t\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\t\t\tif (run->nfree == bin_info->nregs)\n\t\t\t\tarena_dalloc_bin_run(arena, chunk, run, bin);\n\t\t\telse\n\t\t\t\tarena_bin_lower_run(arena, chunk, run, bin);\n\t\t}\n\t\treturn (ret);\n\t}\n\n\tif (run == NULL)\n\t\treturn (NULL);\n\n\tbin->runcur = run;\n\n\tassert(bin->runcur->nfree > 0);\n\n\treturn (arena_run_reg_alloc(bin->runcur, bin_info));\n}\n\nvoid\narena_prof_accum(arena_t *arena, uint64_t accumbytes)\n{\n\n\tcassert(config_prof);\n\n\tif (config_prof && prof_interval != 0) {\n\t\tarena->prof_accumbytes += accumbytes;\n\t\tif (arena->prof_accumbytes >= prof_interval) {\n\t\t\tprof_idump();\n\t\t\tarena->prof_accumbytes -= prof_interval;\n\t\t}\n\t}\n}\n\nvoid\narena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind,\n uint64_t prof_accumbytes)\n{\n\tunsigned i, nfill;\n\tarena_bin_t *bin;\n\tarena_run_t *run;\n\tvoid *ptr;\n\n\tassert(tbin->ncached == 0);\n\n\tif (config_prof) {\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tarena_prof_accum(arena, prof_accumbytes);\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t}\n\tbin = &arena->bins[binind];\n\tmalloc_mutex_lock(&bin->lock);\n\tfor (i = 0, nfill = (tcache_bin_info[binind].ncached_max >>\n\t tbin->lg_fill_div); i < nfill; i++) {\n\t\tif ((run = bin->runcur) != NULL && run->nfree > 0)\n\t\t\tptr = arena_run_reg_alloc(run, &arena_bin_info[binind]);\n\t\telse\n\t\t\tptr = arena_bin_malloc_hard(arena, bin);\n\t\tif (ptr == NULL)\n\t\t\tbreak;\n\t\tif (config_fill && opt_junk) {\n\t\t\tarena_alloc_junk_small(ptr, &arena_bin_info[binind],\n\t\t\t true);\n\t\t}\n\t\t/* Insert such that low regions get used first. */\n\t\ttbin->avail[nfill - 1 - i] = ptr;\n\t}\n\tif (config_stats) {\n\t\tbin->stats.allocated += i * arena_bin_info[binind].reg_size;\n\t\tbin->stats.nmalloc += i;\n\t\tbin->stats.nrequests += tbin->tstats.nrequests;\n\t\tbin->stats.nfills++;\n\t\ttbin->tstats.nrequests = 0;\n\t}\n\tmalloc_mutex_unlock(&bin->lock);\n\ttbin->ncached = i;\n}\n\nvoid\narena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info, bool zero)\n{\n\n\tif (zero) {\n\t\tsize_t redzone_size = bin_info->redzone_size;\n\t\tmemset((void *)((uintptr_t)ptr - redzone_size), 0xa5,\n\t\t redzone_size);\n\t\tmemset((void *)((uintptr_t)ptr + bin_info->reg_size), 0xa5,\n\t\t redzone_size);\n\t} else {\n\t\tmemset((void *)((uintptr_t)ptr - bin_info->redzone_size), 0xa5,\n\t\t bin_info->reg_interval);\n\t}\n}\n\nvoid\narena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info)\n{\n\tsize_t size = bin_info->reg_size;\n\tsize_t redzone_size = bin_info->redzone_size;\n\tsize_t i;\n\tbool error = false;\n\n\tfor (i = 1; i <= redzone_size; i++) {\n\t\tunsigned byte;\n\t\tif ((byte = *(uint8_t *)((uintptr_t)ptr - i)) != 0xa5) {\n\t\t\terror = true;\n\t\t\tmalloc_printf(\": Corrupt redzone \"\n\t\t\t \"%zu byte%s before %p (size %zu), byte=%#x\\n\", i,\n\t\t\t (i == 1) ? \"\" : \"s\", ptr, size, byte);\n\t\t}\n\t}\n\tfor (i = 0; i < redzone_size; i++) {\n\t\tunsigned byte;\n\t\tif ((byte = *(uint8_t *)((uintptr_t)ptr + size + i)) != 0xa5) {\n\t\t\terror = true;\n\t\t\tmalloc_printf(\": Corrupt redzone \"\n\t\t\t \"%zu byte%s after end of %p (size %zu), byte=%#x\\n\",\n\t\t\t i, (i == 1) ? \"\" : \"s\", ptr, size, byte);\n\t\t}\n\t}\n\tif (opt_abort && error)\n\t\tabort();\n\n\tmemset((void *)((uintptr_t)ptr - redzone_size), 0x5a,\n\t bin_info->reg_interval);\n}\n\nvoid *\narena_malloc_small(arena_t *arena, size_t size, bool zero)\n{\n\tvoid *ret;\n\tarena_bin_t *bin;\n\tarena_run_t *run;\n\tsize_t binind;\n\n\tbinind = SMALL_SIZE2BIN(size);\n\tassert(binind < NBINS);\n\tbin = &arena->bins[binind];\n\tsize = arena_bin_info[binind].reg_size;\n\n\tmalloc_mutex_lock(&bin->lock);\n\tif ((run = bin->runcur) != NULL && run->nfree > 0)\n\t\tret = arena_run_reg_alloc(run, &arena_bin_info[binind]);\n\telse\n\t\tret = arena_bin_malloc_hard(arena, bin);\n\n\tif (ret == NULL) {\n\t\tmalloc_mutex_unlock(&bin->lock);\n\t\treturn (NULL);\n\t}\n\n\tif (config_stats) {\n\t\tbin->stats.allocated += size;\n\t\tbin->stats.nmalloc++;\n\t\tbin->stats.nrequests++;\n\t}\n\tmalloc_mutex_unlock(&bin->lock);\n\tif (config_prof && isthreaded == false) {\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tarena_prof_accum(arena, size);\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t}\n\n\tif (zero == false) {\n\t\tif (config_fill) {\n\t\t\tif (opt_junk) {\n\t\t\t\tarena_alloc_junk_small(ret,\n\t\t\t\t &arena_bin_info[binind], false);\n\t\t\t} else if (opt_zero)\n\t\t\t\tmemset(ret, 0, size);\n\t\t}\n\t} else {\n\t\tif (config_fill && opt_junk) {\n\t\t\tarena_alloc_junk_small(ret, &arena_bin_info[binind],\n\t\t\t true);\n\t\t}\n\t\tVALGRIND_MAKE_MEM_UNDEFINED(ret, size);\n\t\tmemset(ret, 0, size);\n\t}\n\n\treturn (ret);\n}\n\nvoid *\narena_malloc_large(arena_t *arena, size_t size, bool zero)\n{\n\tvoid *ret;\n\n\t/* Large allocation. */\n\tsize = PAGE_CEILING(size);\n\tmalloc_mutex_lock(&arena->lock);\n\tret = (void *)arena_run_alloc(arena, size, true, BININD_INVALID, zero);\n\tif (ret == NULL) {\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t\treturn (NULL);\n\t}\n\tif (config_stats) {\n\t\tarena->stats.nmalloc_large++;\n\t\tarena->stats.nrequests_large++;\n\t\tarena->stats.allocated_large += size;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].curruns++;\n\t}\n\tif (config_prof)\n\t\tarena_prof_accum(arena, size);\n\tmalloc_mutex_unlock(&arena->lock);\n\n\tif (zero == false) {\n\t\tif (config_fill) {\n\t\t\tif (opt_junk)\n\t\t\t\tmemset(ret, 0xa5, size);\n\t\t\telse if (opt_zero)\n\t\t\t\tmemset(ret, 0, size);\n\t\t}\n\t}\n\n\treturn (ret);\n}\n\n/* Only handles large allocations that require more than page alignment. */\nvoid *\narena_palloc(arena_t *arena, size_t size, size_t alignment, bool zero)\n{\n\tvoid *ret;\n\tsize_t alloc_size, leadsize, trailsize;\n\tarena_run_t *run;\n\tarena_chunk_t *chunk;\n\n\tassert((size & PAGE_MASK) == 0);\n\n\talignment = PAGE_CEILING(alignment);\n\talloc_size = size + alignment - PAGE;\n\n\tmalloc_mutex_lock(&arena->lock);\n\trun = arena_run_alloc(arena, alloc_size, true, BININD_INVALID, zero);\n\tif (run == NULL) {\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t\treturn (NULL);\n\t}\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);\n\n\tleadsize = ALIGNMENT_CEILING((uintptr_t)run, alignment) -\n\t (uintptr_t)run;\n\tassert(alloc_size >= leadsize + size);\n\ttrailsize = alloc_size - leadsize - size;\n\tret = (void *)((uintptr_t)run + leadsize);\n\tif (leadsize != 0) {\n\t\tarena_run_trim_head(arena, chunk, run, alloc_size, alloc_size -\n\t\t leadsize);\n\t}\n\tif (trailsize != 0) {\n\t\tarena_run_trim_tail(arena, chunk, ret, size + trailsize, size,\n\t\t false);\n\t}\n\n\tif (config_stats) {\n\t\tarena->stats.nmalloc_large++;\n\t\tarena->stats.nrequests_large++;\n\t\tarena->stats.allocated_large += size;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].curruns++;\n\t}\n\tmalloc_mutex_unlock(&arena->lock);\n\n\tif (config_fill && zero == false) {\n\t\tif (opt_junk)\n\t\t\tmemset(ret, 0xa5, size);\n\t\telse if (opt_zero)\n\t\t\tmemset(ret, 0, size);\n\t}\n\treturn (ret);\n}\n\nvoid\narena_prof_promoted(const void *ptr, size_t size)\n{\n\tarena_chunk_t *chunk;\n\tsize_t pageind, binind;\n\n\tcassert(config_prof);\n\tassert(ptr != NULL);\n\tassert(CHUNK_ADDR2BASE(ptr) != ptr);\n\tassert(isalloc(ptr, false) == PAGE);\n\tassert(isalloc(ptr, true) == PAGE);\n\tassert(size <= SMALL_MAXCLASS);\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tbinind = SMALL_SIZE2BIN(size);\n\tassert(binind < NBINS);\n\tarena_mapbits_large_binind_set(chunk, pageind, binind);\n\n\tassert(isalloc(ptr, false) == PAGE);\n\tassert(isalloc(ptr, true) == size);\n}\n\nstatic void\narena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run,\n arena_bin_t *bin)\n{\n\n\t/* Dissociate run from bin. */\n\tif (run == bin->runcur)\n\t\tbin->runcur = NULL;\n\telse {\n\t\tsize_t binind = arena_bin_index(chunk->arena, bin);\n\t\tarena_bin_info_t *bin_info = &arena_bin_info[binind];\n\n\t\tif (bin_info->nregs != 1) {\n\t\t\t/*\n\t\t\t * This block's conditional is necessary because if the\n\t\t\t * run only contains one region, then it never gets\n\t\t\t * inserted into the non-full runs tree.\n\t\t\t */\n\t\t\tarena_bin_runs_remove(bin, run);\n\t\t}\n\t}\n}\n\nstatic void\narena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run,\n arena_bin_t *bin)\n{\n\tsize_t binind;\n\tarena_bin_info_t *bin_info;\n\tsize_t npages, run_ind, past;\n\n\tassert(run != bin->runcur);\n\tassert(arena_run_tree_search(&bin->runs,\n\t arena_mapp_get(chunk, ((uintptr_t)run-(uintptr_t)chunk)>>LG_PAGE))\n\t == NULL);\n\n\tbinind = arena_bin_index(chunk->arena, run->bin);\n\tbin_info = &arena_bin_info[binind];\n\n\tmalloc_mutex_unlock(&bin->lock);\n\t/******************************/\n\tnpages = bin_info->run_size >> LG_PAGE;\n\trun_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE);\n\tpast = (size_t)(PAGE_CEILING((uintptr_t)run +\n\t (uintptr_t)bin_info->reg0_offset + (uintptr_t)(run->nextind *\n\t bin_info->reg_interval - bin_info->redzone_size) -\n\t (uintptr_t)chunk) >> LG_PAGE);\n\tmalloc_mutex_lock(&arena->lock);\n\n\t/*\n\t * If the run was originally clean, and some pages were never touched,\n\t * trim the clean pages before deallocating the dirty portion of the\n\t * run.\n\t */\n\tassert(arena_mapbits_dirty_get(chunk, run_ind) ==\n\t arena_mapbits_dirty_get(chunk, run_ind+npages-1));\n\tif (arena_mapbits_dirty_get(chunk, run_ind) == 0 && past - run_ind <\n\t npages) {\n\t\t/* Trim clean pages. Convert to large run beforehand. */\n\t\tassert(npages > 0);\n\t\tarena_mapbits_large_set(chunk, run_ind, bin_info->run_size, 0);\n\t\tarena_mapbits_large_set(chunk, run_ind+npages-1, 0, 0);\n\t\tarena_run_trim_tail(arena, chunk, run, (npages << LG_PAGE),\n\t\t ((past - run_ind) << LG_PAGE), false);\n\t\t/* npages = past - run_ind; */\n\t}\n\tarena_run_dalloc(arena, run, true, false);\n\tmalloc_mutex_unlock(&arena->lock);\n\t/****************************/\n\tmalloc_mutex_lock(&bin->lock);\n\tif (config_stats)\n\t\tbin->stats.curruns--;\n}\n\nstatic void\narena_bin_lower_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run,\n arena_bin_t *bin)\n{\n\n\t/*\n\t * Make sure that if bin->runcur is non-NULL, it refers to the lowest\n\t * non-full run. It is okay to NULL runcur out rather than proactively\n\t * keeping it pointing at the lowest non-full run.\n\t */\n\tif ((uintptr_t)run < (uintptr_t)bin->runcur) {\n\t\t/* Switch runcur. */\n\t\tif (bin->runcur->nfree > 0)\n\t\t\tarena_bin_runs_insert(bin, bin->runcur);\n\t\tbin->runcur = run;\n\t\tif (config_stats)\n\t\t\tbin->stats.reruns++;\n\t} else\n\t\tarena_bin_runs_insert(bin, run);\n}\n\nvoid\narena_dalloc_bin_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n arena_chunk_map_t *mapelm)\n{\n\tsize_t pageind;\n\tarena_run_t *run;\n\tarena_bin_t *bin;\n\tarena_bin_info_t *bin_info;\n\tsize_t size, binind;\n\n\tpageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\trun = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind -\n\t arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE));\n\tbin = run->bin;\n\tbinind = arena_ptr_small_binind_get(ptr, mapelm->bits);\n\tbin_info = &arena_bin_info[binind];\n\tif (config_fill || config_stats)\n\t\tsize = bin_info->reg_size;\n\n\tif (config_fill && opt_junk)\n\t\tarena_dalloc_junk_small(ptr, bin_info);\n\n\tarena_run_reg_dalloc(run, ptr);\n\tif (run->nfree == bin_info->nregs) {\n\t\tarena_dissociate_bin_run(chunk, run, bin);\n\t\tarena_dalloc_bin_run(arena, chunk, run, bin);\n\t} else if (run->nfree == 1 && run != bin->runcur)\n\t\tarena_bin_lower_run(arena, chunk, run, bin);\n\n\tif (config_stats) {\n\t\tbin->stats.allocated -= size;\n\t\tbin->stats.ndalloc++;\n\t}\n}\n\nvoid\narena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t pageind, arena_chunk_map_t *mapelm)\n{\n\tarena_run_t *run;\n\tarena_bin_t *bin;\n\n\trun = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind -\n\t arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE));\n\tbin = run->bin;\n\tmalloc_mutex_lock(&bin->lock);\n\tarena_dalloc_bin_locked(arena, chunk, ptr, mapelm);\n\tmalloc_mutex_unlock(&bin->lock);\n}\n\nvoid\narena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t pageind)\n{\n\tarena_chunk_map_t *mapelm;\n\n\tif (config_debug) {\n\t\t/* arena_ptr_small_binind_get() does extra sanity checking. */\n\t\tassert(arena_ptr_small_binind_get(ptr, arena_mapbits_get(chunk,\n\t\t pageind)) != BININD_INVALID);\n\t}\n\tmapelm = arena_mapp_get(chunk, pageind);\n\tarena_dalloc_bin(arena, chunk, ptr, pageind, mapelm);\n}\n\nvoid\narena_dalloc_large_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr)\n{\n\n\tif (config_fill || config_stats) {\n\t\tsize_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\t\tsize_t size = arena_mapbits_large_size_get(chunk, pageind);\n\n\t\tif (config_fill && config_stats && opt_junk)\n\t\t\tmemset(ptr, 0x5a, size);\n\t\tif (config_stats) {\n\t\t\tarena->stats.ndalloc_large++;\n\t\t\tarena->stats.allocated_large -= size;\n\t\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].ndalloc++;\n\t\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].curruns--;\n\t\t}\n\t}\n\n\tarena_run_dalloc(arena, (arena_run_t *)ptr, true, false);\n}\n\nvoid\narena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr)\n{\n\n\tmalloc_mutex_lock(&arena->lock);\n\tarena_dalloc_large_locked(arena, chunk, ptr);\n\tmalloc_mutex_unlock(&arena->lock);\n}\n\nstatic void\narena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t oldsize, size_t size)\n{\n\n\tassert(size < oldsize);\n\n\t/*\n\t * Shrink the run, and make trailing pages available for other\n\t * allocations.\n\t */\n\tmalloc_mutex_lock(&arena->lock);\n\tarena_run_trim_tail(arena, chunk, (arena_run_t *)ptr, oldsize, size,\n\t true);\n\tif (config_stats) {\n\t\tarena->stats.ndalloc_large++;\n\t\tarena->stats.allocated_large -= oldsize;\n\t\tarena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++;\n\t\tarena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--;\n\n\t\tarena->stats.nmalloc_large++;\n\t\tarena->stats.nrequests_large++;\n\t\tarena->stats.allocated_large += size;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++;\n\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].curruns++;\n\t}\n\tmalloc_mutex_unlock(&arena->lock);\n}\n\nstatic bool\narena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr,\n size_t oldsize, size_t size, size_t extra, bool zero)\n{\n\tsize_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;\n\tsize_t npages = oldsize >> LG_PAGE;\n\tsize_t followsize;\n\n\tassert(oldsize == arena_mapbits_large_size_get(chunk, pageind));\n\n\t/* Try to extend the run. */\n\tassert(size + extra > oldsize);\n\tmalloc_mutex_lock(&arena->lock);\n\tif (pageind + npages < chunk_npages &&\n\t arena_mapbits_allocated_get(chunk, pageind+npages) == 0 &&\n\t (followsize = arena_mapbits_unallocated_size_get(chunk,\n\t pageind+npages)) >= size - oldsize) {\n\t\t/*\n\t\t * The next run is available and sufficiently large. Split the\n\t\t * following run, then merge the first part with the existing\n\t\t * allocation.\n\t\t */\n\t\tsize_t flag_dirty;\n\t\tsize_t splitsize = (oldsize + followsize <= size + extra)\n\t\t ? followsize : size + extra - oldsize;\n\t\tarena_run_split(arena, (arena_run_t *)((uintptr_t)chunk +\n\t\t ((pageind+npages) << LG_PAGE)), splitsize, true,\n\t\t BININD_INVALID, zero);\n\n\t\tsize = oldsize + splitsize;\n\t\tnpages = size >> LG_PAGE;\n\n\t\t/*\n\t\t * Mark the extended run as dirty if either portion of the run\n\t\t * was dirty before allocation. This is rather pedantic,\n\t\t * because there's not actually any sequence of events that\n\t\t * could cause the resulting run to be passed to\n\t\t * arena_run_dalloc() with the dirty argument set to false\n\t\t * (which is when dirty flag consistency would really matter).\n\t\t */\n\t\tflag_dirty = arena_mapbits_dirty_get(chunk, pageind) |\n\t\t arena_mapbits_dirty_get(chunk, pageind+npages-1);\n\t\tarena_mapbits_large_set(chunk, pageind, size, flag_dirty);\n\t\tarena_mapbits_large_set(chunk, pageind+npages-1, 0, flag_dirty);\n\n\t\tif (config_stats) {\n\t\t\tarena->stats.ndalloc_large++;\n\t\t\tarena->stats.allocated_large -= oldsize;\n\t\t\tarena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++;\n\t\t\tarena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--;\n\n\t\t\tarena->stats.nmalloc_large++;\n\t\t\tarena->stats.nrequests_large++;\n\t\t\tarena->stats.allocated_large += size;\n\t\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++;\n\t\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++;\n\t\t\tarena->stats.lstats[(size >> LG_PAGE) - 1].curruns++;\n\t\t}\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t\treturn (false);\n\t}\n\tmalloc_mutex_unlock(&arena->lock);\n\n\treturn (true);\n}\n\n/*\n * Try to resize a large allocation, in order to avoid copying. This will\n * always fail if growing an object, and the following run is already in use.\n */\nstatic bool\narena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra,\n bool zero)\n{\n\tsize_t psize;\n\n\tpsize = PAGE_CEILING(size + extra);\n\tif (psize == oldsize) {\n\t\t/* Same size class. */\n\t\tif (config_fill && opt_junk && size < oldsize) {\n\t\t\tmemset((void *)((uintptr_t)ptr + size), 0x5a, oldsize -\n\t\t\t size);\n\t\t}\n\t\treturn (false);\n\t} else {\n\t\tarena_chunk_t *chunk;\n\t\tarena_t *arena;\n\n\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\t\tarena = chunk->arena;\n\n\t\tif (psize < oldsize) {\n\t\t\t/* Fill before shrinking in order avoid a race. */\n\t\t\tif (config_fill && opt_junk) {\n\t\t\t\tmemset((void *)((uintptr_t)ptr + size), 0x5a,\n\t\t\t\t oldsize - size);\n\t\t\t}\n\t\t\tarena_ralloc_large_shrink(arena, chunk, ptr, oldsize,\n\t\t\t psize);\n\t\t\treturn (false);\n\t\t} else {\n\t\t\tbool ret = arena_ralloc_large_grow(arena, chunk, ptr,\n\t\t\t oldsize, PAGE_CEILING(size),\n\t\t\t psize - PAGE_CEILING(size), zero);\n\t\t\tif (config_fill && ret == false && zero == false &&\n\t\t\t opt_zero) {\n\t\t\t\tmemset((void *)((uintptr_t)ptr + oldsize), 0,\n\t\t\t\t size - oldsize);\n\t\t\t}\n\t\t\treturn (ret);\n\t\t}\n\t}\n}\n\nvoid *\narena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra,\n bool zero)\n{\n\n\t/*\n\t * Avoid moving the allocation if the size class can be left the same.\n\t */\n\tif (oldsize <= arena_maxclass) {\n\t\tif (oldsize <= SMALL_MAXCLASS) {\n\t\t\tassert(arena_bin_info[SMALL_SIZE2BIN(oldsize)].reg_size\n\t\t\t == oldsize);\n\t\t\tif ((size + extra <= SMALL_MAXCLASS &&\n\t\t\t SMALL_SIZE2BIN(size + extra) ==\n\t\t\t SMALL_SIZE2BIN(oldsize)) || (size <= oldsize &&\n\t\t\t size + extra >= oldsize)) {\n\t\t\t\tif (config_fill && opt_junk && size < oldsize) {\n\t\t\t\t\tmemset((void *)((uintptr_t)ptr + size),\n\t\t\t\t\t 0x5a, oldsize - size);\n\t\t\t\t}\n\t\t\t\treturn (ptr);\n\t\t\t}\n\t\t} else {\n\t\t\tassert(size <= arena_maxclass);\n\t\t\tif (size + extra > SMALL_MAXCLASS) {\n\t\t\t\tif (arena_ralloc_large(ptr, oldsize, size,\n\t\t\t\t extra, zero) == false)\n\t\t\t\t\treturn (ptr);\n\t\t\t}\n\t\t}\n\t}\n\n\t/* Reallocation would require a move. */\n\treturn (NULL);\n}\n\nvoid *\narena_ralloc(arena_t *arena, void *ptr, size_t oldsize, size_t size,\n size_t extra, size_t alignment, bool zero, bool try_tcache_alloc,\n bool try_tcache_dalloc)\n{\n\tvoid *ret;\n\tsize_t copysize;\n\n\t/* Try to avoid moving the allocation. */\n\tret = arena_ralloc_no_move(ptr, oldsize, size, extra, zero);\n\tif (ret != NULL)\n\t\treturn (ret);\n\n\t/*\n\t * size and oldsize are different enough that we need to move the\n\t * object. In that case, fall back to allocating new space and\n\t * copying.\n\t */\n\tif (alignment != 0) {\n\t\tsize_t usize = sa2u(size + extra, alignment);\n\t\tif (usize == 0)\n\t\t\treturn (NULL);\n\t\tret = ipallocx(usize, alignment, zero, try_tcache_alloc, arena);\n\t} else\n\t\tret = arena_malloc(arena, size + extra, zero, try_tcache_alloc);\n\n\tif (ret == NULL) {\n\t\tif (extra == 0)\n\t\t\treturn (NULL);\n\t\t/* Try again, this time without extra. */\n\t\tif (alignment != 0) {\n\t\t\tsize_t usize = sa2u(size, alignment);\n\t\t\tif (usize == 0)\n\t\t\t\treturn (NULL);\n\t\t\tret = ipallocx(usize, alignment, zero, try_tcache_alloc,\n\t\t\t arena);\n\t\t} else\n\t\t\tret = arena_malloc(arena, size, zero, try_tcache_alloc);\n\n\t\tif (ret == NULL)\n\t\t\treturn (NULL);\n\t}\n\n\t/* Junk/zero-filling were already done by ipalloc()/arena_malloc(). */\n\n\t/*\n\t * Copy at most size bytes (not size+extra), since the caller has no\n\t * expectation that the extra bytes will be reliably preserved.\n\t */\n\tcopysize = (size < oldsize) ? size : oldsize;\n\tVALGRIND_MAKE_MEM_UNDEFINED(ret, copysize);\n\tmemcpy(ret, ptr, copysize);\n\tiqallocx(ptr, try_tcache_dalloc);\n\treturn (ret);\n}\n\ndss_prec_t\narena_dss_prec_get(arena_t *arena)\n{\n\tdss_prec_t ret;\n\n\tmalloc_mutex_lock(&arena->lock);\n\tret = arena->dss_prec;\n\tmalloc_mutex_unlock(&arena->lock);\n\treturn (ret);\n}\n\nvoid\narena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec)\n{\n\n\tmalloc_mutex_lock(&arena->lock);\n\tarena->dss_prec = dss_prec;\n\tmalloc_mutex_unlock(&arena->lock);\n}\n\nvoid\narena_stats_merge(arena_t *arena, const char **dss, size_t *nactive,\n size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats,\n malloc_large_stats_t *lstats)\n{\n\tunsigned i;\n\n\tmalloc_mutex_lock(&arena->lock);\n\t*dss = dss_prec_names[arena->dss_prec];\n\t*nactive += arena->nactive;\n\t*ndirty += arena->ndirty;\n\n\tastats->mapped += arena->stats.mapped;\n\tastats->npurge += arena->stats.npurge;\n\tastats->nmadvise += arena->stats.nmadvise;\n\tastats->purged += arena->stats.purged;\n\tastats->allocated_large += arena->stats.allocated_large;\n\tastats->nmalloc_large += arena->stats.nmalloc_large;\n\tastats->ndalloc_large += arena->stats.ndalloc_large;\n\tastats->nrequests_large += arena->stats.nrequests_large;\n\n\tfor (i = 0; i < nlclasses; i++) {\n\t\tlstats[i].nmalloc += arena->stats.lstats[i].nmalloc;\n\t\tlstats[i].ndalloc += arena->stats.lstats[i].ndalloc;\n\t\tlstats[i].nrequests += arena->stats.lstats[i].nrequests;\n\t\tlstats[i].curruns += arena->stats.lstats[i].curruns;\n\t}\n\tmalloc_mutex_unlock(&arena->lock);\n\n\tfor (i = 0; i < NBINS; i++) {\n\t\tarena_bin_t *bin = &arena->bins[i];\n\n\t\tmalloc_mutex_lock(&bin->lock);\n\t\tbstats[i].allocated += bin->stats.allocated;\n\t\tbstats[i].nmalloc += bin->stats.nmalloc;\n\t\tbstats[i].ndalloc += bin->stats.ndalloc;\n\t\tbstats[i].nrequests += bin->stats.nrequests;\n\t\tif (config_tcache) {\n\t\t\tbstats[i].nfills += bin->stats.nfills;\n\t\t\tbstats[i].nflushes += bin->stats.nflushes;\n\t\t}\n\t\tbstats[i].nruns += bin->stats.nruns;\n\t\tbstats[i].reruns += bin->stats.reruns;\n\t\tbstats[i].curruns += bin->stats.curruns;\n\t\tmalloc_mutex_unlock(&bin->lock);\n\t}\n}\n\nbool\narena_new(arena_t *arena, unsigned ind)\n{\n\tunsigned i;\n\tarena_bin_t *bin;\n\n\tarena->ind = ind;\n\tarena->nthreads = 0;\n\n\tif (malloc_mutex_init(&arena->lock))\n\t\treturn (true);\n\n\tif (config_stats) {\n\t\tmemset(&arena->stats, 0, sizeof(arena_stats_t));\n\t\tarena->stats.lstats =\n\t\t (malloc_large_stats_t *)base_alloc(nlclasses *\n\t\t sizeof(malloc_large_stats_t));\n\t\tif (arena->stats.lstats == NULL)\n\t\t\treturn (true);\n\t\tmemset(arena->stats.lstats, 0, nlclasses *\n\t\t sizeof(malloc_large_stats_t));\n\t\tif (config_tcache)\n\t\t\tql_new(&arena->tcache_ql);\n\t}\n\n\tif (config_prof)\n\t\tarena->prof_accumbytes = 0;\n\n\tarena->dss_prec = chunk_dss_prec_get();\n\n\t/* Initialize chunks. */\n\tarena_chunk_dirty_new(&arena->chunks_dirty);\n\tarena->spare = NULL;\n\n\tarena->nactive = 0;\n\tarena->ndirty = 0;\n\tarena->npurgatory = 0;\n\n\tarena_avail_tree_new(&arena->runs_avail);\n\n\t/* Initialize bins. */\n\tfor (i = 0; i < NBINS; i++) {\n\t\tbin = &arena->bins[i];\n\t\tif (malloc_mutex_init(&bin->lock))\n\t\t\treturn (true);\n\t\tbin->runcur = NULL;\n\t\tarena_run_tree_new(&bin->runs);\n\t\tif (config_stats)\n\t\t\tmemset(&bin->stats, 0, sizeof(malloc_bin_stats_t));\n\t}\n\n\treturn (false);\n}\n\n/*\n * Calculate bin_info->run_size such that it meets the following constraints:\n *\n * *) bin_info->run_size >= min_run_size\n * *) bin_info->run_size <= arena_maxclass\n * *) run header overhead <= RUN_MAX_OVRHD (or header overhead relaxed).\n * *) bin_info->nregs <= RUN_MAXREGS\n *\n * bin_info->nregs, bin_info->bitmap_offset, and bin_info->reg0_offset are also\n * calculated here, since these settings are all interdependent.\n */\nstatic size_t\nbin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size)\n{\n\tsize_t pad_size;\n\tsize_t try_run_size, good_run_size;\n\tuint32_t try_nregs, good_nregs;\n\tuint32_t try_hdr_size, good_hdr_size;\n\tuint32_t try_bitmap_offset, good_bitmap_offset;\n\tuint32_t try_ctx0_offset, good_ctx0_offset;\n\tuint32_t try_redzone0_offset, good_redzone0_offset;\n\n\tassert(min_run_size >= PAGE);\n\tassert(min_run_size <= arena_maxclass);\n\n\t/*\n\t * Determine redzone size based on minimum alignment and minimum\n\t * redzone size. Add padding to the end of the run if it is needed to\n\t * align the regions. The padding allows each redzone to be half the\n\t * minimum alignment; without the padding, each redzone would have to\n\t * be twice as large in order to maintain alignment.\n\t */\n\tif (config_fill && opt_redzone) {\n\t\tsize_t align_min = ZU(1) << (ffs(bin_info->reg_size) - 1);\n\t\tif (align_min <= REDZONE_MINSIZE) {\n\t\t\tbin_info->redzone_size = REDZONE_MINSIZE;\n\t\t\tpad_size = 0;\n\t\t} else {\n\t\t\tbin_info->redzone_size = align_min >> 1;\n\t\t\tpad_size = bin_info->redzone_size;\n\t\t}\n\t} else {\n\t\tbin_info->redzone_size = 0;\n\t\tpad_size = 0;\n\t}\n\tbin_info->reg_interval = bin_info->reg_size +\n\t (bin_info->redzone_size << 1);\n\n\t/*\n\t * Calculate known-valid settings before entering the run_size\n\t * expansion loop, so that the first part of the loop always copies\n\t * valid settings.\n\t *\n\t * The do..while loop iteratively reduces the number of regions until\n\t * the run header and the regions no longer overlap. A closed formula\n\t * would be quite messy, since there is an interdependency between the\n\t * header's mask length and the number of regions.\n\t */\n\ttry_run_size = min_run_size;\n\ttry_nregs = ((try_run_size - sizeof(arena_run_t)) /\n\t bin_info->reg_interval)\n\t + 1; /* Counter-act try_nregs-- in loop. */\n\tif (try_nregs > RUN_MAXREGS) {\n\t\ttry_nregs = RUN_MAXREGS\n\t\t + 1; /* Counter-act try_nregs-- in loop. */\n\t}\n\tdo {\n\t\ttry_nregs--;\n\t\ttry_hdr_size = sizeof(arena_run_t);\n\t\t/* Pad to a long boundary. */\n\t\ttry_hdr_size = LONG_CEILING(try_hdr_size);\n\t\ttry_bitmap_offset = try_hdr_size;\n\t\t/* Add space for bitmap. */\n\t\ttry_hdr_size += bitmap_size(try_nregs);\n\t\tif (config_prof && opt_prof && prof_promote == false) {\n\t\t\t/* Pad to a quantum boundary. */\n\t\t\ttry_hdr_size = QUANTUM_CEILING(try_hdr_size);\n\t\t\ttry_ctx0_offset = try_hdr_size;\n\t\t\t/* Add space for one (prof_ctx_t *) per region. */\n\t\t\ttry_hdr_size += try_nregs * sizeof(prof_ctx_t *);\n\t\t} else\n\t\t\ttry_ctx0_offset = 0;\n\t\ttry_redzone0_offset = try_run_size - (try_nregs *\n\t\t bin_info->reg_interval) - pad_size;\n\t} while (try_hdr_size > try_redzone0_offset);\n\n\t/* run_size expansion loop. */\n\tdo {\n\t\t/*\n\t\t * Copy valid settings before trying more aggressive settings.\n\t\t */\n\t\tgood_run_size = try_run_size;\n\t\tgood_nregs = try_nregs;\n\t\tgood_hdr_size = try_hdr_size;\n\t\tgood_bitmap_offset = try_bitmap_offset;\n\t\tgood_ctx0_offset = try_ctx0_offset;\n\t\tgood_redzone0_offset = try_redzone0_offset;\n\n\t\t/* Try more aggressive settings. */\n\t\ttry_run_size += PAGE;\n\t\ttry_nregs = ((try_run_size - sizeof(arena_run_t) - pad_size) /\n\t\t bin_info->reg_interval)\n\t\t + 1; /* Counter-act try_nregs-- in loop. */\n\t\tif (try_nregs > RUN_MAXREGS) {\n\t\t\ttry_nregs = RUN_MAXREGS\n\t\t\t + 1; /* Counter-act try_nregs-- in loop. */\n\t\t}\n\t\tdo {\n\t\t\ttry_nregs--;\n\t\t\ttry_hdr_size = sizeof(arena_run_t);\n\t\t\t/* Pad to a long boundary. */\n\t\t\ttry_hdr_size = LONG_CEILING(try_hdr_size);\n\t\t\ttry_bitmap_offset = try_hdr_size;\n\t\t\t/* Add space for bitmap. */\n\t\t\ttry_hdr_size += bitmap_size(try_nregs);\n\t\t\tif (config_prof && opt_prof && prof_promote == false) {\n\t\t\t\t/* Pad to a quantum boundary. */\n\t\t\t\ttry_hdr_size = QUANTUM_CEILING(try_hdr_size);\n\t\t\t\ttry_ctx0_offset = try_hdr_size;\n\t\t\t\t/*\n\t\t\t\t * Add space for one (prof_ctx_t *) per region.\n\t\t\t\t */\n\t\t\t\ttry_hdr_size += try_nregs *\n\t\t\t\t sizeof(prof_ctx_t *);\n\t\t\t}\n\t\t\ttry_redzone0_offset = try_run_size - (try_nregs *\n\t\t\t bin_info->reg_interval) - pad_size;\n\t\t} while (try_hdr_size > try_redzone0_offset);\n\t} while (try_run_size <= arena_maxclass\n\t && try_run_size <= arena_maxclass\n\t && RUN_MAX_OVRHD * (bin_info->reg_interval << 3) >\n\t RUN_MAX_OVRHD_RELAX\n\t && (try_redzone0_offset << RUN_BFP) > RUN_MAX_OVRHD * try_run_size\n\t && try_nregs < RUN_MAXREGS);\n\n\tassert(good_hdr_size <= good_redzone0_offset);\n\n\t/* Copy final settings. */\n\tbin_info->run_size = good_run_size;\n\tbin_info->nregs = good_nregs;\n\tbin_info->bitmap_offset = good_bitmap_offset;\n\tbin_info->ctx0_offset = good_ctx0_offset;\n\tbin_info->reg0_offset = good_redzone0_offset + bin_info->redzone_size;\n\n\tassert(bin_info->reg0_offset - bin_info->redzone_size + (bin_info->nregs\n\t * bin_info->reg_interval) + pad_size == bin_info->run_size);\n\n\treturn (good_run_size);\n}\n\nstatic void\nbin_info_init(void)\n{\n\tarena_bin_info_t *bin_info;\n\tsize_t prev_run_size = PAGE;\n\n#define\tSIZE_CLASS(bin, delta, size)\t\t\t\t\t\\\n\tbin_info = &arena_bin_info[bin];\t\t\t\t\\\n\tbin_info->reg_size = size;\t\t\t\t\t\\\n\tprev_run_size = bin_info_run_size_calc(bin_info, prev_run_size);\\\n\tbitmap_info_init(&bin_info->bitmap_info, bin_info->nregs);\n\tSIZE_CLASSES\n#undef SIZE_CLASS\n}\n\nvoid\narena_boot(void)\n{\n\tsize_t header_size;\n\tunsigned i;\n\n\t/*\n\t * Compute the header size such that it is large enough to contain the\n\t * page map. The page map is biased to omit entries for the header\n\t * itself, so some iteration is necessary to compute the map bias.\n\t *\n\t * 1) Compute safe header_size and map_bias values that include enough\n\t * space for an unbiased page map.\n\t * 2) Refine map_bias based on (1) to omit the header pages in the page\n\t * map. The resulting map_bias may be one too small.\n\t * 3) Refine map_bias based on (2). The result will be >= the result\n\t * from (2), and will always be correct.\n\t */\n\tmap_bias = 0;\n\tfor (i = 0; i < 3; i++) {\n\t\theader_size = offsetof(arena_chunk_t, map) +\n\t\t (sizeof(arena_chunk_map_t) * (chunk_npages-map_bias));\n\t\tmap_bias = (header_size >> LG_PAGE) + ((header_size & PAGE_MASK)\n\t\t != 0);\n\t}\n\tassert(map_bias > 0);\n\n\tarena_maxclass = chunksize - (map_bias << LG_PAGE);\n\n\tbin_info_init();\n}\n\nvoid\narena_prefork(arena_t *arena)\n{\n\tunsigned i;\n\n\tmalloc_mutex_prefork(&arena->lock);\n\tfor (i = 0; i < NBINS; i++)\n\t\tmalloc_mutex_prefork(&arena->bins[i].lock);\n}\n\nvoid\narena_postfork_parent(arena_t *arena)\n{\n\tunsigned i;\n\n\tfor (i = 0; i < NBINS; i++)\n\t\tmalloc_mutex_postfork_parent(&arena->bins[i].lock);\n\tmalloc_mutex_postfork_parent(&arena->lock);\n}\n\nvoid\narena_postfork_child(arena_t *arena)\n{\n\tunsigned i;\n\n\tfor (i = 0; i < NBINS; i++)\n\t\tmalloc_mutex_postfork_child(&arena->bins[i].lock);\n\tmalloc_mutex_postfork_child(&arena->lock);\n}\n" }, { "path": "deps/jemalloc/src/atomic.c", "content": "#define\tJEMALLOC_ATOMIC_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n" }, { "path": "deps/jemalloc/src/base.c", "content": "#define\tJEMALLOC_BASE_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nstatic malloc_mutex_t\tbase_mtx;\n\n/*\n * Current pages that are being used for internal memory allocations. These\n * pages are carved up in cacheline-size quanta, so that there is no chance of\n * false cache line sharing.\n */\nstatic void\t\t*base_pages;\nstatic void\t\t*base_next_addr;\nstatic void\t\t*base_past_addr; /* Addr immediately past base_pages. */\nstatic extent_node_t\t*base_nodes;\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic bool\tbase_pages_alloc(size_t minsize);\n\n/******************************************************************************/\n\nstatic bool\nbase_pages_alloc(size_t minsize)\n{\n\tsize_t csize;\n\tbool zero;\n\n\tassert(minsize != 0);\n\tcsize = CHUNK_CEILING(minsize);\n\tzero = false;\n\tbase_pages = chunk_alloc(csize, chunksize, true, &zero,\n\t chunk_dss_prec_get());\n\tif (base_pages == NULL)\n\t\treturn (true);\n\tbase_next_addr = base_pages;\n\tbase_past_addr = (void *)((uintptr_t)base_pages + csize);\n\n\treturn (false);\n}\n\nvoid *\nbase_alloc(size_t size)\n{\n\tvoid *ret;\n\tsize_t csize;\n\n\t/* Round size up to nearest multiple of the cacheline size. */\n\tcsize = CACHELINE_CEILING(size);\n\n\tmalloc_mutex_lock(&base_mtx);\n\t/* Make sure there's enough space for the allocation. */\n\tif ((uintptr_t)base_next_addr + csize > (uintptr_t)base_past_addr) {\n\t\tif (base_pages_alloc(csize)) {\n\t\t\tmalloc_mutex_unlock(&base_mtx);\n\t\t\treturn (NULL);\n\t\t}\n\t}\n\t/* Allocate. */\n\tret = base_next_addr;\n\tbase_next_addr = (void *)((uintptr_t)base_next_addr + csize);\n\tmalloc_mutex_unlock(&base_mtx);\n\n\treturn (ret);\n}\n\nvoid *\nbase_calloc(size_t number, size_t size)\n{\n\tvoid *ret = base_alloc(number * size);\n\n\tif (ret != NULL)\n\t\tmemset(ret, 0, number * size);\n\n\treturn (ret);\n}\n\nextent_node_t *\nbase_node_alloc(void)\n{\n\textent_node_t *ret;\n\n\tmalloc_mutex_lock(&base_mtx);\n\tif (base_nodes != NULL) {\n\t\tret = base_nodes;\n\t\tbase_nodes = *(extent_node_t **)ret;\n\t\tmalloc_mutex_unlock(&base_mtx);\n\t} else {\n\t\tmalloc_mutex_unlock(&base_mtx);\n\t\tret = (extent_node_t *)base_alloc(sizeof(extent_node_t));\n\t}\n\n\treturn (ret);\n}\n\nvoid\nbase_node_dealloc(extent_node_t *node)\n{\n\n\tmalloc_mutex_lock(&base_mtx);\n\t*(extent_node_t **)node = base_nodes;\n\tbase_nodes = node;\n\tmalloc_mutex_unlock(&base_mtx);\n}\n\nbool\nbase_boot(void)\n{\n\n\tbase_nodes = NULL;\n\tif (malloc_mutex_init(&base_mtx))\n\t\treturn (true);\n\n\treturn (false);\n}\n\nvoid\nbase_prefork(void)\n{\n\n\tmalloc_mutex_prefork(&base_mtx);\n}\n\nvoid\nbase_postfork_parent(void)\n{\n\n\tmalloc_mutex_postfork_parent(&base_mtx);\n}\n\nvoid\nbase_postfork_child(void)\n{\n\n\tmalloc_mutex_postfork_child(&base_mtx);\n}\n" }, { "path": "deps/jemalloc/src/bitmap.c", "content": "#define JEMALLOC_BITMAP_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic size_t\tbits2groups(size_t nbits);\n\n/******************************************************************************/\n\nstatic size_t\nbits2groups(size_t nbits)\n{\n\n\treturn ((nbits >> LG_BITMAP_GROUP_NBITS) +\n\t !!(nbits & BITMAP_GROUP_NBITS_MASK));\n}\n\nvoid\nbitmap_info_init(bitmap_info_t *binfo, size_t nbits)\n{\n\tunsigned i;\n\tsize_t group_count;\n\n\tassert(nbits > 0);\n\tassert(nbits <= (ZU(1) << LG_BITMAP_MAXBITS));\n\n\t/*\n\t * Compute the number of groups necessary to store nbits bits, and\n\t * progressively work upward through the levels until reaching a level\n\t * that requires only one group.\n\t */\n\tbinfo->levels[0].group_offset = 0;\n\tgroup_count = bits2groups(nbits);\n\tfor (i = 1; group_count > 1; i++) {\n\t\tassert(i < BITMAP_MAX_LEVELS);\n\t\tbinfo->levels[i].group_offset = binfo->levels[i-1].group_offset\n\t\t + group_count;\n\t\tgroup_count = bits2groups(group_count);\n\t}\n\tbinfo->levels[i].group_offset = binfo->levels[i-1].group_offset\n\t + group_count;\n\tbinfo->nlevels = i;\n\tbinfo->nbits = nbits;\n}\n\nsize_t\nbitmap_info_ngroups(const bitmap_info_t *binfo)\n{\n\n\treturn (binfo->levels[binfo->nlevels].group_offset << LG_SIZEOF_BITMAP);\n}\n\nsize_t\nbitmap_size(size_t nbits)\n{\n\tbitmap_info_t binfo;\n\n\tbitmap_info_init(&binfo, nbits);\n\treturn (bitmap_info_ngroups(&binfo));\n}\n\nvoid\nbitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)\n{\n\tsize_t extra;\n\tunsigned i;\n\n\t/*\n\t * Bits are actually inverted with regard to the external bitmap\n\t * interface, so the bitmap starts out with all 1 bits, except for\n\t * trailing unused bits (if any). Note that each group uses bit 0 to\n\t * correspond to the first logical bit in the group, so extra bits\n\t * are the most significant bits of the last group.\n\t */\n\tmemset(bitmap, 0xffU, binfo->levels[binfo->nlevels].group_offset <<\n\t LG_SIZEOF_BITMAP);\n\textra = (BITMAP_GROUP_NBITS - (binfo->nbits & BITMAP_GROUP_NBITS_MASK))\n\t & BITMAP_GROUP_NBITS_MASK;\n\tif (extra != 0)\n\t\tbitmap[binfo->levels[1].group_offset - 1] >>= extra;\n\tfor (i = 1; i < binfo->nlevels; i++) {\n\t\tsize_t group_count = binfo->levels[i].group_offset -\n\t\t binfo->levels[i-1].group_offset;\n\t\textra = (BITMAP_GROUP_NBITS - (group_count &\n\t\t BITMAP_GROUP_NBITS_MASK)) & BITMAP_GROUP_NBITS_MASK;\n\t\tif (extra != 0)\n\t\t\tbitmap[binfo->levels[i+1].group_offset - 1] >>= extra;\n\t}\n}\n" }, { "path": "deps/jemalloc/src/chunk.c", "content": "#define\tJEMALLOC_CHUNK_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nconst char\t*opt_dss = DSS_DEFAULT;\nsize_t\t\topt_lg_chunk = LG_CHUNK_DEFAULT;\n\nmalloc_mutex_t\tchunks_mtx;\nchunk_stats_t\tstats_chunks;\n\n/*\n * Trees of chunks that were previously allocated (trees differ only in node\n * ordering). These are used when allocating chunks, in an attempt to re-use\n * address space. Depending on function, different tree orderings are needed,\n * which is why there are two trees with the same contents.\n */\nstatic extent_tree_t\tchunks_szad_mmap;\nstatic extent_tree_t\tchunks_ad_mmap;\nstatic extent_tree_t\tchunks_szad_dss;\nstatic extent_tree_t\tchunks_ad_dss;\n\nrtree_t\t\t*chunks_rtree;\n\n/* Various chunk-related settings. */\nsize_t\t\tchunksize;\nsize_t\t\tchunksize_mask; /* (chunksize - 1). */\nsize_t\t\tchunk_npages;\nsize_t\t\tmap_bias;\nsize_t\t\tarena_maxclass; /* Max size class for arenas. */\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\t*chunk_recycle(extent_tree_t *chunks_szad,\n extent_tree_t *chunks_ad, size_t size, size_t alignment, bool base,\n bool *zero);\nstatic void\tchunk_record(extent_tree_t *chunks_szad,\n extent_tree_t *chunks_ad, void *chunk, size_t size);\n\n/******************************************************************************/\n\nstatic void *\nchunk_recycle(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, size_t size,\n size_t alignment, bool base, bool *zero)\n{\n\tvoid *ret;\n\textent_node_t *node;\n\textent_node_t key;\n\tsize_t alloc_size, leadsize, trailsize;\n\tbool zeroed;\n\n\tif (base) {\n\t\t/*\n\t\t * This function may need to call base_node_{,de}alloc(), but\n\t\t * the current chunk allocation request is on behalf of the\n\t\t * base allocator. Avoid deadlock (and if that weren't an\n\t\t * issue, potential for infinite recursion) by returning NULL.\n\t\t */\n\t\treturn (NULL);\n\t}\n\n\talloc_size = size + alignment - chunksize;\n\t/* Beware size_t wrap-around. */\n\tif (alloc_size < size)\n\t\treturn (NULL);\n\tkey.addr = NULL;\n\tkey.size = alloc_size;\n\tmalloc_mutex_lock(&chunks_mtx);\n\tnode = extent_tree_szad_nsearch(chunks_szad, &key);\n\tif (node == NULL) {\n\t\tmalloc_mutex_unlock(&chunks_mtx);\n\t\treturn (NULL);\n\t}\n\tleadsize = ALIGNMENT_CEILING((uintptr_t)node->addr, alignment) -\n\t (uintptr_t)node->addr;\n\tassert(node->size >= leadsize + size);\n\ttrailsize = node->size - leadsize - size;\n\tret = (void *)((uintptr_t)node->addr + leadsize);\n\t/* Remove node from the tree. */\n\textent_tree_szad_remove(chunks_szad, node);\n\textent_tree_ad_remove(chunks_ad, node);\n\tif (leadsize != 0) {\n\t\t/* Insert the leading space as a smaller chunk. */\n\t\tnode->size = leadsize;\n\t\textent_tree_szad_insert(chunks_szad, node);\n\t\textent_tree_ad_insert(chunks_ad, node);\n\t\tnode = NULL;\n\t}\n\tif (trailsize != 0) {\n\t\t/* Insert the trailing space as a smaller chunk. */\n\t\tif (node == NULL) {\n\t\t\t/*\n\t\t\t * An additional node is required, but\n\t\t\t * base_node_alloc() can cause a new base chunk to be\n\t\t\t * allocated. Drop chunks_mtx in order to avoid\n\t\t\t * deadlock, and if node allocation fails, deallocate\n\t\t\t * the result before returning an error.\n\t\t\t */\n\t\t\tmalloc_mutex_unlock(&chunks_mtx);\n\t\t\tnode = base_node_alloc();\n\t\t\tif (node == NULL) {\n\t\t\t\tchunk_dealloc(ret, size, true);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tmalloc_mutex_lock(&chunks_mtx);\n\t\t}\n\t\tnode->addr = (void *)((uintptr_t)(ret) + size);\n\t\tnode->size = trailsize;\n\t\textent_tree_szad_insert(chunks_szad, node);\n\t\textent_tree_ad_insert(chunks_ad, node);\n\t\tnode = NULL;\n\t}\n\tmalloc_mutex_unlock(&chunks_mtx);\n\n\tzeroed = false;\n\tif (node != NULL) {\n\t\tif (node->zeroed) {\n\t\t\tzeroed = true;\n\t\t\t*zero = true;\n\t\t}\n\t\tbase_node_dealloc(node);\n\t}\n\tif (zeroed == false && *zero) {\n\t\tVALGRIND_MAKE_MEM_UNDEFINED(ret, size);\n\t\tmemset(ret, 0, size);\n\t}\n\treturn (ret);\n}\n\n/*\n * If the caller specifies (*zero == false), it is still possible to receive\n * zeroed memory, in which case *zero is toggled to true. arena_chunk_alloc()\n * takes advantage of this to avoid demanding zeroed chunks, but taking\n * advantage of them if they are returned.\n */\nvoid *\nchunk_alloc(size_t size, size_t alignment, bool base, bool *zero,\n dss_prec_t dss_prec)\n{\n\tvoid *ret;\n\n\tassert(size != 0);\n\tassert((size & chunksize_mask) == 0);\n\tassert(alignment != 0);\n\tassert((alignment & chunksize_mask) == 0);\n\n\t/* \"primary\" dss. */\n\tif (config_dss && dss_prec == dss_prec_primary) {\n\t\tif ((ret = chunk_recycle(&chunks_szad_dss, &chunks_ad_dss, size,\n\t\t alignment, base, zero)) != NULL)\n\t\t\tgoto label_return;\n\t\tif ((ret = chunk_alloc_dss(size, alignment, zero)) != NULL)\n\t\t\tgoto label_return;\n\t}\n\t/* mmap. */\n\tif ((ret = chunk_recycle(&chunks_szad_mmap, &chunks_ad_mmap, size,\n\t alignment, base, zero)) != NULL)\n\t\tgoto label_return;\n\tif ((ret = chunk_alloc_mmap(size, alignment, zero)) != NULL)\n\t\tgoto label_return;\n\t/* \"secondary\" dss. */\n\tif (config_dss && dss_prec == dss_prec_secondary) {\n\t\tif ((ret = chunk_recycle(&chunks_szad_dss, &chunks_ad_dss, size,\n\t\t alignment, base, zero)) != NULL)\n\t\t\tgoto label_return;\n\t\tif ((ret = chunk_alloc_dss(size, alignment, zero)) != NULL)\n\t\t\tgoto label_return;\n\t}\n\n\t/* All strategies for allocation failed. */\n\tret = NULL;\nlabel_return:\n\tif (config_ivsalloc && base == false && ret != NULL) {\n\t\tif (rtree_set(chunks_rtree, (uintptr_t)ret, ret)) {\n\t\t\tchunk_dealloc(ret, size, true);\n\t\t\treturn (NULL);\n\t\t}\n\t}\n\tif ((config_stats || config_prof) && ret != NULL) {\n\t\tbool gdump;\n\t\tmalloc_mutex_lock(&chunks_mtx);\n\t\tif (config_stats)\n\t\t\tstats_chunks.nchunks += (size / chunksize);\n\t\tstats_chunks.curchunks += (size / chunksize);\n\t\tif (stats_chunks.curchunks > stats_chunks.highchunks) {\n\t\t\tstats_chunks.highchunks = stats_chunks.curchunks;\n\t\t\tif (config_prof)\n\t\t\t\tgdump = true;\n\t\t} else if (config_prof)\n\t\t\tgdump = false;\n\t\tmalloc_mutex_unlock(&chunks_mtx);\n\t\tif (config_prof && opt_prof && opt_prof_gdump && gdump)\n\t\t\tprof_gdump();\n\t}\n\tif (config_debug && *zero && ret != NULL) {\n\t\tsize_t i;\n\t\tsize_t *p = (size_t *)(uintptr_t)ret;\n\n\t\tVALGRIND_MAKE_MEM_DEFINED(ret, size);\n\t\tfor (i = 0; i < size / sizeof(size_t); i++)\n\t\t\tassert(p[i] == 0);\n\t}\n\tassert(CHUNK_ADDR2BASE(ret) == ret);\n\treturn (ret);\n}\n\nstatic void\nchunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,\n size_t size)\n{\n\tbool unzeroed;\n\textent_node_t *xnode, *node, *prev, key;\n\n\tunzeroed = pages_purge(chunk, size);\n\n\t/*\n\t * Allocate a node before acquiring chunks_mtx even though it might not\n\t * be needed, because base_node_alloc() may cause a new base chunk to\n\t * be allocated, which could cause deadlock if chunks_mtx were already\n\t * held.\n\t */\n\txnode = base_node_alloc();\n\n\tmalloc_mutex_lock(&chunks_mtx);\n\tkey.addr = (void *)((uintptr_t)chunk + size);\n\tnode = extent_tree_ad_nsearch(chunks_ad, &key);\n\t/* Try to coalesce forward. */\n\tif (node != NULL && node->addr == key.addr) {\n\t\t/*\n\t\t * Coalesce chunk with the following address range. This does\n\t\t * not change the position within chunks_ad, so only\n\t\t * remove/insert from/into chunks_szad.\n\t\t */\n\t\textent_tree_szad_remove(chunks_szad, node);\n\t\tnode->addr = chunk;\n\t\tnode->size += size;\n\t\tnode->zeroed = (node->zeroed && (unzeroed == false));\n\t\textent_tree_szad_insert(chunks_szad, node);\n\t\tif (xnode != NULL)\n\t\t\tbase_node_dealloc(xnode);\n\t} else {\n\t\t/* Coalescing forward failed, so insert a new node. */\n\t\tif (xnode == NULL) {\n\t\t\t/*\n\t\t\t * base_node_alloc() failed, which is an exceedingly\n\t\t\t * unlikely failure. Leak chunk; its pages have\n\t\t\t * already been purged, so this is only a virtual\n\t\t\t * memory leak.\n\t\t\t */\n\t\t\tmalloc_mutex_unlock(&chunks_mtx);\n\t\t\treturn;\n\t\t}\n\t\tnode = xnode;\n\t\tnode->addr = chunk;\n\t\tnode->size = size;\n\t\tnode->zeroed = (unzeroed == false);\n\t\textent_tree_ad_insert(chunks_ad, node);\n\t\textent_tree_szad_insert(chunks_szad, node);\n\t}\n\n\t/* Try to coalesce backward. */\n\tprev = extent_tree_ad_prev(chunks_ad, node);\n\tif (prev != NULL && (void *)((uintptr_t)prev->addr + prev->size) ==\n\t chunk) {\n\t\t/*\n\t\t * Coalesce chunk with the previous address range. This does\n\t\t * not change the position within chunks_ad, so only\n\t\t * remove/insert node from/into chunks_szad.\n\t\t */\n\t\textent_tree_szad_remove(chunks_szad, prev);\n\t\textent_tree_ad_remove(chunks_ad, prev);\n\n\t\textent_tree_szad_remove(chunks_szad, node);\n\t\tnode->addr = prev->addr;\n\t\tnode->size += prev->size;\n\t\tnode->zeroed = (node->zeroed && prev->zeroed);\n\t\textent_tree_szad_insert(chunks_szad, node);\n\n\t\tbase_node_dealloc(prev);\n\t}\n\tmalloc_mutex_unlock(&chunks_mtx);\n}\n\nvoid\nchunk_unmap(void *chunk, size_t size)\n{\n\tassert(chunk != NULL);\n\tassert(CHUNK_ADDR2BASE(chunk) == chunk);\n\tassert(size != 0);\n\tassert((size & chunksize_mask) == 0);\n\n\tif (config_dss && chunk_in_dss(chunk))\n\t\tchunk_record(&chunks_szad_dss, &chunks_ad_dss, chunk, size);\n\telse if (chunk_dealloc_mmap(chunk, size))\n\t\tchunk_record(&chunks_szad_mmap, &chunks_ad_mmap, chunk, size);\n}\n\nvoid\nchunk_dealloc(void *chunk, size_t size, bool unmap)\n{\n\n\tassert(chunk != NULL);\n\tassert(CHUNK_ADDR2BASE(chunk) == chunk);\n\tassert(size != 0);\n\tassert((size & chunksize_mask) == 0);\n\n\tif (config_ivsalloc)\n\t\trtree_set(chunks_rtree, (uintptr_t)chunk, NULL);\n\tif (config_stats || config_prof) {\n\t\tmalloc_mutex_lock(&chunks_mtx);\n\t\tassert(stats_chunks.curchunks >= (size / chunksize));\n\t\tstats_chunks.curchunks -= (size / chunksize);\n\t\tmalloc_mutex_unlock(&chunks_mtx);\n\t}\n\n\tif (unmap)\n\t\tchunk_unmap(chunk, size);\n}\n\nbool\nchunk_boot(void)\n{\n\n\t/* Set variables according to the value of opt_lg_chunk. */\n\tchunksize = (ZU(1) << opt_lg_chunk);\n\tassert(chunksize >= PAGE);\n\tchunksize_mask = chunksize - 1;\n\tchunk_npages = (chunksize >> LG_PAGE);\n\n\tif (config_stats || config_prof) {\n\t\tif (malloc_mutex_init(&chunks_mtx))\n\t\t\treturn (true);\n\t\tmemset(&stats_chunks, 0, sizeof(chunk_stats_t));\n\t}\n\tif (config_dss && chunk_dss_boot())\n\t\treturn (true);\n\textent_tree_szad_new(&chunks_szad_mmap);\n\textent_tree_ad_new(&chunks_ad_mmap);\n\textent_tree_szad_new(&chunks_szad_dss);\n\textent_tree_ad_new(&chunks_ad_dss);\n\tif (config_ivsalloc) {\n\t\tchunks_rtree = rtree_new((ZU(1) << (LG_SIZEOF_PTR+3)) -\n\t\t opt_lg_chunk);\n\t\tif (chunks_rtree == NULL)\n\t\t\treturn (true);\n\t}\n\n\treturn (false);\n}\n\nvoid\nchunk_prefork(void)\n{\n\n\tmalloc_mutex_lock(&chunks_mtx);\n\tif (config_ivsalloc)\n\t\trtree_prefork(chunks_rtree);\n\tchunk_dss_prefork();\n}\n\nvoid\nchunk_postfork_parent(void)\n{\n\n\tchunk_dss_postfork_parent();\n\tif (config_ivsalloc)\n\t\trtree_postfork_parent(chunks_rtree);\n\tmalloc_mutex_postfork_parent(&chunks_mtx);\n}\n\nvoid\nchunk_postfork_child(void)\n{\n\n\tchunk_dss_postfork_child();\n\tif (config_ivsalloc)\n\t\trtree_postfork_child(chunks_rtree);\n\tmalloc_mutex_postfork_child(&chunks_mtx);\n}\n" }, { "path": "deps/jemalloc/src/chunk_dss.c", "content": "#define\tJEMALLOC_CHUNK_DSS_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n/******************************************************************************/\n/* Data. */\n\nconst char\t*dss_prec_names[] = {\n\t\"disabled\",\n\t\"primary\",\n\t\"secondary\",\n\t\"N/A\"\n};\n\n/* Current dss precedence default, used when creating new arenas. */\nstatic dss_prec_t\tdss_prec_default = DSS_PREC_DEFAULT;\n\n/*\n * Protects sbrk() calls. This avoids malloc races among threads, though it\n * does not protect against races with threads that call sbrk() directly.\n */\nstatic malloc_mutex_t\tdss_mtx;\n\n/* Base address of the DSS. */\nstatic void\t\t*dss_base;\n/* Current end of the DSS, or ((void *)-1) if the DSS is exhausted. */\nstatic void\t\t*dss_prev;\n/* Current upper limit on DSS addresses. */\nstatic void\t\t*dss_max;\n\n/******************************************************************************/\n\n#ifndef JEMALLOC_HAVE_SBRK\nstatic void *\nsbrk(intptr_t increment)\n{\n\n\tnot_implemented();\n\n\treturn (NULL);\n}\n#endif\n\ndss_prec_t\nchunk_dss_prec_get(void)\n{\n\tdss_prec_t ret;\n\n\tif (config_dss == false)\n\t\treturn (dss_prec_disabled);\n\tmalloc_mutex_lock(&dss_mtx);\n\tret = dss_prec_default;\n\tmalloc_mutex_unlock(&dss_mtx);\n\treturn (ret);\n}\n\nbool\nchunk_dss_prec_set(dss_prec_t dss_prec)\n{\n\n\tif (config_dss == false)\n\t\treturn (true);\n\tmalloc_mutex_lock(&dss_mtx);\n\tdss_prec_default = dss_prec;\n\tmalloc_mutex_unlock(&dss_mtx);\n\treturn (false);\n}\n\nvoid *\nchunk_alloc_dss(size_t size, size_t alignment, bool *zero)\n{\n\tvoid *ret;\n\n\tcassert(config_dss);\n\tassert(size > 0 && (size & chunksize_mask) == 0);\n\tassert(alignment > 0 && (alignment & chunksize_mask) == 0);\n\n\t/*\n\t * sbrk() uses a signed increment argument, so take care not to\n\t * interpret a huge allocation request as a negative increment.\n\t */\n\tif ((intptr_t)size < 0)\n\t\treturn (NULL);\n\n\tmalloc_mutex_lock(&dss_mtx);\n\tif (dss_prev != (void *)-1) {\n\t\tsize_t gap_size, cpad_size;\n\t\tvoid *cpad, *dss_next;\n\t\tintptr_t incr;\n\n\t\t/*\n\t\t * The loop is necessary to recover from races with other\n\t\t * threads that are using the DSS for something other than\n\t\t * malloc.\n\t\t */\n\t\tdo {\n\t\t\t/* Get the current end of the DSS. */\n\t\t\tdss_max = sbrk(0);\n\t\t\t/*\n\t\t\t * Calculate how much padding is necessary to\n\t\t\t * chunk-align the end of the DSS.\n\t\t\t */\n\t\t\tgap_size = (chunksize - CHUNK_ADDR2OFFSET(dss_max)) &\n\t\t\t chunksize_mask;\n\t\t\t/*\n\t\t\t * Compute how much chunk-aligned pad space (if any) is\n\t\t\t * necessary to satisfy alignment. This space can be\n\t\t\t * recycled for later use.\n\t\t\t */\n\t\t\tcpad = (void *)((uintptr_t)dss_max + gap_size);\n\t\t\tret = (void *)ALIGNMENT_CEILING((uintptr_t)dss_max,\n\t\t\t alignment);\n\t\t\tcpad_size = (uintptr_t)ret - (uintptr_t)cpad;\n\t\t\tdss_next = (void *)((uintptr_t)ret + size);\n\t\t\tif ((uintptr_t)ret < (uintptr_t)dss_max ||\n\t\t\t (uintptr_t)dss_next < (uintptr_t)dss_max) {\n\t\t\t\t/* Wrap-around. */\n\t\t\t\tmalloc_mutex_unlock(&dss_mtx);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tincr = gap_size + cpad_size + size;\n\t\t\tdss_prev = sbrk(incr);\n\t\t\tif (dss_prev == dss_max) {\n\t\t\t\t/* Success. */\n\t\t\t\tdss_max = dss_next;\n\t\t\t\tmalloc_mutex_unlock(&dss_mtx);\n\t\t\t\tif (cpad_size != 0)\n\t\t\t\t\tchunk_unmap(cpad, cpad_size);\n\t\t\t\tif (*zero) {\n\t\t\t\t\tVALGRIND_MAKE_MEM_UNDEFINED(ret, size);\n\t\t\t\t\tmemset(ret, 0, size);\n\t\t\t\t}\n\t\t\t\treturn (ret);\n\t\t\t}\n\t\t} while (dss_prev != (void *)-1);\n\t}\n\tmalloc_mutex_unlock(&dss_mtx);\n\n\treturn (NULL);\n}\n\nbool\nchunk_in_dss(void *chunk)\n{\n\tbool ret;\n\n\tcassert(config_dss);\n\n\tmalloc_mutex_lock(&dss_mtx);\n\tif ((uintptr_t)chunk >= (uintptr_t)dss_base\n\t && (uintptr_t)chunk < (uintptr_t)dss_max)\n\t\tret = true;\n\telse\n\t\tret = false;\n\tmalloc_mutex_unlock(&dss_mtx);\n\n\treturn (ret);\n}\n\nbool\nchunk_dss_boot(void)\n{\n\n\tcassert(config_dss);\n\n\tif (malloc_mutex_init(&dss_mtx))\n\t\treturn (true);\n\tdss_base = sbrk(0);\n\tdss_prev = dss_base;\n\tdss_max = dss_base;\n\n\treturn (false);\n}\n\nvoid\nchunk_dss_prefork(void)\n{\n\n\tif (config_dss)\n\t\tmalloc_mutex_prefork(&dss_mtx);\n}\n\nvoid\nchunk_dss_postfork_parent(void)\n{\n\n\tif (config_dss)\n\t\tmalloc_mutex_postfork_parent(&dss_mtx);\n}\n\nvoid\nchunk_dss_postfork_child(void)\n{\n\n\tif (config_dss)\n\t\tmalloc_mutex_postfork_child(&dss_mtx);\n}\n\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/src/chunk_mmap.c", "content": "#define\tJEMALLOC_CHUNK_MMAP_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\t*pages_map(void *addr, size_t size);\nstatic void\tpages_unmap(void *addr, size_t size);\nstatic void\t*chunk_alloc_mmap_slow(size_t size, size_t alignment,\n bool *zero);\n\n/******************************************************************************/\n\nstatic void *\npages_map(void *addr, size_t size)\n{\n\tvoid *ret;\n\n\tassert(size != 0);\n\n#ifdef _WIN32\n\t/*\n\t * If VirtualAlloc can't allocate at the given address when one is\n\t * given, it fails and returns NULL.\n\t */\n\tret = VirtualAlloc(addr, size, MEM_COMMIT | MEM_RESERVE,\n\t PAGE_READWRITE);\n#else\n\t/*\n\t * We don't use MAP_FIXED here, because it can cause the *replacement*\n\t * of existing mappings, and we only want to create new mappings.\n\t */\n\tret = mmap(addr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON,\n\t -1, 0);\n\tassert(ret != NULL);\n\n\tif (ret == MAP_FAILED)\n\t\tret = NULL;\n\telse if (addr != NULL && ret != addr) {\n\t\t/*\n\t\t * We succeeded in mapping memory, but not in the right place.\n\t\t */\n\t\tif (munmap(ret, size) == -1) {\n\t\t\tchar buf[BUFERROR_BUF];\n\n\t\t\tbuferror(buf, sizeof(buf));\n\t\t\tmalloc_printf(\": Error in \"\n#ifdef _WIN32\n\t\t \"VirtualFree\"\n#else\n\t\t \"munmap\"\n#endif\n\t\t \"(): %s\\n\", buf);\n\t\tif (opt_abort)\n\t\t\tabort();\n\t}\n}\n\nstatic void *\npages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size)\n{\n\tvoid *ret = (void *)((uintptr_t)addr + leadsize);\n\n\tassert(alloc_size >= leadsize + size);\n#ifdef _WIN32\n\t{\n\t\tvoid *new_addr;\n\n\t\tpages_unmap(addr, alloc_size);\n\t\tnew_addr = pages_map(ret, size);\n\t\tif (new_addr == ret)\n\t\t\treturn (ret);\n\t\tif (new_addr)\n\t\t\tpages_unmap(new_addr, size);\n\t\treturn (NULL);\n\t}\n#else\n\t{\n\t\tsize_t trailsize = alloc_size - leadsize - size;\n\n\t\tif (leadsize != 0)\n\t\t\tpages_unmap(addr, leadsize);\n\t\tif (trailsize != 0)\n\t\t\tpages_unmap((void *)((uintptr_t)ret + size), trailsize);\n\t\treturn (ret);\n\t}\n#endif\n}\n\nbool\npages_purge(void *addr, size_t length)\n{\n\tbool unzeroed;\n\n#ifdef _WIN32\n\tVirtualAlloc(addr, length, MEM_RESET, PAGE_READWRITE);\n\tunzeroed = true;\n#else\n# ifdef JEMALLOC_PURGE_MADVISE_DONTNEED\n# define JEMALLOC_MADV_PURGE MADV_DONTNEED\n# define JEMALLOC_MADV_ZEROS true\n# elif defined(JEMALLOC_PURGE_MADVISE_FREE)\n# define JEMALLOC_MADV_PURGE MADV_FREE\n# define JEMALLOC_MADV_ZEROS false\n# else\n# error \"No method defined for purging unused dirty pages.\"\n# endif\n\tint err = madvise(addr, length, JEMALLOC_MADV_PURGE);\n\tunzeroed = (JEMALLOC_MADV_ZEROS == false || err != 0);\n# undef JEMALLOC_MADV_PURGE\n# undef JEMALLOC_MADV_ZEROS\n#endif\n\treturn (unzeroed);\n}\n\nstatic void *\nchunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero)\n{\n\tvoid *ret, *pages;\n\tsize_t alloc_size, leadsize;\n\n\talloc_size = size + alignment - PAGE;\n\t/* Beware size_t wrap-around. */\n\tif (alloc_size < size)\n\t\treturn (NULL);\n\tdo {\n\t\tpages = pages_map(NULL, alloc_size);\n\t\tif (pages == NULL)\n\t\t\treturn (NULL);\n\t\tleadsize = ALIGNMENT_CEILING((uintptr_t)pages, alignment) -\n\t\t (uintptr_t)pages;\n\t\tret = pages_trim(pages, alloc_size, leadsize, size);\n\t} while (ret == NULL);\n\n\tassert(ret != NULL);\n\t*zero = true;\n\treturn (ret);\n}\n\nvoid *\nchunk_alloc_mmap(size_t size, size_t alignment, bool *zero)\n{\n\tvoid *ret;\n\tsize_t offset;\n\n\t/*\n\t * Ideally, there would be a way to specify alignment to mmap() (like\n\t * NetBSD has), but in the absence of such a feature, we have to work\n\t * hard to efficiently create aligned mappings. The reliable, but\n\t * slow method is to create a mapping that is over-sized, then trim the\n\t * excess. However, that always results in one or two calls to\n\t * pages_unmap().\n\t *\n\t * Optimistically try mapping precisely the right amount before falling\n\t * back to the slow method, with the expectation that the optimistic\n\t * approach works most of the time.\n\t */\n\n\tassert(alignment != 0);\n\tassert((alignment & chunksize_mask) == 0);\n\n\tret = pages_map(NULL, size);\n\tif (ret == NULL)\n\t\treturn (NULL);\n\toffset = ALIGNMENT_ADDR2OFFSET(ret, alignment);\n\tif (offset != 0) {\n\t\tpages_unmap(ret, size);\n\t\treturn (chunk_alloc_mmap_slow(size, alignment, zero));\n\t}\n\n\tassert(ret != NULL);\n\t*zero = true;\n\treturn (ret);\n}\n\nbool\nchunk_dealloc_mmap(void *chunk, size_t size)\n{\n\n\tif (config_munmap)\n\t\tpages_unmap(chunk, size);\n\n\treturn (config_munmap == false);\n}\n" }, { "path": "deps/jemalloc/src/ckh.c", "content": "/*\n *******************************************************************************\n * Implementation of (2^1+,2) cuckoo hashing, where 2^1+ indicates that each\n * hash bucket contains 2^n cells, for n >= 1, and 2 indicates that two hash\n * functions are employed. The original cuckoo hashing algorithm was described\n * in:\n *\n * Pagh, R., F.F. Rodler (2004) Cuckoo Hashing. Journal of Algorithms\n * 51(2):122-144.\n *\n * Generalization of cuckoo hashing was discussed in:\n *\n * Erlingsson, U., M. Manasse, F. McSherry (2006) A cool and practical\n * alternative to traditional hash tables. In Proceedings of the 7th\n * Workshop on Distributed Data and Structures (WDAS'06), Santa Clara, CA,\n * January 2006.\n *\n * This implementation uses precisely two hash functions because that is the\n * fewest that can work, and supporting multiple hashes is an implementation\n * burden. Here is a reproduction of Figure 1 from Erlingsson et al. (2006)\n * that shows approximate expected maximum load factors for various\n * configurations:\n *\n * | #cells/bucket |\n * #hashes | 1 | 2 | 4 | 8 |\n * --------+-------+-------+-------+-------+\n * 1 | 0.006 | 0.006 | 0.03 | 0.12 |\n * 2 | 0.49 | 0.86 |>0.93< |>0.96< |\n * 3 | 0.91 | 0.97 | 0.98 | 0.999 |\n * 4 | 0.97 | 0.99 | 0.999 | |\n *\n * The number of cells per bucket is chosen such that a bucket fits in one cache\n * line. So, on 32- and 64-bit systems, we use (8,2) and (4,2) cuckoo hashing,\n * respectively.\n *\n ******************************************************************************/\n#define\tJEMALLOC_CKH_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic bool\tckh_grow(ckh_t *ckh);\nstatic void\tckh_shrink(ckh_t *ckh);\n\n/******************************************************************************/\n\n/*\n * Search bucket for key and return the cell number if found; SIZE_T_MAX\n * otherwise.\n */\nJEMALLOC_INLINE size_t\nckh_bucket_search(ckh_t *ckh, size_t bucket, const void *key)\n{\n\tckhc_t *cell;\n\tunsigned i;\n\n\tfor (i = 0; i < (ZU(1) << LG_CKH_BUCKET_CELLS); i++) {\n\t\tcell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) + i];\n\t\tif (cell->key != NULL && ckh->keycomp(key, cell->key))\n\t\t\treturn ((bucket << LG_CKH_BUCKET_CELLS) + i);\n\t}\n\n\treturn (SIZE_T_MAX);\n}\n\n/*\n * Search table for key and return cell number if found; SIZE_T_MAX otherwise.\n */\nJEMALLOC_INLINE size_t\nckh_isearch(ckh_t *ckh, const void *key)\n{\n\tsize_t hash1, hash2, bucket, cell;\n\n\tassert(ckh != NULL);\n\n\tckh->hash(key, ckh->lg_curbuckets, &hash1, &hash2);\n\n\t/* Search primary bucket. */\n\tbucket = hash1 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\tcell = ckh_bucket_search(ckh, bucket, key);\n\tif (cell != SIZE_T_MAX)\n\t\treturn (cell);\n\n\t/* Search secondary bucket. */\n\tbucket = hash2 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\tcell = ckh_bucket_search(ckh, bucket, key);\n\treturn (cell);\n}\n\nJEMALLOC_INLINE bool\nckh_try_bucket_insert(ckh_t *ckh, size_t bucket, const void *key,\n const void *data)\n{\n\tckhc_t *cell;\n\tunsigned offset, i;\n\n\t/*\n\t * Cycle through the cells in the bucket, starting at a random position.\n\t * The randomness avoids worst-case search overhead as buckets fill up.\n\t */\n\tprng32(offset, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C);\n\tfor (i = 0; i < (ZU(1) << LG_CKH_BUCKET_CELLS); i++) {\n\t\tcell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) +\n\t\t ((i + offset) & ((ZU(1) << LG_CKH_BUCKET_CELLS) - 1))];\n\t\tif (cell->key == NULL) {\n\t\t\tcell->key = key;\n\t\t\tcell->data = data;\n\t\t\tckh->count++;\n\t\t\treturn (false);\n\t\t}\n\t}\n\n\treturn (true);\n}\n\n/*\n * No space is available in bucket. Randomly evict an item, then try to find an\n * alternate location for that item. Iteratively repeat this\n * eviction/relocation procedure until either success or detection of an\n * eviction/relocation bucket cycle.\n */\nJEMALLOC_INLINE bool\nckh_evict_reloc_insert(ckh_t *ckh, size_t argbucket, void const **argkey,\n void const **argdata)\n{\n\tconst void *key, *data, *tkey, *tdata;\n\tckhc_t *cell;\n\tsize_t hash1, hash2, bucket, tbucket;\n\tunsigned i;\n\n\tbucket = argbucket;\n\tkey = *argkey;\n\tdata = *argdata;\n\twhile (true) {\n\t\t/*\n\t\t * Choose a random item within the bucket to evict. This is\n\t\t * critical to correct function, because without (eventually)\n\t\t * evicting all items within a bucket during iteration, it\n\t\t * would be possible to get stuck in an infinite loop if there\n\t\t * were an item for which both hashes indicated the same\n\t\t * bucket.\n\t\t */\n\t\tprng32(i, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C);\n\t\tcell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) + i];\n\t\tassert(cell->key != NULL);\n\n\t\t/* Swap cell->{key,data} and {key,data} (evict). */\n\t\ttkey = cell->key; tdata = cell->data;\n\t\tcell->key = key; cell->data = data;\n\t\tkey = tkey; data = tdata;\n\n#ifdef CKH_COUNT\n\t\tckh->nrelocs++;\n#endif\n\n\t\t/* Find the alternate bucket for the evicted item. */\n\t\tckh->hash(key, ckh->lg_curbuckets, &hash1, &hash2);\n\t\ttbucket = hash2 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\t\tif (tbucket == bucket) {\n\t\t\ttbucket = hash1 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\t\t\t/*\n\t\t\t * It may be that (tbucket == bucket) still, if the\n\t\t\t * item's hashes both indicate this bucket. However,\n\t\t\t * we are guaranteed to eventually escape this bucket\n\t\t\t * during iteration, assuming pseudo-random item\n\t\t\t * selection (true randomness would make infinite\n\t\t\t * looping a remote possibility). The reason we can\n\t\t\t * never get trapped forever is that there are two\n\t\t\t * cases:\n\t\t\t *\n\t\t\t * 1) This bucket == argbucket, so we will quickly\n\t\t\t * detect an eviction cycle and terminate.\n\t\t\t * 2) An item was evicted to this bucket from another,\n\t\t\t * which means that at least one item in this bucket\n\t\t\t * has hashes that indicate distinct buckets.\n\t\t\t */\n\t\t}\n\t\t/* Check for a cycle. */\n\t\tif (tbucket == argbucket) {\n\t\t\t*argkey = key;\n\t\t\t*argdata = data;\n\t\t\treturn (true);\n\t\t}\n\n\t\tbucket = tbucket;\n\t\tif (ckh_try_bucket_insert(ckh, bucket, key, data) == false)\n\t\t\treturn (false);\n\t}\n}\n\nJEMALLOC_INLINE bool\nckh_try_insert(ckh_t *ckh, void const**argkey, void const**argdata)\n{\n\tsize_t hash1, hash2, bucket;\n\tconst void *key = *argkey;\n\tconst void *data = *argdata;\n\n\tckh->hash(key, ckh->lg_curbuckets, &hash1, &hash2);\n\n\t/* Try to insert in primary bucket. */\n\tbucket = hash1 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\tif (ckh_try_bucket_insert(ckh, bucket, key, data) == false)\n\t\treturn (false);\n\n\t/* Try to insert in secondary bucket. */\n\tbucket = hash2 & ((ZU(1) << ckh->lg_curbuckets) - 1);\n\tif (ckh_try_bucket_insert(ckh, bucket, key, data) == false)\n\t\treturn (false);\n\n\t/*\n\t * Try to find a place for this item via iterative eviction/relocation.\n\t */\n\treturn (ckh_evict_reloc_insert(ckh, bucket, argkey, argdata));\n}\n\n/*\n * Try to rebuild the hash table from scratch by inserting all items from the\n * old table into the new.\n */\nJEMALLOC_INLINE bool\nckh_rebuild(ckh_t *ckh, ckhc_t *aTab)\n{\n\tsize_t count, i, nins;\n\tconst void *key, *data;\n\n\tcount = ckh->count;\n\tckh->count = 0;\n\tfor (i = nins = 0; nins < count; i++) {\n\t\tif (aTab[i].key != NULL) {\n\t\t\tkey = aTab[i].key;\n\t\t\tdata = aTab[i].data;\n\t\t\tif (ckh_try_insert(ckh, &key, &data)) {\n\t\t\t\tckh->count = count;\n\t\t\t\treturn (true);\n\t\t\t}\n\t\t\tnins++;\n\t\t}\n\t}\n\n\treturn (false);\n}\n\nstatic bool\nckh_grow(ckh_t *ckh)\n{\n\tbool ret;\n\tckhc_t *tab, *ttab;\n\tsize_t lg_curcells;\n\tunsigned lg_prevbuckets;\n\n#ifdef CKH_COUNT\n\tckh->ngrows++;\n#endif\n\n\t/*\n\t * It is possible (though unlikely, given well behaved hashes) that the\n\t * table will have to be doubled more than once in order to create a\n\t * usable table.\n\t */\n\tlg_prevbuckets = ckh->lg_curbuckets;\n\tlg_curcells = ckh->lg_curbuckets + LG_CKH_BUCKET_CELLS;\n\twhile (true) {\n\t\tsize_t usize;\n\n\t\tlg_curcells++;\n\t\tusize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);\n\t\tif (usize == 0) {\n\t\t\tret = true;\n\t\t\tgoto label_return;\n\t\t}\n\t\ttab = (ckhc_t *)ipalloc(usize, CACHELINE, true);\n\t\tif (tab == NULL) {\n\t\t\tret = true;\n\t\t\tgoto label_return;\n\t\t}\n\t\t/* Swap in new table. */\n\t\tttab = ckh->tab;\n\t\tckh->tab = tab;\n\t\ttab = ttab;\n\t\tckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;\n\n\t\tif (ckh_rebuild(ckh, tab) == false) {\n\t\t\tidalloc(tab);\n\t\t\tbreak;\n\t\t}\n\n\t\t/* Rebuilding failed, so back out partially rebuilt table. */\n\t\tidalloc(ckh->tab);\n\t\tckh->tab = tab;\n\t\tckh->lg_curbuckets = lg_prevbuckets;\n\t}\n\n\tret = false;\nlabel_return:\n\treturn (ret);\n}\n\nstatic void\nckh_shrink(ckh_t *ckh)\n{\n\tckhc_t *tab, *ttab;\n\tsize_t lg_curcells, usize;\n\tunsigned lg_prevbuckets;\n\n\t/*\n\t * It is possible (though unlikely, given well behaved hashes) that the\n\t * table rebuild will fail.\n\t */\n\tlg_prevbuckets = ckh->lg_curbuckets;\n\tlg_curcells = ckh->lg_curbuckets + LG_CKH_BUCKET_CELLS - 1;\n\tusize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);\n\tif (usize == 0)\n\t\treturn;\n\ttab = (ckhc_t *)ipalloc(usize, CACHELINE, true);\n\tif (tab == NULL) {\n\t\t/*\n\t\t * An OOM error isn't worth propagating, since it doesn't\n\t\t * prevent this or future operations from proceeding.\n\t\t */\n\t\treturn;\n\t}\n\t/* Swap in new table. */\n\tttab = ckh->tab;\n\tckh->tab = tab;\n\ttab = ttab;\n\tckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;\n\n\tif (ckh_rebuild(ckh, tab) == false) {\n\t\tidalloc(tab);\n#ifdef CKH_COUNT\n\t\tckh->nshrinks++;\n#endif\n\t\treturn;\n\t}\n\n\t/* Rebuilding failed, so back out partially rebuilt table. */\n\tidalloc(ckh->tab);\n\tckh->tab = tab;\n\tckh->lg_curbuckets = lg_prevbuckets;\n#ifdef CKH_COUNT\n\tckh->nshrinkfails++;\n#endif\n}\n\nbool\nckh_new(ckh_t *ckh, size_t minitems, ckh_hash_t *hash, ckh_keycomp_t *keycomp)\n{\n\tbool ret;\n\tsize_t mincells, usize;\n\tunsigned lg_mincells;\n\n\tassert(minitems > 0);\n\tassert(hash != NULL);\n\tassert(keycomp != NULL);\n\n#ifdef CKH_COUNT\n\tckh->ngrows = 0;\n\tckh->nshrinks = 0;\n\tckh->nshrinkfails = 0;\n\tckh->ninserts = 0;\n\tckh->nrelocs = 0;\n#endif\n\tckh->prng_state = 42; /* Value doesn't really matter. */\n\tckh->count = 0;\n\n\t/*\n\t * Find the minimum power of 2 that is large enough to fit aBaseCount\n\t * entries. We are using (2+,2) cuckoo hashing, which has an expected\n\t * maximum load factor of at least ~0.86, so 0.75 is a conservative load\n\t * factor that will typically allow 2^aLgMinItems to fit without ever\n\t * growing the table.\n\t */\n\tassert(LG_CKH_BUCKET_CELLS > 0);\n\tmincells = ((minitems + (3 - (minitems % 3))) / 3) << 2;\n\tfor (lg_mincells = LG_CKH_BUCKET_CELLS;\n\t (ZU(1) << lg_mincells) < mincells;\n\t lg_mincells++)\n\t\t; /* Do nothing. */\n\tckh->lg_minbuckets = lg_mincells - LG_CKH_BUCKET_CELLS;\n\tckh->lg_curbuckets = lg_mincells - LG_CKH_BUCKET_CELLS;\n\tckh->hash = hash;\n\tckh->keycomp = keycomp;\n\n\tusize = sa2u(sizeof(ckhc_t) << lg_mincells, CACHELINE);\n\tif (usize == 0) {\n\t\tret = true;\n\t\tgoto label_return;\n\t}\n\tckh->tab = (ckhc_t *)ipalloc(usize, CACHELINE, true);\n\tif (ckh->tab == NULL) {\n\t\tret = true;\n\t\tgoto label_return;\n\t}\n\n\tret = false;\nlabel_return:\n\treturn (ret);\n}\n\nvoid\nckh_delete(ckh_t *ckh)\n{\n\n\tassert(ckh != NULL);\n\n#ifdef CKH_VERBOSE\n\tmalloc_printf(\n\t \"%s(%p): ngrows: %\"PRIu64\", nshrinks: %\"PRIu64\",\"\n\t \" nshrinkfails: %\"PRIu64\", ninserts: %\"PRIu64\",\"\n\t \" nrelocs: %\"PRIu64\"\\n\", __func__, ckh,\n\t (unsigned long long)ckh->ngrows,\n\t (unsigned long long)ckh->nshrinks,\n\t (unsigned long long)ckh->nshrinkfails,\n\t (unsigned long long)ckh->ninserts,\n\t (unsigned long long)ckh->nrelocs);\n#endif\n\n\tidalloc(ckh->tab);\n#ifdef JEMALLOC_DEBUG\n\tmemset(ckh, 0x5a, sizeof(ckh_t));\n#endif\n}\n\nsize_t\nckh_count(ckh_t *ckh)\n{\n\n\tassert(ckh != NULL);\n\n\treturn (ckh->count);\n}\n\nbool\nckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data)\n{\n\tsize_t i, ncells;\n\n\tfor (i = *tabind, ncells = (ZU(1) << (ckh->lg_curbuckets +\n\t LG_CKH_BUCKET_CELLS)); i < ncells; i++) {\n\t\tif (ckh->tab[i].key != NULL) {\n\t\t\tif (key != NULL)\n\t\t\t\t*key = (void *)ckh->tab[i].key;\n\t\t\tif (data != NULL)\n\t\t\t\t*data = (void *)ckh->tab[i].data;\n\t\t\t*tabind = i + 1;\n\t\t\treturn (false);\n\t\t}\n\t}\n\n\treturn (true);\n}\n\nbool\nckh_insert(ckh_t *ckh, const void *key, const void *data)\n{\n\tbool ret;\n\n\tassert(ckh != NULL);\n\tassert(ckh_search(ckh, key, NULL, NULL));\n\n#ifdef CKH_COUNT\n\tckh->ninserts++;\n#endif\n\n\twhile (ckh_try_insert(ckh, &key, &data)) {\n\t\tif (ckh_grow(ckh)) {\n\t\t\tret = true;\n\t\t\tgoto label_return;\n\t\t}\n\t}\n\n\tret = false;\nlabel_return:\n\treturn (ret);\n}\n\nbool\nckh_remove(ckh_t *ckh, const void *searchkey, void **key, void **data)\n{\n\tsize_t cell;\n\n\tassert(ckh != NULL);\n\n\tcell = ckh_isearch(ckh, searchkey);\n\tif (cell != SIZE_T_MAX) {\n\t\tif (key != NULL)\n\t\t\t*key = (void *)ckh->tab[cell].key;\n\t\tif (data != NULL)\n\t\t\t*data = (void *)ckh->tab[cell].data;\n\t\tckh->tab[cell].key = NULL;\n\t\tckh->tab[cell].data = NULL; /* Not necessary. */\n\n\t\tckh->count--;\n\t\t/* Try to halve the table if it is less than 1/4 full. */\n\t\tif (ckh->count < (ZU(1) << (ckh->lg_curbuckets\n\t\t + LG_CKH_BUCKET_CELLS - 2)) && ckh->lg_curbuckets\n\t\t > ckh->lg_minbuckets) {\n\t\t\t/* Ignore error due to OOM. */\n\t\t\tckh_shrink(ckh);\n\t\t}\n\n\t\treturn (false);\n\t}\n\n\treturn (true);\n}\n\nbool\nckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data)\n{\n\tsize_t cell;\n\n\tassert(ckh != NULL);\n\n\tcell = ckh_isearch(ckh, searchkey);\n\tif (cell != SIZE_T_MAX) {\n\t\tif (key != NULL)\n\t\t\t*key = (void *)ckh->tab[cell].key;\n\t\tif (data != NULL)\n\t\t\t*data = (void *)ckh->tab[cell].data;\n\t\treturn (false);\n\t}\n\n\treturn (true);\n}\n\nvoid\nckh_string_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2)\n{\n\tsize_t ret1, ret2;\n\tuint64_t h;\n\n\tassert(minbits <= 32 || (SIZEOF_PTR == 8 && minbits <= 64));\n\tassert(hash1 != NULL);\n\tassert(hash2 != NULL);\n\n\th = hash(key, strlen((const char *)key), UINT64_C(0x94122f335b332aea));\n\tif (minbits <= 32) {\n\t\t/*\n\t\t * Avoid doing multiple hashes, since a single hash provides\n\t\t * enough bits.\n\t\t */\n\t\tret1 = h & ZU(0xffffffffU);\n\t\tret2 = h >> 32;\n\t} else {\n\t\tret1 = h;\n\t\tret2 = hash(key, strlen((const char *)key),\n\t\t UINT64_C(0x8432a476666bbc13));\n\t}\n\n\t*hash1 = ret1;\n\t*hash2 = ret2;\n}\n\nbool\nckh_string_keycomp(const void *k1, const void *k2)\n{\n\n assert(k1 != NULL);\n assert(k2 != NULL);\n\n return (strcmp((char *)k1, (char *)k2) ? false : true);\n}\n\nvoid\nckh_pointer_hash(const void *key, unsigned minbits, size_t *hash1,\n size_t *hash2)\n{\n\tsize_t ret1, ret2;\n\tuint64_t h;\n\tunion {\n\t\tconst void\t*v;\n\t\tuint64_t\ti;\n\t} u;\n\n\tassert(minbits <= 32 || (SIZEOF_PTR == 8 && minbits <= 64));\n\tassert(hash1 != NULL);\n\tassert(hash2 != NULL);\n\n\tassert(sizeof(u.v) == sizeof(u.i));\n#if (LG_SIZEOF_PTR != LG_SIZEOF_INT)\n\tu.i = 0;\n#endif\n\tu.v = key;\n\th = hash(&u.i, sizeof(u.i), UINT64_C(0xd983396e68886082));\n\tif (minbits <= 32) {\n\t\t/*\n\t\t * Avoid doing multiple hashes, since a single hash provides\n\t\t * enough bits.\n\t\t */\n\t\tret1 = h & ZU(0xffffffffU);\n\t\tret2 = h >> 32;\n\t} else {\n\t\tassert(SIZEOF_PTR == 8);\n\t\tret1 = h;\n\t\tret2 = hash(&u.i, sizeof(u.i), UINT64_C(0x5e2be9aff8709a5d));\n\t}\n\n\t*hash1 = ret1;\n\t*hash2 = ret2;\n}\n\nbool\nckh_pointer_keycomp(const void *k1, const void *k2)\n{\n\n\treturn ((k1 == k2) ? true : false);\n}\n" }, { "path": "deps/jemalloc/src/ctl.c", "content": "#define\tJEMALLOC_CTL_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\n/*\n * ctl_mtx protects the following:\n * - ctl_stats.*\n * - opt_prof_active\n */\nstatic malloc_mutex_t\tctl_mtx;\nstatic bool\t\tctl_initialized;\nstatic uint64_t\t\tctl_epoch;\nstatic ctl_stats_t\tctl_stats;\n\n/******************************************************************************/\n/* Helpers for named and indexed nodes. */\n\nstatic inline const ctl_named_node_t *\nctl_named_node(const ctl_node_t *node)\n{\n\n\treturn ((node->named) ? (const ctl_named_node_t *)node : NULL);\n}\n\nstatic inline const ctl_named_node_t *\nctl_named_children(const ctl_named_node_t *node, int index)\n{\n\tconst ctl_named_node_t *children = ctl_named_node(node->children);\n\n\treturn (children ? &children[index] : NULL);\n}\n\nstatic inline const ctl_indexed_node_t *\nctl_indexed_node(const ctl_node_t *node)\n{\n\n\treturn ((node->named == false) ? (const ctl_indexed_node_t *)node :\n\t NULL);\n}\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\n#define\tCTL_PROTO(n)\t\t\t\t\t\t\t\\\nstatic int\tn##_ctl(const size_t *mib, size_t miblen, void *oldp,\t\\\n size_t *oldlenp, void *newp, size_t newlen);\n\n#define\tINDEX_PROTO(n)\t\t\t\t\t\t\t\\\nstatic const ctl_named_node_t\t*n##_index(const size_t *mib,\t\t\\\n size_t miblen, size_t i);\n\nstatic bool\tctl_arena_init(ctl_arena_stats_t *astats);\nstatic void\tctl_arena_clear(ctl_arena_stats_t *astats);\nstatic void\tctl_arena_stats_amerge(ctl_arena_stats_t *cstats,\n arena_t *arena);\nstatic void\tctl_arena_stats_smerge(ctl_arena_stats_t *sstats,\n ctl_arena_stats_t *astats);\nstatic void\tctl_arena_refresh(arena_t *arena, unsigned i);\nstatic bool\tctl_grow(void);\nstatic void\tctl_refresh(void);\nstatic bool\tctl_init(void);\nstatic int\tctl_lookup(const char *name, ctl_node_t const **nodesp,\n size_t *mibp, size_t *depthp);\n\nCTL_PROTO(version)\nCTL_PROTO(epoch)\nCTL_PROTO(thread_tcache_enabled)\nCTL_PROTO(thread_tcache_flush)\nCTL_PROTO(thread_arena)\nCTL_PROTO(thread_allocated)\nCTL_PROTO(thread_allocatedp)\nCTL_PROTO(thread_deallocated)\nCTL_PROTO(thread_deallocatedp)\nCTL_PROTO(config_debug)\nCTL_PROTO(config_dss)\nCTL_PROTO(config_fill)\nCTL_PROTO(config_lazy_lock)\nCTL_PROTO(config_mremap)\nCTL_PROTO(config_munmap)\nCTL_PROTO(config_prof)\nCTL_PROTO(config_prof_libgcc)\nCTL_PROTO(config_prof_libunwind)\nCTL_PROTO(config_stats)\nCTL_PROTO(config_tcache)\nCTL_PROTO(config_tls)\nCTL_PROTO(config_utrace)\nCTL_PROTO(config_valgrind)\nCTL_PROTO(config_xmalloc)\nCTL_PROTO(opt_abort)\nCTL_PROTO(opt_dss)\nCTL_PROTO(opt_lg_chunk)\nCTL_PROTO(opt_narenas)\nCTL_PROTO(opt_lg_dirty_mult)\nCTL_PROTO(opt_stats_print)\nCTL_PROTO(opt_junk)\nCTL_PROTO(opt_zero)\nCTL_PROTO(opt_quarantine)\nCTL_PROTO(opt_redzone)\nCTL_PROTO(opt_utrace)\nCTL_PROTO(opt_valgrind)\nCTL_PROTO(opt_xmalloc)\nCTL_PROTO(opt_tcache)\nCTL_PROTO(opt_lg_tcache_max)\nCTL_PROTO(opt_prof)\nCTL_PROTO(opt_prof_prefix)\nCTL_PROTO(opt_prof_active)\nCTL_PROTO(opt_lg_prof_sample)\nCTL_PROTO(opt_lg_prof_interval)\nCTL_PROTO(opt_prof_gdump)\nCTL_PROTO(opt_prof_final)\nCTL_PROTO(opt_prof_leak)\nCTL_PROTO(opt_prof_accum)\nCTL_PROTO(arena_i_purge)\nstatic void\tarena_purge(unsigned arena_ind);\nCTL_PROTO(arena_i_dss)\nINDEX_PROTO(arena_i)\nCTL_PROTO(arenas_bin_i_size)\nCTL_PROTO(arenas_bin_i_nregs)\nCTL_PROTO(arenas_bin_i_run_size)\nINDEX_PROTO(arenas_bin_i)\nCTL_PROTO(arenas_lrun_i_size)\nINDEX_PROTO(arenas_lrun_i)\nCTL_PROTO(arenas_narenas)\nCTL_PROTO(arenas_initialized)\nCTL_PROTO(arenas_quantum)\nCTL_PROTO(arenas_page)\nCTL_PROTO(arenas_tcache_max)\nCTL_PROTO(arenas_nbins)\nCTL_PROTO(arenas_nhbins)\nCTL_PROTO(arenas_nlruns)\nCTL_PROTO(arenas_purge)\nCTL_PROTO(arenas_extend)\nCTL_PROTO(prof_active)\nCTL_PROTO(prof_dump)\nCTL_PROTO(prof_interval)\nCTL_PROTO(stats_chunks_current)\nCTL_PROTO(stats_chunks_total)\nCTL_PROTO(stats_chunks_high)\nCTL_PROTO(stats_huge_allocated)\nCTL_PROTO(stats_huge_nmalloc)\nCTL_PROTO(stats_huge_ndalloc)\nCTL_PROTO(stats_arenas_i_small_allocated)\nCTL_PROTO(stats_arenas_i_small_nmalloc)\nCTL_PROTO(stats_arenas_i_small_ndalloc)\nCTL_PROTO(stats_arenas_i_small_nrequests)\nCTL_PROTO(stats_arenas_i_large_allocated)\nCTL_PROTO(stats_arenas_i_large_nmalloc)\nCTL_PROTO(stats_arenas_i_large_ndalloc)\nCTL_PROTO(stats_arenas_i_large_nrequests)\nCTL_PROTO(stats_arenas_i_bins_j_allocated)\nCTL_PROTO(stats_arenas_i_bins_j_nmalloc)\nCTL_PROTO(stats_arenas_i_bins_j_ndalloc)\nCTL_PROTO(stats_arenas_i_bins_j_nrequests)\nCTL_PROTO(stats_arenas_i_bins_j_nfills)\nCTL_PROTO(stats_arenas_i_bins_j_nflushes)\nCTL_PROTO(stats_arenas_i_bins_j_nruns)\nCTL_PROTO(stats_arenas_i_bins_j_nreruns)\nCTL_PROTO(stats_arenas_i_bins_j_curruns)\nINDEX_PROTO(stats_arenas_i_bins_j)\nCTL_PROTO(stats_arenas_i_lruns_j_nmalloc)\nCTL_PROTO(stats_arenas_i_lruns_j_ndalloc)\nCTL_PROTO(stats_arenas_i_lruns_j_nrequests)\nCTL_PROTO(stats_arenas_i_lruns_j_curruns)\nINDEX_PROTO(stats_arenas_i_lruns_j)\nCTL_PROTO(stats_arenas_i_nthreads)\nCTL_PROTO(stats_arenas_i_dss)\nCTL_PROTO(stats_arenas_i_pactive)\nCTL_PROTO(stats_arenas_i_pdirty)\nCTL_PROTO(stats_arenas_i_mapped)\nCTL_PROTO(stats_arenas_i_npurge)\nCTL_PROTO(stats_arenas_i_nmadvise)\nCTL_PROTO(stats_arenas_i_purged)\nINDEX_PROTO(stats_arenas_i)\nCTL_PROTO(stats_cactive)\nCTL_PROTO(stats_allocated)\nCTL_PROTO(stats_active)\nCTL_PROTO(stats_mapped)\n\n/******************************************************************************/\n/* mallctl tree. */\n\n/* Maximum tree depth. */\n#define\tCTL_MAX_DEPTH\t6\n\n#define\tNAME(n)\t{true},\tn\n#define\tCHILD(t, c)\t\t\t\t\t\t\t\\\n\tsizeof(c##_node) / sizeof(ctl_##t##_node_t),\t\t\t\\\n\t(ctl_node_t *)c##_node,\t\t\t\t\t\t\\\n\tNULL\n#define\tCTL(c)\t0, NULL, c##_ctl\n\n/*\n * Only handles internal indexed nodes, since there are currently no external\n * ones.\n */\n#define\tINDEX(i)\t{false},\ti##_index\n\nstatic const ctl_named_node_t\ttcache_node[] = {\n\t{NAME(\"enabled\"),\tCTL(thread_tcache_enabled)},\n\t{NAME(\"flush\"),\t\tCTL(thread_tcache_flush)}\n};\n\nstatic const ctl_named_node_t\tthread_node[] = {\n\t{NAME(\"arena\"),\t\tCTL(thread_arena)},\n\t{NAME(\"allocated\"),\tCTL(thread_allocated)},\n\t{NAME(\"allocatedp\"),\tCTL(thread_allocatedp)},\n\t{NAME(\"deallocated\"),\tCTL(thread_deallocated)},\n\t{NAME(\"deallocatedp\"),\tCTL(thread_deallocatedp)},\n\t{NAME(\"tcache\"),\tCHILD(named, tcache)}\n};\n\nstatic const ctl_named_node_t\tconfig_node[] = {\n\t{NAME(\"debug\"),\t\t\tCTL(config_debug)},\n\t{NAME(\"dss\"),\t\t\tCTL(config_dss)},\n\t{NAME(\"fill\"),\t\t\tCTL(config_fill)},\n\t{NAME(\"lazy_lock\"),\t\tCTL(config_lazy_lock)},\n\t{NAME(\"mremap\"),\t\tCTL(config_mremap)},\n\t{NAME(\"munmap\"),\t\tCTL(config_munmap)},\n\t{NAME(\"prof\"),\t\t\tCTL(config_prof)},\n\t{NAME(\"prof_libgcc\"),\t\tCTL(config_prof_libgcc)},\n\t{NAME(\"prof_libunwind\"),\tCTL(config_prof_libunwind)},\n\t{NAME(\"stats\"),\t\t\tCTL(config_stats)},\n\t{NAME(\"tcache\"),\t\tCTL(config_tcache)},\n\t{NAME(\"tls\"),\t\t\tCTL(config_tls)},\n\t{NAME(\"utrace\"),\t\tCTL(config_utrace)},\n\t{NAME(\"valgrind\"),\t\tCTL(config_valgrind)},\n\t{NAME(\"xmalloc\"),\t\tCTL(config_xmalloc)}\n};\n\nstatic const ctl_named_node_t opt_node[] = {\n\t{NAME(\"abort\"),\t\t\tCTL(opt_abort)},\n\t{NAME(\"dss\"),\t\t\tCTL(opt_dss)},\n\t{NAME(\"lg_chunk\"),\t\tCTL(opt_lg_chunk)},\n\t{NAME(\"narenas\"),\t\tCTL(opt_narenas)},\n\t{NAME(\"lg_dirty_mult\"),\t\tCTL(opt_lg_dirty_mult)},\n\t{NAME(\"stats_print\"),\t\tCTL(opt_stats_print)},\n\t{NAME(\"junk\"),\t\t\tCTL(opt_junk)},\n\t{NAME(\"zero\"),\t\t\tCTL(opt_zero)},\n\t{NAME(\"quarantine\"),\t\tCTL(opt_quarantine)},\n\t{NAME(\"redzone\"),\t\tCTL(opt_redzone)},\n\t{NAME(\"utrace\"),\t\tCTL(opt_utrace)},\n\t{NAME(\"valgrind\"),\t\tCTL(opt_valgrind)},\n\t{NAME(\"xmalloc\"),\t\tCTL(opt_xmalloc)},\n\t{NAME(\"tcache\"),\t\tCTL(opt_tcache)},\n\t{NAME(\"lg_tcache_max\"),\t\tCTL(opt_lg_tcache_max)},\n\t{NAME(\"prof\"),\t\t\tCTL(opt_prof)},\n\t{NAME(\"prof_prefix\"),\t\tCTL(opt_prof_prefix)},\n\t{NAME(\"prof_active\"),\t\tCTL(opt_prof_active)},\n\t{NAME(\"lg_prof_sample\"),\tCTL(opt_lg_prof_sample)},\n\t{NAME(\"lg_prof_interval\"),\tCTL(opt_lg_prof_interval)},\n\t{NAME(\"prof_gdump\"),\t\tCTL(opt_prof_gdump)},\n\t{NAME(\"prof_final\"),\t\tCTL(opt_prof_final)},\n\t{NAME(\"prof_leak\"),\t\tCTL(opt_prof_leak)},\n\t{NAME(\"prof_accum\"),\t\tCTL(opt_prof_accum)}\n};\n\nstatic const ctl_named_node_t arena_i_node[] = {\n\t{NAME(\"purge\"),\t\t\tCTL(arena_i_purge)},\n\t{NAME(\"dss\"),\t\t\tCTL(arena_i_dss)}\n};\nstatic const ctl_named_node_t super_arena_i_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, arena_i)}\n};\n\nstatic const ctl_indexed_node_t arena_node[] = {\n\t{INDEX(arena_i)}\n};\n\nstatic const ctl_named_node_t arenas_bin_i_node[] = {\n\t{NAME(\"size\"),\t\t\tCTL(arenas_bin_i_size)},\n\t{NAME(\"nregs\"),\t\t\tCTL(arenas_bin_i_nregs)},\n\t{NAME(\"run_size\"),\t\tCTL(arenas_bin_i_run_size)}\n};\nstatic const ctl_named_node_t super_arenas_bin_i_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, arenas_bin_i)}\n};\n\nstatic const ctl_indexed_node_t arenas_bin_node[] = {\n\t{INDEX(arenas_bin_i)}\n};\n\nstatic const ctl_named_node_t arenas_lrun_i_node[] = {\n\t{NAME(\"size\"),\t\t\tCTL(arenas_lrun_i_size)}\n};\nstatic const ctl_named_node_t super_arenas_lrun_i_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, arenas_lrun_i)}\n};\n\nstatic const ctl_indexed_node_t arenas_lrun_node[] = {\n\t{INDEX(arenas_lrun_i)}\n};\n\nstatic const ctl_named_node_t arenas_node[] = {\n\t{NAME(\"narenas\"),\t\tCTL(arenas_narenas)},\n\t{NAME(\"initialized\"),\t\tCTL(arenas_initialized)},\n\t{NAME(\"quantum\"),\t\tCTL(arenas_quantum)},\n\t{NAME(\"page\"),\t\t\tCTL(arenas_page)},\n\t{NAME(\"tcache_max\"),\t\tCTL(arenas_tcache_max)},\n\t{NAME(\"nbins\"),\t\t\tCTL(arenas_nbins)},\n\t{NAME(\"nhbins\"),\t\tCTL(arenas_nhbins)},\n\t{NAME(\"bin\"),\t\t\tCHILD(indexed, arenas_bin)},\n\t{NAME(\"nlruns\"),\t\tCTL(arenas_nlruns)},\n\t{NAME(\"lrun\"),\t\t\tCHILD(indexed, arenas_lrun)},\n\t{NAME(\"purge\"),\t\t\tCTL(arenas_purge)},\n\t{NAME(\"extend\"),\t\tCTL(arenas_extend)}\n};\n\nstatic const ctl_named_node_t\tprof_node[] = {\n\t{NAME(\"active\"),\tCTL(prof_active)},\n\t{NAME(\"dump\"),\t\tCTL(prof_dump)},\n\t{NAME(\"interval\"),\tCTL(prof_interval)}\n};\n\nstatic const ctl_named_node_t stats_chunks_node[] = {\n\t{NAME(\"current\"),\t\tCTL(stats_chunks_current)},\n\t{NAME(\"total\"),\t\t\tCTL(stats_chunks_total)},\n\t{NAME(\"high\"),\t\t\tCTL(stats_chunks_high)}\n};\n\nstatic const ctl_named_node_t stats_huge_node[] = {\n\t{NAME(\"allocated\"),\t\tCTL(stats_huge_allocated)},\n\t{NAME(\"nmalloc\"),\t\tCTL(stats_huge_nmalloc)},\n\t{NAME(\"ndalloc\"),\t\tCTL(stats_huge_ndalloc)}\n};\n\nstatic const ctl_named_node_t stats_arenas_i_small_node[] = {\n\t{NAME(\"allocated\"),\t\tCTL(stats_arenas_i_small_allocated)},\n\t{NAME(\"nmalloc\"),\t\tCTL(stats_arenas_i_small_nmalloc)},\n\t{NAME(\"ndalloc\"),\t\tCTL(stats_arenas_i_small_ndalloc)},\n\t{NAME(\"nrequests\"),\t\tCTL(stats_arenas_i_small_nrequests)}\n};\n\nstatic const ctl_named_node_t stats_arenas_i_large_node[] = {\n\t{NAME(\"allocated\"),\t\tCTL(stats_arenas_i_large_allocated)},\n\t{NAME(\"nmalloc\"),\t\tCTL(stats_arenas_i_large_nmalloc)},\n\t{NAME(\"ndalloc\"),\t\tCTL(stats_arenas_i_large_ndalloc)},\n\t{NAME(\"nrequests\"),\t\tCTL(stats_arenas_i_large_nrequests)}\n};\n\nstatic const ctl_named_node_t stats_arenas_i_bins_j_node[] = {\n\t{NAME(\"allocated\"),\t\tCTL(stats_arenas_i_bins_j_allocated)},\n\t{NAME(\"nmalloc\"),\t\tCTL(stats_arenas_i_bins_j_nmalloc)},\n\t{NAME(\"ndalloc\"),\t\tCTL(stats_arenas_i_bins_j_ndalloc)},\n\t{NAME(\"nrequests\"),\t\tCTL(stats_arenas_i_bins_j_nrequests)},\n\t{NAME(\"nfills\"),\t\tCTL(stats_arenas_i_bins_j_nfills)},\n\t{NAME(\"nflushes\"),\t\tCTL(stats_arenas_i_bins_j_nflushes)},\n\t{NAME(\"nruns\"),\t\t\tCTL(stats_arenas_i_bins_j_nruns)},\n\t{NAME(\"nreruns\"),\t\tCTL(stats_arenas_i_bins_j_nreruns)},\n\t{NAME(\"curruns\"),\t\tCTL(stats_arenas_i_bins_j_curruns)}\n};\nstatic const ctl_named_node_t super_stats_arenas_i_bins_j_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, stats_arenas_i_bins_j)}\n};\n\nstatic const ctl_indexed_node_t stats_arenas_i_bins_node[] = {\n\t{INDEX(stats_arenas_i_bins_j)}\n};\n\nstatic const ctl_named_node_t stats_arenas_i_lruns_j_node[] = {\n\t{NAME(\"nmalloc\"),\t\tCTL(stats_arenas_i_lruns_j_nmalloc)},\n\t{NAME(\"ndalloc\"),\t\tCTL(stats_arenas_i_lruns_j_ndalloc)},\n\t{NAME(\"nrequests\"),\t\tCTL(stats_arenas_i_lruns_j_nrequests)},\n\t{NAME(\"curruns\"),\t\tCTL(stats_arenas_i_lruns_j_curruns)}\n};\nstatic const ctl_named_node_t super_stats_arenas_i_lruns_j_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, stats_arenas_i_lruns_j)}\n};\n\nstatic const ctl_indexed_node_t stats_arenas_i_lruns_node[] = {\n\t{INDEX(stats_arenas_i_lruns_j)}\n};\n\nstatic const ctl_named_node_t stats_arenas_i_node[] = {\n\t{NAME(\"nthreads\"),\t\tCTL(stats_arenas_i_nthreads)},\n\t{NAME(\"dss\"),\t\t\tCTL(stats_arenas_i_dss)},\n\t{NAME(\"pactive\"),\t\tCTL(stats_arenas_i_pactive)},\n\t{NAME(\"pdirty\"),\t\tCTL(stats_arenas_i_pdirty)},\n\t{NAME(\"mapped\"),\t\tCTL(stats_arenas_i_mapped)},\n\t{NAME(\"npurge\"),\t\tCTL(stats_arenas_i_npurge)},\n\t{NAME(\"nmadvise\"),\t\tCTL(stats_arenas_i_nmadvise)},\n\t{NAME(\"purged\"),\t\tCTL(stats_arenas_i_purged)},\n\t{NAME(\"small\"),\t\t\tCHILD(named, stats_arenas_i_small)},\n\t{NAME(\"large\"),\t\t\tCHILD(named, stats_arenas_i_large)},\n\t{NAME(\"bins\"),\t\t\tCHILD(indexed, stats_arenas_i_bins)},\n\t{NAME(\"lruns\"),\t\t\tCHILD(indexed, stats_arenas_i_lruns)}\n};\nstatic const ctl_named_node_t super_stats_arenas_i_node[] = {\n\t{NAME(\"\"),\t\t\tCHILD(named, stats_arenas_i)}\n};\n\nstatic const ctl_indexed_node_t stats_arenas_node[] = {\n\t{INDEX(stats_arenas_i)}\n};\n\nstatic const ctl_named_node_t stats_node[] = {\n\t{NAME(\"cactive\"),\t\tCTL(stats_cactive)},\n\t{NAME(\"allocated\"),\t\tCTL(stats_allocated)},\n\t{NAME(\"active\"),\t\tCTL(stats_active)},\n\t{NAME(\"mapped\"),\t\tCTL(stats_mapped)},\n\t{NAME(\"chunks\"),\t\tCHILD(named, stats_chunks)},\n\t{NAME(\"huge\"),\t\t\tCHILD(named, stats_huge)},\n\t{NAME(\"arenas\"),\t\tCHILD(indexed, stats_arenas)}\n};\n\nstatic const ctl_named_node_t\troot_node[] = {\n\t{NAME(\"version\"),\tCTL(version)},\n\t{NAME(\"epoch\"),\t\tCTL(epoch)},\n\t{NAME(\"thread\"),\tCHILD(named, thread)},\n\t{NAME(\"config\"),\tCHILD(named, config)},\n\t{NAME(\"opt\"),\t\tCHILD(named, opt)},\n\t{NAME(\"arena\"),\t\tCHILD(indexed, arena)},\n\t{NAME(\"arenas\"),\tCHILD(named, arenas)},\n\t{NAME(\"prof\"),\t\tCHILD(named, prof)},\n\t{NAME(\"stats\"),\t\tCHILD(named, stats)}\n};\nstatic const ctl_named_node_t super_root_node[] = {\n\t{NAME(\"\"),\t\tCHILD(named, root)}\n};\n\n#undef NAME\n#undef CHILD\n#undef CTL\n#undef INDEX\n\n/******************************************************************************/\n\nstatic bool\nctl_arena_init(ctl_arena_stats_t *astats)\n{\n\n\tif (astats->lstats == NULL) {\n\t\tastats->lstats = (malloc_large_stats_t *)base_alloc(nlclasses *\n\t\t sizeof(malloc_large_stats_t));\n\t\tif (astats->lstats == NULL)\n\t\t\treturn (true);\n\t}\n\n\treturn (false);\n}\n\nstatic void\nctl_arena_clear(ctl_arena_stats_t *astats)\n{\n\n\tastats->dss = dss_prec_names[dss_prec_limit];\n\tastats->pactive = 0;\n\tastats->pdirty = 0;\n\tif (config_stats) {\n\t\tmemset(&astats->astats, 0, sizeof(arena_stats_t));\n\t\tastats->allocated_small = 0;\n\t\tastats->nmalloc_small = 0;\n\t\tastats->ndalloc_small = 0;\n\t\tastats->nrequests_small = 0;\n\t\tmemset(astats->bstats, 0, NBINS * sizeof(malloc_bin_stats_t));\n\t\tmemset(astats->lstats, 0, nlclasses *\n\t\t sizeof(malloc_large_stats_t));\n\t}\n}\n\nstatic void\nctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena)\n{\n\tunsigned i;\n\n\tarena_stats_merge(arena, &cstats->dss, &cstats->pactive,\n\t &cstats->pdirty, &cstats->astats, cstats->bstats, cstats->lstats);\n\n\tfor (i = 0; i < NBINS; i++) {\n\t\tcstats->allocated_small += cstats->bstats[i].allocated;\n\t\tcstats->nmalloc_small += cstats->bstats[i].nmalloc;\n\t\tcstats->ndalloc_small += cstats->bstats[i].ndalloc;\n\t\tcstats->nrequests_small += cstats->bstats[i].nrequests;\n\t}\n}\n\nstatic void\nctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats)\n{\n\tunsigned i;\n\n\tsstats->pactive += astats->pactive;\n\tsstats->pdirty += astats->pdirty;\n\n\tsstats->astats.mapped += astats->astats.mapped;\n\tsstats->astats.npurge += astats->astats.npurge;\n\tsstats->astats.nmadvise += astats->astats.nmadvise;\n\tsstats->astats.purged += astats->astats.purged;\n\n\tsstats->allocated_small += astats->allocated_small;\n\tsstats->nmalloc_small += astats->nmalloc_small;\n\tsstats->ndalloc_small += astats->ndalloc_small;\n\tsstats->nrequests_small += astats->nrequests_small;\n\n\tsstats->astats.allocated_large += astats->astats.allocated_large;\n\tsstats->astats.nmalloc_large += astats->astats.nmalloc_large;\n\tsstats->astats.ndalloc_large += astats->astats.ndalloc_large;\n\tsstats->astats.nrequests_large += astats->astats.nrequests_large;\n\n\tfor (i = 0; i < nlclasses; i++) {\n\t\tsstats->lstats[i].nmalloc += astats->lstats[i].nmalloc;\n\t\tsstats->lstats[i].ndalloc += astats->lstats[i].ndalloc;\n\t\tsstats->lstats[i].nrequests += astats->lstats[i].nrequests;\n\t\tsstats->lstats[i].curruns += astats->lstats[i].curruns;\n\t}\n\n\tfor (i = 0; i < NBINS; i++) {\n\t\tsstats->bstats[i].allocated += astats->bstats[i].allocated;\n\t\tsstats->bstats[i].nmalloc += astats->bstats[i].nmalloc;\n\t\tsstats->bstats[i].ndalloc += astats->bstats[i].ndalloc;\n\t\tsstats->bstats[i].nrequests += astats->bstats[i].nrequests;\n\t\tif (config_tcache) {\n\t\t\tsstats->bstats[i].nfills += astats->bstats[i].nfills;\n\t\t\tsstats->bstats[i].nflushes +=\n\t\t\t astats->bstats[i].nflushes;\n\t\t}\n\t\tsstats->bstats[i].nruns += astats->bstats[i].nruns;\n\t\tsstats->bstats[i].reruns += astats->bstats[i].reruns;\n\t\tsstats->bstats[i].curruns += astats->bstats[i].curruns;\n\t}\n}\n\nstatic void\nctl_arena_refresh(arena_t *arena, unsigned i)\n{\n\tctl_arena_stats_t *astats = &ctl_stats.arenas[i];\n\tctl_arena_stats_t *sstats = &ctl_stats.arenas[ctl_stats.narenas];\n\n\tctl_arena_clear(astats);\n\n\tsstats->nthreads += astats->nthreads;\n\tif (config_stats) {\n\t\tctl_arena_stats_amerge(astats, arena);\n\t\t/* Merge into sum stats as well. */\n\t\tctl_arena_stats_smerge(sstats, astats);\n\t} else {\n\t\tastats->pactive += arena->nactive;\n\t\tastats->pdirty += arena->ndirty;\n\t\t/* Merge into sum stats as well. */\n\t\tsstats->pactive += arena->nactive;\n\t\tsstats->pdirty += arena->ndirty;\n\t}\n}\n\nstatic bool\nctl_grow(void)\n{\n\tsize_t astats_size;\n\tctl_arena_stats_t *astats;\n\tarena_t **tarenas;\n\n\t/* Extend arena stats and arenas arrays. */\n\tastats_size = (ctl_stats.narenas + 2) * sizeof(ctl_arena_stats_t);\n\tif (ctl_stats.narenas == narenas_auto) {\n\t\t/* ctl_stats.arenas and arenas came from base_alloc(). */\n\t\tastats = (ctl_arena_stats_t *)imalloc(astats_size);\n\t\tif (astats == NULL)\n\t\t\treturn (true);\n\t\tmemcpy(astats, ctl_stats.arenas, (ctl_stats.narenas + 1) *\n\t\t sizeof(ctl_arena_stats_t));\n\n\t\ttarenas = (arena_t **)imalloc((ctl_stats.narenas + 1) *\n\t\t sizeof(arena_t *));\n\t\tif (tarenas == NULL) {\n\t\t\tidalloc(astats);\n\t\t\treturn (true);\n\t\t}\n\t\tmemcpy(tarenas, arenas, ctl_stats.narenas * sizeof(arena_t *));\n\t} else {\n\t\tastats = (ctl_arena_stats_t *)iralloc(ctl_stats.arenas,\n\t\t astats_size, 0, 0, false, false);\n\t\tif (astats == NULL)\n\t\t\treturn (true);\n\n\t\ttarenas = (arena_t **)iralloc(arenas, (ctl_stats.narenas + 1) *\n\t\t sizeof(arena_t *), 0, 0, false, false);\n\t\tif (tarenas == NULL)\n\t\t\treturn (true);\n\t}\n\t/* Initialize the new astats and arenas elements. */\n\tmemset(&astats[ctl_stats.narenas + 1], 0, sizeof(ctl_arena_stats_t));\n\tif (ctl_arena_init(&astats[ctl_stats.narenas + 1]))\n\t\treturn (true);\n\ttarenas[ctl_stats.narenas] = NULL;\n\t/* Swap merged stats to their new location. */\n\t{\n\t\tctl_arena_stats_t tstats;\n\t\tmemcpy(&tstats, &astats[ctl_stats.narenas],\n\t\t sizeof(ctl_arena_stats_t));\n\t\tmemcpy(&astats[ctl_stats.narenas],\n\t\t &astats[ctl_stats.narenas + 1], sizeof(ctl_arena_stats_t));\n\t\tmemcpy(&astats[ctl_stats.narenas + 1], &tstats,\n\t\t sizeof(ctl_arena_stats_t));\n\t}\n\tctl_stats.arenas = astats;\n\tctl_stats.narenas++;\n\tmalloc_mutex_lock(&arenas_lock);\n\tarenas = tarenas;\n\tnarenas_total++;\n\tarenas_extend(narenas_total - 1);\n\tmalloc_mutex_unlock(&arenas_lock);\n\n\treturn (false);\n}\n\nstatic void\nctl_refresh(void)\n{\n\tunsigned i;\n\tVARIABLE_ARRAY(arena_t *, tarenas, ctl_stats.narenas);\n\n\tif (config_stats) {\n\t\tmalloc_mutex_lock(&chunks_mtx);\n\t\tctl_stats.chunks.current = stats_chunks.curchunks;\n\t\tctl_stats.chunks.total = stats_chunks.nchunks;\n\t\tctl_stats.chunks.high = stats_chunks.highchunks;\n\t\tmalloc_mutex_unlock(&chunks_mtx);\n\n\t\tmalloc_mutex_lock(&huge_mtx);\n\t\tctl_stats.huge.allocated = huge_allocated;\n\t\tctl_stats.huge.nmalloc = huge_nmalloc;\n\t\tctl_stats.huge.ndalloc = huge_ndalloc;\n\t\tmalloc_mutex_unlock(&huge_mtx);\n\t}\n\n\t/*\n\t * Clear sum stats, since they will be merged into by\n\t * ctl_arena_refresh().\n\t */\n\tctl_stats.arenas[ctl_stats.narenas].nthreads = 0;\n\tctl_arena_clear(&ctl_stats.arenas[ctl_stats.narenas]);\n\n\tmalloc_mutex_lock(&arenas_lock);\n\tmemcpy(tarenas, arenas, sizeof(arena_t *) * ctl_stats.narenas);\n\tfor (i = 0; i < ctl_stats.narenas; i++) {\n\t\tif (arenas[i] != NULL)\n\t\t\tctl_stats.arenas[i].nthreads = arenas[i]->nthreads;\n\t\telse\n\t\t\tctl_stats.arenas[i].nthreads = 0;\n\t}\n\tmalloc_mutex_unlock(&arenas_lock);\n\tfor (i = 0; i < ctl_stats.narenas; i++) {\n\t\tbool initialized = (tarenas[i] != NULL);\n\n\t\tctl_stats.arenas[i].initialized = initialized;\n\t\tif (initialized)\n\t\t\tctl_arena_refresh(tarenas[i], i);\n\t}\n\n\tif (config_stats) {\n\t\tctl_stats.allocated =\n\t\t ctl_stats.arenas[ctl_stats.narenas].allocated_small\n\t\t + ctl_stats.arenas[ctl_stats.narenas].astats.allocated_large\n\t\t + ctl_stats.huge.allocated;\n\t\tctl_stats.active =\n\t\t (ctl_stats.arenas[ctl_stats.narenas].pactive << LG_PAGE)\n\t\t + ctl_stats.huge.allocated;\n\t\tctl_stats.mapped = (ctl_stats.chunks.current << opt_lg_chunk);\n\t}\n\n\tctl_epoch++;\n}\n\nstatic bool\nctl_init(void)\n{\n\tbool ret;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tif (ctl_initialized == false) {\n\t\t/*\n\t\t * Allocate space for one extra arena stats element, which\n\t\t * contains summed stats across all arenas.\n\t\t */\n\t\tassert(narenas_auto == narenas_total_get());\n\t\tctl_stats.narenas = narenas_auto;\n\t\tctl_stats.arenas = (ctl_arena_stats_t *)base_alloc(\n\t\t (ctl_stats.narenas + 1) * sizeof(ctl_arena_stats_t));\n\t\tif (ctl_stats.arenas == NULL) {\n\t\t\tret = true;\n\t\t\tgoto label_return;\n\t\t}\n\t\tmemset(ctl_stats.arenas, 0, (ctl_stats.narenas + 1) *\n\t\t sizeof(ctl_arena_stats_t));\n\n\t\t/*\n\t\t * Initialize all stats structures, regardless of whether they\n\t\t * ever get used. Lazy initialization would allow errors to\n\t\t * cause inconsistent state to be viewable by the application.\n\t\t */\n\t\tif (config_stats) {\n\t\t\tunsigned i;\n\t\t\tfor (i = 0; i <= ctl_stats.narenas; i++) {\n\t\t\t\tif (ctl_arena_init(&ctl_stats.arenas[i])) {\n\t\t\t\t\tret = true;\n\t\t\t\t\tgoto label_return;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t\tctl_stats.arenas[ctl_stats.narenas].initialized = true;\n\n\t\tctl_epoch = 0;\n\t\tctl_refresh();\n\t\tctl_initialized = true;\n\t}\n\n\tret = false;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic int\nctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp,\n size_t *depthp)\n{\n\tint ret;\n\tconst char *elm, *tdot, *dot;\n\tsize_t elen, i, j;\n\tconst ctl_named_node_t *node;\n\n\telm = name;\n\t/* Equivalent to strchrnul(). */\n\tdot = ((tdot = strchr(elm, '.')) != NULL) ? tdot : strchr(elm, '\\0');\n\telen = (size_t)((uintptr_t)dot - (uintptr_t)elm);\n\tif (elen == 0) {\n\t\tret = ENOENT;\n\t\tgoto label_return;\n\t}\n\tnode = super_root_node;\n\tfor (i = 0; i < *depthp; i++) {\n\t\tassert(node);\n\t\tassert(node->nchildren > 0);\n\t\tif (ctl_named_node(node->children) != NULL) {\n\t\t\tconst ctl_named_node_t *pnode = node;\n\n\t\t\t/* Children are named. */\n\t\t\tfor (j = 0; j < node->nchildren; j++) {\n\t\t\t\tconst ctl_named_node_t *child =\n\t\t\t\t ctl_named_children(node, j);\n\t\t\t\tif (strlen(child->name) == elen &&\n\t\t\t\t strncmp(elm, child->name, elen) == 0) {\n\t\t\t\t\tnode = child;\n\t\t\t\t\tif (nodesp != NULL)\n\t\t\t\t\t\tnodesp[i] =\n\t\t\t\t\t\t (const ctl_node_t *)node;\n\t\t\t\t\tmibp[i] = j;\n\t\t\t\t\tbreak;\n\t\t\t\t}\n\t\t\t}\n\t\t\tif (node == pnode) {\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\t\t} else {\n\t\t\tuintmax_t index;\n\t\t\tconst ctl_indexed_node_t *inode;\n\n\t\t\t/* Children are indexed. */\n\t\t\tindex = malloc_strtoumax(elm, NULL, 10);\n\t\t\tif (index == UINTMAX_MAX || index > SIZE_T_MAX) {\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\n\t\t\tinode = ctl_indexed_node(node->children);\n\t\t\tnode = inode->index(mibp, *depthp, (size_t)index);\n\t\t\tif (node == NULL) {\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\n\t\t\tif (nodesp != NULL)\n\t\t\t\tnodesp[i] = (const ctl_node_t *)node;\n\t\t\tmibp[i] = (size_t)index;\n\t\t}\n\n\t\tif (node->ctl != NULL) {\n\t\t\t/* Terminal node. */\n\t\t\tif (*dot != '\\0') {\n\t\t\t\t/*\n\t\t\t\t * The name contains more elements than are\n\t\t\t\t * in this path through the tree.\n\t\t\t\t */\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\t\t\t/* Complete lookup successful. */\n\t\t\t*depthp = i + 1;\n\t\t\tbreak;\n\t\t}\n\n\t\t/* Update elm. */\n\t\tif (*dot == '\\0') {\n\t\t\t/* No more elements. */\n\t\t\tret = ENOENT;\n\t\t\tgoto label_return;\n\t\t}\n\t\telm = &dot[1];\n\t\tdot = ((tdot = strchr(elm, '.')) != NULL) ? tdot :\n\t\t strchr(elm, '\\0');\n\t\telen = (size_t)((uintptr_t)dot - (uintptr_t)elm);\n\t}\n\n\tret = 0;\nlabel_return:\n\treturn (ret);\n}\n\nint\nctl_byname(const char *name, void *oldp, size_t *oldlenp, void *newp,\n size_t newlen)\n{\n\tint ret;\n\tsize_t depth;\n\tctl_node_t const *nodes[CTL_MAX_DEPTH];\n\tsize_t mib[CTL_MAX_DEPTH];\n\tconst ctl_named_node_t *node;\n\n\tif (ctl_initialized == false && ctl_init()) {\n\t\tret = EAGAIN;\n\t\tgoto label_return;\n\t}\n\n\tdepth = CTL_MAX_DEPTH;\n\tret = ctl_lookup(name, nodes, mib, &depth);\n\tif (ret != 0)\n\t\tgoto label_return;\n\n\tnode = ctl_named_node(nodes[depth-1]);\n\tif (node != NULL && node->ctl)\n\t\tret = node->ctl(mib, depth, oldp, oldlenp, newp, newlen);\n\telse {\n\t\t/* The name refers to a partial path through the ctl tree. */\n\t\tret = ENOENT;\n\t}\n\nlabel_return:\n\treturn(ret);\n}\n\nint\nctl_nametomib(const char *name, size_t *mibp, size_t *miblenp)\n{\n\tint ret;\n\n\tif (ctl_initialized == false && ctl_init()) {\n\t\tret = EAGAIN;\n\t\tgoto label_return;\n\t}\n\n\tret = ctl_lookup(name, NULL, mibp, miblenp);\nlabel_return:\n\treturn(ret);\n}\n\nint\nctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tconst ctl_named_node_t *node;\n\tsize_t i;\n\n\tif (ctl_initialized == false && ctl_init()) {\n\t\tret = EAGAIN;\n\t\tgoto label_return;\n\t}\n\n\t/* Iterate down the tree. */\n\tnode = super_root_node;\n\tfor (i = 0; i < miblen; i++) {\n\t\tassert(node);\n\t\tassert(node->nchildren > 0);\n\t\tif (ctl_named_node(node->children) != NULL) {\n\t\t\t/* Children are named. */\n\t\t\tif (node->nchildren <= mib[i]) {\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\t\t\tnode = ctl_named_children(node, mib[i]);\n\t\t} else {\n\t\t\tconst ctl_indexed_node_t *inode;\n\n\t\t\t/* Indexed element. */\n\t\t\tinode = ctl_indexed_node(node->children);\n\t\t\tnode = inode->index(mib, miblen, mib[i]);\n\t\t\tif (node == NULL) {\n\t\t\t\tret = ENOENT;\n\t\t\t\tgoto label_return;\n\t\t\t}\n\t\t}\n\t}\n\n\t/* Call the ctl function. */\n\tif (node && node->ctl)\n\t\tret = node->ctl(mib, miblen, oldp, oldlenp, newp, newlen);\n\telse {\n\t\t/* Partial MIB. */\n\t\tret = ENOENT;\n\t}\n\nlabel_return:\n\treturn(ret);\n}\n\nbool\nctl_boot(void)\n{\n\n\tif (malloc_mutex_init(&ctl_mtx))\n\t\treturn (true);\n\n\tctl_initialized = false;\n\n\treturn (false);\n}\n\nvoid\nctl_prefork(void)\n{\n\n\tmalloc_mutex_lock(&ctl_mtx);\n}\n\nvoid\nctl_postfork_parent(void)\n{\n\n\tmalloc_mutex_postfork_parent(&ctl_mtx);\n}\n\nvoid\nctl_postfork_child(void)\n{\n\n\tmalloc_mutex_postfork_child(&ctl_mtx);\n}\n\n/******************************************************************************/\n/* *_ctl() functions. */\n\n#define\tREADONLY()\tdo {\t\t\t\t\t\t\\\n\tif (newp != NULL || newlen != 0) {\t\t\t\t\\\n\t\tret = EPERM;\t\t\t\t\t\t\\\n\t\tgoto label_return;\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tWRITEONLY()\tdo {\t\t\t\t\t\t\\\n\tif (oldp != NULL || oldlenp != NULL) {\t\t\t\t\\\n\t\tret = EPERM;\t\t\t\t\t\t\\\n\t\tgoto label_return;\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tREAD(v, t)\tdo {\t\t\t\t\t\t\\\n\tif (oldp != NULL && oldlenp != NULL) {\t\t\t\t\\\n\t\tif (*oldlenp != sizeof(t)) {\t\t\t\t\\\n\t\t\tsize_t\tcopylen = (sizeof(t) <= *oldlenp)\t\\\n\t\t\t ? sizeof(t) : *oldlenp;\t\t\t\\\n\t\t\tmemcpy(oldp, (void *)&v, copylen);\t\t\\\n\t\t\tret = EINVAL;\t\t\t\t\t\\\n\t\t\tgoto label_return;\t\t\t\t\\\n\t\t} else\t\t\t\t\t\t\t\\\n\t\t\t*(t *)oldp = v;\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tWRITE(v, t)\tdo {\t\t\t\t\t\t\\\n\tif (newp != NULL) {\t\t\t\t\t\t\\\n\t\tif (newlen != sizeof(t)) {\t\t\t\t\\\n\t\t\tret = EINVAL;\t\t\t\t\t\\\n\t\t\tgoto label_return;\t\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t\tv = *(t *)newp;\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n/*\n * There's a lot of code duplication in the following macros due to limitations\n * in how nested cpp macros are expanded.\n */\n#define\tCTL_RO_CLGEN(c, l, n, v, t)\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tt oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif ((c) == false)\t\t\t\t\t\t\\\n\t\treturn (ENOENT);\t\t\t\t\t\\\n\tif (l)\t\t\t\t\t\t\t\t\\\n\t\tmalloc_mutex_lock(&ctl_mtx);\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = v;\t\t\t\t\t\t\t\\\n\tREAD(oldval, t);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\tif (l)\t\t\t\t\t\t\t\t\\\n\t\tmalloc_mutex_unlock(&ctl_mtx);\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n#define\tCTL_RO_CGEN(c, n, v, t)\t\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tt oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif ((c) == false)\t\t\t\t\t\t\\\n\t\treturn (ENOENT);\t\t\t\t\t\\\n\tmalloc_mutex_lock(&ctl_mtx);\t\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = v;\t\t\t\t\t\t\t\\\n\tREAD(oldval, t);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\tmalloc_mutex_unlock(&ctl_mtx);\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n#define\tCTL_RO_GEN(n, v, t)\t\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tt oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tmalloc_mutex_lock(&ctl_mtx);\t\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = v;\t\t\t\t\t\t\t\\\n\tREAD(oldval, t);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\tmalloc_mutex_unlock(&ctl_mtx);\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n/*\n * ctl_mtx is not acquired, under the assumption that no pertinent data will\n * mutate during the call.\n */\n#define\tCTL_RO_NL_CGEN(c, n, v, t)\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tt oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tif ((c) == false)\t\t\t\t\t\t\\\n\t\treturn (ENOENT);\t\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = v;\t\t\t\t\t\t\t\\\n\tREAD(oldval, t);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n#define\tCTL_RO_NL_GEN(n, v, t)\t\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tt oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = v;\t\t\t\t\t\t\t\\\n\tREAD(oldval, t);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\n#define\tCTL_RO_BOOL_CONFIG_GEN(n)\t\t\t\t\t\\\nstatic int\t\t\t\t\t\t\t\t\\\nn##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\t\\\n void *newp, size_t newlen)\t\t\t\t\t\t\\\n{\t\t\t\t\t\t\t\t\t\\\n\tint ret;\t\t\t\t\t\t\t\\\n\tbool oldval;\t\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tREADONLY();\t\t\t\t\t\t\t\\\n\toldval = n;\t\t\t\t\t\t\t\\\n\tREAD(oldval, bool);\t\t\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\tret = 0;\t\t\t\t\t\t\t\\\nlabel_return:\t\t\t\t\t\t\t\t\\\n\treturn (ret);\t\t\t\t\t\t\t\\\n}\n\nCTL_RO_NL_GEN(version, JEMALLOC_VERSION, const char *)\n\nstatic int\nepoch_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tuint64_t newval;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tWRITE(newval, uint64_t);\n\tif (newp != NULL)\n\t\tctl_refresh();\n\tREAD(ctl_epoch, uint64_t);\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic int\nthread_tcache_enabled_ctl(const size_t *mib, size_t miblen, void *oldp,\n size_t *oldlenp, void *newp, size_t newlen)\n{\n\tint ret;\n\tbool oldval;\n\n\tif (config_tcache == false)\n\t\treturn (ENOENT);\n\n\toldval = tcache_enabled_get();\n\tif (newp != NULL) {\n\t\tif (newlen != sizeof(bool)) {\n\t\t\tret = EINVAL;\n\t\t\tgoto label_return;\n\t\t}\n\t\ttcache_enabled_set(*(bool *)newp);\n\t}\n\tREAD(oldval, bool);\n\n\tret = 0;\nlabel_return:\n\treturn (ret);\n}\n\nstatic int\nthread_tcache_flush_ctl(const size_t *mib, size_t miblen, void *oldp,\n size_t *oldlenp, void *newp, size_t newlen)\n{\n\tint ret;\n\n\tif (config_tcache == false)\n\t\treturn (ENOENT);\n\n\tREADONLY();\n\tWRITEONLY();\n\n\ttcache_flush();\n\n\tret = 0;\nlabel_return:\n\treturn (ret);\n}\n\nstatic int\nthread_arena_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tunsigned newind, oldind;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tnewind = oldind = choose_arena(NULL)->ind;\n\tWRITE(newind, unsigned);\n\tREAD(oldind, unsigned);\n\tif (newind != oldind) {\n\t\tarena_t *arena;\n\n\t\tif (newind >= ctl_stats.narenas) {\n\t\t\t/* New arena index is out of range. */\n\t\t\tret = EFAULT;\n\t\t\tgoto label_return;\n\t\t}\n\n\t\t/* Initialize arena if necessary. */\n\t\tmalloc_mutex_lock(&arenas_lock);\n\t\tif ((arena = arenas[newind]) == NULL && (arena =\n\t\t arenas_extend(newind)) == NULL) {\n\t\t\tmalloc_mutex_unlock(&arenas_lock);\n\t\t\tret = EAGAIN;\n\t\t\tgoto label_return;\n\t\t}\n\t\tassert(arena == arenas[newind]);\n\t\tarenas[oldind]->nthreads--;\n\t\tarenas[newind]->nthreads++;\n\t\tmalloc_mutex_unlock(&arenas_lock);\n\n\t\t/* Set new arena association. */\n\t\tif (config_tcache) {\n\t\t\ttcache_t *tcache;\n\t\t\tif ((uintptr_t)(tcache = *tcache_tsd_get()) >\n\t\t\t (uintptr_t)TCACHE_STATE_MAX) {\n\t\t\t\ttcache_arena_dissociate(tcache);\n\t\t\t\ttcache_arena_associate(tcache, arena);\n\t\t\t}\n\t\t}\n\t\tarenas_tsd_set(&arena);\n\t}\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nCTL_RO_NL_CGEN(config_stats, thread_allocated,\n thread_allocated_tsd_get()->allocated, uint64_t)\nCTL_RO_NL_CGEN(config_stats, thread_allocatedp,\n &thread_allocated_tsd_get()->allocated, uint64_t *)\nCTL_RO_NL_CGEN(config_stats, thread_deallocated,\n thread_allocated_tsd_get()->deallocated, uint64_t)\nCTL_RO_NL_CGEN(config_stats, thread_deallocatedp,\n &thread_allocated_tsd_get()->deallocated, uint64_t *)\n\n/******************************************************************************/\n\nCTL_RO_BOOL_CONFIG_GEN(config_debug)\nCTL_RO_BOOL_CONFIG_GEN(config_dss)\nCTL_RO_BOOL_CONFIG_GEN(config_fill)\nCTL_RO_BOOL_CONFIG_GEN(config_lazy_lock)\nCTL_RO_BOOL_CONFIG_GEN(config_mremap)\nCTL_RO_BOOL_CONFIG_GEN(config_munmap)\nCTL_RO_BOOL_CONFIG_GEN(config_prof)\nCTL_RO_BOOL_CONFIG_GEN(config_prof_libgcc)\nCTL_RO_BOOL_CONFIG_GEN(config_prof_libunwind)\nCTL_RO_BOOL_CONFIG_GEN(config_stats)\nCTL_RO_BOOL_CONFIG_GEN(config_tcache)\nCTL_RO_BOOL_CONFIG_GEN(config_tls)\nCTL_RO_BOOL_CONFIG_GEN(config_utrace)\nCTL_RO_BOOL_CONFIG_GEN(config_valgrind)\nCTL_RO_BOOL_CONFIG_GEN(config_xmalloc)\n\n/******************************************************************************/\n\nCTL_RO_NL_GEN(opt_abort, opt_abort, bool)\nCTL_RO_NL_GEN(opt_dss, opt_dss, const char *)\nCTL_RO_NL_GEN(opt_lg_chunk, opt_lg_chunk, size_t)\nCTL_RO_NL_GEN(opt_narenas, opt_narenas, size_t)\nCTL_RO_NL_GEN(opt_lg_dirty_mult, opt_lg_dirty_mult, ssize_t)\nCTL_RO_NL_GEN(opt_stats_print, opt_stats_print, bool)\nCTL_RO_NL_CGEN(config_fill, opt_junk, opt_junk, bool)\nCTL_RO_NL_CGEN(config_fill, opt_zero, opt_zero, bool)\nCTL_RO_NL_CGEN(config_fill, opt_quarantine, opt_quarantine, size_t)\nCTL_RO_NL_CGEN(config_fill, opt_redzone, opt_redzone, bool)\nCTL_RO_NL_CGEN(config_utrace, opt_utrace, opt_utrace, bool)\nCTL_RO_NL_CGEN(config_valgrind, opt_valgrind, opt_valgrind, bool)\nCTL_RO_NL_CGEN(config_xmalloc, opt_xmalloc, opt_xmalloc, bool)\nCTL_RO_NL_CGEN(config_tcache, opt_tcache, opt_tcache, bool)\nCTL_RO_NL_CGEN(config_tcache, opt_lg_tcache_max, opt_lg_tcache_max, ssize_t)\nCTL_RO_NL_CGEN(config_prof, opt_prof, opt_prof, bool)\nCTL_RO_NL_CGEN(config_prof, opt_prof_prefix, opt_prof_prefix, const char *)\nCTL_RO_CGEN(config_prof, opt_prof_active, opt_prof_active, bool) /* Mutable. */\nCTL_RO_NL_CGEN(config_prof, opt_lg_prof_sample, opt_lg_prof_sample, size_t)\nCTL_RO_NL_CGEN(config_prof, opt_lg_prof_interval, opt_lg_prof_interval, ssize_t)\nCTL_RO_NL_CGEN(config_prof, opt_prof_gdump, opt_prof_gdump, bool)\nCTL_RO_NL_CGEN(config_prof, opt_prof_final, opt_prof_final, bool)\nCTL_RO_NL_CGEN(config_prof, opt_prof_leak, opt_prof_leak, bool)\nCTL_RO_NL_CGEN(config_prof, opt_prof_accum, opt_prof_accum, bool)\n\n/******************************************************************************/\n\n/* ctl_mutex must be held during execution of this function. */\nstatic void\narena_purge(unsigned arena_ind)\n{\n\tVARIABLE_ARRAY(arena_t *, tarenas, ctl_stats.narenas);\n\n\tmalloc_mutex_lock(&arenas_lock);\n\tmemcpy(tarenas, arenas, sizeof(arena_t *) * ctl_stats.narenas);\n\tmalloc_mutex_unlock(&arenas_lock);\n\n\tif (arena_ind == ctl_stats.narenas) {\n\t\tunsigned i;\n\t\tfor (i = 0; i < ctl_stats.narenas; i++) {\n\t\t\tif (tarenas[i] != NULL)\n\t\t\t\tarena_purge_all(tarenas[i]);\n\t\t}\n\t} else {\n\t\tassert(arena_ind < ctl_stats.narenas);\n\t\tif (tarenas[arena_ind] != NULL)\n\t\t\tarena_purge_all(tarenas[arena_ind]);\n\t}\n}\n\nstatic int\narena_i_purge_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\n\tREADONLY();\n\tWRITEONLY();\n\tmalloc_mutex_lock(&ctl_mtx);\n\tarena_purge(mib[1]);\n\tmalloc_mutex_unlock(&ctl_mtx);\n\n\tret = 0;\nlabel_return:\n\treturn (ret);\n}\n\nstatic int\narena_i_dss_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret, i;\n\tbool match, err;\n\tconst char *dss;\n\tunsigned arena_ind = mib[1];\n\tdss_prec_t dss_prec_old = dss_prec_limit;\n\tdss_prec_t dss_prec = dss_prec_limit;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tWRITE(dss, const char *);\n\tmatch = false;\n\tfor (i = 0; i < dss_prec_limit; i++) {\n\t\tif (strcmp(dss_prec_names[i], dss) == 0) {\n\t\t\tdss_prec = i;\n\t\t\tmatch = true;\n\t\t\tbreak;\n\t\t}\n\t}\n\tif (match == false) {\n\t\tret = EINVAL;\n\t\tgoto label_return;\n\t}\n\n\tif (arena_ind < ctl_stats.narenas) {\n\t\tarena_t *arena = arenas[arena_ind];\n\t\tif (arena != NULL) {\n\t\t\tdss_prec_old = arena_dss_prec_get(arena);\n\t\t\tarena_dss_prec_set(arena, dss_prec);\n\t\t\terr = false;\n\t\t} else\n\t\t\terr = true;\n\t} else {\n\t\tdss_prec_old = chunk_dss_prec_get();\n\t\terr = chunk_dss_prec_set(dss_prec);\n\t}\n\tdss = dss_prec_names[dss_prec_old];\n\tREAD(dss, const char *);\n\tif (err) {\n\t\tret = EFAULT;\n\t\tgoto label_return;\n\t}\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic const ctl_named_node_t *\narena_i_index(const size_t *mib, size_t miblen, size_t i)\n{\n\tconst ctl_named_node_t * ret;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tif (i > ctl_stats.narenas) {\n\t\tret = NULL;\n\t\tgoto label_return;\n\t}\n\n\tret = super_arena_i_node;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\n\n/******************************************************************************/\n\nCTL_RO_NL_GEN(arenas_bin_i_size, arena_bin_info[mib[2]].reg_size, size_t)\nCTL_RO_NL_GEN(arenas_bin_i_nregs, arena_bin_info[mib[2]].nregs, uint32_t)\nCTL_RO_NL_GEN(arenas_bin_i_run_size, arena_bin_info[mib[2]].run_size, size_t)\nstatic const ctl_named_node_t *\narenas_bin_i_index(const size_t *mib, size_t miblen, size_t i)\n{\n\n\tif (i > NBINS)\n\t\treturn (NULL);\n\treturn (super_arenas_bin_i_node);\n}\n\nCTL_RO_NL_GEN(arenas_lrun_i_size, ((mib[2]+1) << LG_PAGE), size_t)\nstatic const ctl_named_node_t *\narenas_lrun_i_index(const size_t *mib, size_t miblen, size_t i)\n{\n\n\tif (i > nlclasses)\n\t\treturn (NULL);\n\treturn (super_arenas_lrun_i_node);\n}\n\nstatic int\narenas_narenas_ctl(const size_t *mib, size_t miblen, void *oldp,\n size_t *oldlenp, void *newp, size_t newlen)\n{\n\tint ret;\n\tunsigned narenas;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tREADONLY();\n\tif (*oldlenp != sizeof(unsigned)) {\n\t\tret = EINVAL;\n\t\tgoto label_return;\n\t}\n\tnarenas = ctl_stats.narenas;\n\tREAD(narenas, unsigned);\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic int\narenas_initialized_ctl(const size_t *mib, size_t miblen, void *oldp,\n size_t *oldlenp, void *newp, size_t newlen)\n{\n\tint ret;\n\tunsigned nread, i;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tREADONLY();\n\tif (*oldlenp != ctl_stats.narenas * sizeof(bool)) {\n\t\tret = EINVAL;\n\t\tnread = (*oldlenp < ctl_stats.narenas * sizeof(bool))\n\t\t ? (*oldlenp / sizeof(bool)) : ctl_stats.narenas;\n\t} else {\n\t\tret = 0;\n\t\tnread = ctl_stats.narenas;\n\t}\n\n\tfor (i = 0; i < nread; i++)\n\t\t((bool *)oldp)[i] = ctl_stats.arenas[i].initialized;\n\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nCTL_RO_NL_GEN(arenas_quantum, QUANTUM, size_t)\nCTL_RO_NL_GEN(arenas_page, PAGE, size_t)\nCTL_RO_NL_CGEN(config_tcache, arenas_tcache_max, tcache_maxclass, size_t)\nCTL_RO_NL_GEN(arenas_nbins, NBINS, unsigned)\nCTL_RO_NL_CGEN(config_tcache, arenas_nhbins, nhbins, unsigned)\nCTL_RO_NL_GEN(arenas_nlruns, nlclasses, size_t)\n\nstatic int\narenas_purge_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tunsigned arena_ind;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tWRITEONLY();\n\tarena_ind = UINT_MAX;\n\tWRITE(arena_ind, unsigned);\n\tif (newp != NULL && arena_ind >= ctl_stats.narenas)\n\t\tret = EFAULT;\n\telse {\n\t\tif (arena_ind == UINT_MAX)\n\t\t\tarena_ind = ctl_stats.narenas;\n\t\tarena_purge(arena_ind);\n\t\tret = 0;\n\t}\n\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic int\narenas_extend_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tREADONLY();\n\tif (ctl_grow()) {\n\t\tret = EAGAIN;\n\t\tgoto label_return;\n\t}\n\tREAD(ctl_stats.narenas - 1, unsigned);\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\n/******************************************************************************/\n\nstatic int\nprof_active_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tbool oldval;\n\n\tif (config_prof == false)\n\t\treturn (ENOENT);\n\n\tmalloc_mutex_lock(&ctl_mtx); /* Protect opt_prof_active. */\n\toldval = opt_prof_active;\n\tif (newp != NULL) {\n\t\t/*\n\t\t * The memory barriers will tend to make opt_prof_active\n\t\t * propagate faster on systems with weak memory ordering.\n\t\t */\n\t\tmb_write();\n\t\tWRITE(opt_prof_active, bool);\n\t\tmb_write();\n\t}\n\tREAD(oldval, bool);\n\n\tret = 0;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nstatic int\nprof_dump_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\tint ret;\n\tconst char *filename = NULL;\n\n\tif (config_prof == false)\n\t\treturn (ENOENT);\n\n\tWRITEONLY();\n\tWRITE(filename, const char *);\n\n\tif (prof_mdump(filename)) {\n\t\tret = EFAULT;\n\t\tgoto label_return;\n\t}\n\n\tret = 0;\nlabel_return:\n\treturn (ret);\n}\n\nCTL_RO_NL_CGEN(config_prof, prof_interval, prof_interval, uint64_t)\n\n/******************************************************************************/\n\nCTL_RO_CGEN(config_stats, stats_chunks_current, ctl_stats.chunks.current,\n size_t)\nCTL_RO_CGEN(config_stats, stats_chunks_total, ctl_stats.chunks.total, uint64_t)\nCTL_RO_CGEN(config_stats, stats_chunks_high, ctl_stats.chunks.high, size_t)\nCTL_RO_CGEN(config_stats, stats_huge_allocated, huge_allocated, size_t)\nCTL_RO_CGEN(config_stats, stats_huge_nmalloc, huge_nmalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_huge_ndalloc, huge_ndalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_small_allocated,\n ctl_stats.arenas[mib[2]].allocated_small, size_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_small_nmalloc,\n ctl_stats.arenas[mib[2]].nmalloc_small, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_small_ndalloc,\n ctl_stats.arenas[mib[2]].ndalloc_small, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_small_nrequests,\n ctl_stats.arenas[mib[2]].nrequests_small, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_large_allocated,\n ctl_stats.arenas[mib[2]].astats.allocated_large, size_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_large_nmalloc,\n ctl_stats.arenas[mib[2]].astats.nmalloc_large, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_large_ndalloc,\n ctl_stats.arenas[mib[2]].astats.ndalloc_large, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_large_nrequests,\n ctl_stats.arenas[mib[2]].astats.nrequests_large, uint64_t)\n\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_allocated,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].allocated, size_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nmalloc,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].nmalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_ndalloc,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].ndalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nrequests,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].nrequests, uint64_t)\nCTL_RO_CGEN(config_stats && config_tcache, stats_arenas_i_bins_j_nfills,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].nfills, uint64_t)\nCTL_RO_CGEN(config_stats && config_tcache, stats_arenas_i_bins_j_nflushes,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].nflushes, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nruns,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].nruns, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nreruns,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].reruns, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_curruns,\n ctl_stats.arenas[mib[2]].bstats[mib[4]].curruns, size_t)\n\nstatic const ctl_named_node_t *\nstats_arenas_i_bins_j_index(const size_t *mib, size_t miblen, size_t j)\n{\n\n\tif (j > NBINS)\n\t\treturn (NULL);\n\treturn (super_stats_arenas_i_bins_j_node);\n}\n\nCTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_nmalloc,\n ctl_stats.arenas[mib[2]].lstats[mib[4]].nmalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_ndalloc,\n ctl_stats.arenas[mib[2]].lstats[mib[4]].ndalloc, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_nrequests,\n ctl_stats.arenas[mib[2]].lstats[mib[4]].nrequests, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_curruns,\n ctl_stats.arenas[mib[2]].lstats[mib[4]].curruns, size_t)\n\nstatic const ctl_named_node_t *\nstats_arenas_i_lruns_j_index(const size_t *mib, size_t miblen, size_t j)\n{\n\n\tif (j > nlclasses)\n\t\treturn (NULL);\n\treturn (super_stats_arenas_i_lruns_j_node);\n}\n\nCTL_RO_GEN(stats_arenas_i_nthreads, ctl_stats.arenas[mib[2]].nthreads, unsigned)\nCTL_RO_GEN(stats_arenas_i_dss, ctl_stats.arenas[mib[2]].dss, const char *)\nCTL_RO_GEN(stats_arenas_i_pactive, ctl_stats.arenas[mib[2]].pactive, size_t)\nCTL_RO_GEN(stats_arenas_i_pdirty, ctl_stats.arenas[mib[2]].pdirty, size_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_mapped,\n ctl_stats.arenas[mib[2]].astats.mapped, size_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_npurge,\n ctl_stats.arenas[mib[2]].astats.npurge, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_nmadvise,\n ctl_stats.arenas[mib[2]].astats.nmadvise, uint64_t)\nCTL_RO_CGEN(config_stats, stats_arenas_i_purged,\n ctl_stats.arenas[mib[2]].astats.purged, uint64_t)\n\nstatic const ctl_named_node_t *\nstats_arenas_i_index(const size_t *mib, size_t miblen, size_t i)\n{\n\tconst ctl_named_node_t * ret;\n\n\tmalloc_mutex_lock(&ctl_mtx);\n\tif (i > ctl_stats.narenas || ctl_stats.arenas[i].initialized == false) {\n\t\tret = NULL;\n\t\tgoto label_return;\n\t}\n\n\tret = super_stats_arenas_i_node;\nlabel_return:\n\tmalloc_mutex_unlock(&ctl_mtx);\n\treturn (ret);\n}\n\nCTL_RO_CGEN(config_stats, stats_cactive, &stats_cactive, size_t *)\nCTL_RO_CGEN(config_stats, stats_allocated, ctl_stats.allocated, size_t)\nCTL_RO_CGEN(config_stats, stats_active, ctl_stats.active, size_t)\nCTL_RO_CGEN(config_stats, stats_mapped, ctl_stats.mapped, size_t)\n" }, { "path": "deps/jemalloc/src/extent.c", "content": "#define\tJEMALLOC_EXTENT_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n\nstatic inline int\nextent_szad_comp(extent_node_t *a, extent_node_t *b)\n{\n\tint ret;\n\tsize_t a_size = a->size;\n\tsize_t b_size = b->size;\n\n\tret = (a_size > b_size) - (a_size < b_size);\n\tif (ret == 0) {\n\t\tuintptr_t a_addr = (uintptr_t)a->addr;\n\t\tuintptr_t b_addr = (uintptr_t)b->addr;\n\n\t\tret = (a_addr > b_addr) - (a_addr < b_addr);\n\t}\n\n\treturn (ret);\n}\n\n/* Generate red-black tree functions. */\nrb_gen(, extent_tree_szad_, extent_tree_t, extent_node_t, link_szad,\n extent_szad_comp)\n\nstatic inline int\nextent_ad_comp(extent_node_t *a, extent_node_t *b)\n{\n\tuintptr_t a_addr = (uintptr_t)a->addr;\n\tuintptr_t b_addr = (uintptr_t)b->addr;\n\n\treturn ((a_addr > b_addr) - (a_addr < b_addr));\n}\n\n/* Generate red-black tree functions. */\nrb_gen(, extent_tree_ad_, extent_tree_t, extent_node_t, link_ad,\n extent_ad_comp)\n" }, { "path": "deps/jemalloc/src/hash.c", "content": "#define\tJEMALLOC_HASH_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n" }, { "path": "deps/jemalloc/src/huge.c", "content": "#define\tJEMALLOC_HUGE_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nuint64_t\thuge_nmalloc;\nuint64_t\thuge_ndalloc;\nsize_t\t\thuge_allocated;\n\nmalloc_mutex_t\thuge_mtx;\n\n/******************************************************************************/\n\n/* Tree of chunks that are stand-alone huge allocations. */\nstatic extent_tree_t\thuge;\n\nvoid *\nhuge_malloc(size_t size, bool zero)\n{\n\n\treturn (huge_palloc(size, chunksize, zero));\n}\n\nvoid *\nhuge_palloc(size_t size, size_t alignment, bool zero)\n{\n\tvoid *ret;\n\tsize_t csize;\n\textent_node_t *node;\n\tbool is_zeroed;\n\n\t/* Allocate one or more contiguous chunks for this request. */\n\n\tcsize = CHUNK_CEILING(size);\n\tif (csize == 0) {\n\t\t/* size is large enough to cause size_t wrap-around. */\n\t\treturn (NULL);\n\t}\n\n\t/* Allocate an extent node with which to track the chunk. */\n\tnode = base_node_alloc();\n\tif (node == NULL)\n\t\treturn (NULL);\n\n\t/*\n\t * Copy zero into is_zeroed and pass the copy to chunk_alloc(), so that\n\t * it is possible to make correct junk/zero fill decisions below.\n\t */\n\tis_zeroed = zero;\n\tret = chunk_alloc(csize, alignment, false, &is_zeroed,\n\t chunk_dss_prec_get());\n\tif (ret == NULL) {\n\t\tbase_node_dealloc(node);\n\t\treturn (NULL);\n\t}\n\n\t/* Insert node into huge. */\n\tnode->addr = ret;\n\tnode->size = csize;\n\n\tmalloc_mutex_lock(&huge_mtx);\n\textent_tree_ad_insert(&huge, node);\n\tif (config_stats) {\n\t\tstats_cactive_add(csize);\n\t\thuge_nmalloc++;\n\t\thuge_allocated += csize;\n\t}\n\tmalloc_mutex_unlock(&huge_mtx);\n\n\tif (config_fill && zero == false) {\n\t\tif (opt_junk)\n\t\t\tmemset(ret, 0xa5, csize);\n\t\telse if (opt_zero && is_zeroed == false)\n\t\t\tmemset(ret, 0, csize);\n\t}\n\n\treturn (ret);\n}\n\nvoid *\nhuge_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra)\n{\n\n\t/*\n\t * Avoid moving the allocation if the size class can be left the same.\n\t */\n\tif (oldsize > arena_maxclass\n\t && CHUNK_CEILING(oldsize) >= CHUNK_CEILING(size)\n\t && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(size+extra)) {\n\t\tassert(CHUNK_CEILING(oldsize) == oldsize);\n\t\tif (config_fill && opt_junk && size < oldsize) {\n\t\t\tmemset((void *)((uintptr_t)ptr + size), 0x5a,\n\t\t\t oldsize - size);\n\t\t}\n\t\treturn (ptr);\n\t}\n\n\t/* Reallocation would require a move. */\n\treturn (NULL);\n}\n\nvoid *\nhuge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra,\n size_t alignment, bool zero, bool try_tcache_dalloc)\n{\n\tvoid *ret;\n\tsize_t copysize;\n\n\t/* Try to avoid moving the allocation. */\n\tret = huge_ralloc_no_move(ptr, oldsize, size, extra);\n\tif (ret != NULL)\n\t\treturn (ret);\n\n\t/*\n\t * size and oldsize are different enough that we need to use a\n\t * different size class. In that case, fall back to allocating new\n\t * space and copying.\n\t */\n\tif (alignment > chunksize)\n\t\tret = huge_palloc(size + extra, alignment, zero);\n\telse\n\t\tret = huge_malloc(size + extra, zero);\n\n\tif (ret == NULL) {\n\t\tif (extra == 0)\n\t\t\treturn (NULL);\n\t\t/* Try again, this time without extra. */\n\t\tif (alignment > chunksize)\n\t\t\tret = huge_palloc(size, alignment, zero);\n\t\telse\n\t\t\tret = huge_malloc(size, zero);\n\n\t\tif (ret == NULL)\n\t\t\treturn (NULL);\n\t}\n\n\t/*\n\t * Copy at most size bytes (not size+extra), since the caller has no\n\t * expectation that the extra bytes will be reliably preserved.\n\t */\n\tcopysize = (size < oldsize) ? size : oldsize;\n\n#ifdef JEMALLOC_MREMAP\n\t/*\n\t * Use mremap(2) if this is a huge-->huge reallocation, and neither the\n\t * source nor the destination are in dss.\n\t */\n\tif (oldsize >= chunksize && (config_dss == false || (chunk_in_dss(ptr)\n\t == false && chunk_in_dss(ret) == false))) {\n\t\tsize_t newsize = huge_salloc(ret);\n\n\t\t/*\n\t\t * Remove ptr from the tree of huge allocations before\n\t\t * performing the remap operation, in order to avoid the\n\t\t * possibility of another thread acquiring that mapping before\n\t\t * this one removes it from the tree.\n\t\t */\n\t\thuge_dalloc(ptr, false);\n\t\tif (mremap(ptr, oldsize, newsize, MREMAP_MAYMOVE|MREMAP_FIXED,\n\t\t ret) == MAP_FAILED) {\n\t\t\t/*\n\t\t\t * Assuming no chunk management bugs in the allocator,\n\t\t\t * the only documented way an error can occur here is\n\t\t\t * if the application changed the map type for a\n\t\t\t * portion of the old allocation. This is firmly in\n\t\t\t * undefined behavior territory, so write a diagnostic\n\t\t\t * message, and optionally abort.\n\t\t\t */\n\t\t\tchar buf[BUFERROR_BUF];\n\n\t\t\tbuferror(buf, sizeof(buf));\n\t\t\tmalloc_printf(\": Error in mremap(): %s\\n\",\n\t\t\t buf);\n\t\t\tif (opt_abort)\n\t\t\t\tabort();\n\t\t\tmemcpy(ret, ptr, copysize);\n\t\t\tchunk_dealloc_mmap(ptr, oldsize);\n\t\t}\n\t} else\n#endif\n\t{\n\t\tmemcpy(ret, ptr, copysize);\n\t\tiqallocx(ptr, try_tcache_dalloc);\n\t}\n\treturn (ret);\n}\n\nvoid\nhuge_dalloc(void *ptr, bool unmap)\n{\n\textent_node_t *node, key;\n\n\tmalloc_mutex_lock(&huge_mtx);\n\n\t/* Extract from tree of huge allocations. */\n\tkey.addr = ptr;\n\tnode = extent_tree_ad_search(&huge, &key);\n\tassert(node != NULL);\n\tassert(node->addr == ptr);\n\textent_tree_ad_remove(&huge, node);\n\n\tif (config_stats) {\n\t\tstats_cactive_sub(node->size);\n\t\thuge_ndalloc++;\n\t\thuge_allocated -= node->size;\n\t}\n\n\tmalloc_mutex_unlock(&huge_mtx);\n\n\tif (unmap && config_fill && config_dss && opt_junk)\n\t\tmemset(node->addr, 0x5a, node->size);\n\n\tchunk_dealloc(node->addr, node->size, unmap);\n\n\tbase_node_dealloc(node);\n}\n\nsize_t\nhuge_salloc(const void *ptr)\n{\n\tsize_t ret;\n\textent_node_t *node, key;\n\n\tmalloc_mutex_lock(&huge_mtx);\n\n\t/* Extract from tree of huge allocations. */\n\tkey.addr = __DECONST(void *, ptr);\n\tnode = extent_tree_ad_search(&huge, &key);\n\tassert(node != NULL);\n\n\tret = node->size;\n\n\tmalloc_mutex_unlock(&huge_mtx);\n\n\treturn (ret);\n}\n\nprof_ctx_t *\nhuge_prof_ctx_get(const void *ptr)\n{\n\tprof_ctx_t *ret;\n\textent_node_t *node, key;\n\n\tmalloc_mutex_lock(&huge_mtx);\n\n\t/* Extract from tree of huge allocations. */\n\tkey.addr = __DECONST(void *, ptr);\n\tnode = extent_tree_ad_search(&huge, &key);\n\tassert(node != NULL);\n\n\tret = node->prof_ctx;\n\n\tmalloc_mutex_unlock(&huge_mtx);\n\n\treturn (ret);\n}\n\nvoid\nhuge_prof_ctx_set(const void *ptr, prof_ctx_t *ctx)\n{\n\textent_node_t *node, key;\n\n\tmalloc_mutex_lock(&huge_mtx);\n\n\t/* Extract from tree of huge allocations. */\n\tkey.addr = __DECONST(void *, ptr);\n\tnode = extent_tree_ad_search(&huge, &key);\n\tassert(node != NULL);\n\n\tnode->prof_ctx = ctx;\n\n\tmalloc_mutex_unlock(&huge_mtx);\n}\n\nbool\nhuge_boot(void)\n{\n\n\t/* Initialize chunks data. */\n\tif (malloc_mutex_init(&huge_mtx))\n\t\treturn (true);\n\textent_tree_ad_new(&huge);\n\n\tif (config_stats) {\n\t\thuge_nmalloc = 0;\n\t\thuge_ndalloc = 0;\n\t\thuge_allocated = 0;\n\t}\n\n\treturn (false);\n}\n\nvoid\nhuge_prefork(void)\n{\n\n\tmalloc_mutex_prefork(&huge_mtx);\n}\n\nvoid\nhuge_postfork_parent(void)\n{\n\n\tmalloc_mutex_postfork_parent(&huge_mtx);\n}\n\nvoid\nhuge_postfork_child(void)\n{\n\n\tmalloc_mutex_postfork_child(&huge_mtx);\n}\n" }, { "path": "deps/jemalloc/src/jemalloc.c", "content": "#define\tJEMALLOC_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nmalloc_tsd_data(, arenas, arena_t *, NULL)\nmalloc_tsd_data(, thread_allocated, thread_allocated_t,\n THREAD_ALLOCATED_INITIALIZER)\n\n/* Runtime configuration options. */\nconst char\t*je_malloc_conf;\n#ifdef JEMALLOC_DEBUG\nbool\topt_abort = true;\n# ifdef JEMALLOC_FILL\nbool\topt_junk = true;\n# else\nbool\topt_junk = false;\n# endif\n#else\nbool\topt_abort = false;\nbool\topt_junk = false;\n#endif\nsize_t\topt_quarantine = ZU(0);\nbool\topt_redzone = false;\nbool\topt_utrace = false;\nbool\topt_valgrind = false;\nbool\topt_xmalloc = false;\nbool\topt_zero = false;\nsize_t\topt_narenas = 0;\n\nunsigned\tncpus;\n\nmalloc_mutex_t\t\tarenas_lock;\narena_t\t\t\t**arenas;\nunsigned\t\tnarenas_total;\nunsigned\t\tnarenas_auto;\n\n/* Set to true once the allocator has been initialized. */\nstatic bool\t\tmalloc_initialized = false;\n\n#ifdef JEMALLOC_THREADED_INIT\n/* Used to let the initializing thread recursively allocate. */\n# define NO_INITIALIZER\t((unsigned long)0)\n# define INITIALIZER\t\tpthread_self()\n# define IS_INITIALIZER\t(malloc_initializer == pthread_self())\nstatic pthread_t\t\tmalloc_initializer = NO_INITIALIZER;\n#else\n# define NO_INITIALIZER\tfalse\n# define INITIALIZER\t\ttrue\n# define IS_INITIALIZER\tmalloc_initializer\nstatic bool\t\t\tmalloc_initializer = NO_INITIALIZER;\n#endif\n\n/* Used to avoid initialization races. */\n#ifdef _WIN32\nstatic malloc_mutex_t\tinit_lock;\n\nJEMALLOC_ATTR(constructor)\nstatic void WINAPI\n_init_init_lock(void)\n{\n\n\tmalloc_mutex_init(&init_lock);\n}\n\n#ifdef _MSC_VER\n# pragma section(\".CRT$XCU\", read)\nJEMALLOC_SECTION(\".CRT$XCU\") JEMALLOC_ATTR(used)\nstatic const void (WINAPI *init_init_lock)(void) = _init_init_lock;\n#endif\n\n#else\nstatic malloc_mutex_t\tinit_lock = MALLOC_MUTEX_INITIALIZER;\n#endif\n\ntypedef struct {\n\tvoid\t*p;\t/* Input pointer (as in realloc(p, s)). */\n\tsize_t\ts;\t/* Request size. */\n\tvoid\t*r;\t/* Result pointer. */\n} malloc_utrace_t;\n\n#ifdef JEMALLOC_UTRACE\n# define UTRACE(a, b, c) do {\t\t\t\t\t\t\\\n\tif (opt_utrace) {\t\t\t\t\t\t\\\n\t\tmalloc_utrace_t ut;\t\t\t\t\t\\\n\t\tut.p = (a);\t\t\t\t\t\t\\\n\t\tut.s = (b);\t\t\t\t\t\t\\\n\t\tut.r = (c);\t\t\t\t\t\t\\\n\t\tutrace(&ut, sizeof(ut));\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#else\n# define UTRACE(a, b, c)\n#endif\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\tstats_print_atexit(void);\nstatic unsigned\tmalloc_ncpus(void);\nstatic bool\tmalloc_conf_next(char const **opts_p, char const **k_p,\n size_t *klen_p, char const **v_p, size_t *vlen_p);\nstatic void\tmalloc_conf_error(const char *msg, const char *k, size_t klen,\n const char *v, size_t vlen);\nstatic void\tmalloc_conf_init(void);\nstatic bool\tmalloc_init_hard(void);\nstatic int\timemalign(void **memptr, size_t alignment, size_t size,\n size_t min_alignment);\n\n/******************************************************************************/\n/*\n * Begin miscellaneous support functions.\n */\n\n/* Create a new arena and insert it into the arenas array at index ind. */\narena_t *\narenas_extend(unsigned ind)\n{\n\tarena_t *ret;\n\n\tret = (arena_t *)base_alloc(sizeof(arena_t));\n\tif (ret != NULL && arena_new(ret, ind) == false) {\n\t\tarenas[ind] = ret;\n\t\treturn (ret);\n\t}\n\t/* Only reached if there is an OOM error. */\n\n\t/*\n\t * OOM here is quite inconvenient to propagate, since dealing with it\n\t * would require a check for failure in the fast path. Instead, punt\n\t * by using arenas[0]. In practice, this is an extremely unlikely\n\t * failure.\n\t */\n\tmalloc_write(\": Error initializing arena\\n\");\n\tif (opt_abort)\n\t\tabort();\n\n\treturn (arenas[0]);\n}\n\n/* Slow path, called only by choose_arena(). */\narena_t *\nchoose_arena_hard(void)\n{\n\tarena_t *ret;\n\n\tif (narenas_auto > 1) {\n\t\tunsigned i, choose, first_null;\n\n\t\tchoose = 0;\n\t\tfirst_null = narenas_auto;\n\t\tmalloc_mutex_lock(&arenas_lock);\n\t\tassert(arenas[0] != NULL);\n\t\tfor (i = 1; i < narenas_auto; i++) {\n\t\t\tif (arenas[i] != NULL) {\n\t\t\t\t/*\n\t\t\t\t * Choose the first arena that has the lowest\n\t\t\t\t * number of threads assigned to it.\n\t\t\t\t */\n\t\t\t\tif (arenas[i]->nthreads <\n\t\t\t\t arenas[choose]->nthreads)\n\t\t\t\t\tchoose = i;\n\t\t\t} else if (first_null == narenas_auto) {\n\t\t\t\t/*\n\t\t\t\t * Record the index of the first uninitialized\n\t\t\t\t * arena, in case all extant arenas are in use.\n\t\t\t\t *\n\t\t\t\t * NB: It is possible for there to be\n\t\t\t\t * discontinuities in terms of initialized\n\t\t\t\t * versus uninitialized arenas, due to the\n\t\t\t\t * \"thread.arena\" mallctl.\n\t\t\t\t */\n\t\t\t\tfirst_null = i;\n\t\t\t}\n\t\t}\n\n\t\tif (arenas[choose]->nthreads == 0\n\t\t || first_null == narenas_auto) {\n\t\t\t/*\n\t\t\t * Use an unloaded arena, or the least loaded arena if\n\t\t\t * all arenas are already initialized.\n\t\t\t */\n\t\t\tret = arenas[choose];\n\t\t} else {\n\t\t\t/* Initialize a new arena. */\n\t\t\tret = arenas_extend(first_null);\n\t\t}\n\t\tret->nthreads++;\n\t\tmalloc_mutex_unlock(&arenas_lock);\n\t} else {\n\t\tret = arenas[0];\n\t\tmalloc_mutex_lock(&arenas_lock);\n\t\tret->nthreads++;\n\t\tmalloc_mutex_unlock(&arenas_lock);\n\t}\n\n\tarenas_tsd_set(&ret);\n\n\treturn (ret);\n}\n\nstatic void\nstats_print_atexit(void)\n{\n\n\tif (config_tcache && config_stats) {\n\t\tunsigned narenas, i;\n\n\t\t/*\n\t\t * Merge stats from extant threads. This is racy, since\n\t\t * individual threads do not lock when recording tcache stats\n\t\t * events. As a consequence, the final stats may be slightly\n\t\t * out of date by the time they are reported, if other threads\n\t\t * continue to allocate.\n\t\t */\n\t\tfor (i = 0, narenas = narenas_total_get(); i < narenas; i++) {\n\t\t\tarena_t *arena = arenas[i];\n\t\t\tif (arena != NULL) {\n\t\t\t\ttcache_t *tcache;\n\n\t\t\t\t/*\n\t\t\t\t * tcache_stats_merge() locks bins, so if any\n\t\t\t\t * code is introduced that acquires both arena\n\t\t\t\t * and bin locks in the opposite order,\n\t\t\t\t * deadlocks may result.\n\t\t\t\t */\n\t\t\t\tmalloc_mutex_lock(&arena->lock);\n\t\t\t\tql_foreach(tcache, &arena->tcache_ql, link) {\n\t\t\t\t\ttcache_stats_merge(tcache, arena);\n\t\t\t\t}\n\t\t\t\tmalloc_mutex_unlock(&arena->lock);\n\t\t\t}\n\t\t}\n\t}\n\tje_malloc_stats_print(NULL, NULL, NULL);\n}\n\n/*\n * End miscellaneous support functions.\n */\n/******************************************************************************/\n/*\n * Begin initialization functions.\n */\n\nstatic unsigned\nmalloc_ncpus(void)\n{\n\tunsigned ret;\n\tlong result;\n\n#ifdef _WIN32\n\tSYSTEM_INFO si;\n\tGetSystemInfo(&si);\n\tresult = si.dwNumberOfProcessors;\n#else\n\tresult = sysconf(_SC_NPROCESSORS_ONLN);\n#endif\n\tif (result == -1) {\n\t\t/* Error. */\n\t\tret = 1;\n\t} else {\n ret = (unsigned)result;\n }\n\n\treturn (ret);\n}\n\nvoid\narenas_cleanup(void *arg)\n{\n\tarena_t *arena = *(arena_t **)arg;\n\n\tmalloc_mutex_lock(&arenas_lock);\n\tarena->nthreads--;\n\tmalloc_mutex_unlock(&arenas_lock);\n}\n\nstatic inline bool\nmalloc_init(void)\n{\n\n\tif (malloc_initialized == false)\n\t\treturn (malloc_init_hard());\n\n\treturn (false);\n}\n\nstatic bool\nmalloc_conf_next(char const **opts_p, char const **k_p, size_t *klen_p,\n char const **v_p, size_t *vlen_p)\n{\n\tbool accept;\n\tconst char *opts = *opts_p;\n\n\t*k_p = opts;\n\n\tfor (accept = false; accept == false;) {\n\t\tswitch (*opts) {\n\t\tcase 'A': case 'B': case 'C': case 'D': case 'E': case 'F':\n\t\tcase 'G': case 'H': case 'I': case 'J': case 'K': case 'L':\n\t\tcase 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':\n\t\tcase 'S': case 'T': case 'U': case 'V': case 'W': case 'X':\n\t\tcase 'Y': case 'Z':\n\t\tcase 'a': case 'b': case 'c': case 'd': case 'e': case 'f':\n\t\tcase 'g': case 'h': case 'i': case 'j': case 'k': case 'l':\n\t\tcase 'm': case 'n': case 'o': case 'p': case 'q': case 'r':\n\t\tcase 's': case 't': case 'u': case 'v': case 'w': case 'x':\n\t\tcase 'y': case 'z':\n\t\tcase '0': case '1': case '2': case '3': case '4': case '5':\n\t\tcase '6': case '7': case '8': case '9':\n\t\tcase '_':\n\t\t\topts++;\n\t\t\tbreak;\n\t\tcase ':':\n\t\t\topts++;\n\t\t\t*klen_p = (uintptr_t)opts - 1 - (uintptr_t)*k_p;\n\t\t\t*v_p = opts;\n\t\t\taccept = true;\n\t\t\tbreak;\n\t\tcase '\\0':\n\t\t\tif (opts != *opts_p) {\n\t\t\t\tmalloc_write(\": Conf string ends \"\n\t\t\t\t \"with key\\n\");\n\t\t\t}\n\t\t\treturn (true);\n\t\tdefault:\n\t\t\tmalloc_write(\": Malformed conf string\\n\");\n\t\t\treturn (true);\n\t\t}\n\t}\n\n\tfor (accept = false; accept == false;) {\n\t\tswitch (*opts) {\n\t\tcase ',':\n\t\t\topts++;\n\t\t\t/*\n\t\t\t * Look ahead one character here, because the next time\n\t\t\t * this function is called, it will assume that end of\n\t\t\t * input has been cleanly reached if no input remains,\n\t\t\t * but we have optimistically already consumed the\n\t\t\t * comma if one exists.\n\t\t\t */\n\t\t\tif (*opts == '\\0') {\n\t\t\t\tmalloc_write(\": Conf string ends \"\n\t\t\t\t \"with comma\\n\");\n\t\t\t}\n\t\t\t*vlen_p = (uintptr_t)opts - 1 - (uintptr_t)*v_p;\n\t\t\taccept = true;\n\t\t\tbreak;\n\t\tcase '\\0':\n\t\t\t*vlen_p = (uintptr_t)opts - (uintptr_t)*v_p;\n\t\t\taccept = true;\n\t\t\tbreak;\n\t\tdefault:\n\t\t\topts++;\n\t\t\tbreak;\n\t\t}\n\t}\n\n\t*opts_p = opts;\n\treturn (false);\n}\n\nstatic void\nmalloc_conf_error(const char *msg, const char *k, size_t klen, const char *v,\n size_t vlen)\n{\n\n\tmalloc_printf(\": %s: %.*s:%.*s\\n\", msg, (int)klen, k,\n\t (int)vlen, v);\n}\n\nstatic void\nmalloc_conf_init(void)\n{\n\tunsigned i;\n\tchar buf[PATH_MAX + 1];\n\tconst char *opts, *k, *v;\n\tsize_t klen, vlen;\n\n\t/*\n\t * Automatically configure valgrind before processing options. The\n\t * valgrind option remains in jemalloc 3.x for compatibility reasons.\n\t */\n\tif (config_valgrind) {\n\t\topt_valgrind = (RUNNING_ON_VALGRIND != 0) ? true : false;\n\t\tif (config_fill && opt_valgrind) {\n\t\t\topt_junk = false;\n\t\t\tassert(opt_zero == false);\n\t\t\topt_quarantine = JEMALLOC_VALGRIND_QUARANTINE_DEFAULT;\n\t\t\topt_redzone = true;\n\t\t}\n\t\tif (config_tcache && opt_valgrind)\n\t\t\topt_tcache = false;\n\t}\n\n\tfor (i = 0; i < 3; i++) {\n\t\t/* Get runtime configuration. */\n\t\tswitch (i) {\n\t\tcase 0:\n\t\t\tif (je_malloc_conf != NULL) {\n\t\t\t\t/*\n\t\t\t\t * Use options that were compiled into the\n\t\t\t\t * program.\n\t\t\t\t */\n\t\t\t\topts = je_malloc_conf;\n\t\t\t} else {\n\t\t\t\t/* No configuration specified. */\n\t\t\t\tbuf[0] = '\\0';\n\t\t\t\topts = buf;\n\t\t\t}\n\t\t\tbreak;\n\t\tcase 1: {\n#ifndef _WIN32\n\t\t\tint linklen;\n\t\t\tconst char *linkname =\n# ifdef JEMALLOC_PREFIX\n\t\t\t \"/etc/\"JEMALLOC_PREFIX\"malloc.conf\"\n# else\n\t\t\t \"/etc/malloc.conf\"\n# endif\n\t\t\t ;\n\n\t\t\tif ((linklen = readlink(linkname, buf,\n\t\t\t sizeof(buf) - 1)) != -1) {\n\t\t\t\t/*\n\t\t\t\t * Use the contents of the \"/etc/malloc.conf\"\n\t\t\t\t * symbolic link's name.\n\t\t\t\t */\n\t\t\t\tbuf[linklen] = '\\0';\n\t\t\t\topts = buf;\n\t\t\t} else\n#endif\n\t\t\t{\n\t\t\t\t/* No configuration specified. */\n\t\t\t\tbuf[0] = '\\0';\n\t\t\t\topts = buf;\n\t\t\t}\n\t\t\tbreak;\n\t\t} case 2: {\n\t\t\tconst char *envname =\n#ifdef JEMALLOC_PREFIX\n\t\t\t JEMALLOC_CPREFIX\"MALLOC_CONF\"\n#else\n\t\t\t \"MALLOC_CONF\"\n#endif\n\t\t\t ;\n\n\t\t\tif ((opts = getenv(envname)) != NULL) {\n\t\t\t\t/*\n\t\t\t\t * Do nothing; opts is already initialized to\n\t\t\t\t * the value of the MALLOC_CONF environment\n\t\t\t\t * variable.\n\t\t\t\t */\n\t\t\t} else {\n\t\t\t\t/* No configuration specified. */\n\t\t\t\tbuf[0] = '\\0';\n\t\t\t\topts = buf;\n\t\t\t}\n\t\t\tbreak;\n\t\t} default:\n\t\t\t/* NOTREACHED */\n\t\t\tassert(false);\n\t\t\tbuf[0] = '\\0';\n\t\t\topts = buf;\n\t\t}\n\n\t\twhile (*opts != '\\0' && malloc_conf_next(&opts, &k, &klen, &v,\n\t\t &vlen) == false) {\n#define\tCONF_HANDLE_BOOL_HIT(o, n, hit)\t\t\t\t\t\\\n\t\t\tif (sizeof(n)-1 == klen && strncmp(n, k,\t\\\n\t\t\t klen) == 0) {\t\t\t\t\\\n\t\t\t\tif (strncmp(\"true\", v, vlen) == 0 &&\t\\\n\t\t\t\t vlen == sizeof(\"true\")-1)\t\t\\\n\t\t\t\t\to = true;\t\t\t\\\n\t\t\t\telse if (strncmp(\"false\", v, vlen) ==\t\\\n\t\t\t\t 0 && vlen == sizeof(\"false\")-1)\t\\\n\t\t\t\t\to = false;\t\t\t\\\n\t\t\t\telse {\t\t\t\t\t\\\n\t\t\t\t\tmalloc_conf_error(\t\t\\\n\t\t\t\t\t \"Invalid conf value\",\t\\\n\t\t\t\t\t k, klen, v, vlen);\t\t\\\n\t\t\t\t}\t\t\t\t\t\\\n\t\t\t\thit = true;\t\t\t\t\\\n\t\t\t} else\t\t\t\t\t\t\\\n\t\t\t\thit = false;\n#define\tCONF_HANDLE_BOOL(o, n) {\t\t\t\t\t\\\n\t\t\tbool hit;\t\t\t\t\t\\\n\t\t\tCONF_HANDLE_BOOL_HIT(o, n, hit);\t\t\\\n\t\t\tif (hit)\t\t\t\t\t\\\n\t\t\t\tcontinue;\t\t\t\t\\\n}\n#define\tCONF_HANDLE_SIZE_T(o, n, min, max)\t\t\t\t\\\n\t\t\tif (sizeof(n)-1 == klen && strncmp(n, k,\t\\\n\t\t\t klen) == 0) {\t\t\t\t\\\n\t\t\t\tuintmax_t um;\t\t\t\t\\\n\t\t\t\tchar *end;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\t\t\t\tset_errno(0);\t\t\t\t\\\n\t\t\t\tum = malloc_strtoumax(v, &end, 0);\t\\\n\t\t\t\tif (get_errno() != 0 || (uintptr_t)end -\\\n\t\t\t\t (uintptr_t)v != vlen) {\t\t\\\n\t\t\t\t\tmalloc_conf_error(\t\t\\\n\t\t\t\t\t \"Invalid conf value\",\t\\\n\t\t\t\t\t k, klen, v, vlen);\t\t\\\n\t\t\t\t} else if (um < min || um > max) {\t\\\n\t\t\t\t\tmalloc_conf_error(\t\t\\\n\t\t\t\t\t \"Out-of-range conf value\",\t\\\n\t\t\t\t\t k, klen, v, vlen);\t\t\\\n\t\t\t\t} else\t\t\t\t\t\\\n\t\t\t\t\to = um;\t\t\t\t\\\n\t\t\t\tcontinue;\t\t\t\t\\\n\t\t\t}\n#define\tCONF_HANDLE_SSIZE_T(o, n, min, max)\t\t\t\t\\\n\t\t\tif (sizeof(n)-1 == klen && strncmp(n, k,\t\\\n\t\t\t klen) == 0) {\t\t\t\t\\\n\t\t\t\tlong l;\t\t\t\t\t\\\n\t\t\t\tchar *end;\t\t\t\t\\\n\t\t\t\t\t\t\t\t\t\\\n\t\t\t\tset_errno(0);\t\t\t\t\\\n\t\t\t\tl = strtol(v, &end, 0);\t\t\t\\\n\t\t\t\tif (get_errno() != 0 || (uintptr_t)end -\\\n\t\t\t\t (uintptr_t)v != vlen) {\t\t\\\n\t\t\t\t\tmalloc_conf_error(\t\t\\\n\t\t\t\t\t \"Invalid conf value\",\t\\\n\t\t\t\t\t k, klen, v, vlen);\t\t\\\n\t\t\t\t} else if (l < (ssize_t)min || l >\t\\\n\t\t\t\t (ssize_t)max) {\t\t\t\\\n\t\t\t\t\tmalloc_conf_error(\t\t\\\n\t\t\t\t\t \"Out-of-range conf value\",\t\\\n\t\t\t\t\t k, klen, v, vlen);\t\t\\\n\t\t\t\t} else\t\t\t\t\t\\\n\t\t\t\t\to = l;\t\t\t\t\\\n\t\t\t\tcontinue;\t\t\t\t\\\n\t\t\t}\n#define\tCONF_HANDLE_CHAR_P(o, n, d)\t\t\t\t\t\\\n\t\t\tif (sizeof(n)-1 == klen && strncmp(n, k,\t\\\n\t\t\t klen) == 0) {\t\t\t\t\\\n\t\t\t\tsize_t cpylen = (vlen <=\t\t\\\n\t\t\t\t sizeof(o)-1) ? vlen :\t\t\\\n\t\t\t\t sizeof(o)-1;\t\t\t\\\n\t\t\t\tstrncpy(o, v, cpylen);\t\t\t\\\n\t\t\t\to[cpylen] = '\\0';\t\t\t\\\n\t\t\t\tcontinue;\t\t\t\t\\\n\t\t\t}\n\n\t\t\tCONF_HANDLE_BOOL(opt_abort, \"abort\")\n\t\t\t/*\n\t\t\t * Chunks always require at least one header page, plus\n\t\t\t * one data page in the absence of redzones, or three\n\t\t\t * pages in the presence of redzones. In order to\n\t\t\t * simplify options processing, fix the limit based on\n\t\t\t * config_fill.\n\t\t\t */\n\t\t\tCONF_HANDLE_SIZE_T(opt_lg_chunk, \"lg_chunk\", LG_PAGE +\n\t\t\t (config_fill ? 2 : 1), (sizeof(size_t) << 3) - 1)\n\t\t\tif (strncmp(\"dss\", k, klen) == 0) {\n\t\t\t\tint i;\n\t\t\t\tbool match = false;\n\t\t\t\tfor (i = 0; i < dss_prec_limit; i++) {\n\t\t\t\t\tif (strncmp(dss_prec_names[i], v, vlen)\n\t\t\t\t\t == 0) {\n\t\t\t\t\t\tif (chunk_dss_prec_set(i)) {\n\t\t\t\t\t\t\tmalloc_conf_error(\n\t\t\t\t\t\t\t \"Error setting dss\",\n\t\t\t\t\t\t\t k, klen, v, vlen);\n\t\t\t\t\t\t} else {\n\t\t\t\t\t\t\topt_dss =\n\t\t\t\t\t\t\t dss_prec_names[i];\n\t\t\t\t\t\t\tmatch = true;\n\t\t\t\t\t\t\tbreak;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tif (match == false) {\n\t\t\t\t\tmalloc_conf_error(\"Invalid conf value\",\n\t\t\t\t\t k, klen, v, vlen);\n\t\t\t\t}\n\t\t\t\tcontinue;\n\t\t\t}\n\t\t\tCONF_HANDLE_SIZE_T(opt_narenas, \"narenas\", 1,\n\t\t\t SIZE_T_MAX)\n\t\t\tCONF_HANDLE_SSIZE_T(opt_lg_dirty_mult, \"lg_dirty_mult\",\n\t\t\t -1, (sizeof(size_t) << 3) - 1)\n\t\t\tCONF_HANDLE_BOOL(opt_stats_print, \"stats_print\")\n\t\t\tif (config_fill) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_junk, \"junk\")\n\t\t\t\tCONF_HANDLE_SIZE_T(opt_quarantine, \"quarantine\",\n\t\t\t\t 0, SIZE_T_MAX)\n\t\t\t\tCONF_HANDLE_BOOL(opt_redzone, \"redzone\")\n\t\t\t\tCONF_HANDLE_BOOL(opt_zero, \"zero\")\n\t\t\t}\n\t\t\tif (config_utrace) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_utrace, \"utrace\")\n\t\t\t}\n\t\t\tif (config_valgrind) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_valgrind, \"valgrind\")\n\t\t\t}\n\t\t\tif (config_xmalloc) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_xmalloc, \"xmalloc\")\n\t\t\t}\n\t\t\tif (config_tcache) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_tcache, \"tcache\")\n\t\t\t\tCONF_HANDLE_SSIZE_T(opt_lg_tcache_max,\n\t\t\t\t \"lg_tcache_max\", -1,\n\t\t\t\t (sizeof(size_t) << 3) - 1)\n\t\t\t}\n\t\t\tif (config_prof) {\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof, \"prof\")\n\t\t\t\tCONF_HANDLE_CHAR_P(opt_prof_prefix,\n\t\t\t\t \"prof_prefix\", \"jeprof\")\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof_active, \"prof_active\")\n\t\t\t\tCONF_HANDLE_SSIZE_T(opt_lg_prof_sample,\n\t\t\t\t \"lg_prof_sample\", 0,\n\t\t\t\t (sizeof(uint64_t) << 3) - 1)\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof_accum, \"prof_accum\")\n\t\t\t\tCONF_HANDLE_SSIZE_T(opt_lg_prof_interval,\n\t\t\t\t \"lg_prof_interval\", -1,\n\t\t\t\t (sizeof(uint64_t) << 3) - 1)\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof_gdump, \"prof_gdump\")\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof_final, \"prof_final\")\n\t\t\t\tCONF_HANDLE_BOOL(opt_prof_leak, \"prof_leak\")\n\t\t\t}\n\t\t\tmalloc_conf_error(\"Invalid conf pair\", k, klen, v,\n\t\t\t vlen);\n#undef CONF_HANDLE_BOOL\n#undef CONF_HANDLE_SIZE_T\n#undef CONF_HANDLE_SSIZE_T\n#undef CONF_HANDLE_CHAR_P\n\t\t}\n\t}\n}\n\nstatic bool\nmalloc_init_hard(void)\n{\n\tarena_t *init_arenas[1];\n\n\tmalloc_mutex_lock(&init_lock);\n\tif (malloc_initialized || IS_INITIALIZER) {\n\t\t/*\n\t\t * Another thread initialized the allocator before this one\n\t\t * acquired init_lock, or this thread is the initializing\n\t\t * thread, and it is recursively allocating.\n\t\t */\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (false);\n\t}\n#ifdef JEMALLOC_THREADED_INIT\n\tif (malloc_initializer != NO_INITIALIZER && IS_INITIALIZER == false) {\n\t\t/* Busy-wait until the initializing thread completes. */\n\t\tdo {\n\t\t\tmalloc_mutex_unlock(&init_lock);\n\t\t\tCPU_SPINWAIT;\n\t\t\tmalloc_mutex_lock(&init_lock);\n\t\t} while (malloc_initialized == false);\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (false);\n\t}\n#endif\n\tmalloc_initializer = INITIALIZER;\n\n\tmalloc_tsd_boot();\n\tif (config_prof)\n\t\tprof_boot0();\n\n\tmalloc_conf_init();\n\n#if (!defined(JEMALLOC_MUTEX_INIT_CB) && !defined(JEMALLOC_ZONE) \\\n && !defined(_WIN32))\n\t/* Register fork handlers. */\n\tif (pthread_atfork(jemalloc_prefork, jemalloc_postfork_parent,\n\t jemalloc_postfork_child) != 0) {\n\t\tmalloc_write(\": Error in pthread_atfork()\\n\");\n\t\tif (opt_abort)\n\t\t\tabort();\n\t}\n#endif\n\n\tif (opt_stats_print) {\n\t\t/* Print statistics at exit. */\n\t\tif (atexit(stats_print_atexit) != 0) {\n\t\t\tmalloc_write(\": Error in atexit()\\n\");\n\t\t\tif (opt_abort)\n\t\t\t\tabort();\n\t\t}\n\t}\n\n\tif (base_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (chunk_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (ctl_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (config_prof)\n\t\tprof_boot1();\n\n\tarena_boot();\n\n\tif (config_tcache && tcache_boot0()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (huge_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (malloc_mutex_init(&arenas_lock))\n\t\treturn (true);\n\n\t/*\n\t * Create enough scaffolding to allow recursive allocation in\n\t * malloc_ncpus().\n\t */\n\tnarenas_total = narenas_auto = 1;\n\tarenas = init_arenas;\n\tmemset(arenas, 0, sizeof(arena_t *) * narenas_auto);\n\n\t/*\n\t * Initialize one arena here. The rest are lazily created in\n\t * choose_arena_hard().\n\t */\n\tarenas_extend(0);\n\tif (arenas[0] == NULL) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\t/* Initialize allocation counters before any allocations can occur. */\n\tif (config_stats && thread_allocated_tsd_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (arenas_tsd_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (config_tcache && tcache_boot1()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (config_fill && quarantine_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (config_prof && prof_boot2()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\t/* Get number of CPUs. */\n\tmalloc_mutex_unlock(&init_lock);\n\tncpus = malloc_ncpus();\n\tmalloc_mutex_lock(&init_lock);\n\n\tif (mutex_boot()) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\n\tif (opt_narenas == 0) {\n\t\t/*\n\t\t * For SMP systems, create more than one arena per CPU by\n\t\t * default.\n\t\t */\n\t\tif (ncpus > 1)\n\t\t\topt_narenas = ncpus << 2;\n\t\telse\n\t\t\topt_narenas = 1;\n\t}\n\tnarenas_auto = opt_narenas;\n\t/*\n\t * Make sure that the arenas array can be allocated. In practice, this\n\t * limit is enough to allow the allocator to function, but the ctl\n\t * machinery will fail to allocate memory at far lower limits.\n\t */\n\tif (narenas_auto > chunksize / sizeof(arena_t *)) {\n\t\tnarenas_auto = chunksize / sizeof(arena_t *);\n\t\tmalloc_printf(\": Reducing narenas to limit (%d)\\n\",\n\t\t narenas_auto);\n\t}\n\tnarenas_total = narenas_auto;\n\n\t/* Allocate and initialize arenas. */\n\tarenas = (arena_t **)base_alloc(sizeof(arena_t *) * narenas_total);\n\tif (arenas == NULL) {\n\t\tmalloc_mutex_unlock(&init_lock);\n\t\treturn (true);\n\t}\n\t/*\n\t * Zero the array. In practice, this should always be pre-zeroed,\n\t * since it was just mmap()ed, but let's be sure.\n\t */\n\tmemset(arenas, 0, sizeof(arena_t *) * narenas_total);\n\t/* Copy the pointer to the one arena that was already initialized. */\n\tarenas[0] = init_arenas[0];\n\n\tmalloc_initialized = true;\n\tmalloc_mutex_unlock(&init_lock);\n\treturn (false);\n}\n\n/*\n * End initialization functions.\n */\n/******************************************************************************/\n/*\n * Begin malloc(3)-compatible functions.\n */\n\nvoid *\nje_malloc(size_t size)\n{\n\tvoid *ret;\n\tsize_t usize JEMALLOC_CC_SILENCE_INIT(0);\n\tprof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL);\n\n\tif (malloc_init()) {\n\t\tret = NULL;\n\t\tgoto label_oom;\n\t}\n\n\tif (size == 0)\n\t\tsize = 1;\n\n\tif (config_prof && opt_prof) {\n\t\tusize = s2u(size);\n\t\tPROF_ALLOC_PREP(1, usize, cnt);\n\t\tif (cnt == NULL) {\n\t\t\tret = NULL;\n\t\t\tgoto label_oom;\n\t\t}\n\t\tif (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize <=\n\t\t SMALL_MAXCLASS) {\n\t\t\tret = imalloc(SMALL_MAXCLASS+1);\n\t\t\tif (ret != NULL)\n\t\t\t\tarena_prof_promoted(ret, usize);\n\t\t} else\n\t\t\tret = imalloc(size);\n\t} else {\n\t\tif (config_stats || (config_valgrind && opt_valgrind))\n\t\t\tusize = s2u(size);\n\t\tret = imalloc(size);\n\t}\n\nlabel_oom:\n\tif (ret == NULL) {\n\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\tmalloc_write(\": Error in malloc(): \"\n\t\t\t \"out of memory\\n\");\n\t\t\tabort();\n\t\t}\n\t\tset_errno(ENOMEM);\n\t}\n\tif (config_prof && opt_prof && ret != NULL)\n\t\tprof_malloc(ret, usize, cnt);\n\tif (config_stats && ret != NULL) {\n\t\tassert(usize == isalloc(ret, config_prof));\n\t\tthread_allocated_tsd_get()->allocated += usize;\n\t}\n\tUTRACE(0, size, ret);\n\tJEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, usize, false);\n\treturn (ret);\n}\n\nJEMALLOC_ATTR(nonnull(1))\n#ifdef JEMALLOC_PROF\n/*\n * Avoid any uncertainty as to how many backtrace frames to ignore in\n * PROF_ALLOC_PREP().\n */\nJEMALLOC_ATTR(noinline)\n#endif\nstatic int\nimemalign(void **memptr, size_t alignment, size_t size,\n size_t min_alignment)\n{\n\tint ret;\n\tsize_t usize;\n\tvoid *result;\n\tprof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL);\n\n\tassert(min_alignment != 0);\n\n\tif (malloc_init())\n\t\tresult = NULL;\n\telse {\n\t\tif (size == 0)\n\t\t\tsize = 1;\n\n\t\t/* Make sure that alignment is a large enough power of 2. */\n\t\tif (((alignment - 1) & alignment) != 0\n\t\t || (alignment < min_alignment)) {\n\t\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\t\tmalloc_write(\": Error allocating \"\n\t\t\t\t \"aligned memory: invalid alignment\\n\");\n\t\t\t\tabort();\n\t\t\t}\n\t\t\tresult = NULL;\n\t\t\tret = EINVAL;\n\t\t\tgoto label_return;\n\t\t}\n\n\t\tusize = sa2u(size, alignment);\n\t\tif (usize == 0) {\n\t\t\tresult = NULL;\n\t\t\tret = ENOMEM;\n\t\t\tgoto label_return;\n\t\t}\n\n\t\tif (config_prof && opt_prof) {\n\t\t\tPROF_ALLOC_PREP(2, usize, cnt);\n\t\t\tif (cnt == NULL) {\n\t\t\t\tresult = NULL;\n\t\t\t\tret = EINVAL;\n\t\t\t} else {\n\t\t\t\tif (prof_promote && (uintptr_t)cnt !=\n\t\t\t\t (uintptr_t)1U && usize <= SMALL_MAXCLASS) {\n\t\t\t\t\tassert(sa2u(SMALL_MAXCLASS+1,\n\t\t\t\t\t alignment) != 0);\n\t\t\t\t\tresult = ipalloc(sa2u(SMALL_MAXCLASS+1,\n\t\t\t\t\t alignment), alignment, false);\n\t\t\t\t\tif (result != NULL) {\n\t\t\t\t\t\tarena_prof_promoted(result,\n\t\t\t\t\t\t usize);\n\t\t\t\t\t}\n\t\t\t\t} else {\n\t\t\t\t\tresult = ipalloc(usize, alignment,\n\t\t\t\t\t false);\n\t\t\t\t}\n\t\t\t}\n\t\t} else\n\t\t\tresult = ipalloc(usize, alignment, false);\n\t}\n\n\tif (result == NULL) {\n\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\tmalloc_write(\": Error allocating aligned \"\n\t\t\t \"memory: out of memory\\n\");\n\t\t\tabort();\n\t\t}\n\t\tret = ENOMEM;\n\t\tgoto label_return;\n\t}\n\n\t*memptr = result;\n\tret = 0;\n\nlabel_return:\n\tif (config_stats && result != NULL) {\n\t\tassert(usize == isalloc(result, config_prof));\n\t\tthread_allocated_tsd_get()->allocated += usize;\n\t}\n\tif (config_prof && opt_prof && result != NULL)\n\t\tprof_malloc(result, usize, cnt);\n\tUTRACE(0, size, result);\n\treturn (ret);\n}\n\nint\nje_posix_memalign(void **memptr, size_t alignment, size_t size)\n{\n\tint ret = imemalign(memptr, alignment, size, sizeof(void *));\n\tJEMALLOC_VALGRIND_MALLOC(ret == 0, *memptr, isalloc(*memptr,\n\t config_prof), false);\n\treturn (ret);\n}\n\nvoid *\nje_aligned_alloc(size_t alignment, size_t size)\n{\n\tvoid *ret;\n\tint err;\n\n\tif ((err = imemalign(&ret, alignment, size, 1)) != 0) {\n\t\tret = NULL;\n\t\tset_errno(err);\n\t}\n\tJEMALLOC_VALGRIND_MALLOC(err == 0, ret, isalloc(ret, config_prof),\n\t false);\n\treturn (ret);\n}\n\nvoid *\nje_calloc(size_t num, size_t size)\n{\n\tvoid *ret;\n\tsize_t num_size;\n\tsize_t usize JEMALLOC_CC_SILENCE_INIT(0);\n\tprof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL);\n\n\tif (malloc_init()) {\n\t\tnum_size = 0;\n\t\tret = NULL;\n\t\tgoto label_return;\n\t}\n\n\tnum_size = num * size;\n\tif (num_size == 0) {\n\t\tif (num == 0 || size == 0)\n\t\t\tnum_size = 1;\n\t\telse {\n\t\t\tret = NULL;\n\t\t\tgoto label_return;\n\t\t}\n\t/*\n\t * Try to avoid division here. We know that it isn't possible to\n\t * overflow during multiplication if neither operand uses any of the\n\t * most significant half of the bits in a size_t.\n\t */\n\t} else if (((num | size) & (SIZE_T_MAX << (sizeof(size_t) << 2)))\n\t && (num_size / size != num)) {\n\t\t/* size_t overflow. */\n\t\tret = NULL;\n\t\tgoto label_return;\n\t}\n\n\tif (config_prof && opt_prof) {\n\t\tusize = s2u(num_size);\n\t\tPROF_ALLOC_PREP(1, usize, cnt);\n\t\tif (cnt == NULL) {\n\t\t\tret = NULL;\n\t\t\tgoto label_return;\n\t\t}\n\t\tif (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize\n\t\t <= SMALL_MAXCLASS) {\n\t\t\tret = icalloc(SMALL_MAXCLASS+1);\n\t\t\tif (ret != NULL)\n\t\t\t\tarena_prof_promoted(ret, usize);\n\t\t} else\n\t\t\tret = icalloc(num_size);\n\t} else {\n\t\tif (config_stats || (config_valgrind && opt_valgrind))\n\t\t\tusize = s2u(num_size);\n\t\tret = icalloc(num_size);\n\t}\n\nlabel_return:\n\tif (ret == NULL) {\n\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\tmalloc_write(\": Error in calloc(): out of \"\n\t\t\t \"memory\\n\");\n\t\t\tabort();\n\t\t}\n\t\tset_errno(ENOMEM);\n\t}\n\n\tif (config_prof && opt_prof && ret != NULL)\n\t\tprof_malloc(ret, usize, cnt);\n\tif (config_stats && ret != NULL) {\n\t\tassert(usize == isalloc(ret, config_prof));\n\t\tthread_allocated_tsd_get()->allocated += usize;\n\t}\n\tUTRACE(0, num_size, ret);\n\tJEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, usize, true);\n\treturn (ret);\n}\n\nvoid *\nje_realloc(void *ptr, size_t size)\n{\n\tvoid *ret;\n\tsize_t usize JEMALLOC_CC_SILENCE_INIT(0);\n\tsize_t old_size = 0;\n\tsize_t old_rzsize JEMALLOC_CC_SILENCE_INIT(0);\n\tprof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL);\n\tprof_ctx_t *old_ctx JEMALLOC_CC_SILENCE_INIT(NULL);\n\n\tif (size == 0) {\n\t\tif (ptr != NULL) {\n\t\t\t/* realloc(ptr, 0) is equivalent to free(p). */\n\t\t\tif (config_prof) {\n\t\t\t\told_size = isalloc(ptr, true);\n\t\t\t\tif (config_valgrind && opt_valgrind)\n\t\t\t\t\told_rzsize = p2rz(ptr);\n\t\t\t} else if (config_stats) {\n\t\t\t\told_size = isalloc(ptr, false);\n\t\t\t\tif (config_valgrind && opt_valgrind)\n\t\t\t\t\told_rzsize = u2rz(old_size);\n\t\t\t} else if (config_valgrind && opt_valgrind) {\n\t\t\t\told_size = isalloc(ptr, false);\n\t\t\t\told_rzsize = u2rz(old_size);\n\t\t\t}\n\t\t\tif (config_prof && opt_prof) {\n\t\t\t\told_ctx = prof_ctx_get(ptr);\n\t\t\t\tcnt = NULL;\n\t\t\t}\n\t\t\tiqalloc(ptr);\n\t\t\tret = NULL;\n\t\t\tgoto label_return;\n\t\t} else\n\t\t\tsize = 1;\n\t}\n\n\tif (ptr != NULL) {\n\t\tassert(malloc_initialized || IS_INITIALIZER);\n\n\t\tif (config_prof) {\n\t\t\told_size = isalloc(ptr, true);\n\t\t\tif (config_valgrind && opt_valgrind)\n\t\t\t\told_rzsize = p2rz(ptr);\n\t\t} else if (config_stats) {\n\t\t\told_size = isalloc(ptr, false);\n\t\t\tif (config_valgrind && opt_valgrind)\n\t\t\t\told_rzsize = u2rz(old_size);\n\t\t} else if (config_valgrind && opt_valgrind) {\n\t\t\told_size = isalloc(ptr, false);\n\t\t\told_rzsize = u2rz(old_size);\n\t\t}\n\t\tif (config_prof && opt_prof) {\n\t\t\tusize = s2u(size);\n\t\t\told_ctx = prof_ctx_get(ptr);\n\t\t\tPROF_ALLOC_PREP(1, usize, cnt);\n\t\t\tif (cnt == NULL) {\n\t\t\t\told_ctx = NULL;\n\t\t\t\tret = NULL;\n\t\t\t\tgoto label_oom;\n\t\t\t}\n\t\t\tif (prof_promote && (uintptr_t)cnt != (uintptr_t)1U &&\n\t\t\t usize <= SMALL_MAXCLASS) {\n\t\t\t\tret = iralloc(ptr, SMALL_MAXCLASS+1, 0, 0,\n\t\t\t\t false, false);\n\t\t\t\tif (ret != NULL)\n\t\t\t\t\tarena_prof_promoted(ret, usize);\n\t\t\t\telse\n\t\t\t\t\told_ctx = NULL;\n\t\t\t} else {\n\t\t\t\tret = iralloc(ptr, size, 0, 0, false, false);\n\t\t\t\tif (ret == NULL)\n\t\t\t\t\told_ctx = NULL;\n\t\t\t}\n\t\t} else {\n\t\t\tif (config_stats || (config_valgrind && opt_valgrind))\n\t\t\t\tusize = s2u(size);\n\t\t\tret = iralloc(ptr, size, 0, 0, false, false);\n\t\t}\n\nlabel_oom:\n\t\tif (ret == NULL) {\n\t\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\t\tmalloc_write(\": Error in realloc(): \"\n\t\t\t\t \"out of memory\\n\");\n\t\t\t\tabort();\n\t\t\t}\n\t\t\tset_errno(ENOMEM);\n\t\t}\n\t} else {\n\t\t/* realloc(NULL, size) is equivalent to malloc(size). */\n\t\tif (config_prof && opt_prof)\n\t\t\told_ctx = NULL;\n\t\tif (malloc_init()) {\n\t\t\tif (config_prof && opt_prof)\n\t\t\t\tcnt = NULL;\n\t\t\tret = NULL;\n\t\t} else {\n\t\t\tif (config_prof && opt_prof) {\n\t\t\t\tusize = s2u(size);\n\t\t\t\tPROF_ALLOC_PREP(1, usize, cnt);\n\t\t\t\tif (cnt == NULL)\n\t\t\t\t\tret = NULL;\n\t\t\t\telse {\n\t\t\t\t\tif (prof_promote && (uintptr_t)cnt !=\n\t\t\t\t\t (uintptr_t)1U && usize <=\n\t\t\t\t\t SMALL_MAXCLASS) {\n\t\t\t\t\t\tret = imalloc(SMALL_MAXCLASS+1);\n\t\t\t\t\t\tif (ret != NULL) {\n\t\t\t\t\t\t\tarena_prof_promoted(ret,\n\t\t\t\t\t\t\t usize);\n\t\t\t\t\t\t}\n\t\t\t\t\t} else\n\t\t\t\t\t\tret = imalloc(size);\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\tif (config_stats || (config_valgrind &&\n\t\t\t\t opt_valgrind))\n\t\t\t\t\tusize = s2u(size);\n\t\t\t\tret = imalloc(size);\n\t\t\t}\n\t\t}\n\n\t\tif (ret == NULL) {\n\t\t\tif (config_xmalloc && opt_xmalloc) {\n\t\t\t\tmalloc_write(\": Error in realloc(): \"\n\t\t\t\t \"out of memory\\n\");\n\t\t\t\tabort();\n\t\t\t}\n\t\t\tset_errno(ENOMEM);\n\t\t}\n\t}\n\nlabel_return:\n\tif (config_prof && opt_prof)\n\t\tprof_realloc(ret, usize, cnt, old_size, old_ctx);\n\tif (config_stats && ret != NULL) {\n\t\tthread_allocated_t *ta;\n\t\tassert(usize == isalloc(ret, config_prof));\n\t\tta = thread_allocated_tsd_get();\n\t\tta->allocated += usize;\n\t\tta->deallocated += old_size;\n\t}\n\tUTRACE(ptr, size, ret);\n\tJEMALLOC_VALGRIND_REALLOC(ret, usize, ptr, old_size, old_rzsize, false);\n\treturn (ret);\n}\n\nvoid\nje_free(void *ptr)\n{\n\n\tUTRACE(ptr, 0, 0);\n\tif (ptr != NULL) {\n\t\tsize_t usize;\n\t\tsize_t rzsize JEMALLOC_CC_SILENCE_INIT(0);\n\n\t\tassert(malloc_initialized || IS_INITIALIZER);\n\n\t\tif (config_prof && opt_prof) {\n\t\t\tusize = isalloc(ptr, config_prof);\n\t\t\tprof_free(ptr, usize);\n\t\t} else if (config_stats || config_valgrind)\n\t\t\tusize = isalloc(ptr, config_prof);\n\t\tif (config_stats)\n\t\t\tthread_allocated_tsd_get()->deallocated += usize;\n\t\tif (config_valgrind && opt_valgrind)\n\t\t\trzsize = p2rz(ptr);\n\t\tiqalloc(ptr);\n\t\tJEMALLOC_VALGRIND_FREE(ptr, rzsize);\n\t}\n}\n\n/*\n * End malloc(3)-compatible functions.\n */\n/******************************************************************************/\n/*\n * Begin non-standard override functions.\n */\n\n#ifdef JEMALLOC_OVERRIDE_MEMALIGN\nvoid *\nje_memalign(size_t alignment, size_t size)\n{\n\tvoid *ret JEMALLOC_CC_SILENCE_INIT(NULL);\n\timemalign(&ret, alignment, size, 1);\n\tJEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, size, false);\n\treturn (ret);\n}\n#endif\n\n#ifdef JEMALLOC_OVERRIDE_VALLOC\nvoid *\nje_valloc(size_t size)\n{\n\tvoid *ret JEMALLOC_CC_SILENCE_INIT(NULL);\n\timemalign(&ret, PAGE, size, 1);\n\tJEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, size, false);\n\treturn (ret);\n}\n#endif\n\n/*\n * is_malloc(je_malloc) is some macro magic to detect if jemalloc_defs.h has\n * #define je_malloc malloc\n */\n#define\tmalloc_is_malloc 1\n#define\tis_malloc_(a) malloc_is_ ## a\n#define\tis_malloc(a) is_malloc_(a)\n\n#if ((is_malloc(je_malloc) == 1) && defined(__GLIBC__) && !defined(__UCLIBC__))\n/*\n * glibc provides the RTLD_DEEPBIND flag for dlopen which can make it possible\n * to inconsistently reference libc's malloc(3)-compatible functions\n * (https://bugzilla.mozilla.org/show_bug.cgi?id=493541).\n *\n * These definitions interpose hooks in glibc. The functions are actually\n * passed an extra argument for the caller return address, which will be\n * ignored.\n */\nJEMALLOC_EXPORT void (* __free_hook)(void *ptr) = je_free;\nJEMALLOC_EXPORT void *(* __malloc_hook)(size_t size) = je_malloc;\nJEMALLOC_EXPORT void *(* __realloc_hook)(void *ptr, size_t size) = je_realloc;\nJEMALLOC_EXPORT void *(* __memalign_hook)(size_t alignment, size_t size) =\n je_memalign;\n#endif\n\n/*\n * End non-standard override functions.\n */\n/******************************************************************************/\n/*\n * Begin non-standard functions.\n */\n\nsize_t\nje_malloc_usable_size(JEMALLOC_USABLE_SIZE_CONST void *ptr)\n{\n\tsize_t ret;\n\n\tassert(malloc_initialized || IS_INITIALIZER);\n\n\tif (config_ivsalloc)\n\t\tret = ivsalloc(ptr, config_prof);\n\telse\n\t\tret = (ptr != NULL) ? isalloc(ptr, config_prof) : 0;\n\n\treturn (ret);\n}\n\nvoid\nje_malloc_stats_print(void (*write_cb)(void *, const char *), void *cbopaque,\n const char *opts)\n{\n\n\tstats_print(write_cb, cbopaque, opts);\n}\n\nint\nje_mallctl(const char *name, void *oldp, size_t *oldlenp, void *newp,\n size_t newlen)\n{\n\n\tif (malloc_init())\n\t\treturn (EAGAIN);\n\n\treturn (ctl_byname(name, oldp, oldlenp, newp, newlen));\n}\n\nint\nje_mallctlnametomib(const char *name, size_t *mibp, size_t *miblenp)\n{\n\n\tif (malloc_init())\n\t\treturn (EAGAIN);\n\n\treturn (ctl_nametomib(name, mibp, miblenp));\n}\n\nint\nje_mallctlbymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,\n void *newp, size_t newlen)\n{\n\n\tif (malloc_init())\n\t\treturn (EAGAIN);\n\n\treturn (ctl_bymib(mib, miblen, oldp, oldlenp, newp, newlen));\n}\n\n/*\n * End non-standard functions.\n */\n/******************************************************************************/\n/*\n * Begin experimental functions.\n */\n#ifdef JEMALLOC_EXPERIMENTAL\n\nJEMALLOC_INLINE void *\niallocm(size_t usize, size_t alignment, bool zero, bool try_tcache,\n arena_t *arena)\n{\n\n\tassert(usize == ((alignment == 0) ? s2u(usize) : sa2u(usize,\n\t alignment)));\n\n\tif (alignment != 0)\n\t\treturn (ipallocx(usize, alignment, zero, try_tcache, arena));\n\telse if (zero)\n\t\treturn (icallocx(usize, try_tcache, arena));\n\telse\n\t\treturn (imallocx(usize, try_tcache, arena));\n}\n\nint\nje_allocm(void **ptr, size_t *rsize, size_t size, int flags)\n{\n\tvoid *p;\n\tsize_t usize;\n\tsize_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK)\n\t & (SIZE_T_MAX-1));\n\tbool zero = flags & ALLOCM_ZERO;\n\tunsigned arena_ind = ((unsigned)(flags >> 8)) - 1;\n\tarena_t *arena;\n\tbool try_tcache;\n\n\tassert(ptr != NULL);\n\tassert(size != 0);\n\n\tif (malloc_init())\n\t\tgoto label_oom;\n\n\tif (arena_ind != UINT_MAX) {\n\t\tarena = arenas[arena_ind];\n\t\ttry_tcache = false;\n\t} else {\n\t\tarena = NULL;\n\t\ttry_tcache = true;\n\t}\n\n\tusize = (alignment == 0) ? s2u(size) : sa2u(size, alignment);\n\tif (usize == 0)\n\t\tgoto label_oom;\n\n\tif (config_prof && opt_prof) {\n\t\tprof_thr_cnt_t *cnt;\n\n\t\tPROF_ALLOC_PREP(1, usize, cnt);\n\t\tif (cnt == NULL)\n\t\t\tgoto label_oom;\n\t\tif (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize <=\n\t\t SMALL_MAXCLASS) {\n\t\t\tsize_t usize_promoted = (alignment == 0) ?\n\t\t\t s2u(SMALL_MAXCLASS+1) : sa2u(SMALL_MAXCLASS+1,\n\t\t\t alignment);\n\t\t\tassert(usize_promoted != 0);\n\t\t\tp = iallocm(usize_promoted, alignment, zero,\n\t\t\t try_tcache, arena);\n\t\t\tif (p == NULL)\n\t\t\t\tgoto label_oom;\n\t\t\tarena_prof_promoted(p, usize);\n\t\t} else {\n\t\t\tp = iallocm(usize, alignment, zero, try_tcache, arena);\n\t\t\tif (p == NULL)\n\t\t\t\tgoto label_oom;\n\t\t}\n\t\tprof_malloc(p, usize, cnt);\n\t} else {\n\t\tp = iallocm(usize, alignment, zero, try_tcache, arena);\n\t\tif (p == NULL)\n\t\t\tgoto label_oom;\n\t}\n\tif (rsize != NULL)\n\t\t*rsize = usize;\n\n\t*ptr = p;\n\tif (config_stats) {\n\t\tassert(usize == isalloc(p, config_prof));\n\t\tthread_allocated_tsd_get()->allocated += usize;\n\t}\n\tUTRACE(0, size, p);\n\tJEMALLOC_VALGRIND_MALLOC(true, p, usize, zero);\n\treturn (ALLOCM_SUCCESS);\nlabel_oom:\n\tif (config_xmalloc && opt_xmalloc) {\n\t\tmalloc_write(\": Error in allocm(): \"\n\t\t \"out of memory\\n\");\n\t\tabort();\n\t}\n\t*ptr = NULL;\n\tUTRACE(0, size, 0);\n\treturn (ALLOCM_ERR_OOM);\n}\n\nint\nje_rallocm(void **ptr, size_t *rsize, size_t size, size_t extra, int flags)\n{\n\tvoid *p, *q;\n\tsize_t usize;\n\tsize_t old_size;\n\tsize_t old_rzsize JEMALLOC_CC_SILENCE_INIT(0);\n\tsize_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK)\n\t & (SIZE_T_MAX-1));\n\tbool zero = flags & ALLOCM_ZERO;\n\tbool no_move = flags & ALLOCM_NO_MOVE;\n\tunsigned arena_ind = ((unsigned)(flags >> 8)) - 1;\n\tbool try_tcache_alloc, try_tcache_dalloc;\n\tarena_t *arena;\n\n\tassert(ptr != NULL);\n\tassert(*ptr != NULL);\n\tassert(size != 0);\n\tassert(SIZE_T_MAX - size >= extra);\n\tassert(malloc_initialized || IS_INITIALIZER);\n\n\tif (arena_ind != UINT_MAX) {\n\t\tarena_chunk_t *chunk;\n\t\ttry_tcache_alloc = true;\n\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(*ptr);\n\t\ttry_tcache_dalloc = (chunk == *ptr || chunk->arena !=\n\t\t arenas[arena_ind]);\n\t\tarena = arenas[arena_ind];\n\t} else {\n\t\ttry_tcache_alloc = true;\n\t\ttry_tcache_dalloc = true;\n\t\tarena = NULL;\n\t}\n\n\tp = *ptr;\n\tif (config_prof && opt_prof) {\n\t\tprof_thr_cnt_t *cnt;\n\n\t\t/*\n\t\t * usize isn't knowable before iralloc() returns when extra is\n\t\t * non-zero. Therefore, compute its maximum possible value and\n\t\t * use that in PROF_ALLOC_PREP() to decide whether to capture a\n\t\t * backtrace. prof_realloc() will use the actual usize to\n\t\t * decide whether to sample.\n\t\t */\n\t\tsize_t max_usize = (alignment == 0) ? s2u(size+extra) :\n\t\t sa2u(size+extra, alignment);\n\t\tprof_ctx_t *old_ctx = prof_ctx_get(p);\n\t\told_size = isalloc(p, true);\n\t\tif (config_valgrind && opt_valgrind)\n\t\t\told_rzsize = p2rz(p);\n\t\tPROF_ALLOC_PREP(1, max_usize, cnt);\n\t\tif (cnt == NULL)\n\t\t\tgoto label_oom;\n\t\t/*\n\t\t * Use minimum usize to determine whether promotion may happen.\n\t\t */\n\t\tif (prof_promote && (uintptr_t)cnt != (uintptr_t)1U\n\t\t && ((alignment == 0) ? s2u(size) : sa2u(size, alignment))\n\t\t <= SMALL_MAXCLASS) {\n\t\t\tq = irallocx(p, SMALL_MAXCLASS+1, (SMALL_MAXCLASS+1 >=\n\t\t\t size+extra) ? 0 : size+extra - (SMALL_MAXCLASS+1),\n\t\t\t alignment, zero, no_move, try_tcache_alloc,\n\t\t\t try_tcache_dalloc, arena);\n\t\t\tif (q == NULL)\n\t\t\t\tgoto label_err;\n\t\t\tif (max_usize < PAGE) {\n\t\t\t\tusize = max_usize;\n\t\t\t\tarena_prof_promoted(q, usize);\n\t\t\t} else\n\t\t\t\tusize = isalloc(q, config_prof);\n\t\t} else {\n\t\t\tq = irallocx(p, size, extra, alignment, zero, no_move,\n\t\t\t try_tcache_alloc, try_tcache_dalloc, arena);\n\t\t\tif (q == NULL)\n\t\t\t\tgoto label_err;\n\t\t\tusize = isalloc(q, config_prof);\n\t\t}\n\t\tprof_realloc(q, usize, cnt, old_size, old_ctx);\n\t\tif (rsize != NULL)\n\t\t\t*rsize = usize;\n\t} else {\n\t\tif (config_stats) {\n\t\t\told_size = isalloc(p, false);\n\t\t\tif (config_valgrind && opt_valgrind)\n\t\t\t\told_rzsize = u2rz(old_size);\n\t\t} else if (config_valgrind && opt_valgrind) {\n\t\t\told_size = isalloc(p, false);\n\t\t\told_rzsize = u2rz(old_size);\n\t\t}\n\t\tq = irallocx(p, size, extra, alignment, zero, no_move,\n\t\t try_tcache_alloc, try_tcache_dalloc, arena);\n\t\tif (q == NULL)\n\t\t\tgoto label_err;\n\t\tif (config_stats)\n\t\t\tusize = isalloc(q, config_prof);\n\t\tif (rsize != NULL) {\n\t\t\tif (config_stats == false)\n\t\t\t\tusize = isalloc(q, config_prof);\n\t\t\t*rsize = usize;\n\t\t}\n\t}\n\n\t*ptr = q;\n\tif (config_stats) {\n\t\tthread_allocated_t *ta;\n\t\tta = thread_allocated_tsd_get();\n\t\tta->allocated += usize;\n\t\tta->deallocated += old_size;\n\t}\n\tUTRACE(p, size, q);\n\tJEMALLOC_VALGRIND_REALLOC(q, usize, p, old_size, old_rzsize, zero);\n\treturn (ALLOCM_SUCCESS);\nlabel_err:\n\tif (no_move) {\n\t\tUTRACE(p, size, q);\n\t\treturn (ALLOCM_ERR_NOT_MOVED);\n\t}\nlabel_oom:\n\tif (config_xmalloc && opt_xmalloc) {\n\t\tmalloc_write(\": Error in rallocm(): \"\n\t\t \"out of memory\\n\");\n\t\tabort();\n\t}\n\tUTRACE(p, size, 0);\n\treturn (ALLOCM_ERR_OOM);\n}\n\nint\nje_sallocm(const void *ptr, size_t *rsize, int flags)\n{\n\tsize_t sz;\n\n\tassert(malloc_initialized || IS_INITIALIZER);\n\n\tif (config_ivsalloc)\n\t\tsz = ivsalloc(ptr, config_prof);\n\telse {\n\t\tassert(ptr != NULL);\n\t\tsz = isalloc(ptr, config_prof);\n\t}\n\tassert(rsize != NULL);\n\t*rsize = sz;\n\n\treturn (ALLOCM_SUCCESS);\n}\n\nint\nje_dallocm(void *ptr, int flags)\n{\n\tsize_t usize;\n\tsize_t rzsize JEMALLOC_CC_SILENCE_INIT(0);\n\tunsigned arena_ind = ((unsigned)(flags >> 8)) - 1;\n\tbool try_tcache;\n\n\tassert(ptr != NULL);\n\tassert(malloc_initialized || IS_INITIALIZER);\n\n\tif (arena_ind != UINT_MAX) {\n\t\tarena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\t\ttry_tcache = (chunk == ptr || chunk->arena !=\n\t\t arenas[arena_ind]);\n\t} else\n\t\ttry_tcache = true;\n\n\tUTRACE(ptr, 0, 0);\n\tif (config_stats || config_valgrind)\n\t\tusize = isalloc(ptr, config_prof);\n\tif (config_prof && opt_prof) {\n\t\tif (config_stats == false && config_valgrind == false)\n\t\t\tusize = isalloc(ptr, config_prof);\n\t\tprof_free(ptr, usize);\n\t}\n\tif (config_stats)\n\t\tthread_allocated_tsd_get()->deallocated += usize;\n\tif (config_valgrind && opt_valgrind)\n\t\trzsize = p2rz(ptr);\n\tiqallocx(ptr, try_tcache);\n\tJEMALLOC_VALGRIND_FREE(ptr, rzsize);\n\n\treturn (ALLOCM_SUCCESS);\n}\n\nint\nje_nallocm(size_t *rsize, size_t size, int flags)\n{\n\tsize_t usize;\n\tsize_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK)\n\t & (SIZE_T_MAX-1));\n\n\tassert(size != 0);\n\n\tif (malloc_init())\n\t\treturn (ALLOCM_ERR_OOM);\n\n\tusize = (alignment == 0) ? s2u(size) : sa2u(size, alignment);\n\tif (usize == 0)\n\t\treturn (ALLOCM_ERR_OOM);\n\n\tif (rsize != NULL)\n\t\t*rsize = usize;\n\treturn (ALLOCM_SUCCESS);\n}\n\n#endif\n/*\n * End experimental functions.\n */\n/******************************************************************************/\n/*\n * The following functions are used by threading libraries for protection of\n * malloc during fork().\n */\n\n/*\n * If an application creates a thread before doing any allocation in the main\n * thread, then calls fork(2) in the main thread followed by memory allocation\n * in the child process, a race can occur that results in deadlock within the\n * child: the main thread may have forked while the created thread had\n * partially initialized the allocator. Ordinarily jemalloc prevents\n * fork/malloc races via the following functions it registers during\n * initialization using pthread_atfork(), but of course that does no good if\n * the allocator isn't fully initialized at fork time. The following library\n * constructor is a partial solution to this problem. It may still possible to\n * trigger the deadlock described above, but doing so would involve forking via\n * a library constructor that runs before jemalloc's runs.\n */\nJEMALLOC_ATTR(constructor)\nstatic void\njemalloc_constructor(void)\n{\n\n\tmalloc_init();\n}\n\n#ifndef JEMALLOC_MUTEX_INIT_CB\nvoid\njemalloc_prefork(void)\n#else\nJEMALLOC_EXPORT void\n_malloc_prefork(void)\n#endif\n{\n\tunsigned i;\n\n#ifdef JEMALLOC_MUTEX_INIT_CB\n\tif (malloc_initialized == false)\n\t\treturn;\n#endif\n\tassert(malloc_initialized);\n\n\t/* Acquire all mutexes in a safe order. */\n\tctl_prefork();\n\tmalloc_mutex_prefork(&arenas_lock);\n\tfor (i = 0; i < narenas_total; i++) {\n\t\tif (arenas[i] != NULL)\n\t\t\tarena_prefork(arenas[i]);\n\t}\n\tprof_prefork();\n\tchunk_prefork();\n\tbase_prefork();\n\thuge_prefork();\n}\n\n#ifndef JEMALLOC_MUTEX_INIT_CB\nvoid\njemalloc_postfork_parent(void)\n#else\nJEMALLOC_EXPORT void\n_malloc_postfork(void)\n#endif\n{\n\tunsigned i;\n\n#ifdef JEMALLOC_MUTEX_INIT_CB\n\tif (malloc_initialized == false)\n\t\treturn;\n#endif\n\tassert(malloc_initialized);\n\n\t/* Release all mutexes, now that fork() has completed. */\n\thuge_postfork_parent();\n\tbase_postfork_parent();\n\tchunk_postfork_parent();\n\tprof_postfork_parent();\n\tfor (i = 0; i < narenas_total; i++) {\n\t\tif (arenas[i] != NULL)\n\t\t\tarena_postfork_parent(arenas[i]);\n\t}\n\tmalloc_mutex_postfork_parent(&arenas_lock);\n\tctl_postfork_parent();\n}\n\nvoid\njemalloc_postfork_child(void)\n{\n\tunsigned i;\n\n\tassert(malloc_initialized);\n\n\t/* Release all mutexes, now that fork() has completed. */\n\thuge_postfork_child();\n\tbase_postfork_child();\n\tchunk_postfork_child();\n\tprof_postfork_child();\n\tfor (i = 0; i < narenas_total; i++) {\n\t\tif (arenas[i] != NULL)\n\t\t\tarena_postfork_child(arenas[i]);\n\t}\n\tmalloc_mutex_postfork_child(&arenas_lock);\n\tctl_postfork_child();\n}\n\n/******************************************************************************/\n/*\n * The following functions are used for TLS allocation/deallocation in static\n * binaries on FreeBSD. The primary difference between these and i[mcd]alloc()\n * is that these avoid accessing TLS variables.\n */\n\nstatic void *\na0alloc(size_t size, bool zero)\n{\n\n\tif (malloc_init())\n\t\treturn (NULL);\n\n\tif (size == 0)\n\t\tsize = 1;\n\n\tif (size <= arena_maxclass)\n\t\treturn (arena_malloc(arenas[0], size, zero, false));\n\telse\n\t\treturn (huge_malloc(size, zero));\n}\n\nvoid *\na0malloc(size_t size)\n{\n\n\treturn (a0alloc(size, false));\n}\n\nvoid *\na0calloc(size_t num, size_t size)\n{\n\n\treturn (a0alloc(num * size, true));\n}\n\nvoid\na0free(void *ptr)\n{\n\tarena_chunk_t *chunk;\n\n\tif (ptr == NULL)\n\t\treturn;\n\n\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\tif (chunk != ptr)\n\t\tarena_dalloc(chunk->arena, chunk, ptr, false);\n\telse\n\t\thuge_dalloc(ptr, true);\n}\n\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/src/mb.c", "content": "#define\tJEMALLOC_MB_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n" }, { "path": "deps/jemalloc/src/mutex.c", "content": "#define\tJEMALLOC_MUTEX_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)\n#include \n#endif\n\n#ifndef _CRT_SPINCOUNT\n#define _CRT_SPINCOUNT 4000\n#endif\n\n/******************************************************************************/\n/* Data. */\n\n#ifdef JEMALLOC_LAZY_LOCK\nbool isthreaded = false;\n#endif\n#ifdef JEMALLOC_MUTEX_INIT_CB\nstatic bool\t\tpostpone_init = true;\nstatic malloc_mutex_t\t*postponed_mutexes = NULL;\n#endif\n\n#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)\nstatic void\tpthread_create_once(void);\n#endif\n\n/******************************************************************************/\n/*\n * We intercept pthread_create() calls in order to toggle isthreaded if the\n * process goes multi-threaded.\n */\n\n#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32)\nstatic int (*pthread_create_fptr)(pthread_t *__restrict, const pthread_attr_t *,\n void *(*)(void *), void *__restrict);\n\nstatic void\npthread_create_once(void)\n{\n\n\tpthread_create_fptr = dlsym(RTLD_NEXT, \"pthread_create\");\n\tif (pthread_create_fptr == NULL) {\n\t\tmalloc_write(\": Error in dlsym(RTLD_NEXT, \"\n\t\t \"\\\"pthread_create\\\")\\n\");\n\t\tabort();\n\t}\n\n\tisthreaded = true;\n}\n\nJEMALLOC_EXPORT int\npthread_create(pthread_t *__restrict thread,\n const pthread_attr_t *__restrict attr, void *(*start_routine)(void *),\n void *__restrict arg)\n{\n\tstatic pthread_once_t once_control = PTHREAD_ONCE_INIT;\n\n\tpthread_once(&once_control, pthread_create_once);\n\n\treturn (pthread_create_fptr(thread, attr, start_routine, arg));\n}\n#endif\n\n/******************************************************************************/\n\n#ifdef JEMALLOC_MUTEX_INIT_CB\nJEMALLOC_EXPORT int\t_pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex,\n void *(calloc_cb)(size_t, size_t));\n#endif\n\nbool\nmalloc_mutex_init(malloc_mutex_t *mutex)\n{\n\n#ifdef _WIN32\n\tif (!InitializeCriticalSectionAndSpinCount(&mutex->lock,\n\t _CRT_SPINCOUNT))\n\t\treturn (true);\n#elif (defined(JEMALLOC_OSSPIN))\n\tmutex->lock = 0;\n#elif (defined(JEMALLOC_MUTEX_INIT_CB))\n\tif (postpone_init) {\n\t\tmutex->postponed_next = postponed_mutexes;\n\t\tpostponed_mutexes = mutex;\n\t} else {\n\t\tif (_pthread_mutex_init_calloc_cb(&mutex->lock, base_calloc) !=\n\t\t 0)\n\t\t\treturn (true);\n\t}\n#else\n\tpthread_mutexattr_t attr;\n\n\tif (pthread_mutexattr_init(&attr) != 0)\n\t\treturn (true);\n\tpthread_mutexattr_settype(&attr, MALLOC_MUTEX_TYPE);\n\tif (pthread_mutex_init(&mutex->lock, &attr) != 0) {\n\t\tpthread_mutexattr_destroy(&attr);\n\t\treturn (true);\n\t}\n\tpthread_mutexattr_destroy(&attr);\n#endif\n\treturn (false);\n}\n\nvoid\nmalloc_mutex_prefork(malloc_mutex_t *mutex)\n{\n\n\tmalloc_mutex_lock(mutex);\n}\n\nvoid\nmalloc_mutex_postfork_parent(malloc_mutex_t *mutex)\n{\n\n\tmalloc_mutex_unlock(mutex);\n}\n\nvoid\nmalloc_mutex_postfork_child(malloc_mutex_t *mutex)\n{\n\n#ifdef JEMALLOC_MUTEX_INIT_CB\n\tmalloc_mutex_unlock(mutex);\n#else\n\tif (malloc_mutex_init(mutex)) {\n\t\tmalloc_printf(\": Error re-initializing mutex in \"\n\t\t \"child\\n\");\n\t\tif (opt_abort)\n\t\t\tabort();\n\t}\n#endif\n}\n\nbool\nmutex_boot(void)\n{\n\n#ifdef JEMALLOC_MUTEX_INIT_CB\n\tpostpone_init = false;\n\twhile (postponed_mutexes != NULL) {\n\t\tif (_pthread_mutex_init_calloc_cb(&postponed_mutexes->lock,\n\t\t base_calloc) != 0)\n\t\t\treturn (true);\n\t\tpostponed_mutexes = postponed_mutexes->postponed_next;\n\t}\n#endif\n\treturn (false);\n}\n" }, { "path": "deps/jemalloc/src/prof.c", "content": "#define\tJEMALLOC_PROF_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n/******************************************************************************/\n\n#ifdef JEMALLOC_PROF_LIBUNWIND\n#define\tUNW_LOCAL_ONLY\n#include \n#endif\n\n#ifdef JEMALLOC_PROF_LIBGCC\n#include \n#endif\n\n/******************************************************************************/\n/* Data. */\n\nmalloc_tsd_data(, prof_tdata, prof_tdata_t *, NULL)\n\nbool\t\topt_prof = false;\nbool\t\topt_prof_active = true;\nsize_t\t\topt_lg_prof_sample = LG_PROF_SAMPLE_DEFAULT;\nssize_t\t\topt_lg_prof_interval = LG_PROF_INTERVAL_DEFAULT;\nbool\t\topt_prof_gdump = false;\nbool\t\topt_prof_final = true;\nbool\t\topt_prof_leak = false;\nbool\t\topt_prof_accum = false;\nchar\t\topt_prof_prefix[PATH_MAX + 1];\n\nuint64_t\tprof_interval;\nbool\t\tprof_promote;\n\n/*\n * Table of mutexes that are shared among ctx's. These are leaf locks, so\n * there is no problem with using them for more than one ctx at the same time.\n * The primary motivation for this sharing though is that ctx's are ephemeral,\n * and destroying mutexes causes complications for systems that allocate when\n * creating/destroying mutexes.\n */\nstatic malloc_mutex_t\t*ctx_locks;\nstatic unsigned\t\tcum_ctxs; /* Atomic counter. */\n\n/*\n * Global hash of (prof_bt_t *)-->(prof_ctx_t *). This is the master data\n * structure that knows about all backtraces currently captured.\n */\nstatic ckh_t\t\tbt2ctx;\nstatic malloc_mutex_t\tbt2ctx_mtx;\n\nstatic malloc_mutex_t\tprof_dump_seq_mtx;\nstatic uint64_t\t\tprof_dump_seq;\nstatic uint64_t\t\tprof_dump_iseq;\nstatic uint64_t\t\tprof_dump_mseq;\nstatic uint64_t\t\tprof_dump_useq;\n\n/*\n * This buffer is rather large for stack allocation, so use a single buffer for\n * all profile dumps. The buffer is implicitly protected by bt2ctx_mtx, since\n * it must be locked anyway during dumping.\n */\nstatic char\t\tprof_dump_buf[PROF_DUMP_BUFSIZE];\nstatic unsigned\t\tprof_dump_buf_end;\nstatic int\t\tprof_dump_fd;\n\n/* Do not dump any profiles until bootstrapping is complete. */\nstatic bool\t\tprof_booted = false;\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic prof_bt_t\t*bt_dup(prof_bt_t *bt);\nstatic void\tbt_destroy(prof_bt_t *bt);\n#ifdef JEMALLOC_PROF_LIBGCC\nstatic _Unwind_Reason_Code\tprof_unwind_init_callback(\n struct _Unwind_Context *context, void *arg);\nstatic _Unwind_Reason_Code\tprof_unwind_callback(\n struct _Unwind_Context *context, void *arg);\n#endif\nstatic bool\tprof_flush(bool propagate_err);\nstatic bool\tprof_write(bool propagate_err, const char *s);\nstatic bool\tprof_printf(bool propagate_err, const char *format, ...)\n JEMALLOC_ATTR(format(printf, 2, 3));\nstatic void\tprof_ctx_sum(prof_ctx_t *ctx, prof_cnt_t *cnt_all,\n size_t *leak_nctx);\nstatic void\tprof_ctx_destroy(prof_ctx_t *ctx);\nstatic void\tprof_ctx_merge(prof_ctx_t *ctx, prof_thr_cnt_t *cnt);\nstatic bool\tprof_dump_ctx(bool propagate_err, prof_ctx_t *ctx,\n prof_bt_t *bt);\nstatic bool\tprof_dump_maps(bool propagate_err);\nstatic bool\tprof_dump(bool propagate_err, const char *filename,\n bool leakcheck);\nstatic void\tprof_dump_filename(char *filename, char v, int64_t vseq);\nstatic void\tprof_fdump(void);\nstatic void\tprof_bt_hash(const void *key, unsigned minbits, size_t *hash1,\n size_t *hash2);\nstatic bool\tprof_bt_keycomp(const void *k1, const void *k2);\nstatic malloc_mutex_t\t*prof_ctx_mutex_choose(void);\n\n/******************************************************************************/\n\nvoid\nbt_init(prof_bt_t *bt, void **vec)\n{\n\n\tcassert(config_prof);\n\n\tbt->vec = vec;\n\tbt->len = 0;\n}\n\nstatic void\nbt_destroy(prof_bt_t *bt)\n{\n\n\tcassert(config_prof);\n\n\tidalloc(bt);\n}\n\nstatic prof_bt_t *\nbt_dup(prof_bt_t *bt)\n{\n\tprof_bt_t *ret;\n\n\tcassert(config_prof);\n\n\t/*\n\t * Create a single allocation that has space for vec immediately\n\t * following the prof_bt_t structure. The backtraces that get\n\t * stored in the backtrace caches are copied from stack-allocated\n\t * temporary variables, so size is known at creation time. Making this\n\t * a contiguous object improves cache locality.\n\t */\n\tret = (prof_bt_t *)imalloc(QUANTUM_CEILING(sizeof(prof_bt_t)) +\n\t (bt->len * sizeof(void *)));\n\tif (ret == NULL)\n\t\treturn (NULL);\n\tret->vec = (void **)((uintptr_t)ret +\n\t QUANTUM_CEILING(sizeof(prof_bt_t)));\n\tmemcpy(ret->vec, bt->vec, bt->len * sizeof(void *));\n\tret->len = bt->len;\n\n\treturn (ret);\n}\n\nstatic inline void\nprof_enter(prof_tdata_t *prof_tdata)\n{\n\n\tcassert(config_prof);\n\n\tassert(prof_tdata->enq == false);\n\tprof_tdata->enq = true;\n\n\tmalloc_mutex_lock(&bt2ctx_mtx);\n}\n\nstatic inline void\nprof_leave(prof_tdata_t *prof_tdata)\n{\n\tbool idump, gdump;\n\n\tcassert(config_prof);\n\n\tmalloc_mutex_unlock(&bt2ctx_mtx);\n\n\tassert(prof_tdata->enq);\n\tprof_tdata->enq = false;\n\tidump = prof_tdata->enq_idump;\n\tprof_tdata->enq_idump = false;\n\tgdump = prof_tdata->enq_gdump;\n\tprof_tdata->enq_gdump = false;\n\n\tif (idump)\n\t\tprof_idump();\n\tif (gdump)\n\t\tprof_gdump();\n}\n\n#ifdef JEMALLOC_PROF_LIBUNWIND\nvoid\nprof_backtrace(prof_bt_t *bt, unsigned nignore)\n{\n\tunw_context_t uc;\n\tunw_cursor_t cursor;\n\tunsigned i;\n\tint err;\n\n\tcassert(config_prof);\n\tassert(bt->len == 0);\n\tassert(bt->vec != NULL);\n\n\tunw_getcontext(&uc);\n\tunw_init_local(&cursor, &uc);\n\n\t/* Throw away (nignore+1) stack frames, if that many exist. */\n\tfor (i = 0; i < nignore + 1; i++) {\n\t\terr = unw_step(&cursor);\n\t\tif (err <= 0)\n\t\t\treturn;\n\t}\n\n\t/*\n\t * Iterate over stack frames until there are no more, or until no space\n\t * remains in bt.\n\t */\n\tfor (i = 0; i < PROF_BT_MAX; i++) {\n\t\tunw_get_reg(&cursor, UNW_REG_IP, (unw_word_t *)&bt->vec[i]);\n\t\tbt->len++;\n\t\terr = unw_step(&cursor);\n\t\tif (err <= 0)\n\t\t\tbreak;\n\t}\n}\n#elif (defined(JEMALLOC_PROF_LIBGCC))\nstatic _Unwind_Reason_Code\nprof_unwind_init_callback(struct _Unwind_Context *context, void *arg)\n{\n\n\tcassert(config_prof);\n\n\treturn (_URC_NO_REASON);\n}\n\nstatic _Unwind_Reason_Code\nprof_unwind_callback(struct _Unwind_Context *context, void *arg)\n{\n\tprof_unwind_data_t *data = (prof_unwind_data_t *)arg;\n\n\tcassert(config_prof);\n\n\tif (data->nignore > 0)\n\t\tdata->nignore--;\n\telse {\n\t\tdata->bt->vec[data->bt->len] = (void *)_Unwind_GetIP(context);\n\t\tdata->bt->len++;\n\t\tif (data->bt->len == data->max)\n\t\t\treturn (_URC_END_OF_STACK);\n\t}\n\n\treturn (_URC_NO_REASON);\n}\n\nvoid\nprof_backtrace(prof_bt_t *bt, unsigned nignore)\n{\n\tprof_unwind_data_t data = {bt, nignore, PROF_BT_MAX};\n\n\tcassert(config_prof);\n\n\t_Unwind_Backtrace(prof_unwind_callback, &data);\n}\n#elif (defined(JEMALLOC_PROF_GCC))\nvoid\nprof_backtrace(prof_bt_t *bt, unsigned nignore)\n{\n#define\tBT_FRAME(i)\t\t\t\t\t\t\t\\\n\tif ((i) < nignore + PROF_BT_MAX) {\t\t\t\t\\\n\t\tvoid *p;\t\t\t\t\t\t\\\n\t\tif (__builtin_frame_address(i) == 0)\t\t\t\\\n\t\t\treturn;\t\t\t\t\t\t\\\n\t\tp = __builtin_return_address(i);\t\t\t\\\n\t\tif (p == NULL)\t\t\t\t\t\t\\\n\t\t\treturn;\t\t\t\t\t\t\\\n\t\tif (i >= nignore) {\t\t\t\t\t\\\n\t\t\tbt->vec[(i) - nignore] = p;\t\t\t\\\n\t\t\tbt->len = (i) - nignore + 1;\t\t\t\\\n\t\t}\t\t\t\t\t\t\t\\\n\t} else\t\t\t\t\t\t\t\t\\\n\t\treturn;\n\n\tcassert(config_prof);\n\tassert(nignore <= 3);\n\n\tBT_FRAME(0)\n\tBT_FRAME(1)\n\tBT_FRAME(2)\n\tBT_FRAME(3)\n\tBT_FRAME(4)\n\tBT_FRAME(5)\n\tBT_FRAME(6)\n\tBT_FRAME(7)\n\tBT_FRAME(8)\n\tBT_FRAME(9)\n\n\tBT_FRAME(10)\n\tBT_FRAME(11)\n\tBT_FRAME(12)\n\tBT_FRAME(13)\n\tBT_FRAME(14)\n\tBT_FRAME(15)\n\tBT_FRAME(16)\n\tBT_FRAME(17)\n\tBT_FRAME(18)\n\tBT_FRAME(19)\n\n\tBT_FRAME(20)\n\tBT_FRAME(21)\n\tBT_FRAME(22)\n\tBT_FRAME(23)\n\tBT_FRAME(24)\n\tBT_FRAME(25)\n\tBT_FRAME(26)\n\tBT_FRAME(27)\n\tBT_FRAME(28)\n\tBT_FRAME(29)\n\n\tBT_FRAME(30)\n\tBT_FRAME(31)\n\tBT_FRAME(32)\n\tBT_FRAME(33)\n\tBT_FRAME(34)\n\tBT_FRAME(35)\n\tBT_FRAME(36)\n\tBT_FRAME(37)\n\tBT_FRAME(38)\n\tBT_FRAME(39)\n\n\tBT_FRAME(40)\n\tBT_FRAME(41)\n\tBT_FRAME(42)\n\tBT_FRAME(43)\n\tBT_FRAME(44)\n\tBT_FRAME(45)\n\tBT_FRAME(46)\n\tBT_FRAME(47)\n\tBT_FRAME(48)\n\tBT_FRAME(49)\n\n\tBT_FRAME(50)\n\tBT_FRAME(51)\n\tBT_FRAME(52)\n\tBT_FRAME(53)\n\tBT_FRAME(54)\n\tBT_FRAME(55)\n\tBT_FRAME(56)\n\tBT_FRAME(57)\n\tBT_FRAME(58)\n\tBT_FRAME(59)\n\n\tBT_FRAME(60)\n\tBT_FRAME(61)\n\tBT_FRAME(62)\n\tBT_FRAME(63)\n\tBT_FRAME(64)\n\tBT_FRAME(65)\n\tBT_FRAME(66)\n\tBT_FRAME(67)\n\tBT_FRAME(68)\n\tBT_FRAME(69)\n\n\tBT_FRAME(70)\n\tBT_FRAME(71)\n\tBT_FRAME(72)\n\tBT_FRAME(73)\n\tBT_FRAME(74)\n\tBT_FRAME(75)\n\tBT_FRAME(76)\n\tBT_FRAME(77)\n\tBT_FRAME(78)\n\tBT_FRAME(79)\n\n\tBT_FRAME(80)\n\tBT_FRAME(81)\n\tBT_FRAME(82)\n\tBT_FRAME(83)\n\tBT_FRAME(84)\n\tBT_FRAME(85)\n\tBT_FRAME(86)\n\tBT_FRAME(87)\n\tBT_FRAME(88)\n\tBT_FRAME(89)\n\n\tBT_FRAME(90)\n\tBT_FRAME(91)\n\tBT_FRAME(92)\n\tBT_FRAME(93)\n\tBT_FRAME(94)\n\tBT_FRAME(95)\n\tBT_FRAME(96)\n\tBT_FRAME(97)\n\tBT_FRAME(98)\n\tBT_FRAME(99)\n\n\tBT_FRAME(100)\n\tBT_FRAME(101)\n\tBT_FRAME(102)\n\tBT_FRAME(103)\n\tBT_FRAME(104)\n\tBT_FRAME(105)\n\tBT_FRAME(106)\n\tBT_FRAME(107)\n\tBT_FRAME(108)\n\tBT_FRAME(109)\n\n\tBT_FRAME(110)\n\tBT_FRAME(111)\n\tBT_FRAME(112)\n\tBT_FRAME(113)\n\tBT_FRAME(114)\n\tBT_FRAME(115)\n\tBT_FRAME(116)\n\tBT_FRAME(117)\n\tBT_FRAME(118)\n\tBT_FRAME(119)\n\n\tBT_FRAME(120)\n\tBT_FRAME(121)\n\tBT_FRAME(122)\n\tBT_FRAME(123)\n\tBT_FRAME(124)\n\tBT_FRAME(125)\n\tBT_FRAME(126)\n\tBT_FRAME(127)\n\n\t/* Extras to compensate for nignore. */\n\tBT_FRAME(128)\n\tBT_FRAME(129)\n\tBT_FRAME(130)\n#undef BT_FRAME\n}\n#else\nvoid\nprof_backtrace(prof_bt_t *bt, unsigned nignore)\n{\n\n\tcassert(config_prof);\n\tassert(false);\n}\n#endif\n\nprof_thr_cnt_t *\nprof_lookup(prof_bt_t *bt)\n{\n\tunion {\n\t\tprof_thr_cnt_t\t*p;\n\t\tvoid\t\t*v;\n\t} ret;\n\tprof_tdata_t *prof_tdata;\n\n\tcassert(config_prof);\n\n\tprof_tdata = prof_tdata_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)\n\t\treturn (NULL);\n\n\tif (ckh_search(&prof_tdata->bt2cnt, bt, NULL, &ret.v)) {\n\t\tunion {\n\t\t\tprof_bt_t\t*p;\n\t\t\tvoid\t\t*v;\n\t\t} btkey;\n\t\tunion {\n\t\t\tprof_ctx_t\t*p;\n\t\t\tvoid\t\t*v;\n\t\t} ctx;\n\t\tbool new_ctx;\n\n\t\t/*\n\t\t * This thread's cache lacks bt. Look for it in the global\n\t\t * cache.\n\t\t */\n\t\tprof_enter(prof_tdata);\n\t\tif (ckh_search(&bt2ctx, bt, &btkey.v, &ctx.v)) {\n\t\t\t/* bt has never been seen before. Insert it. */\n\t\t\tctx.v = imalloc(sizeof(prof_ctx_t));\n\t\t\tif (ctx.v == NULL) {\n\t\t\t\tprof_leave(prof_tdata);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tbtkey.p = bt_dup(bt);\n\t\t\tif (btkey.v == NULL) {\n\t\t\t\tprof_leave(prof_tdata);\n\t\t\t\tidalloc(ctx.v);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tctx.p->bt = btkey.p;\n\t\t\tctx.p->lock = prof_ctx_mutex_choose();\n\t\t\t/*\n\t\t\t * Set nlimbo to 1, in order to avoid a race condition\n\t\t\t * with prof_ctx_merge()/prof_ctx_destroy().\n\t\t\t */\n\t\t\tctx.p->nlimbo = 1;\n\t\t\tmemset(&ctx.p->cnt_merged, 0, sizeof(prof_cnt_t));\n\t\t\tql_new(&ctx.p->cnts_ql);\n\t\t\tif (ckh_insert(&bt2ctx, btkey.v, ctx.v)) {\n\t\t\t\t/* OOM. */\n\t\t\t\tprof_leave(prof_tdata);\n\t\t\t\tidalloc(btkey.v);\n\t\t\t\tidalloc(ctx.v);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tnew_ctx = true;\n\t\t} else {\n\t\t\t/*\n\t\t\t * Increment nlimbo, in order to avoid a race condition\n\t\t\t * with prof_ctx_merge()/prof_ctx_destroy().\n\t\t\t */\n\t\t\tmalloc_mutex_lock(ctx.p->lock);\n\t\t\tctx.p->nlimbo++;\n\t\t\tmalloc_mutex_unlock(ctx.p->lock);\n\t\t\tnew_ctx = false;\n\t\t}\n\t\tprof_leave(prof_tdata);\n\n\t\t/* Link a prof_thd_cnt_t into ctx for this thread. */\n\t\tif (ckh_count(&prof_tdata->bt2cnt) == PROF_TCMAX) {\n\t\t\tassert(ckh_count(&prof_tdata->bt2cnt) > 0);\n\t\t\t/*\n\t\t\t * Flush the least recently used cnt in order to keep\n\t\t\t * bt2cnt from becoming too large.\n\t\t\t */\n\t\t\tret.p = ql_last(&prof_tdata->lru_ql, lru_link);\n\t\t\tassert(ret.v != NULL);\n\t\t\tif (ckh_remove(&prof_tdata->bt2cnt, ret.p->ctx->bt,\n\t\t\t NULL, NULL))\n\t\t\t\tassert(false);\n\t\t\tql_remove(&prof_tdata->lru_ql, ret.p, lru_link);\n\t\t\tprof_ctx_merge(ret.p->ctx, ret.p);\n\t\t\t/* ret can now be re-used. */\n\t\t} else {\n\t\t\tassert(ckh_count(&prof_tdata->bt2cnt) < PROF_TCMAX);\n\t\t\t/* Allocate and partially initialize a new cnt. */\n\t\t\tret.v = imalloc(sizeof(prof_thr_cnt_t));\n\t\t\tif (ret.p == NULL) {\n\t\t\t\tif (new_ctx)\n\t\t\t\t\tprof_ctx_destroy(ctx.p);\n\t\t\t\treturn (NULL);\n\t\t\t}\n\t\t\tql_elm_new(ret.p, cnts_link);\n\t\t\tql_elm_new(ret.p, lru_link);\n\t\t}\n\t\t/* Finish initializing ret. */\n\t\tret.p->ctx = ctx.p;\n\t\tret.p->epoch = 0;\n\t\tmemset(&ret.p->cnts, 0, sizeof(prof_cnt_t));\n\t\tif (ckh_insert(&prof_tdata->bt2cnt, btkey.v, ret.v)) {\n\t\t\tif (new_ctx)\n\t\t\t\tprof_ctx_destroy(ctx.p);\n\t\t\tidalloc(ret.v);\n\t\t\treturn (NULL);\n\t\t}\n\t\tql_head_insert(&prof_tdata->lru_ql, ret.p, lru_link);\n\t\tmalloc_mutex_lock(ctx.p->lock);\n\t\tql_tail_insert(&ctx.p->cnts_ql, ret.p, cnts_link);\n\t\tctx.p->nlimbo--;\n\t\tmalloc_mutex_unlock(ctx.p->lock);\n\t} else {\n\t\t/* Move ret to the front of the LRU. */\n\t\tql_remove(&prof_tdata->lru_ql, ret.p, lru_link);\n\t\tql_head_insert(&prof_tdata->lru_ql, ret.p, lru_link);\n\t}\n\n\treturn (ret.p);\n}\n\nstatic bool\nprof_flush(bool propagate_err)\n{\n\tbool ret = false;\n\tssize_t err;\n\n\tcassert(config_prof);\n\n\terr = write(prof_dump_fd, prof_dump_buf, prof_dump_buf_end);\n\tif (err == -1) {\n\t\tif (propagate_err == false) {\n\t\t\tmalloc_write(\": write() failed during heap \"\n\t\t\t \"profile flush\\n\");\n\t\t\tif (opt_abort)\n\t\t\t\tabort();\n\t\t}\n\t\tret = true;\n\t}\n\tprof_dump_buf_end = 0;\n\n\treturn (ret);\n}\n\nstatic bool\nprof_write(bool propagate_err, const char *s)\n{\n\tunsigned i, slen, n;\n\n\tcassert(config_prof);\n\n\ti = 0;\n\tslen = strlen(s);\n\twhile (i < slen) {\n\t\t/* Flush the buffer if it is full. */\n\t\tif (prof_dump_buf_end == PROF_DUMP_BUFSIZE)\n\t\t\tif (prof_flush(propagate_err) && propagate_err)\n\t\t\t\treturn (true);\n\n\t\tif (prof_dump_buf_end + slen <= PROF_DUMP_BUFSIZE) {\n\t\t\t/* Finish writing. */\n\t\t\tn = slen - i;\n\t\t} else {\n\t\t\t/* Write as much of s as will fit. */\n\t\t\tn = PROF_DUMP_BUFSIZE - prof_dump_buf_end;\n\t\t}\n\t\tmemcpy(&prof_dump_buf[prof_dump_buf_end], &s[i], n);\n\t\tprof_dump_buf_end += n;\n\t\ti += n;\n\t}\n\n\treturn (false);\n}\n\nJEMALLOC_ATTR(format(printf, 2, 3))\nstatic bool\nprof_printf(bool propagate_err, const char *format, ...)\n{\n\tbool ret;\n\tva_list ap;\n\tchar buf[PROF_PRINTF_BUFSIZE];\n\n\tva_start(ap, format);\n\tmalloc_vsnprintf(buf, sizeof(buf), format, ap);\n\tva_end(ap);\n\tret = prof_write(propagate_err, buf);\n\n\treturn (ret);\n}\n\nstatic void\nprof_ctx_sum(prof_ctx_t *ctx, prof_cnt_t *cnt_all, size_t *leak_nctx)\n{\n\tprof_thr_cnt_t *thr_cnt;\n\tprof_cnt_t tcnt;\n\n\tcassert(config_prof);\n\n\tmalloc_mutex_lock(ctx->lock);\n\n\tmemcpy(&ctx->cnt_summed, &ctx->cnt_merged, sizeof(prof_cnt_t));\n\tql_foreach(thr_cnt, &ctx->cnts_ql, cnts_link) {\n\t\tvolatile unsigned *epoch = &thr_cnt->epoch;\n\n\t\twhile (true) {\n\t\t\tunsigned epoch0 = *epoch;\n\n\t\t\t/* Make sure epoch is even. */\n\t\t\tif (epoch0 & 1U)\n\t\t\t\tcontinue;\n\n\t\t\tmemcpy(&tcnt, &thr_cnt->cnts, sizeof(prof_cnt_t));\n\n\t\t\t/* Terminate if epoch didn't change while reading. */\n\t\t\tif (*epoch == epoch0)\n\t\t\t\tbreak;\n\t\t}\n\n\t\tctx->cnt_summed.curobjs += tcnt.curobjs;\n\t\tctx->cnt_summed.curbytes += tcnt.curbytes;\n\t\tif (opt_prof_accum) {\n\t\t\tctx->cnt_summed.accumobjs += tcnt.accumobjs;\n\t\t\tctx->cnt_summed.accumbytes += tcnt.accumbytes;\n\t\t}\n\t}\n\n\tif (ctx->cnt_summed.curobjs != 0)\n\t\t(*leak_nctx)++;\n\n\t/* Add to cnt_all. */\n\tcnt_all->curobjs += ctx->cnt_summed.curobjs;\n\tcnt_all->curbytes += ctx->cnt_summed.curbytes;\n\tif (opt_prof_accum) {\n\t\tcnt_all->accumobjs += ctx->cnt_summed.accumobjs;\n\t\tcnt_all->accumbytes += ctx->cnt_summed.accumbytes;\n\t}\n\n\tmalloc_mutex_unlock(ctx->lock);\n}\n\nstatic void\nprof_ctx_destroy(prof_ctx_t *ctx)\n{\n\tprof_tdata_t *prof_tdata;\n\n\tcassert(config_prof);\n\n\t/*\n\t * Check that ctx is still unused by any thread cache before destroying\n\t * it. prof_lookup() increments ctx->nlimbo in order to avoid a race\n\t * condition with this function, as does prof_ctx_merge() in order to\n\t * avoid a race between the main body of prof_ctx_merge() and entry\n\t * into this function.\n\t */\n\tprof_tdata = *prof_tdata_tsd_get();\n\tassert((uintptr_t)prof_tdata > (uintptr_t)PROF_TDATA_STATE_MAX);\n\tprof_enter(prof_tdata);\n\tmalloc_mutex_lock(ctx->lock);\n\tif (ql_first(&ctx->cnts_ql) == NULL && ctx->cnt_merged.curobjs == 0 &&\n\t ctx->nlimbo == 1) {\n\t\tassert(ctx->cnt_merged.curbytes == 0);\n\t\tassert(ctx->cnt_merged.accumobjs == 0);\n\t\tassert(ctx->cnt_merged.accumbytes == 0);\n\t\t/* Remove ctx from bt2ctx. */\n\t\tif (ckh_remove(&bt2ctx, ctx->bt, NULL, NULL))\n\t\t\tassert(false);\n\t\tprof_leave(prof_tdata);\n\t\t/* Destroy ctx. */\n\t\tmalloc_mutex_unlock(ctx->lock);\n\t\tbt_destroy(ctx->bt);\n\t\tidalloc(ctx);\n\t} else {\n\t\t/*\n\t\t * Compensate for increment in prof_ctx_merge() or\n\t\t * prof_lookup().\n\t\t */\n\t\tctx->nlimbo--;\n\t\tmalloc_mutex_unlock(ctx->lock);\n\t\tprof_leave(prof_tdata);\n\t}\n}\n\nstatic void\nprof_ctx_merge(prof_ctx_t *ctx, prof_thr_cnt_t *cnt)\n{\n\tbool destroy;\n\n\tcassert(config_prof);\n\n\t/* Merge cnt stats and detach from ctx. */\n\tmalloc_mutex_lock(ctx->lock);\n\tctx->cnt_merged.curobjs += cnt->cnts.curobjs;\n\tctx->cnt_merged.curbytes += cnt->cnts.curbytes;\n\tctx->cnt_merged.accumobjs += cnt->cnts.accumobjs;\n\tctx->cnt_merged.accumbytes += cnt->cnts.accumbytes;\n\tql_remove(&ctx->cnts_ql, cnt, cnts_link);\n\tif (opt_prof_accum == false && ql_first(&ctx->cnts_ql) == NULL &&\n\t ctx->cnt_merged.curobjs == 0 && ctx->nlimbo == 0) {\n\t\t/*\n\t\t * Increment ctx->nlimbo in order to keep another thread from\n\t\t * winning the race to destroy ctx while this one has ctx->lock\n\t\t * dropped. Without this, it would be possible for another\n\t\t * thread to:\n\t\t *\n\t\t * 1) Sample an allocation associated with ctx.\n\t\t * 2) Deallocate the sampled object.\n\t\t * 3) Successfully prof_ctx_destroy(ctx).\n\t\t *\n\t\t * The result would be that ctx no longer exists by the time\n\t\t * this thread accesses it in prof_ctx_destroy().\n\t\t */\n\t\tctx->nlimbo++;\n\t\tdestroy = true;\n\t} else\n\t\tdestroy = false;\n\tmalloc_mutex_unlock(ctx->lock);\n\tif (destroy)\n\t\tprof_ctx_destroy(ctx);\n}\n\nstatic bool\nprof_dump_ctx(bool propagate_err, prof_ctx_t *ctx, prof_bt_t *bt)\n{\n\tunsigned i;\n\n\tcassert(config_prof);\n\n\t/*\n\t * Current statistics can sum to 0 as a result of unmerged per thread\n\t * statistics. Additionally, interval- and growth-triggered dumps can\n\t * occur between the time a ctx is created and when its statistics are\n\t * filled in. Avoid dumping any ctx that is an artifact of either\n\t * implementation detail.\n\t */\n\tif ((opt_prof_accum == false && ctx->cnt_summed.curobjs == 0) ||\n\t (opt_prof_accum && ctx->cnt_summed.accumobjs == 0)) {\n\t\tassert(ctx->cnt_summed.curobjs == 0);\n\t\tassert(ctx->cnt_summed.curbytes == 0);\n\t\tassert(ctx->cnt_summed.accumobjs == 0);\n\t\tassert(ctx->cnt_summed.accumbytes == 0);\n\t\treturn (false);\n\t}\n\n\tif (prof_printf(propagate_err, \"%\"PRId64\": %\"PRId64\n\t \" [%\"PRIu64\": %\"PRIu64\"] @\",\n\t ctx->cnt_summed.curobjs, ctx->cnt_summed.curbytes,\n\t ctx->cnt_summed.accumobjs, ctx->cnt_summed.accumbytes))\n\t\treturn (true);\n\n\tfor (i = 0; i < bt->len; i++) {\n\t\tif (prof_printf(propagate_err, \" %#\"PRIxPTR,\n\t\t (uintptr_t)bt->vec[i]))\n\t\t\treturn (true);\n\t}\n\n\tif (prof_write(propagate_err, \"\\n\"))\n\t\treturn (true);\n\n\treturn (false);\n}\n\nstatic bool\nprof_dump_maps(bool propagate_err)\n{\n\tint mfd;\n\tchar filename[PATH_MAX + 1];\n\n\tcassert(config_prof);\n\n\tmalloc_snprintf(filename, sizeof(filename), \"/proc/%d/maps\",\n\t (int)getpid());\n\tmfd = open(filename, O_RDONLY);\n\tif (mfd != -1) {\n\t\tssize_t nread;\n\n\t\tif (prof_write(propagate_err, \"\\nMAPPED_LIBRARIES:\\n\") &&\n\t\t propagate_err)\n\t\t\treturn (true);\n\t\tnread = 0;\n\t\tdo {\n\t\t\tprof_dump_buf_end += nread;\n\t\t\tif (prof_dump_buf_end == PROF_DUMP_BUFSIZE) {\n\t\t\t\t/* Make space in prof_dump_buf before read(). */\n\t\t\t\tif (prof_flush(propagate_err) && propagate_err)\n\t\t\t\t\treturn (true);\n\t\t\t}\n\t\t\tnread = read(mfd, &prof_dump_buf[prof_dump_buf_end],\n\t\t\t PROF_DUMP_BUFSIZE - prof_dump_buf_end);\n\t\t} while (nread > 0);\n\t\tclose(mfd);\n\t} else\n\t\treturn (true);\n\n\treturn (false);\n}\n\nstatic bool\nprof_dump(bool propagate_err, const char *filename, bool leakcheck)\n{\n\tprof_tdata_t *prof_tdata;\n\tprof_cnt_t cnt_all;\n\tsize_t tabind;\n\tunion {\n\t\tprof_bt_t\t*p;\n\t\tvoid\t\t*v;\n\t} bt;\n\tunion {\n\t\tprof_ctx_t\t*p;\n\t\tvoid\t\t*v;\n\t} ctx;\n\tsize_t leak_nctx;\n\n\tcassert(config_prof);\n\n\tprof_tdata = prof_tdata_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)\n\t\treturn (true);\n\tprof_enter(prof_tdata);\n\tprof_dump_fd = creat(filename, 0644);\n\tif (prof_dump_fd == -1) {\n\t\tif (propagate_err == false) {\n\t\t\tmalloc_printf(\n\t\t\t \": creat(\\\"%s\\\"), 0644) failed\\n\",\n\t\t\t filename);\n\t\t\tif (opt_abort)\n\t\t\t\tabort();\n\t\t}\n\t\tgoto label_error;\n\t}\n\n\t/* Merge per thread profile stats, and sum them in cnt_all. */\n\tmemset(&cnt_all, 0, sizeof(prof_cnt_t));\n\tleak_nctx = 0;\n\tfor (tabind = 0; ckh_iter(&bt2ctx, &tabind, NULL, &ctx.v) == false;)\n\t\tprof_ctx_sum(ctx.p, &cnt_all, &leak_nctx);\n\n\t/* Dump profile header. */\n\tif (opt_lg_prof_sample == 0) {\n\t\tif (prof_printf(propagate_err,\n\t\t \"heap profile: %\"PRId64\": %\"PRId64\n\t\t \" [%\"PRIu64\": %\"PRIu64\"] @ heapprofile\\n\",\n\t\t cnt_all.curobjs, cnt_all.curbytes,\n\t\t cnt_all.accumobjs, cnt_all.accumbytes))\n\t\t\tgoto label_error;\n\t} else {\n\t\tif (prof_printf(propagate_err,\n\t\t \"heap profile: %\"PRId64\": %\"PRId64\n\t\t \" [%\"PRIu64\": %\"PRIu64\"] @ heap_v2/%\"PRIu64\"\\n\",\n\t\t cnt_all.curobjs, cnt_all.curbytes,\n\t\t cnt_all.accumobjs, cnt_all.accumbytes,\n\t\t ((uint64_t)1U << opt_lg_prof_sample)))\n\t\t\tgoto label_error;\n\t}\n\n\t/* Dump per ctx profile stats. */\n\tfor (tabind = 0; ckh_iter(&bt2ctx, &tabind, &bt.v, &ctx.v)\n\t == false;) {\n\t\tif (prof_dump_ctx(propagate_err, ctx.p, bt.p))\n\t\t\tgoto label_error;\n\t}\n\n\t/* Dump /proc//maps if possible. */\n\tif (prof_dump_maps(propagate_err))\n\t\tgoto label_error;\n\n\tif (prof_flush(propagate_err))\n\t\tgoto label_error;\n\tclose(prof_dump_fd);\n\tprof_leave(prof_tdata);\n\n\tif (leakcheck && cnt_all.curbytes != 0) {\n\t\tmalloc_printf(\": Leak summary: %\"PRId64\" byte%s, %\"\n\t\t PRId64\" object%s, %zu context%s\\n\",\n\t\t cnt_all.curbytes, (cnt_all.curbytes != 1) ? \"s\" : \"\",\n\t\t cnt_all.curobjs, (cnt_all.curobjs != 1) ? \"s\" : \"\",\n\t\t leak_nctx, (leak_nctx != 1) ? \"s\" : \"\");\n\t\tmalloc_printf(\n\t\t \": Run pprof on \\\"%s\\\" for leak detail\\n\",\n\t\t filename);\n\t}\n\n\treturn (false);\nlabel_error:\n\tprof_leave(prof_tdata);\n\treturn (true);\n}\n\n#define\tDUMP_FILENAME_BUFSIZE\t(PATH_MAX + 1)\nstatic void\nprof_dump_filename(char *filename, char v, int64_t vseq)\n{\n\n\tcassert(config_prof);\n\n\tif (vseq != UINT64_C(0xffffffffffffffff)) {\n\t /* \"...v.heap\" */\n\t\tmalloc_snprintf(filename, DUMP_FILENAME_BUFSIZE,\n\t\t \"%s.%d.%\"PRIu64\".%c%\"PRId64\".heap\",\n\t\t opt_prof_prefix, (int)getpid(), prof_dump_seq, v, vseq);\n\t} else {\n\t /* \"....heap\" */\n\t\tmalloc_snprintf(filename, DUMP_FILENAME_BUFSIZE,\n\t\t \"%s.%d.%\"PRIu64\".%c.heap\",\n\t\t opt_prof_prefix, (int)getpid(), prof_dump_seq, v);\n\t}\n\tprof_dump_seq++;\n}\n\nstatic void\nprof_fdump(void)\n{\n\tchar filename[DUMP_FILENAME_BUFSIZE];\n\n\tcassert(config_prof);\n\n\tif (prof_booted == false)\n\t\treturn;\n\n\tif (opt_prof_final && opt_prof_prefix[0] != '\\0') {\n\t\tmalloc_mutex_lock(&prof_dump_seq_mtx);\n\t\tprof_dump_filename(filename, 'f', UINT64_C(0xffffffffffffffff));\n\t\tmalloc_mutex_unlock(&prof_dump_seq_mtx);\n\t\tprof_dump(false, filename, opt_prof_leak);\n\t}\n}\n\nvoid\nprof_idump(void)\n{\n\tprof_tdata_t *prof_tdata;\n\tchar filename[PATH_MAX + 1];\n\n\tcassert(config_prof);\n\n\tif (prof_booted == false)\n\t\treturn;\n\t/*\n\t * Don't call prof_tdata_get() here, because it could cause recursive\n\t * allocation.\n\t */\n\tprof_tdata = *prof_tdata_tsd_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)\n\t\treturn;\n\tif (prof_tdata->enq) {\n\t\tprof_tdata->enq_idump = true;\n\t\treturn;\n\t}\n\n\tif (opt_prof_prefix[0] != '\\0') {\n\t\tmalloc_mutex_lock(&prof_dump_seq_mtx);\n\t\tprof_dump_filename(filename, 'i', prof_dump_iseq);\n\t\tprof_dump_iseq++;\n\t\tmalloc_mutex_unlock(&prof_dump_seq_mtx);\n\t\tprof_dump(false, filename, false);\n\t}\n}\n\nbool\nprof_mdump(const char *filename)\n{\n\tchar filename_buf[DUMP_FILENAME_BUFSIZE];\n\n\tcassert(config_prof);\n\n\tif (opt_prof == false || prof_booted == false)\n\t\treturn (true);\n\n\tif (filename == NULL) {\n\t\t/* No filename specified, so automatically generate one. */\n\t\tif (opt_prof_prefix[0] == '\\0')\n\t\t\treturn (true);\n\t\tmalloc_mutex_lock(&prof_dump_seq_mtx);\n\t\tprof_dump_filename(filename_buf, 'm', prof_dump_mseq);\n\t\tprof_dump_mseq++;\n\t\tmalloc_mutex_unlock(&prof_dump_seq_mtx);\n\t\tfilename = filename_buf;\n\t}\n\treturn (prof_dump(true, filename, false));\n}\n\nvoid\nprof_gdump(void)\n{\n\tprof_tdata_t *prof_tdata;\n\tchar filename[DUMP_FILENAME_BUFSIZE];\n\n\tcassert(config_prof);\n\n\tif (prof_booted == false)\n\t\treturn;\n\t/*\n\t * Don't call prof_tdata_get() here, because it could cause recursive\n\t * allocation.\n\t */\n\tprof_tdata = *prof_tdata_tsd_get();\n\tif ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)\n\t\treturn;\n\tif (prof_tdata->enq) {\n\t\tprof_tdata->enq_gdump = true;\n\t\treturn;\n\t}\n\n\tif (opt_prof_prefix[0] != '\\0') {\n\t\tmalloc_mutex_lock(&prof_dump_seq_mtx);\n\t\tprof_dump_filename(filename, 'u', prof_dump_useq);\n\t\tprof_dump_useq++;\n\t\tmalloc_mutex_unlock(&prof_dump_seq_mtx);\n\t\tprof_dump(false, filename, false);\n\t}\n}\n\nstatic void\nprof_bt_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2)\n{\n\tsize_t ret1, ret2;\n\tuint64_t h;\n\tprof_bt_t *bt = (prof_bt_t *)key;\n\n\tcassert(config_prof);\n\tassert(minbits <= 32 || (SIZEOF_PTR == 8 && minbits <= 64));\n\tassert(hash1 != NULL);\n\tassert(hash2 != NULL);\n\n\th = hash(bt->vec, bt->len * sizeof(void *),\n\t UINT64_C(0x94122f335b332aea));\n\tif (minbits <= 32) {\n\t\t/*\n\t\t * Avoid doing multiple hashes, since a single hash provides\n\t\t * enough bits.\n\t\t */\n\t\tret1 = h & ZU(0xffffffffU);\n\t\tret2 = h >> 32;\n\t} else {\n\t\tret1 = h;\n\t\tret2 = hash(bt->vec, bt->len * sizeof(void *),\n\t\t UINT64_C(0x8432a476666bbc13));\n\t}\n\n\t*hash1 = ret1;\n\t*hash2 = ret2;\n}\n\nstatic bool\nprof_bt_keycomp(const void *k1, const void *k2)\n{\n\tconst prof_bt_t *bt1 = (prof_bt_t *)k1;\n\tconst prof_bt_t *bt2 = (prof_bt_t *)k2;\n\n\tcassert(config_prof);\n\n\tif (bt1->len != bt2->len)\n\t\treturn (false);\n\treturn (memcmp(bt1->vec, bt2->vec, bt1->len * sizeof(void *)) == 0);\n}\n\nstatic malloc_mutex_t *\nprof_ctx_mutex_choose(void)\n{\n\tunsigned nctxs = atomic_add_u(&cum_ctxs, 1);\n\n\treturn (&ctx_locks[(nctxs - 1) % PROF_NCTX_LOCKS]);\n}\n\nprof_tdata_t *\nprof_tdata_init(void)\n{\n\tprof_tdata_t *prof_tdata;\n\n\tcassert(config_prof);\n\n\t/* Initialize an empty cache for this thread. */\n\tprof_tdata = (prof_tdata_t *)imalloc(sizeof(prof_tdata_t));\n\tif (prof_tdata == NULL)\n\t\treturn (NULL);\n\n\tif (ckh_new(&prof_tdata->bt2cnt, PROF_CKH_MINITEMS,\n\t prof_bt_hash, prof_bt_keycomp)) {\n\t\tidalloc(prof_tdata);\n\t\treturn (NULL);\n\t}\n\tql_new(&prof_tdata->lru_ql);\n\n\tprof_tdata->vec = imalloc(sizeof(void *) * PROF_BT_MAX);\n\tif (prof_tdata->vec == NULL) {\n\t\tckh_delete(&prof_tdata->bt2cnt);\n\t\tidalloc(prof_tdata);\n\t\treturn (NULL);\n\t}\n\n\tprof_tdata->prng_state = 0;\n\tprof_tdata->threshold = 0;\n\tprof_tdata->accum = 0;\n\n\tprof_tdata->enq = false;\n\tprof_tdata->enq_idump = false;\n\tprof_tdata->enq_gdump = false;\n\n\tprof_tdata_tsd_set(&prof_tdata);\n\n\treturn (prof_tdata);\n}\n\nvoid\nprof_tdata_cleanup(void *arg)\n{\n\tprof_thr_cnt_t *cnt;\n\tprof_tdata_t *prof_tdata = *(prof_tdata_t **)arg;\n\n\tcassert(config_prof);\n\n\tif (prof_tdata == PROF_TDATA_STATE_REINCARNATED) {\n\t\t/*\n\t\t * Another destructor deallocated memory after this destructor\n\t\t * was called. Reset prof_tdata to PROF_TDATA_STATE_PURGATORY\n\t\t * in order to receive another callback.\n\t\t */\n\t\tprof_tdata = PROF_TDATA_STATE_PURGATORY;\n\t\tprof_tdata_tsd_set(&prof_tdata);\n\t} else if (prof_tdata == PROF_TDATA_STATE_PURGATORY) {\n\t\t/*\n\t\t * The previous time this destructor was called, we set the key\n\t\t * to PROF_TDATA_STATE_PURGATORY so that other destructors\n\t\t * wouldn't cause re-creation of the prof_tdata. This time, do\n\t\t * nothing, so that the destructor will not be called again.\n\t\t */\n\t} else if (prof_tdata != NULL) {\n\t\t/*\n\t\t * Delete the hash table. All of its contents can still be\n\t\t * iterated over via the LRU.\n\t\t */\n\t\tckh_delete(&prof_tdata->bt2cnt);\n\t\t/*\n\t\t * Iteratively merge cnt's into the global stats and delete\n\t\t * them.\n\t\t */\n\t\twhile ((cnt = ql_last(&prof_tdata->lru_ql, lru_link)) != NULL) {\n\t\t\tql_remove(&prof_tdata->lru_ql, cnt, lru_link);\n\t\t\tprof_ctx_merge(cnt->ctx, cnt);\n\t\t\tidalloc(cnt);\n\t\t}\n\t\tidalloc(prof_tdata->vec);\n\t\tidalloc(prof_tdata);\n\t\tprof_tdata = PROF_TDATA_STATE_PURGATORY;\n\t\tprof_tdata_tsd_set(&prof_tdata);\n\t}\n}\n\nvoid\nprof_boot0(void)\n{\n\n\tcassert(config_prof);\n\n\tmemcpy(opt_prof_prefix, PROF_PREFIX_DEFAULT,\n\t sizeof(PROF_PREFIX_DEFAULT));\n}\n\nvoid\nprof_boot1(void)\n{\n\n\tcassert(config_prof);\n\n\t/*\n\t * opt_prof and prof_promote must be in their final state before any\n\t * arenas are initialized, so this function must be executed early.\n\t */\n\n\tif (opt_prof_leak && opt_prof == false) {\n\t\t/*\n\t\t * Enable opt_prof, but in such a way that profiles are never\n\t\t * automatically dumped.\n\t\t */\n\t\topt_prof = true;\n\t\topt_prof_gdump = false;\n\t\tprof_interval = 0;\n\t} else if (opt_prof) {\n\t\tif (opt_lg_prof_interval >= 0) {\n\t\t\tprof_interval = (((uint64_t)1U) <<\n\t\t\t opt_lg_prof_interval);\n\t\t} else\n\t\t\tprof_interval = 0;\n\t}\n\n\tprof_promote = (opt_prof && opt_lg_prof_sample > LG_PAGE);\n}\n\nbool\nprof_boot2(void)\n{\n\n\tcassert(config_prof);\n\n\tif (opt_prof) {\n\t\tunsigned i;\n\n\t\tif (ckh_new(&bt2ctx, PROF_CKH_MINITEMS, prof_bt_hash,\n\t\t prof_bt_keycomp))\n\t\t\treturn (true);\n\t\tif (malloc_mutex_init(&bt2ctx_mtx))\n\t\t\treturn (true);\n\t\tif (prof_tdata_tsd_boot()) {\n\t\t\tmalloc_write(\n\t\t\t \": Error in pthread_key_create()\\n\");\n\t\t\tabort();\n\t\t}\n\n\t\tif (malloc_mutex_init(&prof_dump_seq_mtx))\n\t\t\treturn (true);\n\n\t\tif (atexit(prof_fdump) != 0) {\n\t\t\tmalloc_write(\": Error in atexit()\\n\");\n\t\t\tif (opt_abort)\n\t\t\t\tabort();\n\t\t}\n\n\t\tctx_locks = (malloc_mutex_t *)base_alloc(PROF_NCTX_LOCKS *\n\t\t sizeof(malloc_mutex_t));\n\t\tif (ctx_locks == NULL)\n\t\t\treturn (true);\n\t\tfor (i = 0; i < PROF_NCTX_LOCKS; i++) {\n\t\t\tif (malloc_mutex_init(&ctx_locks[i]))\n\t\t\t\treturn (true);\n\t\t}\n\t}\n\n#ifdef JEMALLOC_PROF_LIBGCC\n\t/*\n\t * Cause the backtracing machinery to allocate its internal state\n\t * before enabling profiling.\n\t */\n\t_Unwind_Backtrace(prof_unwind_init_callback, NULL);\n#endif\n\n\tprof_booted = true;\n\n\treturn (false);\n}\n\nvoid\nprof_prefork(void)\n{\n\n\tif (opt_prof) {\n\t\tunsigned i;\n\n\t\tmalloc_mutex_lock(&bt2ctx_mtx);\n\t\tmalloc_mutex_lock(&prof_dump_seq_mtx);\n\t\tfor (i = 0; i < PROF_NCTX_LOCKS; i++)\n\t\t\tmalloc_mutex_lock(&ctx_locks[i]);\n\t}\n}\n\nvoid\nprof_postfork_parent(void)\n{\n\n\tif (opt_prof) {\n\t\tunsigned i;\n\n\t\tfor (i = 0; i < PROF_NCTX_LOCKS; i++)\n\t\t\tmalloc_mutex_postfork_parent(&ctx_locks[i]);\n\t\tmalloc_mutex_postfork_parent(&prof_dump_seq_mtx);\n\t\tmalloc_mutex_postfork_parent(&bt2ctx_mtx);\n\t}\n}\n\nvoid\nprof_postfork_child(void)\n{\n\n\tif (opt_prof) {\n\t\tunsigned i;\n\n\t\tfor (i = 0; i < PROF_NCTX_LOCKS; i++)\n\t\t\tmalloc_mutex_postfork_child(&ctx_locks[i]);\n\t\tmalloc_mutex_postfork_child(&prof_dump_seq_mtx);\n\t\tmalloc_mutex_postfork_child(&bt2ctx_mtx);\n\t}\n}\n\n/******************************************************************************/\n" }, { "path": "deps/jemalloc/src/quarantine.c", "content": "#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/*\n * quarantine pointers close to NULL are used to encode state information that\n * is used for cleaning up during thread shutdown.\n */\n#define\tQUARANTINE_STATE_REINCARNATED\t((quarantine_t *)(uintptr_t)1)\n#define\tQUARANTINE_STATE_PURGATORY\t((quarantine_t *)(uintptr_t)2)\n#define\tQUARANTINE_STATE_MAX\t\tQUARANTINE_STATE_PURGATORY\n\n/******************************************************************************/\n/* Data. */\n\ntypedef struct quarantine_obj_s quarantine_obj_t;\ntypedef struct quarantine_s quarantine_t;\n\nstruct quarantine_obj_s {\n\tvoid\t*ptr;\n\tsize_t\tusize;\n};\n\nstruct quarantine_s {\n\tsize_t\t\t\tcurbytes;\n\tsize_t\t\t\tcurobjs;\n\tsize_t\t\t\tfirst;\n#define\tLG_MAXOBJS_INIT 10\n\tsize_t\t\t\tlg_maxobjs;\n\tquarantine_obj_t\tobjs[1]; /* Dynamically sized ring buffer. */\n};\n\nstatic void\tquarantine_cleanup(void *arg);\n\nmalloc_tsd_data(static, quarantine, quarantine_t *, NULL)\nmalloc_tsd_funcs(JEMALLOC_INLINE, quarantine, quarantine_t *, NULL,\n quarantine_cleanup)\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic quarantine_t\t*quarantine_init(size_t lg_maxobjs);\nstatic quarantine_t\t*quarantine_grow(quarantine_t *quarantine);\nstatic void\tquarantine_drain(quarantine_t *quarantine, size_t upper_bound);\n\n/******************************************************************************/\n\nstatic quarantine_t *\nquarantine_init(size_t lg_maxobjs)\n{\n\tquarantine_t *quarantine;\n\n\tquarantine = (quarantine_t *)imalloc(offsetof(quarantine_t, objs) +\n\t ((ZU(1) << lg_maxobjs) * sizeof(quarantine_obj_t)));\n\tif (quarantine == NULL)\n\t\treturn (NULL);\n\tquarantine->curbytes = 0;\n\tquarantine->curobjs = 0;\n\tquarantine->first = 0;\n\tquarantine->lg_maxobjs = lg_maxobjs;\n\n\tquarantine_tsd_set(&quarantine);\n\n\treturn (quarantine);\n}\n\nstatic quarantine_t *\nquarantine_grow(quarantine_t *quarantine)\n{\n\tquarantine_t *ret;\n\n\tret = quarantine_init(quarantine->lg_maxobjs + 1);\n\tif (ret == NULL)\n\t\treturn (quarantine);\n\n\tret->curbytes = quarantine->curbytes;\n\tret->curobjs = quarantine->curobjs;\n\tif (quarantine->first + quarantine->curobjs <= (ZU(1) <<\n\t quarantine->lg_maxobjs)) {\n\t\t/* objs ring buffer data are contiguous. */\n\t\tmemcpy(ret->objs, &quarantine->objs[quarantine->first],\n\t\t quarantine->curobjs * sizeof(quarantine_obj_t));\n\t} else {\n\t\t/* objs ring buffer data wrap around. */\n\t\tsize_t ncopy_a = (ZU(1) << quarantine->lg_maxobjs) -\n\t\t quarantine->first;\n\t\tsize_t ncopy_b = quarantine->curobjs - ncopy_a;\n\n\t\tmemcpy(ret->objs, &quarantine->objs[quarantine->first], ncopy_a\n\t\t * sizeof(quarantine_obj_t));\n\t\tmemcpy(&ret->objs[ncopy_a], quarantine->objs, ncopy_b *\n\t\t sizeof(quarantine_obj_t));\n\t}\n\n\treturn (ret);\n}\n\nstatic void\nquarantine_drain(quarantine_t *quarantine, size_t upper_bound)\n{\n\n\twhile (quarantine->curbytes > upper_bound && quarantine->curobjs > 0) {\n\t\tquarantine_obj_t *obj = &quarantine->objs[quarantine->first];\n\t\tassert(obj->usize == isalloc(obj->ptr, config_prof));\n\t\tidalloc(obj->ptr);\n\t\tquarantine->curbytes -= obj->usize;\n\t\tquarantine->curobjs--;\n\t\tquarantine->first = (quarantine->first + 1) & ((ZU(1) <<\n\t\t quarantine->lg_maxobjs) - 1);\n\t}\n}\n\nvoid\nquarantine(void *ptr)\n{\n\tquarantine_t *quarantine;\n\tsize_t usize = isalloc(ptr, config_prof);\n\n\tcassert(config_fill);\n\tassert(opt_quarantine);\n\n\tquarantine = *quarantine_tsd_get();\n\tif ((uintptr_t)quarantine <= (uintptr_t)QUARANTINE_STATE_MAX) {\n\t\tif (quarantine == NULL) {\n\t\t\tif ((quarantine = quarantine_init(LG_MAXOBJS_INIT)) ==\n\t\t\t NULL) {\n\t\t\t\tidalloc(ptr);\n\t\t\t\treturn;\n\t\t\t}\n\t\t} else {\n\t\t\tif (quarantine == QUARANTINE_STATE_PURGATORY) {\n\t\t\t\t/*\n\t\t\t\t * Make a note that quarantine() was called\n\t\t\t\t * after quarantine_cleanup() was called.\n\t\t\t\t */\n\t\t\t\tquarantine = QUARANTINE_STATE_REINCARNATED;\n\t\t\t\tquarantine_tsd_set(&quarantine);\n\t\t\t}\n\t\t\tidalloc(ptr);\n\t\t\treturn;\n\t\t}\n\t}\n\t/*\n\t * Drain one or more objects if the quarantine size limit would be\n\t * exceeded by appending ptr.\n\t */\n\tif (quarantine->curbytes + usize > opt_quarantine) {\n\t\tsize_t upper_bound = (opt_quarantine >= usize) ? opt_quarantine\n\t\t - usize : 0;\n\t\tquarantine_drain(quarantine, upper_bound);\n\t}\n\t/* Grow the quarantine ring buffer if it's full. */\n\tif (quarantine->curobjs == (ZU(1) << quarantine->lg_maxobjs))\n\t\tquarantine = quarantine_grow(quarantine);\n\t/* quarantine_grow() must free a slot if it fails to grow. */\n\tassert(quarantine->curobjs < (ZU(1) << quarantine->lg_maxobjs));\n\t/* Append ptr if its size doesn't exceed the quarantine size. */\n\tif (quarantine->curbytes + usize <= opt_quarantine) {\n\t\tsize_t offset = (quarantine->first + quarantine->curobjs) &\n\t\t ((ZU(1) << quarantine->lg_maxobjs) - 1);\n\t\tquarantine_obj_t *obj = &quarantine->objs[offset];\n\t\tobj->ptr = ptr;\n\t\tobj->usize = usize;\n\t\tquarantine->curbytes += usize;\n\t\tquarantine->curobjs++;\n\t\tif (opt_junk)\n\t\t\tmemset(ptr, 0x5a, usize);\n\t} else {\n\t\tassert(quarantine->curbytes == 0);\n\t\tidalloc(ptr);\n\t}\n}\n\nstatic void\nquarantine_cleanup(void *arg)\n{\n\tquarantine_t *quarantine = *(quarantine_t **)arg;\n\n\tif (quarantine == QUARANTINE_STATE_REINCARNATED) {\n\t\t/*\n\t\t * Another destructor deallocated memory after this destructor\n\t\t * was called. Reset quarantine to QUARANTINE_STATE_PURGATORY\n\t\t * in order to receive another callback.\n\t\t */\n\t\tquarantine = QUARANTINE_STATE_PURGATORY;\n\t\tquarantine_tsd_set(&quarantine);\n\t} else if (quarantine == QUARANTINE_STATE_PURGATORY) {\n\t\t/*\n\t\t * The previous time this destructor was called, we set the key\n\t\t * to QUARANTINE_STATE_PURGATORY so that other destructors\n\t\t * wouldn't cause re-creation of the quarantine. This time, do\n\t\t * nothing, so that the destructor will not be called again.\n\t\t */\n\t} else if (quarantine != NULL) {\n\t\tquarantine_drain(quarantine, 0);\n\t\tidalloc(quarantine);\n\t\tquarantine = QUARANTINE_STATE_PURGATORY;\n\t\tquarantine_tsd_set(&quarantine);\n\t}\n}\n\nbool\nquarantine_boot(void)\n{\n\n\tcassert(config_fill);\n\n\tif (quarantine_tsd_boot())\n\t\treturn (true);\n\n\treturn (false);\n}\n" }, { "path": "deps/jemalloc/src/rtree.c", "content": "#define\tJEMALLOC_RTREE_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\nrtree_t *\nrtree_new(unsigned bits)\n{\n\trtree_t *ret;\n\tunsigned bits_per_level, height, i;\n\n\tbits_per_level = ffs(pow2_ceil((RTREE_NODESIZE / sizeof(void *)))) - 1;\n\theight = bits / bits_per_level;\n\tif (height * bits_per_level != bits)\n\t\theight++;\n\tassert(height * bits_per_level >= bits);\n\n\tret = (rtree_t*)base_alloc(offsetof(rtree_t, level2bits) +\n\t (sizeof(unsigned) * height));\n\tif (ret == NULL)\n\t\treturn (NULL);\n\tmemset(ret, 0, offsetof(rtree_t, level2bits) + (sizeof(unsigned) *\n\t height));\n\n\tif (malloc_mutex_init(&ret->mutex)) {\n\t\t/* Leak the rtree. */\n\t\treturn (NULL);\n\t}\n\tret->height = height;\n\tif (bits_per_level * height > bits)\n\t\tret->level2bits[0] = bits % bits_per_level;\n\telse\n\t\tret->level2bits[0] = bits_per_level;\n\tfor (i = 1; i < height; i++)\n\t\tret->level2bits[i] = bits_per_level;\n\n\tret->root = (void**)base_alloc(sizeof(void *) << ret->level2bits[0]);\n\tif (ret->root == NULL) {\n\t\t/*\n\t\t * We leak the rtree here, since there's no generic base\n\t\t * deallocation.\n\t\t */\n\t\treturn (NULL);\n\t}\n\tmemset(ret->root, 0, sizeof(void *) << ret->level2bits[0]);\n\n\treturn (ret);\n}\n\nvoid\nrtree_prefork(rtree_t *rtree)\n{\n\n\tmalloc_mutex_prefork(&rtree->mutex);\n}\n\nvoid\nrtree_postfork_parent(rtree_t *rtree)\n{\n\n\tmalloc_mutex_postfork_parent(&rtree->mutex);\n}\n\nvoid\nrtree_postfork_child(rtree_t *rtree)\n{\n\n\tmalloc_mutex_postfork_child(&rtree->mutex);\n}\n" }, { "path": "deps/jemalloc/src/stats.c", "content": "#define\tJEMALLOC_STATS_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n#define\tCTL_GET(n, v, t) do {\t\t\t\t\t\t\\\n\tsize_t sz = sizeof(t);\t\t\t\t\t\t\\\n\txmallctl(n, v, &sz, NULL, 0);\t\t\t\t\t\\\n} while (0)\n\n#define\tCTL_I_GET(n, v, t) do {\t\t\t\t\t\t\\\n\tsize_t mib[6];\t\t\t\t\t\t\t\\\n\tsize_t miblen = sizeof(mib) / sizeof(size_t);\t\t\t\\\n\tsize_t sz = sizeof(t);\t\t\t\t\t\t\\\n\txmallctlnametomib(n, mib, &miblen);\t\t\t\t\\\n\tmib[2] = i;\t\t\t\t\t\t\t\\\n\txmallctlbymib(mib, miblen, v, &sz, NULL, 0);\t\t\t\\\n} while (0)\n\n#define\tCTL_J_GET(n, v, t) do {\t\t\t\t\t\t\\\n\tsize_t mib[6];\t\t\t\t\t\t\t\\\n\tsize_t miblen = sizeof(mib) / sizeof(size_t);\t\t\t\\\n\tsize_t sz = sizeof(t);\t\t\t\t\t\t\\\n\txmallctlnametomib(n, mib, &miblen);\t\t\t\t\\\n\tmib[2] = j;\t\t\t\t\t\t\t\\\n\txmallctlbymib(mib, miblen, v, &sz, NULL, 0);\t\t\t\\\n} while (0)\n\n#define\tCTL_IJ_GET(n, v, t) do {\t\t\t\t\t\\\n\tsize_t mib[6];\t\t\t\t\t\t\t\\\n\tsize_t miblen = sizeof(mib) / sizeof(size_t);\t\t\t\\\n\tsize_t sz = sizeof(t);\t\t\t\t\t\t\\\n\txmallctlnametomib(n, mib, &miblen);\t\t\t\t\\\n\tmib[2] = i;\t\t\t\t\t\t\t\\\n\tmib[4] = j;\t\t\t\t\t\t\t\\\n\txmallctlbymib(mib, miblen, v, &sz, NULL, 0);\t\t\t\\\n} while (0)\n\n/******************************************************************************/\n/* Data. */\n\nbool\topt_stats_print = false;\n\nsize_t\tstats_cactive = 0;\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\tstats_arena_bins_print(void (*write_cb)(void *, const char *),\n void *cbopaque, unsigned i);\nstatic void\tstats_arena_lruns_print(void (*write_cb)(void *, const char *),\n void *cbopaque, unsigned i);\nstatic void\tstats_arena_print(void (*write_cb)(void *, const char *),\n void *cbopaque, unsigned i, bool bins, bool large);\n\n/******************************************************************************/\n\nstatic void\nstats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,\n unsigned i)\n{\n\tsize_t page;\n\tbool config_tcache;\n\tunsigned nbins, j, gap_start;\n\n\tCTL_GET(\"arenas.page\", &page, size_t);\n\n\tCTL_GET(\"config.tcache\", &config_tcache, bool);\n\tif (config_tcache) {\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"bins: bin size regs pgs allocated nmalloc\"\n\t\t \" ndalloc nrequests nfills nflushes\"\n\t\t \" newruns reruns curruns\\n\");\n\t} else {\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"bins: bin size regs pgs allocated nmalloc\"\n\t\t \" ndalloc newruns reruns curruns\\n\");\n\t}\n\tCTL_GET(\"arenas.nbins\", &nbins, unsigned);\n\tfor (j = 0, gap_start = UINT_MAX; j < nbins; j++) {\n\t\tuint64_t nruns;\n\n\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nruns\", &nruns, uint64_t);\n\t\tif (nruns == 0) {\n\t\t\tif (gap_start == UINT_MAX)\n\t\t\t\tgap_start = j;\n\t\t} else {\n\t\t\tsize_t reg_size, run_size, allocated;\n\t\t\tuint32_t nregs;\n\t\t\tuint64_t nmalloc, ndalloc, nrequests, nfills, nflushes;\n\t\t\tuint64_t reruns;\n\t\t\tsize_t curruns;\n\n\t\t\tif (gap_start != UINT_MAX) {\n\t\t\t\tif (j > gap_start + 1) {\n\t\t\t\t\t/* Gap of more than one size class. */\n\t\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t\t \"[%u..%u]\\n\", gap_start,\n\t\t\t\t\t j - 1);\n\t\t\t\t} else {\n\t\t\t\t\t/* Gap of one size class. */\n\t\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t\t \"[%u]\\n\", gap_start);\n\t\t\t\t}\n\t\t\t\tgap_start = UINT_MAX;\n\t\t\t}\n\t\t\tCTL_J_GET(\"arenas.bin.0.size\", ®_size, size_t);\n\t\t\tCTL_J_GET(\"arenas.bin.0.nregs\", &nregs, uint32_t);\n\t\t\tCTL_J_GET(\"arenas.bin.0.run_size\", &run_size, size_t);\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.allocated\",\n\t\t\t &allocated, size_t);\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nmalloc\",\n\t\t\t &nmalloc, uint64_t);\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.ndalloc\",\n\t\t\t &ndalloc, uint64_t);\n\t\t\tif (config_tcache) {\n\t\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nrequests\",\n\t\t\t\t &nrequests, uint64_t);\n\t\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nfills\",\n\t\t\t\t &nfills, uint64_t);\n\t\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nflushes\",\n\t\t\t\t &nflushes, uint64_t);\n\t\t\t}\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.nreruns\", &reruns,\n\t\t\t uint64_t);\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.bins.0.curruns\", &curruns,\n\t\t\t size_t);\n\t\t\tif (config_tcache) {\n\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t \"%13u %5zu %4u %3zu %12zu %12\"PRIu64\n\t\t\t\t \" %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\n\t\t\t\t \" %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\n\t\t\t\t \" %12zu\\n\",\n\t\t\t\t j, reg_size, nregs, run_size / page,\n\t\t\t\t allocated, nmalloc, ndalloc, nrequests,\n\t\t\t\t nfills, nflushes, nruns, reruns, curruns);\n\t\t\t} else {\n\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t \"%13u %5zu %4u %3zu %12zu %12\"PRIu64\n\t\t\t\t \" %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\n\t\t\t\t \" %12zu\\n\",\n\t\t\t\t j, reg_size, nregs, run_size / page,\n\t\t\t\t allocated, nmalloc, ndalloc, nruns, reruns,\n\t\t\t\t curruns);\n\t\t\t}\n\t\t}\n\t}\n\tif (gap_start != UINT_MAX) {\n\t\tif (j > gap_start + 1) {\n\t\t\t/* Gap of more than one size class. */\n\t\t\tmalloc_cprintf(write_cb, cbopaque, \"[%u..%u]\\n\",\n\t\t\t gap_start, j - 1);\n\t\t} else {\n\t\t\t/* Gap of one size class. */\n\t\t\tmalloc_cprintf(write_cb, cbopaque, \"[%u]\\n\", gap_start);\n\t\t}\n\t}\n}\n\nstatic void\nstats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque,\n unsigned i)\n{\n\tsize_t page, nlruns, j;\n\tssize_t gap_start;\n\n\tCTL_GET(\"arenas.page\", &page, size_t);\n\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"large: size pages nmalloc ndalloc nrequests\"\n\t \" curruns\\n\");\n\tCTL_GET(\"arenas.nlruns\", &nlruns, size_t);\n\tfor (j = 0, gap_start = -1; j < nlruns; j++) {\n\t\tuint64_t nmalloc, ndalloc, nrequests;\n\t\tsize_t run_size, curruns;\n\n\t\tCTL_IJ_GET(\"stats.arenas.0.lruns.0.nmalloc\", &nmalloc,\n\t\t uint64_t);\n\t\tCTL_IJ_GET(\"stats.arenas.0.lruns.0.ndalloc\", &ndalloc,\n\t\t uint64_t);\n\t\tCTL_IJ_GET(\"stats.arenas.0.lruns.0.nrequests\", &nrequests,\n\t\t uint64_t);\n\t\tif (nrequests == 0) {\n\t\t\tif (gap_start == -1)\n\t\t\t\tgap_start = j;\n\t\t} else {\n\t\t\tCTL_J_GET(\"arenas.lrun.0.size\", &run_size, size_t);\n\t\t\tCTL_IJ_GET(\"stats.arenas.0.lruns.0.curruns\", &curruns,\n\t\t\t size_t);\n\t\t\tif (gap_start != -1) {\n\t\t\t\tmalloc_cprintf(write_cb, cbopaque, \"[%zu]\\n\",\n\t\t\t\t j - gap_start);\n\t\t\t\tgap_start = -1;\n\t\t\t}\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t \"%13zu %5zu %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\n\t\t\t \" %12zu\\n\",\n\t\t\t run_size, run_size / page, nmalloc, ndalloc,\n\t\t\t nrequests, curruns);\n\t\t}\n\t}\n\tif (gap_start != -1)\n\t\tmalloc_cprintf(write_cb, cbopaque, \"[%zu]\\n\", j - gap_start);\n}\n\nstatic void\nstats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,\n unsigned i, bool bins, bool large)\n{\n\tunsigned nthreads;\n\tconst char *dss;\n\tsize_t page, pactive, pdirty, mapped;\n\tuint64_t npurge, nmadvise, purged;\n\tsize_t small_allocated;\n\tuint64_t small_nmalloc, small_ndalloc, small_nrequests;\n\tsize_t large_allocated;\n\tuint64_t large_nmalloc, large_ndalloc, large_nrequests;\n\n\tCTL_GET(\"arenas.page\", &page, size_t);\n\n\tCTL_I_GET(\"stats.arenas.0.nthreads\", &nthreads, unsigned);\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"assigned threads: %u\\n\", nthreads);\n\tCTL_I_GET(\"stats.arenas.0.dss\", &dss, const char *);\n\tmalloc_cprintf(write_cb, cbopaque, \"dss allocation precedence: %s\\n\",\n\t dss);\n\tCTL_I_GET(\"stats.arenas.0.pactive\", &pactive, size_t);\n\tCTL_I_GET(\"stats.arenas.0.pdirty\", &pdirty, size_t);\n\tCTL_I_GET(\"stats.arenas.0.npurge\", &npurge, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.nmadvise\", &nmadvise, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.purged\", &purged, uint64_t);\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"dirty pages: %zu:%zu active:dirty, %\"PRIu64\" sweep%s,\"\n\t \" %\"PRIu64\" madvise%s, %\"PRIu64\" purged\\n\",\n\t pactive, pdirty, npurge, npurge == 1 ? \"\" : \"s\",\n\t nmadvise, nmadvise == 1 ? \"\" : \"s\", purged);\n\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \" allocated nmalloc ndalloc nrequests\\n\");\n\tCTL_I_GET(\"stats.arenas.0.small.allocated\", &small_allocated, size_t);\n\tCTL_I_GET(\"stats.arenas.0.small.nmalloc\", &small_nmalloc, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.small.ndalloc\", &small_ndalloc, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.small.nrequests\", &small_nrequests, uint64_t);\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"small: %12zu %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\"\\n\",\n\t small_allocated, small_nmalloc, small_ndalloc, small_nrequests);\n\tCTL_I_GET(\"stats.arenas.0.large.allocated\", &large_allocated, size_t);\n\tCTL_I_GET(\"stats.arenas.0.large.nmalloc\", &large_nmalloc, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.large.ndalloc\", &large_ndalloc, uint64_t);\n\tCTL_I_GET(\"stats.arenas.0.large.nrequests\", &large_nrequests, uint64_t);\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"large: %12zu %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\"\\n\",\n\t large_allocated, large_nmalloc, large_ndalloc, large_nrequests);\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"total: %12zu %12\"PRIu64\" %12\"PRIu64\" %12\"PRIu64\"\\n\",\n\t small_allocated + large_allocated,\n\t small_nmalloc + large_nmalloc,\n\t small_ndalloc + large_ndalloc,\n\t small_nrequests + large_nrequests);\n\tmalloc_cprintf(write_cb, cbopaque, \"active: %12zu\\n\", pactive * page);\n\tCTL_I_GET(\"stats.arenas.0.mapped\", &mapped, size_t);\n\tmalloc_cprintf(write_cb, cbopaque, \"mapped: %12zu\\n\", mapped);\n\n\tif (bins)\n\t\tstats_arena_bins_print(write_cb, cbopaque, i);\n\tif (large)\n\t\tstats_arena_lruns_print(write_cb, cbopaque, i);\n}\n\nvoid\nstats_print(void (*write_cb)(void *, const char *), void *cbopaque,\n const char *opts)\n{\n\tint err;\n\tuint64_t epoch;\n\tsize_t u64sz;\n\tbool general = true;\n\tbool merged = true;\n\tbool unmerged = true;\n\tbool bins = true;\n\tbool large = true;\n\n\t/*\n\t * Refresh stats, in case mallctl() was called by the application.\n\t *\n\t * Check for OOM here, since refreshing the ctl cache can trigger\n\t * allocation. In practice, none of the subsequent mallctl()-related\n\t * calls in this function will cause OOM if this one succeeds.\n\t * */\n\tepoch = 1;\n\tu64sz = sizeof(uint64_t);\n\terr = je_mallctl(\"epoch\", &epoch, &u64sz, &epoch, sizeof(uint64_t));\n\tif (err != 0) {\n\t\tif (err == EAGAIN) {\n\t\t\tmalloc_write(\": Memory allocation failure in \"\n\t\t\t \"mallctl(\\\"epoch\\\", ...)\\n\");\n\t\t\treturn;\n\t\t}\n\t\tmalloc_write(\": Failure in mallctl(\\\"epoch\\\", \"\n\t\t \"...)\\n\");\n\t\tabort();\n\t}\n\n\tif (opts != NULL) {\n\t\tunsigned i;\n\n\t\tfor (i = 0; opts[i] != '\\0'; i++) {\n\t\t\tswitch (opts[i]) {\n\t\t\tcase 'g':\n\t\t\t\tgeneral = false;\n\t\t\t\tbreak;\n\t\t\tcase 'm':\n\t\t\t\tmerged = false;\n\t\t\t\tbreak;\n\t\t\tcase 'a':\n\t\t\t\tunmerged = false;\n\t\t\t\tbreak;\n\t\t\tcase 'b':\n\t\t\t\tbins = false;\n\t\t\t\tbreak;\n\t\t\tcase 'l':\n\t\t\t\tlarge = false;\n\t\t\t\tbreak;\n\t\t\tdefault:;\n\t\t\t}\n\t\t}\n\t}\n\n\tmalloc_cprintf(write_cb, cbopaque,\n\t \"___ Begin jemalloc statistics ___\\n\");\n\tif (general) {\n\t\tint err;\n\t\tconst char *cpv;\n\t\tbool bv;\n\t\tunsigned uv;\n\t\tssize_t ssv;\n\t\tsize_t sv, bsz, ssz, sssz, cpsz;\n\n\t\tbsz = sizeof(bool);\n\t\tssz = sizeof(size_t);\n\t\tsssz = sizeof(ssize_t);\n\t\tcpsz = sizeof(const char *);\n\n\t\tCTL_GET(\"version\", &cpv, const char *);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Version: %s\\n\", cpv);\n\t\tCTL_GET(\"config.debug\", &bv, bool);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Assertions %s\\n\",\n\t\t bv ? \"enabled\" : \"disabled\");\n\n#define OPT_WRITE_BOOL(n)\t\t\t\t\t\t\\\n\t\tif ((err = je_mallctl(\"opt.\"#n, &bv, &bsz, NULL, 0))\t\\\n\t\t == 0) {\t\t\t\t\t\t\\\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\t\t\\\n\t\t\t \" opt.\"#n\": %s\\n\", bv ? \"true\" : \"false\");\t\\\n\t\t}\n#define OPT_WRITE_SIZE_T(n)\t\t\t\t\t\t\\\n\t\tif ((err = je_mallctl(\"opt.\"#n, &sv, &ssz, NULL, 0))\t\\\n\t\t == 0) {\t\t\t\t\t\t\\\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\t\t\\\n\t\t\t\" opt.\"#n\": %zu\\n\", sv);\t\t\t\\\n\t\t}\n#define OPT_WRITE_SSIZE_T(n)\t\t\t\t\t\t\\\n\t\tif ((err = je_mallctl(\"opt.\"#n, &ssv, &sssz, NULL, 0))\t\\\n\t\t == 0) {\t\t\t\t\t\t\\\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\t\t\\\n\t\t\t \" opt.\"#n\": %zd\\n\", ssv);\t\t\t\\\n\t\t}\n#define OPT_WRITE_CHAR_P(n)\t\t\t\t\t\t\\\n\t\tif ((err = je_mallctl(\"opt.\"#n, &cpv, &cpsz, NULL, 0))\t\\\n\t\t == 0) {\t\t\t\t\t\t\\\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\t\t\\\n\t\t\t \" opt.\"#n\": \\\"%s\\\"\\n\", cpv);\t\t\\\n\t\t}\n\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"Run-time option settings:\\n\");\n\t\tOPT_WRITE_BOOL(abort)\n\t\tOPT_WRITE_SIZE_T(lg_chunk)\n\t\tOPT_WRITE_CHAR_P(dss)\n\t\tOPT_WRITE_SIZE_T(narenas)\n\t\tOPT_WRITE_SSIZE_T(lg_dirty_mult)\n\t\tOPT_WRITE_BOOL(stats_print)\n\t\tOPT_WRITE_BOOL(junk)\n\t\tOPT_WRITE_SIZE_T(quarantine)\n\t\tOPT_WRITE_BOOL(redzone)\n\t\tOPT_WRITE_BOOL(zero)\n\t\tOPT_WRITE_BOOL(utrace)\n\t\tOPT_WRITE_BOOL(valgrind)\n\t\tOPT_WRITE_BOOL(xmalloc)\n\t\tOPT_WRITE_BOOL(tcache)\n\t\tOPT_WRITE_SSIZE_T(lg_tcache_max)\n\t\tOPT_WRITE_BOOL(prof)\n\t\tOPT_WRITE_CHAR_P(prof_prefix)\n\t\tOPT_WRITE_BOOL(prof_active)\n\t\tOPT_WRITE_SSIZE_T(lg_prof_sample)\n\t\tOPT_WRITE_BOOL(prof_accum)\n\t\tOPT_WRITE_SSIZE_T(lg_prof_interval)\n\t\tOPT_WRITE_BOOL(prof_gdump)\n\t\tOPT_WRITE_BOOL(prof_final)\n\t\tOPT_WRITE_BOOL(prof_leak)\n\n#undef OPT_WRITE_BOOL\n#undef OPT_WRITE_SIZE_T\n#undef OPT_WRITE_SSIZE_T\n#undef OPT_WRITE_CHAR_P\n\n\t\tmalloc_cprintf(write_cb, cbopaque, \"CPUs: %u\\n\", ncpus);\n\n\t\tCTL_GET(\"arenas.narenas\", &uv, unsigned);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Arenas: %u\\n\", uv);\n\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Pointer size: %zu\\n\",\n\t\t sizeof(void *));\n\n\t\tCTL_GET(\"arenas.quantum\", &sv, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Quantum size: %zu\\n\", sv);\n\n\t\tCTL_GET(\"arenas.page\", &sv, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Page size: %zu\\n\", sv);\n\n\t\tCTL_GET(\"opt.lg_dirty_mult\", &ssv, ssize_t);\n\t\tif (ssv >= 0) {\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t \"Min active:dirty page ratio per arena: %u:1\\n\",\n\t\t\t (1U << ssv));\n\t\t} else {\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t \"Min active:dirty page ratio per arena: N/A\\n\");\n\t\t}\n\t\tif ((err = je_mallctl(\"arenas.tcache_max\", &sv, &ssz, NULL, 0))\n\t\t == 0) {\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t \"Maximum thread-cached size class: %zu\\n\", sv);\n\t\t}\n\t\tif ((err = je_mallctl(\"opt.prof\", &bv, &bsz, NULL, 0)) == 0 &&\n\t\t bv) {\n\t\t\tCTL_GET(\"opt.lg_prof_sample\", &sv, size_t);\n\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t \"Average profile sample interval: %\"PRIu64\n\t\t\t \" (2^%zu)\\n\", (((uint64_t)1U) << sv), sv);\n\n\t\t\tCTL_GET(\"opt.lg_prof_interval\", &ssv, ssize_t);\n\t\t\tif (ssv >= 0) {\n\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t \"Average profile dump interval: %\"PRIu64\n\t\t\t\t \" (2^%zd)\\n\",\n\t\t\t\t (((uint64_t)1U) << ssv), ssv);\n\t\t\t} else {\n\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t \"Average profile dump interval: N/A\\n\");\n\t\t\t}\n\t\t}\n\t\tCTL_GET(\"opt.lg_chunk\", &sv, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"Chunk size: %zu (2^%zu)\\n\",\n\t\t (ZU(1) << sv), sv);\n\t}\n\n\tif (config_stats) {\n\t\tsize_t *cactive;\n\t\tsize_t allocated, active, mapped;\n\t\tsize_t chunks_current, chunks_high;\n\t\tuint64_t chunks_total;\n\t\tsize_t huge_allocated;\n\t\tuint64_t huge_nmalloc, huge_ndalloc;\n\n\t\tCTL_GET(\"stats.cactive\", &cactive, size_t *);\n\t\tCTL_GET(\"stats.allocated\", &allocated, size_t);\n\t\tCTL_GET(\"stats.active\", &active, size_t);\n\t\tCTL_GET(\"stats.mapped\", &mapped, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"Allocated: %zu, active: %zu, mapped: %zu\\n\",\n\t\t allocated, active, mapped);\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"Current active ceiling: %zu\\n\", atomic_read_z(cactive));\n\n\t\t/* Print chunk stats. */\n\t\tCTL_GET(\"stats.chunks.total\", &chunks_total, uint64_t);\n\t\tCTL_GET(\"stats.chunks.high\", &chunks_high, size_t);\n\t\tCTL_GET(\"stats.chunks.current\", &chunks_current, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque, \"chunks: nchunks \"\n\t\t \"highchunks curchunks\\n\");\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \" %13\"PRIu64\" %12zu %12zu\\n\",\n\t\t chunks_total, chunks_high, chunks_current);\n\n\t\t/* Print huge stats. */\n\t\tCTL_GET(\"stats.huge.nmalloc\", &huge_nmalloc, uint64_t);\n\t\tCTL_GET(\"stats.huge.ndalloc\", &huge_ndalloc, uint64_t);\n\t\tCTL_GET(\"stats.huge.allocated\", &huge_allocated, size_t);\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \"huge: nmalloc ndalloc allocated\\n\");\n\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t \" %12\"PRIu64\" %12\"PRIu64\" %12zu\\n\",\n\t\t huge_nmalloc, huge_ndalloc, huge_allocated);\n\n\t\tif (merged) {\n\t\t\tunsigned narenas;\n\n\t\t\tCTL_GET(\"arenas.narenas\", &narenas, unsigned);\n\t\t\t{\n\t\t\t\tVARIABLE_ARRAY(bool, initialized, narenas);\n\t\t\t\tsize_t isz;\n\t\t\t\tunsigned i, ninitialized;\n\n\t\t\t\tisz = sizeof(bool) * narenas;\n\t\t\t\txmallctl(\"arenas.initialized\", initialized,\n\t\t\t\t &isz, NULL, 0);\n\t\t\t\tfor (i = ninitialized = 0; i < narenas; i++) {\n\t\t\t\t\tif (initialized[i])\n\t\t\t\t\t\tninitialized++;\n\t\t\t\t}\n\n\t\t\t\tif (ninitialized > 1 || unmerged == false) {\n\t\t\t\t\t/* Print merged arena stats. */\n\t\t\t\t\tmalloc_cprintf(write_cb, cbopaque,\n\t\t\t\t\t \"\\nMerged arenas stats:\\n\");\n\t\t\t\t\tstats_arena_print(write_cb, cbopaque,\n\t\t\t\t\t narenas, bins, large);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n\t\tif (unmerged) {\n\t\t\tunsigned narenas;\n\n\t\t\t/* Print stats for each arena. */\n\n\t\t\tCTL_GET(\"arenas.narenas\", &narenas, unsigned);\n\t\t\t{\n\t\t\t\tVARIABLE_ARRAY(bool, initialized, narenas);\n\t\t\t\tsize_t isz;\n\t\t\t\tunsigned i;\n\n\t\t\t\tisz = sizeof(bool) * narenas;\n\t\t\t\txmallctl(\"arenas.initialized\", initialized,\n\t\t\t\t &isz, NULL, 0);\n\n\t\t\t\tfor (i = 0; i < narenas; i++) {\n\t\t\t\t\tif (initialized[i]) {\n\t\t\t\t\t\tmalloc_cprintf(write_cb,\n\t\t\t\t\t\t cbopaque,\n\t\t\t\t\t\t \"\\narenas[%u]:\\n\", i);\n\t\t\t\t\t\tstats_arena_print(write_cb,\n\t\t\t\t\t\t cbopaque, i, bins, large);\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\tmalloc_cprintf(write_cb, cbopaque, \"--- End jemalloc statistics ---\\n\");\n}\n" }, { "path": "deps/jemalloc/src/tcache.c", "content": "#define\tJEMALLOC_TCACHE_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nmalloc_tsd_data(, tcache, tcache_t *, NULL)\nmalloc_tsd_data(, tcache_enabled, tcache_enabled_t, tcache_enabled_default)\n\nbool\topt_tcache = true;\nssize_t\topt_lg_tcache_max = LG_TCACHE_MAXCLASS_DEFAULT;\n\ntcache_bin_info_t\t*tcache_bin_info;\nstatic unsigned\t\tstack_nelms; /* Total stack elms per tcache. */\n\nsize_t\t\t\tnhbins;\nsize_t\t\t\ttcache_maxclass;\n\n/******************************************************************************/\n\nsize_t\ttcache_salloc(const void *ptr)\n{\n\n\treturn (arena_salloc(ptr, false));\n}\n\nvoid\ntcache_event_hard(tcache_t *tcache)\n{\n\tsize_t binind = tcache->next_gc_bin;\n\ttcache_bin_t *tbin = &tcache->tbins[binind];\n\ttcache_bin_info_t *tbin_info = &tcache_bin_info[binind];\n\n\tif (tbin->low_water > 0) {\n\t\t/*\n\t\t * Flush (ceiling) 3/4 of the objects below the low water mark.\n\t\t */\n\t\tif (binind < NBINS) {\n\t\t\ttcache_bin_flush_small(tbin, binind, tbin->ncached -\n\t\t\t tbin->low_water + (tbin->low_water >> 2), tcache);\n\t\t} else {\n\t\t\ttcache_bin_flush_large(tbin, binind, tbin->ncached -\n\t\t\t tbin->low_water + (tbin->low_water >> 2), tcache);\n\t\t}\n\t\t/*\n\t\t * Reduce fill count by 2X. Limit lg_fill_div such that the\n\t\t * fill count is always at least 1.\n\t\t */\n\t\tif ((tbin_info->ncached_max >> (tbin->lg_fill_div+1)) >= 1)\n\t\t\ttbin->lg_fill_div++;\n\t} else if (tbin->low_water < 0) {\n\t\t/*\n\t\t * Increase fill count by 2X. Make sure lg_fill_div stays\n\t\t * greater than 0.\n\t\t */\n\t\tif (tbin->lg_fill_div > 1)\n\t\t\ttbin->lg_fill_div--;\n\t}\n\ttbin->low_water = tbin->ncached;\n\n\ttcache->next_gc_bin++;\n\tif (tcache->next_gc_bin == nhbins)\n\t\ttcache->next_gc_bin = 0;\n\ttcache->ev_cnt = 0;\n}\n\nvoid *\ntcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin, size_t binind)\n{\n\tvoid *ret;\n\n\tarena_tcache_fill_small(tcache->arena, tbin, binind,\n\t config_prof ? tcache->prof_accumbytes : 0);\n\tif (config_prof)\n\t\ttcache->prof_accumbytes = 0;\n\tret = tcache_alloc_easy(tbin);\n\n\treturn (ret);\n}\n\nvoid\ntcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem,\n tcache_t *tcache)\n{\n\tvoid *ptr;\n\tunsigned i, nflush, ndeferred;\n\tbool merged_stats = false;\n\n\tassert(binind < NBINS);\n\tassert(rem <= tbin->ncached);\n\n\tfor (nflush = tbin->ncached - rem; nflush > 0; nflush = ndeferred) {\n\t\t/* Lock the arena bin associated with the first object. */\n\t\tarena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(\n\t\t tbin->avail[0]);\n\t\tarena_t *arena = chunk->arena;\n\t\tarena_bin_t *bin = &arena->bins[binind];\n\n\t\tif (config_prof && arena == tcache->arena) {\n\t\t\tmalloc_mutex_lock(&arena->lock);\n\t\t\tarena_prof_accum(arena, tcache->prof_accumbytes);\n\t\t\tmalloc_mutex_unlock(&arena->lock);\n\t\t\ttcache->prof_accumbytes = 0;\n\t\t}\n\n\t\tmalloc_mutex_lock(&bin->lock);\n\t\tif (config_stats && arena == tcache->arena) {\n\t\t\tassert(merged_stats == false);\n\t\t\tmerged_stats = true;\n\t\t\tbin->stats.nflushes++;\n\t\t\tbin->stats.nrequests += tbin->tstats.nrequests;\n\t\t\ttbin->tstats.nrequests = 0;\n\t\t}\n\t\tndeferred = 0;\n\t\tfor (i = 0; i < nflush; i++) {\n\t\t\tptr = tbin->avail[i];\n\t\t\tassert(ptr != NULL);\n\t\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\t\t\tif (chunk->arena == arena) {\n\t\t\t\tsize_t pageind = ((uintptr_t)ptr -\n\t\t\t\t (uintptr_t)chunk) >> LG_PAGE;\n\t\t\t\tarena_chunk_map_t *mapelm =\n\t\t\t\t arena_mapp_get(chunk, pageind);\n\t\t\t\tif (config_fill && opt_junk) {\n\t\t\t\t\tarena_alloc_junk_small(ptr,\n\t\t\t\t\t &arena_bin_info[binind], true);\n\t\t\t\t}\n\t\t\t\tarena_dalloc_bin_locked(arena, chunk, ptr,\n\t\t\t\t mapelm);\n\t\t\t} else {\n\t\t\t\t/*\n\t\t\t\t * This object was allocated via a different\n\t\t\t\t * arena bin than the one that is currently\n\t\t\t\t * locked. Stash the object, so that it can be\n\t\t\t\t * handled in a future pass.\n\t\t\t\t */\n\t\t\t\ttbin->avail[ndeferred] = ptr;\n\t\t\t\tndeferred++;\n\t\t\t}\n\t\t}\n\t\tmalloc_mutex_unlock(&bin->lock);\n\t}\n\tif (config_stats && merged_stats == false) {\n\t\t/*\n\t\t * The flush loop didn't happen to flush to this thread's\n\t\t * arena, so the stats didn't get merged. Manually do so now.\n\t\t */\n\t\tarena_bin_t *bin = &tcache->arena->bins[binind];\n\t\tmalloc_mutex_lock(&bin->lock);\n\t\tbin->stats.nflushes++;\n\t\tbin->stats.nrequests += tbin->tstats.nrequests;\n\t\ttbin->tstats.nrequests = 0;\n\t\tmalloc_mutex_unlock(&bin->lock);\n\t}\n\n\tmemmove(tbin->avail, &tbin->avail[tbin->ncached - rem],\n\t rem * sizeof(void *));\n\ttbin->ncached = rem;\n\tif ((int)tbin->ncached < tbin->low_water)\n\t\ttbin->low_water = tbin->ncached;\n}\n\nvoid\ntcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem,\n tcache_t *tcache)\n{\n\tvoid *ptr;\n\tunsigned i, nflush, ndeferred;\n\tbool merged_stats = false;\n\n\tassert(binind < nhbins);\n\tassert(rem <= tbin->ncached);\n\n\tfor (nflush = tbin->ncached - rem; nflush > 0; nflush = ndeferred) {\n\t\t/* Lock the arena associated with the first object. */\n\t\tarena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(\n\t\t tbin->avail[0]);\n\t\tarena_t *arena = chunk->arena;\n\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tif ((config_prof || config_stats) && arena == tcache->arena) {\n\t\t\tif (config_prof) {\n\t\t\t\tarena_prof_accum(arena,\n\t\t\t\t tcache->prof_accumbytes);\n\t\t\t\ttcache->prof_accumbytes = 0;\n\t\t\t}\n\t\t\tif (config_stats) {\n\t\t\t\tmerged_stats = true;\n\t\t\t\tarena->stats.nrequests_large +=\n\t\t\t\t tbin->tstats.nrequests;\n\t\t\t\tarena->stats.lstats[binind - NBINS].nrequests +=\n\t\t\t\t tbin->tstats.nrequests;\n\t\t\t\ttbin->tstats.nrequests = 0;\n\t\t\t}\n\t\t}\n\t\tndeferred = 0;\n\t\tfor (i = 0; i < nflush; i++) {\n\t\t\tptr = tbin->avail[i];\n\t\t\tassert(ptr != NULL);\n\t\t\tchunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);\n\t\t\tif (chunk->arena == arena)\n\t\t\t\tarena_dalloc_large_locked(arena, chunk, ptr);\n\t\t\telse {\n\t\t\t\t/*\n\t\t\t\t * This object was allocated via a different\n\t\t\t\t * arena than the one that is currently locked.\n\t\t\t\t * Stash the object, so that it can be handled\n\t\t\t\t * in a future pass.\n\t\t\t\t */\n\t\t\t\ttbin->avail[ndeferred] = ptr;\n\t\t\t\tndeferred++;\n\t\t\t}\n\t\t}\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t}\n\tif (config_stats && merged_stats == false) {\n\t\t/*\n\t\t * The flush loop didn't happen to flush to this thread's\n\t\t * arena, so the stats didn't get merged. Manually do so now.\n\t\t */\n\t\tarena_t *arena = tcache->arena;\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tarena->stats.nrequests_large += tbin->tstats.nrequests;\n\t\tarena->stats.lstats[binind - NBINS].nrequests +=\n\t\t tbin->tstats.nrequests;\n\t\ttbin->tstats.nrequests = 0;\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t}\n\n\tmemmove(tbin->avail, &tbin->avail[tbin->ncached - rem],\n\t rem * sizeof(void *));\n\ttbin->ncached = rem;\n\tif ((int)tbin->ncached < tbin->low_water)\n\t\ttbin->low_water = tbin->ncached;\n}\n\nvoid\ntcache_arena_associate(tcache_t *tcache, arena_t *arena)\n{\n\n\tif (config_stats) {\n\t\t/* Link into list of extant tcaches. */\n\t\tmalloc_mutex_lock(&arena->lock);\n\t\tql_elm_new(tcache, link);\n\t\tql_tail_insert(&arena->tcache_ql, tcache, link);\n\t\tmalloc_mutex_unlock(&arena->lock);\n\t}\n\ttcache->arena = arena;\n}\n\nvoid\ntcache_arena_dissociate(tcache_t *tcache)\n{\n\n\tif (config_stats) {\n\t\t/* Unlink from list of extant tcaches. */\n\t\tmalloc_mutex_lock(&tcache->arena->lock);\n\t\tql_remove(&tcache->arena->tcache_ql, tcache, link);\n\t\tmalloc_mutex_unlock(&tcache->arena->lock);\n\t\ttcache_stats_merge(tcache, tcache->arena);\n\t}\n}\n\ntcache_t *\ntcache_create(arena_t *arena)\n{\n\ttcache_t *tcache;\n\tsize_t size, stack_offset;\n\tunsigned i;\n\n\tsize = offsetof(tcache_t, tbins) + (sizeof(tcache_bin_t) * nhbins);\n\t/* Naturally align the pointer stacks. */\n\tsize = PTR_CEILING(size);\n\tstack_offset = size;\n\tsize += stack_nelms * sizeof(void *);\n\t/*\n\t * Round up to the nearest multiple of the cacheline size, in order to\n\t * avoid the possibility of false cacheline sharing.\n\t *\n\t * That this works relies on the same logic as in ipalloc(), but we\n\t * cannot directly call ipalloc() here due to tcache bootstrapping\n\t * issues.\n\t */\n\tsize = (size + CACHELINE_MASK) & (-CACHELINE);\n\n\tif (size <= SMALL_MAXCLASS)\n\t\ttcache = (tcache_t *)arena_malloc_small(arena, size, true);\n\telse if (size <= tcache_maxclass)\n\t\ttcache = (tcache_t *)arena_malloc_large(arena, size, true);\n\telse\n\t\ttcache = (tcache_t *)icallocx(size, false, arena);\n\n\tif (tcache == NULL)\n\t\treturn (NULL);\n\n\ttcache_arena_associate(tcache, arena);\n\n\tassert((TCACHE_NSLOTS_SMALL_MAX & 1U) == 0);\n\tfor (i = 0; i < nhbins; i++) {\n\t\ttcache->tbins[i].lg_fill_div = 1;\n\t\ttcache->tbins[i].avail = (void **)((uintptr_t)tcache +\n\t\t (uintptr_t)stack_offset);\n\t\tstack_offset += tcache_bin_info[i].ncached_max * sizeof(void *);\n\t}\n\n\ttcache_tsd_set(&tcache);\n\n\treturn (tcache);\n}\n\nvoid\ntcache_destroy(tcache_t *tcache)\n{\n\tunsigned i;\n\tsize_t tcache_size;\n\n\ttcache_arena_dissociate(tcache);\n\n\tfor (i = 0; i < NBINS; i++) {\n\t\ttcache_bin_t *tbin = &tcache->tbins[i];\n\t\ttcache_bin_flush_small(tbin, i, 0, tcache);\n\n\t\tif (config_stats && tbin->tstats.nrequests != 0) {\n\t\t\tarena_t *arena = tcache->arena;\n\t\t\tarena_bin_t *bin = &arena->bins[i];\n\t\t\tmalloc_mutex_lock(&bin->lock);\n\t\t\tbin->stats.nrequests += tbin->tstats.nrequests;\n\t\t\tmalloc_mutex_unlock(&bin->lock);\n\t\t}\n\t}\n\n\tfor (; i < nhbins; i++) {\n\t\ttcache_bin_t *tbin = &tcache->tbins[i];\n\t\ttcache_bin_flush_large(tbin, i, 0, tcache);\n\n\t\tif (config_stats && tbin->tstats.nrequests != 0) {\n\t\t\tarena_t *arena = tcache->arena;\n\t\t\tmalloc_mutex_lock(&arena->lock);\n\t\t\tarena->stats.nrequests_large += tbin->tstats.nrequests;\n\t\t\tarena->stats.lstats[i - NBINS].nrequests +=\n\t\t\t tbin->tstats.nrequests;\n\t\t\tmalloc_mutex_unlock(&arena->lock);\n\t\t}\n\t}\n\n\tif (config_prof && tcache->prof_accumbytes > 0) {\n\t\tmalloc_mutex_lock(&tcache->arena->lock);\n\t\tarena_prof_accum(tcache->arena, tcache->prof_accumbytes);\n\t\tmalloc_mutex_unlock(&tcache->arena->lock);\n\t}\n\n\ttcache_size = arena_salloc(tcache, false);\n\tif (tcache_size <= SMALL_MAXCLASS) {\n\t\tarena_chunk_t *chunk = CHUNK_ADDR2BASE(tcache);\n\t\tarena_t *arena = chunk->arena;\n\t\tsize_t pageind = ((uintptr_t)tcache - (uintptr_t)chunk) >>\n\t\t LG_PAGE;\n\t\tarena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind);\n\n\t\tarena_dalloc_bin(arena, chunk, tcache, pageind, mapelm);\n\t} else if (tcache_size <= tcache_maxclass) {\n\t\tarena_chunk_t *chunk = CHUNK_ADDR2BASE(tcache);\n\t\tarena_t *arena = chunk->arena;\n\n\t\tarena_dalloc_large(arena, chunk, tcache);\n\t} else\n\t\tidallocx(tcache, false);\n}\n\nvoid\ntcache_thread_cleanup(void *arg)\n{\n\ttcache_t *tcache = *(tcache_t **)arg;\n\n\tif (tcache == TCACHE_STATE_DISABLED) {\n\t\t/* Do nothing. */\n\t} else if (tcache == TCACHE_STATE_REINCARNATED) {\n\t\t/*\n\t\t * Another destructor called an allocator function after this\n\t\t * destructor was called. Reset tcache to\n\t\t * TCACHE_STATE_PURGATORY in order to receive another callback.\n\t\t */\n\t\ttcache = TCACHE_STATE_PURGATORY;\n\t\ttcache_tsd_set(&tcache);\n\t} else if (tcache == TCACHE_STATE_PURGATORY) {\n\t\t/*\n\t\t * The previous time this destructor was called, we set the key\n\t\t * to TCACHE_STATE_PURGATORY so that other destructors wouldn't\n\t\t * cause re-creation of the tcache. This time, do nothing, so\n\t\t * that the destructor will not be called again.\n\t\t */\n\t} else if (tcache != NULL) {\n\t\tassert(tcache != TCACHE_STATE_PURGATORY);\n\t\ttcache_destroy(tcache);\n\t\ttcache = TCACHE_STATE_PURGATORY;\n\t\ttcache_tsd_set(&tcache);\n\t}\n}\n\nvoid\ntcache_stats_merge(tcache_t *tcache, arena_t *arena)\n{\n\tunsigned i;\n\n\t/* Merge and reset tcache stats. */\n\tfor (i = 0; i < NBINS; i++) {\n\t\tarena_bin_t *bin = &arena->bins[i];\n\t\ttcache_bin_t *tbin = &tcache->tbins[i];\n\t\tmalloc_mutex_lock(&bin->lock);\n\t\tbin->stats.nrequests += tbin->tstats.nrequests;\n\t\tmalloc_mutex_unlock(&bin->lock);\n\t\ttbin->tstats.nrequests = 0;\n\t}\n\n\tfor (; i < nhbins; i++) {\n\t\tmalloc_large_stats_t *lstats = &arena->stats.lstats[i - NBINS];\n\t\ttcache_bin_t *tbin = &tcache->tbins[i];\n\t\tarena->stats.nrequests_large += tbin->tstats.nrequests;\n\t\tlstats->nrequests += tbin->tstats.nrequests;\n\t\ttbin->tstats.nrequests = 0;\n\t}\n}\n\nbool\ntcache_boot0(void)\n{\n\tunsigned i;\n\n\t/*\n\t * If necessary, clamp opt_lg_tcache_max, now that arena_maxclass is\n\t * known.\n\t */\n\tif (opt_lg_tcache_max < 0 || (1U << opt_lg_tcache_max) < SMALL_MAXCLASS)\n\t\ttcache_maxclass = SMALL_MAXCLASS;\n\telse if ((1U << opt_lg_tcache_max) > arena_maxclass)\n\t\ttcache_maxclass = arena_maxclass;\n\telse\n\t\ttcache_maxclass = (1U << opt_lg_tcache_max);\n\n\tnhbins = NBINS + (tcache_maxclass >> LG_PAGE);\n\n\t/* Initialize tcache_bin_info. */\n\ttcache_bin_info = (tcache_bin_info_t *)base_alloc(nhbins *\n\t sizeof(tcache_bin_info_t));\n\tif (tcache_bin_info == NULL)\n\t\treturn (true);\n\tstack_nelms = 0;\n\tfor (i = 0; i < NBINS; i++) {\n\t\tif ((arena_bin_info[i].nregs << 1) <= TCACHE_NSLOTS_SMALL_MAX) {\n\t\t\ttcache_bin_info[i].ncached_max =\n\t\t\t (arena_bin_info[i].nregs << 1);\n\t\t} else {\n\t\t\ttcache_bin_info[i].ncached_max =\n\t\t\t TCACHE_NSLOTS_SMALL_MAX;\n\t\t}\n\t\tstack_nelms += tcache_bin_info[i].ncached_max;\n\t}\n\tfor (; i < nhbins; i++) {\n\t\ttcache_bin_info[i].ncached_max = TCACHE_NSLOTS_LARGE;\n\t\tstack_nelms += tcache_bin_info[i].ncached_max;\n\t}\n\n\treturn (false);\n}\n\nbool\ntcache_boot1(void)\n{\n\n\tif (tcache_tsd_boot() || tcache_enabled_tsd_boot())\n\t\treturn (true);\n\n\treturn (false);\n}\n" }, { "path": "deps/jemalloc/src/tsd.c", "content": "#define\tJEMALLOC_TSD_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Data. */\n\nstatic unsigned ncleanups;\nstatic malloc_tsd_cleanup_t cleanups[MALLOC_TSD_CLEANUPS_MAX];\n\n/******************************************************************************/\n\nvoid *\nmalloc_tsd_malloc(size_t size)\n{\n\n\t/* Avoid choose_arena() in order to dodge bootstrapping issues. */\n\treturn (arena_malloc(arenas[0], size, false, false));\n}\n\nvoid\nmalloc_tsd_dalloc(void *wrapper)\n{\n\n\tidalloc(wrapper);\n}\n\nvoid\nmalloc_tsd_no_cleanup(void *arg)\n{\n\n\tnot_reached();\n}\n\n#if defined(JEMALLOC_MALLOC_THREAD_CLEANUP) || defined(_WIN32)\n#ifndef _WIN32\nJEMALLOC_EXPORT\n#endif\nvoid\n_malloc_thread_cleanup(void)\n{\n\tbool pending[MALLOC_TSD_CLEANUPS_MAX], again;\n\tunsigned i;\n\n\tfor (i = 0; i < ncleanups; i++)\n\t\tpending[i] = true;\n\n\tdo {\n\t\tagain = false;\n\t\tfor (i = 0; i < ncleanups; i++) {\n\t\t\tif (pending[i]) {\n\t\t\t\tpending[i] = cleanups[i]();\n\t\t\t\tif (pending[i])\n\t\t\t\t\tagain = true;\n\t\t\t}\n\t\t}\n\t} while (again);\n}\n#endif\n\nvoid\nmalloc_tsd_cleanup_register(bool (*f)(void))\n{\n\n\tassert(ncleanups < MALLOC_TSD_CLEANUPS_MAX);\n\tcleanups[ncleanups] = f;\n\tncleanups++;\n}\n\nvoid\nmalloc_tsd_boot(void)\n{\n\n\tncleanups = 0;\n}\n\n#ifdef _WIN32\nstatic BOOL WINAPI\n_tls_callback(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)\n{\n\n\tswitch (fdwReason) {\n#ifdef JEMALLOC_LAZY_LOCK\n\tcase DLL_THREAD_ATTACH:\n\t\tisthreaded = true;\n\t\tbreak;\n#endif\n\tcase DLL_THREAD_DETACH:\n\t\t_malloc_thread_cleanup();\n\t\tbreak;\n\tdefault:\n\t\tbreak;\n\t}\n\treturn (true);\n}\n\n#ifdef _MSC_VER\n# ifdef _M_IX86\n# pragma comment(linker, \"/INCLUDE:__tls_used\")\n# else\n# pragma comment(linker, \"/INCLUDE:_tls_used\")\n# endif\n# pragma section(\".CRT$XLY\",long,read)\n#endif\nJEMALLOC_SECTION(\".CRT$XLY\") JEMALLOC_ATTR(used)\nstatic const BOOL\t(WINAPI *tls_callback)(HINSTANCE hinstDLL,\n DWORD fdwReason, LPVOID lpvReserved) = _tls_callback;\n#endif\n" }, { "path": "deps/jemalloc/src/util.c", "content": "#define\tassert(e) do {\t\t\t\t\t\t\t\\\n\tif (config_debug && !(e)) {\t\t\t\t\t\\\n\t\tmalloc_write(\": Failed assertion\\n\");\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tnot_reached() do {\t\t\t\t\t\t\\\n\tif (config_debug) {\t\t\t\t\t\t\\\n\t\tmalloc_write(\": Unreachable code reached\\n\");\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tnot_implemented() do {\t\t\t\t\t\t\\\n\tif (config_debug) {\t\t\t\t\t\t\\\n\t\tmalloc_write(\": Not implemented\\n\");\t\t\\\n\t\tabort();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n#define\tJEMALLOC_UTIL_C_\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic void\twrtmessage(void *cbopaque, const char *s);\n#define\tU2S_BUFSIZE\t((1U << (LG_SIZEOF_INTMAX_T + 3)) + 1)\nstatic char\t*u2s(uintmax_t x, unsigned base, bool uppercase, char *s,\n size_t *slen_p);\n#define\tD2S_BUFSIZE\t(1 + U2S_BUFSIZE)\nstatic char\t*d2s(intmax_t x, char sign, char *s, size_t *slen_p);\n#define\tO2S_BUFSIZE\t(1 + U2S_BUFSIZE)\nstatic char\t*o2s(uintmax_t x, bool alt_form, char *s, size_t *slen_p);\n#define\tX2S_BUFSIZE\t(2 + U2S_BUFSIZE)\nstatic char\t*x2s(uintmax_t x, bool alt_form, bool uppercase, char *s,\n size_t *slen_p);\n\n/******************************************************************************/\n\n/* malloc_message() setup. */\nstatic void\nwrtmessage(void *cbopaque, const char *s)\n{\n\n#ifdef SYS_write\n\t/*\n\t * Use syscall(2) rather than write(2) when possible in order to avoid\n\t * the possibility of memory allocation within libc. This is necessary\n\t * on FreeBSD; most operating systems do not have this problem though.\n\t */\n\tUNUSED int result = syscall(SYS_write, STDERR_FILENO, s, strlen(s));\n#else\n\tUNUSED int result = write(STDERR_FILENO, s, strlen(s));\n#endif\n}\n\nJEMALLOC_EXPORT void\t(*je_malloc_message)(void *, const char *s);\n\n/*\n * Wrapper around malloc_message() that avoids the need for\n * je_malloc_message(...) throughout the code.\n */\nvoid\nmalloc_write(const char *s)\n{\n\n\tif (je_malloc_message != NULL)\n\t\tje_malloc_message(NULL, s);\n\telse\n\t\twrtmessage(NULL, s);\n}\n\n/*\n * glibc provides a non-standard strerror_r() when _GNU_SOURCE is defined, so\n * provide a wrapper.\n */\nint\nbuferror(char *buf, size_t buflen)\n{\n\n#ifdef _WIN32\n\tFormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM, NULL, GetLastError(), 0,\n\t (LPSTR)buf, buflen, NULL);\n\treturn (0);\n#elif defined(_GNU_SOURCE)\n\tchar *b = strerror_r(errno, buf, buflen);\n\tif (b != buf) {\n\t\tstrncpy(buf, b, buflen);\n\t\tbuf[buflen-1] = '\\0';\n\t}\n\treturn (0);\n#else\n\treturn (strerror_r(errno, buf, buflen));\n#endif\n}\n\nuintmax_t\nmalloc_strtoumax(const char *nptr, char **endptr, int base)\n{\n\tuintmax_t ret, digit;\n\tint b;\n\tbool neg;\n\tconst char *p, *ns;\n\n\tif (base < 0 || base == 1 || base > 36) {\n\t\tset_errno(EINVAL);\n\t\treturn (UINTMAX_MAX);\n\t}\n\tb = base;\n\n\t/* Swallow leading whitespace and get sign, if any. */\n\tneg = false;\n\tp = nptr;\n\twhile (true) {\n\t\tswitch (*p) {\n\t\tcase '\\t': case '\\n': case '\\v': case '\\f': case '\\r': case ' ':\n\t\t\tp++;\n\t\t\tbreak;\n\t\tcase '-':\n\t\t\tneg = true;\n\t\t\t/* Fall through. */\n\t\tcase '+':\n\t\t\tp++;\n\t\t\t/* Fall through. */\n\t\tdefault:\n\t\t\tgoto label_prefix;\n\t\t}\n\t}\n\n\t/* Get prefix, if any. */\n\tlabel_prefix:\n\t/*\n\t * Note where the first non-whitespace/sign character is so that it is\n\t * possible to tell whether any digits are consumed (e.g., \" 0\" vs.\n\t * \" -x\").\n\t */\n\tns = p;\n\tif (*p == '0') {\n\t\tswitch (p[1]) {\n\t\tcase '0': case '1': case '2': case '3': case '4': case '5':\n\t\tcase '6': case '7':\n\t\t\tif (b == 0)\n\t\t\t\tb = 8;\n\t\t\tif (b == 8)\n\t\t\t\tp++;\n\t\t\tbreak;\n\t\tcase 'x':\n\t\t\tswitch (p[2]) {\n\t\t\tcase '0': case '1': case '2': case '3': case '4':\n\t\t\tcase '5': case '6': case '7': case '8': case '9':\n\t\t\tcase 'A': case 'B': case 'C': case 'D': case 'E':\n\t\t\tcase 'F':\n\t\t\tcase 'a': case 'b': case 'c': case 'd': case 'e':\n\t\t\tcase 'f':\n\t\t\t\tif (b == 0)\n\t\t\t\t\tb = 16;\n\t\t\t\tif (b == 16)\n\t\t\t\t\tp += 2;\n\t\t\t\tbreak;\n\t\t\tdefault:\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\tbreak;\n\t\tdefault:\n\t\t\tbreak;\n\t\t}\n\t}\n\tif (b == 0)\n\t\tb = 10;\n\n\t/* Convert. */\n\tret = 0;\n\twhile ((*p >= '0' && *p <= '9' && (digit = *p - '0') < b)\n\t || (*p >= 'A' && *p <= 'Z' && (digit = 10 + *p - 'A') < b)\n\t || (*p >= 'a' && *p <= 'z' && (digit = 10 + *p - 'a') < b)) {\n\t\tuintmax_t pret = ret;\n\t\tret *= b;\n\t\tret += digit;\n\t\tif (ret < pret) {\n\t\t\t/* Overflow. */\n\t\t\tset_errno(ERANGE);\n\t\t\treturn (UINTMAX_MAX);\n\t\t}\n\t\tp++;\n\t}\n\tif (neg)\n\t\tret = -ret;\n\n\tif (endptr != NULL) {\n\t\tif (p == ns) {\n\t\t\t/* No characters were converted. */\n\t\t\t*endptr = (char *)nptr;\n\t\t} else\n\t\t\t*endptr = (char *)p;\n\t}\n\n\treturn (ret);\n}\n\nstatic char *\nu2s(uintmax_t x, unsigned base, bool uppercase, char *s, size_t *slen_p)\n{\n\tunsigned i;\n\n\ti = U2S_BUFSIZE - 1;\n\ts[i] = '\\0';\n\tswitch (base) {\n\tcase 10:\n\t\tdo {\n\t\t\ti--;\n\t\t\ts[i] = \"0123456789\"[x % (uint64_t)10];\n\t\t\tx /= (uint64_t)10;\n\t\t} while (x > 0);\n\t\tbreak;\n\tcase 16: {\n\t\tconst char *digits = (uppercase)\n\t\t ? \"0123456789ABCDEF\"\n\t\t : \"0123456789abcdef\";\n\n\t\tdo {\n\t\t\ti--;\n\t\t\ts[i] = digits[x & 0xf];\n\t\t\tx >>= 4;\n\t\t} while (x > 0);\n\t\tbreak;\n\t} default: {\n\t\tconst char *digits = (uppercase)\n\t\t ? \"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ\"\n\t\t : \"0123456789abcdefghijklmnopqrstuvwxyz\";\n\n\t\tassert(base >= 2 && base <= 36);\n\t\tdo {\n\t\t\ti--;\n\t\t\ts[i] = digits[x % (uint64_t)base];\n\t\t\tx /= (uint64_t)base;\n\t\t} while (x > 0);\n\t}}\n\n\t*slen_p = U2S_BUFSIZE - 1 - i;\n\treturn (&s[i]);\n}\n\nstatic char *\nd2s(intmax_t x, char sign, char *s, size_t *slen_p)\n{\n\tbool neg;\n\n\tif ((neg = (x < 0)))\n\t\tx = -x;\n\ts = u2s(x, 10, false, s, slen_p);\n\tif (neg)\n\t\tsign = '-';\n\tswitch (sign) {\n\tcase '-':\n\t\tif (neg == false)\n\t\t\tbreak;\n\t\t/* Fall through. */\n\tcase ' ':\n\tcase '+':\n\t\ts--;\n\t\t(*slen_p)++;\n\t\t*s = sign;\n\t\tbreak;\n\tdefault: not_reached();\n\t}\n\treturn (s);\n}\n\nstatic char *\no2s(uintmax_t x, bool alt_form, char *s, size_t *slen_p)\n{\n\n\ts = u2s(x, 8, false, s, slen_p);\n\tif (alt_form && *s != '0') {\n\t\ts--;\n\t\t(*slen_p)++;\n\t\t*s = '0';\n\t}\n\treturn (s);\n}\n\nstatic char *\nx2s(uintmax_t x, bool alt_form, bool uppercase, char *s, size_t *slen_p)\n{\n\n\ts = u2s(x, 16, uppercase, s, slen_p);\n\tif (alt_form) {\n\t\ts -= 2;\n\t\t(*slen_p) += 2;\n\t\tmemcpy(s, uppercase ? \"0X\" : \"0x\", 2);\n\t}\n\treturn (s);\n}\n\nint\nmalloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)\n{\n\tint ret;\n\tsize_t i;\n\tconst char *f;\n\n#define\tAPPEND_C(c) do {\t\t\t\t\t\t\\\n\tif (i < size)\t\t\t\t\t\t\t\\\n\t\tstr[i] = (c);\t\t\t\t\t\t\\\n\ti++;\t\t\t\t\t\t\t\t\\\n} while (0)\n#define\tAPPEND_S(s, slen) do {\t\t\t\t\t\t\\\n\tif (i < size) {\t\t\t\t\t\t\t\\\n\t\tsize_t cpylen = (slen <= size - i) ? slen : size - i;\t\\\n\t\tmemcpy(&str[i], s, cpylen);\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\ti += slen;\t\t\t\t\t\t\t\\\n} while (0)\n#define\tAPPEND_PADDED_S(s, slen, width, left_justify) do {\t\t\\\n\t/* Left padding. */\t\t\t\t\t\t\\\n\tsize_t pad_len = (width == -1) ? 0 : ((slen < (size_t)width) ?\t\\\n\t (size_t)width - slen : 0);\t\t\t\t\t\\\n\tif (left_justify == false && pad_len != 0) {\t\t\t\\\n\t\tsize_t j;\t\t\t\t\t\t\\\n\t\tfor (j = 0; j < pad_len; j++)\t\t\t\t\\\n\t\t\tAPPEND_C(' ');\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n\t/* Value. */\t\t\t\t\t\t\t\\\n\tAPPEND_S(s, slen);\t\t\t\t\t\t\\\n\t/* Right padding. */\t\t\t\t\t\t\\\n\tif (left_justify && pad_len != 0) {\t\t\t\t\\\n\t\tsize_t j;\t\t\t\t\t\t\\\n\t\tfor (j = 0; j < pad_len; j++)\t\t\t\t\\\n\t\t\tAPPEND_C(' ');\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n#define GET_ARG_NUMERIC(val, len) do {\t\t\t\t\t\\\n\tswitch (len) {\t\t\t\t\t\t\t\\\n\tcase '?':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, int);\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase '?' | 0x80:\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, unsigned int);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'l':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, long);\t\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'l' | 0x80:\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, unsigned long);\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'q':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, long long);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'q' | 0x80:\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, unsigned long long);\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'j':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, intmax_t);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 't':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, ptrdiff_t);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'z':\t\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, ssize_t);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'z' | 0x80:\t\t\t\t\t\t\\\n\t\tval = va_arg(ap, size_t);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tcase 'p': /* Synthetic; used for %p. */\t\t\t\t\\\n\t\tval = va_arg(ap, uintptr_t);\t\t\t\t\\\n\t\tbreak;\t\t\t\t\t\t\t\\\n\tdefault: not_reached();\t\t\t\t\t\t\\\n\t}\t\t\t\t\t\t\t\t\\\n} while (0)\n\n\ti = 0;\n\tf = format;\n\twhile (true) {\n\t\tswitch (*f) {\n\t\tcase '\\0': goto label_out;\n\t\tcase '%': {\n\t\t\tbool alt_form = false;\n\t\t\tbool left_justify = false;\n\t\t\tbool plus_space = false;\n\t\t\tbool plus_plus = false;\n\t\t\tint prec = -1;\n\t\t\tint width = -1;\n\t\t\tunsigned char len = '?';\n\n\t\t\tf++;\n\t\t\tif (*f == '%') {\n\t\t\t\t/* %% */\n\t\t\t\tAPPEND_C(*f);\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\t/* Flags. */\n\t\t\twhile (true) {\n\t\t\t\tswitch (*f) {\n\t\t\t\tcase '#':\n\t\t\t\t\tassert(alt_form == false);\n\t\t\t\t\talt_form = true;\n\t\t\t\t\tbreak;\n\t\t\t\tcase '-':\n\t\t\t\t\tassert(left_justify == false);\n\t\t\t\t\tleft_justify = true;\n\t\t\t\t\tbreak;\n\t\t\t\tcase ' ':\n\t\t\t\t\tassert(plus_space == false);\n\t\t\t\t\tplus_space = true;\n\t\t\t\t\tbreak;\n\t\t\t\tcase '+':\n\t\t\t\t\tassert(plus_plus == false);\n\t\t\t\t\tplus_plus = true;\n\t\t\t\t\tbreak;\n\t\t\t\tdefault: goto label_width;\n\t\t\t\t}\n\t\t\t\tf++;\n\t\t\t}\n\t\t\t/* Width. */\n\t\t\tlabel_width:\n\t\t\tswitch (*f) {\n\t\t\tcase '*':\n\t\t\t\twidth = va_arg(ap, int);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tcase '0': case '1': case '2': case '3': case '4':\n\t\t\tcase '5': case '6': case '7': case '8': case '9': {\n\t\t\t\tuintmax_t uwidth;\n\t\t\t\tset_errno(0);\n\t\t\t\tuwidth = malloc_strtoumax(f, (char **)&f, 10);\n\t\t\t\tassert(uwidth != UINTMAX_MAX || get_errno() !=\n\t\t\t\t ERANGE);\n\t\t\t\twidth = (int)uwidth;\n\t\t\t\tif (*f == '.') {\n\t\t\t\t\tf++;\n\t\t\t\t\tgoto label_precision;\n\t\t\t\t} else\n\t\t\t\t\tgoto label_length;\n\t\t\t\tbreak;\n\t\t\t} case '.':\n\t\t\t\tf++;\n\t\t\t\tgoto label_precision;\n\t\t\tdefault: goto label_length;\n\t\t\t}\n\t\t\t/* Precision. */\n\t\t\tlabel_precision:\n\t\t\tswitch (*f) {\n\t\t\tcase '*':\n\t\t\t\tprec = va_arg(ap, int);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tcase '0': case '1': case '2': case '3': case '4':\n\t\t\tcase '5': case '6': case '7': case '8': case '9': {\n\t\t\t\tuintmax_t uprec;\n\t\t\t\tset_errno(0);\n\t\t\t\tuprec = malloc_strtoumax(f, (char **)&f, 10);\n\t\t\t\tassert(uprec != UINTMAX_MAX || get_errno() !=\n\t\t\t\t ERANGE);\n\t\t\t\tprec = (int)uprec;\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\tdefault: break;\n\t\t\t}\n\t\t\t/* Length. */\n\t\t\tlabel_length:\n\t\t\tswitch (*f) {\n\t\t\tcase 'l':\n\t\t\t\tf++;\n\t\t\t\tif (*f == 'l') {\n\t\t\t\t\tlen = 'q';\n\t\t\t\t\tf++;\n\t\t\t\t} else\n\t\t\t\t\tlen = 'l';\n\t\t\t\tbreak;\n\t\t\tcase 'j':\n\t\t\t\tlen = 'j';\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tcase 't':\n\t\t\t\tlen = 't';\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tcase 'z':\n\t\t\t\tlen = 'z';\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tdefault: break;\n\t\t\t}\n\t\t\t/* Conversion specifier. */\n\t\t\tswitch (*f) {\n\t\t\t\tchar *s;\n\t\t\t\tsize_t slen;\n\t\t\tcase 'd': case 'i': {\n\t\t\t\tintmax_t val JEMALLOC_CC_SILENCE_INIT(0);\n\t\t\t\tchar buf[D2S_BUFSIZE];\n\n\t\t\t\tGET_ARG_NUMERIC(val, len);\n\t\t\t\ts = d2s(val, (plus_plus ? '+' : (plus_space ?\n\t\t\t\t ' ' : '-')), buf, &slen);\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t} case 'o': {\n\t\t\t\tuintmax_t val JEMALLOC_CC_SILENCE_INIT(0);\n\t\t\t\tchar buf[O2S_BUFSIZE];\n\n\t\t\t\tGET_ARG_NUMERIC(val, len | 0x80);\n\t\t\t\ts = o2s(val, alt_form, buf, &slen);\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t} case 'u': {\n\t\t\t\tuintmax_t val JEMALLOC_CC_SILENCE_INIT(0);\n\t\t\t\tchar buf[U2S_BUFSIZE];\n\n\t\t\t\tGET_ARG_NUMERIC(val, len | 0x80);\n\t\t\t\ts = u2s(val, 10, false, buf, &slen);\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t} case 'x': case 'X': {\n\t\t\t\tuintmax_t val JEMALLOC_CC_SILENCE_INIT(0);\n\t\t\t\tchar buf[X2S_BUFSIZE];\n\n\t\t\t\tGET_ARG_NUMERIC(val, len | 0x80);\n\t\t\t\ts = x2s(val, alt_form, *f == 'X', buf, &slen);\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t} case 'c': {\n\t\t\t\tunsigned char val;\n\t\t\t\tchar buf[2];\n\n\t\t\t\tassert(len == '?' || len == 'l');\n\t\t\t\tassert_not_implemented(len != 'l');\n\t\t\t\tval = va_arg(ap, int);\n\t\t\t\tbuf[0] = val;\n\t\t\t\tbuf[1] = '\\0';\n\t\t\t\tAPPEND_PADDED_S(buf, 1, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t} case 's':\n\t\t\t\tassert(len == '?' || len == 'l');\n\t\t\t\tassert_not_implemented(len != 'l');\n\t\t\t\ts = va_arg(ap, char *);\n\t\t\t\tslen = (prec == -1) ? strlen(s) : prec;\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\tcase 'p': {\n\t\t\t\tuintmax_t val;\n\t\t\t\tchar buf[X2S_BUFSIZE];\n\n\t\t\t\tGET_ARG_NUMERIC(val, 'p');\n\t\t\t\ts = x2s(val, true, false, buf, &slen);\n\t\t\t\tAPPEND_PADDED_S(s, slen, width, left_justify);\n\t\t\t\tf++;\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\tdefault: not_implemented();\n\t\t\t}\n\t\t\tbreak;\n\t\t} default: {\n\t\t\tAPPEND_C(*f);\n\t\t\tf++;\n\t\t\tbreak;\n\t\t}}\n\t}\n\tlabel_out:\n\tif (i < size)\n\t\tstr[i] = '\\0';\n\telse\n\t\tstr[size - 1] = '\\0';\n\tret = i;\n\n#undef APPEND_C\n#undef APPEND_S\n#undef APPEND_PADDED_S\n#undef GET_ARG_NUMERIC\n\treturn (ret);\n}\n\nJEMALLOC_ATTR(format(printf, 3, 4))\nint\nmalloc_snprintf(char *str, size_t size, const char *format, ...)\n{\n\tint ret;\n\tva_list ap;\n\n\tva_start(ap, format);\n\tret = malloc_vsnprintf(str, size, format, ap);\n\tva_end(ap);\n\n\treturn (ret);\n}\n\nvoid\nmalloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque,\n const char *format, va_list ap)\n{\n\tchar buf[MALLOC_PRINTF_BUFSIZE];\n\n\tif (write_cb == NULL) {\n\t\t/*\n\t\t * The caller did not provide an alternate write_cb callback\n\t\t * function, so use the default one. malloc_write() is an\n\t\t * inline function, so use malloc_message() directly here.\n\t\t */\n\t\twrite_cb = (je_malloc_message != NULL) ? je_malloc_message :\n\t\t wrtmessage;\n\t\tcbopaque = NULL;\n\t}\n\n\tmalloc_vsnprintf(buf, sizeof(buf), format, ap);\n\twrite_cb(cbopaque, buf);\n}\n\n/*\n * Print to a callback function in such a way as to (hopefully) avoid memory\n * allocation.\n */\nJEMALLOC_ATTR(format(printf, 3, 4))\nvoid\nmalloc_cprintf(void (*write_cb)(void *, const char *), void *cbopaque,\n const char *format, ...)\n{\n\tva_list ap;\n\n\tva_start(ap, format);\n\tmalloc_vcprintf(write_cb, cbopaque, format, ap);\n\tva_end(ap);\n}\n\n/* Print to stderr in such a way as to avoid memory allocation. */\nJEMALLOC_ATTR(format(printf, 1, 2))\nvoid\nmalloc_printf(const char *format, ...)\n{\n\tva_list ap;\n\n\tva_start(ap, format);\n\tmalloc_vcprintf(NULL, NULL, format, ap);\n\tva_end(ap);\n}\n" }, { "path": "deps/jemalloc/src/zone.c", "content": "#include \"jemalloc/internal/jemalloc_internal.h\"\n#ifndef JEMALLOC_ZONE\n# error \"This source file is for zones on Darwin (OS X).\"\n#endif\n\n/*\n * The malloc_default_purgeable_zone function is only available on >= 10.6.\n * We need to check whether it is present at runtime, thus the weak_import.\n */\nextern malloc_zone_t *malloc_default_purgeable_zone(void)\nJEMALLOC_ATTR(weak_import);\n\n/******************************************************************************/\n/* Data. */\n\nstatic malloc_zone_t zone;\nstatic struct malloc_introspection_t zone_introspect;\n\n/******************************************************************************/\n/* Function prototypes for non-inline static functions. */\n\nstatic size_t\tzone_size(malloc_zone_t *zone, void *ptr);\nstatic void\t*zone_malloc(malloc_zone_t *zone, size_t size);\nstatic void\t*zone_calloc(malloc_zone_t *zone, size_t num, size_t size);\nstatic void\t*zone_valloc(malloc_zone_t *zone, size_t size);\nstatic void\tzone_free(malloc_zone_t *zone, void *ptr);\nstatic void\t*zone_realloc(malloc_zone_t *zone, void *ptr, size_t size);\n#if (JEMALLOC_ZONE_VERSION >= 5)\nstatic void\t*zone_memalign(malloc_zone_t *zone, size_t alignment,\n#endif\n#if (JEMALLOC_ZONE_VERSION >= 6)\n size_t size);\nstatic void\tzone_free_definite_size(malloc_zone_t *zone, void *ptr,\n size_t size);\n#endif\nstatic void\t*zone_destroy(malloc_zone_t *zone);\nstatic size_t\tzone_good_size(malloc_zone_t *zone, size_t size);\nstatic void\tzone_force_lock(malloc_zone_t *zone);\nstatic void\tzone_force_unlock(malloc_zone_t *zone);\n\n/******************************************************************************/\n/*\n * Functions.\n */\n\nstatic size_t\nzone_size(malloc_zone_t *zone, void *ptr)\n{\n\n\t/*\n\t * There appear to be places within Darwin (such as setenv(3)) that\n\t * cause calls to this function with pointers that *no* zone owns. If\n\t * we knew that all pointers were owned by *some* zone, we could split\n\t * our zone into two parts, and use one as the default allocator and\n\t * the other as the default deallocator/reallocator. Since that will\n\t * not work in practice, we must check all pointers to assure that they\n\t * reside within a mapped chunk before determining size.\n\t */\n\treturn (ivsalloc(ptr, config_prof));\n}\n\nstatic void *\nzone_malloc(malloc_zone_t *zone, size_t size)\n{\n\n\treturn (je_malloc(size));\n}\n\nstatic void *\nzone_calloc(malloc_zone_t *zone, size_t num, size_t size)\n{\n\n\treturn (je_calloc(num, size));\n}\n\nstatic void *\nzone_valloc(malloc_zone_t *zone, size_t size)\n{\n\tvoid *ret = NULL; /* Assignment avoids useless compiler warning. */\n\n\tje_posix_memalign(&ret, PAGE, size);\n\n\treturn (ret);\n}\n\nstatic void\nzone_free(malloc_zone_t *zone, void *ptr)\n{\n\n\tif (ivsalloc(ptr, config_prof) != 0) {\n\t\tje_free(ptr);\n\t\treturn;\n\t}\n\n\tfree(ptr);\n}\n\nstatic void *\nzone_realloc(malloc_zone_t *zone, void *ptr, size_t size)\n{\n\n\tif (ivsalloc(ptr, config_prof) != 0)\n\t\treturn (je_realloc(ptr, size));\n\n\treturn (realloc(ptr, size));\n}\n\n#if (JEMALLOC_ZONE_VERSION >= 5)\nstatic void *\nzone_memalign(malloc_zone_t *zone, size_t alignment, size_t size)\n{\n\tvoid *ret = NULL; /* Assignment avoids useless compiler warning. */\n\n\tje_posix_memalign(&ret, alignment, size);\n\n\treturn (ret);\n}\n#endif\n\n#if (JEMALLOC_ZONE_VERSION >= 6)\nstatic void\nzone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size)\n{\n\n\tif (ivsalloc(ptr, config_prof) != 0) {\n\t\tassert(ivsalloc(ptr, config_prof) == size);\n\t\tje_free(ptr);\n\t\treturn;\n\t}\n\n\tfree(ptr);\n}\n#endif\n\nstatic void *\nzone_destroy(malloc_zone_t *zone)\n{\n\n\t/* This function should never be called. */\n\tassert(false);\n\treturn (NULL);\n}\n\nstatic size_t\nzone_good_size(malloc_zone_t *zone, size_t size)\n{\n\n\tif (size == 0)\n\t\tsize = 1;\n\treturn (s2u(size));\n}\n\nstatic void\nzone_force_lock(malloc_zone_t *zone)\n{\n\n\tif (isthreaded)\n\t\tjemalloc_prefork();\n}\n\nstatic void\nzone_force_unlock(malloc_zone_t *zone)\n{\n\n\tif (isthreaded)\n\t\tjemalloc_postfork_parent();\n}\n\nJEMALLOC_ATTR(constructor)\nvoid\nregister_zone(void)\n{\n\n\t/*\n\t * If something else replaced the system default zone allocator, don't\n\t * register jemalloc's.\n\t */\n\tmalloc_zone_t *default_zone = malloc_default_zone();\n\tif (!default_zone->zone_name ||\n\t strcmp(default_zone->zone_name, \"DefaultMallocZone\") != 0) {\n\t\treturn;\n\t}\n\n\tzone.size = (void *)zone_size;\n\tzone.malloc = (void *)zone_malloc;\n\tzone.calloc = (void *)zone_calloc;\n\tzone.valloc = (void *)zone_valloc;\n\tzone.free = (void *)zone_free;\n\tzone.realloc = (void *)zone_realloc;\n\tzone.destroy = (void *)zone_destroy;\n\tzone.zone_name = \"jemalloc_zone\";\n\tzone.batch_malloc = NULL;\n\tzone.batch_free = NULL;\n\tzone.introspect = &zone_introspect;\n\tzone.version = JEMALLOC_ZONE_VERSION;\n#if (JEMALLOC_ZONE_VERSION >= 5)\n\tzone.memalign = zone_memalign;\n#endif\n#if (JEMALLOC_ZONE_VERSION >= 6)\n\tzone.free_definite_size = zone_free_definite_size;\n#endif\n#if (JEMALLOC_ZONE_VERSION >= 8)\n\tzone.pressure_relief = NULL;\n#endif\n\n\tzone_introspect.enumerator = NULL;\n\tzone_introspect.good_size = (void *)zone_good_size;\n\tzone_introspect.check = NULL;\n\tzone_introspect.print = NULL;\n\tzone_introspect.log = NULL;\n\tzone_introspect.force_lock = (void *)zone_force_lock;\n\tzone_introspect.force_unlock = (void *)zone_force_unlock;\n\tzone_introspect.statistics = NULL;\n#if (JEMALLOC_ZONE_VERSION >= 6)\n\tzone_introspect.zone_locked = NULL;\n#endif\n#if (JEMALLOC_ZONE_VERSION >= 7)\n\tzone_introspect.enable_discharge_checking = NULL;\n\tzone_introspect.disable_discharge_checking = NULL;\n\tzone_introspect.discharge = NULL;\n#ifdef __BLOCKS__\n\tzone_introspect.enumerate_discharged_pointers = NULL;\n#else\n\tzone_introspect.enumerate_unavailable_without_blocks = NULL;\n#endif\n#endif\n\n\t/*\n\t * The default purgeable zone is created lazily by OSX's libc. It uses\n\t * the default zone when it is created for \"small\" allocations\n\t * (< 15 KiB), but assumes the default zone is a scalable_zone. This\n\t * obviously fails when the default zone is the jemalloc zone, so\n\t * malloc_default_purgeable_zone is called beforehand so that the\n\t * default purgeable zone is created when the default zone is still\n\t * a scalable_zone. As purgeable zones only exist on >= 10.6, we need\n\t * to check for the existence of malloc_default_purgeable_zone() at\n\t * run time.\n\t */\n\tif (malloc_default_purgeable_zone != NULL)\n\t\tmalloc_default_purgeable_zone();\n\n\t/* Register the custom zone. At this point it won't be the default. */\n\tmalloc_zone_register(&zone);\n\n\t/*\n\t * Unregister and reregister the default zone. On OSX >= 10.6,\n\t * unregistering takes the last registered zone and places it at the\n\t * location of the specified zone. Unregistering the default zone thus\n\t * makes the last registered one the default. On OSX < 10.6,\n\t * unregistering shifts all registered zones. The first registered zone\n\t * then becomes the default.\n\t */\n\tdo {\n\t\tdefault_zone = malloc_default_zone();\n\t\tmalloc_zone_unregister(default_zone);\n\t\tmalloc_zone_register(default_zone);\n\t} while (malloc_default_zone() != &zone);\n}\n" }, { "path": "deps/jemalloc/test/ALLOCM_ARENA.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#define\tNTHREADS 10\n\nvoid *\nje_thread_start(void *arg)\n{\n\tunsigned thread_ind = (unsigned)(uintptr_t)arg;\n\tunsigned arena_ind;\n\tint r;\n\tvoid *p;\n\tsize_t rsz, sz;\n\n\tsz = sizeof(arena_ind);\n\tif (mallctl(\"arenas.extend\", &arena_ind, &sz, NULL, 0)\n\t != 0) {\n\t\tmalloc_printf(\"Error in arenas.extend\\n\");\n\t\tabort();\n\t}\n\n\tif (thread_ind % 4 != 3) {\n\t\tsize_t mib[3];\n\t\tsize_t miblen = sizeof(mib) / sizeof(size_t);\n\t\tconst char *dss_precs[] = {\"disabled\", \"primary\", \"secondary\"};\n\t\tconst char *dss = dss_precs[thread_ind % 4];\n\t\tif (mallctlnametomib(\"arena.0.dss\", mib, &miblen) != 0) {\n\t\t\tmalloc_printf(\"Error in mallctlnametomib()\\n\");\n\t\t\tabort();\n\t\t}\n\t\tmib[1] = arena_ind;\n\t\tif (mallctlbymib(mib, miblen, NULL, NULL, (void *)&dss,\n\t\t sizeof(const char *))) {\n\t\t\tmalloc_printf(\"Error in mallctlbymib()\\n\");\n\t\t\tabort();\n\t\t}\n\t}\n\n\tr = allocm(&p, &rsz, 1, ALLOCM_ARENA(arena_ind));\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected allocm() error\\n\");\n\t\tabort();\n\t}\n\n\treturn (NULL);\n}\n\nint\nmain(void)\n{\n\tje_thread_t threads[NTHREADS];\n\tunsigned i;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tfor (i = 0; i < NTHREADS; i++) {\n\t\tje_thread_create(&threads[i], je_thread_start,\n\t\t (void *)(uintptr_t)i);\n\t}\n\n\tfor (i = 0; i < NTHREADS; i++)\n\t\tje_thread_join(threads[i], NULL);\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/ALLOCM_ARENA.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/aligned_alloc.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#define CHUNK 0x400000\n/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */\n#define MAXALIGN ((size_t)0x2000000LU)\n#define NITER 4\n\nint\nmain(void)\n{\n\tsize_t alignment, size, total;\n\tunsigned i;\n\tvoid *p, *ps[NITER];\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\t/* Test error conditions. */\n\talignment = 0;\n\tset_errno(0);\n\tp = aligned_alloc(alignment, 1);\n\tif (p != NULL || get_errno() != EINVAL) {\n\t\tmalloc_printf(\n\t\t \"Expected error for invalid alignment %zu\\n\", alignment);\n\t}\n\n\tfor (alignment = sizeof(size_t); alignment < MAXALIGN;\n\t alignment <<= 1) {\n\t\tset_errno(0);\n\t\tp = aligned_alloc(alignment + 1, 1);\n\t\tif (p != NULL || get_errno() != EINVAL) {\n\t\t\tmalloc_printf(\n\t\t\t \"Expected error for invalid alignment %zu\\n\",\n\t\t\t alignment + 1);\n\t\t}\n\t}\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x8000000000000000);\n\tsize = UINT64_C(0x8000000000000000);\n#else\n\talignment = 0x80000000LU;\n\tsize = 0x80000000LU;\n#endif\n\tset_errno(0);\n\tp = aligned_alloc(alignment, size);\n\tif (p != NULL || get_errno() != ENOMEM) {\n\t\tmalloc_printf(\n\t\t \"Expected error for aligned_alloc(%zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x4000000000000000);\n\tsize = UINT64_C(0x8400000000000001);\n#else\n\talignment = 0x40000000LU;\n\tsize = 0x84000001LU;\n#endif\n\tset_errno(0);\n\tp = aligned_alloc(alignment, size);\n\tif (p != NULL || get_errno() != ENOMEM) {\n\t\tmalloc_printf(\n\t\t \"Expected error for aligned_alloc(%zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n\talignment = 0x10LU;\n#if LG_SIZEOF_PTR == 3\n\tsize = UINT64_C(0xfffffffffffffff0);\n#else\n\tsize = 0xfffffff0LU;\n#endif\n\tset_errno(0);\n\tp = aligned_alloc(alignment, size);\n\tif (p != NULL || get_errno() != ENOMEM) {\n\t\tmalloc_printf(\n\t\t \"Expected error for aligned_alloc(&p, %zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n\tfor (i = 0; i < NITER; i++)\n\t\tps[i] = NULL;\n\n\tfor (alignment = 8;\n\t alignment <= MAXALIGN;\n\t alignment <<= 1) {\n\t\ttotal = 0;\n\t\tmalloc_printf(\"Alignment: %zu\\n\", alignment);\n\t\tfor (size = 1;\n\t\t size < 3 * alignment && size < (1U << 31);\n\t\t size += (alignment >> (LG_SIZEOF_PTR-1)) - 1) {\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\tps[i] = aligned_alloc(alignment, size);\n\t\t\t\tif (ps[i] == NULL) {\n\t\t\t\t\tchar buf[BUFERROR_BUF];\n\n\t\t\t\t\tbuferror(buf, sizeof(buf));\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"Error for size %zu (%#zx): %s\\n\",\n\t\t\t\t\t size, size, buf);\n\t\t\t\t\texit(1);\n\t\t\t\t}\n\t\t\t\ttotal += malloc_usable_size(ps[i]);\n\t\t\t\tif (total >= (MAXALIGN << 1))\n\t\t\t\t\tbreak;\n\t\t\t}\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\tif (ps[i] != NULL) {\n\t\t\t\t\tfree(ps[i]);\n\t\t\t\t\tps[i] = NULL;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/aligned_alloc.exp", "content": "Test begin\nAlignment: 8\nAlignment: 16\nAlignment: 32\nAlignment: 64\nAlignment: 128\nAlignment: 256\nAlignment: 512\nAlignment: 1024\nAlignment: 2048\nAlignment: 4096\nAlignment: 8192\nAlignment: 16384\nAlignment: 32768\nAlignment: 65536\nAlignment: 131072\nAlignment: 262144\nAlignment: 524288\nAlignment: 1048576\nAlignment: 2097152\nAlignment: 4194304\nAlignment: 8388608\nAlignment: 16777216\nAlignment: 33554432\nTest end\n" }, { "path": "deps/jemalloc/test/allocated.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\nvoid *\nje_thread_start(void *arg)\n{\n\tint err;\n\tvoid *p;\n\tuint64_t a0, a1, d0, d1;\n\tuint64_t *ap0, *ap1, *dp0, *dp1;\n\tsize_t sz, usize;\n\n\tsz = sizeof(a0);\n\tif ((err = mallctl(\"thread.allocated\", &a0, &sz, NULL, 0))) {\n\t\tif (err == ENOENT) {\n#ifdef JEMALLOC_STATS\n\t\t\tassert(false);\n#endif\n\t\t\tgoto label_return;\n\t\t}\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\texit(1);\n\t}\n\tsz = sizeof(ap0);\n\tif ((err = mallctl(\"thread.allocatedp\", &ap0, &sz, NULL, 0))) {\n\t\tif (err == ENOENT) {\n#ifdef JEMALLOC_STATS\n\t\t\tassert(false);\n#endif\n\t\t\tgoto label_return;\n\t\t}\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\texit(1);\n\t}\n\tassert(*ap0 == a0);\n\n\tsz = sizeof(d0);\n\tif ((err = mallctl(\"thread.deallocated\", &d0, &sz, NULL, 0))) {\n\t\tif (err == ENOENT) {\n#ifdef JEMALLOC_STATS\n\t\t\tassert(false);\n#endif\n\t\t\tgoto label_return;\n\t\t}\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\texit(1);\n\t}\n\tsz = sizeof(dp0);\n\tif ((err = mallctl(\"thread.deallocatedp\", &dp0, &sz, NULL, 0))) {\n\t\tif (err == ENOENT) {\n#ifdef JEMALLOC_STATS\n\t\t\tassert(false);\n#endif\n\t\t\tgoto label_return;\n\t\t}\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\texit(1);\n\t}\n\tassert(*dp0 == d0);\n\n\tp = malloc(1);\n\tif (p == NULL) {\n\t\tmalloc_printf(\"%s(): Error in malloc()\\n\", __func__);\n\t\texit(1);\n\t}\n\n\tsz = sizeof(a1);\n\tmallctl(\"thread.allocated\", &a1, &sz, NULL, 0);\n\tsz = sizeof(ap1);\n\tmallctl(\"thread.allocatedp\", &ap1, &sz, NULL, 0);\n\tassert(*ap1 == a1);\n\tassert(ap0 == ap1);\n\n\tusize = malloc_usable_size(p);\n\tassert(a0 + usize <= a1);\n\n\tfree(p);\n\n\tsz = sizeof(d1);\n\tmallctl(\"thread.deallocated\", &d1, &sz, NULL, 0);\n\tsz = sizeof(dp1);\n\tmallctl(\"thread.deallocatedp\", &dp1, &sz, NULL, 0);\n\tassert(*dp1 == d1);\n\tassert(dp0 == dp1);\n\n\tassert(d0 + usize <= d1);\n\nlabel_return:\n\treturn (NULL);\n}\n\nint\nmain(void)\n{\n\tint ret = 0;\n\tje_thread_t thread;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tje_thread_start(NULL);\n\n\tje_thread_create(&thread, je_thread_start, NULL);\n\tje_thread_join(thread, (void *)&ret);\n\n\tje_thread_start(NULL);\n\n\tje_thread_create(&thread, je_thread_start, NULL);\n\tje_thread_join(thread, (void *)&ret);\n\n\tje_thread_start(NULL);\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (ret);\n}\n" }, { "path": "deps/jemalloc/test/allocated.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/allocm.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#define CHUNK 0x400000\n/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */\n#define MAXALIGN ((size_t)0x2000000LU)\n#define NITER 4\n\nint\nmain(void)\n{\n\tint r;\n\tvoid *p;\n\tsize_t nsz, rsz, sz, alignment, total;\n\tunsigned i;\n\tvoid *ps[NITER];\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tsz = 42;\n\tnsz = 0;\n\tr = nallocm(&nsz, sz, 0);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected nallocm() error\\n\");\n\t\tabort();\n\t}\n\trsz = 0;\n\tr = allocm(&p, &rsz, sz, 0);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected allocm() error\\n\");\n\t\tabort();\n\t}\n\tif (rsz < sz)\n\t\tmalloc_printf(\"Real size smaller than expected\\n\");\n\tif (nsz != rsz)\n\t\tmalloc_printf(\"nallocm()/allocm() rsize mismatch\\n\");\n\tif (dallocm(p, 0) != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected dallocm() error\\n\");\n\n\tr = allocm(&p, NULL, sz, 0);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected allocm() error\\n\");\n\t\tabort();\n\t}\n\tif (dallocm(p, 0) != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected dallocm() error\\n\");\n\n\tnsz = 0;\n\tr = nallocm(&nsz, sz, ALLOCM_ZERO);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected nallocm() error\\n\");\n\t\tabort();\n\t}\n\trsz = 0;\n\tr = allocm(&p, &rsz, sz, ALLOCM_ZERO);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected allocm() error\\n\");\n\t\tabort();\n\t}\n\tif (nsz != rsz)\n\t\tmalloc_printf(\"nallocm()/allocm() rsize mismatch\\n\");\n\tif (dallocm(p, 0) != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected dallocm() error\\n\");\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x8000000000000000);\n\tsz = UINT64_C(0x8000000000000000);\n#else\n\talignment = 0x80000000LU;\n\tsz = 0x80000000LU;\n#endif\n\tnsz = 0;\n\tr = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r == ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\n\t\t \"Expected error for nallocm(&nsz, %zu, %#x)\\n\",\n\t\t sz, ALLOCM_ALIGN(alignment));\n\t}\n\trsz = 0;\n\tr = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r == ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\n\t\t \"Expected error for allocm(&p, %zu, %#x)\\n\",\n\t\t sz, ALLOCM_ALIGN(alignment));\n\t}\n\tif (nsz != rsz)\n\t\tmalloc_printf(\"nallocm()/allocm() rsize mismatch\\n\");\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x4000000000000000);\n\tsz = UINT64_C(0x8400000000000001);\n#else\n\talignment = 0x40000000LU;\n\tsz = 0x84000001LU;\n#endif\n\tnsz = 0;\n\tr = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected nallocm() error\\n\");\n\trsz = 0;\n\tr = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r == ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\n\t\t \"Expected error for allocm(&p, %zu, %#x)\\n\",\n\t\t sz, ALLOCM_ALIGN(alignment));\n\t}\n\n\talignment = 0x10LU;\n#if LG_SIZEOF_PTR == 3\n\tsz = UINT64_C(0xfffffffffffffff0);\n#else\n\tsz = 0xfffffff0LU;\n#endif\n\tnsz = 0;\n\tr = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r == ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\n\t\t \"Expected error for nallocm(&nsz, %zu, %#x)\\n\",\n\t\t sz, ALLOCM_ALIGN(alignment));\n\t}\n\trsz = 0;\n\tr = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment));\n\tif (r == ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\n\t\t \"Expected error for allocm(&p, %zu, %#x)\\n\",\n\t\t sz, ALLOCM_ALIGN(alignment));\n\t}\n\tif (nsz != rsz)\n\t\tmalloc_printf(\"nallocm()/allocm() rsize mismatch\\n\");\n\n\tfor (i = 0; i < NITER; i++)\n\t\tps[i] = NULL;\n\n\tfor (alignment = 8;\n\t alignment <= MAXALIGN;\n\t alignment <<= 1) {\n\t\ttotal = 0;\n\t\tmalloc_printf(\"Alignment: %zu\\n\", alignment);\n\t\tfor (sz = 1;\n\t\t sz < 3 * alignment && sz < (1U << 31);\n\t\t sz += (alignment >> (LG_SIZEOF_PTR-1)) - 1) {\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\tnsz = 0;\n\t\t\t\tr = nallocm(&nsz, sz,\n\t\t\t\t ALLOCM_ALIGN(alignment) | ALLOCM_ZERO);\n\t\t\t\tif (r != ALLOCM_SUCCESS) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"nallocm() error for size %zu\"\n\t\t\t\t\t \" (%#zx): %d\\n\",\n\t\t\t\t\t sz, sz, r);\n\t\t\t\t\texit(1);\n\t\t\t\t}\n\t\t\t\trsz = 0;\n\t\t\t\tr = allocm(&ps[i], &rsz, sz,\n\t\t\t\t ALLOCM_ALIGN(alignment) | ALLOCM_ZERO);\n\t\t\t\tif (r != ALLOCM_SUCCESS) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"allocm() error for size %zu\"\n\t\t\t\t\t \" (%#zx): %d\\n\",\n\t\t\t\t\t sz, sz, r);\n\t\t\t\t\texit(1);\n\t\t\t\t}\n\t\t\t\tif (rsz < sz) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"Real size smaller than\"\n\t\t\t\t\t \" expected\\n\");\n\t\t\t\t}\n\t\t\t\tif (nsz != rsz) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"nallocm()/allocm() rsize\"\n\t\t\t\t\t \" mismatch\\n\");\n\t\t\t\t}\n\t\t\t\tif ((uintptr_t)p & (alignment-1)) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"%p inadequately aligned for\"\n\t\t\t\t\t \" alignment: %zu\\n\", p, alignment);\n\t\t\t\t}\n\t\t\t\tsallocm(ps[i], &rsz, 0);\n\t\t\t\ttotal += rsz;\n\t\t\t\tif (total >= (MAXALIGN << 1))\n\t\t\t\t\tbreak;\n\t\t\t}\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\tif (ps[i] != NULL) {\n\t\t\t\t\tdallocm(ps[i], 0);\n\t\t\t\t\tps[i] = NULL;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/allocm.exp", "content": "Test begin\nAlignment: 8\nAlignment: 16\nAlignment: 32\nAlignment: 64\nAlignment: 128\nAlignment: 256\nAlignment: 512\nAlignment: 1024\nAlignment: 2048\nAlignment: 4096\nAlignment: 8192\nAlignment: 16384\nAlignment: 32768\nAlignment: 65536\nAlignment: 131072\nAlignment: 262144\nAlignment: 524288\nAlignment: 1048576\nAlignment: 2097152\nAlignment: 4194304\nAlignment: 8388608\nAlignment: 16777216\nAlignment: 33554432\nTest end\n" }, { "path": "deps/jemalloc/test/bitmap.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#if (LG_BITMAP_MAXBITS > 12)\n# define MAXBITS\t4500\n#else\n# define MAXBITS\t(1U << LG_BITMAP_MAXBITS)\n#endif\n\nstatic void\ntest_bitmap_size(void)\n{\n\tsize_t i, prev_size;\n\n\tprev_size = 0;\n\tfor (i = 1; i <= MAXBITS; i++) {\n\t\tsize_t size = bitmap_size(i);\n\t\tassert(size >= prev_size);\n\t\tprev_size = size;\n\t}\n}\n\nstatic void\ntest_bitmap_init(void)\n{\n\tsize_t i;\n\n\tfor (i = 1; i <= MAXBITS; i++) {\n\t\tbitmap_info_t binfo;\n\t\tbitmap_info_init(&binfo, i);\n\t\t{\n\t\t\tsize_t j;\n\t\t\tbitmap_t *bitmap = malloc(sizeof(bitmap_t) *\n\t\t\t\tbitmap_info_ngroups(&binfo));\n\t\t\tbitmap_init(bitmap, &binfo);\n\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tassert(bitmap_get(bitmap, &binfo, j) == false);\n\t\t\tfree(bitmap);\n\n\t\t}\n\t}\n}\n\nstatic void\ntest_bitmap_set(void)\n{\n\tsize_t i;\n\n\tfor (i = 1; i <= MAXBITS; i++) {\n\t\tbitmap_info_t binfo;\n\t\tbitmap_info_init(&binfo, i);\n\t\t{\n\t\t\tsize_t j;\n\t\t\tbitmap_t *bitmap = malloc(sizeof(bitmap_t) *\n\t\t\t\tbitmap_info_ngroups(&binfo));\n\t\t\tbitmap_init(bitmap, &binfo);\n\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tbitmap_set(bitmap, &binfo, j);\n\t\t\tassert(bitmap_full(bitmap, &binfo));\n\t\t\tfree(bitmap);\n\t\t}\n\t}\n}\n\nstatic void\ntest_bitmap_unset(void)\n{\n\tsize_t i;\n\n\tfor (i = 1; i <= MAXBITS; i++) {\n\t\tbitmap_info_t binfo;\n\t\tbitmap_info_init(&binfo, i);\n\t\t{\n\t\t\tsize_t j;\n\t\t\tbitmap_t *bitmap = malloc(sizeof(bitmap_t) *\n\t\t\t\tbitmap_info_ngroups(&binfo));\n\t\t\tbitmap_init(bitmap, &binfo);\n\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tbitmap_set(bitmap, &binfo, j);\n\t\t\tassert(bitmap_full(bitmap, &binfo));\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tbitmap_unset(bitmap, &binfo, j);\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tbitmap_set(bitmap, &binfo, j);\n\t\t\tassert(bitmap_full(bitmap, &binfo));\n\t\t\tfree(bitmap);\n\t\t}\n\t}\n}\n\nstatic void\ntest_bitmap_sfu(void)\n{\n\tsize_t i;\n\n\tfor (i = 1; i <= MAXBITS; i++) {\n\t\tbitmap_info_t binfo;\n\t\tbitmap_info_init(&binfo, i);\n\t\t{\n\t\t\tssize_t j;\n\t\t\tbitmap_t *bitmap = malloc(sizeof(bitmap_t) *\n\t\t\t\tbitmap_info_ngroups(&binfo));\n\t\t\tbitmap_init(bitmap, &binfo);\n\n\t\t\t/* Iteratively set bits starting at the beginning. */\n\t\t\tfor (j = 0; j < i; j++)\n\t\t\t\tassert(bitmap_sfu(bitmap, &binfo) == j);\n\t\t\tassert(bitmap_full(bitmap, &binfo));\n\n\t\t\t/*\n\t\t\t * Iteratively unset bits starting at the end, and\n\t\t\t * verify that bitmap_sfu() reaches the unset bits.\n\t\t\t */\n\t\t\tfor (j = i - 1; j >= 0; j--) {\n\t\t\t\tbitmap_unset(bitmap, &binfo, j);\n\t\t\t\tassert(bitmap_sfu(bitmap, &binfo) == j);\n\t\t\t\tbitmap_unset(bitmap, &binfo, j);\n\t\t\t}\n\t\t\tassert(bitmap_get(bitmap, &binfo, 0) == false);\n\n\t\t\t/*\n\t\t\t * Iteratively set bits starting at the beginning, and\n\t\t\t * verify that bitmap_sfu() looks past them.\n\t\t\t */\n\t\t\tfor (j = 1; j < i; j++) {\n\t\t\t\tbitmap_set(bitmap, &binfo, j - 1);\n\t\t\t\tassert(bitmap_sfu(bitmap, &binfo) == j);\n\t\t\t\tbitmap_unset(bitmap, &binfo, j);\n\t\t\t}\n\t\t\tassert(bitmap_sfu(bitmap, &binfo) == i - 1);\n\t\t\tassert(bitmap_full(bitmap, &binfo));\n\t\t\tfree(bitmap);\n\t\t}\n\t}\n}\n\nint\nmain(void)\n{\n\tmalloc_printf(\"Test begin\\n\");\n\n\ttest_bitmap_size();\n\ttest_bitmap_init();\n\ttest_bitmap_set();\n\ttest_bitmap_unset();\n\ttest_bitmap_sfu();\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/bitmap.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/jemalloc_test.h.in", "content": "/*\n * This header should be included by tests, rather than directly including\n * jemalloc/jemalloc.h, because --with-install-suffix may cause the header to\n * have a different name.\n */\n#include \"jemalloc/jemalloc@install_suffix@.h\"\n#include \"jemalloc/internal/jemalloc_internal.h\"\n\n/* Abstraction layer for threading in tests */\n#ifdef _WIN32\n#include \n\ntypedef HANDLE je_thread_t;\n\nvoid\nje_thread_create(je_thread_t *thread, void *(*proc)(void *), void *arg)\n{\n\tLPTHREAD_START_ROUTINE routine = (LPTHREAD_START_ROUTINE)proc;\n\t*thread = CreateThread(NULL, 0, routine, arg, 0, NULL);\n\tif (*thread == NULL) {\n\t\tmalloc_printf(\"Error in CreateThread()\\n\");\n\t\texit(1);\n\t}\n}\n\nvoid\nje_thread_join(je_thread_t thread, void **ret)\n{\n\tWaitForSingleObject(thread, INFINITE);\n}\n\n#else\n#include \n\ntypedef pthread_t je_thread_t;\n\nvoid\nje_thread_create(je_thread_t *thread, void *(*proc)(void *), void *arg)\n{\n\n\tif (pthread_create(thread, NULL, proc, arg) != 0) {\n\t\tmalloc_printf(\"Error in pthread_create()\\n\");\n\t\texit(1);\n\t}\n}\n\nvoid\nje_thread_join(je_thread_t thread, void **ret)\n{\n\n\tpthread_join(thread, ret);\n}\n#endif\n" }, { "path": "deps/jemalloc/test/mremap.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\nint\nmain(void)\n{\n\tint ret, err;\n\tsize_t sz, lg_chunk, chunksize, i;\n\tchar *p, *q;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tsz = sizeof(lg_chunk);\n\tif ((err = mallctl(\"opt.lg_chunk\", &lg_chunk, &sz, NULL, 0))) {\n\t\tassert(err != ENOENT);\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\tchunksize = ((size_t)1U) << lg_chunk;\n\n\tp = (char *)malloc(chunksize);\n\tif (p == NULL) {\n\t\tmalloc_printf(\"malloc(%zu) --> %p\\n\", chunksize, p);\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\tmemset(p, 'a', chunksize);\n\n\tq = (char *)realloc(p, chunksize * 2);\n\tif (q == NULL) {\n\t\tmalloc_printf(\"realloc(%p, %zu) --> %p\\n\", p, chunksize * 2,\n\t\t q);\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\tfor (i = 0; i < chunksize; i++) {\n\t\tassert(q[i] == 'a');\n\t}\n\n\tp = q;\n\n\tq = (char *)realloc(p, chunksize);\n\tif (q == NULL) {\n\t\tmalloc_printf(\"realloc(%p, %zu) --> %p\\n\", p, chunksize, q);\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\tfor (i = 0; i < chunksize; i++) {\n\t\tassert(q[i] == 'a');\n\t}\n\n\tfree(q);\n\n\tret = 0;\nlabel_return:\n\tmalloc_printf(\"Test end\\n\");\n\treturn (ret);\n}\n" }, { "path": "deps/jemalloc/test/mremap.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/posix_memalign.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#define CHUNK 0x400000\n/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */\n#define MAXALIGN ((size_t)0x2000000LU)\n#define NITER 4\n\nint\nmain(void)\n{\n\tsize_t alignment, size, total;\n\tunsigned i;\n\tint err;\n\tvoid *p, *ps[NITER];\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\t/* Test error conditions. */\n\tfor (alignment = 0; alignment < sizeof(void *); alignment++) {\n\t\terr = posix_memalign(&p, alignment, 1);\n\t\tif (err != EINVAL) {\n\t\t\tmalloc_printf(\n\t\t\t \"Expected error for invalid alignment %zu\\n\",\n\t\t\t alignment);\n\t\t}\n\t}\n\n\tfor (alignment = sizeof(size_t); alignment < MAXALIGN;\n\t alignment <<= 1) {\n\t\terr = posix_memalign(&p, alignment + 1, 1);\n\t\tif (err == 0) {\n\t\t\tmalloc_printf(\n\t\t\t \"Expected error for invalid alignment %zu\\n\",\n\t\t\t alignment + 1);\n\t\t}\n\t}\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x8000000000000000);\n\tsize = UINT64_C(0x8000000000000000);\n#else\n\talignment = 0x80000000LU;\n\tsize = 0x80000000LU;\n#endif\n\terr = posix_memalign(&p, alignment, size);\n\tif (err == 0) {\n\t\tmalloc_printf(\n\t\t \"Expected error for posix_memalign(&p, %zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n#if LG_SIZEOF_PTR == 3\n\talignment = UINT64_C(0x4000000000000000);\n\tsize = UINT64_C(0x8400000000000001);\n#else\n\talignment = 0x40000000LU;\n\tsize = 0x84000001LU;\n#endif\n\terr = posix_memalign(&p, alignment, size);\n\tif (err == 0) {\n\t\tmalloc_printf(\n\t\t \"Expected error for posix_memalign(&p, %zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n\talignment = 0x10LU;\n#if LG_SIZEOF_PTR == 3\n\tsize = UINT64_C(0xfffffffffffffff0);\n#else\n\tsize = 0xfffffff0LU;\n#endif\n\terr = posix_memalign(&p, alignment, size);\n\tif (err == 0) {\n\t\tmalloc_printf(\n\t\t \"Expected error for posix_memalign(&p, %zu, %zu)\\n\",\n\t\t alignment, size);\n\t}\n\n\tfor (i = 0; i < NITER; i++)\n\t\tps[i] = NULL;\n\n\tfor (alignment = 8;\n\t alignment <= MAXALIGN;\n\t alignment <<= 1) {\n\t\ttotal = 0;\n\t\tmalloc_printf(\"Alignment: %zu\\n\", alignment);\n\t\tfor (size = 1;\n\t\t size < 3 * alignment && size < (1U << 31);\n\t\t size += (alignment >> (LG_SIZEOF_PTR-1)) - 1) {\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\terr = posix_memalign(&ps[i],\n\t\t\t\t alignment, size);\n\t\t\t\tif (err) {\n\t\t\t\t\tmalloc_printf(\n\t\t\t\t\t \"Error for size %zu (%#zx): %s\\n\",\n\t\t\t\t\t size, size, strerror(err));\n\t\t\t\t\texit(1);\n\t\t\t\t}\n\t\t\t\ttotal += malloc_usable_size(ps[i]);\n\t\t\t\tif (total >= (MAXALIGN << 1))\n\t\t\t\t\tbreak;\n\t\t\t}\n\t\t\tfor (i = 0; i < NITER; i++) {\n\t\t\t\tif (ps[i] != NULL) {\n\t\t\t\t\tfree(ps[i]);\n\t\t\t\t\tps[i] = NULL;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/posix_memalign.exp", "content": "Test begin\nAlignment: 8\nAlignment: 16\nAlignment: 32\nAlignment: 64\nAlignment: 128\nAlignment: 256\nAlignment: 512\nAlignment: 1024\nAlignment: 2048\nAlignment: 4096\nAlignment: 8192\nAlignment: 16384\nAlignment: 32768\nAlignment: 65536\nAlignment: 131072\nAlignment: 262144\nAlignment: 524288\nAlignment: 1048576\nAlignment: 2097152\nAlignment: 4194304\nAlignment: 8388608\nAlignment: 16777216\nAlignment: 33554432\nTest end\n" }, { "path": "deps/jemalloc/test/rallocm.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\nint\nmain(void)\n{\n\tsize_t pagesize;\n\tvoid *p, *q;\n\tsize_t sz, tsz;\n\tint r;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\t/* Get page size. */\n\t{\n#ifdef _WIN32\n\t\tSYSTEM_INFO si;\n\t\tGetSystemInfo(&si);\n\t\tpagesize = (size_t)si.dwPageSize;\n#else\n\t\tlong result = sysconf(_SC_PAGESIZE);\n\t\tassert(result != -1);\n\t\tpagesize = (size_t)result;\n#endif\n\t}\n\n\tr = allocm(&p, &sz, 42, 0);\n\tif (r != ALLOCM_SUCCESS) {\n\t\tmalloc_printf(\"Unexpected allocm() error\\n\");\n\t\tabort();\n\t}\n\n\tq = p;\n\tr = rallocm(&q, &tsz, sz, 0, ALLOCM_NO_MOVE);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q != p)\n\t\tmalloc_printf(\"Unexpected object move\\n\");\n\tif (tsz != sz) {\n\t\tmalloc_printf(\"Unexpected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\n\tq = p;\n\tr = rallocm(&q, &tsz, sz, 5, ALLOCM_NO_MOVE);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q != p)\n\t\tmalloc_printf(\"Unexpected object move\\n\");\n\tif (tsz != sz) {\n\t\tmalloc_printf(\"Unexpected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\n\tq = p;\n\tr = rallocm(&q, &tsz, sz + 5, 0, ALLOCM_NO_MOVE);\n\tif (r != ALLOCM_ERR_NOT_MOVED)\n\t\tmalloc_printf(\"Unexpected rallocm() result\\n\");\n\tif (q != p)\n\t\tmalloc_printf(\"Unexpected object move\\n\");\n\tif (tsz != sz) {\n\t\tmalloc_printf(\"Unexpected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\n\tq = p;\n\tr = rallocm(&q, &tsz, sz + 5, 0, 0);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q == p)\n\t\tmalloc_printf(\"Expected object move\\n\");\n\tif (tsz == sz) {\n\t\tmalloc_printf(\"Expected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\tp = q;\n\tsz = tsz;\n\n\tr = rallocm(&q, &tsz, pagesize*2, 0, 0);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q == p)\n\t\tmalloc_printf(\"Expected object move\\n\");\n\tif (tsz == sz) {\n\t\tmalloc_printf(\"Expected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\tp = q;\n\tsz = tsz;\n\n\tr = rallocm(&q, &tsz, pagesize*4, 0, 0);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (tsz == sz) {\n\t\tmalloc_printf(\"Expected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\tp = q;\n\tsz = tsz;\n\n\tr = rallocm(&q, &tsz, pagesize*2, 0, ALLOCM_NO_MOVE);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q != p)\n\t\tmalloc_printf(\"Unexpected object move\\n\");\n\tif (tsz == sz) {\n\t\tmalloc_printf(\"Expected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\tsz = tsz;\n\n\tr = rallocm(&q, &tsz, pagesize*4, 0, ALLOCM_NO_MOVE);\n\tif (r != ALLOCM_SUCCESS)\n\t\tmalloc_printf(\"Unexpected rallocm() error\\n\");\n\tif (q != p)\n\t\tmalloc_printf(\"Unexpected object move\\n\");\n\tif (tsz == sz) {\n\t\tmalloc_printf(\"Expected size change: %zu --> %zu\\n\",\n\t\t sz, tsz);\n\t}\n\tsz = tsz;\n\n\tdallocm(p, 0);\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (0);\n}\n" }, { "path": "deps/jemalloc/test/rallocm.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/thread_arena.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\n#define\tNTHREADS 10\n\nvoid *\nje_thread_start(void *arg)\n{\n\tunsigned main_arena_ind = *(unsigned *)arg;\n\tvoid *p;\n\tunsigned arena_ind;\n\tsize_t size;\n\tint err;\n\n\tp = malloc(1);\n\tif (p == NULL) {\n\t\tmalloc_printf(\"%s(): Error in malloc()\\n\", __func__);\n\t\treturn (void *)1;\n\t}\n\n\tsize = sizeof(arena_ind);\n\tif ((err = mallctl(\"thread.arena\", &arena_ind, &size, &main_arena_ind,\n\t sizeof(main_arena_ind)))) {\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\treturn (void *)1;\n\t}\n\n\tsize = sizeof(arena_ind);\n\tif ((err = mallctl(\"thread.arena\", &arena_ind, &size, NULL,\n\t 0))) {\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\treturn (void *)1;\n\t}\n\tassert(arena_ind == main_arena_ind);\n\n\treturn (NULL);\n}\n\nint\nmain(void)\n{\n\tint ret = 0;\n\tvoid *p;\n\tunsigned arena_ind;\n\tsize_t size;\n\tint err;\n\tje_thread_t threads[NTHREADS];\n\tunsigned i;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tp = malloc(1);\n\tif (p == NULL) {\n\t\tmalloc_printf(\"%s(): Error in malloc()\\n\", __func__);\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\n\tsize = sizeof(arena_ind);\n\tif ((err = mallctl(\"thread.arena\", &arena_ind, &size, NULL, 0))) {\n\t\tmalloc_printf(\"%s(): Error in mallctl(): %s\\n\", __func__,\n\t\t strerror(err));\n\t\tret = 1;\n\t\tgoto label_return;\n\t}\n\n\tfor (i = 0; i < NTHREADS; i++) {\n\t\tje_thread_create(&threads[i], je_thread_start,\n\t\t (void *)&arena_ind);\n\t}\n\n\tfor (i = 0; i < NTHREADS; i++)\n\t\tje_thread_join(threads[i], (void *)&ret);\n\nlabel_return:\n\tmalloc_printf(\"Test end\\n\");\n\treturn (ret);\n}\n" }, { "path": "deps/jemalloc/test/thread_arena.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/jemalloc/test/thread_tcache_enabled.c", "content": "#define\tJEMALLOC_MANGLE\n#include \"jemalloc_test.h\"\n\nvoid *\nje_thread_start(void *arg)\n{\n\tint err;\n\tsize_t sz;\n\tbool e0, e1;\n\n\tsz = sizeof(bool);\n\tif ((err = mallctl(\"thread.tcache.enabled\", &e0, &sz, NULL, 0))) {\n\t\tif (err == ENOENT) {\n#ifdef JEMALLOC_TCACHE\n\t\t\tassert(false);\n#endif\n\t\t}\n\t\tgoto label_return;\n\t}\n\n\tif (e0) {\n\t\te1 = false;\n\t\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz)\n\t\t == 0);\n\t\tassert(e0);\n\t}\n\n\te1 = true;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0 == false);\n\n\te1 = true;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0);\n\n\te1 = false;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0);\n\n\te1 = false;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0 == false);\n\n\tfree(malloc(1));\n\te1 = true;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0 == false);\n\n\tfree(malloc(1));\n\te1 = true;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0);\n\n\tfree(malloc(1));\n\te1 = false;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0);\n\n\tfree(malloc(1));\n\te1 = false;\n\tassert(mallctl(\"thread.tcache.enabled\", &e0, &sz, &e1, sz) == 0);\n\tassert(e0 == false);\n\n\tfree(malloc(1));\nlabel_return:\n\treturn (NULL);\n}\n\nint\nmain(void)\n{\n\tint ret = 0;\n\tje_thread_t thread;\n\n\tmalloc_printf(\"Test begin\\n\");\n\n\tje_thread_start(NULL);\n\n\tje_thread_create(&thread, je_thread_start, NULL);\n\tje_thread_join(thread, (void *)&ret);\n\n\tje_thread_start(NULL);\n\n\tje_thread_create(&thread, je_thread_start, NULL);\n\tje_thread_join(thread, (void *)&ret);\n\n\tje_thread_start(NULL);\n\n\tmalloc_printf(\"Test end\\n\");\n\treturn (ret);\n}\n" }, { "path": "deps/jemalloc/test/thread_tcache_enabled.exp", "content": "Test begin\nTest end\n" }, { "path": "deps/linenoise/.gitignore", "content": "linenoise_example*\n" }, { "path": "deps/linenoise/Makefile", "content": "STD=\nWARN= -Wall\nOPT= -Os\n\nR_CFLAGS= $(STD) $(WARN) $(OPT) $(DEBUG) $(CFLAGS)\nR_LDFLAGS= $(LDFLAGS)\nDEBUG= -g\n\nR_CC=$(CC) $(R_CFLAGS)\nR_LD=$(CC) $(R_LDFLAGS)\n\nlinenoise.o: linenoise.h linenoise.c\n\nlinenoise_example: linenoise.o example.o\n\t$(R_LD) -o $@ $^\n\n.c.o:\n\t$(R_CC) -c $<\n\nclean:\n\trm -f linenoise_example *.o\n" }, { "path": "deps/linenoise/README.markdown", "content": "# Linenoise\n\nA minimal, zero-config, BSD licensed, readline replacement.\n\nNews: linenoise is now part of [Android](http://android.git.kernel.org/?p=platform/system/core.git;a=tree;f=liblinenoise;h=56450eaed7f783760e5e6a5993ef75cde2e29dea;hb=HEAD Android)!\n\n## Can a line editing library be 20k lines of code?\n\nLine editing with some support for history is a really important feature for command line utilities. Instead of retyping almost the same stuff again and again it's just much better to hit the up arrow and edit on syntax errors, or in order to try a slightly different command. But apparently code dealing with terminals is some sort of Black Magic: readline is 30k lines of code, libedit 20k. Is it reasonable to link small utilities to huge libraries just to get a minimal support for line editing?\n\nSo what usually happens is either:\n\n * Large programs with configure scripts disabling line editing if readline is not present in the system, or not supporting it at all since readline is GPL licensed and libedit (the BSD clone) is not as known and available as readline is (Readl world example of this problem: Tclsh).\n * Smaller programs not using a configure script not supporting line editing at all (A problem we had with Redis-cli for instance).\n \nThe result is a pollution of binaries without line editing support.\n\nSo I spent more or less two hours doing a reality check resulting in this little library: is it *really* needed for a line editing library to be 20k lines of code? Apparently not, it is possibe to get a very small, zero configuration, trivial to embed library, that solves the problem. Smaller programs will just include this, supporing line editing out of the box. Larger programs may use this little library or just checking with configure if readline/libedit is available and resorting to linenoise if not.\n\n## Terminals, in 2010.\n\nApparently almost every terminal you can happen to use today has some kind of support for VT100 alike escape sequences. So I tried to write a lib using just very basic VT100 features. The resulting library appears to work everywhere I tried to use it.\n\nSince it's so young I guess there are a few bugs, or the lib may not compile or work with some operating system, but it's a matter of a few weeks and eventually we'll get it right, and there will be no excuses for not shipping command line tools without built-in line editing support.\n\nThe library is currently less than 400 lines of code. In order to use it in your project just look at the *example.c* file in the source distribution, it is trivial. Linenoise is BSD code, so you can use both in free software and commercial software.\n\n## Tested with...\n\n * Linux text only console ($TERM = linux)\n * Linux KDE terminal application ($TERM = xterm)\n * Linux xterm ($TERM = xterm)\n * Mac OS X iTerm ($TERM = xterm)\n * Mac OS X default Terminal.app ($TERM = xterm)\n * OpenBSD 4.5 through an OSX Terminal.app ($TERM = screen)\n * IBM AIX 6.1\n * FreeBSD xterm ($TERM = xterm)\n\nPlease test it everywhere you can and report back!\n\n## Let's push this forward!\n\nPlease fork it and add something interesting and send me a pull request. What's especially interesting are fixes, new key bindings, completion.\n\nSend feedbacks to antirez at gmail\n" }, { "path": "deps/linenoise/example.c", "content": "#include \n#include \n#include \"linenoise.h\"\n\n\nvoid completion(const char *buf, linenoiseCompletions *lc) {\n if (buf[0] == 'h') {\n linenoiseAddCompletion(lc,\"hello\");\n linenoiseAddCompletion(lc,\"hello there\");\n }\n}\n\nint main(void) {\n char *line;\n\n linenoiseSetCompletionCallback(completion);\n linenoiseHistoryLoad(\"history.txt\"); /* Load the history at startup */\n while((line = linenoise(\"hello> \")) != NULL) {\n if (line[0] != '\\0') {\n printf(\"echo: '%s'\\n\", line);\n linenoiseHistoryAdd(line);\n linenoiseHistorySave(\"history.txt\"); /* Save every new entry */\n }\n free(line);\n }\n return 0;\n}\n" }, { "path": "deps/linenoise/linenoise.c", "content": "/* linenoise.c -- guerrilla line editing library against the idea that a\n * line editing lib needs to be 20,000 lines of C code.\n *\n * You can find the latest source code at:\n * \n * http://github.com/antirez/linenoise\n *\n * Does a number of crazy assumptions that happen to be true in 99.9999% of\n * the 2010 UNIX computers around.\n *\n * ------------------------------------------------------------------------\n *\n * Copyright (c) 2010-2013, Salvatore Sanfilippo \n * Copyright (c) 2010-2013, Pieter Noordhuis \n *\n * All rights reserved.\n * \n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are\n * met:\n * \n * * Redistributions of source code must retain the above copyright\n * notice, this list of conditions and the following disclaimer.\n *\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * \n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\n * \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\n * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\n * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\n * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\n * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\n * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\n * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\n * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n * \n * ------------------------------------------------------------------------\n *\n * References:\n * - http://invisible-island.net/xterm/ctlseqs/ctlseqs.html\n * - http://www.3waylabs.com/nw/WWW/products/wizcon/vt220.html\n *\n * Todo list:\n * - Filter bogus Ctrl+ combinations.\n * - Win32 support\n *\n * Bloat:\n * - History search like Ctrl+r in readline?\n *\n * List of escape sequences used by this program, we do everything just\n * with three sequences. In order to be so cheap we may have some\n * flickering effect with some slow terminal, but the lesser sequences\n * the more compatible.\n *\n * CHA (Cursor Horizontal Absolute)\n * Sequence: ESC [ n G\n * Effect: moves cursor to column n\n *\n * EL (Erase Line)\n * Sequence: ESC [ n K\n * Effect: if n is 0 or missing, clear from cursor to end of line\n * Effect: if n is 1, clear from beginning of line to cursor\n * Effect: if n is 2, clear entire line\n *\n * CUF (CUrsor Forward)\n * Sequence: ESC [ n C\n * Effect: moves cursor forward of n chars\n *\n * When multi line mode is enabled, we also use an additional escape\n * sequence. However multi line editing is disabled by default.\n *\n * CUU (Cursor Up)\n * Sequence: ESC [ n A\n * Effect: moves cursor up of n chars.\n *\n * CUD (Cursor Down)\n * Sequence: ESC [ n B\n * Effect: moves cursor down of n chars.\n *\n * The following are used to clear the screen: ESC [ H ESC [ 2 J\n * This is actually composed of two sequences:\n *\n * cursorhome\n * Sequence: ESC [ H\n * Effect: moves the cursor to upper left corner\n *\n * ED2 (Clear entire screen)\n * Sequence: ESC [ 2 J\n * Effect: clear the whole screen\n * \n */\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \"linenoise.h\"\n\n#define LINENOISE_DEFAULT_HISTORY_MAX_LEN 100\n#define LINENOISE_MAX_LINE 4096\nstatic char *unsupported_term[] = {\"dumb\",\"cons25\",\"emacs\",NULL};\nstatic linenoiseCompletionCallback *completionCallback = NULL;\n\nstatic struct termios orig_termios; /* In order to restore at exit.*/\nstatic int rawmode = 0; /* For atexit() function to check if restore is needed*/\nstatic int mlmode = 0; /* Multi line mode. Default is single line. */\nstatic int atexit_registered = 0; /* Register atexit just 1 time. */\nstatic int history_max_len = LINENOISE_DEFAULT_HISTORY_MAX_LEN;\nstatic int history_len = 0;\nstatic char **history = NULL;\n\n/* The linenoiseState structure represents the state during line editing.\n * We pass this state to functions implementing specific editing\n * functionalities. */\nstruct linenoiseState {\n int ifd; /* Terminal stdin file descriptor. */\n int ofd; /* Terminal stdout file descriptor. */\n char *buf; /* Edited line buffer. */\n size_t buflen; /* Edited line buffer size. */\n const char *prompt; /* Prompt to display. */\n size_t plen; /* Prompt length. */\n size_t pos; /* Current cursor position. */\n size_t oldpos; /* Previous refresh cursor position. */\n size_t len; /* Current edited line length. */\n size_t cols; /* Number of columns in terminal. */\n size_t maxrows; /* Maximum num of rows used so far (multiline mode) */\n int history_index; /* The history index we are currently editing. */\n};\n\nenum KEY_ACTION{\n\tKEY_NULL = 0,\t /* NULL */\n\tCTRL_A = 1, /* Ctrl+a */\n\tCTRL_B = 2, /* Ctrl-b */\n\tCTRL_C = 3, /* Ctrl-c */\n\tCTRL_D = 4, /* Ctrl-d */\n\tCTRL_E = 5, /* Ctrl-e */\n\tCTRL_F = 6, /* Ctrl-f */\n\tCTRL_H = 8, /* Ctrl-h */\n\tTAB = 9, /* Tab */\n\tCTRL_K = 11, /* Ctrl+k */\n\tCTRL_L = 12, /* Ctrl+l */\n\tENTER = 13, /* Enter */\n\tCTRL_N = 14, /* Ctrl-n */\n\tCTRL_P = 16, /* Ctrl-p */\n\tCTRL_T = 20, /* Ctrl-t */\n\tCTRL_U = 21, /* Ctrl+u */\n\tCTRL_W = 23, /* Ctrl+w */\n\tESC = 27, /* Escape */\n\tBACKSPACE = 127 /* Backspace */\n};\n\nstatic void linenoiseAtExit(void);\nint linenoiseHistoryAdd(const char *line);\nstatic void refreshLine(struct linenoiseState *l);\n\n/* Debugging macro. */\n#if 0\nFILE *lndebug_fp = NULL;\n#define lndebug(...) \\\n do { \\\n if (lndebug_fp == NULL) { \\\n lndebug_fp = fopen(\"/tmp/lndebug.txt\",\"a\"); \\\n fprintf(lndebug_fp, \\\n \"[%d %d %d] p: %d, rows: %d, rpos: %d, max: %d, oldmax: %d\\n\", \\\n (int)l->len,(int)l->pos,(int)l->oldpos,plen,rows,rpos, \\\n (int)l->maxrows,old_rows); \\\n } \\\n fprintf(lndebug_fp, \", \" __VA_ARGS__); \\\n fflush(lndebug_fp); \\\n } while (0)\n#else\n#define lndebug(fmt, ...)\n#endif\n\n/* ======================= Low level terminal handling ====================== */\n\n/* Set if to use or not the multi line mode. */\nvoid linenoiseSetMultiLine(int ml) {\n mlmode = ml;\n}\n\n/* Return true if the terminal name is in the list of terminals we know are\n * not able to understand basic escape sequences. */\nstatic int isUnsupportedTerm(void) {\n char *term = getenv(\"TERM\");\n int j;\n\n if (term == NULL) return 0;\n for (j = 0; unsupported_term[j]; j++)\n if (!strcasecmp(term,unsupported_term[j])) return 1;\n return 0;\n}\n\n/* Raw mode: 1960 magic shit. */\nstatic int enableRawMode(int fd) {\n struct termios raw;\n\n if (!isatty(STDIN_FILENO)) goto fatal;\n if (!atexit_registered) {\n atexit(linenoiseAtExit);\n atexit_registered = 1;\n }\n if (tcgetattr(fd,&orig_termios) == -1) goto fatal;\n\n raw = orig_termios; /* modify the original mode */\n /* input modes: no break, no CR to NL, no parity check, no strip char,\n * no start/stop output control. */\n raw.c_iflag &= ~(BRKINT | ICRNL | INPCK | ISTRIP | IXON);\n /* output modes - disable post processing */\n raw.c_oflag &= ~(OPOST);\n /* control modes - set 8 bit chars */\n raw.c_cflag |= (CS8);\n /* local modes - choing off, canonical off, no extended functions,\n * no signal chars (^Z,^C) */\n raw.c_lflag &= ~(ECHO | ICANON | IEXTEN | ISIG);\n /* control chars - set return condition: min number of bytes and timer.\n * We want read to return every single byte, without timeout. */\n raw.c_cc[VMIN] = 1; raw.c_cc[VTIME] = 0; /* 1 byte, no timer */\n\n /* put terminal in raw mode after flushing */\n if (tcsetattr(fd,TCSAFLUSH,&raw) < 0) goto fatal;\n rawmode = 1;\n return 0;\n\nfatal:\n errno = ENOTTY;\n return -1;\n}\n\nstatic void disableRawMode(int fd) {\n /* Don't even check the return value as it's too late. */\n if (rawmode && tcsetattr(fd,TCSAFLUSH,&orig_termios) != -1)\n rawmode = 0;\n}\n\n/* Use the ESC [6n escape sequence to query the horizontal cursor position\n * and return it. On error -1 is returned, on success the position of the\n * cursor. */\nstatic int getCursorPosition(int ifd, int ofd) {\n char buf[32];\n int cols, rows;\n unsigned int i = 0;\n\n /* Report cursor location */\n if (write(ofd, \"\\x1b[6n\", 4) != 4) return -1;\n\n /* Read the response: ESC [ rows ; cols R */\n while (i < sizeof(buf)-1) {\n if (read(ifd,buf+i,1) != 1) break;\n if (buf[i] == 'R') break;\n i++;\n }\n buf[i] = '\\0';\n\n /* Parse it. */\n if (buf[0] != ESC || buf[1] != '[') return -1;\n if (sscanf(buf+2,\"%d;%d\",&rows,&cols) != 2) return -1;\n return cols;\n}\n\n/* Try to get the number of columns in the current terminal, or assume 80\n * if it fails. */\nstatic int getColumns(int ifd, int ofd) {\n struct winsize ws;\n\n if (ioctl(1, TIOCGWINSZ, &ws) == -1 || ws.ws_col == 0) {\n /* ioctl() failed. Try to query the terminal itself. */\n int start, cols;\n\n /* Get the initial position so we can restore it later. */\n start = getCursorPosition(ifd,ofd);\n if (start == -1) goto failed;\n\n /* Go to right margin and get position. */\n if (write(ofd,\"\\x1b[999C\",6) != 6) goto failed;\n cols = getCursorPosition(ifd,ofd);\n if (cols == -1) goto failed;\n\n /* Restore position. */\n if (cols > start) {\n char seq[32];\n snprintf(seq,32,\"\\x1b[%dD\",cols-start);\n if (write(ofd,seq,strlen(seq)) == -1) {\n /* Can't recover... */\n }\n }\n return cols;\n } else {\n return ws.ws_col;\n }\n\nfailed:\n return 80;\n}\n\n/* Clear the screen. Used to handle ctrl+l */\nvoid linenoiseClearScreen(void) {\n if (write(STDOUT_FILENO,\"\\x1b[H\\x1b[2J\",7) <= 0) {\n /* nothing to do, just to avoid warning. */\n }\n}\n\n/* Beep, used for completion when there is nothing to complete or when all\n * the choices were already shown. */\nstatic void linenoiseBeep(void) {\n fprintf(stderr, \"\\x7\");\n fflush(stderr);\n}\n\n/* ============================== Completion ================================ */\n\n/* Free a list of completion option populated by linenoiseAddCompletion(). */\nstatic void freeCompletions(linenoiseCompletions *lc) {\n size_t i;\n for (i = 0; i < lc->len; i++)\n free(lc->cvec[i]);\n if (lc->cvec != NULL)\n free(lc->cvec);\n}\n\n/* This is an helper function for linenoiseEdit() and is called when the\n * user types the key in order to complete the string currently in the\n * input.\n * \n * The state of the editing is encapsulated into the pointed linenoiseState\n * structure as described in the structure definition. */\nstatic int completeLine(struct linenoiseState *ls) {\n linenoiseCompletions lc = { 0, NULL };\n int nread, nwritten;\n char c = 0;\n\n completionCallback(ls->buf,&lc);\n if (lc.len == 0) {\n linenoiseBeep();\n } else {\n size_t stop = 0, i = 0;\n\n while(!stop) {\n /* Show completion or original buffer */\n if (i < lc.len) {\n struct linenoiseState saved = *ls;\n\n ls->len = ls->pos = strlen(lc.cvec[i]);\n ls->buf = lc.cvec[i];\n refreshLine(ls);\n ls->len = saved.len;\n ls->pos = saved.pos;\n ls->buf = saved.buf;\n } else {\n refreshLine(ls);\n }\n\n nread = read(ls->ifd,&c,1);\n if (nread <= 0) {\n freeCompletions(&lc);\n return -1;\n }\n\n switch(c) {\n case 9: /* tab */\n i = (i+1) % (lc.len+1);\n if (i == lc.len) linenoiseBeep();\n break;\n case 27: /* escape */\n /* Re-show original buffer */\n if (i < lc.len) refreshLine(ls);\n stop = 1;\n break;\n default:\n /* Update buffer and return */\n if (i < lc.len) {\n nwritten = snprintf(ls->buf,ls->buflen,\"%s\",lc.cvec[i]);\n ls->len = ls->pos = nwritten;\n }\n stop = 1;\n break;\n }\n }\n }\n\n freeCompletions(&lc);\n return c; /* Return last read character */\n}\n\n/* Register a callback function to be called for tab-completion. */\nvoid linenoiseSetCompletionCallback(linenoiseCompletionCallback *fn) {\n completionCallback = fn;\n}\n\n/* This function is used by the callback function registered by the user\n * in order to add completion options given the input string when the\n * user typed . See the example.c source code for a very easy to\n * understand example. */\nvoid linenoiseAddCompletion(linenoiseCompletions *lc, const char *str) {\n size_t len = strlen(str);\n char *copy, **cvec;\n\n copy = malloc(len+1);\n if (copy == NULL) return;\n memcpy(copy,str,len+1);\n cvec = realloc(lc->cvec,sizeof(char*)*(lc->len+1));\n if (cvec == NULL) {\n free(copy);\n return;\n }\n lc->cvec = cvec;\n lc->cvec[lc->len++] = copy;\n}\n\n/* =========================== Line editing ================================= */\n\n/* We define a very simple \"append buffer\" structure, that is an heap\n * allocated string where we can append to. This is useful in order to\n * write all the escape sequences in a buffer and flush them to the standard\n * output in a single call, to avoid flickering effects. */\nstruct abuf {\n char *b;\n int len;\n};\n\nstatic void abInit(struct abuf *ab) {\n ab->b = NULL;\n ab->len = 0;\n}\n\nstatic void abAppend(struct abuf *ab, const char *s, int len) {\n char *new = realloc(ab->b,ab->len+len);\n\n if (new == NULL) return;\n memcpy(new+ab->len,s,len);\n ab->b = new;\n ab->len += len;\n}\n\nstatic void abFree(struct abuf *ab) {\n free(ab->b);\n}\n\n/* Single line low level line refresh.\n *\n * Rewrite the currently edited line accordingly to the buffer content,\n * cursor position, and number of columns of the terminal. */\nstatic void refreshSingleLine(struct linenoiseState *l) {\n char seq[64];\n size_t plen = strlen(l->prompt);\n int fd = l->ofd;\n char *buf = l->buf;\n size_t len = l->len;\n size_t pos = l->pos;\n struct abuf ab;\n \n while((plen+pos) >= l->cols) {\n buf++;\n len--;\n pos--;\n }\n while (plen+len > l->cols) {\n len--;\n }\n\n abInit(&ab);\n /* Cursor to left edge */\n snprintf(seq,64,\"\\x1b[0G\");\n abAppend(&ab,seq,strlen(seq));\n /* Write the prompt and the current buffer content */\n abAppend(&ab,l->prompt,strlen(l->prompt));\n abAppend(&ab,buf,len);\n /* Erase to right */\n snprintf(seq,64,\"\\x1b[0K\");\n abAppend(&ab,seq,strlen(seq));\n /* Move cursor to original position. */\n snprintf(seq,64,\"\\x1b[0G\\x1b[%dC\", (int)(pos+plen));\n abAppend(&ab,seq,strlen(seq));\n if (write(fd,ab.b,ab.len) == -1) {} /* Can't recover from write error. */\n abFree(&ab);\n}\n\n/* Multi line low level line refresh.\n *\n * Rewrite the currently edited line accordingly to the buffer content,\n * cursor position, and number of columns of the terminal. */\nstatic void refreshMultiLine(struct linenoiseState *l) {\n char seq[64];\n int plen = strlen(l->prompt);\n int rows = (plen+l->len+l->cols-1)/l->cols; /* rows used by current buf. */\n int rpos = (plen+l->oldpos+l->cols)/l->cols; /* cursor relative row. */\n int rpos2; /* rpos after refresh. */\n int old_rows = l->maxrows;\n int fd = l->ofd, j;\n struct abuf ab;\n\n /* Update maxrows if needed. */\n if (rows > (int)l->maxrows) l->maxrows = rows;\n\n /* First step: clear all the lines used before. To do so start by\n * going to the last row. */\n abInit(&ab);\n if (old_rows-rpos > 0) {\n lndebug(\"go down %d\", old_rows-rpos);\n snprintf(seq,64,\"\\x1b[%dB\", old_rows-rpos);\n abAppend(&ab,seq,strlen(seq));\n }\n\n /* Now for every row clear it, go up. */\n for (j = 0; j < old_rows-1; j++) {\n lndebug(\"clear+up\");\n snprintf(seq,64,\"\\x1b[0G\\x1b[0K\\x1b[1A\");\n abAppend(&ab,seq,strlen(seq));\n }\n\n /* Clean the top line. */\n lndebug(\"clear\");\n snprintf(seq,64,\"\\x1b[0G\\x1b[0K\");\n abAppend(&ab,seq,strlen(seq));\n \n /* Write the prompt and the current buffer content */\n abAppend(&ab,l->prompt,strlen(l->prompt));\n abAppend(&ab,l->buf,l->len);\n\n /* If we are at the very end of the screen with our prompt, we need to\n * emit a newline and move the prompt to the first column. */\n if (l->pos &&\n l->pos == l->len &&\n (l->pos+plen) % l->cols == 0)\n {\n lndebug(\"\");\n abAppend(&ab,\"\\n\",1);\n snprintf(seq,64,\"\\x1b[0G\");\n abAppend(&ab,seq,strlen(seq));\n rows++;\n if (rows > (int)l->maxrows) l->maxrows = rows;\n }\n\n /* Move cursor to right position. */\n rpos2 = (plen+l->pos+l->cols)/l->cols; /* current cursor relative row. */\n lndebug(\"rpos2 %d\", rpos2);\n\n /* Go up till we reach the expected positon. */\n if (rows-rpos2 > 0) {\n lndebug(\"go-up %d\", rows-rpos2);\n snprintf(seq,64,\"\\x1b[%dA\", rows-rpos2);\n abAppend(&ab,seq,strlen(seq));\n }\n\n /* Set column. */\n lndebug(\"set col %d\", 1+((plen+(int)l->pos) % (int)l->cols));\n snprintf(seq,64,\"\\x1b[%dG\", 1+((plen+(int)l->pos) % (int)l->cols));\n abAppend(&ab,seq,strlen(seq));\n\n lndebug(\"\\n\");\n l->oldpos = l->pos;\n\n if (write(fd,ab.b,ab.len) == -1) {} /* Can't recover from write error. */\n abFree(&ab);\n}\n\n/* Calls the two low level functions refreshSingleLine() or\n * refreshMultiLine() according to the selected mode. */\nstatic void refreshLine(struct linenoiseState *l) {\n if (mlmode)\n refreshMultiLine(l);\n else\n refreshSingleLine(l);\n}\n\n/* Insert the character 'c' at cursor current position.\n *\n * On error writing to the terminal -1 is returned, otherwise 0. */\nint linenoiseEditInsert(struct linenoiseState *l, char c) {\n if (l->len < l->buflen) {\n if (l->len == l->pos) {\n l->buf[l->pos] = c;\n l->pos++;\n l->len++;\n l->buf[l->len] = '\\0';\n if ((!mlmode && l->plen+l->len < l->cols) /* || mlmode */) {\n /* Avoid a full update of the line in the\n * trivial case. */\n if (write(l->ofd,&c,1) == -1) return -1;\n } else {\n refreshLine(l);\n }\n } else {\n memmove(l->buf+l->pos+1,l->buf+l->pos,l->len-l->pos);\n l->buf[l->pos] = c;\n l->len++;\n l->pos++;\n l->buf[l->len] = '\\0';\n refreshLine(l);\n }\n }\n return 0;\n}\n\n/* Move cursor on the left. */\nvoid linenoiseEditMoveLeft(struct linenoiseState *l) {\n if (l->pos > 0) {\n l->pos--;\n refreshLine(l);\n }\n}\n\n/* Move cursor on the right. */\nvoid linenoiseEditMoveRight(struct linenoiseState *l) {\n if (l->pos != l->len) {\n l->pos++;\n refreshLine(l);\n }\n}\n\n/* Move cursor to the start of the line. */\nvoid linenoiseEditMoveHome(struct linenoiseState *l) {\n if (l->pos != 0) {\n l->pos = 0;\n refreshLine(l);\n }\n}\n\n/* Move cursor to the end of the line. */\nvoid linenoiseEditMoveEnd(struct linenoiseState *l) {\n if (l->pos != l->len) {\n l->pos = l->len;\n refreshLine(l);\n }\n}\n\n/* Substitute the currently edited line with the next or previous history\n * entry as specified by 'dir'. */\n#define LINENOISE_HISTORY_NEXT 0\n#define LINENOISE_HISTORY_PREV 1\nvoid linenoiseEditHistoryNext(struct linenoiseState *l, int dir) {\n if (history_len > 1) {\n /* Update the current history entry before to\n * overwrite it with the next one. */\n free(history[history_len - 1 - l->history_index]);\n history[history_len - 1 - l->history_index] = strdup(l->buf);\n /* Show the new entry */\n l->history_index += (dir == LINENOISE_HISTORY_PREV) ? 1 : -1;\n if (l->history_index < 0) {\n l->history_index = 0;\n return;\n } else if (l->history_index >= history_len) {\n l->history_index = history_len-1;\n return;\n }\n strncpy(l->buf,history[history_len - 1 - l->history_index],l->buflen);\n l->buf[l->buflen-1] = '\\0';\n l->len = l->pos = strlen(l->buf);\n refreshLine(l);\n }\n}\n\n/* Delete the character at the right of the cursor without altering the cursor\n * position. Basically this is what happens with the \"Delete\" keyboard key. */\nvoid linenoiseEditDelete(struct linenoiseState *l) {\n if (l->len > 0 && l->pos < l->len) {\n memmove(l->buf+l->pos,l->buf+l->pos+1,l->len-l->pos-1);\n l->len--;\n l->buf[l->len] = '\\0';\n refreshLine(l);\n }\n}\n\n/* Backspace implementation. */\nvoid linenoiseEditBackspace(struct linenoiseState *l) {\n if (l->pos > 0 && l->len > 0) {\n memmove(l->buf+l->pos-1,l->buf+l->pos,l->len-l->pos);\n l->pos--;\n l->len--;\n l->buf[l->len] = '\\0';\n refreshLine(l);\n }\n}\n\n/* Delete the previosu word, maintaining the cursor at the start of the\n * current word. */\nvoid linenoiseEditDeletePrevWord(struct linenoiseState *l) {\n size_t old_pos = l->pos;\n size_t diff;\n\n while (l->pos > 0 && l->buf[l->pos-1] == ' ')\n l->pos--;\n while (l->pos > 0 && l->buf[l->pos-1] != ' ')\n l->pos--;\n diff = old_pos - l->pos;\n memmove(l->buf+l->pos,l->buf+old_pos,l->len-old_pos+1);\n l->len -= diff;\n refreshLine(l);\n}\n\n/* This function is the core of the line editing capability of linenoise.\n * It expects 'fd' to be already in \"raw mode\" so that every key pressed\n * will be returned ASAP to read().\n *\n * The resulting string is put into 'buf' when the user type enter, or\n * when ctrl+d is typed.\n *\n * The function returns the length of the current buffer. */\nstatic int linenoiseEdit(int stdin_fd, int stdout_fd, char *buf, size_t buflen, const char *prompt)\n{\n struct linenoiseState l;\n\n /* Populate the linenoise state that we pass to functions implementing\n * specific editing functionalities. */\n l.ifd = stdin_fd;\n l.ofd = stdout_fd;\n l.buf = buf;\n l.buflen = buflen;\n l.prompt = prompt;\n l.plen = strlen(prompt);\n l.oldpos = l.pos = 0;\n l.len = 0;\n l.cols = getColumns(stdin_fd, stdout_fd);\n l.maxrows = 0;\n l.history_index = 0;\n\n /* Buffer starts empty. */\n l.buf[0] = '\\0';\n l.buflen--; /* Make sure there is always space for the nulterm */\n\n /* The latest history entry is always our current buffer, that\n * initially is just an empty string. */\n linenoiseHistoryAdd(\"\");\n \n if (write(l.ofd,prompt,l.plen) == -1) return -1;\n while(1) {\n char c;\n int nread;\n char seq[3];\n\n nread = read(l.ifd,&c,1);\n if (nread <= 0) return l.len;\n\n /* Only autocomplete when the callback is set. It returns < 0 when\n * there was an error reading from fd. Otherwise it will return the\n * character that should be handled next. */\n if (c == 9 && completionCallback != NULL) {\n c = completeLine(&l);\n /* Return on errors */\n if (c < 0) return l.len;\n /* Read next character when 0 */\n if (c == 0) continue;\n }\n\n switch(c) {\n case ENTER: /* enter */\n history_len--;\n free(history[history_len]);\n return (int)l.len;\n case CTRL_C: /* ctrl-c */\n errno = EAGAIN;\n return -1;\n case BACKSPACE: /* backspace */\n case 8: /* ctrl-h */\n linenoiseEditBackspace(&l);\n break;\n case CTRL_D: /* ctrl-d, remove char at right of cursor, or of the\n line is empty, act as end-of-file. */\n if (l.len > 0) {\n linenoiseEditDelete(&l);\n } else {\n history_len--;\n free(history[history_len]);\n return -1;\n }\n break;\n case CTRL_T: /* ctrl-t, swaps current character with previous. */\n if (l.pos > 0 && l.pos < l.len) {\n int aux = buf[l.pos-1];\n buf[l.pos-1] = buf[l.pos];\n buf[l.pos] = aux;\n if (l.pos != l.len-1) l.pos++;\n refreshLine(&l);\n }\n break;\n case CTRL_B: /* ctrl-b */\n linenoiseEditMoveLeft(&l);\n break;\n case CTRL_F: /* ctrl-f */\n linenoiseEditMoveRight(&l);\n break;\n case CTRL_P: /* ctrl-p */\n linenoiseEditHistoryNext(&l, LINENOISE_HISTORY_PREV);\n break;\n case CTRL_N: /* ctrl-n */\n linenoiseEditHistoryNext(&l, LINENOISE_HISTORY_NEXT);\n break;\n case ESC: /* escape sequence */\n /* Read the next two bytes representing the escape sequence.\n * Use two calls to handle slow terminals returning the two\n * chars at different times. */\n if (read(l.ifd,seq,1) == -1) break;\n if (read(l.ifd,seq+1,1) == -1) break;\n\n /* ESC [ sequences. */\n if (seq[0] == '[') {\n if (seq[1] >= '0' && seq[1] <= '9') {\n /* Extended escape, read additional byte. */\n if (read(l.ifd,seq+2,1) == -1) break;\n if (seq[2] == '~') {\n switch(seq[1]) {\n case '3': /* Delete key. */\n linenoiseEditDelete(&l);\n break;\n }\n }\n } else {\n switch(seq[1]) {\n case 'A': /* Up */\n linenoiseEditHistoryNext(&l, LINENOISE_HISTORY_PREV);\n break;\n case 'B': /* Down */\n linenoiseEditHistoryNext(&l, LINENOISE_HISTORY_NEXT);\n break;\n case 'C': /* Right */\n linenoiseEditMoveRight(&l);\n break;\n case 'D': /* Left */\n linenoiseEditMoveLeft(&l);\n break;\n case 'H': /* Home */\n linenoiseEditMoveHome(&l);\n break;\n case 'F': /* End*/\n linenoiseEditMoveEnd(&l);\n break;\n }\n }\n }\n\n /* ESC O sequences. */\n else if (seq[0] == 'O') {\n switch(seq[1]) {\n case 'H': /* Home */\n linenoiseEditMoveHome(&l);\n break;\n case 'F': /* End*/\n linenoiseEditMoveEnd(&l);\n break;\n }\n }\n break;\n default:\n if (linenoiseEditInsert(&l,c)) return -1;\n break;\n case CTRL_U: /* Ctrl+u, delete the whole line. */\n buf[0] = '\\0';\n l.pos = l.len = 0;\n refreshLine(&l);\n break;\n case CTRL_K: /* Ctrl+k, delete from current to end of line. */\n buf[l.pos] = '\\0';\n l.len = l.pos;\n refreshLine(&l);\n break;\n case CTRL_A: /* Ctrl+a, go to the start of the line */\n linenoiseEditMoveHome(&l);\n break;\n case CTRL_E: /* ctrl+e, go to the end of the line */\n linenoiseEditMoveEnd(&l);\n break;\n case CTRL_L: /* ctrl+l, clear screen */\n linenoiseClearScreen();\n refreshLine(&l);\n break;\n case CTRL_W: /* ctrl+w, delete previous word */\n linenoiseEditDeletePrevWord(&l);\n break;\n }\n }\n return l.len;\n}\n\n/* This special mode is used by linenoise in order to print scan codes\n * on screen for debugging / development purposes. It is implemented\n * by the linenoise_example program using the --keycodes option. */\nvoid linenoisePrintKeyCodes(void) {\n char quit[4];\n\n printf(\"Linenoise key codes debugging mode.\\n\"\n \"Press keys to see scan codes. Type 'quit' at any time to exit.\\n\");\n if (enableRawMode(STDIN_FILENO) == -1) return;\n memset(quit,' ',4);\n while(1) {\n char c;\n int nread;\n\n nread = read(STDIN_FILENO,&c,1);\n if (nread <= 0) continue;\n memmove(quit,quit+1,sizeof(quit)-1); /* shift string to left. */\n quit[sizeof(quit)-1] = c; /* Insert current char on the right. */\n if (memcmp(quit,\"quit\",sizeof(quit)) == 0) break;\n\n printf(\"'%c' %02x (%d) (type quit to exit)\\n\",\n isprint(c) ? c : '?', (int)c, (int)c);\n printf(\"\\x1b[0G\"); /* Go left edge manually, we are in raw mode. */\n fflush(stdout);\n }\n disableRawMode(STDIN_FILENO);\n}\n\n/* This function calls the line editing function linenoiseEdit() using\n * the STDIN file descriptor set in raw mode. */\nstatic int linenoiseRaw(char *buf, size_t buflen, const char *prompt) {\n int count;\n\n if (buflen == 0) {\n errno = EINVAL;\n return -1;\n }\n if (!isatty(STDIN_FILENO)) {\n /* Not a tty: read from file / pipe. */\n if (fgets(buf, buflen, stdin) == NULL) return -1;\n count = strlen(buf);\n if (count && buf[count-1] == '\\n') {\n count--;\n buf[count] = '\\0';\n }\n } else {\n /* Interactive editing. */\n if (enableRawMode(STDIN_FILENO) == -1) return -1;\n count = linenoiseEdit(STDIN_FILENO, STDOUT_FILENO, buf, buflen, prompt);\n disableRawMode(STDIN_FILENO);\n printf(\"\\n\");\n }\n return count;\n}\n\n/* The high level function that is the main API of the linenoise library.\n * This function checks if the terminal has basic capabilities, just checking\n * for a blacklist of stupid terminals, and later either calls the line\n * editing function or uses dummy fgets() so that you will be able to type\n * something even in the most desperate of the conditions. */\nchar *linenoise(const char *prompt) {\n char buf[LINENOISE_MAX_LINE];\n int count;\n\n if (isUnsupportedTerm()) {\n size_t len;\n\n printf(\"%s\",prompt);\n fflush(stdout);\n if (fgets(buf,LINENOISE_MAX_LINE,stdin) == NULL) return NULL;\n len = strlen(buf);\n while(len && (buf[len-1] == '\\n' || buf[len-1] == '\\r')) {\n len--;\n buf[len] = '\\0';\n }\n return strdup(buf);\n } else {\n count = linenoiseRaw(buf,LINENOISE_MAX_LINE,prompt);\n if (count == -1) return NULL;\n return strdup(buf);\n }\n}\n\n/* ================================ History ================================= */\n\n/* Free the history, but does not reset it. Only used when we have to\n * exit() to avoid memory leaks are reported by valgrind & co. */\nstatic void freeHistory(void) {\n if (history) {\n int j;\n\n for (j = 0; j < history_len; j++)\n free(history[j]);\n free(history);\n }\n}\n\n/* At exit we'll try to fix the terminal to the initial conditions. */\nstatic void linenoiseAtExit(void) {\n disableRawMode(STDIN_FILENO);\n freeHistory();\n}\n\n/* This is the API call to add a new entry in the linenoise history.\n * It uses a fixed array of char pointers that are shifted (memmoved)\n * when the history max length is reached in order to remove the older\n * entry and make room for the new one, so it is not exactly suitable for huge\n * histories, but will work well for a few hundred of entries.\n *\n * Using a circular buffer is smarter, but a bit more complex to handle. */\nint linenoiseHistoryAdd(const char *line) {\n char *linecopy;\n\n if (history_max_len == 0) return 0;\n\n /* Initialization on first call. */\n if (history == NULL) {\n history = malloc(sizeof(char*)*history_max_len);\n if (history == NULL) return 0;\n memset(history,0,(sizeof(char*)*history_max_len));\n }\n\n /* Don't add duplicated lines. */\n if (history_len && !strcmp(history[history_len-1], line)) return 0;\n\n /* Add an heap allocated copy of the line in the history.\n * If we reached the max length, remove the older line. */\n linecopy = strdup(line);\n if (!linecopy) return 0;\n if (history_len == history_max_len) {\n free(history[0]);\n memmove(history,history+1,sizeof(char*)*(history_max_len-1));\n history_len--;\n }\n history[history_len] = linecopy;\n history_len++;\n return 1;\n}\n\n/* Set the maximum length for the history. This function can be called even\n * if there is already some history, the function will make sure to retain\n * just the latest 'len' elements if the new history length value is smaller\n * than the amount of items already inside the history. */\nint linenoiseHistorySetMaxLen(int len) {\n char **new;\n\n if (len < 1) return 0;\n if (history) {\n int tocopy = history_len;\n\n new = malloc(sizeof(char*)*len);\n if (new == NULL) return 0;\n\n /* If we can't copy everything, free the elements we'll not use. */\n if (len < tocopy) {\n int j;\n\n for (j = 0; j < tocopy-len; j++) free(history[j]);\n tocopy = len;\n }\n memset(new,0,sizeof(char*)*len);\n memcpy(new,history+(history_len-tocopy), sizeof(char*)*tocopy);\n free(history);\n history = new;\n }\n history_max_len = len;\n if (history_len > history_max_len)\n history_len = history_max_len;\n return 1;\n}\n\n/* Save the history in the specified file. On success 0 is returned\n * otherwise -1 is returned. */\nint linenoiseHistorySave(const char *filename) {\n FILE *fp = fopen(filename,\"w\");\n int j;\n \n if (fp == NULL) return -1;\n for (j = 0; j < history_len; j++)\n fprintf(fp,\"%s\\n\",history[j]);\n fclose(fp);\n return 0;\n}\n\n/* Load the history from the specified file. If the file does not exist\n * zero is returned and no operation is performed.\n *\n * If the file exists and the operation succeeded 0 is returned, otherwise\n * on error -1 is returned. */\nint linenoiseHistoryLoad(const char *filename) {\n FILE *fp = fopen(filename,\"r\");\n char buf[LINENOISE_MAX_LINE];\n \n if (fp == NULL) return -1;\n\n while (fgets(buf,LINENOISE_MAX_LINE,fp) != NULL) {\n char *p;\n \n p = strchr(buf,'\\r');\n if (!p) p = strchr(buf,'\\n');\n if (p) *p = '\\0';\n linenoiseHistoryAdd(buf);\n }\n fclose(fp);\n return 0;\n}\n" }, { "path": "deps/linenoise/linenoise.h", "content": "/* linenoise.h -- guerrilla line editing library against the idea that a\n * line editing lib needs to be 20,000 lines of C code.\n *\n * See linenoise.c for more information.\n *\n * ------------------------------------------------------------------------\n *\n * Copyright (c) 2010, Salvatore Sanfilippo \n * Copyright (c) 2010, Pieter Noordhuis \n *\n * All rights reserved.\n * \n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are\n * met:\n * \n * * Redistributions of source code must retain the above copyright\n * notice, this list of conditions and the following disclaimer.\n *\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * \n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\n * \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\n * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\n * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\n * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\n * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\n * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\n * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\n * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __LINENOISE_H\n#define __LINENOISE_H\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\ntypedef struct linenoiseCompletions {\n size_t len;\n char **cvec;\n} linenoiseCompletions;\n\ntypedef void(linenoiseCompletionCallback)(const char *, linenoiseCompletions *);\nvoid linenoiseSetCompletionCallback(linenoiseCompletionCallback *);\nvoid linenoiseAddCompletion(linenoiseCompletions *, const char *);\n\nchar *linenoise(const char *prompt);\nint linenoiseHistoryAdd(const char *line);\nint linenoiseHistorySetMaxLen(int len);\nint linenoiseHistorySave(const char *filename);\nint linenoiseHistoryLoad(const char *filename);\nvoid linenoiseClearScreen(void);\nvoid linenoiseSetMultiLine(int ml);\nvoid linenoisePrintKeyCodes(void);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif /* __LINENOISE_H */\n" }, { "path": "deps/lua/COPYRIGHT", "content": "Lua License\n-----------\n\nLua is licensed under the terms of the MIT license reproduced below.\nThis means that Lua is free software and can be used for both academic\nand commercial purposes at absolutely no cost.\n\nFor details and rationale, see http://www.lua.org/license.html .\n\n===============================================================================\n\nCopyright (C) 1994-2012 Lua.org, PUC-Rio.\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in\nall copies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN\nTHE SOFTWARE.\n\n===============================================================================\n\n(end of COPYRIGHT)\n" }, { "path": "deps/lua/HISTORY", "content": "HISTORY for Lua 5.1\n\n* Changes from version 5.0 to 5.1\n -------------------------------\n Language:\n + new module system.\n + new semantics for control variables of fors.\n + new semantics for setn/getn.\n + new syntax/semantics for varargs.\n + new long strings and comments.\n + new `mod' operator (`%')\n + new length operator #t\n + metatables for all types\n API:\n + new functions: lua_createtable, lua_get(set)field, lua_push(to)integer.\n + user supplies memory allocator (lua_open becomes lua_newstate).\n + luaopen_* functions must be called through Lua.\n Implementation:\n + new configuration scheme via luaconf.h.\n + incremental garbage collection.\n + better handling of end-of-line in the lexer.\n + fully reentrant parser (new Lua function `load')\n + better support for 64-bit machines.\n + native loadlib support for Mac OS X.\n + standard distribution in only one library (lualib.a merged into lua.a)\n\n* Changes from version 4.0 to 5.0\n -------------------------------\n Language:\n + lexical scoping.\n + Lua coroutines.\n + standard libraries now packaged in tables.\n + tags replaced by metatables and tag methods replaced by metamethods,\n stored in metatables.\n + proper tail calls.\n + each function can have its own global table, which can be shared.\n + new __newindex metamethod, called when we insert a new key into a table.\n + new block comments: --[[ ... ]].\n + new generic for.\n + new weak tables.\n + new boolean type.\n + new syntax \"local function\".\n + (f()) returns the first value returned by f.\n + {f()} fills a table with all values returned by f.\n + \\n ignored in [[\\n .\n + fixed and-or priorities.\n + more general syntax for function definition (e.g. function a.x.y:f()...end).\n + more general syntax for function calls (e.g. (print or write)(9)).\n + new functions (time/date, tmpfile, unpack, require, load*, etc.).\n API:\n + chunks are loaded by using lua_load; new luaL_loadfile and luaL_loadbuffer.\n + introduced lightweight userdata, a simple \"void*\" without a metatable.\n + new error handling protocol: the core no longer prints error messages;\n all errors are reported to the caller on the stack.\n + new lua_atpanic for host cleanup.\n + new, signal-safe, hook scheme.\n Implementation:\n + new license: MIT.\n + new, faster, register-based virtual machine.\n + support for external multithreading and coroutines.\n + new and consistent error message format.\n + the core no longer needs \"stdio.h\" for anything (except for a single\n use of sprintf to convert numbers to strings).\n + lua.c now runs the environment variable LUA_INIT, if present. It can\n be \"@filename\", to run a file, or the chunk itself.\n + support for user extensions in lua.c.\n sample implementation given for command line editing.\n + new dynamic loading library, active by default on several platforms.\n + safe garbage-collector metamethods.\n + precompiled bytecodes checked for integrity (secure binary dostring).\n + strings are fully aligned.\n + position capture in string.find.\n + read('*l') can read lines with embedded zeros.\n\n* Changes from version 3.2 to 4.0\n -------------------------------\n Language:\n + new \"break\" and \"for\" statements (both numerical and for tables).\n + uniform treatment of globals: globals are now stored in a Lua table.\n + improved error messages.\n + no more '$debug': full speed *and* full debug information.\n + new read form: read(N) for next N bytes.\n + general read patterns now deprecated.\n (still available with -DCOMPAT_READPATTERNS.)\n + all return values are passed as arguments for the last function\n (old semantics still available with -DLUA_COMPAT_ARGRET)\n + garbage collection tag methods for tables now deprecated.\n + there is now only one tag method for order.\n API:\n + New API: fully re-entrant, simpler, and more efficient.\n + New debug API.\n Implementation:\n + faster than ever: cleaner virtual machine and new hashing algorithm.\n + non-recursive garbage-collector algorithm.\n + reduced memory usage for programs with many strings.\n + improved treatment for memory allocation errors.\n + improved support for 16-bit machines (we hope).\n + code now compiles unmodified as both ANSI C and C++.\n + numbers in bases other than 10 are converted using strtoul.\n + new -f option in Lua to support #! scripts.\n + luac can now combine text and binaries.\n\n* Changes from version 3.1 to 3.2\n -------------------------------\n + redirected all output in Lua's core to _ERRORMESSAGE and _ALERT.\n + increased limit on the number of constants and globals per function\n (from 2^16 to 2^24).\n + debugging info (lua_debug and hooks) moved into lua_state and new API\n functions provided to get and set this info.\n + new debug lib gives full debugging access within Lua.\n + new table functions \"foreachi\", \"sort\", \"tinsert\", \"tremove\", \"getn\".\n + new io functions \"flush\", \"seek\".\n\n* Changes from version 3.0 to 3.1\n -------------------------------\n + NEW FEATURE: anonymous functions with closures (via \"upvalues\").\n + new syntax:\n - local variables in chunks.\n - better scope control with DO block END.\n - constructors can now be also written: { record-part; list-part }.\n - more general syntax for function calls and lvalues, e.g.:\n f(x).y=1\n o:f(x,y):g(z)\n f\"string\" is sugar for f(\"string\")\n + strings may now contain arbitrary binary data (e.g., embedded zeros).\n + major code re-organization and clean-up; reduced module interdependecies.\n + no arbitrary limits on the total number of constants and globals.\n + support for multiple global contexts.\n + better syntax error messages.\n + new traversal functions \"foreach\" and \"foreachvar\".\n + the default for numbers is now double.\n changing it to use floats or longs is easy.\n + complete debug information stored in pre-compiled chunks.\n + sample interpreter now prompts user when run interactively, and also\n handles control-C interruptions gracefully.\n\n* Changes from version 2.5 to 3.0\n -------------------------------\n + NEW CONCEPT: \"tag methods\".\n Tag methods replace fallbacks as the meta-mechanism for extending the\n semantics of Lua. Whereas fallbacks had a global nature, tag methods\n work on objects having the same tag (e.g., groups of tables).\n Existing code that uses fallbacks should work without change.\n + new, general syntax for constructors {[exp] = exp, ... }.\n + support for handling variable number of arguments in functions (varargs).\n + support for conditional compilation ($if ... $else ... $end).\n + cleaner semantics in API simplifies host code.\n + better support for writing libraries (auxlib.h).\n + better type checking and error messages in the standard library.\n + luac can now also undump.\n\n* Changes from version 2.4 to 2.5\n -------------------------------\n + io and string libraries are now based on pattern matching;\n the old libraries are still available for compatibility\n + dofile and dostring can now return values (via return statement)\n + better support for 16- and 64-bit machines\n + expanded documentation, with more examples\n\n* Changes from version 2.2 to 2.4\n -------------------------------\n + external compiler creates portable binary files that can be loaded faster\n + interface for debugging and profiling\n + new \"getglobal\" fallback\n + new functions for handling references to Lua objects\n + new functions in standard lib\n + only one copy of each string is stored\n + expanded documentation, with more examples\n\n* Changes from version 2.1 to 2.2\n -------------------------------\n + functions now may be declared with any \"lvalue\" as a name\n + garbage collection of functions\n + support for pipes\n\n* Changes from version 1.1 to 2.1\n -------------------------------\n + object-oriented support\n + fallbacks\n + simplified syntax for tables\n + many internal improvements\n\n(end of HISTORY)\n" }, { "path": "deps/lua/INSTALL", "content": "INSTALL for Lua 5.1\n\n* Building Lua\n ------------\n Lua is built in the src directory, but the build process can be\n controlled from the top-level Makefile.\n\n Building Lua on Unix systems should be very easy. First do \"make\" and\n see if your platform is listed. If so, just do \"make xxx\", where xxx\n is your platform name. The platforms currently supported are:\n aix ansi bsd freebsd generic linux macosx mingw posix solaris\n\n If your platform is not listed, try the closest one or posix, generic,\n ansi, in this order.\n\n See below for customization instructions and for instructions on how\n to build with other Windows compilers.\n\n If you want to check that Lua has been built correctly, do \"make test\"\n after building Lua. Also, have a look at the example programs in test.\n\n* Installing Lua\n --------------\n Once you have built Lua, you may want to install it in an official\n place in your system. In this case, do \"make install\". The official\n place and the way to install files are defined in Makefile. You must\n have the right permissions to install files.\n\n If you want to build and install Lua in one step, do \"make xxx install\",\n where xxx is your platform name.\n\n If you want to install Lua locally, then do \"make local\". This will\n create directories bin, include, lib, man, and install Lua there as\n follows:\n\n bin:\tlua luac\n include:\tlua.h luaconf.h lualib.h lauxlib.h lua.hpp\n lib:\tliblua.a\n man/man1:\tlua.1 luac.1\n\n These are the only directories you need for development.\n\n There are man pages for lua and luac, in both nroff and html, and a\n reference manual in html in doc, some sample code in test, and some\n useful stuff in etc. You don't need these directories for development.\n\n If you want to install Lua locally, but in some other directory, do\n \"make install INSTALL_TOP=xxx\", where xxx is your chosen directory.\n\n See below for instructions for Windows and other systems.\n\n* Customization\n -------------\n Three things can be customized by editing a file:\n - Where and how to install Lua -- edit Makefile.\n - How to build Lua -- edit src/Makefile.\n - Lua features -- edit src/luaconf.h.\n\n You don't actually need to edit the Makefiles because you may set the\n relevant variables when invoking make.\n\n On the other hand, if you need to select some Lua features, you'll need\n to edit src/luaconf.h. The edited file will be the one installed, and\n it will be used by any Lua clients that you build, to ensure consistency.\n\n We strongly recommend that you enable dynamic loading. This is done\n automatically for all platforms listed above that have this feature\n (and also Windows). See src/luaconf.h and also src/Makefile.\n\n* Building Lua on Windows and other systems\n -----------------------------------------\n If you're not using the usual Unix tools, then the instructions for\n building Lua depend on the compiler you use. You'll need to create\n projects (or whatever your compiler uses) for building the library,\n the interpreter, and the compiler, as follows:\n\n library:\tlapi.c lcode.c ldebug.c ldo.c ldump.c lfunc.c lgc.c llex.c\n\t\tlmem.c lobject.c lopcodes.c lparser.c lstate.c lstring.c\n\t\tltable.c ltm.c lundump.c lvm.c lzio.c\n\t\tlauxlib.c lbaselib.c ldblib.c liolib.c lmathlib.c loslib.c\n\t\tltablib.c lstrlib.c loadlib.c linit.c\n\n interpreter:\tlibrary, lua.c\n\n compiler:\tlibrary, luac.c print.c\n\n If you use Visual Studio .NET, you can use etc/luavs.bat in its\n \"Command Prompt\".\n\n If all you want is to build the Lua interpreter, you may put all .c files\n in a single project, except for luac.c and print.c. Or just use etc/all.c.\n\n To use Lua as a library in your own programs, you'll need to know how to\n create and use libraries with your compiler.\n\n As mentioned above, you may edit luaconf.h to select some features before\n building Lua.\n\n(end of INSTALL)\n" }, { "path": "deps/lua/Makefile", "content": "# makefile for installing Lua\n# see INSTALL for installation instructions\n# see src/Makefile and src/luaconf.h for further customization\n\n# == CHANGE THE SETTINGS BELOW TO SUIT YOUR ENVIRONMENT =======================\n\n# Your platform. See PLATS for possible values.\nPLAT= none\n\n# Where to install. The installation starts in the src and doc directories,\n# so take care if INSTALL_TOP is not an absolute path.\nINSTALL_TOP= /usr/local\nINSTALL_BIN= $(INSTALL_TOP)/bin\nINSTALL_INC= $(INSTALL_TOP)/include\nINSTALL_LIB= $(INSTALL_TOP)/lib\nINSTALL_MAN= $(INSTALL_TOP)/man/man1\n#\n# You probably want to make INSTALL_LMOD and INSTALL_CMOD consistent with\n# LUA_ROOT, LUA_LDIR, and LUA_CDIR in luaconf.h (and also with etc/lua.pc).\nINSTALL_LMOD= $(INSTALL_TOP)/share/lua/$V\nINSTALL_CMOD= $(INSTALL_TOP)/lib/lua/$V\n\n# How to install. If your install program does not support \"-p\", then you\n# may have to run ranlib on the installed liblua.a (do \"make ranlib\").\nINSTALL= install -p\nINSTALL_EXEC= $(INSTALL) -m 0755\nINSTALL_DATA= $(INSTALL) -m 0644\n#\n# If you don't have install you can use cp instead.\n# INSTALL= cp -p\n# INSTALL_EXEC= $(INSTALL)\n# INSTALL_DATA= $(INSTALL)\n\n# Utilities.\nMKDIR= mkdir -p\nRANLIB= ranlib\n\n# == END OF USER SETTINGS. NO NEED TO CHANGE ANYTHING BELOW THIS LINE =========\n\n# Convenience platforms targets.\nPLATS= aix ansi bsd freebsd generic linux macosx mingw posix solaris\n\n# What to install.\nTO_BIN= lua luac\nTO_INC= lua.h luaconf.h lualib.h lauxlib.h ../etc/lua.hpp\nTO_LIB= liblua.a\nTO_MAN= lua.1 luac.1\n\n# Lua version and release.\nV= 5.1\nR= 5.1.5\n\nall:\t$(PLAT)\n\n$(PLATS) clean:\n\tcd src && $(MAKE) $@\n\ntest:\tdummy\n\tsrc/lua test/hello.lua\n\ninstall: dummy\n\tcd src && $(MKDIR) $(INSTALL_BIN) $(INSTALL_INC) $(INSTALL_LIB) $(INSTALL_MAN) $(INSTALL_LMOD) $(INSTALL_CMOD)\n\tcd src && $(INSTALL_EXEC) $(TO_BIN) $(INSTALL_BIN)\n\tcd src && $(INSTALL_DATA) $(TO_INC) $(INSTALL_INC)\n\tcd src && $(INSTALL_DATA) $(TO_LIB) $(INSTALL_LIB)\n\tcd doc && $(INSTALL_DATA) $(TO_MAN) $(INSTALL_MAN)\n\nranlib:\n\tcd src && cd $(INSTALL_LIB) && $(RANLIB) $(TO_LIB)\n\nlocal:\n\t$(MAKE) install INSTALL_TOP=..\n\nnone:\n\t@echo \"Please do\"\n\t@echo \" make PLATFORM\"\n\t@echo \"where PLATFORM is one of these:\"\n\t@echo \" $(PLATS)\"\n\t@echo \"See INSTALL for complete instructions.\"\n\n# make may get confused with test/ and INSTALL in a case-insensitive OS\ndummy:\n\n# echo config parameters\necho:\n\t@echo \"\"\n\t@echo \"These are the parameters currently set in src/Makefile to build Lua $R:\"\n\t@echo \"\"\n\t@cd src && $(MAKE) -s echo\n\t@echo \"\"\n\t@echo \"These are the parameters currently set in Makefile to install Lua $R:\"\n\t@echo \"\"\n\t@echo \"PLAT = $(PLAT)\"\n\t@echo \"INSTALL_TOP = $(INSTALL_TOP)\"\n\t@echo \"INSTALL_BIN = $(INSTALL_BIN)\"\n\t@echo \"INSTALL_INC = $(INSTALL_INC)\"\n\t@echo \"INSTALL_LIB = $(INSTALL_LIB)\"\n\t@echo \"INSTALL_MAN = $(INSTALL_MAN)\"\n\t@echo \"INSTALL_LMOD = $(INSTALL_LMOD)\"\n\t@echo \"INSTALL_CMOD = $(INSTALL_CMOD)\"\n\t@echo \"INSTALL_EXEC = $(INSTALL_EXEC)\"\n\t@echo \"INSTALL_DATA = $(INSTALL_DATA)\"\n\t@echo \"\"\n\t@echo \"See also src/luaconf.h .\"\n\t@echo \"\"\n\n# echo private config parameters\npecho:\n\t@echo \"V = $(V)\"\n\t@echo \"R = $(R)\"\n\t@echo \"TO_BIN = $(TO_BIN)\"\n\t@echo \"TO_INC = $(TO_INC)\"\n\t@echo \"TO_LIB = $(TO_LIB)\"\n\t@echo \"TO_MAN = $(TO_MAN)\"\n\n# echo config parameters as Lua code\n# uncomment the last sed expression if you want nil instead of empty strings\nlecho:\n\t@echo \"-- installation parameters for Lua $R\"\n\t@echo \"VERSION = '$V'\"\n\t@echo \"RELEASE = '$R'\"\n\t@$(MAKE) echo | grep = | sed -e 's/= /= \"/' -e 's/$$/\"/' #-e 's/\"\"/nil/'\n\t@echo \"-- EOF\"\n\n# list targets that do not create files (but not all makes understand .PHONY)\n.PHONY: all $(PLATS) clean test install local none dummy echo pecho lecho\n\n# (end of Makefile)\n" }, { "path": "deps/lua/README", "content": "README for Lua 5.1\n\nSee INSTALL for installation instructions.\nSee HISTORY for a summary of changes since the last released version.\n\n* What is Lua?\n ------------\n Lua is a powerful, light-weight programming language designed for extending\n applications. Lua is also frequently used as a general-purpose, stand-alone\n language. Lua is free software.\n\n For complete information, visit Lua's web site at http://www.lua.org/ .\n For an executive summary, see http://www.lua.org/about.html .\n\n Lua has been used in many different projects around the world.\n For a short list, see http://www.lua.org/uses.html .\n\n* Availability\n ------------\n Lua is freely available for both academic and commercial purposes.\n See COPYRIGHT and http://www.lua.org/license.html for details.\n Lua can be downloaded at http://www.lua.org/download.html .\n\n* Installation\n ------------\n Lua is implemented in pure ANSI C, and compiles unmodified in all known\n platforms that have an ANSI C compiler. In most Unix-like platforms, simply\n do \"make\" with a suitable target. See INSTALL for detailed instructions.\n\n* Origin\n ------\n Lua is developed at Lua.org, a laboratory of the Department of Computer\n Science of PUC-Rio (the Pontifical Catholic University of Rio de Janeiro\n in Brazil).\n For more information about the authors, see http://www.lua.org/authors.html .\n\n(end of README)\n" }, { "path": "deps/lua/doc/contents.html", "content": "\n\n\nLua 5.1 Reference Manual - contents\n\n\n\n\n\n\n\n
\n

\n\"\"\nLua 5.1 Reference Manual\n

\n\n

\nThe reference manual is the official definition of the Lua language.\nFor a complete introduction to Lua programming, see the book\nProgramming in Lua.\n\n

\nThis manual is also available as a book:\n

\n\n\"\"\n\nLua 5.1 Reference Manual\n
by R. Ierusalimschy, L. H. de Figueiredo, W. Celes\n
Lua.org, August 2006\n
ISBN 85-903798-3-3\n
\n
\n\n

\nBuy a copy\nof this book and\nhelp to support\nthe Lua project.\n\n

\nstart\n·\ncontents\n·\nindex\n·\nother versions\n

\n\nCopyright © 2006–2012 Lua.org, PUC-Rio.\nFreely available under the terms of the\nLua license.\n\n\n

Contents

\n\n\n

Index

\n\n\n\n\n\n\n\n
\n

Lua functions

\n_G
\n_VERSION
\n

\n\nassert
\ncollectgarbage
\ndofile
\nerror
\ngetfenv
\ngetmetatable
\nipairs
\nload
\nloadfile
\nloadstring
\nmodule
\nnext
\npairs
\npcall
\nprint
\nrawequal
\nrawget
\nrawset
\nrequire
\nselect
\nsetfenv
\nsetmetatable
\ntonumber
\ntostring
\ntype
\nunpack
\nxpcall
\n

\n\ncoroutine.create
\ncoroutine.resume
\ncoroutine.running
\ncoroutine.status
\ncoroutine.wrap
\ncoroutine.yield
\n

\n\ndebug.debug
\ndebug.getfenv
\ndebug.gethook
\ndebug.getinfo
\ndebug.getlocal
\ndebug.getmetatable
\ndebug.getregistry
\ndebug.getupvalue
\ndebug.setfenv
\ndebug.sethook
\ndebug.setlocal
\ndebug.setmetatable
\ndebug.setupvalue
\ndebug.traceback
\n\n

\n

 

\nfile:close
\nfile:flush
\nfile:lines
\nfile:read
\nfile:seek
\nfile:setvbuf
\nfile:write
\n

\n\nio.close
\nio.flush
\nio.input
\nio.lines
\nio.open
\nio.output
\nio.popen
\nio.read
\nio.stderr
\nio.stdin
\nio.stdout
\nio.tmpfile
\nio.type
\nio.write
\n

\n\nmath.abs
\nmath.acos
\nmath.asin
\nmath.atan
\nmath.atan2
\nmath.ceil
\nmath.cos
\nmath.cosh
\nmath.deg
\nmath.exp
\nmath.floor
\nmath.fmod
\nmath.frexp
\nmath.huge
\nmath.ldexp
\nmath.log
\nmath.log10
\nmath.max
\nmath.min
\nmath.modf
\nmath.pi
\nmath.pow
\nmath.rad
\nmath.random
\nmath.randomseed
\nmath.sin
\nmath.sinh
\nmath.sqrt
\nmath.tan
\nmath.tanh
\n

\n\nos.clock
\nos.date
\nos.difftime
\nos.execute
\nos.exit
\nos.getenv
\nos.remove
\nos.rename
\nos.setlocale
\nos.time
\nos.tmpname
\n

\n\npackage.cpath
\npackage.loaded
\npackage.loaders
\npackage.loadlib
\npackage.path
\npackage.preload
\npackage.seeall
\n

\n\nstring.byte
\nstring.char
\nstring.dump
\nstring.find
\nstring.format
\nstring.gmatch
\nstring.gsub
\nstring.len
\nstring.lower
\nstring.match
\nstring.rep
\nstring.reverse
\nstring.sub
\nstring.upper
\n

\n\ntable.concat
\ntable.insert
\ntable.maxn
\ntable.remove
\ntable.sort
\n\n

\n

C API

\nlua_Alloc
\nlua_CFunction
\nlua_Debug
\nlua_Hook
\nlua_Integer
\nlua_Number
\nlua_Reader
\nlua_State
\nlua_Writer
\n

\n\nlua_atpanic
\nlua_call
\nlua_checkstack
\nlua_close
\nlua_concat
\nlua_cpcall
\nlua_createtable
\nlua_dump
\nlua_equal
\nlua_error
\nlua_gc
\nlua_getallocf
\nlua_getfenv
\nlua_getfield
\nlua_getglobal
\nlua_gethook
\nlua_gethookcount
\nlua_gethookmask
\nlua_getinfo
\nlua_getlocal
\nlua_getmetatable
\nlua_getstack
\nlua_gettable
\nlua_gettop
\nlua_getupvalue
\nlua_insert
\nlua_isboolean
\nlua_iscfunction
\nlua_isfunction
\nlua_islightuserdata
\nlua_isnil
\nlua_isnone
\nlua_isnoneornil
\nlua_isnumber
\nlua_isstring
\nlua_istable
\nlua_isthread
\nlua_isuserdata
\nlua_lessthan
\nlua_load
\nlua_newstate
\nlua_newtable
\nlua_newthread
\nlua_newuserdata
\nlua_next
\nlua_objlen
\nlua_pcall
\nlua_pop
\nlua_pushboolean
\nlua_pushcclosure
\nlua_pushcfunction
\nlua_pushfstring
\nlua_pushinteger
\nlua_pushlightuserdata
\nlua_pushliteral
\nlua_pushlstring
\nlua_pushnil
\nlua_pushnumber
\nlua_pushstring
\nlua_pushthread
\nlua_pushvalue
\nlua_pushvfstring
\nlua_rawequal
\nlua_rawget
\nlua_rawgeti
\nlua_rawset
\nlua_rawseti
\nlua_register
\nlua_remove
\nlua_replace
\nlua_resume
\nlua_setallocf
\nlua_setfenv
\nlua_setfield
\nlua_setglobal
\nlua_sethook
\nlua_setlocal
\nlua_setmetatable
\nlua_settable
\nlua_settop
\nlua_setupvalue
\nlua_status
\nlua_toboolean
\nlua_tocfunction
\nlua_tointeger
\nlua_tolstring
\nlua_tonumber
\nlua_topointer
\nlua_tostring
\nlua_tothread
\nlua_touserdata
\nlua_type
\nlua_typename
\nlua_upvalueindex
\nlua_xmove
\nlua_yield
\n\n

\n

auxiliary library

\nluaL_Buffer
\nluaL_Reg
\n

\n\nluaL_addchar
\nluaL_addlstring
\nluaL_addsize
\nluaL_addstring
\nluaL_addvalue
\nluaL_argcheck
\nluaL_argerror
\nluaL_buffinit
\nluaL_callmeta
\nluaL_checkany
\nluaL_checkint
\nluaL_checkinteger
\nluaL_checklong
\nluaL_checklstring
\nluaL_checknumber
\nluaL_checkoption
\nluaL_checkstack
\nluaL_checkstring
\nluaL_checktype
\nluaL_checkudata
\nluaL_dofile
\nluaL_dostring
\nluaL_error
\nluaL_getmetafield
\nluaL_getmetatable
\nluaL_gsub
\nluaL_loadbuffer
\nluaL_loadfile
\nluaL_loadstring
\nluaL_newmetatable
\nluaL_newstate
\nluaL_openlibs
\nluaL_optint
\nluaL_optinteger
\nluaL_optlong
\nluaL_optlstring
\nluaL_optnumber
\nluaL_optstring
\nluaL_prepbuffer
\nluaL_pushresult
\nluaL_ref
\nluaL_register
\nluaL_typename
\nluaL_typerror
\nluaL_unref
\nluaL_where
\n\n

\n

\n\n

\n\nLast update:\nMon Feb 13 18:53:32 BRST 2012\n\n\n\n\n\n" }, { "path": "deps/lua/doc/lua.1", "content": ".\\\" $Id: lua.man,v 1.11 2006/01/06 16:03:34 lhf Exp $\n.TH LUA 1 \"$Date: 2006/01/06 16:03:34 $\"\n.SH NAME\nlua \\- Lua interpreter\n.SH SYNOPSIS\n.B lua\n[\n.I options\n]\n[\n.I script\n[\n.I args\n]\n]\n.SH DESCRIPTION\n.B lua\nis the stand-alone Lua interpreter.\nIt loads and executes Lua programs,\neither in textual source form or\nin precompiled binary form.\n(Precompiled binaries are output by\n.BR luac ,\nthe Lua compiler.)\n.B lua\ncan be used as a batch interpreter and also interactively.\n.LP\nThe given\n.I options\n(see below)\nare executed and then\nthe Lua program in file\n.I script\nis loaded and executed.\nThe given\n.I args\nare available to\n.I script\nas strings in a global table named\n.BR arg .\nIf these arguments contain spaces or other characters special to the shell,\nthen they should be quoted\n(but note that the quotes will be removed by the shell).\nThe arguments in\n.B arg\nstart at 0,\nwhich contains the string\n.RI ' script '.\nThe index of the last argument is stored in\n.BR arg.n .\nThe arguments given in the command line before\n.IR script ,\nincluding the name of the interpreter,\nare available in negative indices in\n.BR arg .\n.LP\nAt the very start,\nbefore even handling the command line,\n.B lua\nexecutes the contents of the environment variable\n.BR LUA_INIT ,\nif it is defined.\nIf the value of\n.B LUA_INIT\nis of the form\n.RI '@ filename ',\nthen\n.I filename\nis executed.\nOtherwise, the string is assumed to be a Lua statement and is executed.\n.LP\nOptions start with\n.B '\\-'\nand are described below.\nYou can use\n.B \"'\\--'\"\nto signal the end of options.\n.LP\nIf no arguments are given,\nthen\n.B \"\\-v \\-i\"\nis assumed when the standard input is a terminal;\notherwise,\n.B \"\\-\"\nis assumed.\n.LP\nIn interactive mode,\n.B lua\nprompts the user,\nreads lines from the standard input,\nand executes them as they are read.\nIf a line does not contain a complete statement,\nthen a secondary prompt is displayed and\nlines are read until a complete statement is formed or\na syntax error is found.\nSo, one way to interrupt the reading of an incomplete statement is\nto force a syntax error:\nadding a\n.B ';' \nin the middle of a statement is a sure way of forcing a syntax error\n(except inside multiline strings and comments; these must be closed explicitly).\nIf a line starts with\n.BR '=' ,\nthen\n.B lua\ndisplays the values of all the expressions in the remainder of the\nline. The expressions must be separated by commas.\nThe primary prompt is the value of the global variable\n.BR _PROMPT ,\nif this value is a string;\notherwise, the default prompt is used.\nSimilarly, the secondary prompt is the value of the global variable\n.BR _PROMPT2 .\nSo,\nto change the prompts,\nset the corresponding variable to a string of your choice.\nYou can do that after calling the interpreter\nor on the command line\n(but in this case you have to be careful with quotes\nif the prompt string contains a space; otherwise you may confuse the shell.)\nThe default prompts are \"> \" and \">> \".\n.SH OPTIONS\n.TP\n.B \\-\nload and execute the standard input as a file,\nthat is,\nnot interactively,\neven when the standard input is a terminal.\n.TP\n.BI \\-e \" stat\"\nexecute statement\n.IR stat .\nYou need to quote\n.I stat \nif it contains spaces, quotes,\nor other characters special to the shell.\n.TP\n.B \\-i\nenter interactive mode after\n.I script\nis executed.\n.TP\n.BI \\-l \" name\"\ncall\n.BI require(' name ')\nbefore executing\n.IR script .\nTypically used to load libraries.\n.TP\n.B \\-v\nshow version information.\n.SH \"SEE ALSO\"\n.BR luac (1)\n.br\nhttp://www.lua.org/\n.SH DIAGNOSTICS\nError messages should be self explanatory.\n.SH AUTHORS\nR. Ierusalimschy,\nL. H. de Figueiredo,\nand\nW. Celes\n.\\\" EOF\n" }, { "path": "deps/lua/doc/lua.css", "content": "body {\n\tcolor: #000000 ;\n\tbackground-color: #FFFFFF ;\n\tfont-family: Helvetica, Arial, sans-serif ;\n\ttext-align: justify ;\n\tmargin-right: 30px ;\n\tmargin-left: 30px ;\n}\n\nh1, h2, h3, h4 {\n\tfont-family: Verdana, Geneva, sans-serif ;\n\tfont-weight: normal ;\n\tfont-style: italic ;\n}\n\nh2 {\n\tpadding-top: 0.4em ;\n\tpadding-bottom: 0.4em ;\n\tpadding-left: 30px ;\n\tpadding-right: 30px ;\n\tmargin-left: -30px ;\n\tbackground-color: #E0E0FF ;\n}\n\nh3 {\n\tpadding-left: 0.5em ;\n\tborder-left: solid #E0E0FF 1em ;\n}\n\ntable h3 {\n\tpadding-left: 0px ;\n\tborder-left: none ;\n}\n\na:link {\n\tcolor: #000080 ;\n\tbackground-color: inherit ;\n\ttext-decoration: none ;\n}\n\na:visited {\n\tbackground-color: inherit ;\n\ttext-decoration: none ;\n}\n\na:link:hover, a:visited:hover {\n\tcolor: #000080 ;\n\tbackground-color: #E0E0FF ;\n}\n\na:link:active, a:visited:active {\n\tcolor: #FF0000 ;\n}\n\nhr {\n\tborder: 0 ;\n\theight: 1px ;\n\tcolor: #a0a0a0 ;\n\tbackground-color: #a0a0a0 ;\n}\n\n:target {\n\tbackground-color: #F8F8F8 ;\n\tpadding: 8px ;\n\tborder: solid #a0a0a0 2px ;\n}\n\n.footer {\n\tcolor: gray ;\n\tfont-size: small ;\n}\n\ninput[type=text] {\n\tborder: solid #a0a0a0 2px ;\n\tborder-radius: 2em ;\n\t-moz-border-radius: 2em ;\n\tbackground-image: url('images/search.png') ;\n\tbackground-repeat: no-repeat;\n\tbackground-position: 4px center ;\n\tpadding-left: 20px ;\n\theight: 2em ;\n}\n\n" }, { "path": "deps/lua/doc/lua.html", "content": "\n\n\nLUA man page\n\n\n\n\n\n

NAME

\nlua - Lua interpreter\n

SYNOPSIS

\nlua\n[\noptions\n]\n[\nscript\n[\nargs\n]\n]\n

DESCRIPTION

\nlua\nis the stand-alone Lua interpreter.\nIt loads and executes Lua programs,\neither in textual source form or\nin precompiled binary form.\n(Precompiled binaries are output by\nluac,\nthe Lua compiler.)\nlua\ncan be used as a batch interpreter and also interactively.\n

\nThe given\noptions\n(see below)\nare executed and then\nthe Lua program in file\nscript\nis loaded and executed.\nThe given\nargs\nare available to\nscript\nas strings in a global table named\narg.\nIf these arguments contain spaces or other characters special to the shell,\nthen they should be quoted\n(but note that the quotes will be removed by the shell).\nThe arguments in\narg\nstart at 0,\nwhich contains the string\n'script'.\nThe index of the last argument is stored in\narg.n.\nThe arguments given in the command line before\nscript,\nincluding the name of the interpreter,\nare available in negative indices in\narg.\n

\nAt the very start,\nbefore even handling the command line,\nlua\nexecutes the contents of the environment variable\nLUA_INIT,\nif it is defined.\nIf the value of\nLUA_INIT\nis of the form\n'@filename',\nthen\nfilename\nis executed.\nOtherwise, the string is assumed to be a Lua statement and is executed.\n

\nOptions start with\n'-'\nand are described below.\nYou can use\n'--'\nto signal the end of options.\n

\nIf no arguments are given,\nthen\n\"-v -i\"\nis assumed when the standard input is a terminal;\notherwise,\n\"-\"\nis assumed.\n

\nIn interactive mode,\nlua\nprompts the user,\nreads lines from the standard input,\nand executes them as they are read.\nIf a line does not contain a complete statement,\nthen a secondary prompt is displayed and\nlines are read until a complete statement is formed or\na syntax error is found.\nSo, one way to interrupt the reading of an incomplete statement is\nto force a syntax error:\nadding a\n';'\nin the middle of a statement is a sure way of forcing a syntax error\n(except inside multiline strings and comments; these must be closed explicitly).\nIf a line starts with\n'=',\nthen\nlua\ndisplays the values of all the expressions in the remainder of the\nline. The expressions must be separated by commas.\nThe primary prompt is the value of the global variable\n_PROMPT,\nif this value is a string;\notherwise, the default prompt is used.\nSimilarly, the secondary prompt is the value of the global variable\n_PROMPT2.\nSo,\nto change the prompts,\nset the corresponding variable to a string of your choice.\nYou can do that after calling the interpreter\nor on the command line\n(but in this case you have to be careful with quotes\nif the prompt string contains a space; otherwise you may confuse the shell.)\nThe default prompts are \"> \" and \">> \".\n

OPTIONS

\n

\n-\nload and execute the standard input as a file,\nthat is,\nnot interactively,\neven when the standard input is a terminal.\n

\n-e stat\nexecute statement\nstat.\nYou need to quote\nstat \nif it contains spaces, quotes,\nor other characters special to the shell.\n

\n-i\nenter interactive mode after\nscript\nis executed.\n

\n-l name\ncall\nrequire('name')\nbefore executing\nscript.\nTypically used to load libraries.\n

\n-v\nshow version information.\n

SEE ALSO

\nluac(1)\n
\nhttp://www.lua.org/\n

DIAGNOSTICS

\nError messages should be self explanatory.\n

AUTHORS

\nR. Ierusalimschy,\nL. H. de Figueiredo,\nand\nW. Celes\n\n\n\n" }, { "path": "deps/lua/doc/luac.1", "content": ".\\\" $Id: luac.man,v 1.28 2006/01/06 16:03:34 lhf Exp $\n.TH LUAC 1 \"$Date: 2006/01/06 16:03:34 $\"\n.SH NAME\nluac \\- Lua compiler\n.SH SYNOPSIS\n.B luac\n[\n.I options\n] [\n.I filenames\n]\n.SH DESCRIPTION\n.B luac\nis the Lua compiler.\nIt translates programs written in the Lua programming language\ninto binary files that can be later loaded and executed.\n.LP\nThe main advantages of precompiling chunks are:\nfaster loading,\nprotecting source code from accidental user changes,\nand\noff-line syntax checking.\n.LP\nPre-compiling does not imply faster execution\nbecause in Lua chunks are always compiled into bytecodes before being executed.\n.B luac\nsimply allows those bytecodes to be saved in a file for later execution.\n.LP\nPre-compiled chunks are not necessarily smaller than the corresponding source.\nThe main goal in pre-compiling is faster loading.\n.LP\nThe binary files created by\n.B luac\nare portable only among architectures with the same word size and byte order.\n.LP\n.B luac\nproduces a single output file containing the bytecodes\nfor all source files given.\nBy default,\nthe output file is named\n.BR luac.out ,\nbut you can change this with the\n.B \\-o\noption.\n.LP\nIn the command line,\nyou can mix\ntext files containing Lua source and\nbinary files containing precompiled chunks.\nThis is useful to combine several precompiled chunks,\neven from different (but compatible) platforms,\ninto a single precompiled chunk.\n.LP\nYou can use\n.B \"'\\-'\"\nto indicate the standard input as a source file\nand\n.B \"'\\--'\"\nto signal the end of options\n(that is,\nall remaining arguments will be treated as files even if they start with\n.BR \"'\\-'\" ).\n.LP\nThe internal format of the binary files produced by\n.B luac\nis likely to change when a new version of Lua is released.\nSo,\nsave the source files of all Lua programs that you precompile.\n.LP\n.SH OPTIONS\nOptions must be separate.\n.TP\n.B \\-l\nproduce a listing of the compiled bytecode for Lua's virtual machine.\nListing bytecodes is useful to learn about Lua's virtual machine.\nIf no files are given, then\n.B luac\nloads\n.B luac.out\nand lists its contents.\n.TP\n.BI \\-o \" file\"\noutput to\n.IR file ,\ninstead of the default\n.BR luac.out .\n(You can use\n.B \"'\\-'\"\nfor standard output,\nbut not on platforms that open standard output in text mode.)\nThe output file may be a source file because\nall files are loaded before the output file is written.\nBe careful not to overwrite precious files.\n.TP\n.B \\-p\nload files but do not generate any output file.\nUsed mainly for syntax checking and for testing precompiled chunks:\ncorrupted files will probably generate errors when loaded.\nLua always performs a thorough integrity test on precompiled chunks.\nBytecode that passes this test is completely safe,\nin the sense that it will not break the interpreter.\nHowever,\nthere is no guarantee that such code does anything sensible.\n(None can be given, because the halting problem is unsolvable.)\nIf no files are given, then\n.B luac\nloads\n.B luac.out\nand tests its contents.\nNo messages are displayed if the file passes the integrity test.\n.TP\n.B \\-s\nstrip debug information before writing the output file.\nThis saves some space in very large chunks,\nbut if errors occur when running a stripped chunk,\nthen the error messages may not contain the full information they usually do.\nFor instance,\nline numbers and names of local variables are lost.\n.TP\n.B \\-v\nshow version information.\n.SH FILES\n.TP 15\n.B luac.out\ndefault output file\n.SH \"SEE ALSO\"\n.BR lua (1)\n.br\nhttp://www.lua.org/\n.SH DIAGNOSTICS\nError messages should be self explanatory.\n.SH AUTHORS\nL. H. de Figueiredo,\nR. Ierusalimschy and\nW. Celes\n.\\\" EOF\n" }, { "path": "deps/lua/doc/luac.html", "content": "\n\n\nLUAC man page\n\n\n\n\n\n

NAME

\nluac - Lua compiler\n

SYNOPSIS

\nluac\n[\noptions\n] [\nfilenames\n]\n

DESCRIPTION

\nluac\nis the Lua compiler.\nIt translates programs written in the Lua programming language\ninto binary files that can be later loaded and executed.\n

\nThe main advantages of precompiling chunks are:\nfaster loading,\nprotecting source code from accidental user changes,\nand\noff-line syntax checking.\n

\nPrecompiling does not imply faster execution\nbecause in Lua chunks are always compiled into bytecodes before being executed.\nluac\nsimply allows those bytecodes to be saved in a file for later execution.\n

\nPrecompiled chunks are not necessarily smaller than the corresponding source.\nThe main goal in precompiling is faster loading.\n

\nThe binary files created by\nluac\nare portable only among architectures with the same word size and byte order.\n

\nluac\nproduces a single output file containing the bytecodes\nfor all source files given.\nBy default,\nthe output file is named\nluac.out,\nbut you can change this with the\n-o\noption.\n

\nIn the command line,\nyou can mix\ntext files containing Lua source and\nbinary files containing precompiled chunks.\nThis is useful because several precompiled chunks,\neven from different (but compatible) platforms,\ncan be combined into a single precompiled chunk.\n

\nYou can use\n'-'\nto indicate the standard input as a source file\nand\n'--'\nto signal the end of options\n(that is,\nall remaining arguments will be treated as files even if they start with\n'-').\n

\nThe internal format of the binary files produced by\nluac\nis likely to change when a new version of Lua is released.\nSo,\nsave the source files of all Lua programs that you precompile.\n

\n

OPTIONS

\nOptions must be separate.\n

\n-l\nproduce a listing of the compiled bytecode for Lua's virtual machine.\nListing bytecodes is useful to learn about Lua's virtual machine.\nIf no files are given, then\nluac\nloads\nluac.out\nand lists its contents.\n

\n-o file\noutput to\nfile,\ninstead of the default\nluac.out.\n(You can use\n'-'\nfor standard output,\nbut not on platforms that open standard output in text mode.)\nThe output file may be a source file because\nall files are loaded before the output file is written.\nBe careful not to overwrite precious files.\n

\n-p\nload files but do not generate any output file.\nUsed mainly for syntax checking and for testing precompiled chunks:\ncorrupted files will probably generate errors when loaded.\nLua always performs a thorough integrity test on precompiled chunks.\nBytecode that passes this test is completely safe,\nin the sense that it will not break the interpreter.\nHowever,\nthere is no guarantee that such code does anything sensible.\n(None can be given, because the halting problem is unsolvable.)\nIf no files are given, then\nluac\nloads\nluac.out\nand tests its contents.\nNo messages are displayed if the file passes the integrity test.\n

\n-s\nstrip debug information before writing the output file.\nThis saves some space in very large chunks,\nbut if errors occur when running a stripped chunk,\nthen the error messages may not contain the full information they usually do.\nFor instance,\nline numbers and names of local variables are lost.\n

\n-v\nshow version information.\n

FILES

\n

\nluac.out\ndefault output file\n

SEE ALSO

\nlua(1)\n
\nhttp://www.lua.org/\n

DIAGNOSTICS

\nError messages should be self explanatory.\n

AUTHORS

\nL. H. de Figueiredo,\nR. Ierusalimschy and\nW. Celes\n\n\n\n" }, { "path": "deps/lua/doc/manual.css", "content": "h3 code {\n\tfont-family: inherit ;\n\tfont-size: inherit ;\n}\n\npre, code {\n\tfont-size: 12pt ;\n}\n\nspan.apii {\n\tfloat: right ;\n\tfont-family: inherit ;\n\tfont-style: normal ;\n\tfont-size: small ;\n\tcolor: gray ;\n}\n\np+h1, ul+h1 {\n\tpadding-top: 0.4em ;\n\tpadding-bottom: 0.4em ;\n\tpadding-left: 30px ;\n\tmargin-left: -30px ;\n\tbackground-color: #E0E0FF ;\n}\n" }, { "path": "deps/lua/doc/manual.html", "content": "\n\n\n\nLua 5.1 Reference Manual\n\n\n\n\n\n\n\n
\n

\n\"\"\nLua 5.1 Reference Manual\n

\n\nby Roberto Ierusalimschy, Luiz Henrique de Figueiredo, Waldemar Celes\n

\n\nCopyright © 2006–2012 Lua.org, PUC-Rio.\nFreely available under the terms of the\nLua license.\n\n

\n

\n\ncontents\n·\nindex\n·\nother versions\n\n\n

\n\n\n\n\n\n\n

1 - Introduction

\n\n

\nLua is an extension programming language designed to support\ngeneral procedural programming with data description\nfacilities.\nIt also offers good support for object-oriented programming,\nfunctional programming, and data-driven programming.\nLua is intended to be used as a powerful, light-weight\nscripting language for any program that needs one.\nLua is implemented as a library, written in clean C\n(that is, in the common subset of ANSI C and C++).\n\n\n

\nBeing an extension language, Lua has no notion of a \"main\" program:\nit only works embedded in a host client,\ncalled the embedding program or simply the host.\nThis host program can invoke functions to execute a piece of Lua code,\ncan write and read Lua variables,\nand can register C functions to be called by Lua code.\nThrough the use of C functions, Lua can be augmented to cope with\na wide range of different domains,\nthus creating customized programming languages sharing a syntactical framework.\nThe Lua distribution includes a sample host program called lua,\nwhich uses the Lua library to offer a complete, stand-alone Lua interpreter.\n\n\n

\nLua is free software,\nand is provided as usual with no guarantees,\nas stated in its license.\nThe implementation described in this manual is available\nat Lua's official web site, www.lua.org.\n\n\n

\nLike any other reference manual,\nthis document is dry in places.\nFor a discussion of the decisions behind the design of Lua,\nsee the technical papers available at Lua's web site.\nFor a detailed introduction to programming in Lua,\nsee Roberto's book, Programming in Lua (Second Edition).\n\n\n\n

2 - The Language

\n\n

\nThis section describes the lexis, the syntax, and the semantics of Lua.\nIn other words,\nthis section describes\nwhich tokens are valid,\nhow they can be combined,\nand what their combinations mean.\n\n\n

\nThe language constructs will be explained using the usual extended BNF notation,\nin which\n{a} means 0 or more a's, and\n[a] means an optional a.\nNon-terminals are shown like non-terminal,\nkeywords are shown like kword,\nand other terminal symbols are shown like `=´.\nThe complete syntax of Lua can be found in §8\nat the end of this manual.\n\n\n\n

2.1 - Lexical Conventions

\n\n

\nNames\n(also called identifiers)\nin Lua can be any string of letters,\ndigits, and underscores,\nnot beginning with a digit.\nThis coincides with the definition of names in most languages.\n(The definition of letter depends on the current locale:\nany character considered alphabetic by the current locale\ncan be used in an identifier.)\nIdentifiers are used to name variables and table fields.\n\n\n

\nThe following keywords are reserved\nand cannot be used as names:\n\n\n

\n     and       break     do        else      elseif\n     end       false     for       function  if\n     in        local     nil       not       or\n     repeat    return    then      true      until     while\n
\n\n

\nLua is a case-sensitive language:\nand is a reserved word, but And and AND\nare two different, valid names.\nAs a convention, names starting with an underscore followed by\nuppercase letters (such as _VERSION)\nare reserved for internal global variables used by Lua.\n\n\n

\nThe following strings denote other tokens:\n\n

\n     +     -     *     /     %     ^     #\n     ==    ~=    <=    >=    <     >     =\n     (     )     {     }     [     ]\n     ;     :     ,     .     ..    ...\n
\n\n

\nLiteral strings\ncan be delimited by matching single or double quotes,\nand can contain the following C-like escape sequences:\n'\\a' (bell),\n'\\b' (backspace),\n'\\f' (form feed),\n'\\n' (newline),\n'\\r' (carriage return),\n'\\t' (horizontal tab),\n'\\v' (vertical tab),\n'\\\\' (backslash),\n'\\\"' (quotation mark [double quote]),\nand '\\'' (apostrophe [single quote]).\nMoreover, a backslash followed by a real newline\nresults in a newline in the string.\nA character in a string can also be specified by its numerical value\nusing the escape sequence \\ddd,\nwhere ddd is a sequence of up to three decimal digits.\n(Note that if a numerical escape is to be followed by a digit,\nit must be expressed using exactly three digits.)\nStrings in Lua can contain any 8-bit value, including embedded zeros,\nwhich can be specified as '\\0'.\n\n\n

\nLiteral strings can also be defined using a long format\nenclosed by long brackets.\nWe define an opening long bracket of level n as an opening\nsquare bracket followed by n equal signs followed by another\nopening square bracket.\nSo, an opening long bracket of level 0 is written as [[,\nan opening long bracket of level 1 is written as [=[,\nand so on.\nA closing long bracket is defined similarly;\nfor instance, a closing long bracket of level 4 is written as ]====].\nA long string starts with an opening long bracket of any level and\nends at the first closing long bracket of the same level.\nLiterals in this bracketed form can run for several lines,\ndo not interpret any escape sequences,\nand ignore long brackets of any other level.\nThey can contain anything except a closing bracket of the proper level.\n\n\n

\nFor convenience,\nwhen the opening long bracket is immediately followed by a newline,\nthe newline is not included in the string.\nAs an example, in a system using ASCII\n(in which 'a' is coded as 97,\nnewline is coded as 10, and '1' is coded as 49),\nthe five literal strings below denote the same string:\n\n

\n     a = 'alo\\n123\"'\n     a = \"alo\\n123\\\"\"\n     a = '\\97lo\\10\\04923\"'\n     a = [[alo\n     123\"]]\n     a = [==[\n     alo\n     123\"]==]\n
\n\n

\nA numerical constant can be written with an optional decimal part\nand an optional decimal exponent.\nLua also accepts integer hexadecimal constants,\nby prefixing them with 0x.\nExamples of valid numerical constants are\n\n

\n     3   3.0   3.1416   314.16e-2   0.31416E1   0xff   0x56\n
\n\n

\nA comment starts with a double hyphen (--)\nanywhere outside a string.\nIf the text immediately after -- is not an opening long bracket,\nthe comment is a short comment,\nwhich runs until the end of the line.\nOtherwise, it is a long comment,\nwhich runs until the corresponding closing long bracket.\nLong comments are frequently used to disable code temporarily.\n\n\n\n\n\n

2.2 - Values and Types

\n\n

\nLua is a dynamically typed language.\nThis means that\nvariables do not have types; only values do.\nThere are no type definitions in the language.\nAll values carry their own type.\n\n\n

\nAll values in Lua are first-class values.\nThis means that all values can be stored in variables,\npassed as arguments to other functions, and returned as results.\n\n\n

\nThere are eight basic types in Lua:\nnil, boolean, number,\nstring, function, userdata,\nthread, and table.\nNil is the type of the value nil,\nwhose main property is to be different from any other value;\nit usually represents the absence of a useful value.\nBoolean is the type of the values false and true.\nBoth nil and false make a condition false;\nany other value makes it true.\nNumber represents real (double-precision floating-point) numbers.\n(It is easy to build Lua interpreters that use other\ninternal representations for numbers,\nsuch as single-precision float or long integers;\nsee file luaconf.h.)\nString represents arrays of characters.\n\nLua is 8-bit clean:\nstrings can contain any 8-bit character,\nincluding embedded zeros ('\\0') (see §2.1).\n\n\n

\nLua can call (and manipulate) functions written in Lua and\nfunctions written in C\n(see §2.5.8).\n\n\n

\nThe type userdata is provided to allow arbitrary C data to\nbe stored in Lua variables.\nThis type corresponds to a block of raw memory\nand has no pre-defined operations in Lua,\nexcept assignment and identity test.\nHowever, by using metatables,\nthe programmer can define operations for userdata values\n(see §2.8).\nUserdata values cannot be created or modified in Lua,\nonly through the C API.\nThis guarantees the integrity of data owned by the host program.\n\n\n

\nThe type thread represents independent threads of execution\nand it is used to implement coroutines (see §2.11).\nDo not confuse Lua threads with operating-system threads.\nLua supports coroutines on all systems,\neven those that do not support threads.\n\n\n

\nThe type table implements associative arrays,\nthat is, arrays that can be indexed not only with numbers,\nbut with any value (except nil).\nTables can be heterogeneous;\nthat is, they can contain values of all types (except nil).\nTables are the sole data structuring mechanism in Lua;\nthey can be used to represent ordinary arrays,\nsymbol tables, sets, records, graphs, trees, etc.\nTo represent records, Lua uses the field name as an index.\nThe language supports this representation by\nproviding a.name as syntactic sugar for a[\"name\"].\nThere are several convenient ways to create tables in Lua\n(see §2.5.7).\n\n\n

\nLike indices,\nthe value of a table field can be of any type (except nil).\nIn particular,\nbecause functions are first-class values,\ntable fields can contain functions.\nThus tables can also carry methods (see §2.5.9).\n\n\n

\nTables, functions, threads, and (full) userdata values are objects:\nvariables do not actually contain these values,\nonly references to them.\nAssignment, parameter passing, and function returns\nalways manipulate references to such values;\nthese operations do not imply any kind of copy.\n\n\n

\nThe library function type returns a string describing the type\nof a given value.\n\n\n\n

2.2.1 - Coercion

\n\n

\nLua provides automatic conversion between\nstring and number values at run time.\nAny arithmetic operation applied to a string tries to convert\nthis string to a number, following the usual conversion rules.\nConversely, whenever a number is used where a string is expected,\nthe number is converted to a string, in a reasonable format.\nFor complete control over how numbers are converted to strings,\nuse the format function from the string library\n(see string.format).\n\n\n\n\n\n\n\n

2.3 - Variables

\n\n

\nVariables are places that store values.\n\nThere are three kinds of variables in Lua:\nglobal variables, local variables, and table fields.\n\n\n

\nA single name can denote a global variable or a local variable\n(or a function's formal parameter,\nwhich is a particular kind of local variable):\n\n

\n\tvar ::= Name\n

\nName denotes identifiers, as defined in §2.1.\n\n\n

\nAny variable is assumed to be global unless explicitly declared\nas a local (see §2.4.7).\nLocal variables are lexically scoped:\nlocal variables can be freely accessed by functions\ndefined inside their scope (see §2.6).\n\n\n

\nBefore the first assignment to a variable, its value is nil.\n\n\n

\nSquare brackets are used to index a table:\n\n

\n\tvar ::= prefixexp `[´ exp `]´\n

\nThe meaning of accesses to global variables \nand table fields can be changed via metatables.\nAn access to an indexed variable t[i] is equivalent to\na call gettable_event(t,i).\n(See §2.8 for a complete description of the\ngettable_event function.\nThis function is not defined or callable in Lua.\nWe use it here only for explanatory purposes.)\n\n\n

\nThe syntax var.Name is just syntactic sugar for\nvar[\"Name\"]:\n\n

\n\tvar ::= prefixexp `.´ Name\n
\n\n

\nAll global variables live as fields in ordinary Lua tables,\ncalled environment tables or simply\nenvironments (see §2.9).\nEach function has its own reference to an environment,\nso that all global variables in this function\nwill refer to this environment table.\nWhen a function is created,\nit inherits the environment from the function that created it.\nTo get the environment table of a Lua function,\nyou call getfenv.\nTo replace it,\nyou call setfenv.\n(You can only manipulate the environment of C functions\nthrough the debug library; (see §5.9).)\n\n\n

\nAn access to a global variable x\nis equivalent to _env.x,\nwhich in turn is equivalent to\n\n

\n     gettable_event(_env, \"x\")\n

\nwhere _env is the environment of the running function.\n(See §2.8 for a complete description of the\ngettable_event function.\nThis function is not defined or callable in Lua.\nSimilarly, the _env variable is not defined in Lua.\nWe use them here only for explanatory purposes.)\n\n\n\n\n\n

2.4 - Statements

\n\n

\nLua supports an almost conventional set of statements,\nsimilar to those in Pascal or C.\nThis set includes\nassignments, control structures, function calls,\nand variable declarations.\n\n\n\n

2.4.1 - Chunks

\n\n

\nThe unit of execution of Lua is called a chunk.\nA chunk is simply a sequence of statements,\nwhich are executed sequentially.\nEach statement can be optionally followed by a semicolon:\n\n

\n\tchunk ::= {stat [`;´]}\n

\nThere are no empty statements and thus ';;' is not legal.\n\n\n

\nLua handles a chunk as the body of an anonymous function \nwith a variable number of arguments\n(see §2.5.9).\nAs such, chunks can define local variables,\nreceive arguments, and return values.\n\n\n

\nA chunk can be stored in a file or in a string inside the host program.\nTo execute a chunk,\nLua first pre-compiles the chunk into instructions for a virtual machine,\nand then it executes the compiled code\nwith an interpreter for the virtual machine.\n\n\n

\nChunks can also be pre-compiled into binary form;\nsee program luac for details.\nPrograms in source and compiled forms are interchangeable;\nLua automatically detects the file type and acts accordingly.\n\n\n\n\n\n\n

2.4.2 - Blocks

\nA block is a list of statements;\nsyntactically, a block is the same as a chunk:\n\n

\n\tblock ::= chunk\n
\n\n

\nA block can be explicitly delimited to produce a single statement:\n\n

\n\tstat ::= do block end\n

\nExplicit blocks are useful\nto control the scope of variable declarations.\nExplicit blocks are also sometimes used to\nadd a return or break statement in the middle\nof another block (see §2.4.4).\n\n\n\n\n\n

2.4.3 - Assignment

\n\n

\nLua allows multiple assignments.\nTherefore, the syntax for assignment\ndefines a list of variables on the left side\nand a list of expressions on the right side.\nThe elements in both lists are separated by commas:\n\n

\n\tstat ::= varlist `=´ explist\n\tvarlist ::= var {`,´ var}\n\texplist ::= exp {`,´ exp}\n

\nExpressions are discussed in §2.5.\n\n\n

\nBefore the assignment,\nthe list of values is adjusted to the length of\nthe list of variables.\nIf there are more values than needed,\nthe excess values are thrown away.\nIf there are fewer values than needed,\nthe list is extended with as many nil's as needed.\nIf the list of expressions ends with a function call,\nthen all values returned by that call enter the list of values,\nbefore the adjustment\n(except when the call is enclosed in parentheses; see §2.5).\n\n\n

\nThe assignment statement first evaluates all its expressions\nand only then are the assignments performed.\nThus the code\n\n

\n     i = 3\n     i, a[i] = i+1, 20\n

\nsets a[3] to 20, without affecting a[4]\nbecause the i in a[i] is evaluated (to 3)\nbefore it is assigned 4.\nSimilarly, the line\n\n

\n     x, y = y, x\n

\nexchanges the values of x and y,\nand\n\n

\n     x, y, z = y, z, x\n

\ncyclically permutes the values of x, y, and z.\n\n\n

\nThe meaning of assignments to global variables\nand table fields can be changed via metatables.\nAn assignment to an indexed variable t[i] = val is equivalent to\nsettable_event(t,i,val).\n(See §2.8 for a complete description of the\nsettable_event function.\nThis function is not defined or callable in Lua.\nWe use it here only for explanatory purposes.)\n\n\n

\nAn assignment to a global variable x = val\nis equivalent to the assignment\n_env.x = val,\nwhich in turn is equivalent to\n\n

\n     settable_event(_env, \"x\", val)\n

\nwhere _env is the environment of the running function.\n(The _env variable is not defined in Lua.\nWe use it here only for explanatory purposes.)\n\n\n\n\n\n

2.4.4 - Control Structures

\nThe control structures\nif, while, and repeat have the usual meaning and\nfamiliar syntax:\n\n\n\n\n

\n\tstat ::= while exp do block end\n\tstat ::= repeat block until exp\n\tstat ::= if exp then block {elseif exp then block} [else block] end\n

\nLua also has a for statement, in two flavors (see §2.4.5).\n\n\n

\nThe condition expression of a\ncontrol structure can return any value.\nBoth false and nil are considered false.\nAll values different from nil and false are considered true\n(in particular, the number 0 and the empty string are also true).\n\n\n

\nIn the repeat–until loop,\nthe inner block does not end at the until keyword,\nbut only after the condition.\nSo, the condition can refer to local variables\ndeclared inside the loop block.\n\n\n

\nThe return statement is used to return values\nfrom a function or a chunk (which is just a function).\n\nFunctions and chunks can return more than one value,\nand so the syntax for the return statement is\n\n

\n\tstat ::= return [explist]\n
\n\n

\nThe break statement is used to terminate the execution of a\nwhile, repeat, or for loop,\nskipping to the next statement after the loop:\n\n\n

\n\tstat ::= break\n

\nA break ends the innermost enclosing loop.\n\n\n

\nThe return and break\nstatements can only be written as the last statement of a block.\nIf it is really necessary to return or break in the\nmiddle of a block,\nthen an explicit inner block can be used,\nas in the idioms\ndo return end and do break end,\nbecause now return and break are the last statements in\ntheir (inner) blocks.\n\n\n\n\n\n

2.4.5 - For Statement

\n\n

\n\nThe for statement has two forms:\none numeric and one generic.\n\n\n

\nThe numeric for loop repeats a block of code while a\ncontrol variable runs through an arithmetic progression.\nIt has the following syntax:\n\n

\n\tstat ::= for Name `=´ exp `,´ exp [`,´ exp] do block end\n

\nThe block is repeated for name starting at the value of\nthe first exp, until it passes the second exp by steps of the\nthird exp.\nMore precisely, a for statement like\n\n

\n     for v = e1, e2, e3 do block end\n

\nis equivalent to the code:\n\n

\n     do\n       local var, limit, step = tonumber(e1), tonumber(e2), tonumber(e3)\n       if not (var and limit and step) then error() end\n       while (step > 0 and var <= limit) or (step <= 0 and var >= limit) do\n         local v = var\n         block\n         var = var + step\n       end\n     end\n

\nNote the following:\n\n

    \n\n
  • \nAll three control expressions are evaluated only once,\nbefore the loop starts.\nThey must all result in numbers.\n
    • \n\n
  • \nvar, limit, and step are invisible variables.\nThe names shown here are for explanatory purposes only.\n
    • \n\n
  • \nIf the third expression (the step) is absent,\nthen a step of 1 is used.\n
    • \n\n
  • \nYou can use break to exit a for loop.\n
    • \n\n
  • \nThe loop variable v is local to the loop;\nyou cannot use its value after the for ends or is broken.\nIf you need this value,\nassign it to another variable before breaking or exiting the loop.\n
    • \n\n
\n\n

\nThe generic for statement works over functions,\ncalled iterators.\nOn each iteration, the iterator function is called to produce a new value,\nstopping when this new value is nil.\nThe generic for loop has the following syntax:\n\n

\n\tstat ::= for namelist in explist do block end\n\tnamelist ::= Name {`,´ Name}\n

\nA for statement like\n\n

\n     for var_1, ···, var_n in explist do block end\n

\nis equivalent to the code:\n\n

\n     do\n       local f, s, var = explist\n       while true do\n         local var_1, ···, var_n = f(s, var)\n         var = var_1\n         if var == nil then break end\n         block\n       end\n     end\n

\nNote the following:\n\n

    \n\n
  • \nexplist is evaluated only once.\nIts results are an iterator function,\na state,\nand an initial value for the first iterator variable.\n
    • \n\n
  • \nf, s, and var are invisible variables.\nThe names are here for explanatory purposes only.\n
    • \n\n
  • \nYou can use break to exit a for loop.\n
    • \n\n
  • \nThe loop variables var_i are local to the loop;\nyou cannot use their values after the for ends.\nIf you need these values,\nthen assign them to other variables before breaking or exiting the loop.\n
    • \n\n
\n\n\n\n\n

2.4.6 - Function Calls as Statements

\nTo allow possible side-effects,\nfunction calls can be executed as statements:\n\n

\n\tstat ::= functioncall\n

\nIn this case, all returned values are thrown away.\nFunction calls are explained in §2.5.8.\n\n\n\n\n\n

2.4.7 - Local Declarations

\nLocal variables can be declared anywhere inside a block.\nThe declaration can include an initial assignment:\n\n

\n\tstat ::= local namelist [`=´ explist]\n

\nIf present, an initial assignment has the same semantics\nof a multiple assignment (see §2.4.3).\nOtherwise, all variables are initialized with nil.\n\n\n

\nA chunk is also a block (see §2.4.1),\nand so local variables can be declared in a chunk outside any explicit block.\nThe scope of such local variables extends until the end of the chunk.\n\n\n

\nThe visibility rules for local variables are explained in §2.6.\n\n\n\n\n\n\n\n

2.5 - Expressions

\n\n

\nThe basic expressions in Lua are the following:\n\n

\n\texp ::= prefixexp\n\texp ::= nil | false | true\n\texp ::= Number\n\texp ::= String\n\texp ::= function\n\texp ::= tableconstructor\n\texp ::= `...´\n\texp ::= exp binop exp\n\texp ::= unop exp\n\tprefixexp ::= var | functioncall | `(´ exp `)´\n
\n\n

\nNumbers and literal strings are explained in §2.1;\nvariables are explained in §2.3;\nfunction definitions are explained in §2.5.9;\nfunction calls are explained in §2.5.8;\ntable constructors are explained in §2.5.7.\nVararg expressions,\ndenoted by three dots ('...'), can only be used when\ndirectly inside a vararg function;\nthey are explained in §2.5.9.\n\n\n

\nBinary operators comprise arithmetic operators (see §2.5.1),\nrelational operators (see §2.5.2), logical operators (see §2.5.3),\nand the concatenation operator (see §2.5.4).\nUnary operators comprise the unary minus (see §2.5.1),\nthe unary not (see §2.5.3),\nand the unary length operator (see §2.5.5).\n\n\n

\nBoth function calls and vararg expressions can result in multiple values.\nIf an expression is used as a statement\n(only possible for function calls (see §2.4.6)),\nthen its return list is adjusted to zero elements,\nthus discarding all returned values.\nIf an expression is used as the last (or the only) element\nof a list of expressions,\nthen no adjustment is made\n(unless the call is enclosed in parentheses).\nIn all other contexts,\nLua adjusts the result list to one element,\ndiscarding all values except the first one.\n\n\n

\nHere are some examples:\n\n

\n     f()                -- adjusted to 0 results\n     g(f(), x)          -- f() is adjusted to 1 result\n     g(x, f())          -- g gets x plus all results from f()\n     a,b,c = f(), x     -- f() is adjusted to 1 result (c gets nil)\n     a,b = ...          -- a gets the first vararg parameter, b gets\n                        -- the second (both a and b can get nil if there\n                        -- is no corresponding vararg parameter)\n     \n     a,b,c = x, f()     -- f() is adjusted to 2 results\n     a,b,c = f()        -- f() is adjusted to 3 results\n     return f()         -- returns all results from f()\n     return ...         -- returns all received vararg parameters\n     return x,y,f()     -- returns x, y, and all results from f()\n     {f()}              -- creates a list with all results from f()\n     {...}              -- creates a list with all vararg parameters\n     {f(), nil}         -- f() is adjusted to 1 result\n
\n\n

\nAny expression enclosed in parentheses always results in only one value.\nThus,\n(f(x,y,z)) is always a single value,\neven if f returns several values.\n(The value of (f(x,y,z)) is the first value returned by f\nor nil if f does not return any values.)\n\n\n\n

2.5.1 - Arithmetic Operators

\nLua supports the usual arithmetic operators:\nthe binary + (addition),\n- (subtraction), * (multiplication),\n/ (division), % (modulo), and ^ (exponentiation);\nand unary - (negation).\nIf the operands are numbers, or strings that can be converted to\nnumbers (see §2.2.1),\nthen all operations have the usual meaning.\nExponentiation works for any exponent.\nFor instance, x^(-0.5) computes the inverse of the square root of x.\nModulo is defined as\n\n

\n     a % b == a - math.floor(a/b)*b\n

\nThat is, it is the remainder of a division that rounds\nthe quotient towards minus infinity.\n\n\n\n\n\n

2.5.2 - Relational Operators

\nThe relational operators in Lua are\n\n

\n     ==    ~=    <     >     <=    >=\n

\nThese operators always result in false or true.\n\n\n

\nEquality (==) first compares the type of its operands.\nIf the types are different, then the result is false.\nOtherwise, the values of the operands are compared.\nNumbers and strings are compared in the usual way.\nObjects (tables, userdata, threads, and functions)\nare compared by reference:\ntwo objects are considered equal only if they are the same object.\nEvery time you create a new object\n(a table, userdata, thread, or function),\nthis new object is different from any previously existing object.\n\n\n

\nYou can change the way that Lua compares tables and userdata \nby using the \"eq\" metamethod (see §2.8).\n\n\n

\nThe conversion rules of §2.2.1\ndo not apply to equality comparisons.\nThus, \"0\"==0 evaluates to false,\nand t[0] and t[\"0\"] denote different\nentries in a table.\n\n\n

\nThe operator ~= is exactly the negation of equality (==).\n\n\n

\nThe order operators work as follows.\nIf both arguments are numbers, then they are compared as such.\nOtherwise, if both arguments are strings,\nthen their values are compared according to the current locale.\nOtherwise, Lua tries to call the \"lt\" or the \"le\"\nmetamethod (see §2.8).\nA comparison a > b is translated to b < a\nand a >= b is translated to b <= a.\n\n\n\n\n\n

2.5.3 - Logical Operators

\nThe logical operators in Lua are\nand, or, and not.\nLike the control structures (see §2.4.4),\nall logical operators consider both false and nil as false\nand anything else as true.\n\n\n

\nThe negation operator not always returns false or true.\nThe conjunction operator and returns its first argument\nif this value is false or nil;\notherwise, and returns its second argument.\nThe disjunction operator or returns its first argument\nif this value is different from nil and false;\notherwise, or returns its second argument.\nBoth and and or use short-cut evaluation;\nthat is,\nthe second operand is evaluated only if necessary.\nHere are some examples:\n\n

\n     10 or 20            --> 10\n     10 or error()       --> 10\n     nil or \"a\"          --> \"a\"\n     nil and 10          --> nil\n     false and error()   --> false\n     false and nil       --> false\n     false or nil        --> nil\n     10 and 20           --> 20\n

\n(In this manual,\n--> indicates the result of the preceding expression.)\n\n\n\n\n\n

2.5.4 - Concatenation

\nThe string concatenation operator in Lua is\ndenoted by two dots ('..').\nIf both operands are strings or numbers, then they are converted to\nstrings according to the rules mentioned in §2.2.1.\nOtherwise, the \"concat\" metamethod is called (see §2.8).\n\n\n\n\n\n

2.5.5 - The Length Operator

\n\n

\nThe length operator is denoted by the unary operator #.\nThe length of a string is its number of bytes\n(that is, the usual meaning of string length when each\ncharacter is one byte).\n\n\n

\nThe length of a table t is defined to be any\ninteger index n\nsuch that t[n] is not nil and t[n+1] is nil;\nmoreover, if t[1] is nil, n can be zero.\nFor a regular array, with non-nil values from 1 to a given n,\nits length is exactly that n,\nthe index of its last value.\nIf the array has \"holes\"\n(that is, nil values between other non-nil values),\nthen #t can be any of the indices that\ndirectly precedes a nil value\n(that is, it may consider any such nil value as the end of\nthe array). \n\n\n\n\n\n

2.5.6 - Precedence

\nOperator precedence in Lua follows the table below,\nfrom lower to higher priority:\n\n

\n     or\n     and\n     <     >     <=    >=    ~=    ==\n     ..\n     +     -\n     *     /     %\n     not   #     - (unary)\n     ^\n

\nAs usual,\nyou can use parentheses to change the precedences of an expression.\nThe concatenation ('..') and exponentiation ('^')\noperators are right associative.\nAll other binary operators are left associative.\n\n\n\n\n\n

2.5.7 - Table Constructors

\nTable constructors are expressions that create tables.\nEvery time a constructor is evaluated, a new table is created.\nA constructor can be used to create an empty table\nor to create a table and initialize some of its fields.\nThe general syntax for constructors is\n\n

\n\ttableconstructor ::= `{´ [fieldlist] `}´\n\tfieldlist ::= field {fieldsep field} [fieldsep]\n\tfield ::= `[´ exp `]´ `=´ exp | Name `=´ exp | exp\n\tfieldsep ::= `,´ | `;´\n
\n\n

\nEach field of the form [exp1] = exp2 adds to the new table an entry\nwith key exp1 and value exp2.\nA field of the form name = exp is equivalent to\n[\"name\"] = exp.\nFinally, fields of the form exp are equivalent to\n[i] = exp, where i are consecutive numerical integers,\nstarting with 1.\nFields in the other formats do not affect this counting.\nFor example,\n\n

\n     a = { [f(1)] = g; \"x\", \"y\"; x = 1, f(x), [30] = 23; 45 }\n

\nis equivalent to\n\n

\n     do\n       local t = {}\n       t[f(1)] = g\n       t[1] = \"x\"         -- 1st exp\n       t[2] = \"y\"         -- 2nd exp\n       t.x = 1            -- t[\"x\"] = 1\n       t[3] = f(x)        -- 3rd exp\n       t[30] = 23\n       t[4] = 45          -- 4th exp\n       a = t\n     end\n
\n\n

\nIf the last field in the list has the form exp\nand the expression is a function call or a vararg expression,\nthen all values returned by this expression enter the list consecutively\n(see §2.5.8).\nTo avoid this,\nenclose the function call or the vararg expression\nin parentheses (see §2.5).\n\n\n

\nThe field list can have an optional trailing separator,\nas a convenience for machine-generated code.\n\n\n\n\n\n

2.5.8 - Function Calls

\nA function call in Lua has the following syntax:\n\n

\n\tfunctioncall ::= prefixexp args\n

\nIn a function call,\nfirst prefixexp and args are evaluated.\nIf the value of prefixexp has type function,\nthen this function is called\nwith the given arguments.\nOtherwise, the prefixexp \"call\" metamethod is called,\nhaving as first parameter the value of prefixexp,\nfollowed by the original call arguments\n(see §2.8).\n\n\n

\nThe form\n\n

\n\tfunctioncall ::= prefixexp `:´ Name args\n

\ncan be used to call \"methods\".\nA call v:name(args)\nis syntactic sugar for v.name(v,args),\nexcept that v is evaluated only once.\n\n\n

\nArguments have the following syntax:\n\n

\n\targs ::= `(´ [explist] `)´\n\targs ::= tableconstructor\n\targs ::= String\n

\nAll argument expressions are evaluated before the call.\nA call of the form f{fields} is\nsyntactic sugar for f({fields});\nthat is, the argument list is a single new table.\nA call of the form f'string'\n(or f\"string\" or f[[string]])\nis syntactic sugar for f('string');\nthat is, the argument list is a single literal string.\n\n\n

\nAs an exception to the free-format syntax of Lua,\nyou cannot put a line break before the '(' in a function call.\nThis restriction avoids some ambiguities in the language.\nIf you write\n\n

\n     a = f\n     (g).x(a)\n

\nLua would see that as a single statement, a = f(g).x(a).\nSo, if you want two statements, you must add a semi-colon between them.\nIf you actually want to call f,\nyou must remove the line break before (g).\n\n\n

\nA call of the form return functioncall is called\na tail call.\nLua implements proper tail calls\n(or proper tail recursion):\nin a tail call,\nthe called function reuses the stack entry of the calling function.\nTherefore, there is no limit on the number of nested tail calls that\na program can execute.\nHowever, a tail call erases any debug information about the\ncalling function.\nNote that a tail call only happens with a particular syntax,\nwhere the return has one single function call as argument;\nthis syntax makes the calling function return exactly\nthe returns of the called function.\nSo, none of the following examples are tail calls:\n\n

\n     return (f(x))        -- results adjusted to 1\n     return 2 * f(x)\n     return x, f(x)       -- additional results\n     f(x); return         -- results discarded\n     return x or f(x)     -- results adjusted to 1\n
\n\n\n\n\n

2.5.9 - Function Definitions

\n\n

\nThe syntax for function definition is\n\n

\n\tfunction ::= function funcbody\n\tfuncbody ::= `(´ [parlist] `)´ block end\n
\n\n

\nThe following syntactic sugar simplifies function definitions:\n\n

\n\tstat ::= function funcname funcbody\n\tstat ::= local function Name funcbody\n\tfuncname ::= Name {`.´ Name} [`:´ Name]\n

\nThe statement\n\n

\n     function f () body end\n

\ntranslates to\n\n

\n     f = function () body end\n

\nThe statement\n\n

\n     function t.a.b.c.f () body end\n

\ntranslates to\n\n

\n     t.a.b.c.f = function () body end\n

\nThe statement\n\n

\n     local function f () body end\n

\ntranslates to\n\n

\n     local f; f = function () body end\n

\nnot to\n\n

\n     local f = function () body end\n

\n(This only makes a difference when the body of the function\ncontains references to f.)\n\n\n

\nA function definition is an executable expression,\nwhose value has type function.\nWhen Lua pre-compiles a chunk,\nall its function bodies are pre-compiled too.\nThen, whenever Lua executes the function definition,\nthe function is instantiated (or closed).\nThis function instance (or closure)\nis the final value of the expression.\nDifferent instances of the same function\ncan refer to different external local variables\nand can have different environment tables.\n\n\n

\nParameters act as local variables that are\ninitialized with the argument values:\n\n

\n\tparlist ::= namelist [`,´ `...´] | `...´\n

\nWhen a function is called,\nthe list of arguments is adjusted to\nthe length of the list of parameters,\nunless the function is a variadic or vararg function,\nwhich is\nindicated by three dots ('...') at the end of its parameter list.\nA vararg function does not adjust its argument list;\ninstead, it collects all extra arguments and supplies them\nto the function through a vararg expression,\nwhich is also written as three dots.\nThe value of this expression is a list of all actual extra arguments,\nsimilar to a function with multiple results.\nIf a vararg expression is used inside another expression\nor in the middle of a list of expressions,\nthen its return list is adjusted to one element.\nIf the expression is used as the last element of a list of expressions,\nthen no adjustment is made\n(unless that last expression is enclosed in parentheses).\n\n\n

\nAs an example, consider the following definitions:\n\n

\n     function f(a, b) end\n     function g(a, b, ...) end\n     function r() return 1,2,3 end\n

\nThen, we have the following mapping from arguments to parameters and\nto the vararg expression:\n\n

\n     CALL            PARAMETERS\n     \n     f(3)             a=3, b=nil\n     f(3, 4)          a=3, b=4\n     f(3, 4, 5)       a=3, b=4\n     f(r(), 10)       a=1, b=10\n     f(r())           a=1, b=2\n     \n     g(3)             a=3, b=nil, ... -->  (nothing)\n     g(3, 4)          a=3, b=4,   ... -->  (nothing)\n     g(3, 4, 5, 8)    a=3, b=4,   ... -->  5  8\n     g(5, r())        a=5, b=1,   ... -->  2  3\n
\n\n

\nResults are returned using the return statement (see §2.4.4).\nIf control reaches the end of a function\nwithout encountering a return statement,\nthen the function returns with no results.\n\n\n

\nThe colon syntax\nis used for defining methods,\nthat is, functions that have an implicit extra parameter self.\nThus, the statement\n\n

\n     function t.a.b.c:f (params) body end\n

\nis syntactic sugar for\n\n

\n     t.a.b.c.f = function (self, params) body end\n
\n\n\n\n\n\n\n

2.6 - Visibility Rules

\n\n

\n\nLua is a lexically scoped language.\nThe scope of variables begins at the first statement after\ntheir declaration and lasts until the end of the innermost block that\nincludes the declaration.\nConsider the following example:\n\n

\n     x = 10                -- global variable\n     do                    -- new block\n       local x = x         -- new 'x', with value 10\n       print(x)            --> 10\n       x = x+1\n       do                  -- another block\n         local x = x+1     -- another 'x'\n         print(x)          --> 12\n       end\n       print(x)            --> 11\n     end\n     print(x)              --> 10  (the global one)\n
\n\n

\nNotice that, in a declaration like local x = x,\nthe new x being declared is not in scope yet,\nand so the second x refers to the outside variable.\n\n\n

\nBecause of the lexical scoping rules,\nlocal variables can be freely accessed by functions\ndefined inside their scope.\nA local variable used by an inner function is called\nan upvalue, or external local variable,\ninside the inner function.\n\n\n

\nNotice that each execution of a local statement\ndefines new local variables.\nConsider the following example:\n\n

\n     a = {}\n     local x = 20\n     for i=1,10 do\n       local y = 0\n       a[i] = function () y=y+1; return x+y end\n     end\n

\nThe loop creates ten closures\n(that is, ten instances of the anonymous function).\nEach of these closures uses a different y variable,\nwhile all of them share the same x.\n\n\n\n\n\n

2.7 - Error Handling

\n\n

\nBecause Lua is an embedded extension language,\nall Lua actions start from C code in the host program\ncalling a function from the Lua library (see lua_pcall).\nWhenever an error occurs during Lua compilation or execution,\ncontrol returns to C,\nwhich can take appropriate measures\n(such as printing an error message).\n\n\n

\nLua code can explicitly generate an error by calling the\nerror function.\nIf you need to catch errors in Lua,\nyou can use the pcall function.\n\n\n\n\n\n

2.8 - Metatables

\n\n

\nEvery value in Lua can have a metatable.\nThis metatable is an ordinary Lua table\nthat defines the behavior of the original value\nunder certain special operations.\nYou can change several aspects of the behavior\nof operations over a value by setting specific fields in its metatable.\nFor instance, when a non-numeric value is the operand of an addition,\nLua checks for a function in the field \"__add\" in its metatable.\nIf it finds one,\nLua calls this function to perform the addition.\n\n\n

\nWe call the keys in a metatable events\nand the values metamethods.\nIn the previous example, the event is \"add\" \nand the metamethod is the function that performs the addition.\n\n\n

\nYou can query the metatable of any value\nthrough the getmetatable function.\n\n\n

\nYou can replace the metatable of tables\nthrough the setmetatable\nfunction.\nYou cannot change the metatable of other types from Lua\n(except by using the debug library);\nyou must use the C API for that.\n\n\n

\nTables and full userdata have individual metatables\n(although multiple tables and userdata can share their metatables).\nValues of all other types share one single metatable per type;\nthat is, there is one single metatable for all numbers,\none for all strings, etc.\n\n\n

\nA metatable controls how an object behaves in arithmetic operations,\norder comparisons, concatenation, length operation, and indexing.\nA metatable also can define a function to be called when a userdata\nis garbage collected.\nFor each of these operations Lua associates a specific key\ncalled an event.\nWhen Lua performs one of these operations over a value,\nit checks whether this value has a metatable with the corresponding event.\nIf so, the value associated with that key (the metamethod)\ncontrols how Lua will perform the operation.\n\n\n

\nMetatables control the operations listed next.\nEach operation is identified by its corresponding name.\nThe key for each operation is a string with its name prefixed by\ntwo underscores, '__';\nfor instance, the key for operation \"add\" is the\nstring \"__add\".\nThe semantics of these operations is better explained by a Lua function\ndescribing how the interpreter executes the operation.\n\n\n

\nThe code shown here in Lua is only illustrative;\nthe real behavior is hard coded in the interpreter\nand it is much more efficient than this simulation.\nAll functions used in these descriptions\n(rawget, tonumber, etc.)\nare described in §5.1.\nIn particular, to retrieve the metamethod of a given object,\nwe use the expression\n\n

\n     metatable(obj)[event]\n

\nThis should be read as\n\n

\n     rawget(getmetatable(obj) or {}, event)\n

\n\nThat is, the access to a metamethod does not invoke other metamethods,\nand the access to objects with no metatables does not fail\n(it simply results in nil).\n\n\n\n

    \n\n
  • \"add\":\nthe + operation.\n\n\n\n

    \nThe function getbinhandler below defines how Lua chooses a handler\nfor a binary operation.\nFirst, Lua tries the first operand.\nIf its type does not define a handler for the operation,\nthen Lua tries the second operand.\n\n

    \n     function getbinhandler (op1, op2, event)\n       return metatable(op1)[event] or metatable(op2)[event]\n     end\n

    \nBy using this function,\nthe behavior of the op1 + op2 is\n\n

    \n     function add_event (op1, op2)\n       local o1, o2 = tonumber(op1), tonumber(op2)\n       if o1 and o2 then  -- both operands are numeric?\n         return o1 + o2   -- '+' here is the primitive 'add'\n       else  -- at least one of the operands is not numeric\n         local h = getbinhandler(op1, op2, \"__add\")\n         if h then\n           -- call the handler with both operands\n           return (h(op1, op2))\n         else  -- no handler available: default behavior\n           error(···)\n         end\n       end\n     end\n

    \n

    • \n\n
  • \"sub\":\nthe - operation.\n\nBehavior similar to the \"add\" operation.\n
    • \n\n
  • \"mul\":\nthe * operation.\n\nBehavior similar to the \"add\" operation.\n
    • \n\n
  • \"div\":\nthe / operation.\n\nBehavior similar to the \"add\" operation.\n
    • \n\n
  • \"mod\":\nthe % operation.\n\nBehavior similar to the \"add\" operation,\nwith the operation\no1 - floor(o1/o2)*o2 as the primitive operation.\n
    • \n\n
  • \"pow\":\nthe ^ (exponentiation) operation.\n\nBehavior similar to the \"add\" operation,\nwith the function pow (from the C math library)\nas the primitive operation.\n
    • \n\n
  • \"unm\":\nthe unary - operation.\n\n\n
    \n     function unm_event (op)\n       local o = tonumber(op)\n       if o then  -- operand is numeric?\n         return -o  -- '-' here is the primitive 'unm'\n       else  -- the operand is not numeric.\n         -- Try to get a handler from the operand\n         local h = metatable(op).__unm\n         if h then\n           -- call the handler with the operand\n           return (h(op))\n         else  -- no handler available: default behavior\n           error(···)\n         end\n       end\n     end\n

    \n

    • \n\n
  • \"concat\":\nthe .. (concatenation) operation.\n\n\n
    \n     function concat_event (op1, op2)\n       if (type(op1) == \"string\" or type(op1) == \"number\") and\n          (type(op2) == \"string\" or type(op2) == \"number\") then\n         return op1 .. op2  -- primitive string concatenation\n       else\n         local h = getbinhandler(op1, op2, \"__concat\")\n         if h then\n           return (h(op1, op2))\n         else\n           error(···)\n         end\n       end\n     end\n

    \n

    • \n\n
  • \"len\":\nthe # operation.\n\n\n
    \n     function len_event (op)\n       if type(op) == \"string\" then\n         return strlen(op)         -- primitive string length\n       elseif type(op) == \"table\" then\n         return #op                -- primitive table length\n       else\n         local h = metatable(op).__len\n         if h then\n           -- call the handler with the operand\n           return (h(op))\n         else  -- no handler available: default behavior\n           error(···)\n         end\n       end\n     end\n

    \nSee §2.5.5 for a description of the length of a table.\n

    • \n\n
  • \"eq\":\nthe == operation.\n\nThe function getcomphandler defines how Lua chooses a metamethod\nfor comparison operators.\nA metamethod only is selected when both objects\nbeing compared have the same type\nand the same metamethod for the selected operation.\n\n
    \n     function getcomphandler (op1, op2, event)\n       if type(op1) ~= type(op2) then return nil end\n       local mm1 = metatable(op1)[event]\n       local mm2 = metatable(op2)[event]\n       if mm1 == mm2 then return mm1 else return nil end\n     end\n

    \nThe \"eq\" event is defined as follows:\n\n

    \n     function eq_event (op1, op2)\n       if type(op1) ~= type(op2) then  -- different types?\n         return false   -- different objects\n       end\n       if op1 == op2 then   -- primitive equal?\n         return true   -- objects are equal\n       end\n       -- try metamethod\n       local h = getcomphandler(op1, op2, \"__eq\")\n       if h then\n         return (h(op1, op2))\n       else\n         return false\n       end\n     end\n

    \na ~= b is equivalent to not (a == b).\n

    • \n\n
  • \"lt\":\nthe < operation.\n\n\n
    \n     function lt_event (op1, op2)\n       if type(op1) == \"number\" and type(op2) == \"number\" then\n         return op1 < op2   -- numeric comparison\n       elseif type(op1) == \"string\" and type(op2) == \"string\" then\n         return op1 < op2   -- lexicographic comparison\n       else\n         local h = getcomphandler(op1, op2, \"__lt\")\n         if h then\n           return (h(op1, op2))\n         else\n           error(···)\n         end\n       end\n     end\n

    \na > b is equivalent to b < a.\n

    • \n\n
  • \"le\":\nthe <= operation.\n\n\n
    \n     function le_event (op1, op2)\n       if type(op1) == \"number\" and type(op2) == \"number\" then\n         return op1 <= op2   -- numeric comparison\n       elseif type(op1) == \"string\" and type(op2) == \"string\" then\n         return op1 <= op2   -- lexicographic comparison\n       else\n         local h = getcomphandler(op1, op2, \"__le\")\n         if h then\n           return (h(op1, op2))\n         else\n           h = getcomphandler(op1, op2, \"__lt\")\n           if h then\n             return not h(op2, op1)\n           else\n             error(···)\n           end\n         end\n       end\n     end\n

    \na >= b is equivalent to b <= a.\nNote that, in the absence of a \"le\" metamethod,\nLua tries the \"lt\", assuming that a <= b is\nequivalent to not (b < a).\n

    • \n\n
  • \"index\":\nThe indexing access table[key].\n\n\n
    \n     function gettable_event (table, key)\n       local h\n       if type(table) == \"table\" then\n         local v = rawget(table, key)\n         if v ~= nil then return v end\n         h = metatable(table).__index\n         if h == nil then return nil end\n       else\n         h = metatable(table).__index\n         if h == nil then\n           error(···)\n         end\n       end\n       if type(h) == \"function\" then\n         return (h(table, key))     -- call the handler\n       else return h[key]           -- or repeat operation on it\n       end\n     end\n

    \n

    • \n\n
  • \"newindex\":\nThe indexing assignment table[key] = value.\n\n\n
    \n     function settable_event (table, key, value)\n       local h\n       if type(table) == \"table\" then\n         local v = rawget(table, key)\n         if v ~= nil then rawset(table, key, value); return end\n         h = metatable(table).__newindex\n         if h == nil then rawset(table, key, value); return end\n       else\n         h = metatable(table).__newindex\n         if h == nil then\n           error(···)\n         end\n       end\n       if type(h) == \"function\" then\n         h(table, key,value)           -- call the handler\n       else h[key] = value             -- or repeat operation on it\n       end\n     end\n

    \n

    • \n\n
  • \"call\":\ncalled when Lua calls a value.\n\n\n
    \n     function function_event (func, ...)\n       if type(func) == \"function\" then\n         return func(...)   -- primitive call\n       else\n         local h = metatable(func).__call\n         if h then\n           return h(func, ...)\n         else\n           error(···)\n         end\n       end\n     end\n

    \n

    • \n\n
\n\n\n\n\n

2.9 - Environments

\n\n

\nBesides metatables,\nobjects of types thread, function, and userdata\nhave another table associated with them,\ncalled their environment.\nLike metatables, environments are regular tables and\nmultiple objects can share the same environment.\n\n\n

\nThreads are created sharing the environment of the creating thread.\nUserdata and C functions are created sharing the environment\nof the creating C function.\nNon-nested Lua functions\n(created by loadfile, loadstring or load)\nare created sharing the environment of the creating thread.\nNested Lua functions are created sharing the environment of\nthe creating Lua function.\n\n\n

\nEnvironments associated with userdata have no meaning for Lua.\nIt is only a convenience feature for programmers to associate a table to\na userdata.\n\n\n

\nEnvironments associated with threads are called\nglobal environments.\nThey are used as the default environment for threads and\nnon-nested Lua functions created by the thread\nand can be directly accessed by C code (see §3.3).\n\n\n

\nThe environment associated with a C function can be directly\naccessed by C code (see §3.3).\nIt is used as the default environment for other C functions\nand userdata created by the function.\n\n\n

\nEnvironments associated with Lua functions are used to resolve\nall accesses to global variables within the function (see §2.3).\nThey are used as the default environment for nested Lua functions\ncreated by the function.\n\n\n

\nYou can change the environment of a Lua function or the\nrunning thread by calling setfenv.\nYou can get the environment of a Lua function or the running thread\nby calling getfenv.\nTo manipulate the environment of other objects\n(userdata, C functions, other threads) you must\nuse the C API.\n\n\n\n\n\n

2.10 - Garbage Collection

\n\n

\nLua performs automatic memory management.\nThis means that\nyou have to worry neither about allocating memory for new objects\nnor about freeing it when the objects are no longer needed.\nLua manages memory automatically by running\na garbage collector from time to time\nto collect all dead objects\n(that is, objects that are no longer accessible from Lua).\nAll memory used by Lua is subject to automatic management:\ntables, userdata, functions, threads, strings, etc.\n\n\n

\nLua implements an incremental mark-and-sweep collector.\nIt uses two numbers to control its garbage-collection cycles:\nthe garbage-collector pause and\nthe garbage-collector step multiplier.\nBoth use percentage points as units\n(so that a value of 100 means an internal value of 1).\n\n\n

\nThe garbage-collector pause\ncontrols how long the collector waits before starting a new cycle.\nLarger values make the collector less aggressive.\nValues smaller than 100 mean the collector will not wait to\nstart a new cycle.\nA value of 200 means that the collector waits for the total memory in use\nto double before starting a new cycle.\n\n\n

\nThe step multiplier\ncontrols the relative speed of the collector relative to\nmemory allocation.\nLarger values make the collector more aggressive but also increase\nthe size of each incremental step.\nValues smaller than 100 make the collector too slow and\ncan result in the collector never finishing a cycle.\nThe default, 200, means that the collector runs at \"twice\"\nthe speed of memory allocation.\n\n\n

\nYou can change these numbers by calling lua_gc in C\nor collectgarbage in Lua.\nWith these functions you can also control \nthe collector directly (e.g., stop and restart it).\n\n\n\n

2.10.1 - Garbage-Collection Metamethods

\n\n

\nUsing the C API,\nyou can set garbage-collector metamethods for userdata (see §2.8).\nThese metamethods are also called finalizers.\nFinalizers allow you to coordinate Lua's garbage collection\nwith external resource management\n(such as closing files, network or database connections,\nor freeing your own memory).\n\n\n

\nGarbage userdata with a field __gc in their metatables are not\ncollected immediately by the garbage collector.\nInstead, Lua puts them in a list.\nAfter the collection,\nLua does the equivalent of the following function\nfor each userdata in that list:\n\n

\n     function gc_event (udata)\n       local h = metatable(udata).__gc\n       if h then\n         h(udata)\n       end\n     end\n
\n\n

\nAt the end of each garbage-collection cycle,\nthe finalizers for userdata are called in reverse\norder of their creation,\namong those collected in that cycle.\nThat is, the first finalizer to be called is the one associated\nwith the userdata created last in the program.\nThe userdata itself is freed only in the next garbage-collection cycle.\n\n\n\n\n\n

2.10.2 - Weak Tables

\n\n

\nA weak table is a table whose elements are\nweak references.\nA weak reference is ignored by the garbage collector.\nIn other words,\nif the only references to an object are weak references,\nthen the garbage collector will collect this object.\n\n\n

\nA weak table can have weak keys, weak values, or both.\nA table with weak keys allows the collection of its keys,\nbut prevents the collection of its values.\nA table with both weak keys and weak values allows the collection of\nboth keys and values.\nIn any case, if either the key or the value is collected,\nthe whole pair is removed from the table.\nThe weakness of a table is controlled by the\n__mode field of its metatable.\nIf the __mode field is a string containing the character 'k',\nthe keys in the table are weak.\nIf __mode contains 'v',\nthe values in the table are weak.\n\n\n

\nAfter you use a table as a metatable,\nyou should not change the value of its __mode field.\nOtherwise, the weak behavior of the tables controlled by this\nmetatable is undefined.\n\n\n\n\n\n\n\n

2.11 - Coroutines

\n\n

\nLua supports coroutines,\nalso called collaborative multithreading.\nA coroutine in Lua represents an independent thread of execution.\nUnlike threads in multithread systems, however,\na coroutine only suspends its execution by explicitly calling\na yield function.\n\n\n

\nYou create a coroutine with a call to coroutine.create.\nIts sole argument is a function\nthat is the main function of the coroutine.\nThe create function only creates a new coroutine and\nreturns a handle to it (an object of type thread);\nit does not start the coroutine execution.\n\n\n

\nWhen you first call coroutine.resume,\npassing as its first argument\na thread returned by coroutine.create,\nthe coroutine starts its execution,\nat the first line of its main function.\nExtra arguments passed to coroutine.resume are passed on\nto the coroutine main function.\nAfter the coroutine starts running,\nit runs until it terminates or yields.\n\n\n

\nA coroutine can terminate its execution in two ways:\nnormally, when its main function returns\n(explicitly or implicitly, after the last instruction);\nand abnormally, if there is an unprotected error.\nIn the first case, coroutine.resume returns true,\nplus any values returned by the coroutine main function.\nIn case of errors, coroutine.resume returns false\nplus an error message.\n\n\n

\nA coroutine yields by calling coroutine.yield.\nWhen a coroutine yields,\nthe corresponding coroutine.resume returns immediately,\neven if the yield happens inside nested function calls\n(that is, not in the main function,\nbut in a function directly or indirectly called by the main function).\nIn the case of a yield, coroutine.resume also returns true,\nplus any values passed to coroutine.yield.\nThe next time you resume the same coroutine,\nit continues its execution from the point where it yielded,\nwith the call to coroutine.yield returning any extra\narguments passed to coroutine.resume.\n\n\n

\nLike coroutine.create,\nthe coroutine.wrap function also creates a coroutine,\nbut instead of returning the coroutine itself,\nit returns a function that, when called, resumes the coroutine.\nAny arguments passed to this function\ngo as extra arguments to coroutine.resume.\ncoroutine.wrap returns all the values returned by coroutine.resume,\nexcept the first one (the boolean error code).\nUnlike coroutine.resume,\ncoroutine.wrap does not catch errors;\nany error is propagated to the caller.\n\n\n

\nAs an example,\nconsider the following code:\n\n

\n     function foo (a)\n       print(\"foo\", a)\n       return coroutine.yield(2*a)\n     end\n     \n     co = coroutine.create(function (a,b)\n           print(\"co-body\", a, b)\n           local r = foo(a+1)\n           print(\"co-body\", r)\n           local r, s = coroutine.yield(a+b, a-b)\n           print(\"co-body\", r, s)\n           return b, \"end\"\n     end)\n            \n     print(\"main\", coroutine.resume(co, 1, 10))\n     print(\"main\", coroutine.resume(co, \"r\"))\n     print(\"main\", coroutine.resume(co, \"x\", \"y\"))\n     print(\"main\", coroutine.resume(co, \"x\", \"y\"))\n

\nWhen you run it, it produces the following output:\n\n

\n     co-body 1       10\n     foo     2\n     \n     main    true    4\n     co-body r\n     main    true    11      -9\n     co-body x       y\n     main    true    10      end\n     main    false   cannot resume dead coroutine\n
\n\n\n\n\n

3 - The Application Program Interface

\n\n

\n\nThis section describes the C API for Lua, that is,\nthe set of C functions available to the host program to communicate\nwith Lua.\nAll API functions and related types and constants\nare declared in the header file lua.h.\n\n\n

\nEven when we use the term \"function\",\nany facility in the API may be provided as a macro instead.\nAll such macros use each of their arguments exactly once\n(except for the first argument, which is always a Lua state),\nand so do not generate any hidden side-effects.\n\n\n

\nAs in most C libraries,\nthe Lua API functions do not check their arguments for validity or consistency.\nHowever, you can change this behavior by compiling Lua\nwith a proper definition for the macro luai_apicheck,\nin file luaconf.h.\n\n\n\n

3.1 - The Stack

\n\n

\nLua uses a virtual stack to pass values to and from C.\nEach element in this stack represents a Lua value\n(nil, number, string, etc.).\n\n\n

\nWhenever Lua calls C, the called function gets a new stack,\nwhich is independent of previous stacks and of stacks of\nC functions that are still active.\nThis stack initially contains any arguments to the C function\nand it is where the C function pushes its results\nto be returned to the caller (see lua_CFunction).\n\n\n

\nFor convenience,\nmost query operations in the API do not follow a strict stack discipline.\nInstead, they can refer to any element in the stack\nby using an index:\nA positive index represents an absolute stack position\n(starting at 1);\na negative index represents an offset relative to the top of the stack.\nMore specifically, if the stack has n elements,\nthen index 1 represents the first element\n(that is, the element that was pushed onto the stack first)\nand\nindex n represents the last element;\nindex -1 also represents the last element\n(that is, the element at the top)\nand index -n represents the first element.\nWe say that an index is valid\nif it lies between 1 and the stack top\n(that is, if 1 ≤ abs(index) ≤ top).\n \n\n\n\n\n\n

3.2 - Stack Size

\n\n

\nWhen you interact with Lua API,\nyou are responsible for ensuring consistency.\nIn particular,\nyou are responsible for controlling stack overflow.\nYou can use the function lua_checkstack\nto grow the stack size.\n\n\n

\nWhenever Lua calls C,\nit ensures that at least LUA_MINSTACK stack positions are available.\nLUA_MINSTACK is defined as 20,\nso that usually you do not have to worry about stack space\nunless your code has loops pushing elements onto the stack.\n\n\n

\nMost query functions accept as indices any value inside the\navailable stack space, that is, indices up to the maximum stack size\nyou have set through lua_checkstack.\nSuch indices are called acceptable indices.\nMore formally, we define an acceptable index\nas follows:\n\n

\n     (index < 0 && abs(index) <= top) ||\n     (index > 0 && index <= stackspace)\n

\nNote that 0 is never an acceptable index.\n\n\n\n\n\n

3.3 - Pseudo-Indices

\n\n

\nUnless otherwise noted,\nany function that accepts valid indices can also be called with\npseudo-indices,\nwhich represent some Lua values that are accessible to C code\nbut which are not in the stack.\nPseudo-indices are used to access the thread environment,\nthe function environment,\nthe registry,\nand the upvalues of a C function (see §3.4).\n\n\n

\nThe thread environment (where global variables live) is\nalways at pseudo-index LUA_GLOBALSINDEX.\nThe environment of the running C function is always\nat pseudo-index LUA_ENVIRONINDEX.\n\n\n

\nTo access and change the value of global variables,\nyou can use regular table operations over an environment table.\nFor instance, to access the value of a global variable, do\n\n

\n     lua_getfield(L, LUA_GLOBALSINDEX, varname);\n
\n\n\n\n\n

3.4 - C Closures

\n\n

\nWhen a C function is created,\nit is possible to associate some values with it,\nthus creating a C closure;\nthese values are called upvalues and are\naccessible to the function whenever it is called\n(see lua_pushcclosure).\n\n\n

\nWhenever a C function is called,\nits upvalues are located at specific pseudo-indices.\nThese pseudo-indices are produced by the macro\nlua_upvalueindex.\nThe first value associated with a function is at position\nlua_upvalueindex(1), and so on.\nAny access to lua_upvalueindex(n),\nwhere n is greater than the number of upvalues of the\ncurrent function (but not greater than 256),\nproduces an acceptable (but invalid) index.\n\n\n\n\n\n

3.5 - Registry

\n\n

\nLua provides a registry,\na pre-defined table that can be used by any C code to\nstore whatever Lua value it needs to store.\nThis table is always located at pseudo-index\nLUA_REGISTRYINDEX.\nAny C library can store data into this table,\nbut it should take care to choose keys different from those used\nby other libraries, to avoid collisions.\nTypically, you should use as key a string containing your library name\nor a light userdata with the address of a C object in your code.\n\n\n

\nThe integer keys in the registry are used by the reference mechanism,\nimplemented by the auxiliary library,\nand therefore should not be used for other purposes.\n\n\n\n\n\n

3.6 - Error Handling in C

\n\n

\nInternally, Lua uses the C longjmp facility to handle errors.\n(You can also choose to use exceptions if you use C++;\nsee file luaconf.h.)\nWhen Lua faces any error\n(such as memory allocation errors, type errors, syntax errors,\nand runtime errors)\nit raises an error;\nthat is, it does a long jump.\nA protected environment uses setjmp\nto set a recover point;\nany error jumps to the most recent active recover point.\n\n\n

\nMost functions in the API can throw an error,\nfor instance due to a memory allocation error.\nThe documentation for each function indicates whether\nit can throw errors.\n\n\n

\nInside a C function you can throw an error by calling lua_error.\n\n\n\n\n\n

3.7 - Functions and Types

\n\n

\nHere we list all functions and types from the C API in\nalphabetical order.\nEach function has an indicator like this:\n[-o, +p, x]\n\n\n

\nThe first field, o,\nis how many elements the function pops from the stack.\nThe second field, p,\nis how many elements the function pushes onto the stack.\n(Any function always pushes its results after popping its arguments.)\nA field in the form x|y means the function can push (or pop)\nx or y elements,\ndepending on the situation;\nan interrogation mark '?' means that\nwe cannot know how many elements the function pops/pushes\nby looking only at its arguments\n(e.g., they may depend on what is on the stack).\nThe third field, x,\ntells whether the function may throw errors:\n'-' means the function never throws any error;\n'm' means the function may throw an error\nonly due to not enough memory;\n'e' means the function may throw other kinds of errors;\n'v' means the function may throw an error on purpose.\n\n\n\n

lua_Alloc

\n
typedef void * (*lua_Alloc) (void *ud,\n                             void *ptr,\n                             size_t osize,\n                             size_t nsize);
\n\n

\nThe type of the memory-allocation function used by Lua states.\nThe allocator function must provide a\nfunctionality similar to realloc,\nbut not exactly the same.\nIts arguments are\nud, an opaque pointer passed to lua_newstate;\nptr, a pointer to the block being allocated/reallocated/freed;\nosize, the original size of the block;\nnsize, the new size of the block.\nptr is NULL if and only if osize is zero.\nWhen nsize is zero, the allocator must return NULL;\nif osize is not zero,\nit should free the block pointed to by ptr.\nWhen nsize is not zero, the allocator returns NULL\nif and only if it cannot fill the request.\nWhen nsize is not zero and osize is zero,\nthe allocator should behave like malloc.\nWhen nsize and osize are not zero,\nthe allocator behaves like realloc.\nLua assumes that the allocator never fails when\nosize >= nsize.\n\n\n

\nHere is a simple implementation for the allocator function.\nIt is used in the auxiliary library by luaL_newstate.\n\n

\n     static void *l_alloc (void *ud, void *ptr, size_t osize,\n                                                size_t nsize) {\n       (void)ud;  (void)osize;  /* not used */\n       if (nsize == 0) {\n         free(ptr);\n         return NULL;\n       }\n       else\n         return realloc(ptr, nsize);\n     }\n

\nThis code assumes\nthat free(NULL) has no effect and that\nrealloc(NULL, size) is equivalent to malloc(size).\nANSI C ensures both behaviors.\n\n\n\n\n\n

lua_atpanic

\n[-0, +0, -]\n

lua_CFunction lua_atpanic (lua_State *L, lua_CFunction panicf);
\n\n

\nSets a new panic function and returns the old one.\n\n\n

\nIf an error happens outside any protected environment,\nLua calls a panic function\nand then calls exit(EXIT_FAILURE),\nthus exiting the host application.\nYour panic function can avoid this exit by\nnever returning (e.g., doing a long jump).\n\n\n

\nThe panic function can access the error message at the top of the stack.\n\n\n\n\n\n

lua_call

\n[-(nargs + 1), +nresults, e]\n

void lua_call (lua_State *L, int nargs, int nresults);
\n\n

\nCalls a function.\n\n\n

\nTo call a function you must use the following protocol:\nfirst, the function to be called is pushed onto the stack;\nthen, the arguments to the function are pushed\nin direct order;\nthat is, the first argument is pushed first.\nFinally you call lua_call;\nnargs is the number of arguments that you pushed onto the stack.\nAll arguments and the function value are popped from the stack\nwhen the function is called.\nThe function results are pushed onto the stack when the function returns.\nThe number of results is adjusted to nresults,\nunless nresults is LUA_MULTRET.\nIn this case, all results from the function are pushed.\nLua takes care that the returned values fit into the stack space.\nThe function results are pushed onto the stack in direct order\n(the first result is pushed first),\nso that after the call the last result is on the top of the stack.\n\n\n

\nAny error inside the called function is propagated upwards\n(with a longjmp).\n\n\n

\nThe following example shows how the host program can do the\nequivalent to this Lua code:\n\n

\n     a = f(\"how\", t.x, 14)\n

\nHere it is in C:\n\n

\n     lua_getfield(L, LUA_GLOBALSINDEX, \"f\"); /* function to be called */\n     lua_pushstring(L, \"how\");                        /* 1st argument */\n     lua_getfield(L, LUA_GLOBALSINDEX, \"t\");   /* table to be indexed */\n     lua_getfield(L, -1, \"x\");        /* push result of t.x (2nd arg) */\n     lua_remove(L, -2);                  /* remove 't' from the stack */\n     lua_pushinteger(L, 14);                          /* 3rd argument */\n     lua_call(L, 3, 1);     /* call 'f' with 3 arguments and 1 result */\n     lua_setfield(L, LUA_GLOBALSINDEX, \"a\");        /* set global 'a' */\n

\nNote that the code above is \"balanced\":\nat its end, the stack is back to its original configuration.\nThis is considered good programming practice.\n\n\n\n\n\n

lua_CFunction

\n
typedef int (*lua_CFunction) (lua_State *L);
\n\n

\nType for C functions.\n\n\n

\nIn order to communicate properly with Lua,\na C function must use the following protocol,\nwhich defines the way parameters and results are passed:\na C function receives its arguments from Lua in its stack\nin direct order (the first argument is pushed first).\nSo, when the function starts,\nlua_gettop(L) returns the number of arguments received by the function.\nThe first argument (if any) is at index 1\nand its last argument is at index lua_gettop(L).\nTo return values to Lua, a C function just pushes them onto the stack,\nin direct order (the first result is pushed first),\nand returns the number of results.\nAny other value in the stack below the results will be properly\ndiscarded by Lua.\nLike a Lua function, a C function called by Lua can also return\nmany results.\n\n\n

\nAs an example, the following function receives a variable number\nof numerical arguments and returns their average and sum:\n\n

\n     static int foo (lua_State *L) {\n       int n = lua_gettop(L);    /* number of arguments */\n       lua_Number sum = 0;\n       int i;\n       for (i = 1; i <= n; i++) {\n         if (!lua_isnumber(L, i)) {\n           lua_pushstring(L, \"incorrect argument\");\n           lua_error(L);\n         }\n         sum += lua_tonumber(L, i);\n       }\n       lua_pushnumber(L, sum/n);        /* first result */\n       lua_pushnumber(L, sum);         /* second result */\n       return 2;                   /* number of results */\n     }\n
\n\n\n\n\n

lua_checkstack

\n[-0, +0, m]\n

int lua_checkstack (lua_State *L, int extra);
\n\n

\nEnsures that there are at least extra free stack slots in the stack.\nIt returns false if it cannot grow the stack to that size.\nThis function never shrinks the stack;\nif the stack is already larger than the new size,\nit is left unchanged.\n\n\n\n\n\n

lua_close

\n[-0, +0, -]\n

void lua_close (lua_State *L);
\n\n

\nDestroys all objects in the given Lua state\n(calling the corresponding garbage-collection metamethods, if any)\nand frees all dynamic memory used by this state.\nOn several platforms, you may not need to call this function,\nbecause all resources are naturally released when the host program ends.\nOn the other hand, long-running programs,\nsuch as a daemon or a web server,\nmight need to release states as soon as they are not needed,\nto avoid growing too large.\n\n\n\n\n\n

lua_concat

\n[-n, +1, e]\n

void lua_concat (lua_State *L, int n);
\n\n

\nConcatenates the n values at the top of the stack,\npops them, and leaves the result at the top.\nIf n is 1, the result is the single value on the stack\n(that is, the function does nothing);\nif n is 0, the result is the empty string.\nConcatenation is performed following the usual semantics of Lua\n(see §2.5.4).\n\n\n\n\n\n

lua_cpcall

\n[-0, +(0|1), -]\n

int lua_cpcall (lua_State *L, lua_CFunction func, void *ud);
\n\n

\nCalls the C function func in protected mode.\nfunc starts with only one element in its stack,\na light userdata containing ud.\nIn case of errors,\nlua_cpcall returns the same error codes as lua_pcall,\nplus the error object on the top of the stack;\notherwise, it returns zero, and does not change the stack.\nAll values returned by func are discarded.\n\n\n\n\n\n

lua_createtable

\n[-0, +1, m]\n

void lua_createtable (lua_State *L, int narr, int nrec);
\n\n

\nCreates a new empty table and pushes it onto the stack.\nThe new table has space pre-allocated\nfor narr array elements and nrec non-array elements.\nThis pre-allocation is useful when you know exactly how many elements\nthe table will have.\nOtherwise you can use the function lua_newtable.\n\n\n\n\n\n

lua_dump

\n[-0, +0, m]\n

int lua_dump (lua_State *L, lua_Writer writer, void *data);
\n\n

\nDumps a function as a binary chunk.\nReceives a Lua function on the top of the stack\nand produces a binary chunk that,\nif loaded again,\nresults in a function equivalent to the one dumped.\nAs it produces parts of the chunk,\nlua_dump calls function writer (see lua_Writer)\nwith the given data\nto write them.\n\n\n

\nThe value returned is the error code returned by the last\ncall to the writer;\n0 means no errors.\n\n\n

\nThis function does not pop the Lua function from the stack.\n\n\n\n\n\n

lua_equal

\n[-0, +0, e]\n

int lua_equal (lua_State *L, int index1, int index2);
\n\n

\nReturns 1 if the two values in acceptable indices index1 and\nindex2 are equal,\nfollowing the semantics of the Lua == operator\n(that is, may call metamethods).\nOtherwise returns 0.\nAlso returns 0 if any of the indices is non valid.\n\n\n\n\n\n

lua_error

\n[-1, +0, v]\n

int lua_error (lua_State *L);
\n\n

\nGenerates a Lua error.\nThe error message (which can actually be a Lua value of any type)\nmust be on the stack top.\nThis function does a long jump,\nand therefore never returns.\n(see luaL_error).\n\n\n\n\n\n

lua_gc

\n[-0, +0, e]\n

int lua_gc (lua_State *L, int what, int data);
\n\n

\nControls the garbage collector.\n\n\n

\nThis function performs several tasks,\naccording to the value of the parameter what:\n\n

    \n\n
  • LUA_GCSTOP:\nstops the garbage collector.\n
    • \n\n
  • LUA_GCRESTART:\nrestarts the garbage collector.\n
    • \n\n
  • LUA_GCCOLLECT:\nperforms a full garbage-collection cycle.\n
    • \n\n
  • LUA_GCCOUNT:\nreturns the current amount of memory (in Kbytes) in use by Lua.\n
    • \n\n
  • LUA_GCCOUNTB:\nreturns the remainder of dividing the current amount of bytes of\nmemory in use by Lua by 1024.\n
    • \n\n
  • LUA_GCSTEP:\nperforms an incremental step of garbage collection.\nThe step \"size\" is controlled by data\n(larger values mean more steps) in a non-specified way.\nIf you want to control the step size\nyou must experimentally tune the value of data.\nThe function returns 1 if the step finished a\ngarbage-collection cycle.\n
    • \n\n
  • LUA_GCSETPAUSE:\nsets data as the new value\nfor the pause of the collector (see §2.10).\nThe function returns the previous value of the pause.\n
    • \n\n
  • LUA_GCSETSTEPMUL:\nsets data as the new value for the step multiplier of\nthe collector (see §2.10).\nThe function returns the previous value of the step multiplier.\n
    • \n\n
\n\n\n\n\n

lua_getallocf

\n[-0, +0, -]\n

lua_Alloc lua_getallocf (lua_State *L, void **ud);
\n\n

\nReturns the memory-allocation function of a given state.\nIf ud is not NULL, Lua stores in *ud the\nopaque pointer passed to lua_newstate.\n\n\n\n\n\n

lua_getfenv

\n[-0, +1, -]\n

void lua_getfenv (lua_State *L, int index);
\n\n

\nPushes onto the stack the environment table of\nthe value at the given index.\n\n\n\n\n\n

lua_getfield

\n[-0, +1, e]\n

void lua_getfield (lua_State *L, int index, const char *k);
\n\n

\nPushes onto the stack the value t[k],\nwhere t is the value at the given valid index.\nAs in Lua, this function may trigger a metamethod\nfor the \"index\" event (see §2.8).\n\n\n\n\n\n

lua_getglobal

\n[-0, +1, e]\n

void lua_getglobal (lua_State *L, const char *name);
\n\n

\nPushes onto the stack the value of the global name.\nIt is defined as a macro:\n\n

\n     #define lua_getglobal(L,s)  lua_getfield(L, LUA_GLOBALSINDEX, s)\n
\n\n\n\n\n

lua_getmetatable

\n[-0, +(0|1), -]\n

int lua_getmetatable (lua_State *L, int index);
\n\n

\nPushes onto the stack the metatable of the value at the given\nacceptable index.\nIf the index is not valid,\nor if the value does not have a metatable,\nthe function returns 0 and pushes nothing on the stack.\n\n\n\n\n\n

lua_gettable

\n[-1, +1, e]\n

void lua_gettable (lua_State *L, int index);
\n\n

\nPushes onto the stack the value t[k],\nwhere t is the value at the given valid index\nand k is the value at the top of the stack.\n\n\n

\nThis function pops the key from the stack\n(putting the resulting value in its place).\nAs in Lua, this function may trigger a metamethod\nfor the \"index\" event (see §2.8).\n\n\n\n\n\n

lua_gettop

\n[-0, +0, -]\n

int lua_gettop (lua_State *L);
\n\n

\nReturns the index of the top element in the stack.\nBecause indices start at 1,\nthis result is equal to the number of elements in the stack\n(and so 0 means an empty stack).\n\n\n\n\n\n

lua_insert

\n[-1, +1, -]\n

void lua_insert (lua_State *L, int index);
\n\n

\nMoves the top element into the given valid index,\nshifting up the elements above this index to open space.\nCannot be called with a pseudo-index,\nbecause a pseudo-index is not an actual stack position.\n\n\n\n\n\n

lua_Integer

\n
typedef ptrdiff_t lua_Integer;
\n\n

\nThe type used by the Lua API to represent integral values.\n\n\n

\nBy default it is a ptrdiff_t,\nwhich is usually the largest signed integral type the machine handles\n\"comfortably\".\n\n\n\n\n\n

lua_isboolean

\n[-0, +0, -]\n

int lua_isboolean (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index has type boolean,\nand 0 otherwise.\n\n\n\n\n\n

lua_iscfunction

\n[-0, +0, -]\n

int lua_iscfunction (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a C function,\nand 0 otherwise.\n\n\n\n\n\n

lua_isfunction

\n[-0, +0, -]\n

int lua_isfunction (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a function\n(either C or Lua), and 0 otherwise.\n\n\n\n\n\n

lua_islightuserdata

\n[-0, +0, -]\n

int lua_islightuserdata (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a light userdata,\nand 0 otherwise.\n\n\n\n\n\n

lua_isnil

\n[-0, +0, -]\n

int lua_isnil (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is nil,\nand 0 otherwise.\n\n\n\n\n\n

lua_isnone

\n[-0, +0, -]\n

int lua_isnone (lua_State *L, int index);
\n\n

\nReturns 1 if the given acceptable index is not valid\n(that is, it refers to an element outside the current stack),\nand 0 otherwise.\n\n\n\n\n\n

lua_isnoneornil

\n[-0, +0, -]\n

int lua_isnoneornil (lua_State *L, int index);
\n\n

\nReturns 1 if the given acceptable index is not valid\n(that is, it refers to an element outside the current stack)\nor if the value at this index is nil,\nand 0 otherwise.\n\n\n\n\n\n

lua_isnumber

\n[-0, +0, -]\n

int lua_isnumber (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a number\nor a string convertible to a number,\nand 0 otherwise.\n\n\n\n\n\n

lua_isstring

\n[-0, +0, -]\n

int lua_isstring (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a string\nor a number (which is always convertible to a string),\nand 0 otherwise.\n\n\n\n\n\n

lua_istable

\n[-0, +0, -]\n

int lua_istable (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a table,\nand 0 otherwise.\n\n\n\n\n\n

lua_isthread

\n[-0, +0, -]\n

int lua_isthread (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a thread,\nand 0 otherwise.\n\n\n\n\n\n

lua_isuserdata

\n[-0, +0, -]\n

int lua_isuserdata (lua_State *L, int index);
\n\n

\nReturns 1 if the value at the given acceptable index is a userdata\n(either full or light), and 0 otherwise.\n\n\n\n\n\n

lua_lessthan

\n[-0, +0, e]\n

int lua_lessthan (lua_State *L, int index1, int index2);
\n\n

\nReturns 1 if the value at acceptable index index1 is smaller\nthan the value at acceptable index index2,\nfollowing the semantics of the Lua < operator\n(that is, may call metamethods).\nOtherwise returns 0.\nAlso returns 0 if any of the indices is non valid.\n\n\n\n\n\n

lua_load

\n[-0, +1, -]\n

int lua_load (lua_State *L,\n              lua_Reader reader,\n              void *data,\n              const char *chunkname);
\n\n

\nLoads a Lua chunk.\nIf there are no errors,\nlua_load pushes the compiled chunk as a Lua\nfunction on top of the stack.\nOtherwise, it pushes an error message.\nThe return values of lua_load are:\n\n

    \n\n
  • 0: no errors;
    • \n\n
  • LUA_ERRSYNTAX:\nsyntax error during pre-compilation;
    • \n\n
  • LUA_ERRMEM:\nmemory allocation error.
    • \n\n
\n\n

\nThis function only loads a chunk;\nit does not run it.\n\n\n

\nlua_load automatically detects whether the chunk is text or binary,\nand loads it accordingly (see program luac).\n\n\n

\nThe lua_load function uses a user-supplied reader function\nto read the chunk (see lua_Reader).\nThe data argument is an opaque value passed to the reader function.\n\n\n

\nThe chunkname argument gives a name to the chunk,\nwhich is used for error messages and in debug information (see §3.8).\n\n\n\n\n\n

lua_newstate

\n[-0, +0, -]\n

lua_State *lua_newstate (lua_Alloc f, void *ud);
\n\n

\nCreates a new, independent state.\nReturns NULL if cannot create the state\n(due to lack of memory).\nThe argument f is the allocator function;\nLua does all memory allocation for this state through this function.\nThe second argument, ud, is an opaque pointer that Lua\nsimply passes to the allocator in every call.\n\n\n\n\n\n

lua_newtable

\n[-0, +1, m]\n

void lua_newtable (lua_State *L);
\n\n

\nCreates a new empty table and pushes it onto the stack.\nIt is equivalent to lua_createtable(L, 0, 0).\n\n\n\n\n\n

lua_newthread

\n[-0, +1, m]\n

lua_State *lua_newthread (lua_State *L);
\n\n

\nCreates a new thread, pushes it on the stack,\nand returns a pointer to a lua_State that represents this new thread.\nThe new state returned by this function shares with the original state\nall global objects (such as tables),\nbut has an independent execution stack.\n\n\n

\nThere is no explicit function to close or to destroy a thread.\nThreads are subject to garbage collection,\nlike any Lua object.\n\n\n\n\n\n

lua_newuserdata

\n[-0, +1, m]\n

void *lua_newuserdata (lua_State *L, size_t size);
\n\n

\nThis function allocates a new block of memory with the given size,\npushes onto the stack a new full userdata with the block address,\nand returns this address.\n\n\n

\nUserdata represent C values in Lua.\nA full userdata represents a block of memory.\nIt is an object (like a table):\nyou must create it, it can have its own metatable,\nand you can detect when it is being collected.\nA full userdata is only equal to itself (under raw equality).\n\n\n

\nWhen Lua collects a full userdata with a gc metamethod,\nLua calls the metamethod and marks the userdata as finalized.\nWhen this userdata is collected again then\nLua frees its corresponding memory.\n\n\n\n\n\n

lua_next

\n[-1, +(2|0), e]\n

int lua_next (lua_State *L, int index);
\n\n

\nPops a key from the stack,\nand pushes a key-value pair from the table at the given index\n(the \"next\" pair after the given key).\nIf there are no more elements in the table,\nthen lua_next returns 0 (and pushes nothing).\n\n\n

\nA typical traversal looks like this:\n\n

\n     /* table is in the stack at index 't' */\n     lua_pushnil(L);  /* first key */\n     while (lua_next(L, t) != 0) {\n       /* uses 'key' (at index -2) and 'value' (at index -1) */\n       printf(\"%s - %s\\n\",\n              lua_typename(L, lua_type(L, -2)),\n              lua_typename(L, lua_type(L, -1)));\n       /* removes 'value'; keeps 'key' for next iteration */\n       lua_pop(L, 1);\n     }\n
\n\n

\nWhile traversing a table,\ndo not call lua_tolstring directly on a key,\nunless you know that the key is actually a string.\nRecall that lua_tolstring changes\nthe value at the given index;\nthis confuses the next call to lua_next.\n\n\n\n\n\n

lua_Number

\n
typedef double lua_Number;
\n\n

\nThe type of numbers in Lua.\nBy default, it is double, but that can be changed in luaconf.h.\n\n\n

\nThrough the configuration file you can change\nLua to operate with another type for numbers (e.g., float or long).\n\n\n\n\n\n

lua_objlen

\n[-0, +0, -]\n

size_t lua_objlen (lua_State *L, int index);
\n\n

\nReturns the \"length\" of the value at the given acceptable index:\nfor strings, this is the string length;\nfor tables, this is the result of the length operator ('#');\nfor userdata, this is the size of the block of memory allocated\nfor the userdata;\nfor other values, it is 0.\n\n\n\n\n\n

lua_pcall

\n[-(nargs + 1), +(nresults|1), -]\n

int lua_pcall (lua_State *L, int nargs, int nresults, int errfunc);
\n\n

\nCalls a function in protected mode.\n\n\n

\nBoth nargs and nresults have the same meaning as\nin lua_call.\nIf there are no errors during the call,\nlua_pcall behaves exactly like lua_call.\nHowever, if there is any error,\nlua_pcall catches it,\npushes a single value on the stack (the error message),\nand returns an error code.\nLike lua_call,\nlua_pcall always removes the function\nand its arguments from the stack.\n\n\n

\nIf errfunc is 0,\nthen the error message returned on the stack\nis exactly the original error message.\nOtherwise, errfunc is the stack index of an\nerror handler function.\n(In the current implementation, this index cannot be a pseudo-index.)\nIn case of runtime errors,\nthis function will be called with the error message\nand its return value will be the message returned on the stack by lua_pcall.\n\n\n

\nTypically, the error handler function is used to add more debug\ninformation to the error message, such as a stack traceback.\nSuch information cannot be gathered after the return of lua_pcall,\nsince by then the stack has unwound.\n\n\n

\nThe lua_pcall function returns 0 in case of success\nor one of the following error codes\n(defined in lua.h):\n\n

    \n\n
  • LUA_ERRRUN:\na runtime error.\n
    • \n\n
  • LUA_ERRMEM:\nmemory allocation error.\nFor such errors, Lua does not call the error handler function.\n
    • \n\n
  • LUA_ERRERR:\nerror while running the error handler function.\n
    • \n\n
\n\n\n\n\n

lua_pop

\n[-n, +0, -]\n

void lua_pop (lua_State *L, int n);
\n\n

\nPops n elements from the stack.\n\n\n\n\n\n

lua_pushboolean

\n[-0, +1, -]\n

void lua_pushboolean (lua_State *L, int b);
\n\n

\nPushes a boolean value with value b onto the stack.\n\n\n\n\n\n

lua_pushcclosure

\n[-n, +1, m]\n

void lua_pushcclosure (lua_State *L, lua_CFunction fn, int n);
\n\n

\nPushes a new C closure onto the stack.\n\n\n

\nWhen a C function is created,\nit is possible to associate some values with it,\nthus creating a C closure (see §3.4);\nthese values are then accessible to the function whenever it is called.\nTo associate values with a C function,\nfirst these values should be pushed onto the stack\n(when there are multiple values, the first value is pushed first).\nThen lua_pushcclosure\nis called to create and push the C function onto the stack,\nwith the argument n telling how many values should be\nassociated with the function.\nlua_pushcclosure also pops these values from the stack.\n\n\n

\nThe maximum value for n is 255.\n\n\n\n\n\n

lua_pushcfunction

\n[-0, +1, m]\n

void lua_pushcfunction (lua_State *L, lua_CFunction f);
\n\n

\nPushes a C function onto the stack.\nThis function receives a pointer to a C function\nand pushes onto the stack a Lua value of type function that,\nwhen called, invokes the corresponding C function.\n\n\n

\nAny function to be registered in Lua must\nfollow the correct protocol to receive its parameters\nand return its results (see lua_CFunction).\n\n\n

\nlua_pushcfunction is defined as a macro:\n\n

\n     #define lua_pushcfunction(L,f)  lua_pushcclosure(L,f,0)\n
\n\n\n\n\n

lua_pushfstring

\n[-0, +1, m]\n

const char *lua_pushfstring (lua_State *L, const char *fmt, ...);
\n\n

\nPushes onto the stack a formatted string\nand returns a pointer to this string.\nIt is similar to the C function sprintf,\nbut has some important differences:\n\n

    \n\n
  • \nYou do not have to allocate space for the result:\nthe result is a Lua string and Lua takes care of memory allocation\n(and deallocation, through garbage collection).\n
    • \n\n
  • \nThe conversion specifiers are quite restricted.\nThere are no flags, widths, or precisions.\nThe conversion specifiers can only be\n'%%' (inserts a '%' in the string),\n'%s' (inserts a zero-terminated string, with no size restrictions),\n'%f' (inserts a lua_Number),\n'%p' (inserts a pointer as a hexadecimal numeral),\n'%d' (inserts an int), and\n'%c' (inserts an int as a character).\n
    • \n\n
\n\n\n\n\n

lua_pushinteger

\n[-0, +1, -]\n

void lua_pushinteger (lua_State *L, lua_Integer n);
\n\n

\nPushes a number with value n onto the stack.\n\n\n\n\n\n

lua_pushlightuserdata

\n[-0, +1, -]\n

void lua_pushlightuserdata (lua_State *L, void *p);
\n\n

\nPushes a light userdata onto the stack.\n\n\n

\nUserdata represent C values in Lua.\nA light userdata represents a pointer.\nIt is a value (like a number):\nyou do not create it, it has no individual metatable,\nand it is not collected (as it was never created).\nA light userdata is equal to \"any\"\nlight userdata with the same C address.\n\n\n\n\n\n

lua_pushliteral

\n[-0, +1, m]\n

void lua_pushliteral (lua_State *L, const char *s);
\n\n

\nThis macro is equivalent to lua_pushlstring,\nbut can be used only when s is a literal string.\nIn these cases, it automatically provides the string length.\n\n\n\n\n\n

lua_pushlstring

\n[-0, +1, m]\n

void lua_pushlstring (lua_State *L, const char *s, size_t len);
\n\n

\nPushes the string pointed to by s with size len\nonto the stack.\nLua makes (or reuses) an internal copy of the given string,\nso the memory at s can be freed or reused immediately after\nthe function returns.\nThe string can contain embedded zeros.\n\n\n\n\n\n

lua_pushnil

\n[-0, +1, -]\n

void lua_pushnil (lua_State *L);
\n\n

\nPushes a nil value onto the stack.\n\n\n\n\n\n

lua_pushnumber

\n[-0, +1, -]\n

void lua_pushnumber (lua_State *L, lua_Number n);
\n\n

\nPushes a number with value n onto the stack.\n\n\n\n\n\n

lua_pushstring

\n[-0, +1, m]\n

void lua_pushstring (lua_State *L, const char *s);
\n\n

\nPushes the zero-terminated string pointed to by s\nonto the stack.\nLua makes (or reuses) an internal copy of the given string,\nso the memory at s can be freed or reused immediately after\nthe function returns.\nThe string cannot contain embedded zeros;\nit is assumed to end at the first zero.\n\n\n\n\n\n

lua_pushthread

\n[-0, +1, -]\n

int lua_pushthread (lua_State *L);
\n\n

\nPushes the thread represented by L onto the stack.\nReturns 1 if this thread is the main thread of its state.\n\n\n\n\n\n

lua_pushvalue

\n[-0, +1, -]\n

void lua_pushvalue (lua_State *L, int index);
\n\n

\nPushes a copy of the element at the given valid index\nonto the stack.\n\n\n\n\n\n

lua_pushvfstring

\n[-0, +1, m]\n

const char *lua_pushvfstring (lua_State *L,\n                              const char *fmt,\n                              va_list argp);
\n\n

\nEquivalent to lua_pushfstring, except that it receives a va_list\ninstead of a variable number of arguments.\n\n\n\n\n\n

lua_rawequal

\n[-0, +0, -]\n

int lua_rawequal (lua_State *L, int index1, int index2);
\n\n

\nReturns 1 if the two values in acceptable indices index1 and\nindex2 are primitively equal\n(that is, without calling metamethods).\nOtherwise returns 0.\nAlso returns 0 if any of the indices are non valid.\n\n\n\n\n\n

lua_rawget

\n[-1, +1, -]\n

void lua_rawget (lua_State *L, int index);
\n\n

\nSimilar to lua_gettable, but does a raw access\n(i.e., without metamethods).\n\n\n\n\n\n

lua_rawgeti

\n[-0, +1, -]\n

void lua_rawgeti (lua_State *L, int index, int n);
\n\n

\nPushes onto the stack the value t[n],\nwhere t is the value at the given valid index.\nThe access is raw;\nthat is, it does not invoke metamethods.\n\n\n\n\n\n

lua_rawset

\n[-2, +0, m]\n

void lua_rawset (lua_State *L, int index);
\n\n

\nSimilar to lua_settable, but does a raw assignment\n(i.e., without metamethods).\n\n\n\n\n\n

lua_rawseti

\n[-1, +0, m]\n

void lua_rawseti (lua_State *L, int index, int n);
\n\n

\nDoes the equivalent of t[n] = v,\nwhere t is the value at the given valid index\nand v is the value at the top of the stack.\n\n\n

\nThis function pops the value from the stack.\nThe assignment is raw;\nthat is, it does not invoke metamethods.\n\n\n\n\n\n

lua_Reader

\n
typedef const char * (*lua_Reader) (lua_State *L,\n                                    void *data,\n                                    size_t *size);
\n\n

\nThe reader function used by lua_load.\nEvery time it needs another piece of the chunk,\nlua_load calls the reader,\npassing along its data parameter.\nThe reader must return a pointer to a block of memory\nwith a new piece of the chunk\nand set size to the block size.\nThe block must exist until the reader function is called again.\nTo signal the end of the chunk,\nthe reader must return NULL or set size to zero.\nThe reader function may return pieces of any size greater than zero.\n\n\n\n\n\n

lua_register

\n[-0, +0, e]\n

void lua_register (lua_State *L,\n                   const char *name,\n                   lua_CFunction f);
\n\n

\nSets the C function f as the new value of global name.\nIt is defined as a macro:\n\n

\n     #define lua_register(L,n,f) \\\n            (lua_pushcfunction(L, f), lua_setglobal(L, n))\n
\n\n\n\n\n

lua_remove

\n[-1, +0, -]\n

void lua_remove (lua_State *L, int index);
\n\n

\nRemoves the element at the given valid index,\nshifting down the elements above this index to fill the gap.\nCannot be called with a pseudo-index,\nbecause a pseudo-index is not an actual stack position.\n\n\n\n\n\n

lua_replace

\n[-1, +0, -]\n

void lua_replace (lua_State *L, int index);
\n\n

\nMoves the top element into the given position (and pops it),\nwithout shifting any element\n(therefore replacing the value at the given position).\n\n\n\n\n\n

lua_resume

\n[-?, +?, -]\n

int lua_resume (lua_State *L, int narg);
\n\n

\nStarts and resumes a coroutine in a given thread.\n\n\n

\nTo start a coroutine, you first create a new thread\n(see lua_newthread);\nthen you push onto its stack the main function plus any arguments;\nthen you call lua_resume,\nwith narg being the number of arguments.\nThis call returns when the coroutine suspends or finishes its execution.\nWhen it returns, the stack contains all values passed to lua_yield,\nor all values returned by the body function.\nlua_resume returns\nLUA_YIELD if the coroutine yields,\n0 if the coroutine finishes its execution\nwithout errors,\nor an error code in case of errors (see lua_pcall).\nIn case of errors,\nthe stack is not unwound,\nso you can use the debug API over it.\nThe error message is on the top of the stack.\nTo restart a coroutine, you put on its stack only the values to\nbe passed as results from yield,\nand then call lua_resume.\n\n\n\n\n\n

lua_setallocf

\n[-0, +0, -]\n

void lua_setallocf (lua_State *L, lua_Alloc f, void *ud);
\n\n

\nChanges the allocator function of a given state to f\nwith user data ud.\n\n\n\n\n\n

lua_setfenv

\n[-1, +0, -]\n

int lua_setfenv (lua_State *L, int index);
\n\n

\nPops a table from the stack and sets it as\nthe new environment for the value at the given index.\nIf the value at the given index is\nneither a function nor a thread nor a userdata,\nlua_setfenv returns 0.\nOtherwise it returns 1.\n\n\n\n\n\n

lua_setfield

\n[-1, +0, e]\n

void lua_setfield (lua_State *L, int index, const char *k);
\n\n

\nDoes the equivalent to t[k] = v,\nwhere t is the value at the given valid index\nand v is the value at the top of the stack.\n\n\n

\nThis function pops the value from the stack.\nAs in Lua, this function may trigger a metamethod\nfor the \"newindex\" event (see §2.8).\n\n\n\n\n\n

lua_setglobal

\n[-1, +0, e]\n

void lua_setglobal (lua_State *L, const char *name);
\n\n

\nPops a value from the stack and\nsets it as the new value of global name.\nIt is defined as a macro:\n\n

\n     #define lua_setglobal(L,s)   lua_setfield(L, LUA_GLOBALSINDEX, s)\n
\n\n\n\n\n

lua_setmetatable

\n[-1, +0, -]\n

int lua_setmetatable (lua_State *L, int index);
\n\n

\nPops a table from the stack and\nsets it as the new metatable for the value at the given\nacceptable index.\n\n\n\n\n\n

lua_settable

\n[-2, +0, e]\n

void lua_settable (lua_State *L, int index);
\n\n

\nDoes the equivalent to t[k] = v,\nwhere t is the value at the given valid index,\nv is the value at the top of the stack,\nand k is the value just below the top.\n\n\n

\nThis function pops both the key and the value from the stack.\nAs in Lua, this function may trigger a metamethod\nfor the \"newindex\" event (see §2.8).\n\n\n\n\n\n

lua_settop

\n[-?, +?, -]\n

void lua_settop (lua_State *L, int index);
\n\n

\nAccepts any acceptable index, or 0,\nand sets the stack top to this index.\nIf the new top is larger than the old one,\nthen the new elements are filled with nil.\nIf index is 0, then all stack elements are removed.\n\n\n\n\n\n

lua_State

\n
typedef struct lua_State lua_State;
\n\n

\nOpaque structure that keeps the whole state of a Lua interpreter.\nThe Lua library is fully reentrant:\nit has no global variables.\nAll information about a state is kept in this structure.\n\n\n

\nA pointer to this state must be passed as the first argument to\nevery function in the library, except to lua_newstate,\nwhich creates a Lua state from scratch.\n\n\n\n\n\n

lua_status

\n[-0, +0, -]\n

int lua_status (lua_State *L);
\n\n

\nReturns the status of the thread L.\n\n\n

\nThe status can be 0 for a normal thread,\nan error code if the thread finished its execution with an error,\nor LUA_YIELD if the thread is suspended.\n\n\n\n\n\n

lua_toboolean

\n[-0, +0, -]\n

int lua_toboolean (lua_State *L, int index);
\n\n

\nConverts the Lua value at the given acceptable index to a C boolean\nvalue (0 or 1).\nLike all tests in Lua,\nlua_toboolean returns 1 for any Lua value\ndifferent from false and nil;\notherwise it returns 0.\nIt also returns 0 when called with a non-valid index.\n(If you want to accept only actual boolean values,\nuse lua_isboolean to test the value's type.)\n\n\n\n\n\n

lua_tocfunction

\n[-0, +0, -]\n

lua_CFunction lua_tocfunction (lua_State *L, int index);
\n\n

\nConverts a value at the given acceptable index to a C function.\nThat value must be a C function;\notherwise, returns NULL.\n\n\n\n\n\n

lua_tointeger

\n[-0, +0, -]\n

lua_Integer lua_tointeger (lua_State *L, int index);
\n\n

\nConverts the Lua value at the given acceptable index\nto the signed integral type lua_Integer.\nThe Lua value must be a number or a string convertible to a number\n(see §2.2.1);\notherwise, lua_tointeger returns 0.\n\n\n

\nIf the number is not an integer,\nit is truncated in some non-specified way.\n\n\n\n\n\n

lua_tolstring

\n[-0, +0, m]\n

const char *lua_tolstring (lua_State *L, int index, size_t *len);
\n\n

\nConverts the Lua value at the given acceptable index to a C string.\nIf len is not NULL,\nit also sets *len with the string length.\nThe Lua value must be a string or a number;\notherwise, the function returns NULL.\nIf the value is a number,\nthen lua_tolstring also\nchanges the actual value in the stack to a string.\n(This change confuses lua_next\nwhen lua_tolstring is applied to keys during a table traversal.)\n\n\n

\nlua_tolstring returns a fully aligned pointer\nto a string inside the Lua state.\nThis string always has a zero ('\\0')\nafter its last character (as in C),\nbut can contain other zeros in its body.\nBecause Lua has garbage collection,\nthere is no guarantee that the pointer returned by lua_tolstring\nwill be valid after the corresponding value is removed from the stack.\n\n\n\n\n\n

lua_tonumber

\n[-0, +0, -]\n

lua_Number lua_tonumber (lua_State *L, int index);
\n\n

\nConverts the Lua value at the given acceptable index\nto the C type lua_Number (see lua_Number).\nThe Lua value must be a number or a string convertible to a number\n(see §2.2.1);\notherwise, lua_tonumber returns 0.\n\n\n\n\n\n

lua_topointer

\n[-0, +0, -]\n

const void *lua_topointer (lua_State *L, int index);
\n\n

\nConverts the value at the given acceptable index to a generic\nC pointer (void*).\nThe value can be a userdata, a table, a thread, or a function;\notherwise, lua_topointer returns NULL.\nDifferent objects will give different pointers.\nThere is no way to convert the pointer back to its original value.\n\n\n

\nTypically this function is used only for debug information.\n\n\n\n\n\n

lua_tostring

\n[-0, +0, m]\n

const char *lua_tostring (lua_State *L, int index);
\n\n

\nEquivalent to lua_tolstring with len equal to NULL.\n\n\n\n\n\n

lua_tothread

\n[-0, +0, -]\n

lua_State *lua_tothread (lua_State *L, int index);
\n\n

\nConverts the value at the given acceptable index to a Lua thread\n(represented as lua_State*).\nThis value must be a thread;\notherwise, the function returns NULL.\n\n\n\n\n\n

lua_touserdata

\n[-0, +0, -]\n

void *lua_touserdata (lua_State *L, int index);
\n\n

\nIf the value at the given acceptable index is a full userdata,\nreturns its block address.\nIf the value is a light userdata,\nreturns its pointer.\nOtherwise, returns NULL.\n\n\n\n\n\n

lua_type

\n[-0, +0, -]\n

int lua_type (lua_State *L, int index);
\n\n

\nReturns the type of the value in the given acceptable index,\nor LUA_TNONE for a non-valid index\n(that is, an index to an \"empty\" stack position).\nThe types returned by lua_type are coded by the following constants\ndefined in lua.h:\nLUA_TNIL,\nLUA_TNUMBER,\nLUA_TBOOLEAN,\nLUA_TSTRING,\nLUA_TTABLE,\nLUA_TFUNCTION,\nLUA_TUSERDATA,\nLUA_TTHREAD,\nand\nLUA_TLIGHTUSERDATA.\n\n\n\n\n\n

lua_typename

\n[-0, +0, -]\n

const char *lua_typename  (lua_State *L, int tp);
\n\n

\nReturns the name of the type encoded by the value tp,\nwhich must be one the values returned by lua_type.\n\n\n\n\n\n

lua_Writer

\n
typedef int (*lua_Writer) (lua_State *L,\n                           const void* p,\n                           size_t sz,\n                           void* ud);
\n\n

\nThe type of the writer function used by lua_dump.\nEvery time it produces another piece of chunk,\nlua_dump calls the writer,\npassing along the buffer to be written (p),\nits size (sz),\nand the data parameter supplied to lua_dump.\n\n\n

\nThe writer returns an error code:\n0 means no errors;\nany other value means an error and stops lua_dump from\ncalling the writer again.\n\n\n\n\n\n

lua_xmove

\n[-?, +?, -]\n

void lua_xmove (lua_State *from, lua_State *to, int n);
\n\n

\nExchange values between different threads of the same global state.\n\n\n

\nThis function pops n values from the stack from,\nand pushes them onto the stack to.\n\n\n\n\n\n

lua_yield

\n[-?, +?, -]\n

int lua_yield  (lua_State *L, int nresults);
\n\n

\nYields a coroutine.\n\n\n

\nThis function should only be called as the\nreturn expression of a C function, as follows:\n\n

\n     return lua_yield (L, nresults);\n

\nWhen a C function calls lua_yield in that way,\nthe running coroutine suspends its execution,\nand the call to lua_resume that started this coroutine returns.\nThe parameter nresults is the number of values from the stack\nthat are passed as results to lua_resume.\n\n\n\n\n\n\n\n

3.8 - The Debug Interface

\n\n

\nLua has no built-in debugging facilities.\nInstead, it offers a special interface\nby means of functions and hooks.\nThis interface allows the construction of different\nkinds of debuggers, profilers, and other tools\nthat need \"inside information\" from the interpreter.\n\n\n\n

lua_Debug

\n
typedef struct lua_Debug {\n  int event;\n  const char *name;           /* (n) */\n  const char *namewhat;       /* (n) */\n  const char *what;           /* (S) */\n  const char *source;         /* (S) */\n  int currentline;            /* (l) */\n  int nups;                   /* (u) number of upvalues */\n  int linedefined;            /* (S) */\n  int lastlinedefined;        /* (S) */\n  char short_src[LUA_IDSIZE]; /* (S) */\n  /* private part */\n  other fields\n} lua_Debug;
\n\n

\nA structure used to carry different pieces of\ninformation about an active function.\nlua_getstack fills only the private part\nof this structure, for later use.\nTo fill the other fields of lua_Debug with useful information,\ncall lua_getinfo.\n\n\n

\nThe fields of lua_Debug have the following meaning:\n\n

    \n\n
  • source:\nIf the function was defined in a string,\nthen source is that string.\nIf the function was defined in a file,\nthen source starts with a '@' followed by the file name.\n
    • \n\n
  • short_src:\na \"printable\" version of source, to be used in error messages.\n
    • \n\n
  • linedefined:\nthe line number where the definition of the function starts.\n
    • \n\n
  • lastlinedefined:\nthe line number where the definition of the function ends.\n
    • \n\n
  • what:\nthe string \"Lua\" if the function is a Lua function,\n\"C\" if it is a C function,\n\"main\" if it is the main part of a chunk,\nand \"tail\" if it was a function that did a tail call.\nIn the latter case,\nLua has no other information about the function.\n
    • \n\n
  • currentline:\nthe current line where the given function is executing.\nWhen no line information is available,\ncurrentline is set to -1.\n
    • \n\n
  • name:\na reasonable name for the given function.\nBecause functions in Lua are first-class values,\nthey do not have a fixed name:\nsome functions can be the value of multiple global variables,\nwhile others can be stored only in a table field.\nThe lua_getinfo function checks how the function was\ncalled to find a suitable name.\nIf it cannot find a name,\nthen name is set to NULL.\n
    • \n\n
  • namewhat:\nexplains the name field.\nThe value of namewhat can be\n\"global\", \"local\", \"method\",\n\"field\", \"upvalue\", or \"\" (the empty string),\naccording to how the function was called.\n(Lua uses the empty string when no other option seems to apply.)\n
    • \n\n
  • nups:\nthe number of upvalues of the function.\n
    • \n\n
\n\n\n\n\n

lua_gethook

\n[-0, +0, -]\n

lua_Hook lua_gethook (lua_State *L);
\n\n

\nReturns the current hook function.\n\n\n\n\n\n

lua_gethookcount

\n[-0, +0, -]\n

int lua_gethookcount (lua_State *L);
\n\n

\nReturns the current hook count.\n\n\n\n\n\n

lua_gethookmask

\n[-0, +0, -]\n

int lua_gethookmask (lua_State *L);
\n\n

\nReturns the current hook mask.\n\n\n\n\n\n

lua_getinfo

\n[-(0|1), +(0|1|2), m]\n

int lua_getinfo (lua_State *L, const char *what, lua_Debug *ar);
\n\n

\nReturns information about a specific function or function invocation.\n\n\n

\nTo get information about a function invocation,\nthe parameter ar must be a valid activation record that was\nfilled by a previous call to lua_getstack or\ngiven as argument to a hook (see lua_Hook).\n\n\n

\nTo get information about a function you push it onto the stack\nand start the what string with the character '>'.\n(In that case,\nlua_getinfo pops the function in the top of the stack.)\nFor instance, to know in which line a function f was defined,\nyou can write the following code:\n\n

\n     lua_Debug ar;\n     lua_getfield(L, LUA_GLOBALSINDEX, \"f\");  /* get global 'f' */\n     lua_getinfo(L, \">S\", &ar);\n     printf(\"%d\\n\", ar.linedefined);\n
\n\n

\nEach character in the string what\nselects some fields of the structure ar to be filled or\na value to be pushed on the stack:\n\n

    \n\n
  • 'n': fills in the field name and namewhat;\n
    • \n\n
  • 'S':\nfills in the fields source, short_src,\nlinedefined, lastlinedefined, and what;\n
    • \n\n
  • 'l': fills in the field currentline;\n
    • \n\n
  • 'u': fills in the field nups;\n
    • \n\n
  • 'f':\npushes onto the stack the function that is\nrunning at the given level;\n
    • \n\n
  • 'L':\npushes onto the stack a table whose indices are the\nnumbers of the lines that are valid on the function.\n(A valid line is a line with some associated code,\nthat is, a line where you can put a break point.\nNon-valid lines include empty lines and comments.)\n
    • \n\n
\n\n

\nThis function returns 0 on error\n(for instance, an invalid option in what).\n\n\n\n\n\n

lua_getlocal

\n[-0, +(0|1), -]\n

const char *lua_getlocal (lua_State *L, lua_Debug *ar, int n);
\n\n

\nGets information about a local variable of a given activation record.\nThe parameter ar must be a valid activation record that was\nfilled by a previous call to lua_getstack or\ngiven as argument to a hook (see lua_Hook).\nThe index n selects which local variable to inspect\n(1 is the first parameter or active local variable, and so on,\nuntil the last active local variable).\nlua_getlocal pushes the variable's value onto the stack\nand returns its name.\n\n\n

\nVariable names starting with '(' (open parentheses)\nrepresent internal variables\n(loop control variables, temporaries, and C function locals).\n\n\n

\nReturns NULL (and pushes nothing)\nwhen the index is greater than\nthe number of active local variables.\n\n\n\n\n\n

lua_getstack

\n[-0, +0, -]\n

int lua_getstack (lua_State *L, int level, lua_Debug *ar);
\n\n

\nGet information about the interpreter runtime stack.\n\n\n

\nThis function fills parts of a lua_Debug structure with\nan identification of the activation record\nof the function executing at a given level.\nLevel 0 is the current running function,\nwhereas level n+1 is the function that has called level n.\nWhen there are no errors, lua_getstack returns 1;\nwhen called with a level greater than the stack depth,\nit returns 0.\n\n\n\n\n\n

lua_getupvalue

\n[-0, +(0|1), -]\n

const char *lua_getupvalue (lua_State *L, int funcindex, int n);
\n\n

\nGets information about a closure's upvalue.\n(For Lua functions,\nupvalues are the external local variables that the function uses,\nand that are consequently included in its closure.)\nlua_getupvalue gets the index n of an upvalue,\npushes the upvalue's value onto the stack,\nand returns its name.\nfuncindex points to the closure in the stack.\n(Upvalues have no particular order,\nas they are active through the whole function.\nSo, they are numbered in an arbitrary order.)\n\n\n

\nReturns NULL (and pushes nothing)\nwhen the index is greater than the number of upvalues.\nFor C functions, this function uses the empty string \"\"\nas a name for all upvalues.\n\n\n\n\n\n

lua_Hook

\n
typedef void (*lua_Hook) (lua_State *L, lua_Debug *ar);
\n\n

\nType for debugging hook functions.\n\n\n

\nWhenever a hook is called, its ar argument has its field\nevent set to the specific event that triggered the hook.\nLua identifies these events with the following constants:\nLUA_HOOKCALL, LUA_HOOKRET,\nLUA_HOOKTAILRET, LUA_HOOKLINE,\nand LUA_HOOKCOUNT.\nMoreover, for line events, the field currentline is also set.\nTo get the value of any other field in ar,\nthe hook must call lua_getinfo.\nFor return events, event can be LUA_HOOKRET,\nthe normal value, or LUA_HOOKTAILRET.\nIn the latter case, Lua is simulating a return from\na function that did a tail call;\nin this case, it is useless to call lua_getinfo.\n\n\n

\nWhile Lua is running a hook, it disables other calls to hooks.\nTherefore, if a hook calls back Lua to execute a function or a chunk,\nthis execution occurs without any calls to hooks.\n\n\n\n\n\n

lua_sethook

\n[-0, +0, -]\n

int lua_sethook (lua_State *L, lua_Hook f, int mask, int count);
\n\n

\nSets the debugging hook function.\n\n\n

\nArgument f is the hook function.\nmask specifies on which events the hook will be called:\nit is formed by a bitwise or of the constants\nLUA_MASKCALL,\nLUA_MASKRET,\nLUA_MASKLINE,\nand LUA_MASKCOUNT.\nThe count argument is only meaningful when the mask\nincludes LUA_MASKCOUNT.\nFor each event, the hook is called as explained below:\n\n

    \n\n
  • The call hook: is called when the interpreter calls a function.\nThe hook is called just after Lua enters the new function,\nbefore the function gets its arguments.\n
    • \n\n
  • The return hook: is called when the interpreter returns from a function.\nThe hook is called just before Lua leaves the function.\nYou have no access to the values to be returned by the function.\n
    • \n\n
  • The line hook: is called when the interpreter is about to\nstart the execution of a new line of code,\nor when it jumps back in the code (even to the same line).\n(This event only happens while Lua is executing a Lua function.)\n
    • \n\n
  • The count hook: is called after the interpreter executes every\ncount instructions.\n(This event only happens while Lua is executing a Lua function.)\n
    • \n\n
\n\n

\nA hook is disabled by setting mask to zero.\n\n\n\n\n\n

lua_setlocal

\n[-(0|1), +0, -]\n

const char *lua_setlocal (lua_State *L, lua_Debug *ar, int n);
\n\n

\nSets the value of a local variable of a given activation record.\nParameters ar and n are as in lua_getlocal\n(see lua_getlocal).\nlua_setlocal assigns the value at the top of the stack\nto the variable and returns its name.\nIt also pops the value from the stack.\n\n\n

\nReturns NULL (and pops nothing)\nwhen the index is greater than\nthe number of active local variables.\n\n\n\n\n\n

lua_setupvalue

\n[-(0|1), +0, -]\n

const char *lua_setupvalue (lua_State *L, int funcindex, int n);
\n\n

\nSets the value of a closure's upvalue.\nIt assigns the value at the top of the stack\nto the upvalue and returns its name.\nIt also pops the value from the stack.\nParameters funcindex and n are as in the lua_getupvalue\n(see lua_getupvalue).\n\n\n

\nReturns NULL (and pops nothing)\nwhen the index is greater than the number of upvalues.\n\n\n\n\n\n\n\n

4 - The Auxiliary Library

\n\n

\n\nThe auxiliary library provides several convenient functions\nto interface C with Lua.\nWhile the basic API provides the primitive functions for all \ninteractions between C and Lua,\nthe auxiliary library provides higher-level functions for some\ncommon tasks.\n\n\n

\nAll functions from the auxiliary library\nare defined in header file lauxlib.h and\nhave a prefix luaL_.\n\n\n

\nAll functions in the auxiliary library are built on\ntop of the basic API,\nand so they provide nothing that cannot be done with this API.\n\n\n

\nSeveral functions in the auxiliary library are used to\ncheck C function arguments.\nTheir names are always luaL_check* or luaL_opt*.\nAll of these functions throw an error if the check is not satisfied.\nBecause the error message is formatted for arguments\n(e.g., \"bad argument #1\"),\nyou should not use these functions for other stack values.\n\n\n\n

4.1 - Functions and Types

\n\n

\nHere we list all functions and types from the auxiliary library\nin alphabetical order.\n\n\n\n

luaL_addchar

\n[-0, +0, m]\n

void luaL_addchar (luaL_Buffer *B, char c);
\n\n

\nAdds the character c to the buffer B\n(see luaL_Buffer).\n\n\n\n\n\n

luaL_addlstring

\n[-0, +0, m]\n

void luaL_addlstring (luaL_Buffer *B, const char *s, size_t l);
\n\n

\nAdds the string pointed to by s with length l to\nthe buffer B\n(see luaL_Buffer).\nThe string may contain embedded zeros.\n\n\n\n\n\n

luaL_addsize

\n[-0, +0, m]\n

void luaL_addsize (luaL_Buffer *B, size_t n);
\n\n

\nAdds to the buffer B (see luaL_Buffer)\na string of length n previously copied to the\nbuffer area (see luaL_prepbuffer).\n\n\n\n\n\n

luaL_addstring

\n[-0, +0, m]\n

void luaL_addstring (luaL_Buffer *B, const char *s);
\n\n

\nAdds the zero-terminated string pointed to by s\nto the buffer B\n(see luaL_Buffer).\nThe string may not contain embedded zeros.\n\n\n\n\n\n

luaL_addvalue

\n[-1, +0, m]\n

void luaL_addvalue (luaL_Buffer *B);
\n\n

\nAdds the value at the top of the stack\nto the buffer B\n(see luaL_Buffer).\nPops the value.\n\n\n

\nThis is the only function on string buffers that can (and must)\nbe called with an extra element on the stack,\nwhich is the value to be added to the buffer.\n\n\n\n\n\n

luaL_argcheck

\n[-0, +0, v]\n

void luaL_argcheck (lua_State *L,\n                    int cond,\n                    int narg,\n                    const char *extramsg);
\n\n

\nChecks whether cond is true.\nIf not, raises an error with the following message,\nwhere func is retrieved from the call stack:\n\n

\n     bad argument #<narg> to <func> (<extramsg>)\n
\n\n\n\n\n

luaL_argerror

\n[-0, +0, v]\n

int luaL_argerror (lua_State *L, int narg, const char *extramsg);
\n\n

\nRaises an error with the following message,\nwhere func is retrieved from the call stack:\n\n

\n     bad argument #<narg> to <func> (<extramsg>)\n
\n\n

\nThis function never returns,\nbut it is an idiom to use it in C functions\nas return luaL_argerror(args).\n\n\n\n\n\n

luaL_Buffer

\n
typedef struct luaL_Buffer luaL_Buffer;
\n\n

\nType for a string buffer.\n\n\n

\nA string buffer allows C code to build Lua strings piecemeal.\nIts pattern of use is as follows:\n\n

    \n\n
  • First you declare a variable b of type luaL_Buffer.
    • \n\n
  • Then you initialize it with a call luaL_buffinit(L, &b).
    • \n\n
  • \nThen you add string pieces to the buffer calling any of\nthe luaL_add* functions.\n
    • \n\n
  • \nYou finish by calling luaL_pushresult(&b).\nThis call leaves the final string on the top of the stack.\n
    • \n\n
\n\n

\nDuring its normal operation,\na string buffer uses a variable number of stack slots.\nSo, while using a buffer, you cannot assume that you know where\nthe top of the stack is.\nYou can use the stack between successive calls to buffer operations\nas long as that use is balanced;\nthat is,\nwhen you call a buffer operation,\nthe stack is at the same level\nit was immediately after the previous buffer operation.\n(The only exception to this rule is luaL_addvalue.)\nAfter calling luaL_pushresult the stack is back to its\nlevel when the buffer was initialized,\nplus the final string on its top.\n\n\n\n\n\n

luaL_buffinit

\n[-0, +0, -]\n

void luaL_buffinit (lua_State *L, luaL_Buffer *B);
\n\n

\nInitializes a buffer B.\nThis function does not allocate any space;\nthe buffer must be declared as a variable\n(see luaL_Buffer).\n\n\n\n\n\n

luaL_callmeta

\n[-0, +(0|1), e]\n

int luaL_callmeta (lua_State *L, int obj, const char *e);
\n\n

\nCalls a metamethod.\n\n\n

\nIf the object at index obj has a metatable and this\nmetatable has a field e,\nthis function calls this field and passes the object as its only argument.\nIn this case this function returns 1 and pushes onto the\nstack the value returned by the call.\nIf there is no metatable or no metamethod,\nthis function returns 0 (without pushing any value on the stack).\n\n\n\n\n\n

luaL_checkany

\n[-0, +0, v]\n

void luaL_checkany (lua_State *L, int narg);
\n\n

\nChecks whether the function has an argument\nof any type (including nil) at position narg.\n\n\n\n\n\n

luaL_checkint

\n[-0, +0, v]\n

int luaL_checkint (lua_State *L, int narg);
\n\n

\nChecks whether the function argument narg is a number\nand returns this number cast to an int.\n\n\n\n\n\n

luaL_checkinteger

\n[-0, +0, v]\n

lua_Integer luaL_checkinteger (lua_State *L, int narg);
\n\n

\nChecks whether the function argument narg is a number\nand returns this number cast to a lua_Integer.\n\n\n\n\n\n

luaL_checklong

\n[-0, +0, v]\n

long luaL_checklong (lua_State *L, int narg);
\n\n

\nChecks whether the function argument narg is a number\nand returns this number cast to a long.\n\n\n\n\n\n

luaL_checklstring

\n[-0, +0, v]\n

const char *luaL_checklstring (lua_State *L, int narg, size_t *l);
\n\n

\nChecks whether the function argument narg is a string\nand returns this string;\nif l is not NULL fills *l\nwith the string's length.\n\n\n

\nThis function uses lua_tolstring to get its result,\nso all conversions and caveats of that function apply here.\n\n\n\n\n\n

luaL_checknumber

\n[-0, +0, v]\n

lua_Number luaL_checknumber (lua_State *L, int narg);
\n\n

\nChecks whether the function argument narg is a number\nand returns this number.\n\n\n\n\n\n

luaL_checkoption

\n[-0, +0, v]\n

int luaL_checkoption (lua_State *L,\n                      int narg,\n                      const char *def,\n                      const char *const lst[]);
\n\n

\nChecks whether the function argument narg is a string and\nsearches for this string in the array lst\n(which must be NULL-terminated).\nReturns the index in the array where the string was found.\nRaises an error if the argument is not a string or\nif the string cannot be found.\n\n\n

\nIf def is not NULL,\nthe function uses def as a default value when\nthere is no argument narg or if this argument is nil.\n\n\n

\nThis is a useful function for mapping strings to C enums.\n(The usual convention in Lua libraries is\nto use strings instead of numbers to select options.)\n\n\n\n\n\n

luaL_checkstack

\n[-0, +0, v]\n

void luaL_checkstack (lua_State *L, int sz, const char *msg);
\n\n

\nGrows the stack size to top + sz elements,\nraising an error if the stack cannot grow to that size.\nmsg is an additional text to go into the error message.\n\n\n\n\n\n

luaL_checkstring

\n[-0, +0, v]\n

const char *luaL_checkstring (lua_State *L, int narg);
\n\n

\nChecks whether the function argument narg is a string\nand returns this string.\n\n\n

\nThis function uses lua_tolstring to get its result,\nso all conversions and caveats of that function apply here.\n\n\n\n\n\n

luaL_checktype

\n[-0, +0, v]\n

void luaL_checktype (lua_State *L, int narg, int t);
\n\n

\nChecks whether the function argument narg has type t.\nSee lua_type for the encoding of types for t.\n\n\n\n\n\n

luaL_checkudata

\n[-0, +0, v]\n

void *luaL_checkudata (lua_State *L, int narg, const char *tname);
\n\n

\nChecks whether the function argument narg is a userdata\nof the type tname (see luaL_newmetatable).\n\n\n\n\n\n

luaL_dofile

\n[-0, +?, m]\n

int luaL_dofile (lua_State *L, const char *filename);
\n\n

\nLoads and runs the given file.\nIt is defined as the following macro:\n\n

\n     (luaL_loadfile(L, filename) || lua_pcall(L, 0, LUA_MULTRET, 0))\n

\nIt returns 0 if there are no errors\nor 1 in case of errors.\n\n\n\n\n\n

luaL_dostring

\n[-0, +?, m]\n

int luaL_dostring (lua_State *L, const char *str);
\n\n

\nLoads and runs the given string.\nIt is defined as the following macro:\n\n

\n     (luaL_loadstring(L, str) || lua_pcall(L, 0, LUA_MULTRET, 0))\n

\nIt returns 0 if there are no errors\nor 1 in case of errors.\n\n\n\n\n\n

luaL_error

\n[-0, +0, v]\n

int luaL_error (lua_State *L, const char *fmt, ...);
\n\n

\nRaises an error.\nThe error message format is given by fmt\nplus any extra arguments,\nfollowing the same rules of lua_pushfstring.\nIt also adds at the beginning of the message the file name and\nthe line number where the error occurred,\nif this information is available.\n\n\n

\nThis function never returns,\nbut it is an idiom to use it in C functions\nas return luaL_error(args).\n\n\n\n\n\n

luaL_getmetafield

\n[-0, +(0|1), m]\n

int luaL_getmetafield (lua_State *L, int obj, const char *e);
\n\n

\nPushes onto the stack the field e from the metatable\nof the object at index obj.\nIf the object does not have a metatable,\nor if the metatable does not have this field,\nreturns 0 and pushes nothing.\n\n\n\n\n\n

luaL_getmetatable

\n[-0, +1, -]\n

void luaL_getmetatable (lua_State *L, const char *tname);
\n\n

\nPushes onto the stack the metatable associated with name tname\nin the registry (see luaL_newmetatable).\n\n\n\n\n\n

luaL_gsub

\n[-0, +1, m]\n

const char *luaL_gsub (lua_State *L,\n                       const char *s,\n                       const char *p,\n                       const char *r);
\n\n

\nCreates a copy of string s by replacing\nany occurrence of the string p\nwith the string r.\nPushes the resulting string on the stack and returns it.\n\n\n\n\n\n

luaL_loadbuffer

\n[-0, +1, m]\n

int luaL_loadbuffer (lua_State *L,\n                     const char *buff,\n                     size_t sz,\n                     const char *name);
\n\n

\nLoads a buffer as a Lua chunk.\nThis function uses lua_load to load the chunk in the\nbuffer pointed to by buff with size sz.\n\n\n

\nThis function returns the same results as lua_load.\nname is the chunk name,\nused for debug information and error messages.\n\n\n\n\n\n

luaL_loadfile

\n[-0, +1, m]\n

int luaL_loadfile (lua_State *L, const char *filename);
\n\n

\nLoads a file as a Lua chunk.\nThis function uses lua_load to load the chunk in the file\nnamed filename.\nIf filename is NULL,\nthen it loads from the standard input.\nThe first line in the file is ignored if it starts with a #.\n\n\n

\nThis function returns the same results as lua_load,\nbut it has an extra error code LUA_ERRFILE\nif it cannot open/read the file.\n\n\n

\nAs lua_load, this function only loads the chunk;\nit does not run it.\n\n\n\n\n\n

luaL_loadstring

\n[-0, +1, m]\n

int luaL_loadstring (lua_State *L, const char *s);
\n\n

\nLoads a string as a Lua chunk.\nThis function uses lua_load to load the chunk in\nthe zero-terminated string s.\n\n\n

\nThis function returns the same results as lua_load.\n\n\n

\nAlso as lua_load, this function only loads the chunk;\nit does not run it.\n\n\n\n\n\n

luaL_newmetatable

\n[-0, +1, m]\n

int luaL_newmetatable (lua_State *L, const char *tname);
\n\n

\nIf the registry already has the key tname,\nreturns 0.\nOtherwise,\ncreates a new table to be used as a metatable for userdata,\nadds it to the registry with key tname,\nand returns 1.\n\n\n

\nIn both cases pushes onto the stack the final value associated\nwith tname in the registry.\n\n\n\n\n\n

luaL_newstate

\n[-0, +0, -]\n

lua_State *luaL_newstate (void);
\n\n

\nCreates a new Lua state.\nIt calls lua_newstate with an\nallocator based on the standard C realloc function\nand then sets a panic function (see lua_atpanic) that prints\nan error message to the standard error output in case of fatal\nerrors.\n\n\n

\nReturns the new state,\nor NULL if there is a memory allocation error.\n\n\n\n\n\n

luaL_openlibs

\n[-0, +0, m]\n

void luaL_openlibs (lua_State *L);
\n\n

\nOpens all standard Lua libraries into the given state.\n\n\n\n\n\n

luaL_optint

\n[-0, +0, v]\n

int luaL_optint (lua_State *L, int narg, int d);
\n\n

\nIf the function argument narg is a number,\nreturns this number cast to an int.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n\n\n\n

luaL_optinteger

\n[-0, +0, v]\n

lua_Integer luaL_optinteger (lua_State *L,\n                             int narg,\n                             lua_Integer d);
\n\n

\nIf the function argument narg is a number,\nreturns this number cast to a lua_Integer.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n\n\n\n

luaL_optlong

\n[-0, +0, v]\n

long luaL_optlong (lua_State *L, int narg, long d);
\n\n

\nIf the function argument narg is a number,\nreturns this number cast to a long.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n\n\n\n

luaL_optlstring

\n[-0, +0, v]\n

const char *luaL_optlstring (lua_State *L,\n                             int narg,\n                             const char *d,\n                             size_t *l);
\n\n

\nIf the function argument narg is a string,\nreturns this string.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n

\nIf l is not NULL,\nfills the position *l with the results's length.\n\n\n\n\n\n

luaL_optnumber

\n[-0, +0, v]\n

lua_Number luaL_optnumber (lua_State *L, int narg, lua_Number d);
\n\n

\nIf the function argument narg is a number,\nreturns this number.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n\n\n\n

luaL_optstring

\n[-0, +0, v]\n

const char *luaL_optstring (lua_State *L,\n                            int narg,\n                            const char *d);
\n\n

\nIf the function argument narg is a string,\nreturns this string.\nIf this argument is absent or is nil,\nreturns d.\nOtherwise, raises an error.\n\n\n\n\n\n

luaL_prepbuffer

\n[-0, +0, -]\n

char *luaL_prepbuffer (luaL_Buffer *B);
\n\n

\nReturns an address to a space of size LUAL_BUFFERSIZE\nwhere you can copy a string to be added to buffer B\n(see luaL_Buffer).\nAfter copying the string into this space you must call\nluaL_addsize with the size of the string to actually add \nit to the buffer.\n\n\n\n\n\n

luaL_pushresult

\n[-?, +1, m]\n

void luaL_pushresult (luaL_Buffer *B);
\n\n

\nFinishes the use of buffer B leaving the final string on\nthe top of the stack.\n\n\n\n\n\n

luaL_ref

\n[-1, +0, m]\n

int luaL_ref (lua_State *L, int t);
\n\n

\nCreates and returns a reference,\nin the table at index t,\nfor the object at the top of the stack (and pops the object).\n\n\n

\nA reference is a unique integer key.\nAs long as you do not manually add integer keys into table t,\nluaL_ref ensures the uniqueness of the key it returns.\nYou can retrieve an object referred by reference r\nby calling lua_rawgeti(L, t, r).\nFunction luaL_unref frees a reference and its associated object.\n\n\n

\nIf the object at the top of the stack is nil,\nluaL_ref returns the constant LUA_REFNIL.\nThe constant LUA_NOREF is guaranteed to be different\nfrom any reference returned by luaL_ref.\n\n\n\n\n\n

luaL_Reg

\n
typedef struct luaL_Reg {\n  const char *name;\n  lua_CFunction func;\n} luaL_Reg;
\n\n

\nType for arrays of functions to be registered by\nluaL_register.\nname is the function name and func is a pointer to\nthe function.\nAny array of luaL_Reg must end with an sentinel entry\nin which both name and func are NULL.\n\n\n\n\n\n

luaL_register

\n[-(0|1), +1, m]\n

void luaL_register (lua_State *L,\n                    const char *libname,\n                    const luaL_Reg *l);
\n\n

\nOpens a library.\n\n\n

\nWhen called with libname equal to NULL,\nit simply registers all functions in the list l\n(see luaL_Reg) into the table on the top of the stack.\n\n\n

\nWhen called with a non-null libname,\nluaL_register creates a new table t,\nsets it as the value of the global variable libname,\nsets it as the value of package.loaded[libname],\nand registers on it all functions in the list l.\nIf there is a table in package.loaded[libname] or in\nvariable libname,\nreuses this table instead of creating a new one.\n\n\n

\nIn any case the function leaves the table\non the top of the stack.\n\n\n\n\n\n

luaL_typename

\n[-0, +0, -]\n

const char *luaL_typename (lua_State *L, int index);
\n\n

\nReturns the name of the type of the value at the given index.\n\n\n\n\n\n

luaL_typerror

\n[-0, +0, v]\n

int luaL_typerror (lua_State *L, int narg, const char *tname);
\n\n

\nGenerates an error with a message like the following:\n\n

\n     location: bad argument narg to 'func' (tname expected, got rt)\n

\nwhere location is produced by luaL_where,\nfunc is the name of the current function,\nand rt is the type name of the actual argument.\n\n\n\n\n\n

luaL_unref

\n[-0, +0, -]\n

void luaL_unref (lua_State *L, int t, int ref);
\n\n

\nReleases reference ref from the table at index t\n(see luaL_ref).\nThe entry is removed from the table,\nso that the referred object can be collected.\nThe reference ref is also freed to be used again.\n\n\n

\nIf ref is LUA_NOREF or LUA_REFNIL,\nluaL_unref does nothing.\n\n\n\n\n\n

luaL_where

\n[-0, +1, m]\n

void luaL_where (lua_State *L, int lvl);
\n\n

\nPushes onto the stack a string identifying the current position\nof the control at level lvl in the call stack.\nTypically this string has the following format:\n\n

\n     chunkname:currentline:\n

\nLevel 0 is the running function,\nlevel 1 is the function that called the running function,\netc.\n\n\n

\nThis function is used to build a prefix for error messages.\n\n\n\n\n\n\n\n

5 - Standard Libraries

\n\n

\nThe standard Lua libraries provide useful functions\nthat are implemented directly through the C API.\nSome of these functions provide essential services to the language\n(e.g., type and getmetatable);\nothers provide access to \"outside\" services (e.g., I/O);\nand others could be implemented in Lua itself,\nbut are quite useful or have critical performance requirements that\ndeserve an implementation in C (e.g., table.sort).\n\n\n

\nAll libraries are implemented through the official C API\nand are provided as separate C modules.\nCurrently, Lua has the following standard libraries:\n\n

    \n\n
  • basic library, which includes the coroutine sub-library;
    • \n\n
  • package library;
    • \n\n
  • string manipulation;
    • \n\n
  • table manipulation;
    • \n\n
  • mathematical functions (sin, log, etc.);
    • \n\n
  • input and output;
    • \n\n
  • operating system facilities;
    • \n\n
  • debug facilities.
    • \n\n

\nExcept for the basic and package libraries,\neach library provides all its functions as fields of a global table\nor as methods of its objects.\n\n\n

\nTo have access to these libraries,\nthe C host program should call the luaL_openlibs function,\nwhich opens all standard libraries.\nAlternatively,\nit can open them individually by calling\nluaopen_base (for the basic library),\nluaopen_package (for the package library),\nluaopen_string (for the string library),\nluaopen_table (for the table library),\nluaopen_math (for the mathematical library),\nluaopen_io (for the I/O library),\nluaopen_os (for the Operating System library),\nand luaopen_debug (for the debug library).\nThese functions are declared in lualib.h\nand should not be called directly:\nyou must call them like any other Lua C function,\ne.g., by using lua_call.\n\n\n\n

5.1 - Basic Functions

\n\n

\nThe basic library provides some core functions to Lua.\nIf you do not include this library in your application,\nyou should check carefully whether you need to provide \nimplementations for some of its facilities.\n\n\n

\n

assert (v [, message])

\nIssues an error when\nthe value of its argument v is false (i.e., nil or false);\notherwise, returns all its arguments.\nmessage is an error message;\nwhen absent, it defaults to \"assertion failed!\"\n\n\n\n\n

\n

collectgarbage ([opt [, arg]])

\n\n\n

\nThis function is a generic interface to the garbage collector.\nIt performs different functions according to its first argument, opt:\n\n

    \n\n
  • \"collect\":\nperforms a full garbage-collection cycle.\nThis is the default option.\n
    • \n\n
  • \"stop\":\nstops the garbage collector.\n
    • \n\n
  • \"restart\":\nrestarts the garbage collector.\n
    • \n\n
  • \"count\":\nreturns the total memory in use by Lua (in Kbytes).\n
    • \n\n
  • \"step\":\nperforms a garbage-collection step.\nThe step \"size\" is controlled by arg\n(larger values mean more steps) in a non-specified way.\nIf you want to control the step size\nyou must experimentally tune the value of arg.\nReturns true if the step finished a collection cycle.\n
    • \n\n
  • \"setpause\":\nsets arg as the new value for the pause of\nthe collector (see §2.10).\nReturns the previous value for pause.\n
    • \n\n
  • \"setstepmul\":\nsets arg as the new value for the step multiplier of\nthe collector (see §2.10).\nReturns the previous value for step.\n
    • \n\n
\n\n\n\n

\n

dofile ([filename])

\nOpens the named file and executes its contents as a Lua chunk.\nWhen called without arguments,\ndofile executes the contents of the standard input (stdin).\nReturns all values returned by the chunk.\nIn case of errors, dofile propagates the error\nto its caller (that is, dofile does not run in protected mode).\n\n\n\n\n

\n

error (message [, level])

\nTerminates the last protected function called\nand returns message as the error message.\nFunction error never returns.\n\n\n

\nUsually, error adds some information about the error position\nat the beginning of the message.\nThe level argument specifies how to get the error position.\nWith level 1 (the default), the error position is where the\nerror function was called.\nLevel 2 points the error to where the function\nthat called error was called; and so on.\nPassing a level 0 avoids the addition of error position information\nto the message.\n\n\n\n\n

\n

_G

\nA global variable (not a function) that\nholds the global environment (that is, _G._G = _G).\nLua itself does not use this variable;\nchanging its value does not affect any environment,\nnor vice-versa.\n(Use setfenv to change environments.)\n\n\n\n\n

\n

getfenv ([f])

\nReturns the current environment in use by the function.\nf can be a Lua function or a number\nthat specifies the function at that stack level:\nLevel 1 is the function calling getfenv.\nIf the given function is not a Lua function,\nor if f is 0,\ngetfenv returns the global environment.\nThe default for f is 1.\n\n\n\n\n

\n

getmetatable (object)

\n\n\n

\nIf object does not have a metatable, returns nil.\nOtherwise,\nif the object's metatable has a \"__metatable\" field,\nreturns the associated value.\nOtherwise, returns the metatable of the given object.\n\n\n\n\n

\n

ipairs (t)

\n\n\n

\nReturns three values: an iterator function, the table t, and 0,\nso that the construction\n\n

\n     for i,v in ipairs(t) do body end\n

\nwill iterate over the pairs (1,t[1]), (2,t[2]), ···,\nup to the first integer key absent from the table.\n\n\n\n\n

\n

load (func [, chunkname])

\n\n\n

\nLoads a chunk using function func to get its pieces.\nEach call to func must return a string that concatenates\nwith previous results.\nA return of an empty string, nil, or no value signals the end of the chunk.\n\n\n

\nIf there are no errors, \nreturns the compiled chunk as a function;\notherwise, returns nil plus the error message.\nThe environment of the returned function is the global environment.\n\n\n

\nchunkname is used as the chunk name for error messages\nand debug information.\nWhen absent,\nit defaults to \"=(load)\".\n\n\n\n\n

\n

loadfile ([filename])

\n\n\n

\nSimilar to load,\nbut gets the chunk from file filename\nor from the standard input,\nif no file name is given.\n\n\n\n\n

\n

loadstring (string [, chunkname])

\n\n\n

\nSimilar to load,\nbut gets the chunk from the given string.\n\n\n

\nTo load and run a given string, use the idiom\n\n

\n     assert(loadstring(s))()\n
\n\n

\nWhen absent,\nchunkname defaults to the given string.\n\n\n\n\n

\n

next (table [, index])

\n\n\n

\nAllows a program to traverse all fields of a table.\nIts first argument is a table and its second argument\nis an index in this table.\nnext returns the next index of the table\nand its associated value.\nWhen called with nil as its second argument,\nnext returns an initial index\nand its associated value.\nWhen called with the last index,\nor with nil in an empty table,\nnext returns nil.\nIf the second argument is absent, then it is interpreted as nil.\nIn particular,\nyou can use next(t) to check whether a table is empty.\n\n\n

\nThe order in which the indices are enumerated is not specified,\neven for numeric indices.\n(To traverse a table in numeric order,\nuse a numerical for or the ipairs function.)\n\n\n

\nThe behavior of next is undefined if,\nduring the traversal,\nyou assign any value to a non-existent field in the table.\nYou may however modify existing fields.\nIn particular, you may clear existing fields.\n\n\n\n\n

\n

pairs (t)

\n\n\n

\nReturns three values: the next function, the table t, and nil,\nso that the construction\n\n

\n     for k,v in pairs(t) do body end\n

\nwill iterate over all key–value pairs of table t.\n\n\n

\nSee function next for the caveats of modifying\nthe table during its traversal.\n\n\n\n\n

\n

pcall (f, arg1, ···)

\n\n\n

\nCalls function f with\nthe given arguments in protected mode.\nThis means that any error inside f is not propagated;\ninstead, pcall catches the error\nand returns a status code.\nIts first result is the status code (a boolean),\nwhich is true if the call succeeds without errors.\nIn such case, pcall also returns all results from the call,\nafter this first result.\nIn case of any error, pcall returns false plus the error message.\n\n\n\n\n

\n

print (···)

\nReceives any number of arguments,\nand prints their values to stdout,\nusing the tostring function to convert them to strings.\nprint is not intended for formatted output,\nbut only as a quick way to show a value,\ntypically for debugging.\nFor formatted output, use string.format.\n\n\n\n\n

\n

rawequal (v1, v2)

\nChecks whether v1 is equal to v2,\nwithout invoking any metamethod.\nReturns a boolean.\n\n\n\n\n

\n

rawget (table, index)

\nGets the real value of table[index],\nwithout invoking any metamethod.\ntable must be a table;\nindex may be any value.\n\n\n\n\n

\n

rawset (table, index, value)

\nSets the real value of table[index] to value,\nwithout invoking any metamethod.\ntable must be a table,\nindex any value different from nil,\nand value any Lua value.\n\n\n

\nThis function returns table.\n\n\n\n\n

\n

select (index, ···)

\n\n\n

\nIf index is a number,\nreturns all arguments after argument number index.\nOtherwise, index must be the string \"#\",\nand select returns the total number of extra arguments it received.\n\n\n\n\n

\n

setfenv (f, table)

\n\n\n

\nSets the environment to be used by the given function.\nf can be a Lua function or a number\nthat specifies the function at that stack level:\nLevel 1 is the function calling setfenv.\nsetfenv returns the given function.\n\n\n

\nAs a special case, when f is 0 setfenv changes\nthe environment of the running thread.\nIn this case, setfenv returns no values.\n\n\n\n\n

\n

setmetatable (table, metatable)

\n\n\n

\nSets the metatable for the given table.\n(You cannot change the metatable of other types from Lua, only from C.)\nIf metatable is nil,\nremoves the metatable of the given table.\nIf the original metatable has a \"__metatable\" field,\nraises an error.\n\n\n

\nThis function returns table.\n\n\n\n\n

\n

tonumber (e [, base])

\nTries to convert its argument to a number.\nIf the argument is already a number or a string convertible\nto a number, then tonumber returns this number;\notherwise, it returns nil.\n\n\n

\nAn optional argument specifies the base to interpret the numeral.\nThe base may be any integer between 2 and 36, inclusive.\nIn bases above 10, the letter 'A' (in either upper or lower case)\nrepresents 10, 'B' represents 11, and so forth,\nwith 'Z' representing 35.\nIn base 10 (the default), the number can have a decimal part,\nas well as an optional exponent part (see §2.1).\nIn other bases, only unsigned integers are accepted.\n\n\n\n\n

\n

tostring (e)

\nReceives an argument of any type and\nconverts it to a string in a reasonable format.\nFor complete control of how numbers are converted,\nuse string.format.\n\n\n

\nIf the metatable of e has a \"__tostring\" field,\nthen tostring calls the corresponding value\nwith e as argument,\nand uses the result of the call as its result.\n\n\n\n\n

\n

type (v)

\nReturns the type of its only argument, coded as a string.\nThe possible results of this function are\n\"nil\" (a string, not the value nil),\n\"number\",\n\"string\",\n\"boolean\",\n\"table\",\n\"function\",\n\"thread\",\nand \"userdata\".\n\n\n\n\n

\n

unpack (list [, i [, j]])

\nReturns the elements from the given table.\nThis function is equivalent to\n\n
\n     return list[i], list[i+1], ···, list[j]\n

\nexcept that the above code can be written only for a fixed number\nof elements.\nBy default, i is 1 and j is the length of the list,\nas defined by the length operator (see §2.5.5).\n\n\n\n\n

\n

_VERSION

\nA global variable (not a function) that\nholds a string containing the current interpreter version.\nThe current contents of this variable is \"Lua 5.1\".\n\n\n\n\n

\n

xpcall (f, err)

\n\n\n

\nThis function is similar to pcall,\nexcept that you can set a new error handler.\n\n\n

\nxpcall calls function f in protected mode,\nusing err as the error handler.\nAny error inside f is not propagated;\ninstead, xpcall catches the error,\ncalls the err function with the original error object,\nand returns a status code.\nIts first result is the status code (a boolean),\nwhich is true if the call succeeds without errors.\nIn this case, xpcall also returns all results from the call,\nafter this first result.\nIn case of any error,\nxpcall returns false plus the result from err.\n\n\n\n\n\n\n\n

5.2 - Coroutine Manipulation

\n\n

\nThe operations related to coroutines comprise a sub-library of\nthe basic library and come inside the table coroutine.\nSee §2.11 for a general description of coroutines.\n\n\n

\n

coroutine.create (f)

\n\n\n

\nCreates a new coroutine, with body f.\nf must be a Lua function.\nReturns this new coroutine,\nan object with type \"thread\".\n\n\n\n\n

\n

coroutine.resume (co [, val1, ···])

\n\n\n

\nStarts or continues the execution of coroutine co.\nThe first time you resume a coroutine,\nit starts running its body.\nThe values val1, ··· are passed\nas the arguments to the body function.\nIf the coroutine has yielded,\nresume restarts it;\nthe values val1, ··· are passed\nas the results from the yield.\n\n\n

\nIf the coroutine runs without any errors,\nresume returns true plus any values passed to yield\n(if the coroutine yields) or any values returned by the body function\n(if the coroutine terminates).\nIf there is any error,\nresume returns false plus the error message.\n\n\n\n\n

\n

coroutine.running ()

\n\n\n

\nReturns the running coroutine,\nor nil when called by the main thread.\n\n\n\n\n

\n

coroutine.status (co)

\n\n\n

\nReturns the status of coroutine co, as a string:\n\"running\",\nif the coroutine is running (that is, it called status);\n\"suspended\", if the coroutine is suspended in a call to yield,\nor if it has not started running yet;\n\"normal\" if the coroutine is active but not running\n(that is, it has resumed another coroutine);\nand \"dead\" if the coroutine has finished its body function,\nor if it has stopped with an error.\n\n\n\n\n

\n

coroutine.wrap (f)

\n\n\n

\nCreates a new coroutine, with body f.\nf must be a Lua function.\nReturns a function that resumes the coroutine each time it is called.\nAny arguments passed to the function behave as the\nextra arguments to resume.\nReturns the same values returned by resume,\nexcept the first boolean.\nIn case of error, propagates the error.\n\n\n\n\n

\n

coroutine.yield (···)

\n\n\n

\nSuspends the execution of the calling coroutine.\nThe coroutine cannot be running a C function,\na metamethod, or an iterator.\nAny arguments to yield are passed as extra results to resume.\n\n\n\n\n\n\n\n

5.3 - Modules

\n\n

\nThe package library provides basic\nfacilities for loading and building modules in Lua.\nIt exports two of its functions directly in the global environment:\nrequire and module.\nEverything else is exported in a table package.\n\n\n

\n

module (name [, ···])

\n\n\n

\nCreates a module.\nIf there is a table in package.loaded[name],\nthis table is the module.\nOtherwise, if there is a global table t with the given name,\nthis table is the module.\nOtherwise creates a new table t and\nsets it as the value of the global name and\nthe value of package.loaded[name].\nThis function also initializes t._NAME with the given name,\nt._M with the module (t itself),\nand t._PACKAGE with the package name\n(the full module name minus last component; see below).\nFinally, module sets t as the new environment\nof the current function and the new value of package.loaded[name],\nso that require returns t.\n\n\n

\nIf name is a compound name\n(that is, one with components separated by dots),\nmodule creates (or reuses, if they already exist)\ntables for each component.\nFor instance, if name is a.b.c,\nthen module stores the module table in field c of\nfield b of global a.\n\n\n

\nThis function can receive optional options after\nthe module name,\nwhere each option is a function to be applied over the module.\n\n\n\n\n

\n

require (modname)

\n\n\n

\nLoads the given module.\nThe function starts by looking into the package.loaded table\nto determine whether modname is already loaded.\nIf it is, then require returns the value stored\nat package.loaded[modname].\nOtherwise, it tries to find a loader for the module.\n\n\n

\nTo find a loader,\nrequire is guided by the package.loaders array.\nBy changing this array,\nwe can change how require looks for a module.\nThe following explanation is based on the default configuration\nfor package.loaders.\n\n\n

\nFirst require queries package.preload[modname].\nIf it has a value,\nthis value (which should be a function) is the loader.\nOtherwise require searches for a Lua loader using the\npath stored in package.path.\nIf that also fails, it searches for a C loader using the\npath stored in package.cpath.\nIf that also fails,\nit tries an all-in-one loader (see package.loaders).\n\n\n

\nOnce a loader is found,\nrequire calls the loader with a single argument, modname.\nIf the loader returns any value,\nrequire assigns the returned value to package.loaded[modname].\nIf the loader returns no value and\nhas not assigned any value to package.loaded[modname],\nthen require assigns true to this entry.\nIn any case, require returns the\nfinal value of package.loaded[modname].\n\n\n

\nIf there is any error loading or running the module,\nor if it cannot find any loader for the module,\nthen require signals an error. \n\n\n\n\n

\n

package.cpath

\n\n\n

\nThe path used by require to search for a C loader.\n\n\n

\nLua initializes the C path package.cpath in the same way\nit initializes the Lua path package.path,\nusing the environment variable LUA_CPATH\nor a default path defined in luaconf.h.\n\n\n\n\n

\n\n

package.loaded

\n\n\n

\nA table used by require to control which\nmodules are already loaded.\nWhen you require a module modname and\npackage.loaded[modname] is not false,\nrequire simply returns the value stored there.\n\n\n\n\n

\n

package.loaders

\n\n\n

\nA table used by require to control how to load modules.\n\n\n

\nEach entry in this table is a searcher function.\nWhen looking for a module,\nrequire calls each of these searchers in ascending order,\nwith the module name (the argument given to require) as its\nsole parameter.\nThe function can return another function (the module loader)\nor a string explaining why it did not find that module\n(or nil if it has nothing to say).\nLua initializes this table with four functions.\n\n\n

\nThe first searcher simply looks for a loader in the\npackage.preload table.\n\n\n

\nThe second searcher looks for a loader as a Lua library,\nusing the path stored at package.path.\nA path is a sequence of templates separated by semicolons.\nFor each template,\nthe searcher will change each interrogation\nmark in the template by filename,\nwhich is the module name with each dot replaced by a\n\"directory separator\" (such as \"/\" in Unix);\nthen it will try to open the resulting file name.\nSo, for instance, if the Lua path is the string\n\n

\n     \"./?.lua;./?.lc;/usr/local/?/init.lua\"\n

\nthe search for a Lua file for module foo\nwill try to open the files\n./foo.lua, ./foo.lc, and\n/usr/local/foo/init.lua, in that order.\n\n\n

\nThe third searcher looks for a loader as a C library,\nusing the path given by the variable package.cpath.\nFor instance,\nif the C path is the string\n\n

\n     \"./?.so;./?.dll;/usr/local/?/init.so\"\n

\nthe searcher for module foo\nwill try to open the files ./foo.so, ./foo.dll,\nand /usr/local/foo/init.so, in that order.\nOnce it finds a C library,\nthis searcher first uses a dynamic link facility to link the\napplication with the library.\nThen it tries to find a C function inside the library to\nbe used as the loader.\nThe name of this C function is the string \"luaopen_\"\nconcatenated with a copy of the module name where each dot\nis replaced by an underscore.\nMoreover, if the module name has a hyphen,\nits prefix up to (and including) the first hyphen is removed.\nFor instance, if the module name is a.v1-b.c,\nthe function name will be luaopen_b_c.\n\n\n

\nThe fourth searcher tries an all-in-one loader.\nIt searches the C path for a library for\nthe root name of the given module.\nFor instance, when requiring a.b.c,\nit will search for a C library for a.\nIf found, it looks into it for an open function for\nthe submodule;\nin our example, that would be luaopen_a_b_c.\nWith this facility, a package can pack several C submodules\ninto one single library,\nwith each submodule keeping its original open function.\n\n\n\n\n

\n

package.loadlib (libname, funcname)

\n\n\n

\nDynamically links the host program with the C library libname.\nInside this library, looks for a function funcname\nand returns this function as a C function.\n(So, funcname must follow the protocol (see lua_CFunction)).\n\n\n

\nThis is a low-level function.\nIt completely bypasses the package and module system.\nUnlike require,\nit does not perform any path searching and\ndoes not automatically adds extensions.\nlibname must be the complete file name of the C library,\nincluding if necessary a path and extension.\nfuncname must be the exact name exported by the C library\n(which may depend on the C compiler and linker used).\n\n\n

\nThis function is not supported by ANSI C.\nAs such, it is only available on some platforms\n(Windows, Linux, Mac OS X, Solaris, BSD,\nplus other Unix systems that support the dlfcn standard).\n\n\n\n\n

\n

package.path

\n\n\n

\nThe path used by require to search for a Lua loader.\n\n\n

\nAt start-up, Lua initializes this variable with\nthe value of the environment variable LUA_PATH or\nwith a default path defined in luaconf.h,\nif the environment variable is not defined.\nAny \";;\" in the value of the environment variable\nis replaced by the default path.\n\n\n\n\n

\n

package.preload

\n\n\n

\nA table to store loaders for specific modules\n(see require).\n\n\n\n\n

\n

package.seeall (module)

\n\n\n

\nSets a metatable for module with\nits __index field referring to the global environment,\nso that this module inherits values\nfrom the global environment.\nTo be used as an option to function module.\n\n\n\n\n\n\n\n

5.4 - String Manipulation

\n\n

\nThis library provides generic functions for string manipulation,\nsuch as finding and extracting substrings, and pattern matching.\nWhen indexing a string in Lua, the first character is at position 1\n(not at 0, as in C).\nIndices are allowed to be negative and are interpreted as indexing backwards,\nfrom the end of the string.\nThus, the last character is at position -1, and so on.\n\n\n

\nThe string library provides all its functions inside the table\nstring.\nIt also sets a metatable for strings\nwhere the __index field points to the string table.\nTherefore, you can use the string functions in object-oriented style.\nFor instance, string.byte(s, i)\ncan be written as s:byte(i).\n\n\n

\nThe string library assumes one-byte character encodings.\n\n\n

\n

string.byte (s [, i [, j]])

\nReturns the internal numerical codes of the characters s[i],\ns[i+1], ···, s[j].\nThe default value for i is 1;\nthe default value for j is i.\n\n\n

\nNote that numerical codes are not necessarily portable across platforms.\n\n\n\n\n

\n

string.char (···)

\nReceives zero or more integers.\nReturns a string with length equal to the number of arguments,\nin which each character has the internal numerical code equal\nto its corresponding argument.\n\n\n

\nNote that numerical codes are not necessarily portable across platforms.\n\n\n\n\n

\n

string.dump (function)

\n\n\n

\nReturns a string containing a binary representation of the given function,\nso that a later loadstring on this string returns\na copy of the function.\nfunction must be a Lua function without upvalues.\n\n\n\n\n

\n

string.find (s, pattern [, init [, plain]])

\nLooks for the first match of\npattern in the string s.\nIf it finds a match, then find returns the indices of s\nwhere this occurrence starts and ends;\notherwise, it returns nil.\nA third, optional numerical argument init specifies\nwhere to start the search;\nits default value is 1 and can be negative.\nA value of true as a fourth, optional argument plain\nturns off the pattern matching facilities,\nso the function does a plain \"find substring\" operation,\nwith no characters in pattern being considered \"magic\".\nNote that if plain is given, then init must be given as well.\n\n\n

\nIf the pattern has captures,\nthen in a successful match\nthe captured values are also returned,\nafter the two indices.\n\n\n\n\n

\n

string.format (formatstring, ···)

\nReturns a formatted version of its variable number of arguments\nfollowing the description given in its first argument (which must be a string).\nThe format string follows the same rules as the printf family of\nstandard C functions.\nThe only differences are that the options/modifiers\n*, l, L, n, p,\nand h are not supported\nand that there is an extra option, q.\nThe q option formats a string in a form suitable to be safely read\nback by the Lua interpreter:\nthe string is written between double quotes,\nand all double quotes, newlines, embedded zeros,\nand backslashes in the string\nare correctly escaped when written.\nFor instance, the call\n\n
\n     string.format('%q', 'a string with \"quotes\" and \\n new line')\n

\nwill produce the string:\n\n

\n     \"a string with \\\"quotes\\\" and \\\n      new line\"\n
\n\n

\nThe options c, d, E, e, f,\ng, G, i, o, u, X, and x all\nexpect a number as argument,\nwhereas q and s expect a string.\n\n\n

\nThis function does not accept string values\ncontaining embedded zeros,\nexcept as arguments to the q option.\n\n\n\n\n

\n

string.gmatch (s, pattern)

\nReturns an iterator function that,\neach time it is called,\nreturns the next captures from pattern over string s.\nIf pattern specifies no captures,\nthen the whole match is produced in each call.\n\n\n

\nAs an example, the following loop\n\n

\n     s = \"hello world from Lua\"\n     for w in string.gmatch(s, \"%a+\") do\n       print(w)\n     end\n

\nwill iterate over all the words from string s,\nprinting one per line.\nThe next example collects all pairs key=value from the\ngiven string into a table:\n\n

\n     t = {}\n     s = \"from=world, to=Lua\"\n     for k, v in string.gmatch(s, \"(%w+)=(%w+)\") do\n       t[k] = v\n     end\n
\n\n

\nFor this function, a '^' at the start of a pattern does not\nwork as an anchor, as this would prevent the iteration.\n\n\n\n\n

\n

string.gsub (s, pattern, repl [, n])

\nReturns a copy of s\nin which all (or the first n, if given)\noccurrences of the pattern have been\nreplaced by a replacement string specified by repl,\nwhich can be a string, a table, or a function.\ngsub also returns, as its second value,\nthe total number of matches that occurred.\n\n\n

\nIf repl is a string, then its value is used for replacement.\nThe character % works as an escape character:\nany sequence in repl of the form %n,\nwith n between 1 and 9,\nstands for the value of the n-th captured substring (see below).\nThe sequence %0 stands for the whole match.\nThe sequence %% stands for a single %.\n\n\n

\nIf repl is a table, then the table is queried for every match,\nusing the first capture as the key;\nif the pattern specifies no captures,\nthen the whole match is used as the key.\n\n\n

\nIf repl is a function, then this function is called every time a\nmatch occurs, with all captured substrings passed as arguments,\nin order;\nif the pattern specifies no captures,\nthen the whole match is passed as a sole argument.\n\n\n

\nIf the value returned by the table query or by the function call\nis a string or a number,\nthen it is used as the replacement string;\notherwise, if it is false or nil,\nthen there is no replacement\n(that is, the original match is kept in the string).\n\n\n

\nHere are some examples:\n\n

\n     x = string.gsub(\"hello world\", \"(%w+)\", \"%1 %1\")\n     --> x=\"hello hello world world\"\n     \n     x = string.gsub(\"hello world\", \"%w+\", \"%0 %0\", 1)\n     --> x=\"hello hello world\"\n     \n     x = string.gsub(\"hello world from Lua\", \"(%w+)%s*(%w+)\", \"%2 %1\")\n     --> x=\"world hello Lua from\"\n     \n     x = string.gsub(\"home = $HOME, user = $USER\", \"%$(%w+)\", os.getenv)\n     --> x=\"home = /home/roberto, user = roberto\"\n     \n     x = string.gsub(\"4+5 = $return 4+5$\", \"%$(.-)%$\", function (s)\n           return loadstring(s)()\n         end)\n     --> x=\"4+5 = 9\"\n     \n     local t = {name=\"lua\", version=\"5.1\"}\n     x = string.gsub(\"$name-$version.tar.gz\", \"%$(%w+)\", t)\n     --> x=\"lua-5.1.tar.gz\"\n
\n\n\n\n

\n

string.len (s)

\nReceives a string and returns its length.\nThe empty string \"\" has length 0.\nEmbedded zeros are counted,\nso \"a\\000bc\\000\" has length 5.\n\n\n\n\n

\n

string.lower (s)

\nReceives a string and returns a copy of this string with all\nuppercase letters changed to lowercase.\nAll other characters are left unchanged.\nThe definition of what an uppercase letter is depends on the current locale.\n\n\n\n\n

\n

string.match (s, pattern [, init])

\nLooks for the first match of\npattern in the string s.\nIf it finds one, then match returns\nthe captures from the pattern;\notherwise it returns nil.\nIf pattern specifies no captures,\nthen the whole match is returned.\nA third, optional numerical argument init specifies\nwhere to start the search;\nits default value is 1 and can be negative.\n\n\n\n\n

\n

string.rep (s, n)

\nReturns a string that is the concatenation of n copies of\nthe string s.\n\n\n\n\n

\n

string.reverse (s)

\nReturns a string that is the string s reversed.\n\n\n\n\n

\n

string.sub (s, i [, j])

\nReturns the substring of s that\nstarts at i and continues until j;\ni and j can be negative.\nIf j is absent, then it is assumed to be equal to -1\n(which is the same as the string length).\nIn particular,\nthe call string.sub(s,1,j) returns a prefix of s\nwith length j,\nand string.sub(s, -i) returns a suffix of s\nwith length i.\n\n\n\n\n

\n

string.upper (s)

\nReceives a string and returns a copy of this string with all\nlowercase letters changed to uppercase.\nAll other characters are left unchanged.\nThe definition of what a lowercase letter is depends on the current locale.\n\n\n\n

5.4.1 - Patterns

\n\n\n

Character Class:

\nA character class is used to represent a set of characters.\nThe following combinations are allowed in describing a character class:\n\n

    \n\n
  • x:\n(where x is not one of the magic characters\n^$()%.[]*+-?)\nrepresents the character x itself.\n
    • \n\n
  • .: (a dot) represents all characters.
    • \n\n
  • %a: represents all letters.
    • \n\n
  • %c: represents all control characters.
    • \n\n
  • %d: represents all digits.
    • \n\n
  • %l: represents all lowercase letters.
    • \n\n
  • %p: represents all punctuation characters.
    • \n\n
  • %s: represents all space characters.
    • \n\n
  • %u: represents all uppercase letters.
    • \n\n
  • %w: represents all alphanumeric characters.
    • \n\n
  • %x: represents all hexadecimal digits.
    • \n\n
  • %z: represents the character with representation 0.
    • \n\n
  • %x: (where x is any non-alphanumeric character)\nrepresents the character x.\nThis is the standard way to escape the magic characters.\nAny punctuation character (even the non magic)\ncan be preceded by a '%'\nwhen used to represent itself in a pattern.\n
    • \n\n
  • [set]:\nrepresents the class which is the union of all\ncharacters in set.\nA range of characters can be specified by\nseparating the end characters of the range with a '-'.\nAll classes %x described above can also be used as\ncomponents in set.\nAll other characters in set represent themselves.\nFor example, [%w_] (or [_%w])\nrepresents all alphanumeric characters plus the underscore,\n[0-7] represents the octal digits,\nand [0-7%l%-] represents the octal digits plus\nthe lowercase letters plus the '-' character.\n\n\n

    \nThe interaction between ranges and classes is not defined.\nTherefore, patterns like [%a-z] or [a-%%]\nhave no meaning.\n

    • \n\n
  • [^set]:\nrepresents the complement of set,\nwhere set is interpreted as above.\n
    • \n\n

\nFor all classes represented by single letters (%a, %c, etc.),\nthe corresponding uppercase letter represents the complement of the class.\nFor instance, %S represents all non-space characters.\n\n\n

\nThe definitions of letter, space, and other character groups\ndepend on the current locale.\nIn particular, the class [a-z] may not be equivalent to %l.\n\n\n\n\n\n

Pattern Item:

\nA pattern item can be\n\n

    \n\n
  • \na single character class,\nwhich matches any single character in the class;\n
    • \n\n
  • \na single character class followed by '*',\nwhich matches 0 or more repetitions of characters in the class.\nThese repetition items will always match the longest possible sequence;\n
    • \n\n
  • \na single character class followed by '+',\nwhich matches 1 or more repetitions of characters in the class.\nThese repetition items will always match the longest possible sequence;\n
    • \n\n
  • \na single character class followed by '-',\nwhich also matches 0 or more repetitions of characters in the class.\nUnlike '*',\nthese repetition items will always match the shortest possible sequence;\n
    • \n\n
  • \na single character class followed by '?',\nwhich matches 0 or 1 occurrence of a character in the class;\n
    • \n\n
  • \n%n, for n between 1 and 9;\nsuch item matches a substring equal to the n-th captured string\n(see below);\n
    • \n\n
  • \n%bxy, where x and y are two distinct characters;\nsuch item matches strings that start with x, end with y,\nand where the x and y are balanced.\nThis means that, if one reads the string from left to right,\ncounting +1 for an x and -1 for a y,\nthe ending y is the first y where the count reaches 0.\nFor instance, the item %b() matches expressions with\nbalanced parentheses.\n
    • \n\n
\n\n\n\n\n

Pattern:

\nA pattern is a sequence of pattern items.\nA '^' at the beginning of a pattern anchors the match at the\nbeginning of the subject string.\nA '$' at the end of a pattern anchors the match at the\nend of the subject string.\nAt other positions,\n'^' and '$' have no special meaning and represent themselves.\n\n\n\n\n\n

Captures:

\nA pattern can contain sub-patterns enclosed in parentheses;\nthey describe captures.\nWhen a match succeeds, the substrings of the subject string\nthat match captures are stored (captured) for future use.\nCaptures are numbered according to their left parentheses.\nFor instance, in the pattern \"(a*(.)%w(%s*))\",\nthe part of the string matching \"a*(.)%w(%s*)\" is\nstored as the first capture (and therefore has number 1);\nthe character matching \".\" is captured with number 2,\nand the part matching \"%s*\" has number 3.\n\n\n

\nAs a special case, the empty capture () captures\nthe current string position (a number).\nFor instance, if we apply the pattern \"()aa()\" on the\nstring \"flaaap\", there will be two captures: 3 and 5.\n\n\n

\nA pattern cannot contain embedded zeros. Use %z instead.\n\n\n\n\n\n\n\n\n\n\n\n

5.5 - Table Manipulation

\nThis library provides generic functions for table manipulation.\nIt provides all its functions inside the table table.\n\n\n

\nMost functions in the table library assume that the table\nrepresents an array or a list.\nFor these functions, when we talk about the \"length\" of a table\nwe mean the result of the length operator.\n\n\n

\n

table.concat (table [, sep [, i [, j]]])

\nGiven an array where all elements are strings or numbers,\nreturns table[i]..sep..table[i+1] ··· sep..table[j].\nThe default value for sep is the empty string,\nthe default for i is 1,\nand the default for j is the length of the table.\nIf i is greater than j, returns the empty string.\n\n\n\n\n

\n

table.insert (table, [pos,] value)

\n\n\n

\nInserts element value at position pos in table,\nshifting up other elements to open space, if necessary.\nThe default value for pos is n+1,\nwhere n is the length of the table (see §2.5.5),\nso that a call table.insert(t,x) inserts x at the end\nof table t.\n\n\n\n\n

\n

table.maxn (table)

\n\n\n

\nReturns the largest positive numerical index of the given table,\nor zero if the table has no positive numerical indices.\n(To do its job this function does a linear traversal of\nthe whole table.) \n\n\n\n\n

\n

table.remove (table [, pos])

\n\n\n

\nRemoves from table the element at position pos,\nshifting down other elements to close the space, if necessary.\nReturns the value of the removed element.\nThe default value for pos is n,\nwhere n is the length of the table,\nso that a call table.remove(t) removes the last element\nof table t.\n\n\n\n\n

\n

table.sort (table [, comp])

\nSorts table elements in a given order, in-place,\nfrom table[1] to table[n],\nwhere n is the length of the table.\nIf comp is given,\nthen it must be a function that receives two table elements,\nand returns true\nwhen the first is less than the second\n(so that not comp(a[i+1],a[i]) will be true after the sort).\nIf comp is not given,\nthen the standard Lua operator < is used instead.\n\n\n

\nThe sort algorithm is not stable;\nthat is, elements considered equal by the given order\nmay have their relative positions changed by the sort.\n\n\n\n\n\n\n\n

5.6 - Mathematical Functions

\n\n

\nThis library is an interface to the standard C math library.\nIt provides all its functions inside the table math.\n\n\n

\n

math.abs (x)

\n\n\n

\nReturns the absolute value of x.\n\n\n\n\n

\n

math.acos (x)

\n\n\n

\nReturns the arc cosine of x (in radians).\n\n\n\n\n

\n

math.asin (x)

\n\n\n

\nReturns the arc sine of x (in radians).\n\n\n\n\n

\n

math.atan (x)

\n\n\n

\nReturns the arc tangent of x (in radians).\n\n\n\n\n

\n

math.atan2 (y, x)

\n\n\n

\nReturns the arc tangent of y/x (in radians),\nbut uses the signs of both parameters to find the\nquadrant of the result.\n(It also handles correctly the case of x being zero.)\n\n\n\n\n

\n

math.ceil (x)

\n\n\n

\nReturns the smallest integer larger than or equal to x.\n\n\n\n\n

\n

math.cos (x)

\n\n\n

\nReturns the cosine of x (assumed to be in radians).\n\n\n\n\n

\n

math.cosh (x)

\n\n\n

\nReturns the hyperbolic cosine of x.\n\n\n\n\n

\n

math.deg (x)

\n\n\n

\nReturns the angle x (given in radians) in degrees.\n\n\n\n\n

\n

math.exp (x)

\n\n\n

\nReturns the value ex.\n\n\n\n\n

\n

math.floor (x)

\n\n\n

\nReturns the largest integer smaller than or equal to x.\n\n\n\n\n

\n

math.fmod (x, y)

\n\n\n

\nReturns the remainder of the division of x by y\nthat rounds the quotient towards zero.\n\n\n\n\n

\n

math.frexp (x)

\n\n\n

\nReturns m and e such that x = m2e,\ne is an integer and the absolute value of m is\nin the range [0.5, 1)\n(or zero when x is zero).\n\n\n\n\n

\n

math.huge

\n\n\n

\nThe value HUGE_VAL,\na value larger than or equal to any other numerical value.\n\n\n\n\n

\n

math.ldexp (m, e)

\n\n\n

\nReturns m2e (e should be an integer).\n\n\n\n\n

\n

math.log (x)

\n\n\n

\nReturns the natural logarithm of x.\n\n\n\n\n

\n

math.log10 (x)

\n\n\n

\nReturns the base-10 logarithm of x.\n\n\n\n\n

\n

math.max (x, ···)

\n\n\n

\nReturns the maximum value among its arguments.\n\n\n\n\n

\n

math.min (x, ···)

\n\n\n

\nReturns the minimum value among its arguments.\n\n\n\n\n

\n

math.modf (x)

\n\n\n

\nReturns two numbers,\nthe integral part of x and the fractional part of x.\n\n\n\n\n

\n

math.pi

\n\n\n

\nThe value of pi.\n\n\n\n\n

\n

math.pow (x, y)

\n\n\n

\nReturns xy.\n(You can also use the expression x^y to compute this value.)\n\n\n\n\n

\n

math.rad (x)

\n\n\n

\nReturns the angle x (given in degrees) in radians.\n\n\n\n\n

\n

math.random ([m [, n]])

\n\n\n

\nThis function is an interface to the simple\npseudo-random generator function rand provided by ANSI C.\n(No guarantees can be given for its statistical properties.)\n\n\n

\nWhen called without arguments,\nreturns a uniform pseudo-random real number\nin the range [0,1). \nWhen called with an integer number m,\nmath.random returns\na uniform pseudo-random integer in the range [1, m].\nWhen called with two integer numbers m and n,\nmath.random returns a uniform pseudo-random\ninteger in the range [m, n].\n\n\n\n\n

\n

math.randomseed (x)

\n\n\n

\nSets x as the \"seed\"\nfor the pseudo-random generator:\nequal seeds produce equal sequences of numbers.\n\n\n\n\n

\n

math.sin (x)

\n\n\n

\nReturns the sine of x (assumed to be in radians).\n\n\n\n\n

\n

math.sinh (x)

\n\n\n

\nReturns the hyperbolic sine of x.\n\n\n\n\n

\n

math.sqrt (x)

\n\n\n

\nReturns the square root of x.\n(You can also use the expression x^0.5 to compute this value.)\n\n\n\n\n

\n

math.tan (x)

\n\n\n

\nReturns the tangent of x (assumed to be in radians).\n\n\n\n\n

\n

math.tanh (x)

\n\n\n

\nReturns the hyperbolic tangent of x.\n\n\n\n\n\n\n\n

5.7 - Input and Output Facilities

\n\n

\nThe I/O library provides two different styles for file manipulation.\nThe first one uses implicit file descriptors;\nthat is, there are operations to set a default input file and a\ndefault output file,\nand all input/output operations are over these default files.\nThe second style uses explicit file descriptors.\n\n\n

\nWhen using implicit file descriptors,\nall operations are supplied by table io.\nWhen using explicit file descriptors,\nthe operation io.open returns a file descriptor\nand then all operations are supplied as methods of the file descriptor.\n\n\n

\nThe table io also provides\nthree predefined file descriptors with their usual meanings from C:\nio.stdin, io.stdout, and io.stderr.\nThe I/O library never closes these files.\n\n\n

\nUnless otherwise stated,\nall I/O functions return nil on failure\n(plus an error message as a second result and\na system-dependent error code as a third result)\nand some value different from nil on success.\n\n\n

\n

io.close ([file])

\n\n\n

\nEquivalent to file:close().\nWithout a file, closes the default output file.\n\n\n\n\n

\n

io.flush ()

\n\n\n

\nEquivalent to file:flush over the default output file.\n\n\n\n\n

\n

io.input ([file])

\n\n\n

\nWhen called with a file name, it opens the named file (in text mode),\nand sets its handle as the default input file.\nWhen called with a file handle,\nit simply sets this file handle as the default input file.\nWhen called without parameters,\nit returns the current default input file.\n\n\n

\nIn case of errors this function raises the error,\ninstead of returning an error code.\n\n\n\n\n

\n

io.lines ([filename])

\n\n\n

\nOpens the given file name in read mode\nand returns an iterator function that,\neach time it is called,\nreturns a new line from the file.\nTherefore, the construction\n\n

\n     for line in io.lines(filename) do body end\n

\nwill iterate over all lines of the file.\nWhen the iterator function detects the end of file,\nit returns nil (to finish the loop) and automatically closes the file.\n\n\n

\nThe call io.lines() (with no file name) is equivalent\nto io.input():lines();\nthat is, it iterates over the lines of the default input file.\nIn this case it does not close the file when the loop ends.\n\n\n\n\n

\n

io.open (filename [, mode])

\n\n\n

\nThis function opens a file,\nin the mode specified in the string mode.\nIt returns a new file handle,\nor, in case of errors, nil plus an error message.\n\n\n

\nThe mode string can be any of the following:\n\n

    \n
  • \"r\": read mode (the default);
    • \n
  • \"w\": write mode;
    • \n
  • \"a\": append mode;
    • \n
  • \"r+\": update mode, all previous data is preserved;
    • \n
  • \"w+\": update mode, all previous data is erased;
    • \n
  • \"a+\": append update mode, previous data is preserved,\n writing is only allowed at the end of file.
    • \n

\nThe mode string can also have a 'b' at the end,\nwhich is needed in some systems to open the file in binary mode.\nThis string is exactly what is used in the\nstandard C function fopen.\n\n\n\n\n

\n

io.output ([file])

\n\n\n

\nSimilar to io.input, but operates over the default output file.\n\n\n\n\n

\n

io.popen (prog [, mode])

\n\n\n

\nStarts program prog in a separated process and returns\na file handle that you can use to read data from this program\n(if mode is \"r\", the default)\nor to write data to this program\n(if mode is \"w\").\n\n\n

\nThis function is system dependent and is not available\non all platforms.\n\n\n\n\n

\n

io.read (···)

\n\n\n

\nEquivalent to io.input():read.\n\n\n\n\n

\n

io.tmpfile ()

\n\n\n

\nReturns a handle for a temporary file.\nThis file is opened in update mode\nand it is automatically removed when the program ends.\n\n\n\n\n

\n

io.type (obj)

\n\n\n

\nChecks whether obj is a valid file handle.\nReturns the string \"file\" if obj is an open file handle,\n\"closed file\" if obj is a closed file handle,\nor nil if obj is not a file handle.\n\n\n\n\n

\n

io.write (···)

\n\n\n

\nEquivalent to io.output():write.\n\n\n\n\n

\n

file:close ()

\n\n\n

\nCloses file.\nNote that files are automatically closed when\ntheir handles are garbage collected,\nbut that takes an unpredictable amount of time to happen.\n\n\n\n\n

\n

file:flush ()

\n\n\n

\nSaves any written data to file.\n\n\n\n\n

\n

file:lines ()

\n\n\n

\nReturns an iterator function that,\neach time it is called,\nreturns a new line from the file.\nTherefore, the construction\n\n

\n     for line in file:lines() do body end\n

\nwill iterate over all lines of the file.\n(Unlike io.lines, this function does not close the file\nwhen the loop ends.)\n\n\n\n\n

\n

file:read (···)

\n\n\n

\nReads the file file,\naccording to the given formats, which specify what to read.\nFor each format,\nthe function returns a string (or a number) with the characters read,\nor nil if it cannot read data with the specified format.\nWhen called without formats,\nit uses a default format that reads the entire next line\n(see below).\n\n\n

\nThe available formats are\n\n

    \n\n
  • \"*n\":\nreads a number;\nthis is the only format that returns a number instead of a string.\n
    • \n\n
  • \"*a\":\nreads the whole file, starting at the current position.\nOn end of file, it returns the empty string.\n
    • \n\n
  • \"*l\":\nreads the next line (skipping the end of line),\nreturning nil on end of file.\nThis is the default format.\n
    • \n\n
  • number:\nreads a string with up to this number of characters,\nreturning nil on end of file.\nIf number is zero,\nit reads nothing and returns an empty string,\nor nil on end of file.\n
    • \n\n
\n\n\n\n

\n

file:seek ([whence] [, offset])

\n\n\n

\nSets and gets the file position,\nmeasured from the beginning of the file,\nto the position given by offset plus a base\nspecified by the string whence, as follows:\n\n

    \n
  • \"set\": base is position 0 (beginning of the file);
    • \n
  • \"cur\": base is current position;
    • \n
  • \"end\": base is end of file;
    • \n

\nIn case of success, function seek returns the final file position,\nmeasured in bytes from the beginning of the file.\nIf this function fails, it returns nil,\nplus a string describing the error.\n\n\n

\nThe default value for whence is \"cur\",\nand for offset is 0.\nTherefore, the call file:seek() returns the current\nfile position, without changing it;\nthe call file:seek(\"set\") sets the position to the\nbeginning of the file (and returns 0);\nand the call file:seek(\"end\") sets the position to the\nend of the file, and returns its size.\n\n\n\n\n

\n

file:setvbuf (mode [, size])

\n\n\n

\nSets the buffering mode for an output file.\nThere are three available modes:\n\n

    \n\n
  • \"no\":\nno buffering; the result of any output operation appears immediately.\n
    • \n\n
  • \"full\":\nfull buffering; output operation is performed only\nwhen the buffer is full (or when you explicitly flush the file\n(see io.flush)).\n
    • \n\n
  • \"line\":\nline buffering; output is buffered until a newline is output\nor there is any input from some special files\n(such as a terminal device).\n
    • \n\n

\nFor the last two cases, size\nspecifies the size of the buffer, in bytes.\nThe default is an appropriate size.\n\n\n\n\n

\n

file:write (···)

\n\n\n

\nWrites the value of each of its arguments to\nthe file.\nThe arguments must be strings or numbers.\nTo write other values,\nuse tostring or string.format before write.\n\n\n\n\n\n\n\n

5.8 - Operating System Facilities

\n\n

\nThis library is implemented through table os.\n\n\n

\n

os.clock ()

\n\n\n

\nReturns an approximation of the amount in seconds of CPU time\nused by the program.\n\n\n\n\n

\n

os.date ([format [, time]])

\n\n\n

\nReturns a string or a table containing date and time,\nformatted according to the given string format.\n\n\n

\nIf the time argument is present,\nthis is the time to be formatted\n(see the os.time function for a description of this value).\nOtherwise, date formats the current time.\n\n\n

\nIf format starts with '!',\nthen the date is formatted in Coordinated Universal Time.\nAfter this optional character,\nif format is the string \"*t\",\nthen date returns a table with the following fields:\nyear (four digits), month (1--12), day (1--31),\nhour (0--23), min (0--59), sec (0--61),\nwday (weekday, Sunday is 1),\nyday (day of the year),\nand isdst (daylight saving flag, a boolean).\n\n\n

\nIf format is not \"*t\",\nthen date returns the date as a string,\nformatted according to the same rules as the C function strftime.\n\n\n

\nWhen called without arguments,\ndate returns a reasonable date and time representation that depends on\nthe host system and on the current locale\n(that is, os.date() is equivalent to os.date(\"%c\")).\n\n\n\n\n

\n

os.difftime (t2, t1)

\n\n\n

\nReturns the number of seconds from time t1 to time t2.\nIn POSIX, Windows, and some other systems,\nthis value is exactly t2-t1.\n\n\n\n\n

\n

os.execute ([command])

\n\n\n

\nThis function is equivalent to the C function system.\nIt passes command to be executed by an operating system shell.\nIt returns a status code, which is system-dependent.\nIf command is absent, then it returns nonzero if a shell is available\nand zero otherwise.\n\n\n\n\n

\n

os.exit ([code])

\n\n\n

\nCalls the C function exit,\nwith an optional code,\nto terminate the host program.\nThe default value for code is the success code.\n\n\n\n\n

\n

os.getenv (varname)

\n\n\n

\nReturns the value of the process environment variable varname,\nor nil if the variable is not defined.\n\n\n\n\n

\n

os.remove (filename)

\n\n\n

\nDeletes the file or directory with the given name.\nDirectories must be empty to be removed.\nIf this function fails, it returns nil,\nplus a string describing the error.\n\n\n\n\n

\n

os.rename (oldname, newname)

\n\n\n

\nRenames file or directory named oldname to newname.\nIf this function fails, it returns nil,\nplus a string describing the error.\n\n\n\n\n

\n

os.setlocale (locale [, category])

\n\n\n

\nSets the current locale of the program.\nlocale is a string specifying a locale;\ncategory is an optional string describing which category to change:\n\"all\", \"collate\", \"ctype\",\n\"monetary\", \"numeric\", or \"time\";\nthe default category is \"all\".\nThe function returns the name of the new locale,\nor nil if the request cannot be honored.\n\n\n

\nIf locale is the empty string,\nthe current locale is set to an implementation-defined native locale.\nIf locale is the string \"C\",\nthe current locale is set to the standard C locale.\n\n\n

\nWhen called with nil as the first argument,\nthis function only returns the name of the current locale\nfor the given category.\n\n\n\n\n

\n

os.time ([table])

\n\n\n

\nReturns the current time when called without arguments,\nor a time representing the date and time specified by the given table.\nThis table must have fields year, month, and day,\nand may have fields hour, min, sec, and isdst\n(for a description of these fields, see the os.date function).\n\n\n

\nThe returned value is a number, whose meaning depends on your system.\nIn POSIX, Windows, and some other systems, this number counts the number\nof seconds since some given start time (the \"epoch\").\nIn other systems, the meaning is not specified,\nand the number returned by time can be used only as an argument to\ndate and difftime.\n\n\n\n\n

\n

os.tmpname ()

\n\n\n

\nReturns a string with a file name that can\nbe used for a temporary file.\nThe file must be explicitly opened before its use\nand explicitly removed when no longer needed.\n\n\n

\nOn some systems (POSIX),\nthis function also creates a file with that name,\nto avoid security risks.\n(Someone else might create the file with wrong permissions\nin the time between getting the name and creating the file.)\nYou still have to open the file to use it\nand to remove it (even if you do not use it).\n\n\n

\nWhen possible,\nyou may prefer to use io.tmpfile,\nwhich automatically removes the file when the program ends.\n\n\n\n\n\n\n\n

5.9 - The Debug Library

\n\n

\nThis library provides\nthe functionality of the debug interface to Lua programs.\nYou should exert care when using this library.\nThe functions provided here should be used exclusively for debugging\nand similar tasks, such as profiling.\nPlease resist the temptation to use them as a\nusual programming tool:\nthey can be very slow.\nMoreover, several of these functions\nviolate some assumptions about Lua code\n(e.g., that variables local to a function\ncannot be accessed from outside or\nthat userdata metatables cannot be changed by Lua code)\nand therefore can compromise otherwise secure code.\n\n\n

\nAll functions in this library are provided\ninside the debug table.\nAll functions that operate over a thread\nhave an optional first argument which is the\nthread to operate over.\nThe default is always the current thread.\n\n\n

\n

debug.debug ()

\n\n\n

\nEnters an interactive mode with the user,\nrunning each string that the user enters.\nUsing simple commands and other debug facilities,\nthe user can inspect global and local variables,\nchange their values, evaluate expressions, and so on.\nA line containing only the word cont finishes this function,\nso that the caller continues its execution.\n\n\n

\nNote that commands for debug.debug are not lexically nested\nwithin any function, and so have no direct access to local variables.\n\n\n\n\n

\n

debug.getfenv (o)

\nReturns the environment of object o.\n\n\n\n\n

\n

debug.gethook ([thread])

\n\n\n

\nReturns the current hook settings of the thread, as three values:\nthe current hook function, the current hook mask,\nand the current hook count\n(as set by the debug.sethook function).\n\n\n\n\n

\n

debug.getinfo ([thread,] function [, what])

\n\n\n

\nReturns a table with information about a function.\nYou can give the function directly,\nor you can give a number as the value of function,\nwhich means the function running at level function of the call stack\nof the given thread:\nlevel 0 is the current function (getinfo itself);\nlevel 1 is the function that called getinfo;\nand so on.\nIf function is a number larger than the number of active functions,\nthen getinfo returns nil.\n\n\n

\nThe returned table can contain all the fields returned by lua_getinfo,\nwith the string what describing which fields to fill in.\nThe default for what is to get all information available,\nexcept the table of valid lines.\nIf present,\nthe option 'f'\nadds a field named func with the function itself.\nIf present,\nthe option 'L'\nadds a field named activelines with the table of\nvalid lines.\n\n\n

\nFor instance, the expression debug.getinfo(1,\"n\").name returns\na table with a name for the current function,\nif a reasonable name can be found,\nand the expression debug.getinfo(print)\nreturns a table with all available information\nabout the print function.\n\n\n\n\n

\n

debug.getlocal ([thread,] level, local)

\n\n\n

\nThis function returns the name and the value of the local variable\nwith index local of the function at level level of the stack.\n(The first parameter or local variable has index 1, and so on,\nuntil the last active local variable.)\nThe function returns nil if there is no local\nvariable with the given index,\nand raises an error when called with a level out of range.\n(You can call debug.getinfo to check whether the level is valid.)\n\n\n

\nVariable names starting with '(' (open parentheses)\nrepresent internal variables\n(loop control variables, temporaries, and C function locals).\n\n\n\n\n

\n

debug.getmetatable (object)

\n\n\n

\nReturns the metatable of the given object\nor nil if it does not have a metatable.\n\n\n\n\n

\n

debug.getregistry ()

\n\n\n

\nReturns the registry table (see §3.5).\n\n\n\n\n

\n

debug.getupvalue (func, up)

\n\n\n

\nThis function returns the name and the value of the upvalue\nwith index up of the function func.\nThe function returns nil if there is no upvalue with the given index.\n\n\n\n\n

\n

debug.setfenv (object, table)

\n\n\n

\nSets the environment of the given object to the given table.\nReturns object.\n\n\n\n\n

\n

debug.sethook ([thread,] hook, mask [, count])

\n\n\n

\nSets the given function as a hook.\nThe string mask and the number count describe\nwhen the hook will be called.\nThe string mask may have the following characters,\nwith the given meaning:\n\n

    \n
  • \"c\": the hook is called every time Lua calls a function;
    • \n
  • \"r\": the hook is called every time Lua returns from a function;
    • \n
  • \"l\": the hook is called every time Lua enters a new line of code.
    • \n

\nWith a count different from zero,\nthe hook is called after every count instructions.\n\n\n

\nWhen called without arguments,\ndebug.sethook turns off the hook.\n\n\n

\nWhen the hook is called, its first parameter is a string\ndescribing the event that has triggered its call:\n\"call\", \"return\" (or \"tail return\",\nwhen simulating a return from a tail call),\n\"line\", and \"count\".\nFor line events,\nthe hook also gets the new line number as its second parameter.\nInside a hook,\nyou can call getinfo with level 2 to get more information about\nthe running function\n(level 0 is the getinfo function,\nand level 1 is the hook function),\nunless the event is \"tail return\".\nIn this case, Lua is only simulating the return,\nand a call to getinfo will return invalid data.\n\n\n\n\n

\n

debug.setlocal ([thread,] level, local, value)

\n\n\n

\nThis function assigns the value value to the local variable\nwith index local of the function at level level of the stack.\nThe function returns nil if there is no local\nvariable with the given index,\nand raises an error when called with a level out of range.\n(You can call getinfo to check whether the level is valid.)\nOtherwise, it returns the name of the local variable.\n\n\n\n\n

\n

debug.setmetatable (object, table)

\n\n\n

\nSets the metatable for the given object to the given table\n(which can be nil).\n\n\n\n\n

\n

debug.setupvalue (func, up, value)

\n\n\n

\nThis function assigns the value value to the upvalue\nwith index up of the function func.\nThe function returns nil if there is no upvalue\nwith the given index.\nOtherwise, it returns the name of the upvalue.\n\n\n\n\n

\n

debug.traceback ([thread,] [message [, level]])

\n\n\n

\nReturns a string with a traceback of the call stack.\nAn optional message string is appended\nat the beginning of the traceback.\nAn optional level number tells at which level\nto start the traceback\n(default is 1, the function calling traceback).\n\n\n\n\n\n\n\n

6 - Lua Stand-alone

\n\n

\nAlthough Lua has been designed as an extension language,\nto be embedded in a host C program,\nit is also frequently used as a stand-alone language.\nAn interpreter for Lua as a stand-alone language,\ncalled simply lua,\nis provided with the standard distribution.\nThe stand-alone interpreter includes\nall standard libraries, including the debug library.\nIts usage is:\n\n

\n     lua [options] [script [args]]\n

\nThe options are:\n\n

    \n
  • -e stat: executes string stat;
    • \n
  • -l mod: \"requires\" mod;
    • \n
  • -i: enters interactive mode after running script;
    • \n
  • -v: prints version information;
    • \n
  • --: stops handling options;
    • \n
  • -: executes stdin as a file and stops handling options.
    • \n

\nAfter handling its options, lua runs the given script,\npassing to it the given args as string arguments.\nWhen called without arguments,\nlua behaves as lua -v -i\nwhen the standard input (stdin) is a terminal,\nand as lua - otherwise.\n\n\n

\nBefore running any argument,\nthe interpreter checks for an environment variable LUA_INIT.\nIf its format is @filename,\nthen lua executes the file.\nOtherwise, lua executes the string itself.\n\n\n

\nAll options are handled in order, except -i.\nFor instance, an invocation like\n\n

\n     $ lua -e'a=1' -e 'print(a)' script.lua\n

\nwill first set a to 1, then print the value of a (which is '1'),\nand finally run the file script.lua with no arguments.\n(Here $ is the shell prompt. Your prompt may be different.)\n\n\n

\nBefore starting to run the script,\nlua collects all arguments in the command line\nin a global table called arg.\nThe script name is stored at index 0,\nthe first argument after the script name goes to index 1,\nand so on.\nAny arguments before the script name\n(that is, the interpreter name plus the options)\ngo to negative indices.\nFor instance, in the call\n\n

\n     $ lua -la b.lua t1 t2\n

\nthe interpreter first runs the file a.lua,\nthen creates a table\n\n

\n     arg = { [-2] = \"lua\", [-1] = \"-la\",\n             [0] = \"b.lua\",\n             [1] = \"t1\", [2] = \"t2\" }\n

\nand finally runs the file b.lua.\nThe script is called with arg[1], arg[2], ···\nas arguments;\nit can also access these arguments with the vararg expression '...'.\n\n\n

\nIn interactive mode,\nif you write an incomplete statement,\nthe interpreter waits for its completion\nby issuing a different prompt.\n\n\n

\nIf the global variable _PROMPT contains a string,\nthen its value is used as the prompt.\nSimilarly, if the global variable _PROMPT2 contains a string,\nits value is used as the secondary prompt\n(issued during incomplete statements).\nTherefore, both prompts can be changed directly on the command line\nor in any Lua programs by assigning to _PROMPT.\nSee the next example:\n\n

\n     $ lua -e\"_PROMPT='myprompt> '\" -i\n

\n(The outer pair of quotes is for the shell,\nthe inner pair is for Lua.)\nNote the use of -i to enter interactive mode;\notherwise,\nthe program would just end silently\nright after the assignment to _PROMPT.\n\n\n

\nTo allow the use of Lua as a\nscript interpreter in Unix systems,\nthe stand-alone interpreter skips\nthe first line of a chunk if it starts with #.\nTherefore, Lua scripts can be made into executable programs\nby using chmod +x and the #! form,\nas in\n\n

\n     #!/usr/local/bin/lua\n

\n(Of course,\nthe location of the Lua interpreter may be different in your machine.\nIf lua is in your PATH,\nthen \n\n

\n     #!/usr/bin/env lua\n

\nis a more portable solution.) \n\n\n\n

7 - Incompatibilities with the Previous Version

\n\n

\nHere we list the incompatibilities that you may find when moving a program\nfrom Lua 5.0 to Lua 5.1.\nYou can avoid most of the incompatibilities compiling Lua with\nappropriate options (see file luaconf.h).\nHowever,\nall these compatibility options will be removed in the next version of Lua.\n\n\n\n

7.1 - Changes in the Language

\n
    \n\n
  • \nThe vararg system changed from the pseudo-argument arg with a\ntable with the extra arguments to the vararg expression.\n(See compile-time option LUA_COMPAT_VARARG in luaconf.h.)\n
    • \n\n
  • \nThere was a subtle change in the scope of the implicit\nvariables of the for statement and for the repeat statement.\n
    • \n\n
  • \nThe long string/long comment syntax ([[string]])\ndoes not allow nesting.\nYou can use the new syntax ([=[string]=]) in these cases.\n(See compile-time option LUA_COMPAT_LSTR in luaconf.h.)\n
    • \n\n
\n\n\n\n\n

7.2 - Changes in the Libraries

\n
    \n\n
  • \nFunction string.gfind was renamed string.gmatch.\n(See compile-time option LUA_COMPAT_GFIND in luaconf.h.)\n
    • \n\n
  • \nWhen string.gsub is called with a function as its\nthird argument,\nwhenever this function returns nil or false the\nreplacement string is the whole match,\ninstead of the empty string.\n
    • \n\n
  • \nFunction table.setn was deprecated.\nFunction table.getn corresponds\nto the new length operator (#);\nuse the operator instead of the function.\n(See compile-time option LUA_COMPAT_GETN in luaconf.h.)\n
    • \n\n
  • \nFunction loadlib was renamed package.loadlib.\n(See compile-time option LUA_COMPAT_LOADLIB in luaconf.h.)\n
    • \n\n
  • \nFunction math.mod was renamed math.fmod.\n(See compile-time option LUA_COMPAT_MOD in luaconf.h.)\n
    • \n\n
  • \nFunctions table.foreach and table.foreachi are deprecated.\nYou can use a for loop with pairs or ipairs instead.\n
    • \n\n
  • \nThere were substantial changes in function require due to\nthe new module system.\nHowever, the new behavior is mostly compatible with the old,\nbut require gets the path from package.path instead\nof from LUA_PATH.\n
    • \n\n
  • \nFunction collectgarbage has different arguments.\nFunction gcinfo is deprecated;\nuse collectgarbage(\"count\") instead.\n
    • \n\n
\n\n\n\n\n

7.3 - Changes in the API

\n
    \n\n
  • \nThe luaopen_* functions (to open libraries)\ncannot be called directly,\nlike a regular C function.\nThey must be called through Lua,\nlike a Lua function.\n
    • \n\n
  • \nFunction lua_open was replaced by lua_newstate to\nallow the user to set a memory-allocation function.\nYou can use luaL_newstate from the standard library to\ncreate a state with a standard allocation function\n(based on realloc).\n
    • \n\n
  • \nFunctions luaL_getn and luaL_setn\n(from the auxiliary library) are deprecated.\nUse lua_objlen instead of luaL_getn\nand nothing instead of luaL_setn.\n
    • \n\n
  • \nFunction luaL_openlib was replaced by luaL_register.\n
    • \n\n
  • \nFunction luaL_checkudata now throws an error when the given value\nis not a userdata of the expected type.\n(In Lua 5.0 it returned NULL.)\n
    • \n\n
\n\n\n\n\n

8 - The Complete Syntax of Lua

\n\n

\nHere is the complete syntax of Lua in extended BNF.\n(It does not describe operator precedences.)\n\n\n\n\n

\n\n\tchunk ::= {stat [`;´]} [laststat [`;´]]\n\n\tblock ::= chunk\n\n\tstat ::=  varlist `=´ explist | \n\t\t functioncall | \n\t\t do block end | \n\t\t while exp do block end | \n\t\t repeat block until exp | \n\t\t if exp then block {elseif exp then block} [else block] end | \n\t\t for Name `=´ exp `,´ exp [`,´ exp] do block end | \n\t\t for namelist in explist do block end | \n\t\t function funcname funcbody | \n\t\t local function Name funcbody | \n\t\t local namelist [`=´ explist] \n\n\tlaststat ::= return [explist] | break\n\n\tfuncname ::= Name {`.´ Name} [`:´ Name]\n\n\tvarlist ::= var {`,´ var}\n\n\tvar ::=  Name | prefixexp `[´ exp `]´ | prefixexp `.´ Name \n\n\tnamelist ::= Name {`,´ Name}\n\n\texplist ::= {exp `,´} exp\n\n\texp ::=  nil | false | true | Number | String | `...´ | function | \n\t\t prefixexp | tableconstructor | exp binop exp | unop exp \n\n\tprefixexp ::= var | functioncall | `(´ exp `)´\n\n\tfunctioncall ::=  prefixexp args | prefixexp `:´ Name args \n\n\targs ::=  `(´ [explist] `)´ | tableconstructor | String \n\n\tfunction ::= function funcbody\n\n\tfuncbody ::= `(´ [parlist] `)´ block end\n\n\tparlist ::= namelist [`,´ `...´] | `...´\n\n\ttableconstructor ::= `{´ [fieldlist] `}´\n\n\tfieldlist ::= field {fieldsep field} [fieldsep]\n\n\tfield ::= `[´ exp `]´ `=´ exp | Name `=´ exp | exp\n\n\tfieldsep ::= `,´ | `;´\n\n\tbinop ::= `+´ | `-´ | `*´ | `/´ | `^´ | `%´ | `..´ | \n\t\t `<´ | `<=´ | `>´ | `>=´ | `==´ | `~=´ | \n\t\t and | or\n\n\tunop ::= `-´ | not | `#´\n\n
\n\n

\n\n\n\n\n\n\n\n

\n\nLast update:\nMon Feb 13 18:54:19 BRST 2012\n\n\n\n\n\n" }, { "path": "deps/lua/doc/readme.html", "content": "\n\nLua documentation\n\n\n\n\n\n
\n

\n\"Lua\"\nDocumentation\n

\n\nThis is the documentation included in the source distribution of Lua 5.1.5.\n\n\n\nLua's\nofficial web site\ncontains updated documentation,\nespecially the\nreference manual.\n

\n\n

\n\nLast update:\nFri Feb 3 09:44:42 BRST 2012\n\n\n\n\n" }, { "path": "deps/lua/etc/Makefile", "content": "# makefile for Lua etc\n\nTOP= ..\nLIB= $(TOP)/src\nINC= $(TOP)/src\nBIN= $(TOP)/src\nSRC= $(TOP)/src\nTST= $(TOP)/test\n\nCC= gcc\nCFLAGS= -O2 -Wall -I$(INC) $(MYCFLAGS)\nMYCFLAGS= \nMYLDFLAGS= -Wl,-E\nMYLIBS= -lm\n#MYLIBS= -lm -Wl,-E -ldl -lreadline -lhistory -lncurses\nRM= rm -f\n\ndefault:\n\t@echo 'Please choose a target: min noparser one strict clean'\n\nmin:\tmin.c\n\t$(CC) $(CFLAGS) $@.c -L$(LIB) -llua $(MYLIBS)\n\techo 'print\"Hello there!\"' | ./a.out\n\nnoparser: noparser.o\n\t$(CC) noparser.o $(SRC)/lua.o -L$(LIB) -llua $(MYLIBS)\n\t$(BIN)/luac $(TST)/hello.lua\n\t-./a.out luac.out\n\t-./a.out -e'a=1'\n\none:\n\t$(CC) $(CFLAGS) all.c $(MYLIBS)\n\t./a.out $(TST)/hello.lua\n\nstrict:\n\t-$(BIN)/lua -e 'print(a);b=2'\n\t-$(BIN)/lua -lstrict -e 'print(a)'\n\t-$(BIN)/lua -e 'function f() b=2 end f()'\n\t-$(BIN)/lua -lstrict -e 'function f() b=2 end f()'\n\nclean:\n\t$(RM) a.out core core.* *.o luac.out\n\n.PHONY:\tdefault min noparser one strict clean\n" }, { "path": "deps/lua/etc/README", "content": "This directory contains some useful files and code.\nUnlike the code in ../src, everything here is in the public domain.\n\nIf any of the makes fail, you're probably not using the same libraries\nused to build Lua. Set MYLIBS in Makefile accordingly.\n\nall.c\n\tFull Lua interpreter in a single file.\n\tDo \"make one\" for a demo.\n\nlua.hpp\n\tLua header files for C++ using 'extern \"C\"'.\n\nlua.ico\n\tA Lua icon for Windows (and web sites: save as favicon.ico).\n\tDrawn by hand by Markus Gritsch .\n\nlua.pc\n\tpkg-config data for Lua\n\nluavs.bat\n\tScript to build Lua under \"Visual Studio .NET Command Prompt\".\n\tRun it from the toplevel as etc\\luavs.bat.\n\nmin.c\n\tA minimal Lua interpreter.\n\tGood for learning and for starting your own.\n\tDo \"make min\" for a demo.\n\nnoparser.c\n\tLinking with noparser.o avoids loading the parsing modules in lualib.a.\n\tDo \"make noparser\" for a demo.\n\nstrict.lua\n\tTraps uses of undeclared global variables.\n\tDo \"make strict\" for a demo.\n\n" }, { "path": "deps/lua/etc/all.c", "content": "/*\n* all.c -- Lua core, libraries and interpreter in a single file\n*/\n\n#define luaall_c\n\n#include \"lapi.c\"\n#include \"lcode.c\"\n#include \"ldebug.c\"\n#include \"ldo.c\"\n#include \"ldump.c\"\n#include \"lfunc.c\"\n#include \"lgc.c\"\n#include \"llex.c\"\n#include \"lmem.c\"\n#include \"lobject.c\"\n#include \"lopcodes.c\"\n#include \"lparser.c\"\n#include \"lstate.c\"\n#include \"lstring.c\"\n#include \"ltable.c\"\n#include \"ltm.c\"\n#include \"lundump.c\"\n#include \"lvm.c\"\n#include \"lzio.c\"\n\n#include \"lauxlib.c\"\n#include \"lbaselib.c\"\n#include \"ldblib.c\"\n#include \"liolib.c\"\n#include \"linit.c\"\n#include \"lmathlib.c\"\n#include \"loadlib.c\"\n#include \"loslib.c\"\n#include \"lstrlib.c\"\n#include \"ltablib.c\"\n\n#include \"lua.c\"\n" }, { "path": "deps/lua/etc/lua.hpp", "content": "// lua.hpp\n// Lua header files for C++\n// <> not supplied automatically because Lua also compiles as C++\n\nextern \"C\" {\n#include \"lua.h\"\n#include \"lualib.h\"\n#include \"lauxlib.h\"\n}\n" }, { "path": "deps/lua/etc/lua.pc", "content": "# lua.pc -- pkg-config data for Lua\n\n# vars from install Makefile\n\n# grep '^V=' ../Makefile\nV= 5.1\n# grep '^R=' ../Makefile\nR= 5.1.5\n\n# grep '^INSTALL_.*=' ../Makefile | sed 's/INSTALL_TOP/prefix/'\nprefix= /usr/local\nINSTALL_BIN= ${prefix}/bin\nINSTALL_INC= ${prefix}/include\nINSTALL_LIB= ${prefix}/lib\nINSTALL_MAN= ${prefix}/man/man1\nINSTALL_LMOD= ${prefix}/share/lua/${V}\nINSTALL_CMOD= ${prefix}/lib/lua/${V}\n\n# canonical vars\nexec_prefix=${prefix}\nlibdir=${exec_prefix}/lib\nincludedir=${prefix}/include\n\nName: Lua\nDescription: An Extensible Extension Language\nVersion: ${R}\nRequires: \nLibs: -L${libdir} -llua -lm\nCflags: -I${includedir}\n\n# (end of lua.pc)\n" }, { "path": "deps/lua/etc/luavs.bat", "content": "@rem Script to build Lua under \"Visual Studio .NET Command Prompt\".\r\n@rem Do not run from this directory; run it from the toplevel: etc\\luavs.bat .\r\n@rem It creates lua51.dll, lua51.lib, lua.exe, and luac.exe in src.\r\n@rem (contributed by David Manura and Mike Pall)\r\n\r\n@setlocal\r\n@set MYCOMPILE=cl /nologo /MD /O2 /W3 /c /D_CRT_SECURE_NO_DEPRECATE\r\n@set MYLINK=link /nologo\r\n@set MYMT=mt /nologo\r\n\r\ncd src\r\n%MYCOMPILE% /DLUA_BUILD_AS_DLL l*.c\r\ndel lua.obj luac.obj\r\n%MYLINK% /DLL /out:lua51.dll l*.obj\r\nif exist lua51.dll.manifest^\r\n %MYMT% -manifest lua51.dll.manifest -outputresource:lua51.dll;2\r\n%MYCOMPILE% /DLUA_BUILD_AS_DLL lua.c\r\n%MYLINK% /out:lua.exe lua.obj lua51.lib\r\nif exist lua.exe.manifest^\r\n %MYMT% -manifest lua.exe.manifest -outputresource:lua.exe\r\n%MYCOMPILE% l*.c print.c\r\ndel lua.obj linit.obj lbaselib.obj ldblib.obj liolib.obj lmathlib.obj^\r\n loslib.obj ltablib.obj lstrlib.obj loadlib.obj\r\n%MYLINK% /out:luac.exe *.obj\r\nif exist luac.exe.manifest^\r\n %MYMT% -manifest luac.exe.manifest -outputresource:luac.exe\r\ndel *.obj *.manifest\r\ncd ..\r\n" }, { "path": "deps/lua/etc/min.c", "content": "/*\n* min.c -- a minimal Lua interpreter\n* loads stdin only with minimal error handling.\n* no interaction, and no standard library, only a \"print\" function.\n*/\n\n#include \n\n#include \"lua.h\"\n#include \"lauxlib.h\"\n\nstatic int print(lua_State *L)\n{\n int n=lua_gettop(L);\n int i;\n for (i=1; i<=n; i++)\n {\n if (i>1) printf(\"\\t\");\n if (lua_isstring(L,i))\n printf(\"%s\",lua_tostring(L,i));\n else if (lua_isnil(L,i))\n printf(\"%s\",\"nil\");\n else if (lua_isboolean(L,i))\n printf(\"%s\",lua_toboolean(L,i) ? \"true\" : \"false\");\n else\n printf(\"%s:%p\",luaL_typename(L,i),lua_topointer(L,i));\n }\n printf(\"\\n\");\n return 0;\n}\n\nint main(void)\n{\n lua_State *L=lua_open();\n lua_register(L,\"print\",print);\n if (luaL_dofile(L,NULL)!=0) fprintf(stderr,\"%s\\n\",lua_tostring(L,-1));\n lua_close(L);\n return 0;\n}\n" }, { "path": "deps/lua/etc/noparser.c", "content": "/*\n* The code below can be used to make a Lua core that does not contain the\n* parsing modules (lcode, llex, lparser), which represent 35% of the total core.\n* You'll only be able to load binary files and strings, precompiled with luac.\n* (Of course, you'll have to build luac with the original parsing modules!)\n*\n* To use this module, simply compile it (\"make noparser\" does that) and list\n* its object file before the Lua libraries. The linker should then not load\n* the parsing modules. To try it, do \"make luab\".\n*\n* If you also want to avoid the dump module (ldump.o), define NODUMP.\n* #define NODUMP\n*/\n\n#define LUA_CORE\n\n#include \"llex.h\"\n#include \"lparser.h\"\n#include \"lzio.h\"\n\nLUAI_FUNC void luaX_init (lua_State *L) {\n UNUSED(L);\n}\n\nLUAI_FUNC Proto *luaY_parser (lua_State *L, ZIO *z, Mbuffer *buff, const char *name) {\n UNUSED(z);\n UNUSED(buff);\n UNUSED(name);\n lua_pushliteral(L,\"parser not loaded\");\n lua_error(L);\n return NULL;\n}\n\n#ifdef NODUMP\n#include \"lundump.h\"\n\nLUAI_FUNC int luaU_dump (lua_State* L, const Proto* f, lua_Writer w, void* data, int strip) {\n UNUSED(f);\n UNUSED(w);\n UNUSED(data);\n UNUSED(strip);\n#if 1\n UNUSED(L);\n return 0;\n#else\n lua_pushliteral(L,\"dumper not loaded\");\n lua_error(L);\n#endif\n}\n#endif\n" }, { "path": "deps/lua/etc/strict.lua", "content": "--\n-- strict.lua\n-- checks uses of undeclared global variables\n-- All global variables must be 'declared' through a regular assignment\n-- (even assigning nil will do) in a main chunk before being used\n-- anywhere or assigned to inside a function.\n--\n\nlocal getinfo, error, rawset, rawget = debug.getinfo, error, rawset, rawget\n\nlocal mt = getmetatable(_G)\nif mt == nil then\n mt = {}\n setmetatable(_G, mt)\nend\n\nmt.__declared = {}\n\nlocal function what ()\n local d = getinfo(3, \"S\")\n return d and d.what or \"C\"\nend\n\nmt.__newindex = function (t, n, v)\n if not mt.__declared[n] then\n local w = what()\n if w ~= \"main\" and w ~= \"C\" then\n error(\"assign to undeclared variable '\"..n..\"'\", 2)\n end\n mt.__declared[n] = true\n end\n rawset(t, n, v)\nend\n \nmt.__index = function (t, n)\n if not mt.__declared[n] and what() ~= \"C\" then\n error(\"variable '\"..n..\"' is not declared\", 2)\n end\n return rawget(t, n)\nend\n\n" }, { "path": "deps/lua/src/Makefile", "content": "# makefile for building Lua\n# see ../INSTALL for installation instructions\n# see ../Makefile and luaconf.h for further customization\n\n# == CHANGE THE SETTINGS BELOW TO SUIT YOUR ENVIRONMENT =======================\n\n# Your platform. See PLATS for possible values.\nPLAT= none\n\nCC?= gcc\nCFLAGS= -O2 -Wall $(MYCFLAGS)\nAR= ar rcu\nRANLIB= ranlib\nRM= rm -f\nLIBS= -lm $(MYLIBS)\n\nMYCFLAGS=\nMYLDFLAGS=\nMYLIBS=\n\n# == END OF USER SETTINGS. NO NEED TO CHANGE ANYTHING BELOW THIS LINE =========\n\nPLATS= aix ansi bsd freebsd generic linux macosx mingw posix solaris\n\nLUA_A=\tliblua.a\nCORE_O=\tlapi.o lcode.o ldebug.o ldo.o ldump.o lfunc.o lgc.o llex.o lmem.o \\\n\tlobject.o lopcodes.o lparser.o lstate.o lstring.o ltable.o ltm.o \\\n\tlundump.o lvm.o lzio.o strbuf.o\nLIB_O=\tlauxlib.o lbaselib.o ldblib.o liolib.o lmathlib.o loslib.o ltablib.o \\\n\tlstrlib.o loadlib.o linit.o lua_cjson.o lua_struct.o lua_cmsgpack.o\n\nLUA_T=\tlua\nLUA_O=\tlua.o\n\nLUAC_T=\tluac\nLUAC_O=\tluac.o print.o\n\nALL_O= $(CORE_O) $(LIB_O) $(LUA_O) $(LUAC_O)\nALL_T= $(LUA_A) $(LUA_T) $(LUAC_T)\nALL_A= $(LUA_A)\n\ndefault: $(PLAT)\n\nall:\t$(ALL_T)\n\no:\t$(ALL_O)\n\na:\t$(ALL_A)\n\n$(LUA_A): $(CORE_O) $(LIB_O)\n\t$(AR) $@ $(CORE_O) $(LIB_O)\t# DLL needs all object files\n\t$(RANLIB) $@\n\n$(LUA_T): $(LUA_O) $(LUA_A)\n\t$(CC) -o $@ $(MYLDFLAGS) $(LUA_O) $(LUA_A) $(LIBS)\n\n$(LUAC_T): $(LUAC_O) $(LUA_A)\n\t$(CC) -o $@ $(MYLDFLAGS) $(LUAC_O) $(LUA_A) $(LIBS)\n\nclean:\n\t$(RM) $(ALL_T) $(ALL_O)\n\ndepend:\n\t@$(CC) $(CFLAGS) -MM l*.c print.c\n\necho:\n\t@echo \"PLAT = $(PLAT)\"\n\t@echo \"CC = $(CC)\"\n\t@echo \"CFLAGS = $(CFLAGS)\"\n\t@echo \"AR = $(AR)\"\n\t@echo \"RANLIB = $(RANLIB)\"\n\t@echo \"RM = $(RM)\"\n\t@echo \"MYCFLAGS = $(MYCFLAGS)\"\n\t@echo \"MYLDFLAGS = $(MYLDFLAGS)\"\n\t@echo \"MYLIBS = $(MYLIBS)\"\n\n# convenience targets for popular platforms\n\nnone:\n\t@echo \"Please choose a platform:\"\n\t@echo \" $(PLATS)\"\n\naix:\n\t$(MAKE) all CC=\"xlc\" CFLAGS=\"-O2 -DLUA_USE_POSIX -DLUA_USE_DLOPEN\" MYLIBS=\"-ldl\" MYLDFLAGS=\"-brtl -bexpall\"\n\nansi:\n\t$(MAKE) all MYCFLAGS=-DLUA_ANSI\n\nbsd:\n\t$(MAKE) all MYCFLAGS=\"-DLUA_USE_POSIX -DLUA_USE_DLOPEN\" MYLIBS=\"-Wl,-E\"\n\nfreebsd:\n\t$(MAKE) all MYCFLAGS=\"-DLUA_USE_LINUX\" MYLIBS=\"-Wl,-E -lreadline\"\n\ngeneric:\n\t$(MAKE) all MYCFLAGS=\n\nlinux:\n\t$(MAKE) all MYCFLAGS=-DLUA_USE_LINUX MYLIBS=\"-Wl,-E -ldl -lreadline -lhistory -lncurses\"\n\nmacosx:\n\t$(MAKE) all MYCFLAGS=-DLUA_USE_LINUX MYLIBS=\"-lreadline\"\n# use this on Mac OS X 10.3-\n#\t$(MAKE) all MYCFLAGS=-DLUA_USE_MACOSX\n\nmingw:\n\t$(MAKE) \"LUA_A=lua51.dll\" \"LUA_T=lua.exe\" \\\n\t\"AR=$(CC) -shared -o\" \"RANLIB=strip --strip-unneeded\" \\\n\t\"MYCFLAGS=-DLUA_BUILD_AS_DLL\" \"MYLIBS=\" \"MYLDFLAGS=-s\" lua.exe\n\t$(MAKE) \"LUAC_T=luac.exe\" luac.exe\n\nposix:\n\t$(MAKE) all MYCFLAGS=-DLUA_USE_POSIX\n\nsolaris:\n\t$(MAKE) all MYCFLAGS=\"-DLUA_USE_POSIX -DLUA_USE_DLOPEN\" MYLIBS=\"-ldl\"\n\n# list targets that do not create files (but not all makes understand .PHONY)\n.PHONY: all $(PLATS) default o a clean depend echo none\n\n# DO NOT DELETE\n\nlapi.o: lapi.c lua.h luaconf.h lapi.h lobject.h llimits.h ldebug.h \\\n lstate.h ltm.h lzio.h lmem.h ldo.h lfunc.h lgc.h lstring.h ltable.h \\\n lundump.h lvm.h\nlauxlib.o: lauxlib.c lua.h luaconf.h lauxlib.h\nlbaselib.o: lbaselib.c lua.h luaconf.h lauxlib.h lualib.h\nlcode.o: lcode.c lua.h luaconf.h lcode.h llex.h lobject.h llimits.h \\\n lzio.h lmem.h lopcodes.h lparser.h ldebug.h lstate.h ltm.h ldo.h lgc.h \\\n ltable.h\nldblib.o: ldblib.c lua.h luaconf.h lauxlib.h lualib.h\nldebug.o: ldebug.c lua.h luaconf.h lapi.h lobject.h llimits.h lcode.h \\\n llex.h lzio.h lmem.h lopcodes.h lparser.h ldebug.h lstate.h ltm.h ldo.h \\\n lfunc.h lstring.h lgc.h ltable.h lvm.h\nldo.o: ldo.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h ltm.h \\\n lzio.h lmem.h ldo.h lfunc.h lgc.h lopcodes.h lparser.h lstring.h \\\n ltable.h lundump.h lvm.h\nldump.o: ldump.c lua.h luaconf.h lobject.h llimits.h lstate.h ltm.h \\\n lzio.h lmem.h lundump.h\nlfunc.o: lfunc.c lua.h luaconf.h lfunc.h lobject.h llimits.h lgc.h lmem.h \\\n lstate.h ltm.h lzio.h\nlgc.o: lgc.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h ltm.h \\\n lzio.h lmem.h ldo.h lfunc.h lgc.h lstring.h ltable.h\nlinit.o: linit.c lua.h luaconf.h lualib.h lauxlib.h\nliolib.o: liolib.c lua.h luaconf.h lauxlib.h lualib.h\nllex.o: llex.c lua.h luaconf.h ldo.h lobject.h llimits.h lstate.h ltm.h \\\n lzio.h lmem.h llex.h lparser.h lstring.h lgc.h ltable.h\nlmathlib.o: lmathlib.c lua.h luaconf.h lauxlib.h lualib.h\nlmem.o: lmem.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h \\\n ltm.h lzio.h lmem.h ldo.h\nloadlib.o: loadlib.c lua.h luaconf.h lauxlib.h lualib.h\nlobject.o: lobject.c lua.h luaconf.h ldo.h lobject.h llimits.h lstate.h \\\n ltm.h lzio.h lmem.h lstring.h lgc.h lvm.h\nlopcodes.o: lopcodes.c lopcodes.h llimits.h lua.h luaconf.h\nloslib.o: loslib.c lua.h luaconf.h lauxlib.h lualib.h\nlparser.o: lparser.c lua.h luaconf.h lcode.h llex.h lobject.h llimits.h \\\n lzio.h lmem.h lopcodes.h lparser.h ldebug.h lstate.h ltm.h ldo.h \\\n lfunc.h lstring.h lgc.h ltable.h\nlstate.o: lstate.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h \\\n ltm.h lzio.h lmem.h ldo.h lfunc.h lgc.h llex.h lstring.h ltable.h\nlstring.o: lstring.c lua.h luaconf.h lmem.h llimits.h lobject.h lstate.h \\\n ltm.h lzio.h lstring.h lgc.h\nlstrlib.o: lstrlib.c lua.h luaconf.h lauxlib.h lualib.h\nltable.o: ltable.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h \\\n ltm.h lzio.h lmem.h ldo.h lgc.h ltable.h\nltablib.o: ltablib.c lua.h luaconf.h lauxlib.h lualib.h\nltm.o: ltm.c lua.h luaconf.h lobject.h llimits.h lstate.h ltm.h lzio.h \\\n lmem.h lstring.h lgc.h ltable.h\nlua.o: lua.c lua.h luaconf.h lauxlib.h lualib.h\nluac.o: luac.c lua.h luaconf.h lauxlib.h ldo.h lobject.h llimits.h \\\n lstate.h ltm.h lzio.h lmem.h lfunc.h lopcodes.h lstring.h lgc.h \\\n lundump.h\nlundump.o: lundump.c lua.h luaconf.h ldebug.h lstate.h lobject.h \\\n llimits.h ltm.h lzio.h lmem.h ldo.h lfunc.h lstring.h lgc.h lundump.h\nlvm.o: lvm.c lua.h luaconf.h ldebug.h lstate.h lobject.h llimits.h ltm.h \\\n lzio.h lmem.h ldo.h lfunc.h lgc.h lopcodes.h lstring.h ltable.h lvm.h\nlzio.o: lzio.c lua.h luaconf.h llimits.h lmem.h lstate.h lobject.h ltm.h \\\n lzio.h\nprint.o: print.c ldebug.h lstate.h lua.h luaconf.h lobject.h llimits.h \\\n ltm.h lzio.h lmem.h lopcodes.h lundump.h\n\n# (end of Makefile)\n" }, { "path": "deps/lua/src/lapi.c", "content": "/*\n** $Id: lapi.c,v 2.55.1.5 2008/07/04 18:41:18 roberto Exp $\n** Lua API\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n\n#define lapi_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lapi.h\"\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n#include \"lundump.h\"\n#include \"lvm.h\"\n\n\n\nconst char lua_ident[] =\n \"$Lua: \" LUA_RELEASE \" \" LUA_COPYRIGHT \" $\\n\"\n \"$Authors: \" LUA_AUTHORS \" $\\n\"\n \"$URL: www.lua.org $\\n\";\n\n\n\n#define api_checknelems(L, n)\tapi_check(L, (n) <= (L->top - L->base))\n\n#define api_checkvalidindex(L, i)\tapi_check(L, (i) != luaO_nilobject)\n\n#define api_incr_top(L) {api_check(L, L->top < L->ci->top); L->top++;}\n\n\n\nstatic TValue *index2adr (lua_State *L, int idx) {\n if (idx > 0) {\n TValue *o = L->base + (idx - 1);\n api_check(L, idx <= L->ci->top - L->base);\n if (o >= L->top) return cast(TValue *, luaO_nilobject);\n else return o;\n }\n else if (idx > LUA_REGISTRYINDEX) {\n api_check(L, idx != 0 && -idx <= L->top - L->base);\n return L->top + idx;\n }\n else switch (idx) { /* pseudo-indices */\n case LUA_REGISTRYINDEX: return registry(L);\n case LUA_ENVIRONINDEX: {\n Closure *func = curr_func(L);\n sethvalue(L, &L->env, func->c.env);\n return &L->env;\n }\n case LUA_GLOBALSINDEX: return gt(L);\n default: {\n Closure *func = curr_func(L);\n idx = LUA_GLOBALSINDEX - idx;\n return (idx <= func->c.nupvalues)\n ? &func->c.upvalue[idx-1]\n : cast(TValue *, luaO_nilobject);\n }\n }\n}\n\n\nstatic Table *getcurrenv (lua_State *L) {\n if (L->ci == L->base_ci) /* no enclosing function? */\n return hvalue(gt(L)); /* use global table as environment */\n else {\n Closure *func = curr_func(L);\n return func->c.env;\n }\n}\n\n\nvoid luaA_pushobject (lua_State *L, const TValue *o) {\n setobj2s(L, L->top, o);\n api_incr_top(L);\n}\n\n\nLUA_API int lua_checkstack (lua_State *L, int size) {\n int res = 1;\n lua_lock(L);\n if (size > LUAI_MAXCSTACK || (L->top - L->base + size) > LUAI_MAXCSTACK)\n res = 0; /* stack overflow */\n else if (size > 0) {\n luaD_checkstack(L, size);\n if (L->ci->top < L->top + size)\n L->ci->top = L->top + size;\n }\n lua_unlock(L);\n return res;\n}\n\n\nLUA_API void lua_xmove (lua_State *from, lua_State *to, int n) {\n int i;\n if (from == to) return;\n lua_lock(to);\n api_checknelems(from, n);\n api_check(from, G(from) == G(to));\n api_check(from, to->ci->top - to->top >= n);\n from->top -= n;\n for (i = 0; i < n; i++) {\n setobj2s(to, to->top++, from->top + i);\n }\n lua_unlock(to);\n}\n\n\nLUA_API void lua_setlevel (lua_State *from, lua_State *to) {\n to->nCcalls = from->nCcalls;\n}\n\n\nLUA_API lua_CFunction lua_atpanic (lua_State *L, lua_CFunction panicf) {\n lua_CFunction old;\n lua_lock(L);\n old = G(L)->panic;\n G(L)->panic = panicf;\n lua_unlock(L);\n return old;\n}\n\n\nLUA_API lua_State *lua_newthread (lua_State *L) {\n lua_State *L1;\n lua_lock(L);\n luaC_checkGC(L);\n L1 = luaE_newthread(L);\n setthvalue(L, L->top, L1);\n api_incr_top(L);\n lua_unlock(L);\n luai_userstatethread(L, L1);\n return L1;\n}\n\n\n\n/*\n** basic stack manipulation\n*/\n\n\nLUA_API int lua_gettop (lua_State *L) {\n return cast_int(L->top - L->base);\n}\n\n\nLUA_API void lua_settop (lua_State *L, int idx) {\n lua_lock(L);\n if (idx >= 0) {\n api_check(L, idx <= L->stack_last - L->base);\n while (L->top < L->base + idx)\n setnilvalue(L->top++);\n L->top = L->base + idx;\n }\n else {\n api_check(L, -(idx+1) <= (L->top - L->base));\n L->top += idx+1; /* `subtract' index (index is negative) */\n }\n lua_unlock(L);\n}\n\n\nLUA_API void lua_remove (lua_State *L, int idx) {\n StkId p;\n lua_lock(L);\n p = index2adr(L, idx);\n api_checkvalidindex(L, p);\n while (++p < L->top) setobjs2s(L, p-1, p);\n L->top--;\n lua_unlock(L);\n}\n\n\nLUA_API void lua_insert (lua_State *L, int idx) {\n StkId p;\n StkId q;\n lua_lock(L);\n p = index2adr(L, idx);\n api_checkvalidindex(L, p);\n for (q = L->top; q>p; q--) setobjs2s(L, q, q-1);\n setobjs2s(L, p, L->top);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_replace (lua_State *L, int idx) {\n StkId o;\n lua_lock(L);\n /* explicit test for incompatible code */\n if (idx == LUA_ENVIRONINDEX && L->ci == L->base_ci)\n luaG_runerror(L, \"no calling environment\");\n api_checknelems(L, 1);\n o = index2adr(L, idx);\n api_checkvalidindex(L, o);\n if (idx == LUA_ENVIRONINDEX) {\n Closure *func = curr_func(L);\n api_check(L, ttistable(L->top - 1)); \n func->c.env = hvalue(L->top - 1);\n luaC_barrier(L, func, L->top - 1);\n }\n else {\n setobj(L, o, L->top - 1);\n if (idx < LUA_GLOBALSINDEX) /* function upvalue? */\n luaC_barrier(L, curr_func(L), L->top - 1);\n }\n L->top--;\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushvalue (lua_State *L, int idx) {\n lua_lock(L);\n setobj2s(L, L->top, index2adr(L, idx));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\n\n/*\n** access functions (stack -> C)\n*/\n\n\nLUA_API int lua_type (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n return (o == luaO_nilobject) ? LUA_TNONE : ttype(o);\n}\n\n\nLUA_API const char *lua_typename (lua_State *L, int t) {\n UNUSED(L);\n return (t == LUA_TNONE) ? \"no value\" : luaT_typenames[t];\n}\n\n\nLUA_API int lua_iscfunction (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n return iscfunction(o);\n}\n\n\nLUA_API int lua_isnumber (lua_State *L, int idx) {\n TValue n;\n const TValue *o = index2adr(L, idx);\n return tonumber(o, &n);\n}\n\n\nLUA_API int lua_isstring (lua_State *L, int idx) {\n int t = lua_type(L, idx);\n return (t == LUA_TSTRING || t == LUA_TNUMBER);\n}\n\n\nLUA_API int lua_isuserdata (lua_State *L, int idx) {\n const TValue *o = index2adr(L, idx);\n return (ttisuserdata(o) || ttislightuserdata(o));\n}\n\n\nLUA_API int lua_rawequal (lua_State *L, int index1, int index2) {\n StkId o1 = index2adr(L, index1);\n StkId o2 = index2adr(L, index2);\n return (o1 == luaO_nilobject || o2 == luaO_nilobject) ? 0\n : luaO_rawequalObj(o1, o2);\n}\n\n\nLUA_API int lua_equal (lua_State *L, int index1, int index2) {\n StkId o1, o2;\n int i;\n lua_lock(L); /* may call tag method */\n o1 = index2adr(L, index1);\n o2 = index2adr(L, index2);\n i = (o1 == luaO_nilobject || o2 == luaO_nilobject) ? 0 : equalobj(L, o1, o2);\n lua_unlock(L);\n return i;\n}\n\n\nLUA_API int lua_lessthan (lua_State *L, int index1, int index2) {\n StkId o1, o2;\n int i;\n lua_lock(L); /* may call tag method */\n o1 = index2adr(L, index1);\n o2 = index2adr(L, index2);\n i = (o1 == luaO_nilobject || o2 == luaO_nilobject) ? 0\n : luaV_lessthan(L, o1, o2);\n lua_unlock(L);\n return i;\n}\n\n\n\nLUA_API lua_Number lua_tonumber (lua_State *L, int idx) {\n TValue n;\n const TValue *o = index2adr(L, idx);\n if (tonumber(o, &n))\n return nvalue(o);\n else\n return 0;\n}\n\n\nLUA_API lua_Integer lua_tointeger (lua_State *L, int idx) {\n TValue n;\n const TValue *o = index2adr(L, idx);\n if (tonumber(o, &n)) {\n lua_Integer res;\n lua_Number num = nvalue(o);\n lua_number2integer(res, num);\n return res;\n }\n else\n return 0;\n}\n\n\nLUA_API int lua_toboolean (lua_State *L, int idx) {\n const TValue *o = index2adr(L, idx);\n return !l_isfalse(o);\n}\n\n\nLUA_API const char *lua_tolstring (lua_State *L, int idx, size_t *len) {\n StkId o = index2adr(L, idx);\n if (!ttisstring(o)) {\n lua_lock(L); /* `luaV_tostring' may create a new string */\n if (!luaV_tostring(L, o)) { /* conversion failed? */\n if (len != NULL) *len = 0;\n lua_unlock(L);\n return NULL;\n }\n luaC_checkGC(L);\n o = index2adr(L, idx); /* previous call may reallocate the stack */\n lua_unlock(L);\n }\n if (len != NULL) *len = tsvalue(o)->len;\n return svalue(o);\n}\n\n\nLUA_API size_t lua_objlen (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n switch (ttype(o)) {\n case LUA_TSTRING: return tsvalue(o)->len;\n case LUA_TUSERDATA: return uvalue(o)->len;\n case LUA_TTABLE: return luaH_getn(hvalue(o));\n case LUA_TNUMBER: {\n size_t l;\n lua_lock(L); /* `luaV_tostring' may create a new string */\n l = (luaV_tostring(L, o) ? tsvalue(o)->len : 0);\n lua_unlock(L);\n return l;\n }\n default: return 0;\n }\n}\n\n\nLUA_API lua_CFunction lua_tocfunction (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n return (!iscfunction(o)) ? NULL : clvalue(o)->c.f;\n}\n\n\nLUA_API void *lua_touserdata (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n switch (ttype(o)) {\n case LUA_TUSERDATA: return (rawuvalue(o) + 1);\n case LUA_TLIGHTUSERDATA: return pvalue(o);\n default: return NULL;\n }\n}\n\n\nLUA_API lua_State *lua_tothread (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n return (!ttisthread(o)) ? NULL : thvalue(o);\n}\n\n\nLUA_API const void *lua_topointer (lua_State *L, int idx) {\n StkId o = index2adr(L, idx);\n switch (ttype(o)) {\n case LUA_TTABLE: return hvalue(o);\n case LUA_TFUNCTION: return clvalue(o);\n case LUA_TTHREAD: return thvalue(o);\n case LUA_TUSERDATA:\n case LUA_TLIGHTUSERDATA:\n return lua_touserdata(L, idx);\n default: return NULL;\n }\n}\n\n\n\n/*\n** push functions (C -> stack)\n*/\n\n\nLUA_API void lua_pushnil (lua_State *L) {\n lua_lock(L);\n setnilvalue(L->top);\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushnumber (lua_State *L, lua_Number n) {\n lua_lock(L);\n setnvalue(L->top, n);\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushinteger (lua_State *L, lua_Integer n) {\n lua_lock(L);\n setnvalue(L->top, cast_num(n));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushlstring (lua_State *L, const char *s, size_t len) {\n lua_lock(L);\n luaC_checkGC(L);\n setsvalue2s(L, L->top, luaS_newlstr(L, s, len));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushstring (lua_State *L, const char *s) {\n if (s == NULL)\n lua_pushnil(L);\n else\n lua_pushlstring(L, s, strlen(s));\n}\n\n\nLUA_API const char *lua_pushvfstring (lua_State *L, const char *fmt,\n va_list argp) {\n const char *ret;\n lua_lock(L);\n luaC_checkGC(L);\n ret = luaO_pushvfstring(L, fmt, argp);\n lua_unlock(L);\n return ret;\n}\n\n\nLUA_API const char *lua_pushfstring (lua_State *L, const char *fmt, ...) {\n const char *ret;\n va_list argp;\n lua_lock(L);\n luaC_checkGC(L);\n va_start(argp, fmt);\n ret = luaO_pushvfstring(L, fmt, argp);\n va_end(argp);\n lua_unlock(L);\n return ret;\n}\n\n\nLUA_API void lua_pushcclosure (lua_State *L, lua_CFunction fn, int n) {\n Closure *cl;\n lua_lock(L);\n luaC_checkGC(L);\n api_checknelems(L, n);\n cl = luaF_newCclosure(L, n, getcurrenv(L));\n cl->c.f = fn;\n L->top -= n;\n while (n--)\n setobj2n(L, &cl->c.upvalue[n], L->top+n);\n setclvalue(L, L->top, cl);\n lua_assert(iswhite(obj2gco(cl)));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushboolean (lua_State *L, int b) {\n lua_lock(L);\n setbvalue(L->top, (b != 0)); /* ensure that true is 1 */\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_pushlightuserdata (lua_State *L, void *p) {\n lua_lock(L);\n setpvalue(L->top, p);\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API int lua_pushthread (lua_State *L) {\n lua_lock(L);\n setthvalue(L, L->top, L);\n api_incr_top(L);\n lua_unlock(L);\n return (G(L)->mainthread == L);\n}\n\n\n\n/*\n** get functions (Lua -> stack)\n*/\n\n\nLUA_API void lua_gettable (lua_State *L, int idx) {\n StkId t;\n lua_lock(L);\n t = index2adr(L, idx);\n api_checkvalidindex(L, t);\n luaV_gettable(L, t, L->top - 1, L->top - 1);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_getfield (lua_State *L, int idx, const char *k) {\n StkId t;\n TValue key;\n lua_lock(L);\n t = index2adr(L, idx);\n api_checkvalidindex(L, t);\n setsvalue(L, &key, luaS_new(L, k));\n luaV_gettable(L, t, &key, L->top);\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_rawget (lua_State *L, int idx) {\n StkId t;\n lua_lock(L);\n t = index2adr(L, idx);\n api_check(L, ttistable(t));\n setobj2s(L, L->top - 1, luaH_get(hvalue(t), L->top - 1));\n lua_unlock(L);\n}\n\n\nLUA_API void lua_rawgeti (lua_State *L, int idx, int n) {\n StkId o;\n lua_lock(L);\n o = index2adr(L, idx);\n api_check(L, ttistable(o));\n setobj2s(L, L->top, luaH_getnum(hvalue(o), n));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API void lua_createtable (lua_State *L, int narray, int nrec) {\n lua_lock(L);\n luaC_checkGC(L);\n sethvalue(L, L->top, luaH_new(L, narray, nrec));\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\nLUA_API int lua_getmetatable (lua_State *L, int objindex) {\n const TValue *obj;\n Table *mt = NULL;\n int res;\n lua_lock(L);\n obj = index2adr(L, objindex);\n switch (ttype(obj)) {\n case LUA_TTABLE:\n mt = hvalue(obj)->metatable;\n break;\n case LUA_TUSERDATA:\n mt = uvalue(obj)->metatable;\n break;\n default:\n mt = G(L)->mt[ttype(obj)];\n break;\n }\n if (mt == NULL)\n res = 0;\n else {\n sethvalue(L, L->top, mt);\n api_incr_top(L);\n res = 1;\n }\n lua_unlock(L);\n return res;\n}\n\n\nLUA_API void lua_getfenv (lua_State *L, int idx) {\n StkId o;\n lua_lock(L);\n o = index2adr(L, idx);\n api_checkvalidindex(L, o);\n switch (ttype(o)) {\n case LUA_TFUNCTION:\n sethvalue(L, L->top, clvalue(o)->c.env);\n break;\n case LUA_TUSERDATA:\n sethvalue(L, L->top, uvalue(o)->env);\n break;\n case LUA_TTHREAD:\n setobj2s(L, L->top, gt(thvalue(o)));\n break;\n default:\n setnilvalue(L->top);\n break;\n }\n api_incr_top(L);\n lua_unlock(L);\n}\n\n\n/*\n** set functions (stack -> Lua)\n*/\n\n\nLUA_API void lua_settable (lua_State *L, int idx) {\n StkId t;\n lua_lock(L);\n api_checknelems(L, 2);\n t = index2adr(L, idx);\n api_checkvalidindex(L, t);\n luaV_settable(L, t, L->top - 2, L->top - 1);\n L->top -= 2; /* pop index and value */\n lua_unlock(L);\n}\n\n\nLUA_API void lua_setfield (lua_State *L, int idx, const char *k) {\n StkId t;\n TValue key;\n lua_lock(L);\n api_checknelems(L, 1);\n t = index2adr(L, idx);\n api_checkvalidindex(L, t);\n setsvalue(L, &key, luaS_new(L, k));\n luaV_settable(L, t, &key, L->top - 1);\n L->top--; /* pop value */\n lua_unlock(L);\n}\n\n\nLUA_API void lua_rawset (lua_State *L, int idx) {\n StkId t;\n lua_lock(L);\n api_checknelems(L, 2);\n t = index2adr(L, idx);\n api_check(L, ttistable(t));\n setobj2t(L, luaH_set(L, hvalue(t), L->top-2), L->top-1);\n luaC_barriert(L, hvalue(t), L->top-1);\n L->top -= 2;\n lua_unlock(L);\n}\n\n\nLUA_API void lua_rawseti (lua_State *L, int idx, int n) {\n StkId o;\n lua_lock(L);\n api_checknelems(L, 1);\n o = index2adr(L, idx);\n api_check(L, ttistable(o));\n setobj2t(L, luaH_setnum(L, hvalue(o), n), L->top-1);\n luaC_barriert(L, hvalue(o), L->top-1);\n L->top--;\n lua_unlock(L);\n}\n\n\nLUA_API int lua_setmetatable (lua_State *L, int objindex) {\n TValue *obj;\n Table *mt;\n lua_lock(L);\n api_checknelems(L, 1);\n obj = index2adr(L, objindex);\n api_checkvalidindex(L, obj);\n if (ttisnil(L->top - 1))\n mt = NULL;\n else {\n api_check(L, ttistable(L->top - 1));\n mt = hvalue(L->top - 1);\n }\n switch (ttype(obj)) {\n case LUA_TTABLE: {\n hvalue(obj)->metatable = mt;\n if (mt)\n luaC_objbarriert(L, hvalue(obj), mt);\n break;\n }\n case LUA_TUSERDATA: {\n uvalue(obj)->metatable = mt;\n if (mt)\n luaC_objbarrier(L, rawuvalue(obj), mt);\n break;\n }\n default: {\n G(L)->mt[ttype(obj)] = mt;\n break;\n }\n }\n L->top--;\n lua_unlock(L);\n return 1;\n}\n\n\nLUA_API int lua_setfenv (lua_State *L, int idx) {\n StkId o;\n int res = 1;\n lua_lock(L);\n api_checknelems(L, 1);\n o = index2adr(L, idx);\n api_checkvalidindex(L, o);\n api_check(L, ttistable(L->top - 1));\n switch (ttype(o)) {\n case LUA_TFUNCTION:\n clvalue(o)->c.env = hvalue(L->top - 1);\n break;\n case LUA_TUSERDATA:\n uvalue(o)->env = hvalue(L->top - 1);\n break;\n case LUA_TTHREAD:\n sethvalue(L, gt(thvalue(o)), hvalue(L->top - 1));\n break;\n default:\n res = 0;\n break;\n }\n if (res) luaC_objbarrier(L, gcvalue(o), hvalue(L->top - 1));\n L->top--;\n lua_unlock(L);\n return res;\n}\n\n\n/*\n** `load' and `call' functions (run Lua code)\n*/\n\n\n#define adjustresults(L,nres) \\\n { if (nres == LUA_MULTRET && L->top >= L->ci->top) L->ci->top = L->top; }\n\n\n#define checkresults(L,na,nr) \\\n api_check(L, (nr) == LUA_MULTRET || (L->ci->top - L->top >= (nr) - (na)))\n\t\n\nLUA_API void lua_call (lua_State *L, int nargs, int nresults) {\n StkId func;\n lua_lock(L);\n api_checknelems(L, nargs+1);\n checkresults(L, nargs, nresults);\n func = L->top - (nargs+1);\n luaD_call(L, func, nresults);\n adjustresults(L, nresults);\n lua_unlock(L);\n}\n\n\n\n/*\n** Execute a protected call.\n*/\nstruct CallS { /* data to `f_call' */\n StkId func;\n int nresults;\n};\n\n\nstatic void f_call (lua_State *L, void *ud) {\n struct CallS *c = cast(struct CallS *, ud);\n luaD_call(L, c->func, c->nresults);\n}\n\n\n\nLUA_API int lua_pcall (lua_State *L, int nargs, int nresults, int errfunc) {\n struct CallS c;\n int status;\n ptrdiff_t func;\n lua_lock(L);\n api_checknelems(L, nargs+1);\n checkresults(L, nargs, nresults);\n if (errfunc == 0)\n func = 0;\n else {\n StkId o = index2adr(L, errfunc);\n api_checkvalidindex(L, o);\n func = savestack(L, o);\n }\n c.func = L->top - (nargs+1); /* function to be called */\n c.nresults = nresults;\n status = luaD_pcall(L, f_call, &c, savestack(L, c.func), func);\n adjustresults(L, nresults);\n lua_unlock(L);\n return status;\n}\n\n\n/*\n** Execute a protected C call.\n*/\nstruct CCallS { /* data to `f_Ccall' */\n lua_CFunction func;\n void *ud;\n};\n\n\nstatic void f_Ccall (lua_State *L, void *ud) {\n struct CCallS *c = cast(struct CCallS *, ud);\n Closure *cl;\n cl = luaF_newCclosure(L, 0, getcurrenv(L));\n cl->c.f = c->func;\n setclvalue(L, L->top, cl); /* push function */\n api_incr_top(L);\n setpvalue(L->top, c->ud); /* push only argument */\n api_incr_top(L);\n luaD_call(L, L->top - 2, 0);\n}\n\n\nLUA_API int lua_cpcall (lua_State *L, lua_CFunction func, void *ud) {\n struct CCallS c;\n int status;\n lua_lock(L);\n c.func = func;\n c.ud = ud;\n status = luaD_pcall(L, f_Ccall, &c, savestack(L, L->top), 0);\n lua_unlock(L);\n return status;\n}\n\n\nLUA_API int lua_load (lua_State *L, lua_Reader reader, void *data,\n const char *chunkname) {\n ZIO z;\n int status;\n lua_lock(L);\n if (!chunkname) chunkname = \"?\";\n luaZ_init(L, &z, reader, data);\n status = luaD_protectedparser(L, &z, chunkname);\n lua_unlock(L);\n return status;\n}\n\n\nLUA_API int lua_dump (lua_State *L, lua_Writer writer, void *data) {\n int status;\n TValue *o;\n lua_lock(L);\n api_checknelems(L, 1);\n o = L->top - 1;\n if (isLfunction(o))\n status = luaU_dump(L, clvalue(o)->l.p, writer, data, 0);\n else\n status = 1;\n lua_unlock(L);\n return status;\n}\n\n\nLUA_API int lua_status (lua_State *L) {\n return L->status;\n}\n\n\n/*\n** Garbage-collection function\n*/\n\nLUA_API int lua_gc (lua_State *L, int what, int data) {\n int res = 0;\n global_State *g;\n lua_lock(L);\n g = G(L);\n switch (what) {\n case LUA_GCSTOP: {\n g->GCthreshold = MAX_LUMEM;\n break;\n }\n case LUA_GCRESTART: {\n g->GCthreshold = g->totalbytes;\n break;\n }\n case LUA_GCCOLLECT: {\n luaC_fullgc(L);\n break;\n }\n case LUA_GCCOUNT: {\n /* GC values are expressed in Kbytes: #bytes/2^10 */\n res = cast_int(g->totalbytes >> 10);\n break;\n }\n case LUA_GCCOUNTB: {\n res = cast_int(g->totalbytes & 0x3ff);\n break;\n }\n case LUA_GCSTEP: {\n lu_mem a = (cast(lu_mem, data) << 10);\n if (a <= g->totalbytes)\n g->GCthreshold = g->totalbytes - a;\n else\n g->GCthreshold = 0;\n while (g->GCthreshold <= g->totalbytes) {\n luaC_step(L);\n if (g->gcstate == GCSpause) { /* end of cycle? */\n res = 1; /* signal it */\n break;\n }\n }\n break;\n }\n case LUA_GCSETPAUSE: {\n res = g->gcpause;\n g->gcpause = data;\n break;\n }\n case LUA_GCSETSTEPMUL: {\n res = g->gcstepmul;\n g->gcstepmul = data;\n break;\n }\n default: res = -1; /* invalid option */\n }\n lua_unlock(L);\n return res;\n}\n\n\n\n/*\n** miscellaneous functions\n*/\n\n\nLUA_API int lua_error (lua_State *L) {\n lua_lock(L);\n api_checknelems(L, 1);\n luaG_errormsg(L);\n lua_unlock(L);\n return 0; /* to avoid warnings */\n}\n\n\nLUA_API int lua_next (lua_State *L, int idx) {\n StkId t;\n int more;\n lua_lock(L);\n t = index2adr(L, idx);\n api_check(L, ttistable(t));\n more = luaH_next(L, hvalue(t), L->top - 1);\n if (more) {\n api_incr_top(L);\n }\n else /* no more elements */\n L->top -= 1; /* remove key */\n lua_unlock(L);\n return more;\n}\n\n\nLUA_API void lua_concat (lua_State *L, int n) {\n lua_lock(L);\n api_checknelems(L, n);\n if (n >= 2) {\n luaC_checkGC(L);\n luaV_concat(L, n, cast_int(L->top - L->base) - 1);\n L->top -= (n-1);\n }\n else if (n == 0) { /* push empty string */\n setsvalue2s(L, L->top, luaS_newlstr(L, \"\", 0));\n api_incr_top(L);\n }\n /* else n == 1; nothing to do */\n lua_unlock(L);\n}\n\n\nLUA_API lua_Alloc lua_getallocf (lua_State *L, void **ud) {\n lua_Alloc f;\n lua_lock(L);\n if (ud) *ud = G(L)->ud;\n f = G(L)->frealloc;\n lua_unlock(L);\n return f;\n}\n\n\nLUA_API void lua_setallocf (lua_State *L, lua_Alloc f, void *ud) {\n lua_lock(L);\n G(L)->ud = ud;\n G(L)->frealloc = f;\n lua_unlock(L);\n}\n\n\nLUA_API void *lua_newuserdata (lua_State *L, size_t size) {\n Udata *u;\n lua_lock(L);\n luaC_checkGC(L);\n u = luaS_newudata(L, size, getcurrenv(L));\n setuvalue(L, L->top, u);\n api_incr_top(L);\n lua_unlock(L);\n return u + 1;\n}\n\n\n\n\nstatic const char *aux_upvalue (StkId fi, int n, TValue **val) {\n Closure *f;\n if (!ttisfunction(fi)) return NULL;\n f = clvalue(fi);\n if (f->c.isC) {\n if (!(1 <= n && n <= f->c.nupvalues)) return NULL;\n *val = &f->c.upvalue[n-1];\n return \"\";\n }\n else {\n Proto *p = f->l.p;\n if (!(1 <= n && n <= p->sizeupvalues)) return NULL;\n *val = f->l.upvals[n-1]->v;\n return getstr(p->upvalues[n-1]);\n }\n}\n\n\nLUA_API const char *lua_getupvalue (lua_State *L, int funcindex, int n) {\n const char *name;\n TValue *val;\n lua_lock(L);\n name = aux_upvalue(index2adr(L, funcindex), n, &val);\n if (name) {\n setobj2s(L, L->top, val);\n api_incr_top(L);\n }\n lua_unlock(L);\n return name;\n}\n\n\nLUA_API const char *lua_setupvalue (lua_State *L, int funcindex, int n) {\n const char *name;\n TValue *val;\n StkId fi;\n lua_lock(L);\n fi = index2adr(L, funcindex);\n api_checknelems(L, 1);\n name = aux_upvalue(fi, n, &val);\n if (name) {\n L->top--;\n setobj(L, val, L->top);\n luaC_barrier(L, clvalue(fi), L->top);\n }\n lua_unlock(L);\n return name;\n}\n\n" }, { "path": "deps/lua/src/lapi.h", "content": "/*\n** $Id: lapi.h,v 2.2.1.1 2007/12/27 13:02:25 roberto Exp $\n** Auxiliary functions from Lua API\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lapi_h\n#define lapi_h\n\n\n#include \"lobject.h\"\n\n\nLUAI_FUNC void luaA_pushobject (lua_State *L, const TValue *o);\n\n#endif\n" }, { "path": "deps/lua/src/lauxlib.c", "content": "/*\n** $Id: lauxlib.c,v 1.159.1.3 2008/01/21 13:20:51 roberto Exp $\n** Auxiliary functions for building Lua libraries\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n\n/* This file uses only the official API of Lua.\n** Any function declared here could be written as an application function.\n*/\n\n#define lauxlib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n\n\n#define FREELIST_REF\t0\t/* free list of references */\n\n\n/* convert a stack index to positive */\n#define abs_index(L, i)\t\t((i) > 0 || (i) <= LUA_REGISTRYINDEX ? (i) : \\\n\t\t\t\t\tlua_gettop(L) + (i) + 1)\n\n\n/*\n** {======================================================\n** Error-report functions\n** =======================================================\n*/\n\n\nLUALIB_API int luaL_argerror (lua_State *L, int narg, const char *extramsg) {\n lua_Debug ar;\n if (!lua_getstack(L, 0, &ar)) /* no stack frame? */\n return luaL_error(L, \"bad argument #%d (%s)\", narg, extramsg);\n lua_getinfo(L, \"n\", &ar);\n if (strcmp(ar.namewhat, \"method\") == 0) {\n narg--; /* do not count `self' */\n if (narg == 0) /* error is in the self argument itself? */\n return luaL_error(L, \"calling \" LUA_QS \" on bad self (%s)\",\n ar.name, extramsg);\n }\n if (ar.name == NULL)\n ar.name = \"?\";\n return luaL_error(L, \"bad argument #%d to \" LUA_QS \" (%s)\",\n narg, ar.name, extramsg);\n}\n\n\nLUALIB_API int luaL_typerror (lua_State *L, int narg, const char *tname) {\n const char *msg = lua_pushfstring(L, \"%s expected, got %s\",\n tname, luaL_typename(L, narg));\n return luaL_argerror(L, narg, msg);\n}\n\n\nstatic void tag_error (lua_State *L, int narg, int tag) {\n luaL_typerror(L, narg, lua_typename(L, tag));\n}\n\n\nLUALIB_API void luaL_where (lua_State *L, int level) {\n lua_Debug ar;\n if (lua_getstack(L, level, &ar)) { /* check function at level */\n lua_getinfo(L, \"Sl\", &ar); /* get info about it */\n if (ar.currentline > 0) { /* is there info? */\n lua_pushfstring(L, \"%s:%d: \", ar.short_src, ar.currentline);\n return;\n }\n }\n lua_pushliteral(L, \"\"); /* else, no information available... */\n}\n\n\nLUALIB_API int luaL_error (lua_State *L, const char *fmt, ...) {\n va_list argp;\n va_start(argp, fmt);\n luaL_where(L, 1);\n lua_pushvfstring(L, fmt, argp);\n va_end(argp);\n lua_concat(L, 2);\n return lua_error(L);\n}\n\n/* }====================================================== */\n\n\nLUALIB_API int luaL_checkoption (lua_State *L, int narg, const char *def,\n const char *const lst[]) {\n const char *name = (def) ? luaL_optstring(L, narg, def) :\n luaL_checkstring(L, narg);\n int i;\n for (i=0; lst[i]; i++)\n if (strcmp(lst[i], name) == 0)\n return i;\n return luaL_argerror(L, narg,\n lua_pushfstring(L, \"invalid option \" LUA_QS, name));\n}\n\n\nLUALIB_API int luaL_newmetatable (lua_State *L, const char *tname) {\n lua_getfield(L, LUA_REGISTRYINDEX, tname); /* get registry.name */\n if (!lua_isnil(L, -1)) /* name already in use? */\n return 0; /* leave previous value on top, but return 0 */\n lua_pop(L, 1);\n lua_newtable(L); /* create metatable */\n lua_pushvalue(L, -1);\n lua_setfield(L, LUA_REGISTRYINDEX, tname); /* registry.name = metatable */\n return 1;\n}\n\n\nLUALIB_API void *luaL_checkudata (lua_State *L, int ud, const char *tname) {\n void *p = lua_touserdata(L, ud);\n if (p != NULL) { /* value is a userdata? */\n if (lua_getmetatable(L, ud)) { /* does it have a metatable? */\n lua_getfield(L, LUA_REGISTRYINDEX, tname); /* get correct metatable */\n if (lua_rawequal(L, -1, -2)) { /* does it have the correct mt? */\n lua_pop(L, 2); /* remove both metatables */\n return p;\n }\n }\n }\n luaL_typerror(L, ud, tname); /* else error */\n return NULL; /* to avoid warnings */\n}\n\n\nLUALIB_API void luaL_checkstack (lua_State *L, int space, const char *mes) {\n if (!lua_checkstack(L, space))\n luaL_error(L, \"stack overflow (%s)\", mes);\n}\n\n\nLUALIB_API void luaL_checktype (lua_State *L, int narg, int t) {\n if (lua_type(L, narg) != t)\n tag_error(L, narg, t);\n}\n\n\nLUALIB_API void luaL_checkany (lua_State *L, int narg) {\n if (lua_type(L, narg) == LUA_TNONE)\n luaL_argerror(L, narg, \"value expected\");\n}\n\n\nLUALIB_API const char *luaL_checklstring (lua_State *L, int narg, size_t *len) {\n const char *s = lua_tolstring(L, narg, len);\n if (!s) tag_error(L, narg, LUA_TSTRING);\n return s;\n}\n\n\nLUALIB_API const char *luaL_optlstring (lua_State *L, int narg,\n const char *def, size_t *len) {\n if (lua_isnoneornil(L, narg)) {\n if (len)\n *len = (def ? strlen(def) : 0);\n return def;\n }\n else return luaL_checklstring(L, narg, len);\n}\n\n\nLUALIB_API lua_Number luaL_checknumber (lua_State *L, int narg) {\n lua_Number d = lua_tonumber(L, narg);\n if (d == 0 && !lua_isnumber(L, narg)) /* avoid extra test when d is not 0 */\n tag_error(L, narg, LUA_TNUMBER);\n return d;\n}\n\n\nLUALIB_API lua_Number luaL_optnumber (lua_State *L, int narg, lua_Number def) {\n return luaL_opt(L, luaL_checknumber, narg, def);\n}\n\n\nLUALIB_API lua_Integer luaL_checkinteger (lua_State *L, int narg) {\n lua_Integer d = lua_tointeger(L, narg);\n if (d == 0 && !lua_isnumber(L, narg)) /* avoid extra test when d is not 0 */\n tag_error(L, narg, LUA_TNUMBER);\n return d;\n}\n\n\nLUALIB_API lua_Integer luaL_optinteger (lua_State *L, int narg,\n lua_Integer def) {\n return luaL_opt(L, luaL_checkinteger, narg, def);\n}\n\n\nLUALIB_API int luaL_getmetafield (lua_State *L, int obj, const char *event) {\n if (!lua_getmetatable(L, obj)) /* no metatable? */\n return 0;\n lua_pushstring(L, event);\n lua_rawget(L, -2);\n if (lua_isnil(L, -1)) {\n lua_pop(L, 2); /* remove metatable and metafield */\n return 0;\n }\n else {\n lua_remove(L, -2); /* remove only metatable */\n return 1;\n }\n}\n\n\nLUALIB_API int luaL_callmeta (lua_State *L, int obj, const char *event) {\n obj = abs_index(L, obj);\n if (!luaL_getmetafield(L, obj, event)) /* no metafield? */\n return 0;\n lua_pushvalue(L, obj);\n lua_call(L, 1, 1);\n return 1;\n}\n\n\nLUALIB_API void (luaL_register) (lua_State *L, const char *libname,\n const luaL_Reg *l) {\n luaI_openlib(L, libname, l, 0);\n}\n\n\nstatic int libsize (const luaL_Reg *l) {\n int size = 0;\n for (; l->name; l++) size++;\n return size;\n}\n\n\nLUALIB_API void luaI_openlib (lua_State *L, const char *libname,\n const luaL_Reg *l, int nup) {\n if (libname) {\n int size = libsize(l);\n /* check whether lib already exists */\n luaL_findtable(L, LUA_REGISTRYINDEX, \"_LOADED\", 1);\n lua_getfield(L, -1, libname); /* get _LOADED[libname] */\n if (!lua_istable(L, -1)) { /* not found? */\n lua_pop(L, 1); /* remove previous result */\n /* try global variable (and create one if it does not exist) */\n if (luaL_findtable(L, LUA_GLOBALSINDEX, libname, size) != NULL)\n luaL_error(L, \"name conflict for module \" LUA_QS, libname);\n lua_pushvalue(L, -1);\n lua_setfield(L, -3, libname); /* _LOADED[libname] = new table */\n }\n lua_remove(L, -2); /* remove _LOADED table */\n lua_insert(L, -(nup+1)); /* move library table to below upvalues */\n }\n for (; l->name; l++) {\n int i;\n for (i=0; ifunc, nup);\n lua_setfield(L, -(nup+2), l->name);\n }\n lua_pop(L, nup); /* remove upvalues */\n}\n\n\n\n/*\n** {======================================================\n** getn-setn: size for arrays\n** =======================================================\n*/\n\n#if defined(LUA_COMPAT_GETN)\n\nstatic int checkint (lua_State *L, int topop) {\n int n = (lua_type(L, -1) == LUA_TNUMBER) ? lua_tointeger(L, -1) : -1;\n lua_pop(L, topop);\n return n;\n}\n\n\nstatic void getsizes (lua_State *L) {\n lua_getfield(L, LUA_REGISTRYINDEX, \"LUA_SIZES\");\n if (lua_isnil(L, -1)) { /* no `size' table? */\n lua_pop(L, 1); /* remove nil */\n lua_newtable(L); /* create it */\n lua_pushvalue(L, -1); /* `size' will be its own metatable */\n lua_setmetatable(L, -2);\n lua_pushliteral(L, \"kv\");\n lua_setfield(L, -2, \"__mode\"); /* metatable(N).__mode = \"kv\" */\n lua_pushvalue(L, -1);\n lua_setfield(L, LUA_REGISTRYINDEX, \"LUA_SIZES\"); /* store in register */\n }\n}\n\n\nLUALIB_API void luaL_setn (lua_State *L, int t, int n) {\n t = abs_index(L, t);\n lua_pushliteral(L, \"n\");\n lua_rawget(L, t);\n if (checkint(L, 1) >= 0) { /* is there a numeric field `n'? */\n lua_pushliteral(L, \"n\"); /* use it */\n lua_pushinteger(L, n);\n lua_rawset(L, t);\n }\n else { /* use `sizes' */\n getsizes(L);\n lua_pushvalue(L, t);\n lua_pushinteger(L, n);\n lua_rawset(L, -3); /* sizes[t] = n */\n lua_pop(L, 1); /* remove `sizes' */\n }\n}\n\n\nLUALIB_API int luaL_getn (lua_State *L, int t) {\n int n;\n t = abs_index(L, t);\n lua_pushliteral(L, \"n\"); /* try t.n */\n lua_rawget(L, t);\n if ((n = checkint(L, 1)) >= 0) return n;\n getsizes(L); /* else try sizes[t] */\n lua_pushvalue(L, t);\n lua_rawget(L, -2);\n if ((n = checkint(L, 2)) >= 0) return n;\n return (int)lua_objlen(L, t);\n}\n\n#endif\n\n/* }====================================================== */\n\n\n\nLUALIB_API const char *luaL_gsub (lua_State *L, const char *s, const char *p,\n const char *r) {\n const char *wild;\n size_t l = strlen(p);\n luaL_Buffer b;\n luaL_buffinit(L, &b);\n while ((wild = strstr(s, p)) != NULL) {\n luaL_addlstring(&b, s, wild - s); /* push prefix */\n luaL_addstring(&b, r); /* push replacement in place of pattern */\n s = wild + l; /* continue after `p' */\n }\n luaL_addstring(&b, s); /* push last suffix */\n luaL_pushresult(&b);\n return lua_tostring(L, -1);\n}\n\n\nLUALIB_API const char *luaL_findtable (lua_State *L, int idx,\n const char *fname, int szhint) {\n const char *e;\n lua_pushvalue(L, idx);\n do {\n e = strchr(fname, '.');\n if (e == NULL) e = fname + strlen(fname);\n lua_pushlstring(L, fname, e - fname);\n lua_rawget(L, -2);\n if (lua_isnil(L, -1)) { /* no such field? */\n lua_pop(L, 1); /* remove this nil */\n lua_createtable(L, 0, (*e == '.' ? 1 : szhint)); /* new table for field */\n lua_pushlstring(L, fname, e - fname);\n lua_pushvalue(L, -2);\n lua_settable(L, -4); /* set new table into field */\n }\n else if (!lua_istable(L, -1)) { /* field has a non-table value? */\n lua_pop(L, 2); /* remove table and value */\n return fname; /* return problematic part of the name */\n }\n lua_remove(L, -2); /* remove previous table */\n fname = e + 1;\n } while (*e == '.');\n return NULL;\n}\n\n\n\n/*\n** {======================================================\n** Generic Buffer manipulation\n** =======================================================\n*/\n\n\n#define bufflen(B)\t((B)->p - (B)->buffer)\n#define bufffree(B)\t((size_t)(LUAL_BUFFERSIZE - bufflen(B)))\n\n#define LIMIT\t(LUA_MINSTACK/2)\n\n\nstatic int emptybuffer (luaL_Buffer *B) {\n size_t l = bufflen(B);\n if (l == 0) return 0; /* put nothing on stack */\n else {\n lua_pushlstring(B->L, B->buffer, l);\n B->p = B->buffer;\n B->lvl++;\n return 1;\n }\n}\n\n\nstatic void adjuststack (luaL_Buffer *B) {\n if (B->lvl > 1) {\n lua_State *L = B->L;\n int toget = 1; /* number of levels to concat */\n size_t toplen = lua_strlen(L, -1);\n do {\n size_t l = lua_strlen(L, -(toget+1));\n if (B->lvl - toget + 1 >= LIMIT || toplen > l) {\n toplen += l;\n toget++;\n }\n else break;\n } while (toget < B->lvl);\n lua_concat(L, toget);\n B->lvl = B->lvl - toget + 1;\n }\n}\n\n\nLUALIB_API char *luaL_prepbuffer (luaL_Buffer *B) {\n if (emptybuffer(B))\n adjuststack(B);\n return B->buffer;\n}\n\n\nLUALIB_API void luaL_addlstring (luaL_Buffer *B, const char *s, size_t l) {\n while (l--)\n luaL_addchar(B, *s++);\n}\n\n\nLUALIB_API void luaL_addstring (luaL_Buffer *B, const char *s) {\n luaL_addlstring(B, s, strlen(s));\n}\n\n\nLUALIB_API void luaL_pushresult (luaL_Buffer *B) {\n emptybuffer(B);\n lua_concat(B->L, B->lvl);\n B->lvl = 1;\n}\n\n\nLUALIB_API void luaL_addvalue (luaL_Buffer *B) {\n lua_State *L = B->L;\n size_t vl;\n const char *s = lua_tolstring(L, -1, &vl);\n if (vl <= bufffree(B)) { /* fit into buffer? */\n memcpy(B->p, s, vl); /* put it there */\n B->p += vl;\n lua_pop(L, 1); /* remove from stack */\n }\n else {\n if (emptybuffer(B))\n lua_insert(L, -2); /* put buffer before new value */\n B->lvl++; /* add new value into B stack */\n adjuststack(B);\n }\n}\n\n\nLUALIB_API void luaL_buffinit (lua_State *L, luaL_Buffer *B) {\n B->L = L;\n B->p = B->buffer;\n B->lvl = 0;\n}\n\n/* }====================================================== */\n\n\nLUALIB_API int luaL_ref (lua_State *L, int t) {\n int ref;\n t = abs_index(L, t);\n if (lua_isnil(L, -1)) {\n lua_pop(L, 1); /* remove from stack */\n return LUA_REFNIL; /* `nil' has a unique fixed reference */\n }\n lua_rawgeti(L, t, FREELIST_REF); /* get first free element */\n ref = (int)lua_tointeger(L, -1); /* ref = t[FREELIST_REF] */\n lua_pop(L, 1); /* remove it from stack */\n if (ref != 0) { /* any free element? */\n lua_rawgeti(L, t, ref); /* remove it from list */\n lua_rawseti(L, t, FREELIST_REF); /* (t[FREELIST_REF] = t[ref]) */\n }\n else { /* no free elements */\n ref = (int)lua_objlen(L, t);\n ref++; /* create new reference */\n }\n lua_rawseti(L, t, ref);\n return ref;\n}\n\n\nLUALIB_API void luaL_unref (lua_State *L, int t, int ref) {\n if (ref >= 0) {\n t = abs_index(L, t);\n lua_rawgeti(L, t, FREELIST_REF);\n lua_rawseti(L, t, ref); /* t[ref] = t[FREELIST_REF] */\n lua_pushinteger(L, ref);\n lua_rawseti(L, t, FREELIST_REF); /* t[FREELIST_REF] = ref */\n }\n}\n\n\n\n/*\n** {======================================================\n** Load functions\n** =======================================================\n*/\n\ntypedef struct LoadF {\n int extraline;\n FILE *f;\n char buff[LUAL_BUFFERSIZE];\n} LoadF;\n\n\nstatic const char *getF (lua_State *L, void *ud, size_t *size) {\n LoadF *lf = (LoadF *)ud;\n (void)L;\n if (lf->extraline) {\n lf->extraline = 0;\n *size = 1;\n return \"\\n\";\n }\n if (feof(lf->f)) return NULL;\n *size = fread(lf->buff, 1, sizeof(lf->buff), lf->f);\n return (*size > 0) ? lf->buff : NULL;\n}\n\n\nstatic int errfile (lua_State *L, const char *what, int fnameindex) {\n const char *serr = strerror(errno);\n const char *filename = lua_tostring(L, fnameindex) + 1;\n lua_pushfstring(L, \"cannot %s %s: %s\", what, filename, serr);\n lua_remove(L, fnameindex);\n return LUA_ERRFILE;\n}\n\n\nLUALIB_API int luaL_loadfile (lua_State *L, const char *filename) {\n LoadF lf;\n int status, readstatus;\n int c;\n int fnameindex = lua_gettop(L) + 1; /* index of filename on the stack */\n lf.extraline = 0;\n if (filename == NULL) {\n lua_pushliteral(L, \"=stdin\");\n lf.f = stdin;\n }\n else {\n lua_pushfstring(L, \"@%s\", filename);\n lf.f = fopen(filename, \"r\");\n if (lf.f == NULL) return errfile(L, \"open\", fnameindex);\n }\n c = getc(lf.f);\n if (c == '#') { /* Unix exec. file? */\n lf.extraline = 1;\n while ((c = getc(lf.f)) != EOF && c != '\\n') ; /* skip first line */\n if (c == '\\n') c = getc(lf.f);\n }\n if (c == LUA_SIGNATURE[0] && filename) { /* binary file? */\n lf.f = freopen(filename, \"rb\", lf.f); /* reopen in binary mode */\n if (lf.f == NULL) return errfile(L, \"reopen\", fnameindex);\n /* skip eventual `#!...' */\n while ((c = getc(lf.f)) != EOF && c != LUA_SIGNATURE[0]) ;\n lf.extraline = 0;\n }\n ungetc(c, lf.f);\n status = lua_load(L, getF, &lf, lua_tostring(L, -1));\n readstatus = ferror(lf.f);\n if (filename) fclose(lf.f); /* close file (even in case of errors) */\n if (readstatus) {\n lua_settop(L, fnameindex); /* ignore results from `lua_load' */\n return errfile(L, \"read\", fnameindex);\n }\n lua_remove(L, fnameindex);\n return status;\n}\n\n\ntypedef struct LoadS {\n const char *s;\n size_t size;\n} LoadS;\n\n\nstatic const char *getS (lua_State *L, void *ud, size_t *size) {\n LoadS *ls = (LoadS *)ud;\n (void)L;\n if (ls->size == 0) return NULL;\n *size = ls->size;\n ls->size = 0;\n return ls->s;\n}\n\n\nLUALIB_API int luaL_loadbuffer (lua_State *L, const char *buff, size_t size,\n const char *name) {\n LoadS ls;\n ls.s = buff;\n ls.size = size;\n return lua_load(L, getS, &ls, name);\n}\n\n\nLUALIB_API int (luaL_loadstring) (lua_State *L, const char *s) {\n return luaL_loadbuffer(L, s, strlen(s), s);\n}\n\n\n\n/* }====================================================== */\n\n\nstatic void *l_alloc (void *ud, void *ptr, size_t osize, size_t nsize) {\n (void)ud;\n (void)osize;\n if (nsize == 0) {\n free(ptr);\n return NULL;\n }\n else\n return realloc(ptr, nsize);\n}\n\n\nstatic int panic (lua_State *L) {\n (void)L; /* to avoid warnings */\n fprintf(stderr, \"PANIC: unprotected error in call to Lua API (%s)\\n\",\n lua_tostring(L, -1));\n return 0;\n}\n\n\nLUALIB_API lua_State *luaL_newstate (void) {\n lua_State *L = lua_newstate(l_alloc, NULL);\n if (L) lua_atpanic(L, &panic);\n return L;\n}\n\n" }, { "path": "deps/lua/src/lauxlib.h", "content": "/*\n** $Id: lauxlib.h,v 1.88.1.1 2007/12/27 13:02:25 roberto Exp $\n** Auxiliary functions for building Lua libraries\n** See Copyright Notice in lua.h\n*/\n\n\n#ifndef lauxlib_h\n#define lauxlib_h\n\n\n#include \n#include \n\n#include \"lua.h\"\n\n\n#if defined(LUA_COMPAT_GETN)\nLUALIB_API int (luaL_getn) (lua_State *L, int t);\nLUALIB_API void (luaL_setn) (lua_State *L, int t, int n);\n#else\n#define luaL_getn(L,i) ((int)lua_objlen(L, i))\n#define luaL_setn(L,i,j) ((void)0) /* no op! */\n#endif\n\n#if defined(LUA_COMPAT_OPENLIB)\n#define luaI_openlib\tluaL_openlib\n#endif\n\n\n/* extra error code for `luaL_load' */\n#define LUA_ERRFILE (LUA_ERRERR+1)\n\n\ntypedef struct luaL_Reg {\n const char *name;\n lua_CFunction func;\n} luaL_Reg;\n\n\n\nLUALIB_API void (luaI_openlib) (lua_State *L, const char *libname,\n const luaL_Reg *l, int nup);\nLUALIB_API void (luaL_register) (lua_State *L, const char *libname,\n const luaL_Reg *l);\nLUALIB_API int (luaL_getmetafield) (lua_State *L, int obj, const char *e);\nLUALIB_API int (luaL_callmeta) (lua_State *L, int obj, const char *e);\nLUALIB_API int (luaL_typerror) (lua_State *L, int narg, const char *tname);\nLUALIB_API int (luaL_argerror) (lua_State *L, int numarg, const char *extramsg);\nLUALIB_API const char *(luaL_checklstring) (lua_State *L, int numArg,\n size_t *l);\nLUALIB_API const char *(luaL_optlstring) (lua_State *L, int numArg,\n const char *def, size_t *l);\nLUALIB_API lua_Number (luaL_checknumber) (lua_State *L, int numArg);\nLUALIB_API lua_Number (luaL_optnumber) (lua_State *L, int nArg, lua_Number def);\n\nLUALIB_API lua_Integer (luaL_checkinteger) (lua_State *L, int numArg);\nLUALIB_API lua_Integer (luaL_optinteger) (lua_State *L, int nArg,\n lua_Integer def);\n\nLUALIB_API void (luaL_checkstack) (lua_State *L, int sz, const char *msg);\nLUALIB_API void (luaL_checktype) (lua_State *L, int narg, int t);\nLUALIB_API void (luaL_checkany) (lua_State *L, int narg);\n\nLUALIB_API int (luaL_newmetatable) (lua_State *L, const char *tname);\nLUALIB_API void *(luaL_checkudata) (lua_State *L, int ud, const char *tname);\n\nLUALIB_API void (luaL_where) (lua_State *L, int lvl);\nLUALIB_API int (luaL_error) (lua_State *L, const char *fmt, ...);\n\nLUALIB_API int (luaL_checkoption) (lua_State *L, int narg, const char *def,\n const char *const lst[]);\n\nLUALIB_API int (luaL_ref) (lua_State *L, int t);\nLUALIB_API void (luaL_unref) (lua_State *L, int t, int ref);\n\nLUALIB_API int (luaL_loadfile) (lua_State *L, const char *filename);\nLUALIB_API int (luaL_loadbuffer) (lua_State *L, const char *buff, size_t sz,\n const char *name);\nLUALIB_API int (luaL_loadstring) (lua_State *L, const char *s);\n\nLUALIB_API lua_State *(luaL_newstate) (void);\n\n\nLUALIB_API const char *(luaL_gsub) (lua_State *L, const char *s, const char *p,\n const char *r);\n\nLUALIB_API const char *(luaL_findtable) (lua_State *L, int idx,\n const char *fname, int szhint);\n\n\n\n\n/*\n** ===============================================================\n** some useful macros\n** ===============================================================\n*/\n\n#define luaL_argcheck(L, cond,numarg,extramsg)\t\\\n\t\t((void)((cond) || luaL_argerror(L, (numarg), (extramsg))))\n#define luaL_checkstring(L,n)\t(luaL_checklstring(L, (n), NULL))\n#define luaL_optstring(L,n,d)\t(luaL_optlstring(L, (n), (d), NULL))\n#define luaL_checkint(L,n)\t((int)luaL_checkinteger(L, (n)))\n#define luaL_optint(L,n,d)\t((int)luaL_optinteger(L, (n), (d)))\n#define luaL_checklong(L,n)\t((long)luaL_checkinteger(L, (n)))\n#define luaL_optlong(L,n,d)\t((long)luaL_optinteger(L, (n), (d)))\n\n#define luaL_typename(L,i)\tlua_typename(L, lua_type(L,(i)))\n\n#define luaL_dofile(L, fn) \\\n\t(luaL_loadfile(L, fn) || lua_pcall(L, 0, LUA_MULTRET, 0))\n\n#define luaL_dostring(L, s) \\\n\t(luaL_loadstring(L, s) || lua_pcall(L, 0, LUA_MULTRET, 0))\n\n#define luaL_getmetatable(L,n)\t(lua_getfield(L, LUA_REGISTRYINDEX, (n)))\n\n#define luaL_opt(L,f,n,d)\t(lua_isnoneornil(L,(n)) ? (d) : f(L,(n)))\n\n/*\n** {======================================================\n** Generic Buffer manipulation\n** =======================================================\n*/\n\n\n\ntypedef struct luaL_Buffer {\n char *p;\t\t\t/* current position in buffer */\n int lvl; /* number of strings in the stack (level) */\n lua_State *L;\n char buffer[LUAL_BUFFERSIZE];\n} luaL_Buffer;\n\n#define luaL_addchar(B,c) \\\n ((void)((B)->p < ((B)->buffer+LUAL_BUFFERSIZE) || luaL_prepbuffer(B)), \\\n (*(B)->p++ = (char)(c)))\n\n/* compatibility only */\n#define luaL_putchar(B,c)\tluaL_addchar(B,c)\n\n#define luaL_addsize(B,n)\t((B)->p += (n))\n\nLUALIB_API void (luaL_buffinit) (lua_State *L, luaL_Buffer *B);\nLUALIB_API char *(luaL_prepbuffer) (luaL_Buffer *B);\nLUALIB_API void (luaL_addlstring) (luaL_Buffer *B, const char *s, size_t l);\nLUALIB_API void (luaL_addstring) (luaL_Buffer *B, const char *s);\nLUALIB_API void (luaL_addvalue) (luaL_Buffer *B);\nLUALIB_API void (luaL_pushresult) (luaL_Buffer *B);\n\n\n/* }====================================================== */\n\n\n/* compatibility with ref system */\n\n/* pre-defined references */\n#define LUA_NOREF (-2)\n#define LUA_REFNIL (-1)\n\n#define lua_ref(L,lock) ((lock) ? luaL_ref(L, LUA_REGISTRYINDEX) : \\\n (lua_pushstring(L, \"unlocked references are obsolete\"), lua_error(L), 0))\n\n#define lua_unref(L,ref) luaL_unref(L, LUA_REGISTRYINDEX, (ref))\n\n#define lua_getref(L,ref) lua_rawgeti(L, LUA_REGISTRYINDEX, (ref))\n\n\n#define luaL_reg\tluaL_Reg\n\n#endif\n\n\n" }, { "path": "deps/lua/src/lbaselib.c", "content": "/*\n** $Id: lbaselib.c,v 1.191.1.6 2008/02/14 16:46:22 roberto Exp $\n** Basic library\n** See Copyright Notice in lua.h\n*/\n\n\n\n#include \n#include \n#include \n#include \n\n#define lbaselib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n\n\n/*\n** If your system does not support `stdout', you can just remove this function.\n** If you need, you can define your own `print' function, following this\n** model but changing `fputs' to put the strings at a proper place\n** (a console window or a log file, for instance).\n*/\nstatic int luaB_print (lua_State *L) {\n int n = lua_gettop(L); /* number of arguments */\n int i;\n lua_getglobal(L, \"tostring\");\n for (i=1; i<=n; i++) {\n const char *s;\n lua_pushvalue(L, -1); /* function to be called */\n lua_pushvalue(L, i); /* value to print */\n lua_call(L, 1, 1);\n s = lua_tostring(L, -1); /* get result */\n if (s == NULL)\n return luaL_error(L, LUA_QL(\"tostring\") \" must return a string to \"\n LUA_QL(\"print\"));\n if (i>1) fputs(\"\\t\", stdout);\n fputs(s, stdout);\n lua_pop(L, 1); /* pop result */\n }\n fputs(\"\\n\", stdout);\n return 0;\n}\n\n\nstatic int luaB_tonumber (lua_State *L) {\n int base = luaL_optint(L, 2, 10);\n if (base == 10) { /* standard conversion */\n luaL_checkany(L, 1);\n if (lua_isnumber(L, 1)) {\n lua_pushnumber(L, lua_tonumber(L, 1));\n return 1;\n }\n }\n else {\n const char *s1 = luaL_checkstring(L, 1);\n char *s2;\n unsigned long n;\n luaL_argcheck(L, 2 <= base && base <= 36, 2, \"base out of range\");\n n = strtoul(s1, &s2, base);\n if (s1 != s2) { /* at least one valid digit? */\n while (isspace((unsigned char)(*s2))) s2++; /* skip trailing spaces */\n if (*s2 == '\\0') { /* no invalid trailing characters? */\n lua_pushnumber(L, (lua_Number)n);\n return 1;\n }\n }\n }\n lua_pushnil(L); /* else not a number */\n return 1;\n}\n\n\nstatic int luaB_error (lua_State *L) {\n int level = luaL_optint(L, 2, 1);\n lua_settop(L, 1);\n if (lua_isstring(L, 1) && level > 0) { /* add extra information? */\n luaL_where(L, level);\n lua_pushvalue(L, 1);\n lua_concat(L, 2);\n }\n return lua_error(L);\n}\n\n\nstatic int luaB_getmetatable (lua_State *L) {\n luaL_checkany(L, 1);\n if (!lua_getmetatable(L, 1)) {\n lua_pushnil(L);\n return 1; /* no metatable */\n }\n luaL_getmetafield(L, 1, \"__metatable\");\n return 1; /* returns either __metatable field (if present) or metatable */\n}\n\n\nstatic int luaB_setmetatable (lua_State *L) {\n int t = lua_type(L, 2);\n luaL_checktype(L, 1, LUA_TTABLE);\n luaL_argcheck(L, t == LUA_TNIL || t == LUA_TTABLE, 2,\n \"nil or table expected\");\n if (luaL_getmetafield(L, 1, \"__metatable\"))\n luaL_error(L, \"cannot change a protected metatable\");\n lua_settop(L, 2);\n lua_setmetatable(L, 1);\n return 1;\n}\n\n\nstatic void getfunc (lua_State *L, int opt) {\n if (lua_isfunction(L, 1)) lua_pushvalue(L, 1);\n else {\n lua_Debug ar;\n int level = opt ? luaL_optint(L, 1, 1) : luaL_checkint(L, 1);\n luaL_argcheck(L, level >= 0, 1, \"level must be non-negative\");\n if (lua_getstack(L, level, &ar) == 0)\n luaL_argerror(L, 1, \"invalid level\");\n lua_getinfo(L, \"f\", &ar);\n if (lua_isnil(L, -1))\n luaL_error(L, \"no function environment for tail call at level %d\",\n level);\n }\n}\n\n\nstatic int luaB_getfenv (lua_State *L) {\n getfunc(L, 1);\n if (lua_iscfunction(L, -1)) /* is a C function? */\n lua_pushvalue(L, LUA_GLOBALSINDEX); /* return the thread's global env. */\n else\n lua_getfenv(L, -1);\n return 1;\n}\n\n\nstatic int luaB_setfenv (lua_State *L) {\n luaL_checktype(L, 2, LUA_TTABLE);\n getfunc(L, 0);\n lua_pushvalue(L, 2);\n if (lua_isnumber(L, 1) && lua_tonumber(L, 1) == 0) {\n /* change environment of current thread */\n lua_pushthread(L);\n lua_insert(L, -2);\n lua_setfenv(L, -2);\n return 0;\n }\n else if (lua_iscfunction(L, -2) || lua_setfenv(L, -2) == 0)\n luaL_error(L,\n LUA_QL(\"setfenv\") \" cannot change environment of given object\");\n return 1;\n}\n\n\nstatic int luaB_rawequal (lua_State *L) {\n luaL_checkany(L, 1);\n luaL_checkany(L, 2);\n lua_pushboolean(L, lua_rawequal(L, 1, 2));\n return 1;\n}\n\n\nstatic int luaB_rawget (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n luaL_checkany(L, 2);\n lua_settop(L, 2);\n lua_rawget(L, 1);\n return 1;\n}\n\nstatic int luaB_rawset (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n luaL_checkany(L, 2);\n luaL_checkany(L, 3);\n lua_settop(L, 3);\n lua_rawset(L, 1);\n return 1;\n}\n\n\nstatic int luaB_gcinfo (lua_State *L) {\n lua_pushinteger(L, lua_getgccount(L));\n return 1;\n}\n\n\nstatic int luaB_collectgarbage (lua_State *L) {\n static const char *const opts[] = {\"stop\", \"restart\", \"collect\",\n \"count\", \"step\", \"setpause\", \"setstepmul\", NULL};\n static const int optsnum[] = {LUA_GCSTOP, LUA_GCRESTART, LUA_GCCOLLECT,\n LUA_GCCOUNT, LUA_GCSTEP, LUA_GCSETPAUSE, LUA_GCSETSTEPMUL};\n int o = luaL_checkoption(L, 1, \"collect\", opts);\n int ex = luaL_optint(L, 2, 0);\n int res = lua_gc(L, optsnum[o], ex);\n switch (optsnum[o]) {\n case LUA_GCCOUNT: {\n int b = lua_gc(L, LUA_GCCOUNTB, 0);\n lua_pushnumber(L, res + ((lua_Number)b/1024));\n return 1;\n }\n case LUA_GCSTEP: {\n lua_pushboolean(L, res);\n return 1;\n }\n default: {\n lua_pushnumber(L, res);\n return 1;\n }\n }\n}\n\n\nstatic int luaB_type (lua_State *L) {\n luaL_checkany(L, 1);\n lua_pushstring(L, luaL_typename(L, 1));\n return 1;\n}\n\n\nstatic int luaB_next (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n lua_settop(L, 2); /* create a 2nd argument if there isn't one */\n if (lua_next(L, 1))\n return 2;\n else {\n lua_pushnil(L);\n return 1;\n }\n}\n\n\nstatic int luaB_pairs (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n lua_pushvalue(L, lua_upvalueindex(1)); /* return generator, */\n lua_pushvalue(L, 1); /* state, */\n lua_pushnil(L); /* and initial value */\n return 3;\n}\n\n\nstatic int ipairsaux (lua_State *L) {\n int i = luaL_checkint(L, 2);\n luaL_checktype(L, 1, LUA_TTABLE);\n i++; /* next value */\n lua_pushinteger(L, i);\n lua_rawgeti(L, 1, i);\n return (lua_isnil(L, -1)) ? 0 : 2;\n}\n\n\nstatic int luaB_ipairs (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n lua_pushvalue(L, lua_upvalueindex(1)); /* return generator, */\n lua_pushvalue(L, 1); /* state, */\n lua_pushinteger(L, 0); /* and initial value */\n return 3;\n}\n\n\nstatic int load_aux (lua_State *L, int status) {\n if (status == 0) /* OK? */\n return 1;\n else {\n lua_pushnil(L);\n lua_insert(L, -2); /* put before error message */\n return 2; /* return nil plus error message */\n }\n}\n\n\nstatic int luaB_loadstring (lua_State *L) {\n size_t l;\n const char *s = luaL_checklstring(L, 1, &l);\n const char *chunkname = luaL_optstring(L, 2, s);\n return load_aux(L, luaL_loadbuffer(L, s, l, chunkname));\n}\n\n\nstatic int luaB_loadfile (lua_State *L) {\n const char *fname = luaL_optstring(L, 1, NULL);\n return load_aux(L, luaL_loadfile(L, fname));\n}\n\n\n/*\n** Reader for generic `load' function: `lua_load' uses the\n** stack for internal stuff, so the reader cannot change the\n** stack top. Instead, it keeps its resulting string in a\n** reserved slot inside the stack.\n*/\nstatic const char *generic_reader (lua_State *L, void *ud, size_t *size) {\n (void)ud; /* to avoid warnings */\n luaL_checkstack(L, 2, \"too many nested functions\");\n lua_pushvalue(L, 1); /* get function */\n lua_call(L, 0, 1); /* call it */\n if (lua_isnil(L, -1)) {\n *size = 0;\n return NULL;\n }\n else if (lua_isstring(L, -1)) {\n lua_replace(L, 3); /* save string in a reserved stack slot */\n return lua_tolstring(L, 3, size);\n }\n else luaL_error(L, \"reader function must return a string\");\n return NULL; /* to avoid warnings */\n}\n\n\nstatic int luaB_load (lua_State *L) {\n int status;\n const char *cname = luaL_optstring(L, 2, \"=(load)\");\n luaL_checktype(L, 1, LUA_TFUNCTION);\n lua_settop(L, 3); /* function, eventual name, plus one reserved slot */\n status = lua_load(L, generic_reader, NULL, cname);\n return load_aux(L, status);\n}\n\n\nstatic int luaB_dofile (lua_State *L) {\n const char *fname = luaL_optstring(L, 1, NULL);\n int n = lua_gettop(L);\n if (luaL_loadfile(L, fname) != 0) lua_error(L);\n lua_call(L, 0, LUA_MULTRET);\n return lua_gettop(L) - n;\n}\n\n\nstatic int luaB_assert (lua_State *L) {\n luaL_checkany(L, 1);\n if (!lua_toboolean(L, 1))\n return luaL_error(L, \"%s\", luaL_optstring(L, 2, \"assertion failed!\"));\n return lua_gettop(L);\n}\n\n\nstatic int luaB_unpack (lua_State *L) {\n int i, e, n;\n luaL_checktype(L, 1, LUA_TTABLE);\n i = luaL_optint(L, 2, 1);\n e = luaL_opt(L, luaL_checkint, 3, luaL_getn(L, 1));\n if (i > e) return 0; /* empty range */\n n = e - i + 1; /* number of elements */\n if (n <= 0 || !lua_checkstack(L, n)) /* n <= 0 means arith. overflow */\n return luaL_error(L, \"too many results to unpack\");\n lua_rawgeti(L, 1, i); /* push arg[i] (avoiding overflow problems) */\n while (i++ < e) /* push arg[i + 1...e] */\n lua_rawgeti(L, 1, i);\n return n;\n}\n\n\nstatic int luaB_select (lua_State *L) {\n int n = lua_gettop(L);\n if (lua_type(L, 1) == LUA_TSTRING && *lua_tostring(L, 1) == '#') {\n lua_pushinteger(L, n-1);\n return 1;\n }\n else {\n int i = luaL_checkint(L, 1);\n if (i < 0) i = n + i;\n else if (i > n) i = n;\n luaL_argcheck(L, 1 <= i, 1, \"index out of range\");\n return n - i;\n }\n}\n\n\nstatic int luaB_pcall (lua_State *L) {\n int status;\n luaL_checkany(L, 1);\n status = lua_pcall(L, lua_gettop(L) - 1, LUA_MULTRET, 0);\n lua_pushboolean(L, (status == 0));\n lua_insert(L, 1);\n return lua_gettop(L); /* return status + all results */\n}\n\n\nstatic int luaB_xpcall (lua_State *L) {\n int status;\n luaL_checkany(L, 2);\n lua_settop(L, 2);\n lua_insert(L, 1); /* put error function under function to be called */\n status = lua_pcall(L, 0, LUA_MULTRET, 1);\n lua_pushboolean(L, (status == 0));\n lua_replace(L, 1);\n return lua_gettop(L); /* return status + all results */\n}\n\n\nstatic int luaB_tostring (lua_State *L) {\n luaL_checkany(L, 1);\n if (luaL_callmeta(L, 1, \"__tostring\")) /* is there a metafield? */\n return 1; /* use its value */\n switch (lua_type(L, 1)) {\n case LUA_TNUMBER:\n lua_pushstring(L, lua_tostring(L, 1));\n break;\n case LUA_TSTRING:\n lua_pushvalue(L, 1);\n break;\n case LUA_TBOOLEAN:\n lua_pushstring(L, (lua_toboolean(L, 1) ? \"true\" : \"false\"));\n break;\n case LUA_TNIL:\n lua_pushliteral(L, \"nil\");\n break;\n default:\n lua_pushfstring(L, \"%s: %p\", luaL_typename(L, 1), lua_topointer(L, 1));\n break;\n }\n return 1;\n}\n\n\nstatic int luaB_newproxy (lua_State *L) {\n lua_settop(L, 1);\n lua_newuserdata(L, 0); /* create proxy */\n if (lua_toboolean(L, 1) == 0)\n return 1; /* no metatable */\n else if (lua_isboolean(L, 1)) {\n lua_newtable(L); /* create a new metatable `m' ... */\n lua_pushvalue(L, -1); /* ... and mark `m' as a valid metatable */\n lua_pushboolean(L, 1);\n lua_rawset(L, lua_upvalueindex(1)); /* weaktable[m] = true */\n }\n else {\n int validproxy = 0; /* to check if weaktable[metatable(u)] == true */\n if (lua_getmetatable(L, 1)) {\n lua_rawget(L, lua_upvalueindex(1));\n validproxy = lua_toboolean(L, -1);\n lua_pop(L, 1); /* remove value */\n }\n luaL_argcheck(L, validproxy, 1, \"boolean or proxy expected\");\n lua_getmetatable(L, 1); /* metatable is valid; get it */\n }\n lua_setmetatable(L, 2);\n return 1;\n}\n\n\nstatic const luaL_Reg base_funcs[] = {\n {\"assert\", luaB_assert},\n {\"collectgarbage\", luaB_collectgarbage},\n {\"dofile\", luaB_dofile},\n {\"error\", luaB_error},\n {\"gcinfo\", luaB_gcinfo},\n {\"getfenv\", luaB_getfenv},\n {\"getmetatable\", luaB_getmetatable},\n {\"loadfile\", luaB_loadfile},\n {\"load\", luaB_load},\n {\"loadstring\", luaB_loadstring},\n {\"next\", luaB_next},\n {\"pcall\", luaB_pcall},\n {\"print\", luaB_print},\n {\"rawequal\", luaB_rawequal},\n {\"rawget\", luaB_rawget},\n {\"rawset\", luaB_rawset},\n {\"select\", luaB_select},\n {\"setfenv\", luaB_setfenv},\n {\"setmetatable\", luaB_setmetatable},\n {\"tonumber\", luaB_tonumber},\n {\"tostring\", luaB_tostring},\n {\"type\", luaB_type},\n {\"unpack\", luaB_unpack},\n {\"xpcall\", luaB_xpcall},\n {NULL, NULL}\n};\n\n\n/*\n** {======================================================\n** Coroutine library\n** =======================================================\n*/\n\n#define CO_RUN\t0\t/* running */\n#define CO_SUS\t1\t/* suspended */\n#define CO_NOR\t2\t/* 'normal' (it resumed another coroutine) */\n#define CO_DEAD\t3\n\nstatic const char *const statnames[] =\n {\"running\", \"suspended\", \"normal\", \"dead\"};\n\nstatic int costatus (lua_State *L, lua_State *co) {\n if (L == co) return CO_RUN;\n switch (lua_status(co)) {\n case LUA_YIELD:\n return CO_SUS;\n case 0: {\n lua_Debug ar;\n if (lua_getstack(co, 0, &ar) > 0) /* does it have frames? */\n return CO_NOR; /* it is running */\n else if (lua_gettop(co) == 0)\n return CO_DEAD;\n else\n return CO_SUS; /* initial state */\n }\n default: /* some error occured */\n return CO_DEAD;\n }\n}\n\n\nstatic int luaB_costatus (lua_State *L) {\n lua_State *co = lua_tothread(L, 1);\n luaL_argcheck(L, co, 1, \"coroutine expected\");\n lua_pushstring(L, statnames[costatus(L, co)]);\n return 1;\n}\n\n\nstatic int auxresume (lua_State *L, lua_State *co, int narg) {\n int status = costatus(L, co);\n if (!lua_checkstack(co, narg))\n luaL_error(L, \"too many arguments to resume\");\n if (status != CO_SUS) {\n lua_pushfstring(L, \"cannot resume %s coroutine\", statnames[status]);\n return -1; /* error flag */\n }\n lua_xmove(L, co, narg);\n lua_setlevel(L, co);\n status = lua_resume(co, narg);\n if (status == 0 || status == LUA_YIELD) {\n int nres = lua_gettop(co);\n if (!lua_checkstack(L, nres + 1))\n luaL_error(L, \"too many results to resume\");\n lua_xmove(co, L, nres); /* move yielded values */\n return nres;\n }\n else {\n lua_xmove(co, L, 1); /* move error message */\n return -1; /* error flag */\n }\n}\n\n\nstatic int luaB_coresume (lua_State *L) {\n lua_State *co = lua_tothread(L, 1);\n int r;\n luaL_argcheck(L, co, 1, \"coroutine expected\");\n r = auxresume(L, co, lua_gettop(L) - 1);\n if (r < 0) {\n lua_pushboolean(L, 0);\n lua_insert(L, -2);\n return 2; /* return false + error message */\n }\n else {\n lua_pushboolean(L, 1);\n lua_insert(L, -(r + 1));\n return r + 1; /* return true + `resume' returns */\n }\n}\n\n\nstatic int luaB_auxwrap (lua_State *L) {\n lua_State *co = lua_tothread(L, lua_upvalueindex(1));\n int r = auxresume(L, co, lua_gettop(L));\n if (r < 0) {\n if (lua_isstring(L, -1)) { /* error object is a string? */\n luaL_where(L, 1); /* add extra info */\n lua_insert(L, -2);\n lua_concat(L, 2);\n }\n lua_error(L); /* propagate error */\n }\n return r;\n}\n\n\nstatic int luaB_cocreate (lua_State *L) {\n lua_State *NL = lua_newthread(L);\n luaL_argcheck(L, lua_isfunction(L, 1) && !lua_iscfunction(L, 1), 1,\n \"Lua function expected\");\n lua_pushvalue(L, 1); /* move function to top */\n lua_xmove(L, NL, 1); /* move function from L to NL */\n return 1;\n}\n\n\nstatic int luaB_cowrap (lua_State *L) {\n luaB_cocreate(L);\n lua_pushcclosure(L, luaB_auxwrap, 1);\n return 1;\n}\n\n\nstatic int luaB_yield (lua_State *L) {\n return lua_yield(L, lua_gettop(L));\n}\n\n\nstatic int luaB_corunning (lua_State *L) {\n if (lua_pushthread(L))\n lua_pushnil(L); /* main thread is not a coroutine */\n return 1;\n}\n\n\nstatic const luaL_Reg co_funcs[] = {\n {\"create\", luaB_cocreate},\n {\"resume\", luaB_coresume},\n {\"running\", luaB_corunning},\n {\"status\", luaB_costatus},\n {\"wrap\", luaB_cowrap},\n {\"yield\", luaB_yield},\n {NULL, NULL}\n};\n\n/* }====================================================== */\n\n\nstatic void auxopen (lua_State *L, const char *name,\n lua_CFunction f, lua_CFunction u) {\n lua_pushcfunction(L, u);\n lua_pushcclosure(L, f, 1);\n lua_setfield(L, -2, name);\n}\n\n\nstatic void base_open (lua_State *L) {\n /* set global _G */\n lua_pushvalue(L, LUA_GLOBALSINDEX);\n lua_setglobal(L, \"_G\");\n /* open lib into global table */\n luaL_register(L, \"_G\", base_funcs);\n lua_pushliteral(L, LUA_VERSION);\n lua_setglobal(L, \"_VERSION\"); /* set global _VERSION */\n /* `ipairs' and `pairs' need auxiliary functions as upvalues */\n auxopen(L, \"ipairs\", luaB_ipairs, ipairsaux);\n auxopen(L, \"pairs\", luaB_pairs, luaB_next);\n /* `newproxy' needs a weaktable as upvalue */\n lua_createtable(L, 0, 1); /* new table `w' */\n lua_pushvalue(L, -1); /* `w' will be its own metatable */\n lua_setmetatable(L, -2);\n lua_pushliteral(L, \"kv\");\n lua_setfield(L, -2, \"__mode\"); /* metatable(w).__mode = \"kv\" */\n lua_pushcclosure(L, luaB_newproxy, 1);\n lua_setglobal(L, \"newproxy\"); /* set global `newproxy' */\n}\n\n\nLUALIB_API int luaopen_base (lua_State *L) {\n base_open(L);\n luaL_register(L, LUA_COLIBNAME, co_funcs);\n return 2;\n}\n\n" }, { "path": "deps/lua/src/lcode.c", "content": "/*\n** $Id: lcode.c,v 2.25.1.5 2011/01/31 14:53:16 roberto Exp $\n** Code generator for Lua\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lcode_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lcode.h\"\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lgc.h\"\n#include \"llex.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lparser.h\"\n#include \"ltable.h\"\n\n\n#define hasjumps(e)\t((e)->t != (e)->f)\n\n\nstatic int isnumeral(expdesc *e) {\n return (e->k == VKNUM && e->t == NO_JUMP && e->f == NO_JUMP);\n}\n\n\nvoid luaK_nil (FuncState *fs, int from, int n) {\n Instruction *previous;\n if (fs->pc > fs->lasttarget) { /* no jumps to current position? */\n if (fs->pc == 0) { /* function start? */\n if (from >= fs->nactvar)\n return; /* positions are already clean */\n }\n else {\n previous = &fs->f->code[fs->pc-1];\n if (GET_OPCODE(*previous) == OP_LOADNIL) {\n int pfrom = GETARG_A(*previous);\n int pto = GETARG_B(*previous);\n if (pfrom <= from && from <= pto+1) { /* can connect both? */\n if (from+n-1 > pto)\n SETARG_B(*previous, from+n-1);\n return;\n }\n }\n }\n }\n luaK_codeABC(fs, OP_LOADNIL, from, from+n-1, 0); /* else no optimization */\n}\n\n\nint luaK_jump (FuncState *fs) {\n int jpc = fs->jpc; /* save list of jumps to here */\n int j;\n fs->jpc = NO_JUMP;\n j = luaK_codeAsBx(fs, OP_JMP, 0, NO_JUMP);\n luaK_concat(fs, &j, jpc); /* keep them on hold */\n return j;\n}\n\n\nvoid luaK_ret (FuncState *fs, int first, int nret) {\n luaK_codeABC(fs, OP_RETURN, first, nret+1, 0);\n}\n\n\nstatic int condjump (FuncState *fs, OpCode op, int A, int B, int C) {\n luaK_codeABC(fs, op, A, B, C);\n return luaK_jump(fs);\n}\n\n\nstatic void fixjump (FuncState *fs, int pc, int dest) {\n Instruction *jmp = &fs->f->code[pc];\n int offset = dest-(pc+1);\n lua_assert(dest != NO_JUMP);\n if (abs(offset) > MAXARG_sBx)\n luaX_syntaxerror(fs->ls, \"control structure too long\");\n SETARG_sBx(*jmp, offset);\n}\n\n\n/*\n** returns current `pc' and marks it as a jump target (to avoid wrong\n** optimizations with consecutive instructions not in the same basic block).\n*/\nint luaK_getlabel (FuncState *fs) {\n fs->lasttarget = fs->pc;\n return fs->pc;\n}\n\n\nstatic int getjump (FuncState *fs, int pc) {\n int offset = GETARG_sBx(fs->f->code[pc]);\n if (offset == NO_JUMP) /* point to itself represents end of list */\n return NO_JUMP; /* end of list */\n else\n return (pc+1)+offset; /* turn offset into absolute position */\n}\n\n\nstatic Instruction *getjumpcontrol (FuncState *fs, int pc) {\n Instruction *pi = &fs->f->code[pc];\n if (pc >= 1 && testTMode(GET_OPCODE(*(pi-1))))\n return pi-1;\n else\n return pi;\n}\n\n\n/*\n** check whether list has any jump that do not produce a value\n** (or produce an inverted value)\n*/\nstatic int need_value (FuncState *fs, int list) {\n for (; list != NO_JUMP; list = getjump(fs, list)) {\n Instruction i = *getjumpcontrol(fs, list);\n if (GET_OPCODE(i) != OP_TESTSET) return 1;\n }\n return 0; /* not found */\n}\n\n\nstatic int patchtestreg (FuncState *fs, int node, int reg) {\n Instruction *i = getjumpcontrol(fs, node);\n if (GET_OPCODE(*i) != OP_TESTSET)\n return 0; /* cannot patch other instructions */\n if (reg != NO_REG && reg != GETARG_B(*i))\n SETARG_A(*i, reg);\n else /* no register to put value or register already has the value */\n *i = CREATE_ABC(OP_TEST, GETARG_B(*i), 0, GETARG_C(*i));\n\n return 1;\n}\n\n\nstatic void removevalues (FuncState *fs, int list) {\n for (; list != NO_JUMP; list = getjump(fs, list))\n patchtestreg(fs, list, NO_REG);\n}\n\n\nstatic void patchlistaux (FuncState *fs, int list, int vtarget, int reg,\n int dtarget) {\n while (list != NO_JUMP) {\n int next = getjump(fs, list);\n if (patchtestreg(fs, list, reg))\n fixjump(fs, list, vtarget);\n else\n fixjump(fs, list, dtarget); /* jump to default target */\n list = next;\n }\n}\n\n\nstatic void dischargejpc (FuncState *fs) {\n patchlistaux(fs, fs->jpc, fs->pc, NO_REG, fs->pc);\n fs->jpc = NO_JUMP;\n}\n\n\nvoid luaK_patchlist (FuncState *fs, int list, int target) {\n if (target == fs->pc)\n luaK_patchtohere(fs, list);\n else {\n lua_assert(target < fs->pc);\n patchlistaux(fs, list, target, NO_REG, target);\n }\n}\n\n\nvoid luaK_patchtohere (FuncState *fs, int list) {\n luaK_getlabel(fs);\n luaK_concat(fs, &fs->jpc, list);\n}\n\n\nvoid luaK_concat (FuncState *fs, int *l1, int l2) {\n if (l2 == NO_JUMP) return;\n else if (*l1 == NO_JUMP)\n *l1 = l2;\n else {\n int list = *l1;\n int next;\n while ((next = getjump(fs, list)) != NO_JUMP) /* find last element */\n list = next;\n fixjump(fs, list, l2);\n }\n}\n\n\nvoid luaK_checkstack (FuncState *fs, int n) {\n int newstack = fs->freereg + n;\n if (newstack > fs->f->maxstacksize) {\n if (newstack >= MAXSTACK)\n luaX_syntaxerror(fs->ls, \"function or expression too complex\");\n fs->f->maxstacksize = cast_byte(newstack);\n }\n}\n\n\nvoid luaK_reserveregs (FuncState *fs, int n) {\n luaK_checkstack(fs, n);\n fs->freereg += n;\n}\n\n\nstatic void freereg (FuncState *fs, int reg) {\n if (!ISK(reg) && reg >= fs->nactvar) {\n fs->freereg--;\n lua_assert(reg == fs->freereg);\n }\n}\n\n\nstatic void freeexp (FuncState *fs, expdesc *e) {\n if (e->k == VNONRELOC)\n freereg(fs, e->u.s.info);\n}\n\n\nstatic int addk (FuncState *fs, TValue *k, TValue *v) {\n lua_State *L = fs->L;\n TValue *idx = luaH_set(L, fs->h, k);\n Proto *f = fs->f;\n int oldsize = f->sizek;\n if (ttisnumber(idx)) {\n lua_assert(luaO_rawequalObj(&fs->f->k[cast_int(nvalue(idx))], v));\n return cast_int(nvalue(idx));\n }\n else { /* constant not found; create a new entry */\n setnvalue(idx, cast_num(fs->nk));\n luaM_growvector(L, f->k, fs->nk, f->sizek, TValue,\n MAXARG_Bx, \"constant table overflow\");\n while (oldsize < f->sizek) setnilvalue(&f->k[oldsize++]);\n setobj(L, &f->k[fs->nk], v);\n luaC_barrier(L, f, v);\n return fs->nk++;\n }\n}\n\n\nint luaK_stringK (FuncState *fs, TString *s) {\n TValue o;\n setsvalue(fs->L, &o, s);\n return addk(fs, &o, &o);\n}\n\n\nint luaK_numberK (FuncState *fs, lua_Number r) {\n TValue o;\n setnvalue(&o, r);\n return addk(fs, &o, &o);\n}\n\n\nstatic int boolK (FuncState *fs, int b) {\n TValue o;\n setbvalue(&o, b);\n return addk(fs, &o, &o);\n}\n\n\nstatic int nilK (FuncState *fs) {\n TValue k, v;\n setnilvalue(&v);\n /* cannot use nil as key; instead use table itself to represent nil */\n sethvalue(fs->L, &k, fs->h);\n return addk(fs, &k, &v);\n}\n\n\nvoid luaK_setreturns (FuncState *fs, expdesc *e, int nresults) {\n if (e->k == VCALL) { /* expression is an open function call? */\n SETARG_C(getcode(fs, e), nresults+1);\n }\n else if (e->k == VVARARG) {\n SETARG_B(getcode(fs, e), nresults+1);\n SETARG_A(getcode(fs, e), fs->freereg);\n luaK_reserveregs(fs, 1);\n }\n}\n\n\nvoid luaK_setoneret (FuncState *fs, expdesc *e) {\n if (e->k == VCALL) { /* expression is an open function call? */\n e->k = VNONRELOC;\n e->u.s.info = GETARG_A(getcode(fs, e));\n }\n else if (e->k == VVARARG) {\n SETARG_B(getcode(fs, e), 2);\n e->k = VRELOCABLE; /* can relocate its simple result */\n }\n}\n\n\nvoid luaK_dischargevars (FuncState *fs, expdesc *e) {\n switch (e->k) {\n case VLOCAL: {\n e->k = VNONRELOC;\n break;\n }\n case VUPVAL: {\n e->u.s.info = luaK_codeABC(fs, OP_GETUPVAL, 0, e->u.s.info, 0);\n e->k = VRELOCABLE;\n break;\n }\n case VGLOBAL: {\n e->u.s.info = luaK_codeABx(fs, OP_GETGLOBAL, 0, e->u.s.info);\n e->k = VRELOCABLE;\n break;\n }\n case VINDEXED: {\n freereg(fs, e->u.s.aux);\n freereg(fs, e->u.s.info);\n e->u.s.info = luaK_codeABC(fs, OP_GETTABLE, 0, e->u.s.info, e->u.s.aux);\n e->k = VRELOCABLE;\n break;\n }\n case VVARARG:\n case VCALL: {\n luaK_setoneret(fs, e);\n break;\n }\n default: break; /* there is one value available (somewhere) */\n }\n}\n\n\nstatic int code_label (FuncState *fs, int A, int b, int jump) {\n luaK_getlabel(fs); /* those instructions may be jump targets */\n return luaK_codeABC(fs, OP_LOADBOOL, A, b, jump);\n}\n\n\nstatic void discharge2reg (FuncState *fs, expdesc *e, int reg) {\n luaK_dischargevars(fs, e);\n switch (e->k) {\n case VNIL: {\n luaK_nil(fs, reg, 1);\n break;\n }\n case VFALSE: case VTRUE: {\n luaK_codeABC(fs, OP_LOADBOOL, reg, e->k == VTRUE, 0);\n break;\n }\n case VK: {\n luaK_codeABx(fs, OP_LOADK, reg, e->u.s.info);\n break;\n }\n case VKNUM: {\n luaK_codeABx(fs, OP_LOADK, reg, luaK_numberK(fs, e->u.nval));\n break;\n }\n case VRELOCABLE: {\n Instruction *pc = &getcode(fs, e);\n SETARG_A(*pc, reg);\n break;\n }\n case VNONRELOC: {\n if (reg != e->u.s.info)\n luaK_codeABC(fs, OP_MOVE, reg, e->u.s.info, 0);\n break;\n }\n default: {\n lua_assert(e->k == VVOID || e->k == VJMP);\n return; /* nothing to do... */\n }\n }\n e->u.s.info = reg;\n e->k = VNONRELOC;\n}\n\n\nstatic void discharge2anyreg (FuncState *fs, expdesc *e) {\n if (e->k != VNONRELOC) {\n luaK_reserveregs(fs, 1);\n discharge2reg(fs, e, fs->freereg-1);\n }\n}\n\n\nstatic void exp2reg (FuncState *fs, expdesc *e, int reg) {\n discharge2reg(fs, e, reg);\n if (e->k == VJMP)\n luaK_concat(fs, &e->t, e->u.s.info); /* put this jump in `t' list */\n if (hasjumps(e)) {\n int final; /* position after whole expression */\n int p_f = NO_JUMP; /* position of an eventual LOAD false */\n int p_t = NO_JUMP; /* position of an eventual LOAD true */\n if (need_value(fs, e->t) || need_value(fs, e->f)) {\n int fj = (e->k == VJMP) ? NO_JUMP : luaK_jump(fs);\n p_f = code_label(fs, reg, 0, 1);\n p_t = code_label(fs, reg, 1, 0);\n luaK_patchtohere(fs, fj);\n }\n final = luaK_getlabel(fs);\n patchlistaux(fs, e->f, final, reg, p_f);\n patchlistaux(fs, e->t, final, reg, p_t);\n }\n e->f = e->t = NO_JUMP;\n e->u.s.info = reg;\n e->k = VNONRELOC;\n}\n\n\nvoid luaK_exp2nextreg (FuncState *fs, expdesc *e) {\n luaK_dischargevars(fs, e);\n freeexp(fs, e);\n luaK_reserveregs(fs, 1);\n exp2reg(fs, e, fs->freereg - 1);\n}\n\n\nint luaK_exp2anyreg (FuncState *fs, expdesc *e) {\n luaK_dischargevars(fs, e);\n if (e->k == VNONRELOC) {\n if (!hasjumps(e)) return e->u.s.info; /* exp is already in a register */\n if (e->u.s.info >= fs->nactvar) { /* reg. is not a local? */\n exp2reg(fs, e, e->u.s.info); /* put value on it */\n return e->u.s.info;\n }\n }\n luaK_exp2nextreg(fs, e); /* default */\n return e->u.s.info;\n}\n\n\nvoid luaK_exp2val (FuncState *fs, expdesc *e) {\n if (hasjumps(e))\n luaK_exp2anyreg(fs, e);\n else\n luaK_dischargevars(fs, e);\n}\n\n\nint luaK_exp2RK (FuncState *fs, expdesc *e) {\n luaK_exp2val(fs, e);\n switch (e->k) {\n case VKNUM:\n case VTRUE:\n case VFALSE:\n case VNIL: {\n if (fs->nk <= MAXINDEXRK) { /* constant fit in RK operand? */\n e->u.s.info = (e->k == VNIL) ? nilK(fs) :\n (e->k == VKNUM) ? luaK_numberK(fs, e->u.nval) :\n boolK(fs, (e->k == VTRUE));\n e->k = VK;\n return RKASK(e->u.s.info);\n }\n else break;\n }\n case VK: {\n if (e->u.s.info <= MAXINDEXRK) /* constant fit in argC? */\n return RKASK(e->u.s.info);\n else break;\n }\n default: break;\n }\n /* not a constant in the right range: put it in a register */\n return luaK_exp2anyreg(fs, e);\n}\n\n\nvoid luaK_storevar (FuncState *fs, expdesc *var, expdesc *ex) {\n switch (var->k) {\n case VLOCAL: {\n freeexp(fs, ex);\n exp2reg(fs, ex, var->u.s.info);\n return;\n }\n case VUPVAL: {\n int e = luaK_exp2anyreg(fs, ex);\n luaK_codeABC(fs, OP_SETUPVAL, e, var->u.s.info, 0);\n break;\n }\n case VGLOBAL: {\n int e = luaK_exp2anyreg(fs, ex);\n luaK_codeABx(fs, OP_SETGLOBAL, e, var->u.s.info);\n break;\n }\n case VINDEXED: {\n int e = luaK_exp2RK(fs, ex);\n luaK_codeABC(fs, OP_SETTABLE, var->u.s.info, var->u.s.aux, e);\n break;\n }\n default: {\n lua_assert(0); /* invalid var kind to store */\n break;\n }\n }\n freeexp(fs, ex);\n}\n\n\nvoid luaK_self (FuncState *fs, expdesc *e, expdesc *key) {\n int func;\n luaK_exp2anyreg(fs, e);\n freeexp(fs, e);\n func = fs->freereg;\n luaK_reserveregs(fs, 2);\n luaK_codeABC(fs, OP_SELF, func, e->u.s.info, luaK_exp2RK(fs, key));\n freeexp(fs, key);\n e->u.s.info = func;\n e->k = VNONRELOC;\n}\n\n\nstatic void invertjump (FuncState *fs, expdesc *e) {\n Instruction *pc = getjumpcontrol(fs, e->u.s.info);\n lua_assert(testTMode(GET_OPCODE(*pc)) && GET_OPCODE(*pc) != OP_TESTSET &&\n GET_OPCODE(*pc) != OP_TEST);\n SETARG_A(*pc, !(GETARG_A(*pc)));\n}\n\n\nstatic int jumponcond (FuncState *fs, expdesc *e, int cond) {\n if (e->k == VRELOCABLE) {\n Instruction ie = getcode(fs, e);\n if (GET_OPCODE(ie) == OP_NOT) {\n fs->pc--; /* remove previous OP_NOT */\n return condjump(fs, OP_TEST, GETARG_B(ie), 0, !cond);\n }\n /* else go through */\n }\n discharge2anyreg(fs, e);\n freeexp(fs, e);\n return condjump(fs, OP_TESTSET, NO_REG, e->u.s.info, cond);\n}\n\n\nvoid luaK_goiftrue (FuncState *fs, expdesc *e) {\n int pc; /* pc of last jump */\n luaK_dischargevars(fs, e);\n switch (e->k) {\n case VK: case VKNUM: case VTRUE: {\n pc = NO_JUMP; /* always true; do nothing */\n break;\n }\n case VJMP: {\n invertjump(fs, e);\n pc = e->u.s.info;\n break;\n }\n default: {\n pc = jumponcond(fs, e, 0);\n break;\n }\n }\n luaK_concat(fs, &e->f, pc); /* insert last jump in `f' list */\n luaK_patchtohere(fs, e->t);\n e->t = NO_JUMP;\n}\n\n\nstatic void luaK_goiffalse (FuncState *fs, expdesc *e) {\n int pc; /* pc of last jump */\n luaK_dischargevars(fs, e);\n switch (e->k) {\n case VNIL: case VFALSE: {\n pc = NO_JUMP; /* always false; do nothing */\n break;\n }\n case VJMP: {\n pc = e->u.s.info;\n break;\n }\n default: {\n pc = jumponcond(fs, e, 1);\n break;\n }\n }\n luaK_concat(fs, &e->t, pc); /* insert last jump in `t' list */\n luaK_patchtohere(fs, e->f);\n e->f = NO_JUMP;\n}\n\n\nstatic void codenot (FuncState *fs, expdesc *e) {\n luaK_dischargevars(fs, e);\n switch (e->k) {\n case VNIL: case VFALSE: {\n e->k = VTRUE;\n break;\n }\n case VK: case VKNUM: case VTRUE: {\n e->k = VFALSE;\n break;\n }\n case VJMP: {\n invertjump(fs, e);\n break;\n }\n case VRELOCABLE:\n case VNONRELOC: {\n discharge2anyreg(fs, e);\n freeexp(fs, e);\n e->u.s.info = luaK_codeABC(fs, OP_NOT, 0, e->u.s.info, 0);\n e->k = VRELOCABLE;\n break;\n }\n default: {\n lua_assert(0); /* cannot happen */\n break;\n }\n }\n /* interchange true and false lists */\n { int temp = e->f; e->f = e->t; e->t = temp; }\n removevalues(fs, e->f);\n removevalues(fs, e->t);\n}\n\n\nvoid luaK_indexed (FuncState *fs, expdesc *t, expdesc *k) {\n t->u.s.aux = luaK_exp2RK(fs, k);\n t->k = VINDEXED;\n}\n\n\nstatic int constfolding (OpCode op, expdesc *e1, expdesc *e2) {\n lua_Number v1, v2, r;\n if (!isnumeral(e1) || !isnumeral(e2)) return 0;\n v1 = e1->u.nval;\n v2 = e2->u.nval;\n switch (op) {\n case OP_ADD: r = luai_numadd(v1, v2); break;\n case OP_SUB: r = luai_numsub(v1, v2); break;\n case OP_MUL: r = luai_nummul(v1, v2); break;\n case OP_DIV:\n if (v2 == 0) return 0; /* do not attempt to divide by 0 */\n r = luai_numdiv(v1, v2); break;\n case OP_MOD:\n if (v2 == 0) return 0; /* do not attempt to divide by 0 */\n r = luai_nummod(v1, v2); break;\n case OP_POW: r = luai_numpow(v1, v2); break;\n case OP_UNM: r = luai_numunm(v1); break;\n case OP_LEN: return 0; /* no constant folding for 'len' */\n default: lua_assert(0); r = 0; break;\n }\n if (luai_numisnan(r)) return 0; /* do not attempt to produce NaN */\n e1->u.nval = r;\n return 1;\n}\n\n\nstatic void codearith (FuncState *fs, OpCode op, expdesc *e1, expdesc *e2) {\n if (constfolding(op, e1, e2))\n return;\n else {\n int o2 = (op != OP_UNM && op != OP_LEN) ? luaK_exp2RK(fs, e2) : 0;\n int o1 = luaK_exp2RK(fs, e1);\n if (o1 > o2) {\n freeexp(fs, e1);\n freeexp(fs, e2);\n }\n else {\n freeexp(fs, e2);\n freeexp(fs, e1);\n }\n e1->u.s.info = luaK_codeABC(fs, op, 0, o1, o2);\n e1->k = VRELOCABLE;\n }\n}\n\n\nstatic void codecomp (FuncState *fs, OpCode op, int cond, expdesc *e1,\n expdesc *e2) {\n int o1 = luaK_exp2RK(fs, e1);\n int o2 = luaK_exp2RK(fs, e2);\n freeexp(fs, e2);\n freeexp(fs, e1);\n if (cond == 0 && op != OP_EQ) {\n int temp; /* exchange args to replace by `<' or `<=' */\n temp = o1; o1 = o2; o2 = temp; /* o1 <==> o2 */\n cond = 1;\n }\n e1->u.s.info = condjump(fs, op, cond, o1, o2);\n e1->k = VJMP;\n}\n\n\nvoid luaK_prefix (FuncState *fs, UnOpr op, expdesc *e) {\n expdesc e2;\n e2.t = e2.f = NO_JUMP; e2.k = VKNUM; e2.u.nval = 0;\n switch (op) {\n case OPR_MINUS: {\n if (!isnumeral(e))\n luaK_exp2anyreg(fs, e); /* cannot operate on non-numeric constants */\n codearith(fs, OP_UNM, e, &e2);\n break;\n }\n case OPR_NOT: codenot(fs, e); break;\n case OPR_LEN: {\n luaK_exp2anyreg(fs, e); /* cannot operate on constants */\n codearith(fs, OP_LEN, e, &e2);\n break;\n }\n default: lua_assert(0);\n }\n}\n\n\nvoid luaK_infix (FuncState *fs, BinOpr op, expdesc *v) {\n switch (op) {\n case OPR_AND: {\n luaK_goiftrue(fs, v);\n break;\n }\n case OPR_OR: {\n luaK_goiffalse(fs, v);\n break;\n }\n case OPR_CONCAT: {\n luaK_exp2nextreg(fs, v); /* operand must be on the `stack' */\n break;\n }\n case OPR_ADD: case OPR_SUB: case OPR_MUL: case OPR_DIV:\n case OPR_MOD: case OPR_POW: {\n if (!isnumeral(v)) luaK_exp2RK(fs, v);\n break;\n }\n default: {\n luaK_exp2RK(fs, v);\n break;\n }\n }\n}\n\n\nvoid luaK_posfix (FuncState *fs, BinOpr op, expdesc *e1, expdesc *e2) {\n switch (op) {\n case OPR_AND: {\n lua_assert(e1->t == NO_JUMP); /* list must be closed */\n luaK_dischargevars(fs, e2);\n luaK_concat(fs, &e2->f, e1->f);\n *e1 = *e2;\n break;\n }\n case OPR_OR: {\n lua_assert(e1->f == NO_JUMP); /* list must be closed */\n luaK_dischargevars(fs, e2);\n luaK_concat(fs, &e2->t, e1->t);\n *e1 = *e2;\n break;\n }\n case OPR_CONCAT: {\n luaK_exp2val(fs, e2);\n if (e2->k == VRELOCABLE && GET_OPCODE(getcode(fs, e2)) == OP_CONCAT) {\n lua_assert(e1->u.s.info == GETARG_B(getcode(fs, e2))-1);\n freeexp(fs, e1);\n SETARG_B(getcode(fs, e2), e1->u.s.info);\n e1->k = VRELOCABLE; e1->u.s.info = e2->u.s.info;\n }\n else {\n luaK_exp2nextreg(fs, e2); /* operand must be on the 'stack' */\n codearith(fs, OP_CONCAT, e1, e2);\n }\n break;\n }\n case OPR_ADD: codearith(fs, OP_ADD, e1, e2); break;\n case OPR_SUB: codearith(fs, OP_SUB, e1, e2); break;\n case OPR_MUL: codearith(fs, OP_MUL, e1, e2); break;\n case OPR_DIV: codearith(fs, OP_DIV, e1, e2); break;\n case OPR_MOD: codearith(fs, OP_MOD, e1, e2); break;\n case OPR_POW: codearith(fs, OP_POW, e1, e2); break;\n case OPR_EQ: codecomp(fs, OP_EQ, 1, e1, e2); break;\n case OPR_NE: codecomp(fs, OP_EQ, 0, e1, e2); break;\n case OPR_LT: codecomp(fs, OP_LT, 1, e1, e2); break;\n case OPR_LE: codecomp(fs, OP_LE, 1, e1, e2); break;\n case OPR_GT: codecomp(fs, OP_LT, 0, e1, e2); break;\n case OPR_GE: codecomp(fs, OP_LE, 0, e1, e2); break;\n default: lua_assert(0);\n }\n}\n\n\nvoid luaK_fixline (FuncState *fs, int line) {\n fs->f->lineinfo[fs->pc - 1] = line;\n}\n\n\nstatic int luaK_code (FuncState *fs, Instruction i, int line) {\n Proto *f = fs->f;\n dischargejpc(fs); /* `pc' will change */\n /* put new instruction in code array */\n luaM_growvector(fs->L, f->code, fs->pc, f->sizecode, Instruction,\n MAX_INT, \"code size overflow\");\n f->code[fs->pc] = i;\n /* save corresponding line information */\n luaM_growvector(fs->L, f->lineinfo, fs->pc, f->sizelineinfo, int,\n MAX_INT, \"code size overflow\");\n f->lineinfo[fs->pc] = line;\n return fs->pc++;\n}\n\n\nint luaK_codeABC (FuncState *fs, OpCode o, int a, int b, int c) {\n lua_assert(getOpMode(o) == iABC);\n lua_assert(getBMode(o) != OpArgN || b == 0);\n lua_assert(getCMode(o) != OpArgN || c == 0);\n return luaK_code(fs, CREATE_ABC(o, a, b, c), fs->ls->lastline);\n}\n\n\nint luaK_codeABx (FuncState *fs, OpCode o, int a, unsigned int bc) {\n lua_assert(getOpMode(o) == iABx || getOpMode(o) == iAsBx);\n lua_assert(getCMode(o) == OpArgN);\n return luaK_code(fs, CREATE_ABx(o, a, bc), fs->ls->lastline);\n}\n\n\nvoid luaK_setlist (FuncState *fs, int base, int nelems, int tostore) {\n int c = (nelems - 1)/LFIELDS_PER_FLUSH + 1;\n int b = (tostore == LUA_MULTRET) ? 0 : tostore;\n lua_assert(tostore != 0);\n if (c <= MAXARG_C)\n luaK_codeABC(fs, OP_SETLIST, base, b, c);\n else {\n luaK_codeABC(fs, OP_SETLIST, base, b, 0);\n luaK_code(fs, cast(Instruction, c), fs->ls->lastline);\n }\n fs->freereg = base + 1; /* free registers with list values */\n}\n\n" }, { "path": "deps/lua/src/lcode.h", "content": "/*\n** $Id: lcode.h,v 1.48.1.1 2007/12/27 13:02:25 roberto Exp $\n** Code generator for Lua\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lcode_h\n#define lcode_h\n\n#include \"llex.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lparser.h\"\n\n\n/*\n** Marks the end of a patch list. It is an invalid value both as an absolute\n** address, and as a list link (would link an element to itself).\n*/\n#define NO_JUMP (-1)\n\n\n/*\n** grep \"ORDER OPR\" if you change these enums\n*/\ntypedef enum BinOpr {\n OPR_ADD, OPR_SUB, OPR_MUL, OPR_DIV, OPR_MOD, OPR_POW,\n OPR_CONCAT,\n OPR_NE, OPR_EQ,\n OPR_LT, OPR_LE, OPR_GT, OPR_GE,\n OPR_AND, OPR_OR,\n OPR_NOBINOPR\n} BinOpr;\n\n\ntypedef enum UnOpr { OPR_MINUS, OPR_NOT, OPR_LEN, OPR_NOUNOPR } UnOpr;\n\n\n#define getcode(fs,e)\t((fs)->f->code[(e)->u.s.info])\n\n#define luaK_codeAsBx(fs,o,A,sBx)\tluaK_codeABx(fs,o,A,(sBx)+MAXARG_sBx)\n\n#define luaK_setmultret(fs,e)\tluaK_setreturns(fs, e, LUA_MULTRET)\n\nLUAI_FUNC int luaK_codeABx (FuncState *fs, OpCode o, int A, unsigned int Bx);\nLUAI_FUNC int luaK_codeABC (FuncState *fs, OpCode o, int A, int B, int C);\nLUAI_FUNC void luaK_fixline (FuncState *fs, int line);\nLUAI_FUNC void luaK_nil (FuncState *fs, int from, int n);\nLUAI_FUNC void luaK_reserveregs (FuncState *fs, int n);\nLUAI_FUNC void luaK_checkstack (FuncState *fs, int n);\nLUAI_FUNC int luaK_stringK (FuncState *fs, TString *s);\nLUAI_FUNC int luaK_numberK (FuncState *fs, lua_Number r);\nLUAI_FUNC void luaK_dischargevars (FuncState *fs, expdesc *e);\nLUAI_FUNC int luaK_exp2anyreg (FuncState *fs, expdesc *e);\nLUAI_FUNC void luaK_exp2nextreg (FuncState *fs, expdesc *e);\nLUAI_FUNC void luaK_exp2val (FuncState *fs, expdesc *e);\nLUAI_FUNC int luaK_exp2RK (FuncState *fs, expdesc *e);\nLUAI_FUNC void luaK_self (FuncState *fs, expdesc *e, expdesc *key);\nLUAI_FUNC void luaK_indexed (FuncState *fs, expdesc *t, expdesc *k);\nLUAI_FUNC void luaK_goiftrue (FuncState *fs, expdesc *e);\nLUAI_FUNC void luaK_storevar (FuncState *fs, expdesc *var, expdesc *e);\nLUAI_FUNC void luaK_setreturns (FuncState *fs, expdesc *e, int nresults);\nLUAI_FUNC void luaK_setoneret (FuncState *fs, expdesc *e);\nLUAI_FUNC int luaK_jump (FuncState *fs);\nLUAI_FUNC void luaK_ret (FuncState *fs, int first, int nret);\nLUAI_FUNC void luaK_patchlist (FuncState *fs, int list, int target);\nLUAI_FUNC void luaK_patchtohere (FuncState *fs, int list);\nLUAI_FUNC void luaK_concat (FuncState *fs, int *l1, int l2);\nLUAI_FUNC int luaK_getlabel (FuncState *fs);\nLUAI_FUNC void luaK_prefix (FuncState *fs, UnOpr op, expdesc *v);\nLUAI_FUNC void luaK_infix (FuncState *fs, BinOpr op, expdesc *v);\nLUAI_FUNC void luaK_posfix (FuncState *fs, BinOpr op, expdesc *v1, expdesc *v2);\nLUAI_FUNC void luaK_setlist (FuncState *fs, int base, int nelems, int tostore);\n\n\n#endif\n" }, { "path": "deps/lua/src/ldblib.c", "content": "/*\n** $Id: ldblib.c,v 1.104.1.4 2009/08/04 18:50:18 roberto Exp $\n** Interface from Lua to its debug API\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n\n#define ldblib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n\nstatic int db_getregistry (lua_State *L) {\n lua_pushvalue(L, LUA_REGISTRYINDEX);\n return 1;\n}\n\n\nstatic int db_getmetatable (lua_State *L) {\n luaL_checkany(L, 1);\n if (!lua_getmetatable(L, 1)) {\n lua_pushnil(L); /* no metatable */\n }\n return 1;\n}\n\n\nstatic int db_setmetatable (lua_State *L) {\n int t = lua_type(L, 2);\n luaL_argcheck(L, t == LUA_TNIL || t == LUA_TTABLE, 2,\n \"nil or table expected\");\n lua_settop(L, 2);\n lua_pushboolean(L, lua_setmetatable(L, 1));\n return 1;\n}\n\n\nstatic int db_getfenv (lua_State *L) {\n luaL_checkany(L, 1);\n lua_getfenv(L, 1);\n return 1;\n}\n\n\nstatic int db_setfenv (lua_State *L) {\n luaL_checktype(L, 2, LUA_TTABLE);\n lua_settop(L, 2);\n if (lua_setfenv(L, 1) == 0)\n luaL_error(L, LUA_QL(\"setfenv\")\n \" cannot change environment of given object\");\n return 1;\n}\n\n\nstatic void settabss (lua_State *L, const char *i, const char *v) {\n lua_pushstring(L, v);\n lua_setfield(L, -2, i);\n}\n\n\nstatic void settabsi (lua_State *L, const char *i, int v) {\n lua_pushinteger(L, v);\n lua_setfield(L, -2, i);\n}\n\n\nstatic lua_State *getthread (lua_State *L, int *arg) {\n if (lua_isthread(L, 1)) {\n *arg = 1;\n return lua_tothread(L, 1);\n }\n else {\n *arg = 0;\n return L;\n }\n}\n\n\nstatic void treatstackoption (lua_State *L, lua_State *L1, const char *fname) {\n if (L == L1) {\n lua_pushvalue(L, -2);\n lua_remove(L, -3);\n }\n else\n lua_xmove(L1, L, 1);\n lua_setfield(L, -2, fname);\n}\n\n\nstatic int db_getinfo (lua_State *L) {\n lua_Debug ar;\n int arg;\n lua_State *L1 = getthread(L, &arg);\n const char *options = luaL_optstring(L, arg+2, \"flnSu\");\n if (lua_isnumber(L, arg+1)) {\n if (!lua_getstack(L1, (int)lua_tointeger(L, arg+1), &ar)) {\n lua_pushnil(L); /* level out of range */\n return 1;\n }\n }\n else if (lua_isfunction(L, arg+1)) {\n lua_pushfstring(L, \">%s\", options);\n options = lua_tostring(L, -1);\n lua_pushvalue(L, arg+1);\n lua_xmove(L, L1, 1);\n }\n else\n return luaL_argerror(L, arg+1, \"function or level expected\");\n if (!lua_getinfo(L1, options, &ar))\n return luaL_argerror(L, arg+2, \"invalid option\");\n lua_createtable(L, 0, 2);\n if (strchr(options, 'S')) {\n settabss(L, \"source\", ar.source);\n settabss(L, \"short_src\", ar.short_src);\n settabsi(L, \"linedefined\", ar.linedefined);\n settabsi(L, \"lastlinedefined\", ar.lastlinedefined);\n settabss(L, \"what\", ar.what);\n }\n if (strchr(options, 'l'))\n settabsi(L, \"currentline\", ar.currentline);\n if (strchr(options, 'u'))\n settabsi(L, \"nups\", ar.nups);\n if (strchr(options, 'n')) {\n settabss(L, \"name\", ar.name);\n settabss(L, \"namewhat\", ar.namewhat);\n }\n if (strchr(options, 'L'))\n treatstackoption(L, L1, \"activelines\");\n if (strchr(options, 'f'))\n treatstackoption(L, L1, \"func\");\n return 1; /* return table */\n}\n \n\nstatic int db_getlocal (lua_State *L) {\n int arg;\n lua_State *L1 = getthread(L, &arg);\n lua_Debug ar;\n const char *name;\n if (!lua_getstack(L1, luaL_checkint(L, arg+1), &ar)) /* out of range? */\n return luaL_argerror(L, arg+1, \"level out of range\");\n name = lua_getlocal(L1, &ar, luaL_checkint(L, arg+2));\n if (name) {\n lua_xmove(L1, L, 1);\n lua_pushstring(L, name);\n lua_pushvalue(L, -2);\n return 2;\n }\n else {\n lua_pushnil(L);\n return 1;\n }\n}\n\n\nstatic int db_setlocal (lua_State *L) {\n int arg;\n lua_State *L1 = getthread(L, &arg);\n lua_Debug ar;\n if (!lua_getstack(L1, luaL_checkint(L, arg+1), &ar)) /* out of range? */\n return luaL_argerror(L, arg+1, \"level out of range\");\n luaL_checkany(L, arg+3);\n lua_settop(L, arg+3);\n lua_xmove(L, L1, 1);\n lua_pushstring(L, lua_setlocal(L1, &ar, luaL_checkint(L, arg+2)));\n return 1;\n}\n\n\nstatic int auxupvalue (lua_State *L, int get) {\n const char *name;\n int n = luaL_checkint(L, 2);\n luaL_checktype(L, 1, LUA_TFUNCTION);\n if (lua_iscfunction(L, 1)) return 0; /* cannot touch C upvalues from Lua */\n name = get ? lua_getupvalue(L, 1, n) : lua_setupvalue(L, 1, n);\n if (name == NULL) return 0;\n lua_pushstring(L, name);\n lua_insert(L, -(get+1));\n return get + 1;\n}\n\n\nstatic int db_getupvalue (lua_State *L) {\n return auxupvalue(L, 1);\n}\n\n\nstatic int db_setupvalue (lua_State *L) {\n luaL_checkany(L, 3);\n return auxupvalue(L, 0);\n}\n\n\n\nstatic const char KEY_HOOK = 'h';\n\n\nstatic void hookf (lua_State *L, lua_Debug *ar) {\n static const char *const hooknames[] =\n {\"call\", \"return\", \"line\", \"count\", \"tail return\"};\n lua_pushlightuserdata(L, (void *)&KEY_HOOK);\n lua_rawget(L, LUA_REGISTRYINDEX);\n lua_pushlightuserdata(L, L);\n lua_rawget(L, -2);\n if (lua_isfunction(L, -1)) {\n lua_pushstring(L, hooknames[(int)ar->event]);\n if (ar->currentline >= 0)\n lua_pushinteger(L, ar->currentline);\n else lua_pushnil(L);\n lua_assert(lua_getinfo(L, \"lS\", ar));\n lua_call(L, 2, 0);\n }\n}\n\n\nstatic int makemask (const char *smask, int count) {\n int mask = 0;\n if (strchr(smask, 'c')) mask |= LUA_MASKCALL;\n if (strchr(smask, 'r')) mask |= LUA_MASKRET;\n if (strchr(smask, 'l')) mask |= LUA_MASKLINE;\n if (count > 0) mask |= LUA_MASKCOUNT;\n return mask;\n}\n\n\nstatic char *unmakemask (int mask, char *smask) {\n int i = 0;\n if (mask & LUA_MASKCALL) smask[i++] = 'c';\n if (mask & LUA_MASKRET) smask[i++] = 'r';\n if (mask & LUA_MASKLINE) smask[i++] = 'l';\n smask[i] = '\\0';\n return smask;\n}\n\n\nstatic void gethooktable (lua_State *L) {\n lua_pushlightuserdata(L, (void *)&KEY_HOOK);\n lua_rawget(L, LUA_REGISTRYINDEX);\n if (!lua_istable(L, -1)) {\n lua_pop(L, 1);\n lua_createtable(L, 0, 1);\n lua_pushlightuserdata(L, (void *)&KEY_HOOK);\n lua_pushvalue(L, -2);\n lua_rawset(L, LUA_REGISTRYINDEX);\n }\n}\n\n\nstatic int db_sethook (lua_State *L) {\n int arg, mask, count;\n lua_Hook func;\n lua_State *L1 = getthread(L, &arg);\n if (lua_isnoneornil(L, arg+1)) {\n lua_settop(L, arg+1);\n func = NULL; mask = 0; count = 0; /* turn off hooks */\n }\n else {\n const char *smask = luaL_checkstring(L, arg+2);\n luaL_checktype(L, arg+1, LUA_TFUNCTION);\n count = luaL_optint(L, arg+3, 0);\n func = hookf; mask = makemask(smask, count);\n }\n gethooktable(L);\n lua_pushlightuserdata(L, L1);\n lua_pushvalue(L, arg+1);\n lua_rawset(L, -3); /* set new hook */\n lua_pop(L, 1); /* remove hook table */\n lua_sethook(L1, func, mask, count); /* set hooks */\n return 0;\n}\n\n\nstatic int db_gethook (lua_State *L) {\n int arg;\n lua_State *L1 = getthread(L, &arg);\n char buff[5];\n int mask = lua_gethookmask(L1);\n lua_Hook hook = lua_gethook(L1);\n if (hook != NULL && hook != hookf) /* external hook? */\n lua_pushliteral(L, \"external hook\");\n else {\n gethooktable(L);\n lua_pushlightuserdata(L, L1);\n lua_rawget(L, -2); /* get hook */\n lua_remove(L, -2); /* remove hook table */\n }\n lua_pushstring(L, unmakemask(mask, buff));\n lua_pushinteger(L, lua_gethookcount(L1));\n return 3;\n}\n\n\nstatic int db_debug (lua_State *L) {\n for (;;) {\n char buffer[250];\n fputs(\"lua_debug> \", stderr);\n if (fgets(buffer, sizeof(buffer), stdin) == 0 ||\n strcmp(buffer, \"cont\\n\") == 0)\n return 0;\n if (luaL_loadbuffer(L, buffer, strlen(buffer), \"=(debug command)\") ||\n lua_pcall(L, 0, 0, 0)) {\n fputs(lua_tostring(L, -1), stderr);\n fputs(\"\\n\", stderr);\n }\n lua_settop(L, 0); /* remove eventual returns */\n }\n}\n\n\n#define LEVELS1\t12\t/* size of the first part of the stack */\n#define LEVELS2\t10\t/* size of the second part of the stack */\n\nstatic int db_errorfb (lua_State *L) {\n int level;\n int firstpart = 1; /* still before eventual `...' */\n int arg;\n lua_State *L1 = getthread(L, &arg);\n lua_Debug ar;\n if (lua_isnumber(L, arg+2)) {\n level = (int)lua_tointeger(L, arg+2);\n lua_pop(L, 1);\n }\n else\n level = (L == L1) ? 1 : 0; /* level 0 may be this own function */\n if (lua_gettop(L) == arg)\n lua_pushliteral(L, \"\");\n else if (!lua_isstring(L, arg+1)) return 1; /* message is not a string */\n else lua_pushliteral(L, \"\\n\");\n lua_pushliteral(L, \"stack traceback:\");\n while (lua_getstack(L1, level++, &ar)) {\n if (level > LEVELS1 && firstpart) {\n /* no more than `LEVELS2' more levels? */\n if (!lua_getstack(L1, level+LEVELS2, &ar))\n level--; /* keep going */\n else {\n lua_pushliteral(L, \"\\n\\t...\"); /* too many levels */\n while (lua_getstack(L1, level+LEVELS2, &ar)) /* find last levels */\n level++;\n }\n firstpart = 0;\n continue;\n }\n lua_pushliteral(L, \"\\n\\t\");\n lua_getinfo(L1, \"Snl\", &ar);\n lua_pushfstring(L, \"%s:\", ar.short_src);\n if (ar.currentline > 0)\n lua_pushfstring(L, \"%d:\", ar.currentline);\n if (*ar.namewhat != '\\0') /* is there a name? */\n lua_pushfstring(L, \" in function \" LUA_QS, ar.name);\n else {\n if (*ar.what == 'm') /* main? */\n lua_pushfstring(L, \" in main chunk\");\n else if (*ar.what == 'C' || *ar.what == 't')\n lua_pushliteral(L, \" ?\"); /* C function or tail call */\n else\n lua_pushfstring(L, \" in function <%s:%d>\",\n ar.short_src, ar.linedefined);\n }\n lua_concat(L, lua_gettop(L) - arg);\n }\n lua_concat(L, lua_gettop(L) - arg);\n return 1;\n}\n\n\nstatic const luaL_Reg dblib[] = {\n {\"debug\", db_debug},\n {\"getfenv\", db_getfenv},\n {\"gethook\", db_gethook},\n {\"getinfo\", db_getinfo},\n {\"getlocal\", db_getlocal},\n {\"getregistry\", db_getregistry},\n {\"getmetatable\", db_getmetatable},\n {\"getupvalue\", db_getupvalue},\n {\"setfenv\", db_setfenv},\n {\"sethook\", db_sethook},\n {\"setlocal\", db_setlocal},\n {\"setmetatable\", db_setmetatable},\n {\"setupvalue\", db_setupvalue},\n {\"traceback\", db_errorfb},\n {NULL, NULL}\n};\n\n\nLUALIB_API int luaopen_debug (lua_State *L) {\n luaL_register(L, LUA_DBLIBNAME, dblib);\n return 1;\n}\n\n" }, { "path": "deps/lua/src/ldebug.c", "content": "/*\n** $Id: ldebug.c,v 2.29.1.6 2008/05/08 16:56:26 roberto Exp $\n** Debug Interface\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n\n\n#define ldebug_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lapi.h\"\n#include \"lcode.h\"\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n#include \"lvm.h\"\n\n\n\nstatic const char *getfuncname (lua_State *L, CallInfo *ci, const char **name);\n\n\nstatic int currentpc (lua_State *L, CallInfo *ci) {\n if (!isLua(ci)) return -1; /* function is not a Lua function? */\n if (ci == L->ci)\n ci->savedpc = L->savedpc;\n return pcRel(ci->savedpc, ci_func(ci)->l.p);\n}\n\n\nstatic int currentline (lua_State *L, CallInfo *ci) {\n int pc = currentpc(L, ci);\n if (pc < 0)\n return -1; /* only active lua functions have current-line information */\n else\n return getline(ci_func(ci)->l.p, pc);\n}\n\n\n/*\n** this function can be called asynchronous (e.g. during a signal)\n*/\nLUA_API int lua_sethook (lua_State *L, lua_Hook func, int mask, int count) {\n if (func == NULL || mask == 0) { /* turn off hooks? */\n mask = 0;\n func = NULL;\n }\n L->hook = func;\n L->basehookcount = count;\n resethookcount(L);\n L->hookmask = cast_byte(mask);\n return 1;\n}\n\n\nLUA_API lua_Hook lua_gethook (lua_State *L) {\n return L->hook;\n}\n\n\nLUA_API int lua_gethookmask (lua_State *L) {\n return L->hookmask;\n}\n\n\nLUA_API int lua_gethookcount (lua_State *L) {\n return L->basehookcount;\n}\n\n\nLUA_API int lua_getstack (lua_State *L, int level, lua_Debug *ar) {\n int status;\n CallInfo *ci;\n lua_lock(L);\n for (ci = L->ci; level > 0 && ci > L->base_ci; ci--) {\n level--;\n if (f_isLua(ci)) /* Lua function? */\n level -= ci->tailcalls; /* skip lost tail calls */\n }\n if (level == 0 && ci > L->base_ci) { /* level found? */\n status = 1;\n ar->i_ci = cast_int(ci - L->base_ci);\n }\n else if (level < 0) { /* level is of a lost tail call? */\n status = 1;\n ar->i_ci = 0;\n }\n else status = 0; /* no such level */\n lua_unlock(L);\n return status;\n}\n\n\nstatic Proto *getluaproto (CallInfo *ci) {\n return (isLua(ci) ? ci_func(ci)->l.p : NULL);\n}\n\n\nstatic const char *findlocal (lua_State *L, CallInfo *ci, int n) {\n const char *name;\n Proto *fp = getluaproto(ci);\n if (fp && (name = luaF_getlocalname(fp, n, currentpc(L, ci))) != NULL)\n return name; /* is a local variable in a Lua function */\n else {\n StkId limit = (ci == L->ci) ? L->top : (ci+1)->func;\n if (limit - ci->base >= n && n > 0) /* is 'n' inside 'ci' stack? */\n return \"(*temporary)\";\n else\n return NULL;\n }\n}\n\n\nLUA_API const char *lua_getlocal (lua_State *L, const lua_Debug *ar, int n) {\n CallInfo *ci = L->base_ci + ar->i_ci;\n const char *name = findlocal(L, ci, n);\n lua_lock(L);\n if (name)\n luaA_pushobject(L, ci->base + (n - 1));\n lua_unlock(L);\n return name;\n}\n\n\nLUA_API const char *lua_setlocal (lua_State *L, const lua_Debug *ar, int n) {\n CallInfo *ci = L->base_ci + ar->i_ci;\n const char *name = findlocal(L, ci, n);\n lua_lock(L);\n if (name)\n setobjs2s(L, ci->base + (n - 1), L->top - 1);\n L->top--; /* pop value */\n lua_unlock(L);\n return name;\n}\n\n\nstatic void funcinfo (lua_Debug *ar, Closure *cl) {\n if (cl->c.isC) {\n ar->source = \"=[C]\";\n ar->linedefined = -1;\n ar->lastlinedefined = -1;\n ar->what = \"C\";\n }\n else {\n ar->source = getstr(cl->l.p->source);\n ar->linedefined = cl->l.p->linedefined;\n ar->lastlinedefined = cl->l.p->lastlinedefined;\n ar->what = (ar->linedefined == 0) ? \"main\" : \"Lua\";\n }\n luaO_chunkid(ar->short_src, ar->source, LUA_IDSIZE);\n}\n\n\nstatic void info_tailcall (lua_Debug *ar) {\n ar->name = ar->namewhat = \"\";\n ar->what = \"tail\";\n ar->lastlinedefined = ar->linedefined = ar->currentline = -1;\n ar->source = \"=(tail call)\";\n luaO_chunkid(ar->short_src, ar->source, LUA_IDSIZE);\n ar->nups = 0;\n}\n\n\nstatic void collectvalidlines (lua_State *L, Closure *f) {\n if (f == NULL || f->c.isC) {\n setnilvalue(L->top);\n }\n else {\n Table *t = luaH_new(L, 0, 0);\n int *lineinfo = f->l.p->lineinfo;\n int i;\n for (i=0; il.p->sizelineinfo; i++)\n setbvalue(luaH_setnum(L, t, lineinfo[i]), 1);\n sethvalue(L, L->top, t); \n }\n incr_top(L);\n}\n\n\nstatic int auxgetinfo (lua_State *L, const char *what, lua_Debug *ar,\n Closure *f, CallInfo *ci) {\n int status = 1;\n if (f == NULL) {\n info_tailcall(ar);\n return status;\n }\n for (; *what; what++) {\n switch (*what) {\n case 'S': {\n funcinfo(ar, f);\n break;\n }\n case 'l': {\n ar->currentline = (ci) ? currentline(L, ci) : -1;\n break;\n }\n case 'u': {\n ar->nups = f->c.nupvalues;\n break;\n }\n case 'n': {\n ar->namewhat = (ci) ? getfuncname(L, ci, &ar->name) : NULL;\n if (ar->namewhat == NULL) {\n ar->namewhat = \"\"; /* not found */\n ar->name = NULL;\n }\n break;\n }\n case 'L':\n case 'f': /* handled by lua_getinfo */\n break;\n default: status = 0; /* invalid option */\n }\n }\n return status;\n}\n\n\nLUA_API int lua_getinfo (lua_State *L, const char *what, lua_Debug *ar) {\n int status;\n Closure *f = NULL;\n CallInfo *ci = NULL;\n lua_lock(L);\n if (*what == '>') {\n StkId func = L->top - 1;\n luai_apicheck(L, ttisfunction(func));\n what++; /* skip the '>' */\n f = clvalue(func);\n L->top--; /* pop function */\n }\n else if (ar->i_ci != 0) { /* no tail call? */\n ci = L->base_ci + ar->i_ci;\n lua_assert(ttisfunction(ci->func));\n f = clvalue(ci->func);\n }\n status = auxgetinfo(L, what, ar, f, ci);\n if (strchr(what, 'f')) {\n if (f == NULL) setnilvalue(L->top);\n else setclvalue(L, L->top, f);\n incr_top(L);\n }\n if (strchr(what, 'L'))\n collectvalidlines(L, f);\n lua_unlock(L);\n return status;\n}\n\n\n/*\n** {======================================================\n** Symbolic Execution and code checker\n** =======================================================\n*/\n\n#define check(x)\t\tif (!(x)) return 0;\n\n#define checkjump(pt,pc)\tcheck(0 <= pc && pc < pt->sizecode)\n\n#define checkreg(pt,reg)\tcheck((reg) < (pt)->maxstacksize)\n\n\n\nstatic int precheck (const Proto *pt) {\n check(pt->maxstacksize <= MAXSTACK);\n check(pt->numparams+(pt->is_vararg & VARARG_HASARG) <= pt->maxstacksize);\n check(!(pt->is_vararg & VARARG_NEEDSARG) ||\n (pt->is_vararg & VARARG_HASARG));\n check(pt->sizeupvalues <= pt->nups);\n check(pt->sizelineinfo == pt->sizecode || pt->sizelineinfo == 0);\n check(pt->sizecode > 0 && GET_OPCODE(pt->code[pt->sizecode-1]) == OP_RETURN);\n return 1;\n}\n\n\n#define checkopenop(pt,pc)\tluaG_checkopenop((pt)->code[(pc)+1])\n\nint luaG_checkopenop (Instruction i) {\n switch (GET_OPCODE(i)) {\n case OP_CALL:\n case OP_TAILCALL:\n case OP_RETURN:\n case OP_SETLIST: {\n check(GETARG_B(i) == 0);\n return 1;\n }\n default: return 0; /* invalid instruction after an open call */\n }\n}\n\n\nstatic int checkArgMode (const Proto *pt, int r, enum OpArgMask mode) {\n switch (mode) {\n case OpArgN: check(r == 0); break;\n case OpArgU: break;\n case OpArgR: checkreg(pt, r); break;\n case OpArgK:\n check(ISK(r) ? INDEXK(r) < pt->sizek : r < pt->maxstacksize);\n break;\n }\n return 1;\n}\n\n\nstatic Instruction symbexec (const Proto *pt, int lastpc, int reg) {\n int pc;\n int last; /* stores position of last instruction that changed `reg' */\n last = pt->sizecode-1; /* points to final return (a `neutral' instruction) */\n check(precheck(pt));\n for (pc = 0; pc < lastpc; pc++) {\n Instruction i = pt->code[pc];\n OpCode op = GET_OPCODE(i);\n int a = GETARG_A(i);\n int b = 0;\n int c = 0;\n check(op < NUM_OPCODES);\n checkreg(pt, a);\n switch (getOpMode(op)) {\n case iABC: {\n b = GETARG_B(i);\n c = GETARG_C(i);\n check(checkArgMode(pt, b, getBMode(op)));\n check(checkArgMode(pt, c, getCMode(op)));\n break;\n }\n case iABx: {\n b = GETARG_Bx(i);\n if (getBMode(op) == OpArgK) check(b < pt->sizek);\n break;\n }\n case iAsBx: {\n b = GETARG_sBx(i);\n if (getBMode(op) == OpArgR) {\n int dest = pc+1+b;\n check(0 <= dest && dest < pt->sizecode);\n if (dest > 0) {\n int j;\n /* check that it does not jump to a setlist count; this\n is tricky, because the count from a previous setlist may\n have the same value of an invalid setlist; so, we must\n go all the way back to the first of them (if any) */\n for (j = 0; j < dest; j++) {\n Instruction d = pt->code[dest-1-j];\n if (!(GET_OPCODE(d) == OP_SETLIST && GETARG_C(d) == 0)) break;\n }\n /* if 'j' is even, previous value is not a setlist (even if\n it looks like one) */\n check((j&1) == 0);\n }\n }\n break;\n }\n }\n if (testAMode(op)) {\n if (a == reg) last = pc; /* change register `a' */\n }\n if (testTMode(op)) {\n check(pc+2 < pt->sizecode); /* check skip */\n check(GET_OPCODE(pt->code[pc+1]) == OP_JMP);\n }\n switch (op) {\n case OP_LOADBOOL: {\n if (c == 1) { /* does it jump? */\n check(pc+2 < pt->sizecode); /* check its jump */\n check(GET_OPCODE(pt->code[pc+1]) != OP_SETLIST ||\n GETARG_C(pt->code[pc+1]) != 0);\n }\n break;\n }\n case OP_LOADNIL: {\n if (a <= reg && reg <= b)\n last = pc; /* set registers from `a' to `b' */\n break;\n }\n case OP_GETUPVAL:\n case OP_SETUPVAL: {\n check(b < pt->nups);\n break;\n }\n case OP_GETGLOBAL:\n case OP_SETGLOBAL: {\n check(ttisstring(&pt->k[b]));\n break;\n }\n case OP_SELF: {\n checkreg(pt, a+1);\n if (reg == a+1) last = pc;\n break;\n }\n case OP_CONCAT: {\n check(b < c); /* at least two operands */\n break;\n }\n case OP_TFORLOOP: {\n check(c >= 1); /* at least one result (control variable) */\n checkreg(pt, a+2+c); /* space for results */\n if (reg >= a+2) last = pc; /* affect all regs above its base */\n break;\n }\n case OP_FORLOOP:\n case OP_FORPREP:\n checkreg(pt, a+3);\n /* go through */\n case OP_JMP: {\n int dest = pc+1+b;\n /* not full check and jump is forward and do not skip `lastpc'? */\n if (reg != NO_REG && pc < dest && dest <= lastpc)\n pc += b; /* do the jump */\n break;\n }\n case OP_CALL:\n case OP_TAILCALL: {\n if (b != 0) {\n checkreg(pt, a+b-1);\n }\n c--; /* c = num. returns */\n if (c == LUA_MULTRET) {\n check(checkopenop(pt, pc));\n }\n else if (c != 0)\n checkreg(pt, a+c-1);\n if (reg >= a) last = pc; /* affect all registers above base */\n break;\n }\n case OP_RETURN: {\n b--; /* b = num. returns */\n if (b > 0) checkreg(pt, a+b-1);\n break;\n }\n case OP_SETLIST: {\n if (b > 0) checkreg(pt, a + b);\n if (c == 0) {\n pc++;\n check(pc < pt->sizecode - 1);\n }\n break;\n }\n case OP_CLOSURE: {\n int nup, j;\n check(b < pt->sizep);\n nup = pt->p[b]->nups;\n check(pc + nup < pt->sizecode);\n for (j = 1; j <= nup; j++) {\n OpCode op1 = GET_OPCODE(pt->code[pc + j]);\n check(op1 == OP_GETUPVAL || op1 == OP_MOVE);\n }\n if (reg != NO_REG) /* tracing? */\n pc += nup; /* do not 'execute' these pseudo-instructions */\n break;\n }\n case OP_VARARG: {\n check((pt->is_vararg & VARARG_ISVARARG) &&\n !(pt->is_vararg & VARARG_NEEDSARG));\n b--;\n if (b == LUA_MULTRET) check(checkopenop(pt, pc));\n checkreg(pt, a+b-1);\n break;\n }\n default: break;\n }\n }\n return pt->code[last];\n}\n\n#undef check\n#undef checkjump\n#undef checkreg\n\n/* }====================================================== */\n\n\nint luaG_checkcode (const Proto *pt) {\n return (symbexec(pt, pt->sizecode, NO_REG) != 0);\n}\n\n\nstatic const char *kname (Proto *p, int c) {\n if (ISK(c) && ttisstring(&p->k[INDEXK(c)]))\n return svalue(&p->k[INDEXK(c)]);\n else\n return \"?\";\n}\n\n\nstatic const char *getobjname (lua_State *L, CallInfo *ci, int stackpos,\n const char **name) {\n if (isLua(ci)) { /* a Lua function? */\n Proto *p = ci_func(ci)->l.p;\n int pc = currentpc(L, ci);\n Instruction i;\n *name = luaF_getlocalname(p, stackpos+1, pc);\n if (*name) /* is a local? */\n return \"local\";\n i = symbexec(p, pc, stackpos); /* try symbolic execution */\n lua_assert(pc != -1);\n switch (GET_OPCODE(i)) {\n case OP_GETGLOBAL: {\n int g = GETARG_Bx(i); /* global index */\n lua_assert(ttisstring(&p->k[g]));\n *name = svalue(&p->k[g]);\n return \"global\";\n }\n case OP_MOVE: {\n int a = GETARG_A(i);\n int b = GETARG_B(i); /* move from `b' to `a' */\n if (b < a)\n return getobjname(L, ci, b, name); /* get name for `b' */\n break;\n }\n case OP_GETTABLE: {\n int k = GETARG_C(i); /* key index */\n *name = kname(p, k);\n return \"field\";\n }\n case OP_GETUPVAL: {\n int u = GETARG_B(i); /* upvalue index */\n *name = p->upvalues ? getstr(p->upvalues[u]) : \"?\";\n return \"upvalue\";\n }\n case OP_SELF: {\n int k = GETARG_C(i); /* key index */\n *name = kname(p, k);\n return \"method\";\n }\n default: break;\n }\n }\n return NULL; /* no useful name found */\n}\n\n\nstatic const char *getfuncname (lua_State *L, CallInfo *ci, const char **name) {\n Instruction i;\n if ((isLua(ci) && ci->tailcalls > 0) || !isLua(ci - 1))\n return NULL; /* calling function is not Lua (or is unknown) */\n ci--; /* calling function */\n i = ci_func(ci)->l.p->code[currentpc(L, ci)];\n if (GET_OPCODE(i) == OP_CALL || GET_OPCODE(i) == OP_TAILCALL ||\n GET_OPCODE(i) == OP_TFORLOOP)\n return getobjname(L, ci, GETARG_A(i), name);\n else\n return NULL; /* no useful name can be found */\n}\n\n\n/* only ANSI way to check whether a pointer points to an array */\nstatic int isinstack (CallInfo *ci, const TValue *o) {\n StkId p;\n for (p = ci->base; p < ci->top; p++)\n if (o == p) return 1;\n return 0;\n}\n\n\nvoid luaG_typeerror (lua_State *L, const TValue *o, const char *op) {\n const char *name = NULL;\n const char *t = luaT_typenames[ttype(o)];\n const char *kind = (isinstack(L->ci, o)) ?\n getobjname(L, L->ci, cast_int(o - L->base), &name) :\n NULL;\n if (kind)\n luaG_runerror(L, \"attempt to %s %s \" LUA_QS \" (a %s value)\",\n op, kind, name, t);\n else\n luaG_runerror(L, \"attempt to %s a %s value\", op, t);\n}\n\n\nvoid luaG_concaterror (lua_State *L, StkId p1, StkId p2) {\n if (ttisstring(p1) || ttisnumber(p1)) p1 = p2;\n lua_assert(!ttisstring(p1) && !ttisnumber(p1));\n luaG_typeerror(L, p1, \"concatenate\");\n}\n\n\nvoid luaG_aritherror (lua_State *L, const TValue *p1, const TValue *p2) {\n TValue temp;\n if (luaV_tonumber(p1, &temp) == NULL)\n p2 = p1; /* first operand is wrong */\n luaG_typeerror(L, p2, \"perform arithmetic on\");\n}\n\n\nint luaG_ordererror (lua_State *L, const TValue *p1, const TValue *p2) {\n const char *t1 = luaT_typenames[ttype(p1)];\n const char *t2 = luaT_typenames[ttype(p2)];\n if (t1[2] == t2[2])\n luaG_runerror(L, \"attempt to compare two %s values\", t1);\n else\n luaG_runerror(L, \"attempt to compare %s with %s\", t1, t2);\n return 0;\n}\n\n\nstatic void addinfo (lua_State *L, const char *msg) {\n CallInfo *ci = L->ci;\n if (isLua(ci)) { /* is Lua code? */\n char buff[LUA_IDSIZE]; /* add file:line information */\n int line = currentline(L, ci);\n luaO_chunkid(buff, getstr(getluaproto(ci)->source), LUA_IDSIZE);\n luaO_pushfstring(L, \"%s:%d: %s\", buff, line, msg);\n }\n}\n\n\nvoid luaG_errormsg (lua_State *L) {\n if (L->errfunc != 0) { /* is there an error handling function? */\n StkId errfunc = restorestack(L, L->errfunc);\n if (!ttisfunction(errfunc)) luaD_throw(L, LUA_ERRERR);\n setobjs2s(L, L->top, L->top - 1); /* move argument */\n setobjs2s(L, L->top - 1, errfunc); /* push function */\n incr_top(L);\n luaD_call(L, L->top - 2, 1); /* call it */\n }\n luaD_throw(L, LUA_ERRRUN);\n}\n\n\nvoid luaG_runerror (lua_State *L, const char *fmt, ...) {\n va_list argp;\n va_start(argp, fmt);\n addinfo(L, luaO_pushvfstring(L, fmt, argp));\n va_end(argp);\n luaG_errormsg(L);\n}\n\n" }, { "path": "deps/lua/src/ldebug.h", "content": "/*\n** $Id: ldebug.h,v 2.3.1.1 2007/12/27 13:02:25 roberto Exp $\n** Auxiliary functions from Debug Interface module\n** See Copyright Notice in lua.h\n*/\n\n#ifndef ldebug_h\n#define ldebug_h\n\n\n#include \"lstate.h\"\n\n\n#define pcRel(pc, p)\t(cast(int, (pc) - (p)->code) - 1)\n\n#define getline(f,pc)\t(((f)->lineinfo) ? (f)->lineinfo[pc] : 0)\n\n#define resethookcount(L)\t(L->hookcount = L->basehookcount)\n\n\nLUAI_FUNC void luaG_typeerror (lua_State *L, const TValue *o,\n const char *opname);\nLUAI_FUNC void luaG_concaterror (lua_State *L, StkId p1, StkId p2);\nLUAI_FUNC void luaG_aritherror (lua_State *L, const TValue *p1,\n const TValue *p2);\nLUAI_FUNC int luaG_ordererror (lua_State *L, const TValue *p1,\n const TValue *p2);\nLUAI_FUNC void luaG_runerror (lua_State *L, const char *fmt, ...);\nLUAI_FUNC void luaG_errormsg (lua_State *L);\nLUAI_FUNC int luaG_checkcode (const Proto *pt);\nLUAI_FUNC int luaG_checkopenop (Instruction i);\n\n#endif\n" }, { "path": "deps/lua/src/ldo.c", "content": "/*\n** $Id: ldo.c,v 2.38.1.4 2012/01/18 02:27:10 roberto Exp $\n** Stack and Call structure of Lua\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n\n#define ldo_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lparser.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n#include \"lundump.h\"\n#include \"lvm.h\"\n#include \"lzio.h\"\n\n\n\n\n/*\n** {======================================================\n** Error-recovery functions\n** =======================================================\n*/\n\n\n/* chain list of long jump buffers */\nstruct lua_longjmp {\n struct lua_longjmp *previous;\n luai_jmpbuf b;\n volatile int status; /* error code */\n};\n\n\nvoid luaD_seterrorobj (lua_State *L, int errcode, StkId oldtop) {\n switch (errcode) {\n case LUA_ERRMEM: {\n setsvalue2s(L, oldtop, luaS_newliteral(L, MEMERRMSG));\n break;\n }\n case LUA_ERRERR: {\n setsvalue2s(L, oldtop, luaS_newliteral(L, \"error in error handling\"));\n break;\n }\n case LUA_ERRSYNTAX:\n case LUA_ERRRUN: {\n setobjs2s(L, oldtop, L->top - 1); /* error message on current top */\n break;\n }\n }\n L->top = oldtop + 1;\n}\n\n\nstatic void restore_stack_limit (lua_State *L) {\n lua_assert(L->stack_last - L->stack == L->stacksize - EXTRA_STACK - 1);\n if (L->size_ci > LUAI_MAXCALLS) { /* there was an overflow? */\n int inuse = cast_int(L->ci - L->base_ci);\n if (inuse + 1 < LUAI_MAXCALLS) /* can `undo' overflow? */\n luaD_reallocCI(L, LUAI_MAXCALLS);\n }\n}\n\n\nstatic void resetstack (lua_State *L, int status) {\n L->ci = L->base_ci;\n L->base = L->ci->base;\n luaF_close(L, L->base); /* close eventual pending closures */\n luaD_seterrorobj(L, status, L->base);\n L->nCcalls = L->baseCcalls;\n L->allowhook = 1;\n restore_stack_limit(L);\n L->errfunc = 0;\n L->errorJmp = NULL;\n}\n\n\nvoid luaD_throw (lua_State *L, int errcode) {\n if (L->errorJmp) {\n L->errorJmp->status = errcode;\n LUAI_THROW(L, L->errorJmp);\n }\n else {\n L->status = cast_byte(errcode);\n if (G(L)->panic) {\n resetstack(L, errcode);\n lua_unlock(L);\n G(L)->panic(L);\n }\n exit(EXIT_FAILURE);\n }\n}\n\n\nint luaD_rawrunprotected (lua_State *L, Pfunc f, void *ud) {\n struct lua_longjmp lj;\n lj.status = 0;\n lj.previous = L->errorJmp; /* chain new error handler */\n L->errorJmp = &lj;\n LUAI_TRY(L, &lj,\n (*f)(L, ud);\n );\n L->errorJmp = lj.previous; /* restore old error handler */\n return lj.status;\n}\n\n/* }====================================================== */\n\n\nstatic void correctstack (lua_State *L, TValue *oldstack) {\n CallInfo *ci;\n GCObject *up;\n L->top = (L->top - oldstack) + L->stack;\n for (up = L->openupval; up != NULL; up = up->gch.next)\n gco2uv(up)->v = (gco2uv(up)->v - oldstack) + L->stack;\n for (ci = L->base_ci; ci <= L->ci; ci++) {\n ci->top = (ci->top - oldstack) + L->stack;\n ci->base = (ci->base - oldstack) + L->stack;\n ci->func = (ci->func - oldstack) + L->stack;\n }\n L->base = (L->base - oldstack) + L->stack;\n}\n\n\nvoid luaD_reallocstack (lua_State *L, int newsize) {\n TValue *oldstack = L->stack;\n int realsize = newsize + 1 + EXTRA_STACK;\n lua_assert(L->stack_last - L->stack == L->stacksize - EXTRA_STACK - 1);\n luaM_reallocvector(L, L->stack, L->stacksize, realsize, TValue);\n L->stacksize = realsize;\n L->stack_last = L->stack+newsize;\n correctstack(L, oldstack);\n}\n\n\nvoid luaD_reallocCI (lua_State *L, int newsize) {\n CallInfo *oldci = L->base_ci;\n luaM_reallocvector(L, L->base_ci, L->size_ci, newsize, CallInfo);\n L->size_ci = newsize;\n L->ci = (L->ci - oldci) + L->base_ci;\n L->end_ci = L->base_ci + L->size_ci - 1;\n}\n\n\nvoid luaD_growstack (lua_State *L, int n) {\n if (n <= L->stacksize) /* double size is enough? */\n luaD_reallocstack(L, 2*L->stacksize);\n else\n luaD_reallocstack(L, L->stacksize + n);\n}\n\n\nstatic CallInfo *growCI (lua_State *L) {\n if (L->size_ci > LUAI_MAXCALLS) /* overflow while handling overflow? */\n luaD_throw(L, LUA_ERRERR);\n else {\n luaD_reallocCI(L, 2*L->size_ci);\n if (L->size_ci > LUAI_MAXCALLS)\n luaG_runerror(L, \"stack overflow\");\n }\n return ++L->ci;\n}\n\n\nvoid luaD_callhook (lua_State *L, int event, int line) {\n lua_Hook hook = L->hook;\n if (hook && L->allowhook) {\n ptrdiff_t top = savestack(L, L->top);\n ptrdiff_t ci_top = savestack(L, L->ci->top);\n lua_Debug ar;\n ar.event = event;\n ar.currentline = line;\n if (event == LUA_HOOKTAILRET)\n ar.i_ci = 0; /* tail call; no debug information about it */\n else\n ar.i_ci = cast_int(L->ci - L->base_ci);\n luaD_checkstack(L, LUA_MINSTACK); /* ensure minimum stack size */\n L->ci->top = L->top + LUA_MINSTACK;\n lua_assert(L->ci->top <= L->stack_last);\n L->allowhook = 0; /* cannot call hooks inside a hook */\n lua_unlock(L);\n (*hook)(L, &ar);\n lua_lock(L);\n lua_assert(!L->allowhook);\n L->allowhook = 1;\n L->ci->top = restorestack(L, ci_top);\n L->top = restorestack(L, top);\n }\n}\n\n\nstatic StkId adjust_varargs (lua_State *L, Proto *p, int actual) {\n int i;\n int nfixargs = p->numparams;\n Table *htab = NULL;\n StkId base, fixed;\n for (; actual < nfixargs; ++actual)\n setnilvalue(L->top++);\n#if defined(LUA_COMPAT_VARARG)\n if (p->is_vararg & VARARG_NEEDSARG) { /* compat. with old-style vararg? */\n int nvar = actual - nfixargs; /* number of extra arguments */\n lua_assert(p->is_vararg & VARARG_HASARG);\n luaC_checkGC(L);\n luaD_checkstack(L, p->maxstacksize);\n htab = luaH_new(L, nvar, 1); /* create `arg' table */\n for (i=0; itop - nvar + i);\n /* store counter in field `n' */\n setnvalue(luaH_setstr(L, htab, luaS_newliteral(L, \"n\")), cast_num(nvar));\n }\n#endif\n /* move fixed parameters to final position */\n fixed = L->top - actual; /* first fixed argument */\n base = L->top; /* final position of first argument */\n for (i=0; itop++, fixed+i);\n setnilvalue(fixed+i);\n }\n /* add `arg' parameter */\n if (htab) {\n sethvalue(L, L->top++, htab);\n lua_assert(iswhite(obj2gco(htab)));\n }\n return base;\n}\n\n\nstatic StkId tryfuncTM (lua_State *L, StkId func) {\n const TValue *tm = luaT_gettmbyobj(L, func, TM_CALL);\n StkId p;\n ptrdiff_t funcr = savestack(L, func);\n if (!ttisfunction(tm))\n luaG_typeerror(L, func, \"call\");\n /* Open a hole inside the stack at `func' */\n for (p = L->top; p > func; p--) setobjs2s(L, p, p-1);\n incr_top(L);\n func = restorestack(L, funcr); /* previous call may change stack */\n setobj2s(L, func, tm); /* tag method is the new function to be called */\n return func;\n}\n\n\n\n#define inc_ci(L) \\\n ((L->ci == L->end_ci) ? growCI(L) : \\\n (condhardstacktests(luaD_reallocCI(L, L->size_ci)), ++L->ci))\n\n\nint luaD_precall (lua_State *L, StkId func, int nresults) {\n LClosure *cl;\n ptrdiff_t funcr;\n if (!ttisfunction(func)) /* `func' is not a function? */\n func = tryfuncTM(L, func); /* check the `function' tag method */\n funcr = savestack(L, func);\n cl = &clvalue(func)->l;\n L->ci->savedpc = L->savedpc;\n if (!cl->isC) { /* Lua function? prepare its call */\n CallInfo *ci;\n StkId st, base;\n Proto *p = cl->p;\n luaD_checkstack(L, p->maxstacksize);\n func = restorestack(L, funcr);\n if (!p->is_vararg) { /* no varargs? */\n base = func + 1;\n if (L->top > base + p->numparams)\n L->top = base + p->numparams;\n }\n else { /* vararg function */\n int nargs = cast_int(L->top - func) - 1;\n base = adjust_varargs(L, p, nargs);\n func = restorestack(L, funcr); /* previous call may change the stack */\n }\n ci = inc_ci(L); /* now `enter' new function */\n ci->func = func;\n L->base = ci->base = base;\n ci->top = L->base + p->maxstacksize;\n lua_assert(ci->top <= L->stack_last);\n L->savedpc = p->code; /* starting point */\n ci->tailcalls = 0;\n ci->nresults = nresults;\n for (st = L->top; st < ci->top; st++)\n setnilvalue(st);\n L->top = ci->top;\n if (L->hookmask & LUA_MASKCALL) {\n L->savedpc++; /* hooks assume 'pc' is already incremented */\n luaD_callhook(L, LUA_HOOKCALL, -1);\n L->savedpc--; /* correct 'pc' */\n }\n return PCRLUA;\n }\n else { /* if is a C function, call it */\n CallInfo *ci;\n int n;\n luaD_checkstack(L, LUA_MINSTACK); /* ensure minimum stack size */\n ci = inc_ci(L); /* now `enter' new function */\n ci->func = restorestack(L, funcr);\n L->base = ci->base = ci->func + 1;\n ci->top = L->top + LUA_MINSTACK;\n lua_assert(ci->top <= L->stack_last);\n ci->nresults = nresults;\n if (L->hookmask & LUA_MASKCALL)\n luaD_callhook(L, LUA_HOOKCALL, -1);\n lua_unlock(L);\n n = (*curr_func(L)->c.f)(L); /* do the actual call */\n lua_lock(L);\n if (n < 0) /* yielding? */\n return PCRYIELD;\n else {\n luaD_poscall(L, L->top - n);\n return PCRC;\n }\n }\n}\n\n\nstatic StkId callrethooks (lua_State *L, StkId firstResult) {\n ptrdiff_t fr = savestack(L, firstResult); /* next call may change stack */\n luaD_callhook(L, LUA_HOOKRET, -1);\n if (f_isLua(L->ci)) { /* Lua function? */\n while ((L->hookmask & LUA_MASKRET) && L->ci->tailcalls--) /* tail calls */\n luaD_callhook(L, LUA_HOOKTAILRET, -1);\n }\n return restorestack(L, fr);\n}\n\n\nint luaD_poscall (lua_State *L, StkId firstResult) {\n StkId res;\n int wanted, i;\n CallInfo *ci;\n if (L->hookmask & LUA_MASKRET)\n firstResult = callrethooks(L, firstResult);\n ci = L->ci--;\n res = ci->func; /* res == final position of 1st result */\n wanted = ci->nresults;\n L->base = (ci - 1)->base; /* restore base */\n L->savedpc = (ci - 1)->savedpc; /* restore savedpc */\n /* move results to correct place */\n for (i = wanted; i != 0 && firstResult < L->top; i--)\n setobjs2s(L, res++, firstResult++);\n while (i-- > 0)\n setnilvalue(res++);\n L->top = res;\n return (wanted - LUA_MULTRET); /* 0 iff wanted == LUA_MULTRET */\n}\n\n\n/*\n** Call a function (C or Lua). The function to be called is at *func.\n** The arguments are on the stack, right after the function.\n** When returns, all the results are on the stack, starting at the original\n** function position.\n*/ \nvoid luaD_call (lua_State *L, StkId func, int nResults) {\n if (++L->nCcalls >= LUAI_MAXCCALLS) {\n if (L->nCcalls == LUAI_MAXCCALLS)\n luaG_runerror(L, \"C stack overflow\");\n else if (L->nCcalls >= (LUAI_MAXCCALLS + (LUAI_MAXCCALLS>>3)))\n luaD_throw(L, LUA_ERRERR); /* error while handing stack error */\n }\n if (luaD_precall(L, func, nResults) == PCRLUA) /* is a Lua function? */\n luaV_execute(L, 1); /* call it */\n L->nCcalls--;\n luaC_checkGC(L);\n}\n\n\nstatic void resume (lua_State *L, void *ud) {\n StkId firstArg = cast(StkId, ud);\n CallInfo *ci = L->ci;\n if (L->status == 0) { /* start coroutine? */\n lua_assert(ci == L->base_ci && firstArg > L->base);\n if (luaD_precall(L, firstArg - 1, LUA_MULTRET) != PCRLUA)\n return;\n }\n else { /* resuming from previous yield */\n lua_assert(L->status == LUA_YIELD);\n L->status = 0;\n if (!f_isLua(ci)) { /* `common' yield? */\n /* finish interrupted execution of `OP_CALL' */\n lua_assert(GET_OPCODE(*((ci-1)->savedpc - 1)) == OP_CALL ||\n GET_OPCODE(*((ci-1)->savedpc - 1)) == OP_TAILCALL);\n if (luaD_poscall(L, firstArg)) /* complete it... */\n L->top = L->ci->top; /* and correct top if not multiple results */\n }\n else /* yielded inside a hook: just continue its execution */\n L->base = L->ci->base;\n }\n luaV_execute(L, cast_int(L->ci - L->base_ci));\n}\n\n\nstatic int resume_error (lua_State *L, const char *msg) {\n L->top = L->ci->base;\n setsvalue2s(L, L->top, luaS_new(L, msg));\n incr_top(L);\n lua_unlock(L);\n return LUA_ERRRUN;\n}\n\n\nLUA_API int lua_resume (lua_State *L, int nargs) {\n int status;\n lua_lock(L);\n if (L->status != LUA_YIELD && (L->status != 0 || L->ci != L->base_ci))\n return resume_error(L, \"cannot resume non-suspended coroutine\");\n if (L->nCcalls >= LUAI_MAXCCALLS)\n return resume_error(L, \"C stack overflow\");\n luai_userstateresume(L, nargs);\n lua_assert(L->errfunc == 0);\n L->baseCcalls = ++L->nCcalls;\n status = luaD_rawrunprotected(L, resume, L->top - nargs);\n if (status != 0) { /* error? */\n L->status = cast_byte(status); /* mark thread as `dead' */\n luaD_seterrorobj(L, status, L->top);\n L->ci->top = L->top;\n }\n else {\n lua_assert(L->nCcalls == L->baseCcalls);\n status = L->status;\n }\n --L->nCcalls;\n lua_unlock(L);\n return status;\n}\n\n\nLUA_API int lua_yield (lua_State *L, int nresults) {\n luai_userstateyield(L, nresults);\n lua_lock(L);\n if (L->nCcalls > L->baseCcalls)\n luaG_runerror(L, \"attempt to yield across metamethod/C-call boundary\");\n L->base = L->top - nresults; /* protect stack slots below */\n L->status = LUA_YIELD;\n lua_unlock(L);\n return -1;\n}\n\n\nint luaD_pcall (lua_State *L, Pfunc func, void *u,\n ptrdiff_t old_top, ptrdiff_t ef) {\n int status;\n unsigned short oldnCcalls = L->nCcalls;\n ptrdiff_t old_ci = saveci(L, L->ci);\n lu_byte old_allowhooks = L->allowhook;\n ptrdiff_t old_errfunc = L->errfunc;\n L->errfunc = ef;\n status = luaD_rawrunprotected(L, func, u);\n if (status != 0) { /* an error occurred? */\n StkId oldtop = restorestack(L, old_top);\n luaF_close(L, oldtop); /* close eventual pending closures */\n luaD_seterrorobj(L, status, oldtop);\n L->nCcalls = oldnCcalls;\n L->ci = restoreci(L, old_ci);\n L->base = L->ci->base;\n L->savedpc = L->ci->savedpc;\n L->allowhook = old_allowhooks;\n restore_stack_limit(L);\n }\n L->errfunc = old_errfunc;\n return status;\n}\n\n\n\n/*\n** Execute a protected parser.\n*/\nstruct SParser { /* data to `f_parser' */\n ZIO *z;\n Mbuffer buff; /* buffer to be used by the scanner */\n const char *name;\n};\n\nstatic void f_parser (lua_State *L, void *ud) {\n int i;\n Proto *tf;\n Closure *cl;\n struct SParser *p = cast(struct SParser *, ud);\n int c = luaZ_lookahead(p->z);\n luaC_checkGC(L);\n tf = ((c == LUA_SIGNATURE[0]) ? luaU_undump : luaY_parser)(L, p->z,\n &p->buff, p->name);\n cl = luaF_newLclosure(L, tf->nups, hvalue(gt(L)));\n cl->l.p = tf;\n for (i = 0; i < tf->nups; i++) /* initialize eventual upvalues */\n cl->l.upvals[i] = luaF_newupval(L);\n setclvalue(L, L->top, cl);\n incr_top(L);\n}\n\n\nint luaD_protectedparser (lua_State *L, ZIO *z, const char *name) {\n struct SParser p;\n int status;\n p.z = z; p.name = name;\n luaZ_initbuffer(L, &p.buff);\n status = luaD_pcall(L, f_parser, &p, savestack(L, L->top), L->errfunc);\n luaZ_freebuffer(L, &p.buff);\n return status;\n}\n\n\n" }, { "path": "deps/lua/src/ldo.h", "content": "/*\n** $Id: ldo.h,v 2.7.1.1 2007/12/27 13:02:25 roberto Exp $\n** Stack and Call structure of Lua\n** See Copyright Notice in lua.h\n*/\n\n#ifndef ldo_h\n#define ldo_h\n\n\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lzio.h\"\n\n\n#define luaD_checkstack(L,n)\t\\\n if ((char *)L->stack_last - (char *)L->top <= (n)*(int)sizeof(TValue)) \\\n luaD_growstack(L, n); \\\n else condhardstacktests(luaD_reallocstack(L, L->stacksize - EXTRA_STACK - 1));\n\n\n#define incr_top(L) {luaD_checkstack(L,1); L->top++;}\n\n#define savestack(L,p)\t\t((char *)(p) - (char *)L->stack)\n#define restorestack(L,n)\t((TValue *)((char *)L->stack + (n)))\n\n#define saveci(L,p)\t\t((char *)(p) - (char *)L->base_ci)\n#define restoreci(L,n)\t\t((CallInfo *)((char *)L->base_ci + (n)))\n\n\n/* results from luaD_precall */\n#define PCRLUA\t\t0\t/* initiated a call to a Lua function */\n#define PCRC\t\t1\t/* did a call to a C function */\n#define PCRYIELD\t2\t/* C funtion yielded */\n\n\n/* type of protected functions, to be ran by `runprotected' */\ntypedef void (*Pfunc) (lua_State *L, void *ud);\n\nLUAI_FUNC int luaD_protectedparser (lua_State *L, ZIO *z, const char *name);\nLUAI_FUNC void luaD_callhook (lua_State *L, int event, int line);\nLUAI_FUNC int luaD_precall (lua_State *L, StkId func, int nresults);\nLUAI_FUNC void luaD_call (lua_State *L, StkId func, int nResults);\nLUAI_FUNC int luaD_pcall (lua_State *L, Pfunc func, void *u,\n ptrdiff_t oldtop, ptrdiff_t ef);\nLUAI_FUNC int luaD_poscall (lua_State *L, StkId firstResult);\nLUAI_FUNC void luaD_reallocCI (lua_State *L, int newsize);\nLUAI_FUNC void luaD_reallocstack (lua_State *L, int newsize);\nLUAI_FUNC void luaD_growstack (lua_State *L, int n);\n\nLUAI_FUNC void luaD_throw (lua_State *L, int errcode);\nLUAI_FUNC int luaD_rawrunprotected (lua_State *L, Pfunc f, void *ud);\n\nLUAI_FUNC void luaD_seterrorobj (lua_State *L, int errcode, StkId oldtop);\n\n#endif\n\n" }, { "path": "deps/lua/src/ldump.c", "content": "/*\n** $Id: ldump.c,v 2.8.1.1 2007/12/27 13:02:25 roberto Exp $\n** save precompiled Lua chunks\n** See Copyright Notice in lua.h\n*/\n\n#include \n\n#define ldump_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lundump.h\"\n\ntypedef struct {\n lua_State* L;\n lua_Writer writer;\n void* data;\n int strip;\n int status;\n} DumpState;\n\n#define DumpMem(b,n,size,D)\tDumpBlock(b,(n)*(size),D)\n#define DumpVar(x,D)\t \tDumpMem(&x,1,sizeof(x),D)\n\nstatic void DumpBlock(const void* b, size_t size, DumpState* D)\n{\n if (D->status==0)\n {\n lua_unlock(D->L);\n D->status=(*D->writer)(D->L,b,size,D->data);\n lua_lock(D->L);\n }\n}\n\nstatic void DumpChar(int y, DumpState* D)\n{\n char x=(char)y;\n DumpVar(x,D);\n}\n\nstatic void DumpInt(int x, DumpState* D)\n{\n DumpVar(x,D);\n}\n\nstatic void DumpNumber(lua_Number x, DumpState* D)\n{\n DumpVar(x,D);\n}\n\nstatic void DumpVector(const void* b, int n, size_t size, DumpState* D)\n{\n DumpInt(n,D);\n DumpMem(b,n,size,D);\n}\n\nstatic void DumpString(const TString* s, DumpState* D)\n{\n if (s==NULL || getstr(s)==NULL)\n {\n size_t size=0;\n DumpVar(size,D);\n }\n else\n {\n size_t size=s->tsv.len+1;\t\t/* include trailing '\\0' */\n DumpVar(size,D);\n DumpBlock(getstr(s),size,D);\n }\n}\n\n#define DumpCode(f,D)\t DumpVector(f->code,f->sizecode,sizeof(Instruction),D)\n\nstatic void DumpFunction(const Proto* f, const TString* p, DumpState* D);\n\nstatic void DumpConstants(const Proto* f, DumpState* D)\n{\n int i,n=f->sizek;\n DumpInt(n,D);\n for (i=0; ik[i];\n DumpChar(ttype(o),D);\n switch (ttype(o))\n {\n case LUA_TNIL:\n\tbreak;\n case LUA_TBOOLEAN:\n\tDumpChar(bvalue(o),D);\n\tbreak;\n case LUA_TNUMBER:\n\tDumpNumber(nvalue(o),D);\n\tbreak;\n case LUA_TSTRING:\n\tDumpString(rawtsvalue(o),D);\n\tbreak;\n default:\n\tlua_assert(0);\t\t\t/* cannot happen */\n\tbreak;\n }\n }\n n=f->sizep;\n DumpInt(n,D);\n for (i=0; ip[i],f->source,D);\n}\n\nstatic void DumpDebug(const Proto* f, DumpState* D)\n{\n int i,n;\n n= (D->strip) ? 0 : f->sizelineinfo;\n DumpVector(f->lineinfo,n,sizeof(int),D);\n n= (D->strip) ? 0 : f->sizelocvars;\n DumpInt(n,D);\n for (i=0; ilocvars[i].varname,D);\n DumpInt(f->locvars[i].startpc,D);\n DumpInt(f->locvars[i].endpc,D);\n }\n n= (D->strip) ? 0 : f->sizeupvalues;\n DumpInt(n,D);\n for (i=0; iupvalues[i],D);\n}\n\nstatic void DumpFunction(const Proto* f, const TString* p, DumpState* D)\n{\n DumpString((f->source==p || D->strip) ? NULL : f->source,D);\n DumpInt(f->linedefined,D);\n DumpInt(f->lastlinedefined,D);\n DumpChar(f->nups,D);\n DumpChar(f->numparams,D);\n DumpChar(f->is_vararg,D);\n DumpChar(f->maxstacksize,D);\n DumpCode(f,D);\n DumpConstants(f,D);\n DumpDebug(f,D);\n}\n\nstatic void DumpHeader(DumpState* D)\n{\n char h[LUAC_HEADERSIZE];\n luaU_header(h);\n DumpBlock(h,LUAC_HEADERSIZE,D);\n}\n\n/*\n** dump Lua function as precompiled chunk\n*/\nint luaU_dump (lua_State* L, const Proto* f, lua_Writer w, void* data, int strip)\n{\n DumpState D;\n D.L=L;\n D.writer=w;\n D.data=data;\n D.strip=strip;\n D.status=0;\n DumpHeader(&D);\n DumpFunction(f,NULL,&D);\n return D.status;\n}\n" }, { "path": "deps/lua/src/lfunc.c", "content": "/*\n** $Id: lfunc.c,v 2.12.1.2 2007/12/28 14:58:43 roberto Exp $\n** Auxiliary functions to manipulate prototypes and closures\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lfunc_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n\n\n\nClosure *luaF_newCclosure (lua_State *L, int nelems, Table *e) {\n Closure *c = cast(Closure *, luaM_malloc(L, sizeCclosure(nelems)));\n luaC_link(L, obj2gco(c), LUA_TFUNCTION);\n c->c.isC = 1;\n c->c.env = e;\n c->c.nupvalues = cast_byte(nelems);\n return c;\n}\n\n\nClosure *luaF_newLclosure (lua_State *L, int nelems, Table *e) {\n Closure *c = cast(Closure *, luaM_malloc(L, sizeLclosure(nelems)));\n luaC_link(L, obj2gco(c), LUA_TFUNCTION);\n c->l.isC = 0;\n c->l.env = e;\n c->l.nupvalues = cast_byte(nelems);\n while (nelems--) c->l.upvals[nelems] = NULL;\n return c;\n}\n\n\nUpVal *luaF_newupval (lua_State *L) {\n UpVal *uv = luaM_new(L, UpVal);\n luaC_link(L, obj2gco(uv), LUA_TUPVAL);\n uv->v = &uv->u.value;\n setnilvalue(uv->v);\n return uv;\n}\n\n\nUpVal *luaF_findupval (lua_State *L, StkId level) {\n global_State *g = G(L);\n GCObject **pp = &L->openupval;\n UpVal *p;\n UpVal *uv;\n while (*pp != NULL && (p = ngcotouv(*pp))->v >= level) {\n lua_assert(p->v != &p->u.value);\n if (p->v == level) { /* found a corresponding upvalue? */\n if (isdead(g, obj2gco(p))) /* is it dead? */\n changewhite(obj2gco(p)); /* ressurect it */\n return p;\n }\n pp = &p->next;\n }\n uv = luaM_new(L, UpVal); /* not found: create a new one */\n uv->tt = LUA_TUPVAL;\n uv->marked = luaC_white(g);\n uv->v = level; /* current value lives in the stack */\n uv->next = *pp; /* chain it in the proper position */\n *pp = obj2gco(uv);\n uv->u.l.prev = &g->uvhead; /* double link it in `uvhead' list */\n uv->u.l.next = g->uvhead.u.l.next;\n uv->u.l.next->u.l.prev = uv;\n g->uvhead.u.l.next = uv;\n lua_assert(uv->u.l.next->u.l.prev == uv && uv->u.l.prev->u.l.next == uv);\n return uv;\n}\n\n\nstatic void unlinkupval (UpVal *uv) {\n lua_assert(uv->u.l.next->u.l.prev == uv && uv->u.l.prev->u.l.next == uv);\n uv->u.l.next->u.l.prev = uv->u.l.prev; /* remove from `uvhead' list */\n uv->u.l.prev->u.l.next = uv->u.l.next;\n}\n\n\nvoid luaF_freeupval (lua_State *L, UpVal *uv) {\n if (uv->v != &uv->u.value) /* is it open? */\n unlinkupval(uv); /* remove from open list */\n luaM_free(L, uv); /* free upvalue */\n}\n\n\nvoid luaF_close (lua_State *L, StkId level) {\n UpVal *uv;\n global_State *g = G(L);\n while (L->openupval != NULL && (uv = ngcotouv(L->openupval))->v >= level) {\n GCObject *o = obj2gco(uv);\n lua_assert(!isblack(o) && uv->v != &uv->u.value);\n L->openupval = uv->next; /* remove from `open' list */\n if (isdead(g, o))\n luaF_freeupval(L, uv); /* free upvalue */\n else {\n unlinkupval(uv);\n setobj(L, &uv->u.value, uv->v);\n uv->v = &uv->u.value; /* now current value lives here */\n luaC_linkupval(L, uv); /* link upvalue into `gcroot' list */\n }\n }\n}\n\n\nProto *luaF_newproto (lua_State *L) {\n Proto *f = luaM_new(L, Proto);\n luaC_link(L, obj2gco(f), LUA_TPROTO);\n f->k = NULL;\n f->sizek = 0;\n f->p = NULL;\n f->sizep = 0;\n f->code = NULL;\n f->sizecode = 0;\n f->sizelineinfo = 0;\n f->sizeupvalues = 0;\n f->nups = 0;\n f->upvalues = NULL;\n f->numparams = 0;\n f->is_vararg = 0;\n f->maxstacksize = 0;\n f->lineinfo = NULL;\n f->sizelocvars = 0;\n f->locvars = NULL;\n f->linedefined = 0;\n f->lastlinedefined = 0;\n f->source = NULL;\n return f;\n}\n\n\nvoid luaF_freeproto (lua_State *L, Proto *f) {\n luaM_freearray(L, f->code, f->sizecode, Instruction);\n luaM_freearray(L, f->p, f->sizep, Proto *);\n luaM_freearray(L, f->k, f->sizek, TValue);\n luaM_freearray(L, f->lineinfo, f->sizelineinfo, int);\n luaM_freearray(L, f->locvars, f->sizelocvars, struct LocVar);\n luaM_freearray(L, f->upvalues, f->sizeupvalues, TString *);\n luaM_free(L, f);\n}\n\n\nvoid luaF_freeclosure (lua_State *L, Closure *c) {\n int size = (c->c.isC) ? sizeCclosure(c->c.nupvalues) :\n sizeLclosure(c->l.nupvalues);\n luaM_freemem(L, c, size);\n}\n\n\n/*\n** Look for n-th local variable at line `line' in function `func'.\n** Returns NULL if not found.\n*/\nconst char *luaF_getlocalname (const Proto *f, int local_number, int pc) {\n int i;\n for (i = 0; isizelocvars && f->locvars[i].startpc <= pc; i++) {\n if (pc < f->locvars[i].endpc) { /* is variable active? */\n local_number--;\n if (local_number == 0)\n return getstr(f->locvars[i].varname);\n }\n }\n return NULL; /* not found */\n}\n\n" }, { "path": "deps/lua/src/lfunc.h", "content": "/*\n** $Id: lfunc.h,v 2.4.1.1 2007/12/27 13:02:25 roberto Exp $\n** Auxiliary functions to manipulate prototypes and closures\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lfunc_h\n#define lfunc_h\n\n\n#include \"lobject.h\"\n\n\n#define sizeCclosure(n)\t(cast(int, sizeof(CClosure)) + \\\n cast(int, sizeof(TValue)*((n)-1)))\n\n#define sizeLclosure(n)\t(cast(int, sizeof(LClosure)) + \\\n cast(int, sizeof(TValue *)*((n)-1)))\n\n\nLUAI_FUNC Proto *luaF_newproto (lua_State *L);\nLUAI_FUNC Closure *luaF_newCclosure (lua_State *L, int nelems, Table *e);\nLUAI_FUNC Closure *luaF_newLclosure (lua_State *L, int nelems, Table *e);\nLUAI_FUNC UpVal *luaF_newupval (lua_State *L);\nLUAI_FUNC UpVal *luaF_findupval (lua_State *L, StkId level);\nLUAI_FUNC void luaF_close (lua_State *L, StkId level);\nLUAI_FUNC void luaF_freeproto (lua_State *L, Proto *f);\nLUAI_FUNC void luaF_freeclosure (lua_State *L, Closure *c);\nLUAI_FUNC void luaF_freeupval (lua_State *L, UpVal *uv);\nLUAI_FUNC const char *luaF_getlocalname (const Proto *func, int local_number,\n int pc);\n\n\n#endif\n" }, { "path": "deps/lua/src/lgc.c", "content": "/*\n** $Id: lgc.c,v 2.38.1.2 2011/03/18 18:05:38 roberto Exp $\n** Garbage Collector\n** See Copyright Notice in lua.h\n*/\n\n#include \n\n#define lgc_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n\n\n#define GCSTEPSIZE\t1024u\n#define GCSWEEPMAX\t40\n#define GCSWEEPCOST\t10\n#define GCFINALIZECOST\t100\n\n\n#define maskmarks\tcast_byte(~(bitmask(BLACKBIT)|WHITEBITS))\n\n#define makewhite(g,x)\t\\\n ((x)->gch.marked = cast_byte(((x)->gch.marked & maskmarks) | luaC_white(g)))\n\n#define white2gray(x)\treset2bits((x)->gch.marked, WHITE0BIT, WHITE1BIT)\n#define black2gray(x)\tresetbit((x)->gch.marked, BLACKBIT)\n\n#define stringmark(s)\treset2bits((s)->tsv.marked, WHITE0BIT, WHITE1BIT)\n\n\n#define isfinalized(u)\t\ttestbit((u)->marked, FINALIZEDBIT)\n#define markfinalized(u)\tl_setbit((u)->marked, FINALIZEDBIT)\n\n\n#define KEYWEAK bitmask(KEYWEAKBIT)\n#define VALUEWEAK bitmask(VALUEWEAKBIT)\n\n\n\n#define markvalue(g,o) { checkconsistency(o); \\\n if (iscollectable(o) && iswhite(gcvalue(o))) reallymarkobject(g,gcvalue(o)); }\n\n#define markobject(g,t) { if (iswhite(obj2gco(t))) \\\n\t\treallymarkobject(g, obj2gco(t)); }\n\n\n#define setthreshold(g) (g->GCthreshold = (g->estimate/100) * g->gcpause)\n\n\nstatic void removeentry (Node *n) {\n lua_assert(ttisnil(gval(n)));\n if (iscollectable(gkey(n)))\n setttype(gkey(n), LUA_TDEADKEY); /* dead key; remove it */\n}\n\n\nstatic void reallymarkobject (global_State *g, GCObject *o) {\n lua_assert(iswhite(o) && !isdead(g, o));\n white2gray(o);\n switch (o->gch.tt) {\n case LUA_TSTRING: {\n return;\n }\n case LUA_TUSERDATA: {\n Table *mt = gco2u(o)->metatable;\n gray2black(o); /* udata are never gray */\n if (mt) markobject(g, mt);\n markobject(g, gco2u(o)->env);\n return;\n }\n case LUA_TUPVAL: {\n UpVal *uv = gco2uv(o);\n markvalue(g, uv->v);\n if (uv->v == &uv->u.value) /* closed? */\n gray2black(o); /* open upvalues are never black */\n return;\n }\n case LUA_TFUNCTION: {\n gco2cl(o)->c.gclist = g->gray;\n g->gray = o;\n break;\n }\n case LUA_TTABLE: {\n gco2h(o)->gclist = g->gray;\n g->gray = o;\n break;\n }\n case LUA_TTHREAD: {\n gco2th(o)->gclist = g->gray;\n g->gray = o;\n break;\n }\n case LUA_TPROTO: {\n gco2p(o)->gclist = g->gray;\n g->gray = o;\n break;\n }\n default: lua_assert(0);\n }\n}\n\n\nstatic void marktmu (global_State *g) {\n GCObject *u = g->tmudata;\n if (u) {\n do {\n u = u->gch.next;\n makewhite(g, u); /* may be marked, if left from previous GC */\n reallymarkobject(g, u);\n } while (u != g->tmudata);\n }\n}\n\n\n/* move `dead' udata that need finalization to list `tmudata' */\nsize_t luaC_separateudata (lua_State *L, int all) {\n global_State *g = G(L);\n size_t deadmem = 0;\n GCObject **p = &g->mainthread->next;\n GCObject *curr;\n while ((curr = *p) != NULL) {\n if (!(iswhite(curr) || all) || isfinalized(gco2u(curr)))\n p = &curr->gch.next; /* don't bother with them */\n else if (fasttm(L, gco2u(curr)->metatable, TM_GC) == NULL) {\n markfinalized(gco2u(curr)); /* don't need finalization */\n p = &curr->gch.next;\n }\n else { /* must call its gc method */\n deadmem += sizeudata(gco2u(curr));\n markfinalized(gco2u(curr));\n *p = curr->gch.next;\n /* link `curr' at the end of `tmudata' list */\n if (g->tmudata == NULL) /* list is empty? */\n g->tmudata = curr->gch.next = curr; /* creates a circular list */\n else {\n curr->gch.next = g->tmudata->gch.next;\n g->tmudata->gch.next = curr;\n g->tmudata = curr;\n }\n }\n }\n return deadmem;\n}\n\n\nstatic int traversetable (global_State *g, Table *h) {\n int i;\n int weakkey = 0;\n int weakvalue = 0;\n const TValue *mode;\n if (h->metatable)\n markobject(g, h->metatable);\n mode = gfasttm(g, h->metatable, TM_MODE);\n if (mode && ttisstring(mode)) { /* is there a weak mode? */\n weakkey = (strchr(svalue(mode), 'k') != NULL);\n weakvalue = (strchr(svalue(mode), 'v') != NULL);\n if (weakkey || weakvalue) { /* is really weak? */\n h->marked &= ~(KEYWEAK | VALUEWEAK); /* clear bits */\n h->marked |= cast_byte((weakkey << KEYWEAKBIT) |\n (weakvalue << VALUEWEAKBIT));\n h->gclist = g->weak; /* must be cleared after GC, ... */\n g->weak = obj2gco(h); /* ... so put in the appropriate list */\n }\n }\n if (weakkey && weakvalue) return 1;\n if (!weakvalue) {\n i = h->sizearray;\n while (i--)\n markvalue(g, &h->array[i]);\n }\n i = sizenode(h);\n while (i--) {\n Node *n = gnode(h, i);\n lua_assert(ttype(gkey(n)) != LUA_TDEADKEY || ttisnil(gval(n)));\n if (ttisnil(gval(n)))\n removeentry(n); /* remove empty entries */\n else {\n lua_assert(!ttisnil(gkey(n)));\n if (!weakkey) markvalue(g, gkey(n));\n if (!weakvalue) markvalue(g, gval(n));\n }\n }\n return weakkey || weakvalue;\n}\n\n\n/*\n** All marks are conditional because a GC may happen while the\n** prototype is still being created\n*/\nstatic void traverseproto (global_State *g, Proto *f) {\n int i;\n if (f->source) stringmark(f->source);\n for (i=0; isizek; i++) /* mark literals */\n markvalue(g, &f->k[i]);\n for (i=0; isizeupvalues; i++) { /* mark upvalue names */\n if (f->upvalues[i])\n stringmark(f->upvalues[i]);\n }\n for (i=0; isizep; i++) { /* mark nested protos */\n if (f->p[i])\n markobject(g, f->p[i]);\n }\n for (i=0; isizelocvars; i++) { /* mark local-variable names */\n if (f->locvars[i].varname)\n stringmark(f->locvars[i].varname);\n }\n}\n\n\n\nstatic void traverseclosure (global_State *g, Closure *cl) {\n markobject(g, cl->c.env);\n if (cl->c.isC) {\n int i;\n for (i=0; ic.nupvalues; i++) /* mark its upvalues */\n markvalue(g, &cl->c.upvalue[i]);\n }\n else {\n int i;\n lua_assert(cl->l.nupvalues == cl->l.p->nups);\n markobject(g, cl->l.p);\n for (i=0; il.nupvalues; i++) /* mark its upvalues */\n markobject(g, cl->l.upvals[i]);\n }\n}\n\n\nstatic void checkstacksizes (lua_State *L, StkId max) {\n int ci_used = cast_int(L->ci - L->base_ci); /* number of `ci' in use */\n int s_used = cast_int(max - L->stack); /* part of stack in use */\n if (L->size_ci > LUAI_MAXCALLS) /* handling overflow? */\n return; /* do not touch the stacks */\n if (4*ci_used < L->size_ci && 2*BASIC_CI_SIZE < L->size_ci)\n luaD_reallocCI(L, L->size_ci/2); /* still big enough... */\n condhardstacktests(luaD_reallocCI(L, ci_used + 1));\n if (4*s_used < L->stacksize &&\n 2*(BASIC_STACK_SIZE+EXTRA_STACK) < L->stacksize)\n luaD_reallocstack(L, L->stacksize/2); /* still big enough... */\n condhardstacktests(luaD_reallocstack(L, s_used));\n}\n\n\nstatic void traversestack (global_State *g, lua_State *l) {\n StkId o, lim;\n CallInfo *ci;\n markvalue(g, gt(l));\n lim = l->top;\n for (ci = l->base_ci; ci <= l->ci; ci++) {\n lua_assert(ci->top <= l->stack_last);\n if (lim < ci->top) lim = ci->top;\n }\n for (o = l->stack; o < l->top; o++)\n markvalue(g, o);\n for (; o <= lim; o++)\n setnilvalue(o);\n checkstacksizes(l, lim);\n}\n\n\n/*\n** traverse one gray object, turning it to black.\n** Returns `quantity' traversed.\n*/\nstatic l_mem propagatemark (global_State *g) {\n GCObject *o = g->gray;\n lua_assert(isgray(o));\n gray2black(o);\n switch (o->gch.tt) {\n case LUA_TTABLE: {\n Table *h = gco2h(o);\n g->gray = h->gclist;\n if (traversetable(g, h)) /* table is weak? */\n black2gray(o); /* keep it gray */\n return sizeof(Table) + sizeof(TValue) * h->sizearray +\n sizeof(Node) * sizenode(h);\n }\n case LUA_TFUNCTION: {\n Closure *cl = gco2cl(o);\n g->gray = cl->c.gclist;\n traverseclosure(g, cl);\n return (cl->c.isC) ? sizeCclosure(cl->c.nupvalues) :\n sizeLclosure(cl->l.nupvalues);\n }\n case LUA_TTHREAD: {\n lua_State *th = gco2th(o);\n g->gray = th->gclist;\n th->gclist = g->grayagain;\n g->grayagain = o;\n black2gray(o);\n traversestack(g, th);\n return sizeof(lua_State) + sizeof(TValue) * th->stacksize +\n sizeof(CallInfo) * th->size_ci;\n }\n case LUA_TPROTO: {\n Proto *p = gco2p(o);\n g->gray = p->gclist;\n traverseproto(g, p);\n return sizeof(Proto) + sizeof(Instruction) * p->sizecode +\n sizeof(Proto *) * p->sizep +\n sizeof(TValue) * p->sizek + \n sizeof(int) * p->sizelineinfo +\n sizeof(LocVar) * p->sizelocvars +\n sizeof(TString *) * p->sizeupvalues;\n }\n default: lua_assert(0); return 0;\n }\n}\n\n\nstatic size_t propagateall (global_State *g) {\n size_t m = 0;\n while (g->gray) m += propagatemark(g);\n return m;\n}\n\n\n/*\n** The next function tells whether a key or value can be cleared from\n** a weak table. Non-collectable objects are never removed from weak\n** tables. Strings behave as `values', so are never removed too. for\n** other objects: if really collected, cannot keep them; for userdata\n** being finalized, keep them in keys, but not in values\n*/\nstatic int iscleared (const TValue *o, int iskey) {\n if (!iscollectable(o)) return 0;\n if (ttisstring(o)) {\n stringmark(rawtsvalue(o)); /* strings are `values', so are never weak */\n return 0;\n }\n return iswhite(gcvalue(o)) ||\n (ttisuserdata(o) && (!iskey && isfinalized(uvalue(o))));\n}\n\n\n/*\n** clear collected entries from weaktables\n*/\nstatic void cleartable (GCObject *l) {\n while (l) {\n Table *h = gco2h(l);\n int i = h->sizearray;\n lua_assert(testbit(h->marked, VALUEWEAKBIT) ||\n testbit(h->marked, KEYWEAKBIT));\n if (testbit(h->marked, VALUEWEAKBIT)) {\n while (i--) {\n TValue *o = &h->array[i];\n if (iscleared(o, 0)) /* value was collected? */\n setnilvalue(o); /* remove value */\n }\n }\n i = sizenode(h);\n while (i--) {\n Node *n = gnode(h, i);\n if (!ttisnil(gval(n)) && /* non-empty entry? */\n (iscleared(key2tval(n), 1) || iscleared(gval(n), 0))) {\n setnilvalue(gval(n)); /* remove value ... */\n removeentry(n); /* remove entry from table */\n }\n }\n l = h->gclist;\n }\n}\n\n\nstatic void freeobj (lua_State *L, GCObject *o) {\n switch (o->gch.tt) {\n case LUA_TPROTO: luaF_freeproto(L, gco2p(o)); break;\n case LUA_TFUNCTION: luaF_freeclosure(L, gco2cl(o)); break;\n case LUA_TUPVAL: luaF_freeupval(L, gco2uv(o)); break;\n case LUA_TTABLE: luaH_free(L, gco2h(o)); break;\n case LUA_TTHREAD: {\n lua_assert(gco2th(o) != L && gco2th(o) != G(L)->mainthread);\n luaE_freethread(L, gco2th(o));\n break;\n }\n case LUA_TSTRING: {\n G(L)->strt.nuse--;\n luaM_freemem(L, o, sizestring(gco2ts(o)));\n break;\n }\n case LUA_TUSERDATA: {\n luaM_freemem(L, o, sizeudata(gco2u(o)));\n break;\n }\n default: lua_assert(0);\n }\n}\n\n\n\n#define sweepwholelist(L,p)\tsweeplist(L,p,MAX_LUMEM)\n\n\nstatic GCObject **sweeplist (lua_State *L, GCObject **p, lu_mem count) {\n GCObject *curr;\n global_State *g = G(L);\n int deadmask = otherwhite(g);\n while ((curr = *p) != NULL && count-- > 0) {\n if (curr->gch.tt == LUA_TTHREAD) /* sweep open upvalues of each thread */\n sweepwholelist(L, &gco2th(curr)->openupval);\n if ((curr->gch.marked ^ WHITEBITS) & deadmask) { /* not dead? */\n lua_assert(!isdead(g, curr) || testbit(curr->gch.marked, FIXEDBIT));\n makewhite(g, curr); /* make it white (for next cycle) */\n p = &curr->gch.next;\n }\n else { /* must erase `curr' */\n lua_assert(isdead(g, curr) || deadmask == bitmask(SFIXEDBIT));\n *p = curr->gch.next;\n if (curr == g->rootgc) /* is the first element of the list? */\n g->rootgc = curr->gch.next; /* adjust first */\n freeobj(L, curr);\n }\n }\n return p;\n}\n\n\nstatic void checkSizes (lua_State *L) {\n global_State *g = G(L);\n /* check size of string hash */\n if (g->strt.nuse < cast(lu_int32, g->strt.size/4) &&\n g->strt.size > MINSTRTABSIZE*2)\n luaS_resize(L, g->strt.size/2); /* table is too big */\n /* check size of buffer */\n if (luaZ_sizebuffer(&g->buff) > LUA_MINBUFFER*2) { /* buffer too big? */\n size_t newsize = luaZ_sizebuffer(&g->buff) / 2;\n luaZ_resizebuffer(L, &g->buff, newsize);\n }\n}\n\n\nstatic void GCTM (lua_State *L) {\n global_State *g = G(L);\n GCObject *o = g->tmudata->gch.next; /* get first element */\n Udata *udata = rawgco2u(o);\n const TValue *tm;\n /* remove udata from `tmudata' */\n if (o == g->tmudata) /* last element? */\n g->tmudata = NULL;\n else\n g->tmudata->gch.next = udata->uv.next;\n udata->uv.next = g->mainthread->next; /* return it to `root' list */\n g->mainthread->next = o;\n makewhite(g, o);\n tm = fasttm(L, udata->uv.metatable, TM_GC);\n if (tm != NULL) {\n lu_byte oldah = L->allowhook;\n lu_mem oldt = g->GCthreshold;\n L->allowhook = 0; /* stop debug hooks during GC tag method */\n g->GCthreshold = 2*g->totalbytes; /* avoid GC steps */\n setobj2s(L, L->top, tm);\n setuvalue(L, L->top+1, udata);\n L->top += 2;\n luaD_call(L, L->top - 2, 0);\n L->allowhook = oldah; /* restore hooks */\n g->GCthreshold = oldt; /* restore threshold */\n }\n}\n\n\n/*\n** Call all GC tag methods\n*/\nvoid luaC_callGCTM (lua_State *L) {\n while (G(L)->tmudata)\n GCTM(L);\n}\n\n\nvoid luaC_freeall (lua_State *L) {\n global_State *g = G(L);\n int i;\n g->currentwhite = WHITEBITS | bitmask(SFIXEDBIT); /* mask to collect all elements */\n sweepwholelist(L, &g->rootgc);\n for (i = 0; i < g->strt.size; i++) /* free all string lists */\n sweepwholelist(L, &g->strt.hash[i]);\n}\n\n\nstatic void markmt (global_State *g) {\n int i;\n for (i=0; imt[i]) markobject(g, g->mt[i]);\n}\n\n\n/* mark root set */\nstatic void markroot (lua_State *L) {\n global_State *g = G(L);\n g->gray = NULL;\n g->grayagain = NULL;\n g->weak = NULL;\n markobject(g, g->mainthread);\n /* make global table be traversed before main stack */\n markvalue(g, gt(g->mainthread));\n markvalue(g, registry(L));\n markmt(g);\n g->gcstate = GCSpropagate;\n}\n\n\nstatic void remarkupvals (global_State *g) {\n UpVal *uv;\n for (uv = g->uvhead.u.l.next; uv != &g->uvhead; uv = uv->u.l.next) {\n lua_assert(uv->u.l.next->u.l.prev == uv && uv->u.l.prev->u.l.next == uv);\n if (isgray(obj2gco(uv)))\n markvalue(g, uv->v);\n }\n}\n\n\nstatic void atomic (lua_State *L) {\n global_State *g = G(L);\n size_t udsize; /* total size of userdata to be finalized */\n /* remark occasional upvalues of (maybe) dead threads */\n remarkupvals(g);\n /* traverse objects cautch by write barrier and by 'remarkupvals' */\n propagateall(g);\n /* remark weak tables */\n g->gray = g->weak;\n g->weak = NULL;\n lua_assert(!iswhite(obj2gco(g->mainthread)));\n markobject(g, L); /* mark running thread */\n markmt(g); /* mark basic metatables (again) */\n propagateall(g);\n /* remark gray again */\n g->gray = g->grayagain;\n g->grayagain = NULL;\n propagateall(g);\n udsize = luaC_separateudata(L, 0); /* separate userdata to be finalized */\n marktmu(g); /* mark `preserved' userdata */\n udsize += propagateall(g); /* remark, to propagate `preserveness' */\n cleartable(g->weak); /* remove collected objects from weak tables */\n /* flip current white */\n g->currentwhite = cast_byte(otherwhite(g));\n g->sweepstrgc = 0;\n g->sweepgc = &g->rootgc;\n g->gcstate = GCSsweepstring;\n g->estimate = g->totalbytes - udsize; /* first estimate */\n}\n\n\nstatic l_mem singlestep (lua_State *L) {\n global_State *g = G(L);\n /*lua_checkmemory(L);*/\n switch (g->gcstate) {\n case GCSpause: {\n markroot(L); /* start a new collection */\n return 0;\n }\n case GCSpropagate: {\n if (g->gray)\n return propagatemark(g);\n else { /* no more `gray' objects */\n atomic(L); /* finish mark phase */\n return 0;\n }\n }\n case GCSsweepstring: {\n lu_mem old = g->totalbytes;\n sweepwholelist(L, &g->strt.hash[g->sweepstrgc++]);\n if (g->sweepstrgc >= g->strt.size) /* nothing more to sweep? */\n g->gcstate = GCSsweep; /* end sweep-string phase */\n lua_assert(old >= g->totalbytes);\n g->estimate -= old - g->totalbytes;\n return GCSWEEPCOST;\n }\n case GCSsweep: {\n lu_mem old = g->totalbytes;\n g->sweepgc = sweeplist(L, g->sweepgc, GCSWEEPMAX);\n if (*g->sweepgc == NULL) { /* nothing more to sweep? */\n checkSizes(L);\n g->gcstate = GCSfinalize; /* end sweep phase */\n }\n lua_assert(old >= g->totalbytes);\n g->estimate -= old - g->totalbytes;\n return GCSWEEPMAX*GCSWEEPCOST;\n }\n case GCSfinalize: {\n if (g->tmudata) {\n GCTM(L);\n if (g->estimate > GCFINALIZECOST)\n g->estimate -= GCFINALIZECOST;\n return GCFINALIZECOST;\n }\n else {\n g->gcstate = GCSpause; /* end collection */\n g->gcdept = 0;\n return 0;\n }\n }\n default: lua_assert(0); return 0;\n }\n}\n\n\nvoid luaC_step (lua_State *L) {\n global_State *g = G(L);\n l_mem lim = (GCSTEPSIZE/100) * g->gcstepmul;\n if (lim == 0)\n lim = (MAX_LUMEM-1)/2; /* no limit */\n g->gcdept += g->totalbytes - g->GCthreshold;\n do {\n lim -= singlestep(L);\n if (g->gcstate == GCSpause)\n break;\n } while (lim > 0);\n if (g->gcstate != GCSpause) {\n if (g->gcdept < GCSTEPSIZE)\n g->GCthreshold = g->totalbytes + GCSTEPSIZE; /* - lim/g->gcstepmul;*/\n else {\n g->gcdept -= GCSTEPSIZE;\n g->GCthreshold = g->totalbytes;\n }\n }\n else {\n setthreshold(g);\n }\n}\n\n\nvoid luaC_fullgc (lua_State *L) {\n global_State *g = G(L);\n if (g->gcstate <= GCSpropagate) {\n /* reset sweep marks to sweep all elements (returning them to white) */\n g->sweepstrgc = 0;\n g->sweepgc = &g->rootgc;\n /* reset other collector lists */\n g->gray = NULL;\n g->grayagain = NULL;\n g->weak = NULL;\n g->gcstate = GCSsweepstring;\n }\n lua_assert(g->gcstate != GCSpause && g->gcstate != GCSpropagate);\n /* finish any pending sweep phase */\n while (g->gcstate != GCSfinalize) {\n lua_assert(g->gcstate == GCSsweepstring || g->gcstate == GCSsweep);\n singlestep(L);\n }\n markroot(L);\n while (g->gcstate != GCSpause) {\n singlestep(L);\n }\n setthreshold(g);\n}\n\n\nvoid luaC_barrierf (lua_State *L, GCObject *o, GCObject *v) {\n global_State *g = G(L);\n lua_assert(isblack(o) && iswhite(v) && !isdead(g, v) && !isdead(g, o));\n lua_assert(g->gcstate != GCSfinalize && g->gcstate != GCSpause);\n lua_assert(ttype(&o->gch) != LUA_TTABLE);\n /* must keep invariant? */\n if (g->gcstate == GCSpropagate)\n reallymarkobject(g, v); /* restore invariant */\n else /* don't mind */\n makewhite(g, o); /* mark as white just to avoid other barriers */\n}\n\n\nvoid luaC_barrierback (lua_State *L, Table *t) {\n global_State *g = G(L);\n GCObject *o = obj2gco(t);\n lua_assert(isblack(o) && !isdead(g, o));\n lua_assert(g->gcstate != GCSfinalize && g->gcstate != GCSpause);\n black2gray(o); /* make table gray (again) */\n t->gclist = g->grayagain;\n g->grayagain = o;\n}\n\n\nvoid luaC_link (lua_State *L, GCObject *o, lu_byte tt) {\n global_State *g = G(L);\n o->gch.next = g->rootgc;\n g->rootgc = o;\n o->gch.marked = luaC_white(g);\n o->gch.tt = tt;\n}\n\n\nvoid luaC_linkupval (lua_State *L, UpVal *uv) {\n global_State *g = G(L);\n GCObject *o = obj2gco(uv);\n o->gch.next = g->rootgc; /* link upvalue into `rootgc' list */\n g->rootgc = o;\n if (isgray(o)) { \n if (g->gcstate == GCSpropagate) {\n gray2black(o); /* closed upvalues need barrier */\n luaC_barrier(L, uv, uv->v);\n }\n else { /* sweep phase: sweep it (turning it into white) */\n makewhite(g, o);\n lua_assert(g->gcstate != GCSfinalize && g->gcstate != GCSpause);\n }\n }\n}\n\n" }, { "path": "deps/lua/src/lgc.h", "content": "/*\n** $Id: lgc.h,v 2.15.1.1 2007/12/27 13:02:25 roberto Exp $\n** Garbage Collector\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lgc_h\n#define lgc_h\n\n\n#include \"lobject.h\"\n\n\n/*\n** Possible states of the Garbage Collector\n*/\n#define GCSpause\t0\n#define GCSpropagate\t1\n#define GCSsweepstring\t2\n#define GCSsweep\t3\n#define GCSfinalize\t4\n\n\n/*\n** some userful bit tricks\n*/\n#define resetbits(x,m)\t((x) &= cast(lu_byte, ~(m)))\n#define setbits(x,m)\t((x) |= (m))\n#define testbits(x,m)\t((x) & (m))\n#define bitmask(b)\t(1<<(b))\n#define bit2mask(b1,b2)\t(bitmask(b1) | bitmask(b2))\n#define l_setbit(x,b)\tsetbits(x, bitmask(b))\n#define resetbit(x,b)\tresetbits(x, bitmask(b))\n#define testbit(x,b)\ttestbits(x, bitmask(b))\n#define set2bits(x,b1,b2)\tsetbits(x, (bit2mask(b1, b2)))\n#define reset2bits(x,b1,b2)\tresetbits(x, (bit2mask(b1, b2)))\n#define test2bits(x,b1,b2)\ttestbits(x, (bit2mask(b1, b2)))\n\n\n\n/*\n** Layout for bit use in `marked' field:\n** bit 0 - object is white (type 0)\n** bit 1 - object is white (type 1)\n** bit 2 - object is black\n** bit 3 - for userdata: has been finalized\n** bit 3 - for tables: has weak keys\n** bit 4 - for tables: has weak values\n** bit 5 - object is fixed (should not be collected)\n** bit 6 - object is \"super\" fixed (only the main thread)\n*/\n\n\n#define WHITE0BIT\t0\n#define WHITE1BIT\t1\n#define BLACKBIT\t2\n#define FINALIZEDBIT\t3\n#define KEYWEAKBIT\t3\n#define VALUEWEAKBIT\t4\n#define FIXEDBIT\t5\n#define SFIXEDBIT\t6\n#define WHITEBITS\tbit2mask(WHITE0BIT, WHITE1BIT)\n\n\n#define iswhite(x) test2bits((x)->gch.marked, WHITE0BIT, WHITE1BIT)\n#define isblack(x) testbit((x)->gch.marked, BLACKBIT)\n#define isgray(x)\t(!isblack(x) && !iswhite(x))\n\n#define otherwhite(g)\t(g->currentwhite ^ WHITEBITS)\n#define isdead(g,v)\t((v)->gch.marked & otherwhite(g) & WHITEBITS)\n\n#define changewhite(x)\t((x)->gch.marked ^= WHITEBITS)\n#define gray2black(x)\tl_setbit((x)->gch.marked, BLACKBIT)\n\n#define valiswhite(x)\t(iscollectable(x) && iswhite(gcvalue(x)))\n\n#define luaC_white(g)\tcast(lu_byte, (g)->currentwhite & WHITEBITS)\n\n\n#define luaC_checkGC(L) { \\\n condhardstacktests(luaD_reallocstack(L, L->stacksize - EXTRA_STACK - 1)); \\\n if (G(L)->totalbytes >= G(L)->GCthreshold) \\\n\tluaC_step(L); }\n\n\n#define luaC_barrier(L,p,v) { if (valiswhite(v) && isblack(obj2gco(p))) \\\n\tluaC_barrierf(L,obj2gco(p),gcvalue(v)); }\n\n#define luaC_barriert(L,t,v) { if (valiswhite(v) && isblack(obj2gco(t))) \\\n\tluaC_barrierback(L,t); }\n\n#define luaC_objbarrier(L,p,o) \\\n\t{ if (iswhite(obj2gco(o)) && isblack(obj2gco(p))) \\\n\t\tluaC_barrierf(L,obj2gco(p),obj2gco(o)); }\n\n#define luaC_objbarriert(L,t,o) \\\n { if (iswhite(obj2gco(o)) && isblack(obj2gco(t))) luaC_barrierback(L,t); }\n\nLUAI_FUNC size_t luaC_separateudata (lua_State *L, int all);\nLUAI_FUNC void luaC_callGCTM (lua_State *L);\nLUAI_FUNC void luaC_freeall (lua_State *L);\nLUAI_FUNC void luaC_step (lua_State *L);\nLUAI_FUNC void luaC_fullgc (lua_State *L);\nLUAI_FUNC void luaC_link (lua_State *L, GCObject *o, lu_byte tt);\nLUAI_FUNC void luaC_linkupval (lua_State *L, UpVal *uv);\nLUAI_FUNC void luaC_barrierf (lua_State *L, GCObject *o, GCObject *v);\nLUAI_FUNC void luaC_barrierback (lua_State *L, Table *t);\n\n\n#endif\n" }, { "path": "deps/lua/src/linit.c", "content": "/*\n** $Id: linit.c,v 1.14.1.1 2007/12/27 13:02:25 roberto Exp $\n** Initialization of libraries for lua.c\n** See Copyright Notice in lua.h\n*/\n\n\n#define linit_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lualib.h\"\n#include \"lauxlib.h\"\n\n\nstatic const luaL_Reg lualibs[] = {\n {\"\", luaopen_base},\n {LUA_LOADLIBNAME, luaopen_package},\n {LUA_TABLIBNAME, luaopen_table},\n {LUA_IOLIBNAME, luaopen_io},\n {LUA_OSLIBNAME, luaopen_os},\n {LUA_STRLIBNAME, luaopen_string},\n {LUA_MATHLIBNAME, luaopen_math},\n {LUA_DBLIBNAME, luaopen_debug},\n {NULL, NULL}\n};\n\n\nLUALIB_API void luaL_openlibs (lua_State *L) {\n const luaL_Reg *lib = lualibs;\n for (; lib->func; lib++) {\n lua_pushcfunction(L, lib->func);\n lua_pushstring(L, lib->name);\n lua_call(L, 1, 0);\n }\n}\n\n" }, { "path": "deps/lua/src/liolib.c", "content": "/*\n** $Id: liolib.c,v 2.73.1.4 2010/05/14 15:33:51 roberto Exp $\n** Standard I/O (and system) library\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n\n#define liolib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n\n#define IO_INPUT\t1\n#define IO_OUTPUT\t2\n\n\nstatic const char *const fnames[] = {\"input\", \"output\"};\n\n\nstatic int pushresult (lua_State *L, int i, const char *filename) {\n int en = errno; /* calls to Lua API may change this value */\n if (i) {\n lua_pushboolean(L, 1);\n return 1;\n }\n else {\n lua_pushnil(L);\n if (filename)\n lua_pushfstring(L, \"%s: %s\", filename, strerror(en));\n else\n lua_pushfstring(L, \"%s\", strerror(en));\n lua_pushinteger(L, en);\n return 3;\n }\n}\n\n\nstatic void fileerror (lua_State *L, int arg, const char *filename) {\n lua_pushfstring(L, \"%s: %s\", filename, strerror(errno));\n luaL_argerror(L, arg, lua_tostring(L, -1));\n}\n\n\n#define tofilep(L)\t((FILE **)luaL_checkudata(L, 1, LUA_FILEHANDLE))\n\n\nstatic int io_type (lua_State *L) {\n void *ud;\n luaL_checkany(L, 1);\n ud = lua_touserdata(L, 1);\n lua_getfield(L, LUA_REGISTRYINDEX, LUA_FILEHANDLE);\n if (ud == NULL || !lua_getmetatable(L, 1) || !lua_rawequal(L, -2, -1))\n lua_pushnil(L); /* not a file */\n else if (*((FILE **)ud) == NULL)\n lua_pushliteral(L, \"closed file\");\n else\n lua_pushliteral(L, \"file\");\n return 1;\n}\n\n\nstatic FILE *tofile (lua_State *L) {\n FILE **f = tofilep(L);\n if (*f == NULL)\n luaL_error(L, \"attempt to use a closed file\");\n return *f;\n}\n\n\n\n/*\n** When creating file handles, always creates a `closed' file handle\n** before opening the actual file; so, if there is a memory error, the\n** file is not left opened.\n*/\nstatic FILE **newfile (lua_State *L) {\n FILE **pf = (FILE **)lua_newuserdata(L, sizeof(FILE *));\n *pf = NULL; /* file handle is currently `closed' */\n luaL_getmetatable(L, LUA_FILEHANDLE);\n lua_setmetatable(L, -2);\n return pf;\n}\n\n\n/*\n** function to (not) close the standard files stdin, stdout, and stderr\n*/\nstatic int io_noclose (lua_State *L) {\n lua_pushnil(L);\n lua_pushliteral(L, \"cannot close standard file\");\n return 2;\n}\n\n\n/*\n** function to close 'popen' files\n*/\nstatic int io_pclose (lua_State *L) {\n FILE **p = tofilep(L);\n int ok = lua_pclose(L, *p);\n *p = NULL;\n return pushresult(L, ok, NULL);\n}\n\n\n/*\n** function to close regular files\n*/\nstatic int io_fclose (lua_State *L) {\n FILE **p = tofilep(L);\n int ok = (fclose(*p) == 0);\n *p = NULL;\n return pushresult(L, ok, NULL);\n}\n\n\nstatic int aux_close (lua_State *L) {\n lua_getfenv(L, 1);\n lua_getfield(L, -1, \"__close\");\n return (lua_tocfunction(L, -1))(L);\n}\n\n\nstatic int io_close (lua_State *L) {\n if (lua_isnone(L, 1))\n lua_rawgeti(L, LUA_ENVIRONINDEX, IO_OUTPUT);\n tofile(L); /* make sure argument is a file */\n return aux_close(L);\n}\n\n\nstatic int io_gc (lua_State *L) {\n FILE *f = *tofilep(L);\n /* ignore closed files */\n if (f != NULL)\n aux_close(L);\n return 0;\n}\n\n\nstatic int io_tostring (lua_State *L) {\n FILE *f = *tofilep(L);\n if (f == NULL)\n lua_pushliteral(L, \"file (closed)\");\n else\n lua_pushfstring(L, \"file (%p)\", f);\n return 1;\n}\n\n\nstatic int io_open (lua_State *L) {\n const char *filename = luaL_checkstring(L, 1);\n const char *mode = luaL_optstring(L, 2, \"r\");\n FILE **pf = newfile(L);\n *pf = fopen(filename, mode);\n return (*pf == NULL) ? pushresult(L, 0, filename) : 1;\n}\n\n\n/*\n** this function has a separated environment, which defines the\n** correct __close for 'popen' files\n*/\nstatic int io_popen (lua_State *L) {\n const char *filename = luaL_checkstring(L, 1);\n const char *mode = luaL_optstring(L, 2, \"r\");\n FILE **pf = newfile(L);\n *pf = lua_popen(L, filename, mode);\n return (*pf == NULL) ? pushresult(L, 0, filename) : 1;\n}\n\n\nstatic int io_tmpfile (lua_State *L) {\n FILE **pf = newfile(L);\n *pf = tmpfile();\n return (*pf == NULL) ? pushresult(L, 0, NULL) : 1;\n}\n\n\nstatic FILE *getiofile (lua_State *L, int findex) {\n FILE *f;\n lua_rawgeti(L, LUA_ENVIRONINDEX, findex);\n f = *(FILE **)lua_touserdata(L, -1);\n if (f == NULL)\n luaL_error(L, \"standard %s file is closed\", fnames[findex - 1]);\n return f;\n}\n\n\nstatic int g_iofile (lua_State *L, int f, const char *mode) {\n if (!lua_isnoneornil(L, 1)) {\n const char *filename = lua_tostring(L, 1);\n if (filename) {\n FILE **pf = newfile(L);\n *pf = fopen(filename, mode);\n if (*pf == NULL)\n fileerror(L, 1, filename);\n }\n else {\n tofile(L); /* check that it's a valid file handle */\n lua_pushvalue(L, 1);\n }\n lua_rawseti(L, LUA_ENVIRONINDEX, f);\n }\n /* return current value */\n lua_rawgeti(L, LUA_ENVIRONINDEX, f);\n return 1;\n}\n\n\nstatic int io_input (lua_State *L) {\n return g_iofile(L, IO_INPUT, \"r\");\n}\n\n\nstatic int io_output (lua_State *L) {\n return g_iofile(L, IO_OUTPUT, \"w\");\n}\n\n\nstatic int io_readline (lua_State *L);\n\n\nstatic void aux_lines (lua_State *L, int idx, int toclose) {\n lua_pushvalue(L, idx);\n lua_pushboolean(L, toclose); /* close/not close file when finished */\n lua_pushcclosure(L, io_readline, 2);\n}\n\n\nstatic int f_lines (lua_State *L) {\n tofile(L); /* check that it's a valid file handle */\n aux_lines(L, 1, 0);\n return 1;\n}\n\n\nstatic int io_lines (lua_State *L) {\n if (lua_isnoneornil(L, 1)) { /* no arguments? */\n /* will iterate over default input */\n lua_rawgeti(L, LUA_ENVIRONINDEX, IO_INPUT);\n return f_lines(L);\n }\n else {\n const char *filename = luaL_checkstring(L, 1);\n FILE **pf = newfile(L);\n *pf = fopen(filename, \"r\");\n if (*pf == NULL)\n fileerror(L, 1, filename);\n aux_lines(L, lua_gettop(L), 1);\n return 1;\n }\n}\n\n\n/*\n** {======================================================\n** READ\n** =======================================================\n*/\n\n\nstatic int read_number (lua_State *L, FILE *f) {\n lua_Number d;\n if (fscanf(f, LUA_NUMBER_SCAN, &d) == 1) {\n lua_pushnumber(L, d);\n return 1;\n }\n else {\n lua_pushnil(L); /* \"result\" to be removed */\n return 0; /* read fails */\n }\n}\n\n\nstatic int test_eof (lua_State *L, FILE *f) {\n int c = getc(f);\n ungetc(c, f);\n lua_pushlstring(L, NULL, 0);\n return (c != EOF);\n}\n\n\nstatic int read_line (lua_State *L, FILE *f) {\n luaL_Buffer b;\n luaL_buffinit(L, &b);\n for (;;) {\n size_t l;\n char *p = luaL_prepbuffer(&b);\n if (fgets(p, LUAL_BUFFERSIZE, f) == NULL) { /* eof? */\n luaL_pushresult(&b); /* close buffer */\n return (lua_objlen(L, -1) > 0); /* check whether read something */\n }\n l = strlen(p);\n if (l == 0 || p[l-1] != '\\n')\n luaL_addsize(&b, l);\n else {\n luaL_addsize(&b, l - 1); /* do not include `eol' */\n luaL_pushresult(&b); /* close buffer */\n return 1; /* read at least an `eol' */\n }\n }\n}\n\n\nstatic int read_chars (lua_State *L, FILE *f, size_t n) {\n size_t rlen; /* how much to read */\n size_t nr; /* number of chars actually read */\n luaL_Buffer b;\n luaL_buffinit(L, &b);\n rlen = LUAL_BUFFERSIZE; /* try to read that much each time */\n do {\n char *p = luaL_prepbuffer(&b);\n if (rlen > n) rlen = n; /* cannot read more than asked */\n nr = fread(p, sizeof(char), rlen, f);\n luaL_addsize(&b, nr);\n n -= nr; /* still have to read `n' chars */\n } while (n > 0 && nr == rlen); /* until end of count or eof */\n luaL_pushresult(&b); /* close buffer */\n return (n == 0 || lua_objlen(L, -1) > 0);\n}\n\n\nstatic int g_read (lua_State *L, FILE *f, int first) {\n int nargs = lua_gettop(L) - 1;\n int success;\n int n;\n clearerr(f);\n if (nargs == 0) { /* no arguments? */\n success = read_line(L, f);\n n = first+1; /* to return 1 result */\n }\n else { /* ensure stack space for all results and for auxlib's buffer */\n luaL_checkstack(L, nargs+LUA_MINSTACK, \"too many arguments\");\n success = 1;\n for (n = first; nargs-- && success; n++) {\n if (lua_type(L, n) == LUA_TNUMBER) {\n size_t l = (size_t)lua_tointeger(L, n);\n success = (l == 0) ? test_eof(L, f) : read_chars(L, f, l);\n }\n else {\n const char *p = lua_tostring(L, n);\n luaL_argcheck(L, p && p[0] == '*', n, \"invalid option\");\n switch (p[1]) {\n case 'n': /* number */\n success = read_number(L, f);\n break;\n case 'l': /* line */\n success = read_line(L, f);\n break;\n case 'a': /* file */\n read_chars(L, f, ~((size_t)0)); /* read MAX_SIZE_T chars */\n success = 1; /* always success */\n break;\n default:\n return luaL_argerror(L, n, \"invalid format\");\n }\n }\n }\n }\n if (ferror(f))\n return pushresult(L, 0, NULL);\n if (!success) {\n lua_pop(L, 1); /* remove last result */\n lua_pushnil(L); /* push nil instead */\n }\n return n - first;\n}\n\n\nstatic int io_read (lua_State *L) {\n return g_read(L, getiofile(L, IO_INPUT), 1);\n}\n\n\nstatic int f_read (lua_State *L) {\n return g_read(L, tofile(L), 2);\n}\n\n\nstatic int io_readline (lua_State *L) {\n FILE *f = *(FILE **)lua_touserdata(L, lua_upvalueindex(1));\n int sucess;\n if (f == NULL) /* file is already closed? */\n luaL_error(L, \"file is already closed\");\n sucess = read_line(L, f);\n if (ferror(f))\n return luaL_error(L, \"%s\", strerror(errno));\n if (sucess) return 1;\n else { /* EOF */\n if (lua_toboolean(L, lua_upvalueindex(2))) { /* generator created file? */\n lua_settop(L, 0);\n lua_pushvalue(L, lua_upvalueindex(1));\n aux_close(L); /* close it */\n }\n return 0;\n }\n}\n\n/* }====================================================== */\n\n\nstatic int g_write (lua_State *L, FILE *f, int arg) {\n int nargs = lua_gettop(L) - 1;\n int status = 1;\n for (; nargs--; arg++) {\n if (lua_type(L, arg) == LUA_TNUMBER) {\n /* optimization: could be done exactly as for strings */\n status = status &&\n fprintf(f, LUA_NUMBER_FMT, lua_tonumber(L, arg)) > 0;\n }\n else {\n size_t l;\n const char *s = luaL_checklstring(L, arg, &l);\n status = status && (fwrite(s, sizeof(char), l, f) == l);\n }\n }\n return pushresult(L, status, NULL);\n}\n\n\nstatic int io_write (lua_State *L) {\n return g_write(L, getiofile(L, IO_OUTPUT), 1);\n}\n\n\nstatic int f_write (lua_State *L) {\n return g_write(L, tofile(L), 2);\n}\n\n\nstatic int f_seek (lua_State *L) {\n static const int mode[] = {SEEK_SET, SEEK_CUR, SEEK_END};\n static const char *const modenames[] = {\"set\", \"cur\", \"end\", NULL};\n FILE *f = tofile(L);\n int op = luaL_checkoption(L, 2, \"cur\", modenames);\n long offset = luaL_optlong(L, 3, 0);\n op = fseek(f, offset, mode[op]);\n if (op)\n return pushresult(L, 0, NULL); /* error */\n else {\n lua_pushinteger(L, ftell(f));\n return 1;\n }\n}\n\n\nstatic int f_setvbuf (lua_State *L) {\n static const int mode[] = {_IONBF, _IOFBF, _IOLBF};\n static const char *const modenames[] = {\"no\", \"full\", \"line\", NULL};\n FILE *f = tofile(L);\n int op = luaL_checkoption(L, 2, NULL, modenames);\n lua_Integer sz = luaL_optinteger(L, 3, LUAL_BUFFERSIZE);\n int res = setvbuf(f, NULL, mode[op], sz);\n return pushresult(L, res == 0, NULL);\n}\n\n\n\nstatic int io_flush (lua_State *L) {\n return pushresult(L, fflush(getiofile(L, IO_OUTPUT)) == 0, NULL);\n}\n\n\nstatic int f_flush (lua_State *L) {\n return pushresult(L, fflush(tofile(L)) == 0, NULL);\n}\n\n\nstatic const luaL_Reg iolib[] = {\n {\"close\", io_close},\n {\"flush\", io_flush},\n {\"input\", io_input},\n {\"lines\", io_lines},\n {\"open\", io_open},\n {\"output\", io_output},\n {\"popen\", io_popen},\n {\"read\", io_read},\n {\"tmpfile\", io_tmpfile},\n {\"type\", io_type},\n {\"write\", io_write},\n {NULL, NULL}\n};\n\n\nstatic const luaL_Reg flib[] = {\n {\"close\", io_close},\n {\"flush\", f_flush},\n {\"lines\", f_lines},\n {\"read\", f_read},\n {\"seek\", f_seek},\n {\"setvbuf\", f_setvbuf},\n {\"write\", f_write},\n {\"__gc\", io_gc},\n {\"__tostring\", io_tostring},\n {NULL, NULL}\n};\n\n\nstatic void createmeta (lua_State *L) {\n luaL_newmetatable(L, LUA_FILEHANDLE); /* create metatable for file handles */\n lua_pushvalue(L, -1); /* push metatable */\n lua_setfield(L, -2, \"__index\"); /* metatable.__index = metatable */\n luaL_register(L, NULL, flib); /* file methods */\n}\n\n\nstatic void createstdfile (lua_State *L, FILE *f, int k, const char *fname) {\n *newfile(L) = f;\n if (k > 0) {\n lua_pushvalue(L, -1);\n lua_rawseti(L, LUA_ENVIRONINDEX, k);\n }\n lua_pushvalue(L, -2); /* copy environment */\n lua_setfenv(L, -2); /* set it */\n lua_setfield(L, -3, fname);\n}\n\n\nstatic void newfenv (lua_State *L, lua_CFunction cls) {\n lua_createtable(L, 0, 1);\n lua_pushcfunction(L, cls);\n lua_setfield(L, -2, \"__close\");\n}\n\n\nLUALIB_API int luaopen_io (lua_State *L) {\n createmeta(L);\n /* create (private) environment (with fields IO_INPUT, IO_OUTPUT, __close) */\n newfenv(L, io_fclose);\n lua_replace(L, LUA_ENVIRONINDEX);\n /* open library */\n luaL_register(L, LUA_IOLIBNAME, iolib);\n /* create (and set) default files */\n newfenv(L, io_noclose); /* close function for default files */\n createstdfile(L, stdin, IO_INPUT, \"stdin\");\n createstdfile(L, stdout, IO_OUTPUT, \"stdout\");\n createstdfile(L, stderr, 0, \"stderr\");\n lua_pop(L, 1); /* pop environment for default files */\n lua_getfield(L, -1, \"popen\");\n newfenv(L, io_pclose); /* create environment for 'popen' */\n lua_setfenv(L, -2); /* set fenv for 'popen' */\n lua_pop(L, 1); /* pop 'popen' */\n return 1;\n}\n\n" }, { "path": "deps/lua/src/llex.c", "content": "/*\n** $Id: llex.c,v 2.20.1.2 2009/11/23 14:58:22 roberto Exp $\n** Lexical Analyzer\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n\n#define llex_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldo.h\"\n#include \"llex.h\"\n#include \"lobject.h\"\n#include \"lparser.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"lzio.h\"\n\n\n\n#define next(ls) (ls->current = zgetc(ls->z))\n\n\n\n\n#define currIsNewline(ls)\t(ls->current == '\\n' || ls->current == '\\r')\n\n\n/* ORDER RESERVED */\nconst char *const luaX_tokens [] = {\n \"and\", \"break\", \"do\", \"else\", \"elseif\",\n \"end\", \"false\", \"for\", \"function\", \"if\",\n \"in\", \"local\", \"nil\", \"not\", \"or\", \"repeat\",\n \"return\", \"then\", \"true\", \"until\", \"while\",\n \"..\", \"...\", \"==\", \">=\", \"<=\", \"~=\",\n \"\", \"\", \"\", \"\",\n NULL\n};\n\n\n#define save_and_next(ls) (save(ls, ls->current), next(ls))\n\n\nstatic void save (LexState *ls, int c) {\n Mbuffer *b = ls->buff;\n if (b->n + 1 > b->buffsize) {\n size_t newsize;\n if (b->buffsize >= MAX_SIZET/2)\n luaX_lexerror(ls, \"lexical element too long\", 0);\n newsize = b->buffsize * 2;\n luaZ_resizebuffer(ls->L, b, newsize);\n }\n b->buffer[b->n++] = cast(char, c);\n}\n\n\nvoid luaX_init (lua_State *L) {\n int i;\n for (i=0; itsv.reserved = cast_byte(i+1); /* reserved word */\n }\n}\n\n\n#define MAXSRC 80\n\n\nconst char *luaX_token2str (LexState *ls, int token) {\n if (token < FIRST_RESERVED) {\n lua_assert(token == cast(unsigned char, token));\n return (iscntrl(token)) ? luaO_pushfstring(ls->L, \"char(%d)\", token) :\n luaO_pushfstring(ls->L, \"%c\", token);\n }\n else\n return luaX_tokens[token-FIRST_RESERVED];\n}\n\n\nstatic const char *txtToken (LexState *ls, int token) {\n switch (token) {\n case TK_NAME:\n case TK_STRING:\n case TK_NUMBER:\n save(ls, '\\0');\n return luaZ_buffer(ls->buff);\n default:\n return luaX_token2str(ls, token);\n }\n}\n\n\nvoid luaX_lexerror (LexState *ls, const char *msg, int token) {\n char buff[MAXSRC];\n luaO_chunkid(buff, getstr(ls->source), MAXSRC);\n msg = luaO_pushfstring(ls->L, \"%s:%d: %s\", buff, ls->linenumber, msg);\n if (token)\n luaO_pushfstring(ls->L, \"%s near \" LUA_QS, msg, txtToken(ls, token));\n luaD_throw(ls->L, LUA_ERRSYNTAX);\n}\n\n\nvoid luaX_syntaxerror (LexState *ls, const char *msg) {\n luaX_lexerror(ls, msg, ls->t.token);\n}\n\n\nTString *luaX_newstring (LexState *ls, const char *str, size_t l) {\n lua_State *L = ls->L;\n TString *ts = luaS_newlstr(L, str, l);\n TValue *o = luaH_setstr(L, ls->fs->h, ts); /* entry for `str' */\n if (ttisnil(o)) {\n setbvalue(o, 1); /* make sure `str' will not be collected */\n luaC_checkGC(L);\n }\n return ts;\n}\n\n\nstatic void inclinenumber (LexState *ls) {\n int old = ls->current;\n lua_assert(currIsNewline(ls));\n next(ls); /* skip `\\n' or `\\r' */\n if (currIsNewline(ls) && ls->current != old)\n next(ls); /* skip `\\n\\r' or `\\r\\n' */\n if (++ls->linenumber >= MAX_INT)\n luaX_syntaxerror(ls, \"chunk has too many lines\");\n}\n\n\nvoid luaX_setinput (lua_State *L, LexState *ls, ZIO *z, TString *source) {\n ls->decpoint = '.';\n ls->L = L;\n ls->lookahead.token = TK_EOS; /* no look-ahead token */\n ls->z = z;\n ls->fs = NULL;\n ls->linenumber = 1;\n ls->lastline = 1;\n ls->source = source;\n luaZ_resizebuffer(ls->L, ls->buff, LUA_MINBUFFER); /* initialize buffer */\n next(ls); /* read first char */\n}\n\n\n\n/*\n** =======================================================\n** LEXICAL ANALYZER\n** =======================================================\n*/\n\n\n\nstatic int check_next (LexState *ls, const char *set) {\n if (!strchr(set, ls->current))\n return 0;\n save_and_next(ls);\n return 1;\n}\n\n\nstatic void buffreplace (LexState *ls, char from, char to) {\n size_t n = luaZ_bufflen(ls->buff);\n char *p = luaZ_buffer(ls->buff);\n while (n--)\n if (p[n] == from) p[n] = to;\n}\n\n\nstatic void trydecpoint (LexState *ls, SemInfo *seminfo) {\n /* format error: try to update decimal point separator */\n struct lconv *cv = localeconv();\n char old = ls->decpoint;\n ls->decpoint = (cv ? cv->decimal_point[0] : '.');\n buffreplace(ls, old, ls->decpoint); /* try updated decimal separator */\n if (!luaO_str2d(luaZ_buffer(ls->buff), &seminfo->r)) {\n /* format error with correct decimal point: no more options */\n buffreplace(ls, ls->decpoint, '.'); /* undo change (for error message) */\n luaX_lexerror(ls, \"malformed number\", TK_NUMBER);\n }\n}\n\n\n/* LUA_NUMBER */\nstatic void read_numeral (LexState *ls, SemInfo *seminfo) {\n lua_assert(isdigit(ls->current));\n do {\n save_and_next(ls);\n } while (isdigit(ls->current) || ls->current == '.');\n if (check_next(ls, \"Ee\")) /* `E'? */\n check_next(ls, \"+-\"); /* optional exponent sign */\n while (isalnum(ls->current) || ls->current == '_')\n save_and_next(ls);\n save(ls, '\\0');\n buffreplace(ls, '.', ls->decpoint); /* follow locale for decimal point */\n if (!luaO_str2d(luaZ_buffer(ls->buff), &seminfo->r)) /* format error? */\n trydecpoint(ls, seminfo); /* try to update decimal point separator */\n}\n\n\nstatic int skip_sep (LexState *ls) {\n int count = 0;\n int s = ls->current;\n lua_assert(s == '[' || s == ']');\n save_and_next(ls);\n while (ls->current == '=') {\n save_and_next(ls);\n count++;\n }\n return (ls->current == s) ? count : (-count) - 1;\n}\n\n\nstatic void read_long_string (LexState *ls, SemInfo *seminfo, int sep) {\n int cont = 0;\n (void)(cont); /* avoid warnings when `cont' is not used */\n save_and_next(ls); /* skip 2nd `[' */\n if (currIsNewline(ls)) /* string starts with a newline? */\n inclinenumber(ls); /* skip it */\n for (;;) {\n switch (ls->current) {\n case EOZ:\n luaX_lexerror(ls, (seminfo) ? \"unfinished long string\" :\n \"unfinished long comment\", TK_EOS);\n break; /* to avoid warnings */\n#if defined(LUA_COMPAT_LSTR)\n case '[': {\n if (skip_sep(ls) == sep) {\n save_and_next(ls); /* skip 2nd `[' */\n cont++;\n#if LUA_COMPAT_LSTR == 1\n if (sep == 0)\n luaX_lexerror(ls, \"nesting of [[...]] is deprecated\", '[');\n#endif\n }\n break;\n }\n#endif\n case ']': {\n if (skip_sep(ls) == sep) {\n save_and_next(ls); /* skip 2nd `]' */\n#if defined(LUA_COMPAT_LSTR) && LUA_COMPAT_LSTR == 2\n cont--;\n if (sep == 0 && cont >= 0) break;\n#endif\n goto endloop;\n }\n break;\n }\n case '\\n':\n case '\\r': {\n save(ls, '\\n');\n inclinenumber(ls);\n if (!seminfo) luaZ_resetbuffer(ls->buff); /* avoid wasting space */\n break;\n }\n default: {\n if (seminfo) save_and_next(ls);\n else next(ls);\n }\n }\n } endloop:\n if (seminfo)\n seminfo->ts = luaX_newstring(ls, luaZ_buffer(ls->buff) + (2 + sep),\n luaZ_bufflen(ls->buff) - 2*(2 + sep));\n}\n\n\nstatic void read_string (LexState *ls, int del, SemInfo *seminfo) {\n save_and_next(ls);\n while (ls->current != del) {\n switch (ls->current) {\n case EOZ:\n luaX_lexerror(ls, \"unfinished string\", TK_EOS);\n continue; /* to avoid warnings */\n case '\\n':\n case '\\r':\n luaX_lexerror(ls, \"unfinished string\", TK_STRING);\n continue; /* to avoid warnings */\n case '\\\\': {\n int c;\n next(ls); /* do not save the `\\' */\n switch (ls->current) {\n case 'a': c = '\\a'; break;\n case 'b': c = '\\b'; break;\n case 'f': c = '\\f'; break;\n case 'n': c = '\\n'; break;\n case 'r': c = '\\r'; break;\n case 't': c = '\\t'; break;\n case 'v': c = '\\v'; break;\n case '\\n': /* go through */\n case '\\r': save(ls, '\\n'); inclinenumber(ls); continue;\n case EOZ: continue; /* will raise an error next loop */\n default: {\n if (!isdigit(ls->current))\n save_and_next(ls); /* handles \\\\, \\\", \\', and \\? */\n else { /* \\xxx */\n int i = 0;\n c = 0;\n do {\n c = 10*c + (ls->current-'0');\n next(ls);\n } while (++i<3 && isdigit(ls->current));\n if (c > UCHAR_MAX)\n luaX_lexerror(ls, \"escape sequence too large\", TK_STRING);\n save(ls, c);\n }\n continue;\n }\n }\n save(ls, c);\n next(ls);\n continue;\n }\n default:\n save_and_next(ls);\n }\n }\n save_and_next(ls); /* skip delimiter */\n seminfo->ts = luaX_newstring(ls, luaZ_buffer(ls->buff) + 1,\n luaZ_bufflen(ls->buff) - 2);\n}\n\n\nstatic int llex (LexState *ls, SemInfo *seminfo) {\n luaZ_resetbuffer(ls->buff);\n for (;;) {\n switch (ls->current) {\n case '\\n':\n case '\\r': {\n inclinenumber(ls);\n continue;\n }\n case '-': {\n next(ls);\n if (ls->current != '-') return '-';\n /* else is a comment */\n next(ls);\n if (ls->current == '[') {\n int sep = skip_sep(ls);\n luaZ_resetbuffer(ls->buff); /* `skip_sep' may dirty the buffer */\n if (sep >= 0) {\n read_long_string(ls, NULL, sep); /* long comment */\n luaZ_resetbuffer(ls->buff);\n continue;\n }\n }\n /* else short comment */\n while (!currIsNewline(ls) && ls->current != EOZ)\n next(ls);\n continue;\n }\n case '[': {\n int sep = skip_sep(ls);\n if (sep >= 0) {\n read_long_string(ls, seminfo, sep);\n return TK_STRING;\n }\n else if (sep == -1) return '[';\n else luaX_lexerror(ls, \"invalid long string delimiter\", TK_STRING);\n }\n case '=': {\n next(ls);\n if (ls->current != '=') return '=';\n else { next(ls); return TK_EQ; }\n }\n case '<': {\n next(ls);\n if (ls->current != '=') return '<';\n else { next(ls); return TK_LE; }\n }\n case '>': {\n next(ls);\n if (ls->current != '=') return '>';\n else { next(ls); return TK_GE; }\n }\n case '~': {\n next(ls);\n if (ls->current != '=') return '~';\n else { next(ls); return TK_NE; }\n }\n case '\"':\n case '\\'': {\n read_string(ls, ls->current, seminfo);\n return TK_STRING;\n }\n case '.': {\n save_and_next(ls);\n if (check_next(ls, \".\")) {\n if (check_next(ls, \".\"))\n return TK_DOTS; /* ... */\n else return TK_CONCAT; /* .. */\n }\n else if (!isdigit(ls->current)) return '.';\n else {\n read_numeral(ls, seminfo);\n return TK_NUMBER;\n }\n }\n case EOZ: {\n return TK_EOS;\n }\n default: {\n if (isspace(ls->current)) {\n lua_assert(!currIsNewline(ls));\n next(ls);\n continue;\n }\n else if (isdigit(ls->current)) {\n read_numeral(ls, seminfo);\n return TK_NUMBER;\n }\n else if (isalpha(ls->current) || ls->current == '_') {\n /* identifier or reserved word */\n TString *ts;\n do {\n save_and_next(ls);\n } while (isalnum(ls->current) || ls->current == '_');\n ts = luaX_newstring(ls, luaZ_buffer(ls->buff),\n luaZ_bufflen(ls->buff));\n if (ts->tsv.reserved > 0) /* reserved word? */\n return ts->tsv.reserved - 1 + FIRST_RESERVED;\n else {\n seminfo->ts = ts;\n return TK_NAME;\n }\n }\n else {\n int c = ls->current;\n next(ls);\n return c; /* single-char tokens (+ - / ...) */\n }\n }\n }\n }\n}\n\n\nvoid luaX_next (LexState *ls) {\n ls->lastline = ls->linenumber;\n if (ls->lookahead.token != TK_EOS) { /* is there a look-ahead token? */\n ls->t = ls->lookahead; /* use this one */\n ls->lookahead.token = TK_EOS; /* and discharge it */\n }\n else\n ls->t.token = llex(ls, &ls->t.seminfo); /* read next token */\n}\n\n\nvoid luaX_lookahead (LexState *ls) {\n lua_assert(ls->lookahead.token == TK_EOS);\n ls->lookahead.token = llex(ls, &ls->lookahead.seminfo);\n}\n\n" }, { "path": "deps/lua/src/llex.h", "content": "/*\n** $Id: llex.h,v 1.58.1.1 2007/12/27 13:02:25 roberto Exp $\n** Lexical Analyzer\n** See Copyright Notice in lua.h\n*/\n\n#ifndef llex_h\n#define llex_h\n\n#include \"lobject.h\"\n#include \"lzio.h\"\n\n\n#define FIRST_RESERVED\t257\n\n/* maximum length of a reserved word */\n#define TOKEN_LEN\t(sizeof(\"function\")/sizeof(char))\n\n\n/*\n* WARNING: if you change the order of this enumeration,\n* grep \"ORDER RESERVED\"\n*/\nenum RESERVED {\n /* terminal symbols denoted by reserved words */\n TK_AND = FIRST_RESERVED, TK_BREAK,\n TK_DO, TK_ELSE, TK_ELSEIF, TK_END, TK_FALSE, TK_FOR, TK_FUNCTION,\n TK_IF, TK_IN, TK_LOCAL, TK_NIL, TK_NOT, TK_OR, TK_REPEAT,\n TK_RETURN, TK_THEN, TK_TRUE, TK_UNTIL, TK_WHILE,\n /* other terminal symbols */\n TK_CONCAT, TK_DOTS, TK_EQ, TK_GE, TK_LE, TK_NE, TK_NUMBER,\n TK_NAME, TK_STRING, TK_EOS\n};\n\n/* number of reserved words */\n#define NUM_RESERVED\t(cast(int, TK_WHILE-FIRST_RESERVED+1))\n\n\n/* array with token `names' */\nLUAI_DATA const char *const luaX_tokens [];\n\n\ntypedef union {\n lua_Number r;\n TString *ts;\n} SemInfo; /* semantics information */\n\n\ntypedef struct Token {\n int token;\n SemInfo seminfo;\n} Token;\n\n\ntypedef struct LexState {\n int current; /* current character (charint) */\n int linenumber; /* input line counter */\n int lastline; /* line of last token `consumed' */\n Token t; /* current token */\n Token lookahead; /* look ahead token */\n struct FuncState *fs; /* `FuncState' is private to the parser */\n struct lua_State *L;\n ZIO *z; /* input stream */\n Mbuffer *buff; /* buffer for tokens */\n TString *source; /* current source name */\n char decpoint; /* locale decimal point */\n} LexState;\n\n\nLUAI_FUNC void luaX_init (lua_State *L);\nLUAI_FUNC void luaX_setinput (lua_State *L, LexState *ls, ZIO *z,\n TString *source);\nLUAI_FUNC TString *luaX_newstring (LexState *ls, const char *str, size_t l);\nLUAI_FUNC void luaX_next (LexState *ls);\nLUAI_FUNC void luaX_lookahead (LexState *ls);\nLUAI_FUNC void luaX_lexerror (LexState *ls, const char *msg, int token);\nLUAI_FUNC void luaX_syntaxerror (LexState *ls, const char *s);\nLUAI_FUNC const char *luaX_token2str (LexState *ls, int token);\n\n\n#endif\n" }, { "path": "deps/lua/src/llimits.h", "content": "/*\n** $Id: llimits.h,v 1.69.1.1 2007/12/27 13:02:25 roberto Exp $\n** Limits, basic types, and some other `installation-dependent' definitions\n** See Copyright Notice in lua.h\n*/\n\n#ifndef llimits_h\n#define llimits_h\n\n\n#include \n#include \n\n\n#include \"lua.h\"\n\n\ntypedef LUAI_UINT32 lu_int32;\n\ntypedef LUAI_UMEM lu_mem;\n\ntypedef LUAI_MEM l_mem;\n\n\n\n/* chars used as small naturals (so that `char' is reserved for characters) */\ntypedef unsigned char lu_byte;\n\n\n#define MAX_SIZET\t((size_t)(~(size_t)0)-2)\n\n#define MAX_LUMEM\t((lu_mem)(~(lu_mem)0)-2)\n\n\n#define MAX_INT (INT_MAX-2) /* maximum value of an int (-2 for safety) */\n\n/*\n** conversion of pointer to integer\n** this is for hashing only; there is no problem if the integer\n** cannot hold the whole pointer value\n*/\n#define IntPoint(p) ((unsigned int)(lu_mem)(p))\n\n\n\n/* type to ensure maximum alignment */\ntypedef LUAI_USER_ALIGNMENT_T L_Umaxalign;\n\n\n/* result of a `usual argument conversion' over lua_Number */\ntypedef LUAI_UACNUMBER l_uacNumber;\n\n\n/* internal assertions for in-house debugging */\n#ifdef lua_assert\n\n#define check_exp(c,e)\t\t(lua_assert(c), (e))\n#define api_check(l,e)\t\tlua_assert(e)\n\n#else\n\n#define lua_assert(c)\t\t((void)0)\n#define check_exp(c,e)\t\t(e)\n#define api_check\t\tluai_apicheck\n\n#endif\n\n\n#ifndef UNUSED\n#define UNUSED(x)\t((void)(x))\t/* to avoid warnings */\n#endif\n\n\n#ifndef cast\n#define cast(t, exp)\t((t)(exp))\n#endif\n\n#define cast_byte(i)\tcast(lu_byte, (i))\n#define cast_num(i)\tcast(lua_Number, (i))\n#define cast_int(i)\tcast(int, (i))\n\n\n\n/*\n** type for virtual-machine instructions\n** must be an unsigned with (at least) 4 bytes (see details in lopcodes.h)\n*/\ntypedef lu_int32 Instruction;\n\n\n\n/* maximum stack for a Lua function */\n#define MAXSTACK\t250\n\n\n\n/* minimum size for the string table (must be power of 2) */\n#ifndef MINSTRTABSIZE\n#define MINSTRTABSIZE\t32\n#endif\n\n\n/* minimum size for string buffer */\n#ifndef LUA_MINBUFFER\n#define LUA_MINBUFFER\t32\n#endif\n\n\n#ifndef lua_lock\n#define lua_lock(L) ((void) 0) \n#define lua_unlock(L) ((void) 0)\n#endif\n\n#ifndef luai_threadyield\n#define luai_threadyield(L) {lua_unlock(L); lua_lock(L);}\n#endif\n\n\n/*\n** macro to control inclusion of some hard tests on stack reallocation\n*/ \n#ifndef HARDSTACKTESTS\n#define condhardstacktests(x)\t((void)0)\n#else\n#define condhardstacktests(x)\tx\n#endif\n\n#endif\n" }, { "path": "deps/lua/src/lmathlib.c", "content": "/*\n** $Id: lmathlib.c,v 1.67.1.1 2007/12/27 13:02:25 roberto Exp $\n** Standard mathematical library\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n\n#define lmathlib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n#undef PI\n#define PI (3.14159265358979323846)\n#define RADIANS_PER_DEGREE (PI/180.0)\n\n\n\nstatic int math_abs (lua_State *L) {\n lua_pushnumber(L, fabs(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_sin (lua_State *L) {\n lua_pushnumber(L, sin(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_sinh (lua_State *L) {\n lua_pushnumber(L, sinh(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_cos (lua_State *L) {\n lua_pushnumber(L, cos(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_cosh (lua_State *L) {\n lua_pushnumber(L, cosh(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_tan (lua_State *L) {\n lua_pushnumber(L, tan(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_tanh (lua_State *L) {\n lua_pushnumber(L, tanh(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_asin (lua_State *L) {\n lua_pushnumber(L, asin(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_acos (lua_State *L) {\n lua_pushnumber(L, acos(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_atan (lua_State *L) {\n lua_pushnumber(L, atan(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_atan2 (lua_State *L) {\n lua_pushnumber(L, atan2(luaL_checknumber(L, 1), luaL_checknumber(L, 2)));\n return 1;\n}\n\nstatic int math_ceil (lua_State *L) {\n lua_pushnumber(L, ceil(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_floor (lua_State *L) {\n lua_pushnumber(L, floor(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_fmod (lua_State *L) {\n lua_pushnumber(L, fmod(luaL_checknumber(L, 1), luaL_checknumber(L, 2)));\n return 1;\n}\n\nstatic int math_modf (lua_State *L) {\n double ip;\n double fp = modf(luaL_checknumber(L, 1), &ip);\n lua_pushnumber(L, ip);\n lua_pushnumber(L, fp);\n return 2;\n}\n\nstatic int math_sqrt (lua_State *L) {\n lua_pushnumber(L, sqrt(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_pow (lua_State *L) {\n lua_pushnumber(L, pow(luaL_checknumber(L, 1), luaL_checknumber(L, 2)));\n return 1;\n}\n\nstatic int math_log (lua_State *L) {\n lua_pushnumber(L, log(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_log10 (lua_State *L) {\n lua_pushnumber(L, log10(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_exp (lua_State *L) {\n lua_pushnumber(L, exp(luaL_checknumber(L, 1)));\n return 1;\n}\n\nstatic int math_deg (lua_State *L) {\n lua_pushnumber(L, luaL_checknumber(L, 1)/RADIANS_PER_DEGREE);\n return 1;\n}\n\nstatic int math_rad (lua_State *L) {\n lua_pushnumber(L, luaL_checknumber(L, 1)*RADIANS_PER_DEGREE);\n return 1;\n}\n\nstatic int math_frexp (lua_State *L) {\n int e;\n lua_pushnumber(L, frexp(luaL_checknumber(L, 1), &e));\n lua_pushinteger(L, e);\n return 2;\n}\n\nstatic int math_ldexp (lua_State *L) {\n lua_pushnumber(L, ldexp(luaL_checknumber(L, 1), luaL_checkint(L, 2)));\n return 1;\n}\n\n\n\nstatic int math_min (lua_State *L) {\n int n = lua_gettop(L); /* number of arguments */\n lua_Number dmin = luaL_checknumber(L, 1);\n int i;\n for (i=2; i<=n; i++) {\n lua_Number d = luaL_checknumber(L, i);\n if (d < dmin)\n dmin = d;\n }\n lua_pushnumber(L, dmin);\n return 1;\n}\n\n\nstatic int math_max (lua_State *L) {\n int n = lua_gettop(L); /* number of arguments */\n lua_Number dmax = luaL_checknumber(L, 1);\n int i;\n for (i=2; i<=n; i++) {\n lua_Number d = luaL_checknumber(L, i);\n if (d > dmax)\n dmax = d;\n }\n lua_pushnumber(L, dmax);\n return 1;\n}\n\n\nstatic int math_random (lua_State *L) {\n /* the `%' avoids the (rare) case of r==1, and is needed also because on\n some systems (SunOS!) `rand()' may return a value larger than RAND_MAX */\n lua_Number r = (lua_Number)(rand()%RAND_MAX) / (lua_Number)RAND_MAX;\n switch (lua_gettop(L)) { /* check number of arguments */\n case 0: { /* no arguments */\n lua_pushnumber(L, r); /* Number between 0 and 1 */\n break;\n }\n case 1: { /* only upper limit */\n int u = luaL_checkint(L, 1);\n luaL_argcheck(L, 1<=u, 1, \"interval is empty\");\n lua_pushnumber(L, floor(r*u)+1); /* int between 1 and `u' */\n break;\n }\n case 2: { /* lower and upper limits */\n int l = luaL_checkint(L, 1);\n int u = luaL_checkint(L, 2);\n luaL_argcheck(L, l<=u, 2, \"interval is empty\");\n lua_pushnumber(L, floor(r*(u-l+1))+l); /* int between `l' and `u' */\n break;\n }\n default: return luaL_error(L, \"wrong number of arguments\");\n }\n return 1;\n}\n\n\nstatic int math_randomseed (lua_State *L) {\n srand(luaL_checkint(L, 1));\n return 0;\n}\n\n\nstatic const luaL_Reg mathlib[] = {\n {\"abs\", math_abs},\n {\"acos\", math_acos},\n {\"asin\", math_asin},\n {\"atan2\", math_atan2},\n {\"atan\", math_atan},\n {\"ceil\", math_ceil},\n {\"cosh\", math_cosh},\n {\"cos\", math_cos},\n {\"deg\", math_deg},\n {\"exp\", math_exp},\n {\"floor\", math_floor},\n {\"fmod\", math_fmod},\n {\"frexp\", math_frexp},\n {\"ldexp\", math_ldexp},\n {\"log10\", math_log10},\n {\"log\", math_log},\n {\"max\", math_max},\n {\"min\", math_min},\n {\"modf\", math_modf},\n {\"pow\", math_pow},\n {\"rad\", math_rad},\n {\"random\", math_random},\n {\"randomseed\", math_randomseed},\n {\"sinh\", math_sinh},\n {\"sin\", math_sin},\n {\"sqrt\", math_sqrt},\n {\"tanh\", math_tanh},\n {\"tan\", math_tan},\n {NULL, NULL}\n};\n\n\n/*\n** Open math library\n*/\nLUALIB_API int luaopen_math (lua_State *L) {\n luaL_register(L, LUA_MATHLIBNAME, mathlib);\n lua_pushnumber(L, PI);\n lua_setfield(L, -2, \"pi\");\n lua_pushnumber(L, HUGE_VAL);\n lua_setfield(L, -2, \"huge\");\n#if defined(LUA_COMPAT_MOD)\n lua_getfield(L, -1, \"fmod\");\n lua_setfield(L, -2, \"mod\");\n#endif\n return 1;\n}\n\n" }, { "path": "deps/lua/src/lmem.c", "content": "/*\n** $Id: lmem.c,v 1.70.1.1 2007/12/27 13:02:25 roberto Exp $\n** Interface to Memory Manager\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lmem_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n\n\n\n/*\n** About the realloc function:\n** void * frealloc (void *ud, void *ptr, size_t osize, size_t nsize);\n** (`osize' is the old size, `nsize' is the new size)\n**\n** Lua ensures that (ptr == NULL) iff (osize == 0).\n**\n** * frealloc(ud, NULL, 0, x) creates a new block of size `x'\n**\n** * frealloc(ud, p, x, 0) frees the block `p'\n** (in this specific case, frealloc must return NULL).\n** particularly, frealloc(ud, NULL, 0, 0) does nothing\n** (which is equivalent to free(NULL) in ANSI C)\n**\n** frealloc returns NULL if it cannot create or reallocate the area\n** (any reallocation to an equal or smaller size cannot fail!)\n*/\n\n\n\n#define MINSIZEARRAY\t4\n\n\nvoid *luaM_growaux_ (lua_State *L, void *block, int *size, size_t size_elems,\n int limit, const char *errormsg) {\n void *newblock;\n int newsize;\n if (*size >= limit/2) { /* cannot double it? */\n if (*size >= limit) /* cannot grow even a little? */\n luaG_runerror(L, errormsg);\n newsize = limit; /* still have at least one free place */\n }\n else {\n newsize = (*size)*2;\n if (newsize < MINSIZEARRAY)\n newsize = MINSIZEARRAY; /* minimum size */\n }\n newblock = luaM_reallocv(L, block, *size, newsize, size_elems);\n *size = newsize; /* update only when everything else is OK */\n return newblock;\n}\n\n\nvoid *luaM_toobig (lua_State *L) {\n luaG_runerror(L, \"memory allocation error: block too big\");\n return NULL; /* to avoid warnings */\n}\n\n\n\n/*\n** generic allocation routine.\n*/\nvoid *luaM_realloc_ (lua_State *L, void *block, size_t osize, size_t nsize) {\n global_State *g = G(L);\n lua_assert((osize == 0) == (block == NULL));\n block = (*g->frealloc)(g->ud, block, osize, nsize);\n if (block == NULL && nsize > 0)\n luaD_throw(L, LUA_ERRMEM);\n lua_assert((nsize == 0) == (block == NULL));\n g->totalbytes = (g->totalbytes - osize) + nsize;\n return block;\n}\n\n" }, { "path": "deps/lua/src/lmem.h", "content": "/*\n** $Id: lmem.h,v 1.31.1.1 2007/12/27 13:02:25 roberto Exp $\n** Interface to Memory Manager\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lmem_h\n#define lmem_h\n\n\n#include \n\n#include \"llimits.h\"\n#include \"lua.h\"\n\n#define MEMERRMSG\t\"not enough memory\"\n\n\n#define luaM_reallocv(L,b,on,n,e) \\\n\t((cast(size_t, (n)+1) <= MAX_SIZET/(e)) ? /* +1 to avoid warnings */ \\\n\t\tluaM_realloc_(L, (b), (on)*(e), (n)*(e)) : \\\n\t\tluaM_toobig(L))\n\n#define luaM_freemem(L, b, s)\tluaM_realloc_(L, (b), (s), 0)\n#define luaM_free(L, b)\t\tluaM_realloc_(L, (b), sizeof(*(b)), 0)\n#define luaM_freearray(L, b, n, t) luaM_reallocv(L, (b), n, 0, sizeof(t))\n\n#define luaM_malloc(L,t)\tluaM_realloc_(L, NULL, 0, (t))\n#define luaM_new(L,t)\t\tcast(t *, luaM_malloc(L, sizeof(t)))\n#define luaM_newvector(L,n,t) \\\n\t\tcast(t *, luaM_reallocv(L, NULL, 0, n, sizeof(t)))\n\n#define luaM_growvector(L,v,nelems,size,t,limit,e) \\\n if ((nelems)+1 > (size)) \\\n ((v)=cast(t *, luaM_growaux_(L,v,&(size),sizeof(t),limit,e)))\n\n#define luaM_reallocvector(L, v,oldn,n,t) \\\n ((v)=cast(t *, luaM_reallocv(L, v, oldn, n, sizeof(t))))\n\n\nLUAI_FUNC void *luaM_realloc_ (lua_State *L, void *block, size_t oldsize,\n size_t size);\nLUAI_FUNC void *luaM_toobig (lua_State *L);\nLUAI_FUNC void *luaM_growaux_ (lua_State *L, void *block, int *size,\n size_t size_elem, int limit,\n const char *errormsg);\n\n#endif\n\n" }, { "path": "deps/lua/src/loadlib.c", "content": "/*\n** $Id: loadlib.c,v 1.52.1.4 2009/09/09 13:17:16 roberto Exp $\n** Dynamic library loader for Lua\n** See Copyright Notice in lua.h\n**\n** This module contains an implementation of loadlib for Unix systems\n** that have dlfcn, an implementation for Darwin (Mac OS X), an\n** implementation for Windows, and a stub for other systems.\n*/\n\n\n#include \n#include \n\n\n#define loadlib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n/* prefix for open functions in C libraries */\n#define LUA_POF\t\t\"luaopen_\"\n\n/* separator for open functions in C libraries */\n#define LUA_OFSEP\t\"_\"\n\n\n#define LIBPREFIX\t\"LOADLIB: \"\n\n#define POF\t\tLUA_POF\n#define LIB_FAIL\t\"open\"\n\n\n/* error codes for ll_loadfunc */\n#define ERRLIB\t\t1\n#define ERRFUNC\t\t2\n\n#define setprogdir(L)\t\t((void)0)\n\n\nstatic void ll_unloadlib (void *lib);\nstatic void *ll_load (lua_State *L, const char *path);\nstatic lua_CFunction ll_sym (lua_State *L, void *lib, const char *sym);\n\n\n\n#if defined(LUA_DL_DLOPEN)\n/*\n** {========================================================================\n** This is an implementation of loadlib based on the dlfcn interface.\n** The dlfcn interface is available in Linux, SunOS, Solaris, IRIX, FreeBSD,\n** NetBSD, AIX 4.2, HPUX 11, and probably most other Unix flavors, at least\n** as an emulation layer on top of native functions.\n** =========================================================================\n*/\n\n#include \n\nstatic void ll_unloadlib (void *lib) {\n dlclose(lib);\n}\n\n\nstatic void *ll_load (lua_State *L, const char *path) {\n void *lib = dlopen(path, RTLD_NOW);\n if (lib == NULL) lua_pushstring(L, dlerror());\n return lib;\n}\n\n\nstatic lua_CFunction ll_sym (lua_State *L, void *lib, const char *sym) {\n lua_CFunction f = (lua_CFunction)dlsym(lib, sym);\n if (f == NULL) lua_pushstring(L, dlerror());\n return f;\n}\n\n/* }====================================================== */\n\n\n\n#elif defined(LUA_DL_DLL)\n/*\n** {======================================================================\n** This is an implementation of loadlib for Windows using native functions.\n** =======================================================================\n*/\n\n#include \n\n\n#undef setprogdir\n\nstatic void setprogdir (lua_State *L) {\n char buff[MAX_PATH + 1];\n char *lb;\n DWORD nsize = sizeof(buff)/sizeof(char);\n DWORD n = GetModuleFileNameA(NULL, buff, nsize);\n if (n == 0 || n == nsize || (lb = strrchr(buff, '\\\\')) == NULL)\n luaL_error(L, \"unable to get ModuleFileName\");\n else {\n *lb = '\\0';\n luaL_gsub(L, lua_tostring(L, -1), LUA_EXECDIR, buff);\n lua_remove(L, -2); /* remove original string */\n }\n}\n\n\nstatic void pusherror (lua_State *L) {\n int error = GetLastError();\n char buffer[128];\n if (FormatMessageA(FORMAT_MESSAGE_IGNORE_INSERTS | FORMAT_MESSAGE_FROM_SYSTEM,\n NULL, error, 0, buffer, sizeof(buffer), NULL))\n lua_pushstring(L, buffer);\n else\n lua_pushfstring(L, \"system error %d\\n\", error);\n}\n\nstatic void ll_unloadlib (void *lib) {\n FreeLibrary((HINSTANCE)lib);\n}\n\n\nstatic void *ll_load (lua_State *L, const char *path) {\n HINSTANCE lib = LoadLibraryA(path);\n if (lib == NULL) pusherror(L);\n return lib;\n}\n\n\nstatic lua_CFunction ll_sym (lua_State *L, void *lib, const char *sym) {\n lua_CFunction f = (lua_CFunction)GetProcAddress((HINSTANCE)lib, sym);\n if (f == NULL) pusherror(L);\n return f;\n}\n\n/* }====================================================== */\n\n\n\n#elif defined(LUA_DL_DYLD)\n/*\n** {======================================================================\n** Native Mac OS X / Darwin Implementation\n** =======================================================================\n*/\n\n#include \n\n\n/* Mac appends a `_' before C function names */\n#undef POF\n#define POF\t\"_\" LUA_POF\n\n\nstatic void pusherror (lua_State *L) {\n const char *err_str;\n const char *err_file;\n NSLinkEditErrors err;\n int err_num;\n NSLinkEditError(&err, &err_num, &err_file, &err_str);\n lua_pushstring(L, err_str);\n}\n\n\nstatic const char *errorfromcode (NSObjectFileImageReturnCode ret) {\n switch (ret) {\n case NSObjectFileImageInappropriateFile:\n return \"file is not a bundle\";\n case NSObjectFileImageArch:\n return \"library is for wrong CPU type\";\n case NSObjectFileImageFormat:\n return \"bad format\";\n case NSObjectFileImageAccess:\n return \"cannot access file\";\n case NSObjectFileImageFailure:\n default:\n return \"unable to load library\";\n }\n}\n\n\nstatic void ll_unloadlib (void *lib) {\n NSUnLinkModule((NSModule)lib, NSUNLINKMODULE_OPTION_RESET_LAZY_REFERENCES);\n}\n\n\nstatic void *ll_load (lua_State *L, const char *path) {\n NSObjectFileImage img;\n NSObjectFileImageReturnCode ret;\n /* this would be a rare case, but prevents crashing if it happens */\n if(!_dyld_present()) {\n lua_pushliteral(L, \"dyld not present\");\n return NULL;\n }\n ret = NSCreateObjectFileImageFromFile(path, &img);\n if (ret == NSObjectFileImageSuccess) {\n NSModule mod = NSLinkModule(img, path, NSLINKMODULE_OPTION_PRIVATE |\n NSLINKMODULE_OPTION_RETURN_ON_ERROR);\n NSDestroyObjectFileImage(img);\n if (mod == NULL) pusherror(L);\n return mod;\n }\n lua_pushstring(L, errorfromcode(ret));\n return NULL;\n}\n\n\nstatic lua_CFunction ll_sym (lua_State *L, void *lib, const char *sym) {\n NSSymbol nss = NSLookupSymbolInModule((NSModule)lib, sym);\n if (nss == NULL) {\n lua_pushfstring(L, \"symbol \" LUA_QS \" not found\", sym);\n return NULL;\n }\n return (lua_CFunction)NSAddressOfSymbol(nss);\n}\n\n/* }====================================================== */\n\n\n\n#else\n/*\n** {======================================================\n** Fallback for other systems\n** =======================================================\n*/\n\n#undef LIB_FAIL\n#define LIB_FAIL\t\"absent\"\n\n\n#define DLMSG\t\"dynamic libraries not enabled; check your Lua installation\"\n\n\nstatic void ll_unloadlib (void *lib) {\n (void)lib; /* to avoid warnings */\n}\n\n\nstatic void *ll_load (lua_State *L, const char *path) {\n (void)path; /* to avoid warnings */\n lua_pushliteral(L, DLMSG);\n return NULL;\n}\n\n\nstatic lua_CFunction ll_sym (lua_State *L, void *lib, const char *sym) {\n (void)lib; (void)sym; /* to avoid warnings */\n lua_pushliteral(L, DLMSG);\n return NULL;\n}\n\n/* }====================================================== */\n#endif\n\n\n\nstatic void **ll_register (lua_State *L, const char *path) {\n void **plib;\n lua_pushfstring(L, \"%s%s\", LIBPREFIX, path);\n lua_gettable(L, LUA_REGISTRYINDEX); /* check library in registry? */\n if (!lua_isnil(L, -1)) /* is there an entry? */\n plib = (void **)lua_touserdata(L, -1);\n else { /* no entry yet; create one */\n lua_pop(L, 1);\n plib = (void **)lua_newuserdata(L, sizeof(const void *));\n *plib = NULL;\n luaL_getmetatable(L, \"_LOADLIB\");\n lua_setmetatable(L, -2);\n lua_pushfstring(L, \"%s%s\", LIBPREFIX, path);\n lua_pushvalue(L, -2);\n lua_settable(L, LUA_REGISTRYINDEX);\n }\n return plib;\n}\n\n\n/*\n** __gc tag method: calls library's `ll_unloadlib' function with the lib\n** handle\n*/\nstatic int gctm (lua_State *L) {\n void **lib = (void **)luaL_checkudata(L, 1, \"_LOADLIB\");\n if (*lib) ll_unloadlib(*lib);\n *lib = NULL; /* mark library as closed */\n return 0;\n}\n\n\nstatic int ll_loadfunc (lua_State *L, const char *path, const char *sym) {\n void **reg = ll_register(L, path);\n if (*reg == NULL) *reg = ll_load(L, path);\n if (*reg == NULL)\n return ERRLIB; /* unable to load library */\n else {\n lua_CFunction f = ll_sym(L, *reg, sym);\n if (f == NULL)\n return ERRFUNC; /* unable to find function */\n lua_pushcfunction(L, f);\n return 0; /* return function */\n }\n}\n\n\nstatic int ll_loadlib (lua_State *L) {\n const char *path = luaL_checkstring(L, 1);\n const char *init = luaL_checkstring(L, 2);\n int stat = ll_loadfunc(L, path, init);\n if (stat == 0) /* no errors? */\n return 1; /* return the loaded function */\n else { /* error; error message is on stack top */\n lua_pushnil(L);\n lua_insert(L, -2);\n lua_pushstring(L, (stat == ERRLIB) ? LIB_FAIL : \"init\");\n return 3; /* return nil, error message, and where */\n }\n}\n\n\n\n/*\n** {======================================================\n** 'require' function\n** =======================================================\n*/\n\n\nstatic int readable (const char *filename) {\n FILE *f = fopen(filename, \"r\"); /* try to open file */\n if (f == NULL) return 0; /* open failed */\n fclose(f);\n return 1;\n}\n\n\nstatic const char *pushnexttemplate (lua_State *L, const char *path) {\n const char *l;\n while (*path == *LUA_PATHSEP) path++; /* skip separators */\n if (*path == '\\0') return NULL; /* no more templates */\n l = strchr(path, *LUA_PATHSEP); /* find next separator */\n if (l == NULL) l = path + strlen(path);\n lua_pushlstring(L, path, l - path); /* template */\n return l;\n}\n\n\nstatic const char *findfile (lua_State *L, const char *name,\n const char *pname) {\n const char *path;\n name = luaL_gsub(L, name, \".\", LUA_DIRSEP);\n lua_getfield(L, LUA_ENVIRONINDEX, pname);\n path = lua_tostring(L, -1);\n if (path == NULL)\n luaL_error(L, LUA_QL(\"package.%s\") \" must be a string\", pname);\n lua_pushliteral(L, \"\"); /* error accumulator */\n while ((path = pushnexttemplate(L, path)) != NULL) {\n const char *filename;\n filename = luaL_gsub(L, lua_tostring(L, -1), LUA_PATH_MARK, name);\n lua_remove(L, -2); /* remove path template */\n if (readable(filename)) /* does file exist and is readable? */\n return filename; /* return that file name */\n lua_pushfstring(L, \"\\n\\tno file \" LUA_QS, filename);\n lua_remove(L, -2); /* remove file name */\n lua_concat(L, 2); /* add entry to possible error message */\n }\n return NULL; /* not found */\n}\n\n\nstatic void loaderror (lua_State *L, const char *filename) {\n luaL_error(L, \"error loading module \" LUA_QS \" from file \" LUA_QS \":\\n\\t%s\",\n lua_tostring(L, 1), filename, lua_tostring(L, -1));\n}\n\n\nstatic int loader_Lua (lua_State *L) {\n const char *filename;\n const char *name = luaL_checkstring(L, 1);\n filename = findfile(L, name, \"path\");\n if (filename == NULL) return 1; /* library not found in this path */\n if (luaL_loadfile(L, filename) != 0)\n loaderror(L, filename);\n return 1; /* library loaded successfully */\n}\n\n\nstatic const char *mkfuncname (lua_State *L, const char *modname) {\n const char *funcname;\n const char *mark = strchr(modname, *LUA_IGMARK);\n if (mark) modname = mark + 1;\n funcname = luaL_gsub(L, modname, \".\", LUA_OFSEP);\n funcname = lua_pushfstring(L, POF\"%s\", funcname);\n lua_remove(L, -2); /* remove 'gsub' result */\n return funcname;\n}\n\n\nstatic int loader_C (lua_State *L) {\n const char *funcname;\n const char *name = luaL_checkstring(L, 1);\n const char *filename = findfile(L, name, \"cpath\");\n if (filename == NULL) return 1; /* library not found in this path */\n funcname = mkfuncname(L, name);\n if (ll_loadfunc(L, filename, funcname) != 0)\n loaderror(L, filename);\n return 1; /* library loaded successfully */\n}\n\n\nstatic int loader_Croot (lua_State *L) {\n const char *funcname;\n const char *filename;\n const char *name = luaL_checkstring(L, 1);\n const char *p = strchr(name, '.');\n int stat;\n if (p == NULL) return 0; /* is root */\n lua_pushlstring(L, name, p - name);\n filename = findfile(L, lua_tostring(L, -1), \"cpath\");\n if (filename == NULL) return 1; /* root not found */\n funcname = mkfuncname(L, name);\n if ((stat = ll_loadfunc(L, filename, funcname)) != 0) {\n if (stat != ERRFUNC) loaderror(L, filename); /* real error */\n lua_pushfstring(L, \"\\n\\tno module \" LUA_QS \" in file \" LUA_QS,\n name, filename);\n return 1; /* function not found */\n }\n return 1;\n}\n\n\nstatic int loader_preload (lua_State *L) {\n const char *name = luaL_checkstring(L, 1);\n lua_getfield(L, LUA_ENVIRONINDEX, \"preload\");\n if (!lua_istable(L, -1))\n luaL_error(L, LUA_QL(\"package.preload\") \" must be a table\");\n lua_getfield(L, -1, name);\n if (lua_isnil(L, -1)) /* not found? */\n lua_pushfstring(L, \"\\n\\tno field package.preload['%s']\", name);\n return 1;\n}\n\n\nstatic const int sentinel_ = 0;\n#define sentinel\t((void *)&sentinel_)\n\n\nstatic int ll_require (lua_State *L) {\n const char *name = luaL_checkstring(L, 1);\n int i;\n lua_settop(L, 1); /* _LOADED table will be at index 2 */\n lua_getfield(L, LUA_REGISTRYINDEX, \"_LOADED\");\n lua_getfield(L, 2, name);\n if (lua_toboolean(L, -1)) { /* is it there? */\n if (lua_touserdata(L, -1) == sentinel) /* check loops */\n luaL_error(L, \"loop or previous error loading module \" LUA_QS, name);\n return 1; /* package is already loaded */\n }\n /* else must load it; iterate over available loaders */\n lua_getfield(L, LUA_ENVIRONINDEX, \"loaders\");\n if (!lua_istable(L, -1))\n luaL_error(L, LUA_QL(\"package.loaders\") \" must be a table\");\n lua_pushliteral(L, \"\"); /* error message accumulator */\n for (i=1; ; i++) {\n lua_rawgeti(L, -2, i); /* get a loader */\n if (lua_isnil(L, -1))\n luaL_error(L, \"module \" LUA_QS \" not found:%s\",\n name, lua_tostring(L, -2));\n lua_pushstring(L, name);\n lua_call(L, 1, 1); /* call it */\n if (lua_isfunction(L, -1)) /* did it find module? */\n break; /* module loaded successfully */\n else if (lua_isstring(L, -1)) /* loader returned error message? */\n lua_concat(L, 2); /* accumulate it */\n else\n lua_pop(L, 1);\n }\n lua_pushlightuserdata(L, sentinel);\n lua_setfield(L, 2, name); /* _LOADED[name] = sentinel */\n lua_pushstring(L, name); /* pass name as argument to module */\n lua_call(L, 1, 1); /* run loaded module */\n if (!lua_isnil(L, -1)) /* non-nil return? */\n lua_setfield(L, 2, name); /* _LOADED[name] = returned value */\n lua_getfield(L, 2, name);\n if (lua_touserdata(L, -1) == sentinel) { /* module did not set a value? */\n lua_pushboolean(L, 1); /* use true as result */\n lua_pushvalue(L, -1); /* extra copy to be returned */\n lua_setfield(L, 2, name); /* _LOADED[name] = true */\n }\n return 1;\n}\n\n/* }====================================================== */\n\n\n\n/*\n** {======================================================\n** 'module' function\n** =======================================================\n*/\n \n\nstatic void setfenv (lua_State *L) {\n lua_Debug ar;\n if (lua_getstack(L, 1, &ar) == 0 ||\n lua_getinfo(L, \"f\", &ar) == 0 || /* get calling function */\n lua_iscfunction(L, -1))\n luaL_error(L, LUA_QL(\"module\") \" not called from a Lua function\");\n lua_pushvalue(L, -2);\n lua_setfenv(L, -2);\n lua_pop(L, 1);\n}\n\n\nstatic void dooptions (lua_State *L, int n) {\n int i;\n for (i = 2; i <= n; i++) {\n lua_pushvalue(L, i); /* get option (a function) */\n lua_pushvalue(L, -2); /* module */\n lua_call(L, 1, 0);\n }\n}\n\n\nstatic void modinit (lua_State *L, const char *modname) {\n const char *dot;\n lua_pushvalue(L, -1);\n lua_setfield(L, -2, \"_M\"); /* module._M = module */\n lua_pushstring(L, modname);\n lua_setfield(L, -2, \"_NAME\");\n dot = strrchr(modname, '.'); /* look for last dot in module name */\n if (dot == NULL) dot = modname;\n else dot++;\n /* set _PACKAGE as package name (full module name minus last part) */\n lua_pushlstring(L, modname, dot - modname);\n lua_setfield(L, -2, \"_PACKAGE\");\n}\n\n\nstatic int ll_module (lua_State *L) {\n const char *modname = luaL_checkstring(L, 1);\n int loaded = lua_gettop(L) + 1; /* index of _LOADED table */\n lua_getfield(L, LUA_REGISTRYINDEX, \"_LOADED\");\n lua_getfield(L, loaded, modname); /* get _LOADED[modname] */\n if (!lua_istable(L, -1)) { /* not found? */\n lua_pop(L, 1); /* remove previous result */\n /* try global variable (and create one if it does not exist) */\n if (luaL_findtable(L, LUA_GLOBALSINDEX, modname, 1) != NULL)\n return luaL_error(L, \"name conflict for module \" LUA_QS, modname);\n lua_pushvalue(L, -1);\n lua_setfield(L, loaded, modname); /* _LOADED[modname] = new table */\n }\n /* check whether table already has a _NAME field */\n lua_getfield(L, -1, \"_NAME\");\n if (!lua_isnil(L, -1)) /* is table an initialized module? */\n lua_pop(L, 1);\n else { /* no; initialize it */\n lua_pop(L, 1);\n modinit(L, modname);\n }\n lua_pushvalue(L, -1);\n setfenv(L);\n dooptions(L, loaded - 1);\n return 0;\n}\n\n\nstatic int ll_seeall (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n if (!lua_getmetatable(L, 1)) {\n lua_createtable(L, 0, 1); /* create new metatable */\n lua_pushvalue(L, -1);\n lua_setmetatable(L, 1);\n }\n lua_pushvalue(L, LUA_GLOBALSINDEX);\n lua_setfield(L, -2, \"__index\"); /* mt.__index = _G */\n return 0;\n}\n\n\n/* }====================================================== */\n\n\n\n/* auxiliary mark (for internal use) */\n#define AUXMARK\t\t\"\\1\"\n\nstatic void setpath (lua_State *L, const char *fieldname, const char *envname,\n const char *def) {\n const char *path = getenv(envname);\n if (path == NULL) /* no environment variable? */\n lua_pushstring(L, def); /* use default */\n else {\n /* replace \";;\" by \";AUXMARK;\" and then AUXMARK by default path */\n path = luaL_gsub(L, path, LUA_PATHSEP LUA_PATHSEP,\n LUA_PATHSEP AUXMARK LUA_PATHSEP);\n luaL_gsub(L, path, AUXMARK, def);\n lua_remove(L, -2);\n }\n setprogdir(L);\n lua_setfield(L, -2, fieldname);\n}\n\n\nstatic const luaL_Reg pk_funcs[] = {\n {\"loadlib\", ll_loadlib},\n {\"seeall\", ll_seeall},\n {NULL, NULL}\n};\n\n\nstatic const luaL_Reg ll_funcs[] = {\n {\"module\", ll_module},\n {\"require\", ll_require},\n {NULL, NULL}\n};\n\n\nstatic const lua_CFunction loaders[] =\n {loader_preload, loader_Lua, loader_C, loader_Croot, NULL};\n\n\nLUALIB_API int luaopen_package (lua_State *L) {\n int i;\n /* create new type _LOADLIB */\n luaL_newmetatable(L, \"_LOADLIB\");\n lua_pushcfunction(L, gctm);\n lua_setfield(L, -2, \"__gc\");\n /* create `package' table */\n luaL_register(L, LUA_LOADLIBNAME, pk_funcs);\n#if defined(LUA_COMPAT_LOADLIB) \n lua_getfield(L, -1, \"loadlib\");\n lua_setfield(L, LUA_GLOBALSINDEX, \"loadlib\");\n#endif\n lua_pushvalue(L, -1);\n lua_replace(L, LUA_ENVIRONINDEX);\n /* create `loaders' table */\n lua_createtable(L, sizeof(loaders)/sizeof(loaders[0]) - 1, 0);\n /* fill it with pre-defined loaders */\n for (i=0; loaders[i] != NULL; i++) {\n lua_pushcfunction(L, loaders[i]);\n lua_rawseti(L, -2, i+1);\n }\n lua_setfield(L, -2, \"loaders\"); /* put it in field `loaders' */\n setpath(L, \"path\", LUA_PATH, LUA_PATH_DEFAULT); /* set field `path' */\n setpath(L, \"cpath\", LUA_CPATH, LUA_CPATH_DEFAULT); /* set field `cpath' */\n /* store config information */\n lua_pushliteral(L, LUA_DIRSEP \"\\n\" LUA_PATHSEP \"\\n\" LUA_PATH_MARK \"\\n\"\n LUA_EXECDIR \"\\n\" LUA_IGMARK);\n lua_setfield(L, -2, \"config\");\n /* set field `loaded' */\n luaL_findtable(L, LUA_REGISTRYINDEX, \"_LOADED\", 2);\n lua_setfield(L, -2, \"loaded\");\n /* set field `preload' */\n lua_newtable(L);\n lua_setfield(L, -2, \"preload\");\n lua_pushvalue(L, LUA_GLOBALSINDEX);\n luaL_register(L, NULL, ll_funcs); /* open lib into global table */\n lua_pop(L, 1);\n return 1; /* return 'package' table */\n}\n\n" }, { "path": "deps/lua/src/lobject.c", "content": "/*\n** $Id: lobject.c,v 2.22.1.1 2007/12/27 13:02:25 roberto Exp $\n** Some generic functions over Lua objects\n** See Copyright Notice in lua.h\n*/\n\n#include \n#include \n#include \n#include \n#include \n\n#define lobject_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldo.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"lvm.h\"\n\n\n\nconst TValue luaO_nilobject_ = {{NULL}, LUA_TNIL};\n\n\n/*\n** converts an integer to a \"floating point byte\", represented as\n** (eeeeexxx), where the real value is (1xxx) * 2^(eeeee - 1) if\n** eeeee != 0 and (xxx) otherwise.\n*/\nint luaO_int2fb (unsigned int x) {\n int e = 0; /* expoent */\n while (x >= 16) {\n x = (x+1) >> 1;\n e++;\n }\n if (x < 8) return x;\n else return ((e+1) << 3) | (cast_int(x) - 8);\n}\n\n\n/* converts back */\nint luaO_fb2int (int x) {\n int e = (x >> 3) & 31;\n if (e == 0) return x;\n else return ((x & 7)+8) << (e - 1);\n}\n\n\nint luaO_log2 (unsigned int x) {\n static const lu_byte log_2[256] = {\n 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,\n 6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,\n 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,\n 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,\n 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\n 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\n 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,\n 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8\n };\n int l = -1;\n while (x >= 256) { l += 8; x >>= 8; }\n return l + log_2[x];\n\n}\n\n\nint luaO_rawequalObj (const TValue *t1, const TValue *t2) {\n if (ttype(t1) != ttype(t2)) return 0;\n else switch (ttype(t1)) {\n case LUA_TNIL:\n return 1;\n case LUA_TNUMBER:\n return luai_numeq(nvalue(t1), nvalue(t2));\n case LUA_TBOOLEAN:\n return bvalue(t1) == bvalue(t2); /* boolean true must be 1 !! */\n case LUA_TLIGHTUSERDATA:\n return pvalue(t1) == pvalue(t2);\n default:\n lua_assert(iscollectable(t1));\n return gcvalue(t1) == gcvalue(t2);\n }\n}\n\n\nint luaO_str2d (const char *s, lua_Number *result) {\n char *endptr;\n *result = lua_str2number(s, &endptr);\n if (endptr == s) return 0; /* conversion failed */\n if (*endptr == 'x' || *endptr == 'X') /* maybe an hexadecimal constant? */\n *result = cast_num(strtoul(s, &endptr, 16));\n if (*endptr == '\\0') return 1; /* most common case */\n while (isspace(cast(unsigned char, *endptr))) endptr++;\n if (*endptr != '\\0') return 0; /* invalid trailing characters? */\n return 1;\n}\n\n\n\nstatic void pushstr (lua_State *L, const char *str) {\n setsvalue2s(L, L->top, luaS_new(L, str));\n incr_top(L);\n}\n\n\n/* this function handles only `%d', `%c', %f, %p, and `%s' formats */\nconst char *luaO_pushvfstring (lua_State *L, const char *fmt, va_list argp) {\n int n = 1;\n pushstr(L, \"\");\n for (;;) {\n const char *e = strchr(fmt, '%');\n if (e == NULL) break;\n setsvalue2s(L, L->top, luaS_newlstr(L, fmt, e-fmt));\n incr_top(L);\n switch (*(e+1)) {\n case 's': {\n const char *s = va_arg(argp, char *);\n if (s == NULL) s = \"(null)\";\n pushstr(L, s);\n break;\n }\n case 'c': {\n char buff[2];\n buff[0] = cast(char, va_arg(argp, int));\n buff[1] = '\\0';\n pushstr(L, buff);\n break;\n }\n case 'd': {\n setnvalue(L->top, cast_num(va_arg(argp, int)));\n incr_top(L);\n break;\n }\n case 'f': {\n setnvalue(L->top, cast_num(va_arg(argp, l_uacNumber)));\n incr_top(L);\n break;\n }\n case 'p': {\n char buff[4*sizeof(void *) + 8]; /* should be enough space for a `%p' */\n sprintf(buff, \"%p\", va_arg(argp, void *));\n pushstr(L, buff);\n break;\n }\n case '%': {\n pushstr(L, \"%\");\n break;\n }\n default: {\n char buff[3];\n buff[0] = '%';\n buff[1] = *(e+1);\n buff[2] = '\\0';\n pushstr(L, buff);\n break;\n }\n }\n n += 2;\n fmt = e+2;\n }\n pushstr(L, fmt);\n luaV_concat(L, n+1, cast_int(L->top - L->base) - 1);\n L->top -= n;\n return svalue(L->top - 1);\n}\n\n\nconst char *luaO_pushfstring (lua_State *L, const char *fmt, ...) {\n const char *msg;\n va_list argp;\n va_start(argp, fmt);\n msg = luaO_pushvfstring(L, fmt, argp);\n va_end(argp);\n return msg;\n}\n\n\nvoid luaO_chunkid (char *out, const char *source, size_t bufflen) {\n if (*source == '=') {\n strncpy(out, source+1, bufflen); /* remove first char */\n out[bufflen-1] = '\\0'; /* ensures null termination */\n }\n else { /* out = \"source\", or \"...source\" */\n if (*source == '@') {\n size_t l;\n source++; /* skip the `@' */\n bufflen -= sizeof(\" '...' \");\n l = strlen(source);\n strcpy(out, \"\");\n if (l > bufflen) {\n source += (l-bufflen); /* get last part of file name */\n strcat(out, \"...\");\n }\n strcat(out, source);\n }\n else { /* out = [string \"string\"] */\n size_t len = strcspn(source, \"\\n\\r\"); /* stop at first newline */\n bufflen -= sizeof(\" [string \\\"...\\\"] \");\n if (len > bufflen) len = bufflen;\n strcpy(out, \"[string \\\"\");\n if (source[len] != '\\0') { /* must truncate? */\n strncat(out, source, len);\n strcat(out, \"...\");\n }\n else\n strcat(out, source);\n strcat(out, \"\\\"]\");\n }\n }\n}\n" }, { "path": "deps/lua/src/lobject.h", "content": "/*\n** $Id: lobject.h,v 2.20.1.2 2008/08/06 13:29:48 roberto Exp $\n** Type definitions for Lua objects\n** See Copyright Notice in lua.h\n*/\n\n\n#ifndef lobject_h\n#define lobject_h\n\n\n#include \n\n\n#include \"llimits.h\"\n#include \"lua.h\"\n\n\n/* tags for values visible from Lua */\n#define LAST_TAG\tLUA_TTHREAD\n\n#define NUM_TAGS\t(LAST_TAG+1)\n\n\n/*\n** Extra tags for non-values\n*/\n#define LUA_TPROTO\t(LAST_TAG+1)\n#define LUA_TUPVAL\t(LAST_TAG+2)\n#define LUA_TDEADKEY\t(LAST_TAG+3)\n\n\n/*\n** Union of all collectable objects\n*/\ntypedef union GCObject GCObject;\n\n\n/*\n** Common Header for all collectable objects (in macro form, to be\n** included in other objects)\n*/\n#define CommonHeader\tGCObject *next; lu_byte tt; lu_byte marked\n\n\n/*\n** Common header in struct form\n*/\ntypedef struct GCheader {\n CommonHeader;\n} GCheader;\n\n\n\n\n/*\n** Union of all Lua values\n*/\ntypedef union {\n GCObject *gc;\n void *p;\n lua_Number n;\n int b;\n} Value;\n\n\n/*\n** Tagged Values\n*/\n\n#define TValuefields\tValue value; int tt\n\ntypedef struct lua_TValue {\n TValuefields;\n} TValue;\n\n\n/* Macros to test type */\n#define ttisnil(o)\t(ttype(o) == LUA_TNIL)\n#define ttisnumber(o)\t(ttype(o) == LUA_TNUMBER)\n#define ttisstring(o)\t(ttype(o) == LUA_TSTRING)\n#define ttistable(o)\t(ttype(o) == LUA_TTABLE)\n#define ttisfunction(o)\t(ttype(o) == LUA_TFUNCTION)\n#define ttisboolean(o)\t(ttype(o) == LUA_TBOOLEAN)\n#define ttisuserdata(o)\t(ttype(o) == LUA_TUSERDATA)\n#define ttisthread(o)\t(ttype(o) == LUA_TTHREAD)\n#define ttislightuserdata(o)\t(ttype(o) == LUA_TLIGHTUSERDATA)\n\n/* Macros to access values */\n#define ttype(o)\t((o)->tt)\n#define gcvalue(o)\tcheck_exp(iscollectable(o), (o)->value.gc)\n#define pvalue(o)\tcheck_exp(ttislightuserdata(o), (o)->value.p)\n#define nvalue(o)\tcheck_exp(ttisnumber(o), (o)->value.n)\n#define rawtsvalue(o)\tcheck_exp(ttisstring(o), &(o)->value.gc->ts)\n#define tsvalue(o)\t(&rawtsvalue(o)->tsv)\n#define rawuvalue(o)\tcheck_exp(ttisuserdata(o), &(o)->value.gc->u)\n#define uvalue(o)\t(&rawuvalue(o)->uv)\n#define clvalue(o)\tcheck_exp(ttisfunction(o), &(o)->value.gc->cl)\n#define hvalue(o)\tcheck_exp(ttistable(o), &(o)->value.gc->h)\n#define bvalue(o)\tcheck_exp(ttisboolean(o), (o)->value.b)\n#define thvalue(o)\tcheck_exp(ttisthread(o), &(o)->value.gc->th)\n\n#define l_isfalse(o)\t(ttisnil(o) || (ttisboolean(o) && bvalue(o) == 0))\n\n/*\n** for internal debug only\n*/\n#define checkconsistency(obj) \\\n lua_assert(!iscollectable(obj) || (ttype(obj) == (obj)->value.gc->gch.tt))\n\n#define checkliveness(g,obj) \\\n lua_assert(!iscollectable(obj) || \\\n ((ttype(obj) == (obj)->value.gc->gch.tt) && !isdead(g, (obj)->value.gc)))\n\n\n/* Macros to set values */\n#define setnilvalue(obj) ((obj)->tt=LUA_TNIL)\n\n#define setnvalue(obj,x) \\\n { TValue *i_o=(obj); i_o->value.n=(x); i_o->tt=LUA_TNUMBER; }\n\n#define setpvalue(obj,x) \\\n { TValue *i_o=(obj); i_o->value.p=(x); i_o->tt=LUA_TLIGHTUSERDATA; }\n\n#define setbvalue(obj,x) \\\n { TValue *i_o=(obj); i_o->value.b=(x); i_o->tt=LUA_TBOOLEAN; }\n\n#define setsvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TSTRING; \\\n checkliveness(G(L),i_o); }\n\n#define setuvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TUSERDATA; \\\n checkliveness(G(L),i_o); }\n\n#define setthvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TTHREAD; \\\n checkliveness(G(L),i_o); }\n\n#define setclvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TFUNCTION; \\\n checkliveness(G(L),i_o); }\n\n#define sethvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TTABLE; \\\n checkliveness(G(L),i_o); }\n\n#define setptvalue(L,obj,x) \\\n { TValue *i_o=(obj); \\\n i_o->value.gc=cast(GCObject *, (x)); i_o->tt=LUA_TPROTO; \\\n checkliveness(G(L),i_o); }\n\n\n\n\n#define setobj(L,obj1,obj2) \\\n { const TValue *o2=(obj2); TValue *o1=(obj1); \\\n o1->value = o2->value; o1->tt=o2->tt; \\\n checkliveness(G(L),o1); }\n\n\n/*\n** different types of sets, according to destination\n*/\n\n/* from stack to (same) stack */\n#define setobjs2s\tsetobj\n/* to stack (not from same stack) */\n#define setobj2s\tsetobj\n#define setsvalue2s\tsetsvalue\n#define sethvalue2s\tsethvalue\n#define setptvalue2s\tsetptvalue\n/* from table to same table */\n#define setobjt2t\tsetobj\n/* to table */\n#define setobj2t\tsetobj\n/* to new object */\n#define setobj2n\tsetobj\n#define setsvalue2n\tsetsvalue\n\n#define setttype(obj, tt) (ttype(obj) = (tt))\n\n\n#define iscollectable(o)\t(ttype(o) >= LUA_TSTRING)\n\n\n\ntypedef TValue *StkId; /* index to stack elements */\n\n\n/*\n** String headers for string table\n*/\ntypedef union TString {\n L_Umaxalign dummy; /* ensures maximum alignment for strings */\n struct {\n CommonHeader;\n lu_byte reserved;\n unsigned int hash;\n size_t len;\n } tsv;\n} TString;\n\n\n#define getstr(ts)\tcast(const char *, (ts) + 1)\n#define svalue(o) getstr(rawtsvalue(o))\n\n\n\ntypedef union Udata {\n L_Umaxalign dummy; /* ensures maximum alignment for `local' udata */\n struct {\n CommonHeader;\n struct Table *metatable;\n struct Table *env;\n size_t len;\n } uv;\n} Udata;\n\n\n\n\n/*\n** Function Prototypes\n*/\ntypedef struct Proto {\n CommonHeader;\n TValue *k; /* constants used by the function */\n Instruction *code;\n struct Proto **p; /* functions defined inside the function */\n int *lineinfo; /* map from opcodes to source lines */\n struct LocVar *locvars; /* information about local variables */\n TString **upvalues; /* upvalue names */\n TString *source;\n int sizeupvalues;\n int sizek; /* size of `k' */\n int sizecode;\n int sizelineinfo;\n int sizep; /* size of `p' */\n int sizelocvars;\n int linedefined;\n int lastlinedefined;\n GCObject *gclist;\n lu_byte nups; /* number of upvalues */\n lu_byte numparams;\n lu_byte is_vararg;\n lu_byte maxstacksize;\n} Proto;\n\n\n/* masks for new-style vararg */\n#define VARARG_HASARG\t\t1\n#define VARARG_ISVARARG\t\t2\n#define VARARG_NEEDSARG\t\t4\n\n\ntypedef struct LocVar {\n TString *varname;\n int startpc; /* first point where variable is active */\n int endpc; /* first point where variable is dead */\n} LocVar;\n\n\n\n/*\n** Upvalues\n*/\n\ntypedef struct UpVal {\n CommonHeader;\n TValue *v; /* points to stack or to its own value */\n union {\n TValue value; /* the value (when closed) */\n struct { /* double linked list (when open) */\n struct UpVal *prev;\n struct UpVal *next;\n } l;\n } u;\n} UpVal;\n\n\n/*\n** Closures\n*/\n\n#define ClosureHeader \\\n\tCommonHeader; lu_byte isC; lu_byte nupvalues; GCObject *gclist; \\\n\tstruct Table *env\n\ntypedef struct CClosure {\n ClosureHeader;\n lua_CFunction f;\n TValue upvalue[1];\n} CClosure;\n\n\ntypedef struct LClosure {\n ClosureHeader;\n struct Proto *p;\n UpVal *upvals[1];\n} LClosure;\n\n\ntypedef union Closure {\n CClosure c;\n LClosure l;\n} Closure;\n\n\n#define iscfunction(o)\t(ttype(o) == LUA_TFUNCTION && clvalue(o)->c.isC)\n#define isLfunction(o)\t(ttype(o) == LUA_TFUNCTION && !clvalue(o)->c.isC)\n\n\n/*\n** Tables\n*/\n\ntypedef union TKey {\n struct {\n TValuefields;\n struct Node *next; /* for chaining */\n } nk;\n TValue tvk;\n} TKey;\n\n\ntypedef struct Node {\n TValue i_val;\n TKey i_key;\n} Node;\n\n\ntypedef struct Table {\n CommonHeader;\n lu_byte flags; /* 1<

lsizenode))\n\n\n#define luaO_nilobject\t\t(&luaO_nilobject_)\n\nLUAI_DATA const TValue luaO_nilobject_;\n\n#define ceillog2(x)\t(luaO_log2((x)-1) + 1)\n\nLUAI_FUNC int luaO_log2 (unsigned int x);\nLUAI_FUNC int luaO_int2fb (unsigned int x);\nLUAI_FUNC int luaO_fb2int (int x);\nLUAI_FUNC int luaO_rawequalObj (const TValue *t1, const TValue *t2);\nLUAI_FUNC int luaO_str2d (const char *s, lua_Number *result);\nLUAI_FUNC const char *luaO_pushvfstring (lua_State *L, const char *fmt,\n va_list argp);\nLUAI_FUNC const char *luaO_pushfstring (lua_State *L, const char *fmt, ...);\nLUAI_FUNC void luaO_chunkid (char *out, const char *source, size_t len);\n\n\n#endif\n\n" }, { "path": "deps/lua/src/lopcodes.c", "content": "/*\n** $Id: lopcodes.c,v 1.37.1.1 2007/12/27 13:02:25 roberto Exp $\n** See Copyright Notice in lua.h\n*/\n\n\n#define lopcodes_c\n#define LUA_CORE\n\n\n#include \"lopcodes.h\"\n\n\n/* ORDER OP */\n\nconst char *const luaP_opnames[NUM_OPCODES+1] = {\n \"MOVE\",\n \"LOADK\",\n \"LOADBOOL\",\n \"LOADNIL\",\n \"GETUPVAL\",\n \"GETGLOBAL\",\n \"GETTABLE\",\n \"SETGLOBAL\",\n \"SETUPVAL\",\n \"SETTABLE\",\n \"NEWTABLE\",\n \"SELF\",\n \"ADD\",\n \"SUB\",\n \"MUL\",\n \"DIV\",\n \"MOD\",\n \"POW\",\n \"UNM\",\n \"NOT\",\n \"LEN\",\n \"CONCAT\",\n \"JMP\",\n \"EQ\",\n \"LT\",\n \"LE\",\n \"TEST\",\n \"TESTSET\",\n \"CALL\",\n \"TAILCALL\",\n \"RETURN\",\n \"FORLOOP\",\n \"FORPREP\",\n \"TFORLOOP\",\n \"SETLIST\",\n \"CLOSE\",\n \"CLOSURE\",\n \"VARARG\",\n NULL\n};\n\n\n#define opmode(t,a,b,c,m) (((t)<<7) | ((a)<<6) | ((b)<<4) | ((c)<<2) | (m))\n\nconst lu_byte luaP_opmodes[NUM_OPCODES] = {\n/* T A B C mode\t\t opcode\t*/\n opmode(0, 1, OpArgR, OpArgN, iABC) \t\t/* OP_MOVE */\n ,opmode(0, 1, OpArgK, OpArgN, iABx)\t\t/* OP_LOADK */\n ,opmode(0, 1, OpArgU, OpArgU, iABC)\t\t/* OP_LOADBOOL */\n ,opmode(0, 1, OpArgR, OpArgN, iABC)\t\t/* OP_LOADNIL */\n ,opmode(0, 1, OpArgU, OpArgN, iABC)\t\t/* OP_GETUPVAL */\n ,opmode(0, 1, OpArgK, OpArgN, iABx)\t\t/* OP_GETGLOBAL */\n ,opmode(0, 1, OpArgR, OpArgK, iABC)\t\t/* OP_GETTABLE */\n ,opmode(0, 0, OpArgK, OpArgN, iABx)\t\t/* OP_SETGLOBAL */\n ,opmode(0, 0, OpArgU, OpArgN, iABC)\t\t/* OP_SETUPVAL */\n ,opmode(0, 0, OpArgK, OpArgK, iABC)\t\t/* OP_SETTABLE */\n ,opmode(0, 1, OpArgU, OpArgU, iABC)\t\t/* OP_NEWTABLE */\n ,opmode(0, 1, OpArgR, OpArgK, iABC)\t\t/* OP_SELF */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_ADD */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_SUB */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_MUL */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_DIV */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_MOD */\n ,opmode(0, 1, OpArgK, OpArgK, iABC)\t\t/* OP_POW */\n ,opmode(0, 1, OpArgR, OpArgN, iABC)\t\t/* OP_UNM */\n ,opmode(0, 1, OpArgR, OpArgN, iABC)\t\t/* OP_NOT */\n ,opmode(0, 1, OpArgR, OpArgN, iABC)\t\t/* OP_LEN */\n ,opmode(0, 1, OpArgR, OpArgR, iABC)\t\t/* OP_CONCAT */\n ,opmode(0, 0, OpArgR, OpArgN, iAsBx)\t\t/* OP_JMP */\n ,opmode(1, 0, OpArgK, OpArgK, iABC)\t\t/* OP_EQ */\n ,opmode(1, 0, OpArgK, OpArgK, iABC)\t\t/* OP_LT */\n ,opmode(1, 0, OpArgK, OpArgK, iABC)\t\t/* OP_LE */\n ,opmode(1, 1, OpArgR, OpArgU, iABC)\t\t/* OP_TEST */\n ,opmode(1, 1, OpArgR, OpArgU, iABC)\t\t/* OP_TESTSET */\n ,opmode(0, 1, OpArgU, OpArgU, iABC)\t\t/* OP_CALL */\n ,opmode(0, 1, OpArgU, OpArgU, iABC)\t\t/* OP_TAILCALL */\n ,opmode(0, 0, OpArgU, OpArgN, iABC)\t\t/* OP_RETURN */\n ,opmode(0, 1, OpArgR, OpArgN, iAsBx)\t\t/* OP_FORLOOP */\n ,opmode(0, 1, OpArgR, OpArgN, iAsBx)\t\t/* OP_FORPREP */\n ,opmode(1, 0, OpArgN, OpArgU, iABC)\t\t/* OP_TFORLOOP */\n ,opmode(0, 0, OpArgU, OpArgU, iABC)\t\t/* OP_SETLIST */\n ,opmode(0, 0, OpArgN, OpArgN, iABC)\t\t/* OP_CLOSE */\n ,opmode(0, 1, OpArgU, OpArgN, iABx)\t\t/* OP_CLOSURE */\n ,opmode(0, 1, OpArgU, OpArgN, iABC)\t\t/* OP_VARARG */\n};\n\n" }, { "path": "deps/lua/src/lopcodes.h", "content": "/*\n** $Id: lopcodes.h,v 1.125.1.1 2007/12/27 13:02:25 roberto Exp $\n** Opcodes for Lua virtual machine\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lopcodes_h\n#define lopcodes_h\n\n#include \"llimits.h\"\n\n\n/*===========================================================================\n We assume that instructions are unsigned numbers.\n All instructions have an opcode in the first 6 bits.\n Instructions can have the following fields:\n\t`A' : 8 bits\n\t`B' : 9 bits\n\t`C' : 9 bits\n\t`Bx' : 18 bits (`B' and `C' together)\n\t`sBx' : signed Bx\n\n A signed argument is represented in excess K; that is, the number\n value is the unsigned value minus K. K is exactly the maximum value\n for that argument (so that -max is represented by 0, and +max is\n represented by 2*max), which is half the maximum for the corresponding\n unsigned argument.\n===========================================================================*/\n\n\nenum OpMode {iABC, iABx, iAsBx}; /* basic instruction format */\n\n\n/*\n** size and position of opcode arguments.\n*/\n#define SIZE_C\t\t9\n#define SIZE_B\t\t9\n#define SIZE_Bx\t\t(SIZE_C + SIZE_B)\n#define SIZE_A\t\t8\n\n#define SIZE_OP\t\t6\n\n#define POS_OP\t\t0\n#define POS_A\t\t(POS_OP + SIZE_OP)\n#define POS_C\t\t(POS_A + SIZE_A)\n#define POS_B\t\t(POS_C + SIZE_C)\n#define POS_Bx\t\tPOS_C\n\n\n/*\n** limits for opcode arguments.\n** we use (signed) int to manipulate most arguments,\n** so they must fit in LUAI_BITSINT-1 bits (-1 for sign)\n*/\n#if SIZE_Bx < LUAI_BITSINT-1\n#define MAXARG_Bx ((1<>1) /* `sBx' is signed */\n#else\n#define MAXARG_Bx MAX_INT\n#define MAXARG_sBx MAX_INT\n#endif\n\n\n#define MAXARG_A ((1<>POS_OP) & MASK1(SIZE_OP,0)))\n#define SET_OPCODE(i,o)\t((i) = (((i)&MASK0(SIZE_OP,POS_OP)) | \\\n\t\t((cast(Instruction, o)<>POS_A) & MASK1(SIZE_A,0)))\n#define SETARG_A(i,u)\t((i) = (((i)&MASK0(SIZE_A,POS_A)) | \\\n\t\t((cast(Instruction, u)<>POS_B) & MASK1(SIZE_B,0)))\n#define SETARG_B(i,b)\t((i) = (((i)&MASK0(SIZE_B,POS_B)) | \\\n\t\t((cast(Instruction, b)<>POS_C) & MASK1(SIZE_C,0)))\n#define SETARG_C(i,b)\t((i) = (((i)&MASK0(SIZE_C,POS_C)) | \\\n\t\t((cast(Instruction, b)<>POS_Bx) & MASK1(SIZE_Bx,0)))\n#define SETARG_Bx(i,b)\t((i) = (((i)&MASK0(SIZE_Bx,POS_Bx)) | \\\n\t\t((cast(Instruction, b)< C) then pc++\t\t\t*/ \nOP_TESTSET,/*\tA B C\tif (R(B) <=> C) then R(A) := R(B) else pc++\t*/ \n\nOP_CALL,/*\tA B C\tR(A), ... ,R(A+C-2) := R(A)(R(A+1), ... ,R(A+B-1)) */\nOP_TAILCALL,/*\tA B C\treturn R(A)(R(A+1), ... ,R(A+B-1))\t\t*/\nOP_RETURN,/*\tA B\treturn R(A), ... ,R(A+B-2)\t(see note)\t*/\n\nOP_FORLOOP,/*\tA sBx\tR(A)+=R(A+2);\n\t\t\tif R(A) =) R(A)*/\nOP_CLOSURE,/*\tA Bx\tR(A) := closure(KPROTO[Bx], R(A), ... ,R(A+n))\t*/\n\nOP_VARARG/*\tA B\tR(A), R(A+1), ..., R(A+B-1) = vararg\t\t*/\n} OpCode;\n\n\n#define NUM_OPCODES\t(cast(int, OP_VARARG) + 1)\n\n\n\n/*===========================================================================\n Notes:\n (*) In OP_CALL, if (B == 0) then B = top. C is the number of returns - 1,\n and can be 0: OP_CALL then sets `top' to last_result+1, so\n next open instruction (OP_CALL, OP_RETURN, OP_SETLIST) may use `top'.\n\n (*) In OP_VARARG, if (B == 0) then use actual number of varargs and\n set top (like in OP_CALL with C == 0).\n\n (*) In OP_RETURN, if (B == 0) then return up to `top'\n\n (*) In OP_SETLIST, if (B == 0) then B = `top';\n if (C == 0) then next `instruction' is real C\n\n (*) For comparisons, A specifies what condition the test should accept\n (true or false).\n\n (*) All `skips' (pc++) assume that next instruction is a jump\n===========================================================================*/\n\n\n/*\n** masks for instruction properties. The format is:\n** bits 0-1: op mode\n** bits 2-3: C arg mode\n** bits 4-5: B arg mode\n** bit 6: instruction set register A\n** bit 7: operator is a test\n*/ \n\nenum OpArgMask {\n OpArgN, /* argument is not used */\n OpArgU, /* argument is used */\n OpArgR, /* argument is a register or a jump offset */\n OpArgK /* argument is a constant or register/constant */\n};\n\nLUAI_DATA const lu_byte luaP_opmodes[NUM_OPCODES];\n\n#define getOpMode(m)\t(cast(enum OpMode, luaP_opmodes[m] & 3))\n#define getBMode(m)\t(cast(enum OpArgMask, (luaP_opmodes[m] >> 4) & 3))\n#define getCMode(m)\t(cast(enum OpArgMask, (luaP_opmodes[m] >> 2) & 3))\n#define testAMode(m)\t(luaP_opmodes[m] & (1 << 6))\n#define testTMode(m)\t(luaP_opmodes[m] & (1 << 7))\n\n\nLUAI_DATA const char *const luaP_opnames[NUM_OPCODES+1]; /* opcode names */\n\n\n/* number of list items to accumulate before a SETLIST instruction */\n#define LFIELDS_PER_FLUSH\t50\n\n\n#endif\n" }, { "path": "deps/lua/src/loslib.c", "content": "/*\n** $Id: loslib.c,v 1.19.1.3 2008/01/18 16:38:18 roberto Exp $\n** Standard Operating System library\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n#include \n\n#define loslib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\nstatic int os_pushresult (lua_State *L, int i, const char *filename) {\n int en = errno; /* calls to Lua API may change this value */\n if (i) {\n lua_pushboolean(L, 1);\n return 1;\n }\n else {\n lua_pushnil(L);\n lua_pushfstring(L, \"%s: %s\", filename, strerror(en));\n lua_pushinteger(L, en);\n return 3;\n }\n}\n\n\nstatic int os_execute (lua_State *L) {\n lua_pushinteger(L, system(luaL_optstring(L, 1, NULL)));\n return 1;\n}\n\n\nstatic int os_remove (lua_State *L) {\n const char *filename = luaL_checkstring(L, 1);\n return os_pushresult(L, remove(filename) == 0, filename);\n}\n\n\nstatic int os_rename (lua_State *L) {\n const char *fromname = luaL_checkstring(L, 1);\n const char *toname = luaL_checkstring(L, 2);\n return os_pushresult(L, rename(fromname, toname) == 0, fromname);\n}\n\n\nstatic int os_tmpname (lua_State *L) {\n char buff[LUA_TMPNAMBUFSIZE];\n int err;\n lua_tmpnam(buff, err);\n if (err)\n return luaL_error(L, \"unable to generate a unique filename\");\n lua_pushstring(L, buff);\n return 1;\n}\n\n\nstatic int os_getenv (lua_State *L) {\n lua_pushstring(L, getenv(luaL_checkstring(L, 1))); /* if NULL push nil */\n return 1;\n}\n\n\nstatic int os_clock (lua_State *L) {\n lua_pushnumber(L, ((lua_Number)clock())/(lua_Number)CLOCKS_PER_SEC);\n return 1;\n}\n\n\n/*\n** {======================================================\n** Time/Date operations\n** { year=%Y, month=%m, day=%d, hour=%H, min=%M, sec=%S,\n** wday=%w+1, yday=%j, isdst=? }\n** =======================================================\n*/\n\nstatic void setfield (lua_State *L, const char *key, int value) {\n lua_pushinteger(L, value);\n lua_setfield(L, -2, key);\n}\n\nstatic void setboolfield (lua_State *L, const char *key, int value) {\n if (value < 0) /* undefined? */\n return; /* does not set field */\n lua_pushboolean(L, value);\n lua_setfield(L, -2, key);\n}\n\nstatic int getboolfield (lua_State *L, const char *key) {\n int res;\n lua_getfield(L, -1, key);\n res = lua_isnil(L, -1) ? -1 : lua_toboolean(L, -1);\n lua_pop(L, 1);\n return res;\n}\n\n\nstatic int getfield (lua_State *L, const char *key, int d) {\n int res;\n lua_getfield(L, -1, key);\n if (lua_isnumber(L, -1))\n res = (int)lua_tointeger(L, -1);\n else {\n if (d < 0)\n return luaL_error(L, \"field \" LUA_QS \" missing in date table\", key);\n res = d;\n }\n lua_pop(L, 1);\n return res;\n}\n\n\nstatic int os_date (lua_State *L) {\n const char *s = luaL_optstring(L, 1, \"%c\");\n time_t t = luaL_opt(L, (time_t)luaL_checknumber, 2, time(NULL));\n struct tm *stm;\n if (*s == '!') { /* UTC? */\n stm = gmtime(&t);\n s++; /* skip `!' */\n }\n else\n stm = localtime(&t);\n if (stm == NULL) /* invalid date? */\n lua_pushnil(L);\n else if (strcmp(s, \"*t\") == 0) {\n lua_createtable(L, 0, 9); /* 9 = number of fields */\n setfield(L, \"sec\", stm->tm_sec);\n setfield(L, \"min\", stm->tm_min);\n setfield(L, \"hour\", stm->tm_hour);\n setfield(L, \"day\", stm->tm_mday);\n setfield(L, \"month\", stm->tm_mon+1);\n setfield(L, \"year\", stm->tm_year+1900);\n setfield(L, \"wday\", stm->tm_wday+1);\n setfield(L, \"yday\", stm->tm_yday+1);\n setboolfield(L, \"isdst\", stm->tm_isdst);\n }\n else {\n char cc[3];\n luaL_Buffer b;\n cc[0] = '%'; cc[2] = '\\0';\n luaL_buffinit(L, &b);\n for (; *s; s++) {\n if (*s != '%' || *(s + 1) == '\\0') /* no conversion specifier? */\n luaL_addchar(&b, *s);\n else {\n size_t reslen;\n char buff[200]; /* should be big enough for any conversion result */\n cc[1] = *(++s);\n reslen = strftime(buff, sizeof(buff), cc, stm);\n luaL_addlstring(&b, buff, reslen);\n }\n }\n luaL_pushresult(&b);\n }\n return 1;\n}\n\n\nstatic int os_time (lua_State *L) {\n time_t t;\n if (lua_isnoneornil(L, 1)) /* called without args? */\n t = time(NULL); /* get current time */\n else {\n struct tm ts;\n luaL_checktype(L, 1, LUA_TTABLE);\n lua_settop(L, 1); /* make sure table is at the top */\n ts.tm_sec = getfield(L, \"sec\", 0);\n ts.tm_min = getfield(L, \"min\", 0);\n ts.tm_hour = getfield(L, \"hour\", 12);\n ts.tm_mday = getfield(L, \"day\", -1);\n ts.tm_mon = getfield(L, \"month\", -1) - 1;\n ts.tm_year = getfield(L, \"year\", -1) - 1900;\n ts.tm_isdst = getboolfield(L, \"isdst\");\n t = mktime(&ts);\n }\n if (t == (time_t)(-1))\n lua_pushnil(L);\n else\n lua_pushnumber(L, (lua_Number)t);\n return 1;\n}\n\n\nstatic int os_difftime (lua_State *L) {\n lua_pushnumber(L, difftime((time_t)(luaL_checknumber(L, 1)),\n (time_t)(luaL_optnumber(L, 2, 0))));\n return 1;\n}\n\n/* }====================================================== */\n\n\nstatic int os_setlocale (lua_State *L) {\n static const int cat[] = {LC_ALL, LC_COLLATE, LC_CTYPE, LC_MONETARY,\n LC_NUMERIC, LC_TIME};\n static const char *const catnames[] = {\"all\", \"collate\", \"ctype\", \"monetary\",\n \"numeric\", \"time\", NULL};\n const char *l = luaL_optstring(L, 1, NULL);\n int op = luaL_checkoption(L, 2, \"all\", catnames);\n lua_pushstring(L, setlocale(cat[op], l));\n return 1;\n}\n\n\nstatic int os_exit (lua_State *L) {\n exit(luaL_optint(L, 1, EXIT_SUCCESS));\n}\n\nstatic const luaL_Reg syslib[] = {\n {\"clock\", os_clock},\n {\"date\", os_date},\n {\"difftime\", os_difftime},\n {\"execute\", os_execute},\n {\"exit\", os_exit},\n {\"getenv\", os_getenv},\n {\"remove\", os_remove},\n {\"rename\", os_rename},\n {\"setlocale\", os_setlocale},\n {\"time\", os_time},\n {\"tmpname\", os_tmpname},\n {NULL, NULL}\n};\n\n/* }====================================================== */\n\n\n\nLUALIB_API int luaopen_os (lua_State *L) {\n luaL_register(L, LUA_OSLIBNAME, syslib);\n return 1;\n}\n\n" }, { "path": "deps/lua/src/lparser.c", "content": "/*\n** $Id: lparser.c,v 2.42.1.4 2011/10/21 19:31:42 roberto Exp $\n** Lua Parser\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lparser_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lcode.h\"\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"llex.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lparser.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n\n\n\n#define hasmultret(k)\t\t((k) == VCALL || (k) == VVARARG)\n\n#define getlocvar(fs, i)\t((fs)->f->locvars[(fs)->actvar[i]])\n\n#define luaY_checklimit(fs,v,l,m)\tif ((v)>(l)) errorlimit(fs,l,m)\n\n\n/*\n** nodes for block list (list of active blocks)\n*/\ntypedef struct BlockCnt {\n struct BlockCnt *previous; /* chain */\n int breaklist; /* list of jumps out of this loop */\n lu_byte nactvar; /* # active locals outside the breakable structure */\n lu_byte upval; /* true if some variable in the block is an upvalue */\n lu_byte isbreakable; /* true if `block' is a loop */\n} BlockCnt;\n\n\n\n/*\n** prototypes for recursive non-terminal functions\n*/\nstatic void chunk (LexState *ls);\nstatic void expr (LexState *ls, expdesc *v);\n\n\nstatic void anchor_token (LexState *ls) {\n if (ls->t.token == TK_NAME || ls->t.token == TK_STRING) {\n TString *ts = ls->t.seminfo.ts;\n luaX_newstring(ls, getstr(ts), ts->tsv.len);\n }\n}\n\n\nstatic void error_expected (LexState *ls, int token) {\n luaX_syntaxerror(ls,\n luaO_pushfstring(ls->L, LUA_QS \" expected\", luaX_token2str(ls, token)));\n}\n\n\nstatic void errorlimit (FuncState *fs, int limit, const char *what) {\n const char *msg = (fs->f->linedefined == 0) ?\n luaO_pushfstring(fs->L, \"main function has more than %d %s\", limit, what) :\n luaO_pushfstring(fs->L, \"function at line %d has more than %d %s\",\n fs->f->linedefined, limit, what);\n luaX_lexerror(fs->ls, msg, 0);\n}\n\n\nstatic int testnext (LexState *ls, int c) {\n if (ls->t.token == c) {\n luaX_next(ls);\n return 1;\n }\n else return 0;\n}\n\n\nstatic void check (LexState *ls, int c) {\n if (ls->t.token != c)\n error_expected(ls, c);\n}\n\nstatic void checknext (LexState *ls, int c) {\n check(ls, c);\n luaX_next(ls);\n}\n\n\n#define check_condition(ls,c,msg)\t{ if (!(c)) luaX_syntaxerror(ls, msg); }\n\n\n\nstatic void check_match (LexState *ls, int what, int who, int where) {\n if (!testnext(ls, what)) {\n if (where == ls->linenumber)\n error_expected(ls, what);\n else {\n luaX_syntaxerror(ls, luaO_pushfstring(ls->L,\n LUA_QS \" expected (to close \" LUA_QS \" at line %d)\",\n luaX_token2str(ls, what), luaX_token2str(ls, who), where));\n }\n }\n}\n\n\nstatic TString *str_checkname (LexState *ls) {\n TString *ts;\n check(ls, TK_NAME);\n ts = ls->t.seminfo.ts;\n luaX_next(ls);\n return ts;\n}\n\n\nstatic void init_exp (expdesc *e, expkind k, int i) {\n e->f = e->t = NO_JUMP;\n e->k = k;\n e->u.s.info = i;\n}\n\n\nstatic void codestring (LexState *ls, expdesc *e, TString *s) {\n init_exp(e, VK, luaK_stringK(ls->fs, s));\n}\n\n\nstatic void checkname(LexState *ls, expdesc *e) {\n codestring(ls, e, str_checkname(ls));\n}\n\n\nstatic int registerlocalvar (LexState *ls, TString *varname) {\n FuncState *fs = ls->fs;\n Proto *f = fs->f;\n int oldsize = f->sizelocvars;\n luaM_growvector(ls->L, f->locvars, fs->nlocvars, f->sizelocvars,\n LocVar, SHRT_MAX, \"too many local variables\");\n while (oldsize < f->sizelocvars) f->locvars[oldsize++].varname = NULL;\n f->locvars[fs->nlocvars].varname = varname;\n luaC_objbarrier(ls->L, f, varname);\n return fs->nlocvars++;\n}\n\n\n#define new_localvarliteral(ls,v,n) \\\n new_localvar(ls, luaX_newstring(ls, \"\" v, (sizeof(v)/sizeof(char))-1), n)\n\n\nstatic void new_localvar (LexState *ls, TString *name, int n) {\n FuncState *fs = ls->fs;\n luaY_checklimit(fs, fs->nactvar+n+1, LUAI_MAXVARS, \"local variables\");\n fs->actvar[fs->nactvar+n] = cast(unsigned short, registerlocalvar(ls, name));\n}\n\n\nstatic void adjustlocalvars (LexState *ls, int nvars) {\n FuncState *fs = ls->fs;\n fs->nactvar = cast_byte(fs->nactvar + nvars);\n for (; nvars; nvars--) {\n getlocvar(fs, fs->nactvar - nvars).startpc = fs->pc;\n }\n}\n\n\nstatic void removevars (LexState *ls, int tolevel) {\n FuncState *fs = ls->fs;\n while (fs->nactvar > tolevel)\n getlocvar(fs, --fs->nactvar).endpc = fs->pc;\n}\n\n\nstatic int indexupvalue (FuncState *fs, TString *name, expdesc *v) {\n int i;\n Proto *f = fs->f;\n int oldsize = f->sizeupvalues;\n for (i=0; inups; i++) {\n if (fs->upvalues[i].k == v->k && fs->upvalues[i].info == v->u.s.info) {\n lua_assert(f->upvalues[i] == name);\n return i;\n }\n }\n /* new one */\n luaY_checklimit(fs, f->nups + 1, LUAI_MAXUPVALUES, \"upvalues\");\n luaM_growvector(fs->L, f->upvalues, f->nups, f->sizeupvalues,\n TString *, MAX_INT, \"\");\n while (oldsize < f->sizeupvalues) f->upvalues[oldsize++] = NULL;\n f->upvalues[f->nups] = name;\n luaC_objbarrier(fs->L, f, name);\n lua_assert(v->k == VLOCAL || v->k == VUPVAL);\n fs->upvalues[f->nups].k = cast_byte(v->k);\n fs->upvalues[f->nups].info = cast_byte(v->u.s.info);\n return f->nups++;\n}\n\n\nstatic int searchvar (FuncState *fs, TString *n) {\n int i;\n for (i=fs->nactvar-1; i >= 0; i--) {\n if (n == getlocvar(fs, i).varname)\n return i;\n }\n return -1; /* not found */\n}\n\n\nstatic void markupval (FuncState *fs, int level) {\n BlockCnt *bl = fs->bl;\n while (bl && bl->nactvar > level) bl = bl->previous;\n if (bl) bl->upval = 1;\n}\n\n\nstatic int singlevaraux (FuncState *fs, TString *n, expdesc *var, int base) {\n if (fs == NULL) { /* no more levels? */\n init_exp(var, VGLOBAL, NO_REG); /* default is global variable */\n return VGLOBAL;\n }\n else {\n int v = searchvar(fs, n); /* look up at current level */\n if (v >= 0) {\n init_exp(var, VLOCAL, v);\n if (!base)\n markupval(fs, v); /* local will be used as an upval */\n return VLOCAL;\n }\n else { /* not found at current level; try upper one */\n if (singlevaraux(fs->prev, n, var, 0) == VGLOBAL)\n return VGLOBAL;\n var->u.s.info = indexupvalue(fs, n, var); /* else was LOCAL or UPVAL */\n var->k = VUPVAL; /* upvalue in this level */\n return VUPVAL;\n }\n }\n}\n\n\nstatic void singlevar (LexState *ls, expdesc *var) {\n TString *varname = str_checkname(ls);\n FuncState *fs = ls->fs;\n if (singlevaraux(fs, varname, var, 1) == VGLOBAL)\n var->u.s.info = luaK_stringK(fs, varname); /* info points to global name */\n}\n\n\nstatic void adjust_assign (LexState *ls, int nvars, int nexps, expdesc *e) {\n FuncState *fs = ls->fs;\n int extra = nvars - nexps;\n if (hasmultret(e->k)) {\n extra++; /* includes call itself */\n if (extra < 0) extra = 0;\n luaK_setreturns(fs, e, extra); /* last exp. provides the difference */\n if (extra > 1) luaK_reserveregs(fs, extra-1);\n }\n else {\n if (e->k != VVOID) luaK_exp2nextreg(fs, e); /* close last expression */\n if (extra > 0) {\n int reg = fs->freereg;\n luaK_reserveregs(fs, extra);\n luaK_nil(fs, reg, extra);\n }\n }\n}\n\n\nstatic void enterlevel (LexState *ls) {\n if (++ls->L->nCcalls > LUAI_MAXCCALLS)\n\tluaX_lexerror(ls, \"chunk has too many syntax levels\", 0);\n}\n\n\n#define leavelevel(ls)\t((ls)->L->nCcalls--)\n\n\nstatic void enterblock (FuncState *fs, BlockCnt *bl, lu_byte isbreakable) {\n bl->breaklist = NO_JUMP;\n bl->isbreakable = isbreakable;\n bl->nactvar = fs->nactvar;\n bl->upval = 0;\n bl->previous = fs->bl;\n fs->bl = bl;\n lua_assert(fs->freereg == fs->nactvar);\n}\n\n\nstatic void leaveblock (FuncState *fs) {\n BlockCnt *bl = fs->bl;\n fs->bl = bl->previous;\n removevars(fs->ls, bl->nactvar);\n if (bl->upval)\n luaK_codeABC(fs, OP_CLOSE, bl->nactvar, 0, 0);\n /* a block either controls scope or breaks (never both) */\n lua_assert(!bl->isbreakable || !bl->upval);\n lua_assert(bl->nactvar == fs->nactvar);\n fs->freereg = fs->nactvar; /* free registers */\n luaK_patchtohere(fs, bl->breaklist);\n}\n\n\nstatic void pushclosure (LexState *ls, FuncState *func, expdesc *v) {\n FuncState *fs = ls->fs;\n Proto *f = fs->f;\n int oldsize = f->sizep;\n int i;\n luaM_growvector(ls->L, f->p, fs->np, f->sizep, Proto *,\n MAXARG_Bx, \"constant table overflow\");\n while (oldsize < f->sizep) f->p[oldsize++] = NULL;\n f->p[fs->np++] = func->f;\n luaC_objbarrier(ls->L, f, func->f);\n init_exp(v, VRELOCABLE, luaK_codeABx(fs, OP_CLOSURE, 0, fs->np-1));\n for (i=0; if->nups; i++) {\n OpCode o = (func->upvalues[i].k == VLOCAL) ? OP_MOVE : OP_GETUPVAL;\n luaK_codeABC(fs, o, 0, func->upvalues[i].info, 0);\n }\n}\n\n\nstatic void open_func (LexState *ls, FuncState *fs) {\n lua_State *L = ls->L;\n Proto *f = luaF_newproto(L);\n fs->f = f;\n fs->prev = ls->fs; /* linked list of funcstates */\n fs->ls = ls;\n fs->L = L;\n ls->fs = fs;\n fs->pc = 0;\n fs->lasttarget = -1;\n fs->jpc = NO_JUMP;\n fs->freereg = 0;\n fs->nk = 0;\n fs->np = 0;\n fs->nlocvars = 0;\n fs->nactvar = 0;\n fs->bl = NULL;\n f->source = ls->source;\n f->maxstacksize = 2; /* registers 0/1 are always valid */\n fs->h = luaH_new(L, 0, 0);\n /* anchor table of constants and prototype (to avoid being collected) */\n sethvalue2s(L, L->top, fs->h);\n incr_top(L);\n setptvalue2s(L, L->top, f);\n incr_top(L);\n}\n\n\nstatic void close_func (LexState *ls) {\n lua_State *L = ls->L;\n FuncState *fs = ls->fs;\n Proto *f = fs->f;\n removevars(ls, 0);\n luaK_ret(fs, 0, 0); /* final return */\n luaM_reallocvector(L, f->code, f->sizecode, fs->pc, Instruction);\n f->sizecode = fs->pc;\n luaM_reallocvector(L, f->lineinfo, f->sizelineinfo, fs->pc, int);\n f->sizelineinfo = fs->pc;\n luaM_reallocvector(L, f->k, f->sizek, fs->nk, TValue);\n f->sizek = fs->nk;\n luaM_reallocvector(L, f->p, f->sizep, fs->np, Proto *);\n f->sizep = fs->np;\n luaM_reallocvector(L, f->locvars, f->sizelocvars, fs->nlocvars, LocVar);\n f->sizelocvars = fs->nlocvars;\n luaM_reallocvector(L, f->upvalues, f->sizeupvalues, f->nups, TString *);\n f->sizeupvalues = f->nups;\n lua_assert(luaG_checkcode(f));\n lua_assert(fs->bl == NULL);\n ls->fs = fs->prev;\n /* last token read was anchored in defunct function; must reanchor it */\n if (fs) anchor_token(ls);\n L->top -= 2; /* remove table and prototype from the stack */\n}\n\n\nProto *luaY_parser (lua_State *L, ZIO *z, Mbuffer *buff, const char *name) {\n struct LexState lexstate;\n struct FuncState funcstate;\n lexstate.buff = buff;\n luaX_setinput(L, &lexstate, z, luaS_new(L, name));\n open_func(&lexstate, &funcstate);\n funcstate.f->is_vararg = VARARG_ISVARARG; /* main func. is always vararg */\n luaX_next(&lexstate); /* read first token */\n chunk(&lexstate);\n check(&lexstate, TK_EOS);\n close_func(&lexstate);\n lua_assert(funcstate.prev == NULL);\n lua_assert(funcstate.f->nups == 0);\n lua_assert(lexstate.fs == NULL);\n return funcstate.f;\n}\n\n\n\n/*============================================================*/\n/* GRAMMAR RULES */\n/*============================================================*/\n\n\nstatic void field (LexState *ls, expdesc *v) {\n /* field -> ['.' | ':'] NAME */\n FuncState *fs = ls->fs;\n expdesc key;\n luaK_exp2anyreg(fs, v);\n luaX_next(ls); /* skip the dot or colon */\n checkname(ls, &key);\n luaK_indexed(fs, v, &key);\n}\n\n\nstatic void yindex (LexState *ls, expdesc *v) {\n /* index -> '[' expr ']' */\n luaX_next(ls); /* skip the '[' */\n expr(ls, v);\n luaK_exp2val(ls->fs, v);\n checknext(ls, ']');\n}\n\n\n/*\n** {======================================================================\n** Rules for Constructors\n** =======================================================================\n*/\n\n\nstruct ConsControl {\n expdesc v; /* last list item read */\n expdesc *t; /* table descriptor */\n int nh; /* total number of `record' elements */\n int na; /* total number of array elements */\n int tostore; /* number of array elements pending to be stored */\n};\n\n\nstatic void recfield (LexState *ls, struct ConsControl *cc) {\n /* recfield -> (NAME | `['exp1`]') = exp1 */\n FuncState *fs = ls->fs;\n int reg = ls->fs->freereg;\n expdesc key, val;\n int rkkey;\n if (ls->t.token == TK_NAME) {\n luaY_checklimit(fs, cc->nh, MAX_INT, \"items in a constructor\");\n checkname(ls, &key);\n }\n else /* ls->t.token == '[' */\n yindex(ls, &key);\n cc->nh++;\n checknext(ls, '=');\n rkkey = luaK_exp2RK(fs, &key);\n expr(ls, &val);\n luaK_codeABC(fs, OP_SETTABLE, cc->t->u.s.info, rkkey, luaK_exp2RK(fs, &val));\n fs->freereg = reg; /* free registers */\n}\n\n\nstatic void closelistfield (FuncState *fs, struct ConsControl *cc) {\n if (cc->v.k == VVOID) return; /* there is no list item */\n luaK_exp2nextreg(fs, &cc->v);\n cc->v.k = VVOID;\n if (cc->tostore == LFIELDS_PER_FLUSH) {\n luaK_setlist(fs, cc->t->u.s.info, cc->na, cc->tostore); /* flush */\n cc->tostore = 0; /* no more items pending */\n }\n}\n\n\nstatic void lastlistfield (FuncState *fs, struct ConsControl *cc) {\n if (cc->tostore == 0) return;\n if (hasmultret(cc->v.k)) {\n luaK_setmultret(fs, &cc->v);\n luaK_setlist(fs, cc->t->u.s.info, cc->na, LUA_MULTRET);\n cc->na--; /* do not count last expression (unknown number of elements) */\n }\n else {\n if (cc->v.k != VVOID)\n luaK_exp2nextreg(fs, &cc->v);\n luaK_setlist(fs, cc->t->u.s.info, cc->na, cc->tostore);\n }\n}\n\n\nstatic void listfield (LexState *ls, struct ConsControl *cc) {\n expr(ls, &cc->v);\n luaY_checklimit(ls->fs, cc->na, MAX_INT, \"items in a constructor\");\n cc->na++;\n cc->tostore++;\n}\n\n\nstatic void constructor (LexState *ls, expdesc *t) {\n /* constructor -> ?? */\n FuncState *fs = ls->fs;\n int line = ls->linenumber;\n int pc = luaK_codeABC(fs, OP_NEWTABLE, 0, 0, 0);\n struct ConsControl cc;\n cc.na = cc.nh = cc.tostore = 0;\n cc.t = t;\n init_exp(t, VRELOCABLE, pc);\n init_exp(&cc.v, VVOID, 0); /* no value (yet) */\n luaK_exp2nextreg(ls->fs, t); /* fix it at stack top (for gc) */\n checknext(ls, '{');\n do {\n lua_assert(cc.v.k == VVOID || cc.tostore > 0);\n if (ls->t.token == '}') break;\n closelistfield(fs, &cc);\n switch(ls->t.token) {\n case TK_NAME: { /* may be listfields or recfields */\n luaX_lookahead(ls);\n if (ls->lookahead.token != '=') /* expression? */\n listfield(ls, &cc);\n else\n recfield(ls, &cc);\n break;\n }\n case '[': { /* constructor_item -> recfield */\n recfield(ls, &cc);\n break;\n }\n default: { /* constructor_part -> listfield */\n listfield(ls, &cc);\n break;\n }\n }\n } while (testnext(ls, ',') || testnext(ls, ';'));\n check_match(ls, '}', '{', line);\n lastlistfield(fs, &cc);\n SETARG_B(fs->f->code[pc], luaO_int2fb(cc.na)); /* set initial array size */\n SETARG_C(fs->f->code[pc], luaO_int2fb(cc.nh)); /* set initial table size */\n}\n\n/* }====================================================================== */\n\n\n\nstatic void parlist (LexState *ls) {\n /* parlist -> [ param { `,' param } ] */\n FuncState *fs = ls->fs;\n Proto *f = fs->f;\n int nparams = 0;\n f->is_vararg = 0;\n if (ls->t.token != ')') { /* is `parlist' not empty? */\n do {\n switch (ls->t.token) {\n case TK_NAME: { /* param -> NAME */\n new_localvar(ls, str_checkname(ls), nparams++);\n break;\n }\n case TK_DOTS: { /* param -> `...' */\n luaX_next(ls);\n#if defined(LUA_COMPAT_VARARG)\n /* use `arg' as default name */\n new_localvarliteral(ls, \"arg\", nparams++);\n f->is_vararg = VARARG_HASARG | VARARG_NEEDSARG;\n#endif\n f->is_vararg |= VARARG_ISVARARG;\n break;\n }\n default: luaX_syntaxerror(ls, \" or \" LUA_QL(\"...\") \" expected\");\n }\n } while (!f->is_vararg && testnext(ls, ','));\n }\n adjustlocalvars(ls, nparams);\n f->numparams = cast_byte(fs->nactvar - (f->is_vararg & VARARG_HASARG));\n luaK_reserveregs(fs, fs->nactvar); /* reserve register for parameters */\n}\n\n\nstatic void body (LexState *ls, expdesc *e, int needself, int line) {\n /* body -> `(' parlist `)' chunk END */\n FuncState new_fs;\n open_func(ls, &new_fs);\n new_fs.f->linedefined = line;\n checknext(ls, '(');\n if (needself) {\n new_localvarliteral(ls, \"self\", 0);\n adjustlocalvars(ls, 1);\n }\n parlist(ls);\n checknext(ls, ')');\n chunk(ls);\n new_fs.f->lastlinedefined = ls->linenumber;\n check_match(ls, TK_END, TK_FUNCTION, line);\n close_func(ls);\n pushclosure(ls, &new_fs, e);\n}\n\n\nstatic int explist1 (LexState *ls, expdesc *v) {\n /* explist1 -> expr { `,' expr } */\n int n = 1; /* at least one expression */\n expr(ls, v);\n while (testnext(ls, ',')) {\n luaK_exp2nextreg(ls->fs, v);\n expr(ls, v);\n n++;\n }\n return n;\n}\n\n\nstatic void funcargs (LexState *ls, expdesc *f) {\n FuncState *fs = ls->fs;\n expdesc args;\n int base, nparams;\n int line = ls->linenumber;\n switch (ls->t.token) {\n case '(': { /* funcargs -> `(' [ explist1 ] `)' */\n if (line != ls->lastline)\n luaX_syntaxerror(ls,\"ambiguous syntax (function call x new statement)\");\n luaX_next(ls);\n if (ls->t.token == ')') /* arg list is empty? */\n args.k = VVOID;\n else {\n explist1(ls, &args);\n luaK_setmultret(fs, &args);\n }\n check_match(ls, ')', '(', line);\n break;\n }\n case '{': { /* funcargs -> constructor */\n constructor(ls, &args);\n break;\n }\n case TK_STRING: { /* funcargs -> STRING */\n codestring(ls, &args, ls->t.seminfo.ts);\n luaX_next(ls); /* must use `seminfo' before `next' */\n break;\n }\n default: {\n luaX_syntaxerror(ls, \"function arguments expected\");\n return;\n }\n }\n lua_assert(f->k == VNONRELOC);\n base = f->u.s.info; /* base register for call */\n if (hasmultret(args.k))\n nparams = LUA_MULTRET; /* open call */\n else {\n if (args.k != VVOID)\n luaK_exp2nextreg(fs, &args); /* close last argument */\n nparams = fs->freereg - (base+1);\n }\n init_exp(f, VCALL, luaK_codeABC(fs, OP_CALL, base, nparams+1, 2));\n luaK_fixline(fs, line);\n fs->freereg = base+1; /* call remove function and arguments and leaves\n (unless changed) one result */\n}\n\n\n\n\n/*\n** {======================================================================\n** Expression parsing\n** =======================================================================\n*/\n\n\nstatic void prefixexp (LexState *ls, expdesc *v) {\n /* prefixexp -> NAME | '(' expr ')' */\n switch (ls->t.token) {\n case '(': {\n int line = ls->linenumber;\n luaX_next(ls);\n expr(ls, v);\n check_match(ls, ')', '(', line);\n luaK_dischargevars(ls->fs, v);\n return;\n }\n case TK_NAME: {\n singlevar(ls, v);\n return;\n }\n default: {\n luaX_syntaxerror(ls, \"unexpected symbol\");\n return;\n }\n }\n}\n\n\nstatic void primaryexp (LexState *ls, expdesc *v) {\n /* primaryexp ->\n prefixexp { `.' NAME | `[' exp `]' | `:' NAME funcargs | funcargs } */\n FuncState *fs = ls->fs;\n prefixexp(ls, v);\n for (;;) {\n switch (ls->t.token) {\n case '.': { /* field */\n field(ls, v);\n break;\n }\n case '[': { /* `[' exp1 `]' */\n expdesc key;\n luaK_exp2anyreg(fs, v);\n yindex(ls, &key);\n luaK_indexed(fs, v, &key);\n break;\n }\n case ':': { /* `:' NAME funcargs */\n expdesc key;\n luaX_next(ls);\n checkname(ls, &key);\n luaK_self(fs, v, &key);\n funcargs(ls, v);\n break;\n }\n case '(': case TK_STRING: case '{': { /* funcargs */\n luaK_exp2nextreg(fs, v);\n funcargs(ls, v);\n break;\n }\n default: return;\n }\n }\n}\n\n\nstatic void simpleexp (LexState *ls, expdesc *v) {\n /* simpleexp -> NUMBER | STRING | NIL | true | false | ... |\n constructor | FUNCTION body | primaryexp */\n switch (ls->t.token) {\n case TK_NUMBER: {\n init_exp(v, VKNUM, 0);\n v->u.nval = ls->t.seminfo.r;\n break;\n }\n case TK_STRING: {\n codestring(ls, v, ls->t.seminfo.ts);\n break;\n }\n case TK_NIL: {\n init_exp(v, VNIL, 0);\n break;\n }\n case TK_TRUE: {\n init_exp(v, VTRUE, 0);\n break;\n }\n case TK_FALSE: {\n init_exp(v, VFALSE, 0);\n break;\n }\n case TK_DOTS: { /* vararg */\n FuncState *fs = ls->fs;\n check_condition(ls, fs->f->is_vararg,\n \"cannot use \" LUA_QL(\"...\") \" outside a vararg function\");\n fs->f->is_vararg &= ~VARARG_NEEDSARG; /* don't need 'arg' */\n init_exp(v, VVARARG, luaK_codeABC(fs, OP_VARARG, 0, 1, 0));\n break;\n }\n case '{': { /* constructor */\n constructor(ls, v);\n return;\n }\n case TK_FUNCTION: {\n luaX_next(ls);\n body(ls, v, 0, ls->linenumber);\n return;\n }\n default: {\n primaryexp(ls, v);\n return;\n }\n }\n luaX_next(ls);\n}\n\n\nstatic UnOpr getunopr (int op) {\n switch (op) {\n case TK_NOT: return OPR_NOT;\n case '-': return OPR_MINUS;\n case '#': return OPR_LEN;\n default: return OPR_NOUNOPR;\n }\n}\n\n\nstatic BinOpr getbinopr (int op) {\n switch (op) {\n case '+': return OPR_ADD;\n case '-': return OPR_SUB;\n case '*': return OPR_MUL;\n case '/': return OPR_DIV;\n case '%': return OPR_MOD;\n case '^': return OPR_POW;\n case TK_CONCAT: return OPR_CONCAT;\n case TK_NE: return OPR_NE;\n case TK_EQ: return OPR_EQ;\n case '<': return OPR_LT;\n case TK_LE: return OPR_LE;\n case '>': return OPR_GT;\n case TK_GE: return OPR_GE;\n case TK_AND: return OPR_AND;\n case TK_OR: return OPR_OR;\n default: return OPR_NOBINOPR;\n }\n}\n\n\nstatic const struct {\n lu_byte left; /* left priority for each binary operator */\n lu_byte right; /* right priority */\n} priority[] = { /* ORDER OPR */\n {6, 6}, {6, 6}, {7, 7}, {7, 7}, {7, 7}, /* `+' `-' `/' `%' */\n {10, 9}, {5, 4}, /* power and concat (right associative) */\n {3, 3}, {3, 3}, /* equality and inequality */\n {3, 3}, {3, 3}, {3, 3}, {3, 3}, /* order */\n {2, 2}, {1, 1} /* logical (and/or) */\n};\n\n#define UNARY_PRIORITY\t8 /* priority for unary operators */\n\n\n/*\n** subexpr -> (simpleexp | unop subexpr) { binop subexpr }\n** where `binop' is any binary operator with a priority higher than `limit'\n*/\nstatic BinOpr subexpr (LexState *ls, expdesc *v, unsigned int limit) {\n BinOpr op;\n UnOpr uop;\n enterlevel(ls);\n uop = getunopr(ls->t.token);\n if (uop != OPR_NOUNOPR) {\n luaX_next(ls);\n subexpr(ls, v, UNARY_PRIORITY);\n luaK_prefix(ls->fs, uop, v);\n }\n else simpleexp(ls, v);\n /* expand while operators have priorities higher than `limit' */\n op = getbinopr(ls->t.token);\n while (op != OPR_NOBINOPR && priority[op].left > limit) {\n expdesc v2;\n BinOpr nextop;\n luaX_next(ls);\n luaK_infix(ls->fs, op, v);\n /* read sub-expression with higher priority */\n nextop = subexpr(ls, &v2, priority[op].right);\n luaK_posfix(ls->fs, op, v, &v2);\n op = nextop;\n }\n leavelevel(ls);\n return op; /* return first untreated operator */\n}\n\n\nstatic void expr (LexState *ls, expdesc *v) {\n subexpr(ls, v, 0);\n}\n\n/* }==================================================================== */\n\n\n\n/*\n** {======================================================================\n** Rules for Statements\n** =======================================================================\n*/\n\n\nstatic int block_follow (int token) {\n switch (token) {\n case TK_ELSE: case TK_ELSEIF: case TK_END:\n case TK_UNTIL: case TK_EOS:\n return 1;\n default: return 0;\n }\n}\n\n\nstatic void block (LexState *ls) {\n /* block -> chunk */\n FuncState *fs = ls->fs;\n BlockCnt bl;\n enterblock(fs, &bl, 0);\n chunk(ls);\n lua_assert(bl.breaklist == NO_JUMP);\n leaveblock(fs);\n}\n\n\n/*\n** structure to chain all variables in the left-hand side of an\n** assignment\n*/\nstruct LHS_assign {\n struct LHS_assign *prev;\n expdesc v; /* variable (global, local, upvalue, or indexed) */\n};\n\n\n/*\n** check whether, in an assignment to a local variable, the local variable\n** is needed in a previous assignment (to a table). If so, save original\n** local value in a safe place and use this safe copy in the previous\n** assignment.\n*/\nstatic void check_conflict (LexState *ls, struct LHS_assign *lh, expdesc *v) {\n FuncState *fs = ls->fs;\n int extra = fs->freereg; /* eventual position to save local variable */\n int conflict = 0;\n for (; lh; lh = lh->prev) {\n if (lh->v.k == VINDEXED) {\n if (lh->v.u.s.info == v->u.s.info) { /* conflict? */\n conflict = 1;\n lh->v.u.s.info = extra; /* previous assignment will use safe copy */\n }\n if (lh->v.u.s.aux == v->u.s.info) { /* conflict? */\n conflict = 1;\n lh->v.u.s.aux = extra; /* previous assignment will use safe copy */\n }\n }\n }\n if (conflict) {\n luaK_codeABC(fs, OP_MOVE, fs->freereg, v->u.s.info, 0); /* make copy */\n luaK_reserveregs(fs, 1);\n }\n}\n\n\nstatic void assignment (LexState *ls, struct LHS_assign *lh, int nvars) {\n expdesc e;\n check_condition(ls, VLOCAL <= lh->v.k && lh->v.k <= VINDEXED,\n \"syntax error\");\n if (testnext(ls, ',')) { /* assignment -> `,' primaryexp assignment */\n struct LHS_assign nv;\n nv.prev = lh;\n primaryexp(ls, &nv.v);\n if (nv.v.k == VLOCAL)\n check_conflict(ls, lh, &nv.v);\n luaY_checklimit(ls->fs, nvars, LUAI_MAXCCALLS - ls->L->nCcalls,\n \"variables in assignment\");\n assignment(ls, &nv, nvars+1);\n }\n else { /* assignment -> `=' explist1 */\n int nexps;\n checknext(ls, '=');\n nexps = explist1(ls, &e);\n if (nexps != nvars) {\n adjust_assign(ls, nvars, nexps, &e);\n if (nexps > nvars)\n ls->fs->freereg -= nexps - nvars; /* remove extra values */\n }\n else {\n luaK_setoneret(ls->fs, &e); /* close last expression */\n luaK_storevar(ls->fs, &lh->v, &e);\n return; /* avoid default */\n }\n }\n init_exp(&e, VNONRELOC, ls->fs->freereg-1); /* default assignment */\n luaK_storevar(ls->fs, &lh->v, &e);\n}\n\n\nstatic int cond (LexState *ls) {\n /* cond -> exp */\n expdesc v;\n expr(ls, &v); /* read condition */\n if (v.k == VNIL) v.k = VFALSE; /* `falses' are all equal here */\n luaK_goiftrue(ls->fs, &v);\n return v.f;\n}\n\n\nstatic void breakstat (LexState *ls) {\n FuncState *fs = ls->fs;\n BlockCnt *bl = fs->bl;\n int upval = 0;\n while (bl && !bl->isbreakable) {\n upval |= bl->upval;\n bl = bl->previous;\n }\n if (!bl)\n luaX_syntaxerror(ls, \"no loop to break\");\n if (upval)\n luaK_codeABC(fs, OP_CLOSE, bl->nactvar, 0, 0);\n luaK_concat(fs, &bl->breaklist, luaK_jump(fs));\n}\n\n\nstatic void whilestat (LexState *ls, int line) {\n /* whilestat -> WHILE cond DO block END */\n FuncState *fs = ls->fs;\n int whileinit;\n int condexit;\n BlockCnt bl;\n luaX_next(ls); /* skip WHILE */\n whileinit = luaK_getlabel(fs);\n condexit = cond(ls);\n enterblock(fs, &bl, 1);\n checknext(ls, TK_DO);\n block(ls);\n luaK_patchlist(fs, luaK_jump(fs), whileinit);\n check_match(ls, TK_END, TK_WHILE, line);\n leaveblock(fs);\n luaK_patchtohere(fs, condexit); /* false conditions finish the loop */\n}\n\n\nstatic void repeatstat (LexState *ls, int line) {\n /* repeatstat -> REPEAT block UNTIL cond */\n int condexit;\n FuncState *fs = ls->fs;\n int repeat_init = luaK_getlabel(fs);\n BlockCnt bl1, bl2;\n enterblock(fs, &bl1, 1); /* loop block */\n enterblock(fs, &bl2, 0); /* scope block */\n luaX_next(ls); /* skip REPEAT */\n chunk(ls);\n check_match(ls, TK_UNTIL, TK_REPEAT, line);\n condexit = cond(ls); /* read condition (inside scope block) */\n if (!bl2.upval) { /* no upvalues? */\n leaveblock(fs); /* finish scope */\n luaK_patchlist(ls->fs, condexit, repeat_init); /* close the loop */\n }\n else { /* complete semantics when there are upvalues */\n breakstat(ls); /* if condition then break */\n luaK_patchtohere(ls->fs, condexit); /* else... */\n leaveblock(fs); /* finish scope... */\n luaK_patchlist(ls->fs, luaK_jump(fs), repeat_init); /* and repeat */\n }\n leaveblock(fs); /* finish loop */\n}\n\n\nstatic int exp1 (LexState *ls) {\n expdesc e;\n int k;\n expr(ls, &e);\n k = e.k;\n luaK_exp2nextreg(ls->fs, &e);\n return k;\n}\n\n\nstatic void forbody (LexState *ls, int base, int line, int nvars, int isnum) {\n /* forbody -> DO block */\n BlockCnt bl;\n FuncState *fs = ls->fs;\n int prep, endfor;\n adjustlocalvars(ls, 3); /* control variables */\n checknext(ls, TK_DO);\n prep = isnum ? luaK_codeAsBx(fs, OP_FORPREP, base, NO_JUMP) : luaK_jump(fs);\n enterblock(fs, &bl, 0); /* scope for declared variables */\n adjustlocalvars(ls, nvars);\n luaK_reserveregs(fs, nvars);\n block(ls);\n leaveblock(fs); /* end of scope for declared variables */\n luaK_patchtohere(fs, prep);\n endfor = (isnum) ? luaK_codeAsBx(fs, OP_FORLOOP, base, NO_JUMP) :\n luaK_codeABC(fs, OP_TFORLOOP, base, 0, nvars);\n luaK_fixline(fs, line); /* pretend that `OP_FOR' starts the loop */\n luaK_patchlist(fs, (isnum ? endfor : luaK_jump(fs)), prep + 1);\n}\n\n\nstatic void fornum (LexState *ls, TString *varname, int line) {\n /* fornum -> NAME = exp1,exp1[,exp1] forbody */\n FuncState *fs = ls->fs;\n int base = fs->freereg;\n new_localvarliteral(ls, \"(for index)\", 0);\n new_localvarliteral(ls, \"(for limit)\", 1);\n new_localvarliteral(ls, \"(for step)\", 2);\n new_localvar(ls, varname, 3);\n checknext(ls, '=');\n exp1(ls); /* initial value */\n checknext(ls, ',');\n exp1(ls); /* limit */\n if (testnext(ls, ','))\n exp1(ls); /* optional step */\n else { /* default step = 1 */\n luaK_codeABx(fs, OP_LOADK, fs->freereg, luaK_numberK(fs, 1));\n luaK_reserveregs(fs, 1);\n }\n forbody(ls, base, line, 1, 1);\n}\n\n\nstatic void forlist (LexState *ls, TString *indexname) {\n /* forlist -> NAME {,NAME} IN explist1 forbody */\n FuncState *fs = ls->fs;\n expdesc e;\n int nvars = 0;\n int line;\n int base = fs->freereg;\n /* create control variables */\n new_localvarliteral(ls, \"(for generator)\", nvars++);\n new_localvarliteral(ls, \"(for state)\", nvars++);\n new_localvarliteral(ls, \"(for control)\", nvars++);\n /* create declared variables */\n new_localvar(ls, indexname, nvars++);\n while (testnext(ls, ','))\n new_localvar(ls, str_checkname(ls), nvars++);\n checknext(ls, TK_IN);\n line = ls->linenumber;\n adjust_assign(ls, 3, explist1(ls, &e), &e);\n luaK_checkstack(fs, 3); /* extra space to call generator */\n forbody(ls, base, line, nvars - 3, 0);\n}\n\n\nstatic void forstat (LexState *ls, int line) {\n /* forstat -> FOR (fornum | forlist) END */\n FuncState *fs = ls->fs;\n TString *varname;\n BlockCnt bl;\n enterblock(fs, &bl, 1); /* scope for loop and control variables */\n luaX_next(ls); /* skip `for' */\n varname = str_checkname(ls); /* first variable name */\n switch (ls->t.token) {\n case '=': fornum(ls, varname, line); break;\n case ',': case TK_IN: forlist(ls, varname); break;\n default: luaX_syntaxerror(ls, LUA_QL(\"=\") \" or \" LUA_QL(\"in\") \" expected\");\n }\n check_match(ls, TK_END, TK_FOR, line);\n leaveblock(fs); /* loop scope (`break' jumps to this point) */\n}\n\n\nstatic int test_then_block (LexState *ls) {\n /* test_then_block -> [IF | ELSEIF] cond THEN block */\n int condexit;\n luaX_next(ls); /* skip IF or ELSEIF */\n condexit = cond(ls);\n checknext(ls, TK_THEN);\n block(ls); /* `then' part */\n return condexit;\n}\n\n\nstatic void ifstat (LexState *ls, int line) {\n /* ifstat -> IF cond THEN block {ELSEIF cond THEN block} [ELSE block] END */\n FuncState *fs = ls->fs;\n int flist;\n int escapelist = NO_JUMP;\n flist = test_then_block(ls); /* IF cond THEN block */\n while (ls->t.token == TK_ELSEIF) {\n luaK_concat(fs, &escapelist, luaK_jump(fs));\n luaK_patchtohere(fs, flist);\n flist = test_then_block(ls); /* ELSEIF cond THEN block */\n }\n if (ls->t.token == TK_ELSE) {\n luaK_concat(fs, &escapelist, luaK_jump(fs));\n luaK_patchtohere(fs, flist);\n luaX_next(ls); /* skip ELSE (after patch, for correct line info) */\n block(ls); /* `else' part */\n }\n else\n luaK_concat(fs, &escapelist, flist);\n luaK_patchtohere(fs, escapelist);\n check_match(ls, TK_END, TK_IF, line);\n}\n\n\nstatic void localfunc (LexState *ls) {\n expdesc v, b;\n FuncState *fs = ls->fs;\n new_localvar(ls, str_checkname(ls), 0);\n init_exp(&v, VLOCAL, fs->freereg);\n luaK_reserveregs(fs, 1);\n adjustlocalvars(ls, 1);\n body(ls, &b, 0, ls->linenumber);\n luaK_storevar(fs, &v, &b);\n /* debug information will only see the variable after this point! */\n getlocvar(fs, fs->nactvar - 1).startpc = fs->pc;\n}\n\n\nstatic void localstat (LexState *ls) {\n /* stat -> LOCAL NAME {`,' NAME} [`=' explist1] */\n int nvars = 0;\n int nexps;\n expdesc e;\n do {\n new_localvar(ls, str_checkname(ls), nvars++);\n } while (testnext(ls, ','));\n if (testnext(ls, '='))\n nexps = explist1(ls, &e);\n else {\n e.k = VVOID;\n nexps = 0;\n }\n adjust_assign(ls, nvars, nexps, &e);\n adjustlocalvars(ls, nvars);\n}\n\n\nstatic int funcname (LexState *ls, expdesc *v) {\n /* funcname -> NAME {field} [`:' NAME] */\n int needself = 0;\n singlevar(ls, v);\n while (ls->t.token == '.')\n field(ls, v);\n if (ls->t.token == ':') {\n needself = 1;\n field(ls, v);\n }\n return needself;\n}\n\n\nstatic void funcstat (LexState *ls, int line) {\n /* funcstat -> FUNCTION funcname body */\n int needself;\n expdesc v, b;\n luaX_next(ls); /* skip FUNCTION */\n needself = funcname(ls, &v);\n body(ls, &b, needself, line);\n luaK_storevar(ls->fs, &v, &b);\n luaK_fixline(ls->fs, line); /* definition `happens' in the first line */\n}\n\n\nstatic void exprstat (LexState *ls) {\n /* stat -> func | assignment */\n FuncState *fs = ls->fs;\n struct LHS_assign v;\n primaryexp(ls, &v.v);\n if (v.v.k == VCALL) /* stat -> func */\n SETARG_C(getcode(fs, &v.v), 1); /* call statement uses no results */\n else { /* stat -> assignment */\n v.prev = NULL;\n assignment(ls, &v, 1);\n }\n}\n\n\nstatic void retstat (LexState *ls) {\n /* stat -> RETURN explist */\n FuncState *fs = ls->fs;\n expdesc e;\n int first, nret; /* registers with returned values */\n luaX_next(ls); /* skip RETURN */\n if (block_follow(ls->t.token) || ls->t.token == ';')\n first = nret = 0; /* return no values */\n else {\n nret = explist1(ls, &e); /* optional return values */\n if (hasmultret(e.k)) {\n luaK_setmultret(fs, &e);\n if (e.k == VCALL && nret == 1) { /* tail call? */\n SET_OPCODE(getcode(fs,&e), OP_TAILCALL);\n lua_assert(GETARG_A(getcode(fs,&e)) == fs->nactvar);\n }\n first = fs->nactvar;\n nret = LUA_MULTRET; /* return all values */\n }\n else {\n if (nret == 1) /* only one single value? */\n first = luaK_exp2anyreg(fs, &e);\n else {\n luaK_exp2nextreg(fs, &e); /* values must go to the `stack' */\n first = fs->nactvar; /* return all `active' values */\n lua_assert(nret == fs->freereg - first);\n }\n }\n }\n luaK_ret(fs, first, nret);\n}\n\n\nstatic int statement (LexState *ls) {\n int line = ls->linenumber; /* may be needed for error messages */\n switch (ls->t.token) {\n case TK_IF: { /* stat -> ifstat */\n ifstat(ls, line);\n return 0;\n }\n case TK_WHILE: { /* stat -> whilestat */\n whilestat(ls, line);\n return 0;\n }\n case TK_DO: { /* stat -> DO block END */\n luaX_next(ls); /* skip DO */\n block(ls);\n check_match(ls, TK_END, TK_DO, line);\n return 0;\n }\n case TK_FOR: { /* stat -> forstat */\n forstat(ls, line);\n return 0;\n }\n case TK_REPEAT: { /* stat -> repeatstat */\n repeatstat(ls, line);\n return 0;\n }\n case TK_FUNCTION: {\n funcstat(ls, line); /* stat -> funcstat */\n return 0;\n }\n case TK_LOCAL: { /* stat -> localstat */\n luaX_next(ls); /* skip LOCAL */\n if (testnext(ls, TK_FUNCTION)) /* local function? */\n localfunc(ls);\n else\n localstat(ls);\n return 0;\n }\n case TK_RETURN: { /* stat -> retstat */\n retstat(ls);\n return 1; /* must be last statement */\n }\n case TK_BREAK: { /* stat -> breakstat */\n luaX_next(ls); /* skip BREAK */\n breakstat(ls);\n return 1; /* must be last statement */\n }\n default: {\n exprstat(ls);\n return 0; /* to avoid warnings */\n }\n }\n}\n\n\nstatic void chunk (LexState *ls) {\n /* chunk -> { stat [`;'] } */\n int islast = 0;\n enterlevel(ls);\n while (!islast && !block_follow(ls->t.token)) {\n islast = statement(ls);\n testnext(ls, ';');\n lua_assert(ls->fs->f->maxstacksize >= ls->fs->freereg &&\n ls->fs->freereg >= ls->fs->nactvar);\n ls->fs->freereg = ls->fs->nactvar; /* free registers */\n }\n leavelevel(ls);\n}\n\n/* }====================================================================== */\n" }, { "path": "deps/lua/src/lparser.h", "content": "/*\n** $Id: lparser.h,v 1.57.1.1 2007/12/27 13:02:25 roberto Exp $\n** Lua Parser\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lparser_h\n#define lparser_h\n\n#include \"llimits.h\"\n#include \"lobject.h\"\n#include \"lzio.h\"\n\n\n/*\n** Expression descriptor\n*/\n\ntypedef enum {\n VVOID,\t/* no value */\n VNIL,\n VTRUE,\n VFALSE,\n VK,\t\t/* info = index of constant in `k' */\n VKNUM,\t/* nval = numerical value */\n VLOCAL,\t/* info = local register */\n VUPVAL, /* info = index of upvalue in `upvalues' */\n VGLOBAL,\t/* info = index of table; aux = index of global name in `k' */\n VINDEXED,\t/* info = table register; aux = index register (or `k') */\n VJMP,\t\t/* info = instruction pc */\n VRELOCABLE,\t/* info = instruction pc */\n VNONRELOC,\t/* info = result register */\n VCALL,\t/* info = instruction pc */\n VVARARG\t/* info = instruction pc */\n} expkind;\n\ntypedef struct expdesc {\n expkind k;\n union {\n struct { int info, aux; } s;\n lua_Number nval;\n } u;\n int t; /* patch list of `exit when true' */\n int f; /* patch list of `exit when false' */\n} expdesc;\n\n\ntypedef struct upvaldesc {\n lu_byte k;\n lu_byte info;\n} upvaldesc;\n\n\nstruct BlockCnt; /* defined in lparser.c */\n\n\n/* state needed to generate code for a given function */\ntypedef struct FuncState {\n Proto *f; /* current function header */\n Table *h; /* table to find (and reuse) elements in `k' */\n struct FuncState *prev; /* enclosing function */\n struct LexState *ls; /* lexical state */\n struct lua_State *L; /* copy of the Lua state */\n struct BlockCnt *bl; /* chain of current blocks */\n int pc; /* next position to code (equivalent to `ncode') */\n int lasttarget; /* `pc' of last `jump target' */\n int jpc; /* list of pending jumps to `pc' */\n int freereg; /* first free register */\n int nk; /* number of elements in `k' */\n int np; /* number of elements in `p' */\n short nlocvars; /* number of elements in `locvars' */\n lu_byte nactvar; /* number of active local variables */\n upvaldesc upvalues[LUAI_MAXUPVALUES]; /* upvalues */\n unsigned short actvar[LUAI_MAXVARS]; /* declared-variable stack */\n} FuncState;\n\n\nLUAI_FUNC Proto *luaY_parser (lua_State *L, ZIO *z, Mbuffer *buff,\n const char *name);\n\n\n#endif\n" }, { "path": "deps/lua/src/lstate.c", "content": "/*\n** $Id: lstate.c,v 2.36.1.2 2008/01/03 15:20:39 roberto Exp $\n** Global State\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lstate_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"llex.h\"\n#include \"lmem.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n\n\n#define state_size(x)\t(sizeof(x) + LUAI_EXTRASPACE)\n#define fromstate(l)\t(cast(lu_byte *, (l)) - LUAI_EXTRASPACE)\n#define tostate(l) (cast(lua_State *, cast(lu_byte *, l) + LUAI_EXTRASPACE))\n\n\n/*\n** Main thread combines a thread state and the global state\n*/\ntypedef struct LG {\n lua_State l;\n global_State g;\n} LG;\n \n\n\nstatic void stack_init (lua_State *L1, lua_State *L) {\n /* initialize CallInfo array */\n L1->base_ci = luaM_newvector(L, BASIC_CI_SIZE, CallInfo);\n L1->ci = L1->base_ci;\n L1->size_ci = BASIC_CI_SIZE;\n L1->end_ci = L1->base_ci + L1->size_ci - 1;\n /* initialize stack array */\n L1->stack = luaM_newvector(L, BASIC_STACK_SIZE + EXTRA_STACK, TValue);\n L1->stacksize = BASIC_STACK_SIZE + EXTRA_STACK;\n L1->top = L1->stack;\n L1->stack_last = L1->stack+(L1->stacksize - EXTRA_STACK)-1;\n /* initialize first ci */\n L1->ci->func = L1->top;\n setnilvalue(L1->top++); /* `function' entry for this `ci' */\n L1->base = L1->ci->base = L1->top;\n L1->ci->top = L1->top + LUA_MINSTACK;\n}\n\n\nstatic void freestack (lua_State *L, lua_State *L1) {\n luaM_freearray(L, L1->base_ci, L1->size_ci, CallInfo);\n luaM_freearray(L, L1->stack, L1->stacksize, TValue);\n}\n\n\n/*\n** open parts that may cause memory-allocation errors\n*/\nstatic void f_luaopen (lua_State *L, void *ud) {\n global_State *g = G(L);\n UNUSED(ud);\n stack_init(L, L); /* init stack */\n sethvalue(L, gt(L), luaH_new(L, 0, 2)); /* table of globals */\n sethvalue(L, registry(L), luaH_new(L, 0, 2)); /* registry */\n luaS_resize(L, MINSTRTABSIZE); /* initial size of string table */\n luaT_init(L);\n luaX_init(L);\n luaS_fix(luaS_newliteral(L, MEMERRMSG));\n g->GCthreshold = 4*g->totalbytes;\n}\n\n\nstatic void preinit_state (lua_State *L, global_State *g) {\n G(L) = g;\n L->stack = NULL;\n L->stacksize = 0;\n L->errorJmp = NULL;\n L->hook = NULL;\n L->hookmask = 0;\n L->basehookcount = 0;\n L->allowhook = 1;\n resethookcount(L);\n L->openupval = NULL;\n L->size_ci = 0;\n L->nCcalls = L->baseCcalls = 0;\n L->status = 0;\n L->base_ci = L->ci = NULL;\n L->savedpc = NULL;\n L->errfunc = 0;\n setnilvalue(gt(L));\n}\n\n\nstatic void close_state (lua_State *L) {\n global_State *g = G(L);\n luaF_close(L, L->stack); /* close all upvalues for this thread */\n luaC_freeall(L); /* collect all objects */\n lua_assert(g->rootgc == obj2gco(L));\n lua_assert(g->strt.nuse == 0);\n luaM_freearray(L, G(L)->strt.hash, G(L)->strt.size, TString *);\n luaZ_freebuffer(L, &g->buff);\n freestack(L, L);\n lua_assert(g->totalbytes == sizeof(LG));\n (*g->frealloc)(g->ud, fromstate(L), state_size(LG), 0);\n}\n\n\nlua_State *luaE_newthread (lua_State *L) {\n lua_State *L1 = tostate(luaM_malloc(L, state_size(lua_State)));\n luaC_link(L, obj2gco(L1), LUA_TTHREAD);\n preinit_state(L1, G(L));\n stack_init(L1, L); /* init stack */\n setobj2n(L, gt(L1), gt(L)); /* share table of globals */\n L1->hookmask = L->hookmask;\n L1->basehookcount = L->basehookcount;\n L1->hook = L->hook;\n resethookcount(L1);\n lua_assert(iswhite(obj2gco(L1)));\n return L1;\n}\n\n\nvoid luaE_freethread (lua_State *L, lua_State *L1) {\n luaF_close(L1, L1->stack); /* close all upvalues for this thread */\n lua_assert(L1->openupval == NULL);\n luai_userstatefree(L1);\n freestack(L, L1);\n luaM_freemem(L, fromstate(L1), state_size(lua_State));\n}\n\n\nLUA_API lua_State *lua_newstate (lua_Alloc f, void *ud) {\n int i;\n lua_State *L;\n global_State *g;\n void *l = (*f)(ud, NULL, 0, state_size(LG));\n if (l == NULL) return NULL;\n L = tostate(l);\n g = &((LG *)L)->g;\n L->next = NULL;\n L->tt = LUA_TTHREAD;\n g->currentwhite = bit2mask(WHITE0BIT, FIXEDBIT);\n L->marked = luaC_white(g);\n set2bits(L->marked, FIXEDBIT, SFIXEDBIT);\n preinit_state(L, g);\n g->frealloc = f;\n g->ud = ud;\n g->mainthread = L;\n g->uvhead.u.l.prev = &g->uvhead;\n g->uvhead.u.l.next = &g->uvhead;\n g->GCthreshold = 0; /* mark it as unfinished state */\n g->strt.size = 0;\n g->strt.nuse = 0;\n g->strt.hash = NULL;\n setnilvalue(registry(L));\n luaZ_initbuffer(L, &g->buff);\n g->panic = NULL;\n g->gcstate = GCSpause;\n g->rootgc = obj2gco(L);\n g->sweepstrgc = 0;\n g->sweepgc = &g->rootgc;\n g->gray = NULL;\n g->grayagain = NULL;\n g->weak = NULL;\n g->tmudata = NULL;\n g->totalbytes = sizeof(LG);\n g->gcpause = LUAI_GCPAUSE;\n g->gcstepmul = LUAI_GCMUL;\n g->gcdept = 0;\n for (i=0; imt[i] = NULL;\n if (luaD_rawrunprotected(L, f_luaopen, NULL) != 0) {\n /* memory allocation error: free partial state */\n close_state(L);\n L = NULL;\n }\n else\n luai_userstateopen(L);\n return L;\n}\n\n\nstatic void callallgcTM (lua_State *L, void *ud) {\n UNUSED(ud);\n luaC_callGCTM(L); /* call GC metamethods for all udata */\n}\n\n\nLUA_API void lua_close (lua_State *L) {\n L = G(L)->mainthread; /* only the main thread can be closed */\n lua_lock(L);\n luaF_close(L, L->stack); /* close all upvalues for this thread */\n luaC_separateudata(L, 1); /* separate udata that have GC metamethods */\n L->errfunc = 0; /* no error function during GC metamethods */\n do { /* repeat until no more errors */\n L->ci = L->base_ci;\n L->base = L->top = L->ci->base;\n L->nCcalls = L->baseCcalls = 0;\n } while (luaD_rawrunprotected(L, callallgcTM, NULL) != 0);\n lua_assert(G(L)->tmudata == NULL);\n luai_userstateclose(L);\n close_state(L);\n}\n\n" }, { "path": "deps/lua/src/lstate.h", "content": "/*\n** $Id: lstate.h,v 2.24.1.2 2008/01/03 15:20:39 roberto Exp $\n** Global State\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lstate_h\n#define lstate_h\n\n#include \"lua.h\"\n\n#include \"lobject.h\"\n#include \"ltm.h\"\n#include \"lzio.h\"\n\n\n\nstruct lua_longjmp; /* defined in ldo.c */\n\n\n/* table of globals */\n#define gt(L)\t(&L->l_gt)\n\n/* registry */\n#define registry(L)\t(&G(L)->l_registry)\n\n\n/* extra stack space to handle TM calls and some other extras */\n#define EXTRA_STACK 5\n\n\n#define BASIC_CI_SIZE 8\n\n#define BASIC_STACK_SIZE (2*LUA_MINSTACK)\n\n\n\ntypedef struct stringtable {\n GCObject **hash;\n lu_int32 nuse; /* number of elements */\n int size;\n} stringtable;\n\n\n/*\n** informations about a call\n*/\ntypedef struct CallInfo {\n StkId base; /* base for this function */\n StkId func; /* function index in the stack */\n StkId\ttop; /* top for this function */\n const Instruction *savedpc;\n int nresults; /* expected number of results from this function */\n int tailcalls; /* number of tail calls lost under this entry */\n} CallInfo;\n\n\n\n#define curr_func(L)\t(clvalue(L->ci->func))\n#define ci_func(ci)\t(clvalue((ci)->func))\n#define f_isLua(ci)\t(!ci_func(ci)->c.isC)\n#define isLua(ci)\t(ttisfunction((ci)->func) && f_isLua(ci))\n\n\n/*\n** `global state', shared by all threads of this state\n*/\ntypedef struct global_State {\n stringtable strt; /* hash table for strings */\n lua_Alloc frealloc; /* function to reallocate memory */\n void *ud; /* auxiliary data to `frealloc' */\n lu_byte currentwhite;\n lu_byte gcstate; /* state of garbage collector */\n int sweepstrgc; /* position of sweep in `strt' */\n GCObject *rootgc; /* list of all collectable objects */\n GCObject **sweepgc; /* position of sweep in `rootgc' */\n GCObject *gray; /* list of gray objects */\n GCObject *grayagain; /* list of objects to be traversed atomically */\n GCObject *weak; /* list of weak tables (to be cleared) */\n GCObject *tmudata; /* last element of list of userdata to be GC */\n Mbuffer buff; /* temporary buffer for string concatentation */\n lu_mem GCthreshold;\n lu_mem totalbytes; /* number of bytes currently allocated */\n lu_mem estimate; /* an estimate of number of bytes actually in use */\n lu_mem gcdept; /* how much GC is `behind schedule' */\n int gcpause; /* size of pause between successive GCs */\n int gcstepmul; /* GC `granularity' */\n lua_CFunction panic; /* to be called in unprotected errors */\n TValue l_registry;\n struct lua_State *mainthread;\n UpVal uvhead; /* head of double-linked list of all open upvalues */\n struct Table *mt[NUM_TAGS]; /* metatables for basic types */\n TString *tmname[TM_N]; /* array with tag-method names */\n} global_State;\n\n\n/*\n** `per thread' state\n*/\nstruct lua_State {\n CommonHeader;\n lu_byte status;\n StkId top; /* first free slot in the stack */\n StkId base; /* base of current function */\n global_State *l_G;\n CallInfo *ci; /* call info for current function */\n const Instruction *savedpc; /* `savedpc' of current function */\n StkId stack_last; /* last free slot in the stack */\n StkId stack; /* stack base */\n CallInfo *end_ci; /* points after end of ci array*/\n CallInfo *base_ci; /* array of CallInfo's */\n int stacksize;\n int size_ci; /* size of array `base_ci' */\n unsigned short nCcalls; /* number of nested C calls */\n unsigned short baseCcalls; /* nested C calls when resuming coroutine */\n lu_byte hookmask;\n lu_byte allowhook;\n int basehookcount;\n int hookcount;\n lua_Hook hook;\n TValue l_gt; /* table of globals */\n TValue env; /* temporary place for environments */\n GCObject *openupval; /* list of open upvalues in this stack */\n GCObject *gclist;\n struct lua_longjmp *errorJmp; /* current error recover point */\n ptrdiff_t errfunc; /* current error handling function (stack index) */\n};\n\n\n#define G(L)\t(L->l_G)\n\n\n/*\n** Union of all collectable objects\n*/\nunion GCObject {\n GCheader gch;\n union TString ts;\n union Udata u;\n union Closure cl;\n struct Table h;\n struct Proto p;\n struct UpVal uv;\n struct lua_State th; /* thread */\n};\n\n\n/* macros to convert a GCObject into a specific value */\n#define rawgco2ts(o)\tcheck_exp((o)->gch.tt == LUA_TSTRING, &((o)->ts))\n#define gco2ts(o)\t(&rawgco2ts(o)->tsv)\n#define rawgco2u(o)\tcheck_exp((o)->gch.tt == LUA_TUSERDATA, &((o)->u))\n#define gco2u(o)\t(&rawgco2u(o)->uv)\n#define gco2cl(o)\tcheck_exp((o)->gch.tt == LUA_TFUNCTION, &((o)->cl))\n#define gco2h(o)\tcheck_exp((o)->gch.tt == LUA_TTABLE, &((o)->h))\n#define gco2p(o)\tcheck_exp((o)->gch.tt == LUA_TPROTO, &((o)->p))\n#define gco2uv(o)\tcheck_exp((o)->gch.tt == LUA_TUPVAL, &((o)->uv))\n#define ngcotouv(o) \\\n\tcheck_exp((o) == NULL || (o)->gch.tt == LUA_TUPVAL, &((o)->uv))\n#define gco2th(o)\tcheck_exp((o)->gch.tt == LUA_TTHREAD, &((o)->th))\n\n/* macro to convert any Lua object into a GCObject */\n#define obj2gco(v)\t(cast(GCObject *, (v)))\n\n\nLUAI_FUNC lua_State *luaE_newthread (lua_State *L);\nLUAI_FUNC void luaE_freethread (lua_State *L, lua_State *L1);\n\n#endif\n\n" }, { "path": "deps/lua/src/lstring.c", "content": "/*\n** $Id: lstring.c,v 2.8.1.1 2007/12/27 13:02:25 roberto Exp $\n** String table (keeps all strings handled by Lua)\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lstring_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n\n\n\nvoid luaS_resize (lua_State *L, int newsize) {\n GCObject **newhash;\n stringtable *tb;\n int i;\n if (G(L)->gcstate == GCSsweepstring)\n return; /* cannot resize during GC traverse */\n newhash = luaM_newvector(L, newsize, GCObject *);\n tb = &G(L)->strt;\n for (i=0; isize; i++) {\n GCObject *p = tb->hash[i];\n while (p) { /* for each node in the list */\n GCObject *next = p->gch.next; /* save next */\n unsigned int h = gco2ts(p)->hash;\n int h1 = lmod(h, newsize); /* new position */\n lua_assert(cast_int(h%newsize) == lmod(h, newsize));\n p->gch.next = newhash[h1]; /* chain it */\n newhash[h1] = p;\n p = next;\n }\n }\n luaM_freearray(L, tb->hash, tb->size, TString *);\n tb->size = newsize;\n tb->hash = newhash;\n}\n\n\nstatic TString *newlstr (lua_State *L, const char *str, size_t l,\n unsigned int h) {\n TString *ts;\n stringtable *tb;\n if (l+1 > (MAX_SIZET - sizeof(TString))/sizeof(char))\n luaM_toobig(L);\n ts = cast(TString *, luaM_malloc(L, (l+1)*sizeof(char)+sizeof(TString)));\n ts->tsv.len = l;\n ts->tsv.hash = h;\n ts->tsv.marked = luaC_white(G(L));\n ts->tsv.tt = LUA_TSTRING;\n ts->tsv.reserved = 0;\n memcpy(ts+1, str, l*sizeof(char));\n ((char *)(ts+1))[l] = '\\0'; /* ending 0 */\n tb = &G(L)->strt;\n h = lmod(h, tb->size);\n ts->tsv.next = tb->hash[h]; /* chain new entry */\n tb->hash[h] = obj2gco(ts);\n tb->nuse++;\n if (tb->nuse > cast(lu_int32, tb->size) && tb->size <= MAX_INT/2)\n luaS_resize(L, tb->size*2); /* too crowded */\n return ts;\n}\n\n\nTString *luaS_newlstr (lua_State *L, const char *str, size_t l) {\n GCObject *o;\n unsigned int h = cast(unsigned int, l); /* seed */\n size_t step = (l>>5)+1; /* if string is too long, don't hash all its chars */\n size_t l1;\n for (l1=l; l1>=step; l1-=step) /* compute hash */\n h = h ^ ((h<<5)+(h>>2)+cast(unsigned char, str[l1-1]));\n for (o = G(L)->strt.hash[lmod(h, G(L)->strt.size)];\n o != NULL;\n o = o->gch.next) {\n TString *ts = rawgco2ts(o);\n if (ts->tsv.len == l && (memcmp(str, getstr(ts), l) == 0)) {\n /* string may be dead */\n if (isdead(G(L), o)) changewhite(o);\n return ts;\n }\n }\n return newlstr(L, str, l, h); /* not found */\n}\n\n\nUdata *luaS_newudata (lua_State *L, size_t s, Table *e) {\n Udata *u;\n if (s > MAX_SIZET - sizeof(Udata))\n luaM_toobig(L);\n u = cast(Udata *, luaM_malloc(L, s + sizeof(Udata)));\n u->uv.marked = luaC_white(G(L)); /* is not finalized */\n u->uv.tt = LUA_TUSERDATA;\n u->uv.len = s;\n u->uv.metatable = NULL;\n u->uv.env = e;\n /* chain it on udata list (after main thread) */\n u->uv.next = G(L)->mainthread->next;\n G(L)->mainthread->next = obj2gco(u);\n return u;\n}\n\n" }, { "path": "deps/lua/src/lstring.h", "content": "/*\n** $Id: lstring.h,v 1.43.1.1 2007/12/27 13:02:25 roberto Exp $\n** String table (keep all strings handled by Lua)\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lstring_h\n#define lstring_h\n\n\n#include \"lgc.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n\n\n#define sizestring(s)\t(sizeof(union TString)+((s)->len+1)*sizeof(char))\n\n#define sizeudata(u)\t(sizeof(union Udata)+(u)->len)\n\n#define luaS_new(L, s)\t(luaS_newlstr(L, s, strlen(s)))\n#define luaS_newliteral(L, s)\t(luaS_newlstr(L, \"\" s, \\\n (sizeof(s)/sizeof(char))-1))\n\n#define luaS_fix(s)\tl_setbit((s)->tsv.marked, FIXEDBIT)\n\nLUAI_FUNC void luaS_resize (lua_State *L, int newsize);\nLUAI_FUNC Udata *luaS_newudata (lua_State *L, size_t s, Table *e);\nLUAI_FUNC TString *luaS_newlstr (lua_State *L, const char *str, size_t l);\n\n\n#endif\n" }, { "path": "deps/lua/src/lstrlib.c", "content": "/*\n** $Id: lstrlib.c,v 1.132.1.5 2010/05/14 15:34:19 roberto Exp $\n** Standard library for string operations and pattern-matching\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n#include \n\n#define lstrlib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n/* macro to `unsign' a character */\n#define uchar(c) ((unsigned char)(c))\n\n\n\nstatic int str_len (lua_State *L) {\n size_t l;\n luaL_checklstring(L, 1, &l);\n lua_pushinteger(L, l);\n return 1;\n}\n\n\nstatic ptrdiff_t posrelat (ptrdiff_t pos, size_t len) {\n /* relative string position: negative means back from end */\n if (pos < 0) pos += (ptrdiff_t)len + 1;\n return (pos >= 0) ? pos : 0;\n}\n\n\nstatic int str_sub (lua_State *L) {\n size_t l;\n const char *s = luaL_checklstring(L, 1, &l);\n ptrdiff_t start = posrelat(luaL_checkinteger(L, 2), l);\n ptrdiff_t end = posrelat(luaL_optinteger(L, 3, -1), l);\n if (start < 1) start = 1;\n if (end > (ptrdiff_t)l) end = (ptrdiff_t)l;\n if (start <= end)\n lua_pushlstring(L, s+start-1, end-start+1);\n else lua_pushliteral(L, \"\");\n return 1;\n}\n\n\nstatic int str_reverse (lua_State *L) {\n size_t l;\n luaL_Buffer b;\n const char *s = luaL_checklstring(L, 1, &l);\n luaL_buffinit(L, &b);\n while (l--) luaL_addchar(&b, s[l]);\n luaL_pushresult(&b);\n return 1;\n}\n\n\nstatic int str_lower (lua_State *L) {\n size_t l;\n size_t i;\n luaL_Buffer b;\n const char *s = luaL_checklstring(L, 1, &l);\n luaL_buffinit(L, &b);\n for (i=0; i 0)\n luaL_addlstring(&b, s, l);\n luaL_pushresult(&b);\n return 1;\n}\n\n\nstatic int str_byte (lua_State *L) {\n size_t l;\n const char *s = luaL_checklstring(L, 1, &l);\n ptrdiff_t posi = posrelat(luaL_optinteger(L, 2, 1), l);\n ptrdiff_t pose = posrelat(luaL_optinteger(L, 3, posi), l);\n int n, i;\n if (posi <= 0) posi = 1;\n if ((size_t)pose > l) pose = l;\n if (posi > pose) return 0; /* empty interval; return no values */\n n = (int)(pose - posi + 1);\n if (posi + n <= pose) /* overflow? */\n luaL_error(L, \"string slice too long\");\n luaL_checkstack(L, n, \"string slice too long\");\n for (i=0; i= ms->level || ms->capture[l].len == CAP_UNFINISHED)\n return luaL_error(ms->L, \"invalid capture index\");\n return l;\n}\n\n\nstatic int capture_to_close (MatchState *ms) {\n int level = ms->level;\n for (level--; level>=0; level--)\n if (ms->capture[level].len == CAP_UNFINISHED) return level;\n return luaL_error(ms->L, \"invalid pattern capture\");\n}\n\n\nstatic const char *classend (MatchState *ms, const char *p) {\n switch (*p++) {\n case L_ESC: {\n if (*p == '\\0')\n luaL_error(ms->L, \"malformed pattern (ends with \" LUA_QL(\"%%\") \")\");\n return p+1;\n }\n case '[': {\n if (*p == '^') p++;\n do { /* look for a `]' */\n if (*p == '\\0')\n luaL_error(ms->L, \"malformed pattern (missing \" LUA_QL(\"]\") \")\");\n if (*(p++) == L_ESC && *p != '\\0')\n p++; /* skip escapes (e.g. `%]') */\n } while (*p != ']');\n return p+1;\n }\n default: {\n return p;\n }\n }\n}\n\n\nstatic int match_class (int c, int cl) {\n int res;\n switch (tolower(cl)) {\n case 'a' : res = isalpha(c); break;\n case 'c' : res = iscntrl(c); break;\n case 'd' : res = isdigit(c); break;\n case 'l' : res = islower(c); break;\n case 'p' : res = ispunct(c); break;\n case 's' : res = isspace(c); break;\n case 'u' : res = isupper(c); break;\n case 'w' : res = isalnum(c); break;\n case 'x' : res = isxdigit(c); break;\n case 'z' : res = (c == 0); break;\n default: return (cl == c);\n }\n return (islower(cl) ? res : !res);\n}\n\n\nstatic int matchbracketclass (int c, const char *p, const char *ec) {\n int sig = 1;\n if (*(p+1) == '^') {\n sig = 0;\n p++; /* skip the `^' */\n }\n while (++p < ec) {\n if (*p == L_ESC) {\n p++;\n if (match_class(c, uchar(*p)))\n return sig;\n }\n else if ((*(p+1) == '-') && (p+2 < ec)) {\n p+=2;\n if (uchar(*(p-2)) <= c && c <= uchar(*p))\n return sig;\n }\n else if (uchar(*p) == c) return sig;\n }\n return !sig;\n}\n\n\nstatic int singlematch (int c, const char *p, const char *ep) {\n switch (*p) {\n case '.': return 1; /* matches any char */\n case L_ESC: return match_class(c, uchar(*(p+1)));\n case '[': return matchbracketclass(c, p, ep-1);\n default: return (uchar(*p) == c);\n }\n}\n\n\nstatic const char *match (MatchState *ms, const char *s, const char *p);\n\n\nstatic const char *matchbalance (MatchState *ms, const char *s,\n const char *p) {\n if (*p == 0 || *(p+1) == 0)\n luaL_error(ms->L, \"unbalanced pattern\");\n if (*s != *p) return NULL;\n else {\n int b = *p;\n int e = *(p+1);\n int cont = 1;\n while (++s < ms->src_end) {\n if (*s == e) {\n if (--cont == 0) return s+1;\n }\n else if (*s == b) cont++;\n }\n }\n return NULL; /* string ends out of balance */\n}\n\n\nstatic const char *max_expand (MatchState *ms, const char *s,\n const char *p, const char *ep) {\n ptrdiff_t i = 0; /* counts maximum expand for item */\n while ((s+i)src_end && singlematch(uchar(*(s+i)), p, ep))\n i++;\n /* keeps trying to match with the maximum repetitions */\n while (i>=0) {\n const char *res = match(ms, (s+i), ep+1);\n if (res) return res;\n i--; /* else didn't match; reduce 1 repetition to try again */\n }\n return NULL;\n}\n\n\nstatic const char *min_expand (MatchState *ms, const char *s,\n const char *p, const char *ep) {\n for (;;) {\n const char *res = match(ms, s, ep+1);\n if (res != NULL)\n return res;\n else if (ssrc_end && singlematch(uchar(*s), p, ep))\n s++; /* try with one more repetition */\n else return NULL;\n }\n}\n\n\nstatic const char *start_capture (MatchState *ms, const char *s,\n const char *p, int what) {\n const char *res;\n int level = ms->level;\n if (level >= LUA_MAXCAPTURES) luaL_error(ms->L, \"too many captures\");\n ms->capture[level].init = s;\n ms->capture[level].len = what;\n ms->level = level+1;\n if ((res=match(ms, s, p)) == NULL) /* match failed? */\n ms->level--; /* undo capture */\n return res;\n}\n\n\nstatic const char *end_capture (MatchState *ms, const char *s,\n const char *p) {\n int l = capture_to_close(ms);\n const char *res;\n ms->capture[l].len = s - ms->capture[l].init; /* close capture */\n if ((res = match(ms, s, p)) == NULL) /* match failed? */\n ms->capture[l].len = CAP_UNFINISHED; /* undo capture */\n return res;\n}\n\n\nstatic const char *match_capture (MatchState *ms, const char *s, int l) {\n size_t len;\n l = check_capture(ms, l);\n len = ms->capture[l].len;\n if ((size_t)(ms->src_end-s) >= len &&\n memcmp(ms->capture[l].init, s, len) == 0)\n return s+len;\n else return NULL;\n}\n\n\nstatic const char *match (MatchState *ms, const char *s, const char *p) {\n init: /* using goto's to optimize tail recursion */\n switch (*p) {\n case '(': { /* start capture */\n if (*(p+1) == ')') /* position capture? */\n return start_capture(ms, s, p+2, CAP_POSITION);\n else\n return start_capture(ms, s, p+1, CAP_UNFINISHED);\n }\n case ')': { /* end capture */\n return end_capture(ms, s, p+1);\n }\n case L_ESC: {\n switch (*(p+1)) {\n case 'b': { /* balanced string? */\n s = matchbalance(ms, s, p+2);\n if (s == NULL) return NULL;\n p+=4; goto init; /* else return match(ms, s, p+4); */\n }\n case 'f': { /* frontier? */\n const char *ep; char previous;\n p += 2;\n if (*p != '[')\n luaL_error(ms->L, \"missing \" LUA_QL(\"[\") \" after \"\n LUA_QL(\"%%f\") \" in pattern\");\n ep = classend(ms, p); /* points to what is next */\n previous = (s == ms->src_init) ? '\\0' : *(s-1);\n if (matchbracketclass(uchar(previous), p, ep-1) ||\n !matchbracketclass(uchar(*s), p, ep-1)) return NULL;\n p=ep; goto init; /* else return match(ms, s, ep); */\n }\n default: {\n if (isdigit(uchar(*(p+1)))) { /* capture results (%0-%9)? */\n s = match_capture(ms, s, uchar(*(p+1)));\n if (s == NULL) return NULL;\n p+=2; goto init; /* else return match(ms, s, p+2) */\n }\n goto dflt; /* case default */\n }\n }\n }\n case '\\0': { /* end of pattern */\n return s; /* match succeeded */\n }\n case '$': {\n if (*(p+1) == '\\0') /* is the `$' the last char in pattern? */\n return (s == ms->src_end) ? s : NULL; /* check end of string */\n else goto dflt;\n }\n default: dflt: { /* it is a pattern item */\n const char *ep = classend(ms, p); /* points to what is next */\n int m = ssrc_end && singlematch(uchar(*s), p, ep);\n switch (*ep) {\n case '?': { /* optional */\n const char *res;\n if (m && ((res=match(ms, s+1, ep+1)) != NULL))\n return res;\n p=ep+1; goto init; /* else return match(ms, s, ep+1); */\n }\n case '*': { /* 0 or more repetitions */\n return max_expand(ms, s, p, ep);\n }\n case '+': { /* 1 or more repetitions */\n return (m ? max_expand(ms, s+1, p, ep) : NULL);\n }\n case '-': { /* 0 or more repetitions (minimum) */\n return min_expand(ms, s, p, ep);\n }\n default: {\n if (!m) return NULL;\n s++; p=ep; goto init; /* else return match(ms, s+1, ep); */\n }\n }\n }\n }\n}\n\n\n\nstatic const char *lmemfind (const char *s1, size_t l1,\n const char *s2, size_t l2) {\n if (l2 == 0) return s1; /* empty strings are everywhere */\n else if (l2 > l1) return NULL; /* avoids a negative `l1' */\n else {\n const char *init; /* to search for a `*s2' inside `s1' */\n l2--; /* 1st char will be checked by `memchr' */\n l1 = l1-l2; /* `s2' cannot be found after that */\n while (l1 > 0 && (init = (const char *)memchr(s1, *s2, l1)) != NULL) {\n init++; /* 1st char is already checked */\n if (memcmp(init, s2+1, l2) == 0)\n return init-1;\n else { /* correct `l1' and `s1' to try again */\n l1 -= init-s1;\n s1 = init;\n }\n }\n return NULL; /* not found */\n }\n}\n\n\nstatic void push_onecapture (MatchState *ms, int i, const char *s,\n const char *e) {\n if (i >= ms->level) {\n if (i == 0) /* ms->level == 0, too */\n lua_pushlstring(ms->L, s, e - s); /* add whole match */\n else\n luaL_error(ms->L, \"invalid capture index\");\n }\n else {\n ptrdiff_t l = ms->capture[i].len;\n if (l == CAP_UNFINISHED) luaL_error(ms->L, \"unfinished capture\");\n if (l == CAP_POSITION)\n lua_pushinteger(ms->L, ms->capture[i].init - ms->src_init + 1);\n else\n lua_pushlstring(ms->L, ms->capture[i].init, l);\n }\n}\n\n\nstatic int push_captures (MatchState *ms, const char *s, const char *e) {\n int i;\n int nlevels = (ms->level == 0 && s) ? 1 : ms->level;\n luaL_checkstack(ms->L, nlevels, \"too many captures\");\n for (i = 0; i < nlevels; i++)\n push_onecapture(ms, i, s, e);\n return nlevels; /* number of strings pushed */\n}\n\n\nstatic int str_find_aux (lua_State *L, int find) {\n size_t l1, l2;\n const char *s = luaL_checklstring(L, 1, &l1);\n const char *p = luaL_checklstring(L, 2, &l2);\n ptrdiff_t init = posrelat(luaL_optinteger(L, 3, 1), l1) - 1;\n if (init < 0) init = 0;\n else if ((size_t)(init) > l1) init = (ptrdiff_t)l1;\n if (find && (lua_toboolean(L, 4) || /* explicit request? */\n strpbrk(p, SPECIALS) == NULL)) { /* or no special characters? */\n /* do a plain search */\n const char *s2 = lmemfind(s+init, l1-init, p, l2);\n if (s2) {\n lua_pushinteger(L, s2-s+1);\n lua_pushinteger(L, s2-s+l2);\n return 2;\n }\n }\n else {\n MatchState ms;\n int anchor = (*p == '^') ? (p++, 1) : 0;\n const char *s1=s+init;\n ms.L = L;\n ms.src_init = s;\n ms.src_end = s+l1;\n do {\n const char *res;\n ms.level = 0;\n if ((res=match(&ms, s1, p)) != NULL) {\n if (find) {\n lua_pushinteger(L, s1-s+1); /* start */\n lua_pushinteger(L, res-s); /* end */\n return push_captures(&ms, NULL, 0) + 2;\n }\n else\n return push_captures(&ms, s1, res);\n }\n } while (s1++ < ms.src_end && !anchor);\n }\n lua_pushnil(L); /* not found */\n return 1;\n}\n\n\nstatic int str_find (lua_State *L) {\n return str_find_aux(L, 1);\n}\n\n\nstatic int str_match (lua_State *L) {\n return str_find_aux(L, 0);\n}\n\n\nstatic int gmatch_aux (lua_State *L) {\n MatchState ms;\n size_t ls;\n const char *s = lua_tolstring(L, lua_upvalueindex(1), &ls);\n const char *p = lua_tostring(L, lua_upvalueindex(2));\n const char *src;\n ms.L = L;\n ms.src_init = s;\n ms.src_end = s+ls;\n for (src = s + (size_t)lua_tointeger(L, lua_upvalueindex(3));\n src <= ms.src_end;\n src++) {\n const char *e;\n ms.level = 0;\n if ((e = match(&ms, src, p)) != NULL) {\n lua_Integer newstart = e-s;\n if (e == src) newstart++; /* empty match? go at least one position */\n lua_pushinteger(L, newstart);\n lua_replace(L, lua_upvalueindex(3));\n return push_captures(&ms, src, e);\n }\n }\n return 0; /* not found */\n}\n\n\nstatic int gmatch (lua_State *L) {\n luaL_checkstring(L, 1);\n luaL_checkstring(L, 2);\n lua_settop(L, 2);\n lua_pushinteger(L, 0);\n lua_pushcclosure(L, gmatch_aux, 3);\n return 1;\n}\n\n\nstatic int gfind_nodef (lua_State *L) {\n return luaL_error(L, LUA_QL(\"string.gfind\") \" was renamed to \"\n LUA_QL(\"string.gmatch\"));\n}\n\n\nstatic void add_s (MatchState *ms, luaL_Buffer *b, const char *s,\n const char *e) {\n size_t l, i;\n const char *news = lua_tolstring(ms->L, 3, &l);\n for (i = 0; i < l; i++) {\n if (news[i] != L_ESC)\n luaL_addchar(b, news[i]);\n else {\n i++; /* skip ESC */\n if (!isdigit(uchar(news[i])))\n luaL_addchar(b, news[i]);\n else if (news[i] == '0')\n luaL_addlstring(b, s, e - s);\n else {\n push_onecapture(ms, news[i] - '1', s, e);\n luaL_addvalue(b); /* add capture to accumulated result */\n }\n }\n }\n}\n\n\nstatic void add_value (MatchState *ms, luaL_Buffer *b, const char *s,\n const char *e) {\n lua_State *L = ms->L;\n switch (lua_type(L, 3)) {\n case LUA_TNUMBER:\n case LUA_TSTRING: {\n add_s(ms, b, s, e);\n return;\n }\n case LUA_TFUNCTION: {\n int n;\n lua_pushvalue(L, 3);\n n = push_captures(ms, s, e);\n lua_call(L, n, 1);\n break;\n }\n case LUA_TTABLE: {\n push_onecapture(ms, 0, s, e);\n lua_gettable(L, 3);\n break;\n }\n }\n if (!lua_toboolean(L, -1)) { /* nil or false? */\n lua_pop(L, 1);\n lua_pushlstring(L, s, e - s); /* keep original text */\n }\n else if (!lua_isstring(L, -1))\n luaL_error(L, \"invalid replacement value (a %s)\", luaL_typename(L, -1)); \n luaL_addvalue(b); /* add result to accumulator */\n}\n\n\nstatic int str_gsub (lua_State *L) {\n size_t srcl;\n const char *src = luaL_checklstring(L, 1, &srcl);\n const char *p = luaL_checkstring(L, 2);\n int tr = lua_type(L, 3);\n int max_s = luaL_optint(L, 4, srcl+1);\n int anchor = (*p == '^') ? (p++, 1) : 0;\n int n = 0;\n MatchState ms;\n luaL_Buffer b;\n luaL_argcheck(L, tr == LUA_TNUMBER || tr == LUA_TSTRING ||\n tr == LUA_TFUNCTION || tr == LUA_TTABLE, 3,\n \"string/function/table expected\");\n luaL_buffinit(L, &b);\n ms.L = L;\n ms.src_init = src;\n ms.src_end = src+srcl;\n while (n < max_s) {\n const char *e;\n ms.level = 0;\n e = match(&ms, src, p);\n if (e) {\n n++;\n add_value(&ms, &b, src, e);\n }\n if (e && e>src) /* non empty match? */\n src = e; /* skip it */\n else if (src < ms.src_end)\n luaL_addchar(&b, *src++);\n else break;\n if (anchor) break;\n }\n luaL_addlstring(&b, src, ms.src_end-src);\n luaL_pushresult(&b);\n lua_pushinteger(L, n); /* number of substitutions */\n return 2;\n}\n\n/* }====================================================== */\n\n\n/* maximum size of each formatted item (> len(format('%99.99f', -1e308))) */\n#define MAX_ITEM\t512\n/* valid flags in a format specification */\n#define FLAGS\t\"-+ #0\"\n/*\n** maximum size of each format specification (such as '%-099.99d')\n** (+10 accounts for %99.99x plus margin of error)\n*/\n#define MAX_FORMAT\t(sizeof(FLAGS) + sizeof(LUA_INTFRMLEN) + 10)\n\n\nstatic void addquoted (lua_State *L, luaL_Buffer *b, int arg) {\n size_t l;\n const char *s = luaL_checklstring(L, arg, &l);\n luaL_addchar(b, '\"');\n while (l--) {\n switch (*s) {\n case '\"': case '\\\\': case '\\n': {\n luaL_addchar(b, '\\\\');\n luaL_addchar(b, *s);\n break;\n }\n case '\\r': {\n luaL_addlstring(b, \"\\\\r\", 2);\n break;\n }\n case '\\0': {\n luaL_addlstring(b, \"\\\\000\", 4);\n break;\n }\n default: {\n luaL_addchar(b, *s);\n break;\n }\n }\n s++;\n }\n luaL_addchar(b, '\"');\n}\n\nstatic const char *scanformat (lua_State *L, const char *strfrmt, char *form) {\n const char *p = strfrmt;\n while (*p != '\\0' && strchr(FLAGS, *p) != NULL) p++; /* skip flags */\n if ((size_t)(p - strfrmt) >= sizeof(FLAGS))\n luaL_error(L, \"invalid format (repeated flags)\");\n if (isdigit(uchar(*p))) p++; /* skip width */\n if (isdigit(uchar(*p))) p++; /* (2 digits at most) */\n if (*p == '.') {\n p++;\n if (isdigit(uchar(*p))) p++; /* skip precision */\n if (isdigit(uchar(*p))) p++; /* (2 digits at most) */\n }\n if (isdigit(uchar(*p)))\n luaL_error(L, \"invalid format (width or precision too long)\");\n *(form++) = '%';\n strncpy(form, strfrmt, p - strfrmt + 1);\n form += p - strfrmt + 1;\n *form = '\\0';\n return p;\n}\n\n\nstatic void addintlen (char *form) {\n size_t l = strlen(form);\n char spec = form[l - 1];\n strcpy(form + l - 1, LUA_INTFRMLEN);\n form[l + sizeof(LUA_INTFRMLEN) - 2] = spec;\n form[l + sizeof(LUA_INTFRMLEN) - 1] = '\\0';\n}\n\n\nstatic int str_format (lua_State *L) {\n int top = lua_gettop(L);\n int arg = 1;\n size_t sfl;\n const char *strfrmt = luaL_checklstring(L, arg, &sfl);\n const char *strfrmt_end = strfrmt+sfl;\n luaL_Buffer b;\n luaL_buffinit(L, &b);\n while (strfrmt < strfrmt_end) {\n if (*strfrmt != L_ESC)\n luaL_addchar(&b, *strfrmt++);\n else if (*++strfrmt == L_ESC)\n luaL_addchar(&b, *strfrmt++); /* %% */\n else { /* format item */\n char form[MAX_FORMAT]; /* to store the format (`%...') */\n char buff[MAX_ITEM]; /* to store the formatted item */\n if (++arg > top)\n luaL_argerror(L, arg, \"no value\");\n strfrmt = scanformat(L, strfrmt, form);\n switch (*strfrmt++) {\n case 'c': {\n sprintf(buff, form, (int)luaL_checknumber(L, arg));\n break;\n }\n case 'd': case 'i': {\n addintlen(form);\n sprintf(buff, form, (LUA_INTFRM_T)luaL_checknumber(L, arg));\n break;\n }\n case 'o': case 'u': case 'x': case 'X': {\n addintlen(form);\n sprintf(buff, form, (unsigned LUA_INTFRM_T)luaL_checknumber(L, arg));\n break;\n }\n case 'e': case 'E': case 'f':\n case 'g': case 'G': {\n sprintf(buff, form, (double)luaL_checknumber(L, arg));\n break;\n }\n case 'q': {\n addquoted(L, &b, arg);\n continue; /* skip the 'addsize' at the end */\n }\n case 's': {\n size_t l;\n const char *s = luaL_checklstring(L, arg, &l);\n if (!strchr(form, '.') && l >= 100) {\n /* no precision and string is too long to be formatted;\n keep original string */\n lua_pushvalue(L, arg);\n luaL_addvalue(&b);\n continue; /* skip the `addsize' at the end */\n }\n else {\n sprintf(buff, form, s);\n break;\n }\n }\n default: { /* also treat cases `pnLlh' */\n return luaL_error(L, \"invalid option \" LUA_QL(\"%%%c\") \" to \"\n LUA_QL(\"format\"), *(strfrmt - 1));\n }\n }\n luaL_addlstring(&b, buff, strlen(buff));\n }\n }\n luaL_pushresult(&b);\n return 1;\n}\n\n\nstatic const luaL_Reg strlib[] = {\n {\"byte\", str_byte},\n {\"char\", str_char},\n {\"dump\", str_dump},\n {\"find\", str_find},\n {\"format\", str_format},\n {\"gfind\", gfind_nodef},\n {\"gmatch\", gmatch},\n {\"gsub\", str_gsub},\n {\"len\", str_len},\n {\"lower\", str_lower},\n {\"match\", str_match},\n {\"rep\", str_rep},\n {\"reverse\", str_reverse},\n {\"sub\", str_sub},\n {\"upper\", str_upper},\n {NULL, NULL}\n};\n\n\nstatic void createmetatable (lua_State *L) {\n lua_createtable(L, 0, 1); /* create metatable for strings */\n lua_pushliteral(L, \"\"); /* dummy string */\n lua_pushvalue(L, -2);\n lua_setmetatable(L, -2); /* set string metatable */\n lua_pop(L, 1); /* pop dummy string */\n lua_pushvalue(L, -2); /* string library... */\n lua_setfield(L, -2, \"__index\"); /* ...is the __index metamethod */\n lua_pop(L, 1); /* pop metatable */\n}\n\n\n/*\n** Open string library\n*/\nLUALIB_API int luaopen_string (lua_State *L) {\n luaL_register(L, LUA_STRLIBNAME, strlib);\n#if defined(LUA_COMPAT_GFIND)\n lua_getfield(L, -1, \"gmatch\");\n lua_setfield(L, -2, \"gfind\");\n#endif\n createmetatable(L);\n return 1;\n}\n\n" }, { "path": "deps/lua/src/ltable.c", "content": "/*\n** $Id: ltable.c,v 2.32.1.2 2007/12/28 15:32:23 roberto Exp $\n** Lua tables (hash)\n** See Copyright Notice in lua.h\n*/\n\n\n/*\n** Implementation of tables (aka arrays, objects, or hash tables).\n** Tables keep its elements in two parts: an array part and a hash part.\n** Non-negative integer keys are all candidates to be kept in the array\n** part. The actual size of the array is the largest `n' such that at\n** least half the slots between 0 and n are in use.\n** Hash uses a mix of chained scatter table with Brent's variation.\n** A main invariant of these tables is that, if an element is not\n** in its main position (i.e. the `original' position that its hash gives\n** to it), then the colliding element is in its own main position.\n** Hence even when the load factor reaches 100%, performance remains good.\n*/\n\n#include \n#include \n\n#define ltable_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lgc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"ltable.h\"\n\n\n/*\n** max size of array part is 2^MAXBITS\n*/\n#if LUAI_BITSINT > 26\n#define MAXBITS\t\t26\n#else\n#define MAXBITS\t\t(LUAI_BITSINT-2)\n#endif\n\n#define MAXASIZE\t(1 << MAXBITS)\n\n\n#define hashpow2(t,n) (gnode(t, lmod((n), sizenode(t))))\n \n#define hashstr(t,str) hashpow2(t, (str)->tsv.hash)\n#define hashboolean(t,p) hashpow2(t, p)\n\n\n/*\n** for some types, it is better to avoid modulus by power of 2, as\n** they tend to have many 2 factors.\n*/\n#define hashmod(t,n)\t(gnode(t, ((n) % ((sizenode(t)-1)|1))))\n\n\n#define hashpointer(t,p)\thashmod(t, IntPoint(p))\n\n\n/*\n** number of ints inside a lua_Number\n*/\n#define numints\t\tcast_int(sizeof(lua_Number)/sizeof(int))\n\n\n\n#define dummynode\t\t(&dummynode_)\n\nstatic const Node dummynode_ = {\n {{NULL}, LUA_TNIL}, /* value */\n {{{NULL}, LUA_TNIL, NULL}} /* key */\n};\n\n\n/*\n** hash for lua_Numbers\n*/\nstatic Node *hashnum (const Table *t, lua_Number n) {\n unsigned int a[numints];\n int i;\n if (luai_numeq(n, 0)) /* avoid problems with -0 */\n return gnode(t, 0);\n memcpy(a, &n, sizeof(a));\n for (i = 1; i < numints; i++) a[0] += a[i];\n return hashmod(t, a[0]);\n}\n\n\n\n/*\n** returns the `main' position of an element in a table (that is, the index\n** of its hash value)\n*/\nstatic Node *mainposition (const Table *t, const TValue *key) {\n switch (ttype(key)) {\n case LUA_TNUMBER:\n return hashnum(t, nvalue(key));\n case LUA_TSTRING:\n return hashstr(t, rawtsvalue(key));\n case LUA_TBOOLEAN:\n return hashboolean(t, bvalue(key));\n case LUA_TLIGHTUSERDATA:\n return hashpointer(t, pvalue(key));\n default:\n return hashpointer(t, gcvalue(key));\n }\n}\n\n\n/*\n** returns the index for `key' if `key' is an appropriate key to live in\n** the array part of the table, -1 otherwise.\n*/\nstatic int arrayindex (const TValue *key) {\n if (ttisnumber(key)) {\n lua_Number n = nvalue(key);\n int k;\n lua_number2int(k, n);\n if (luai_numeq(cast_num(k), n))\n return k;\n }\n return -1; /* `key' did not match some condition */\n}\n\n\n/*\n** returns the index of a `key' for table traversals. First goes all\n** elements in the array part, then elements in the hash part. The\n** beginning of a traversal is signalled by -1.\n*/\nstatic int findindex (lua_State *L, Table *t, StkId key) {\n int i;\n if (ttisnil(key)) return -1; /* first iteration */\n i = arrayindex(key);\n if (0 < i && i <= t->sizearray) /* is `key' inside array part? */\n return i-1; /* yes; that's the index (corrected to C) */\n else {\n Node *n = mainposition(t, key);\n do { /* check whether `key' is somewhere in the chain */\n /* key may be dead already, but it is ok to use it in `next' */\n if (luaO_rawequalObj(key2tval(n), key) ||\n (ttype(gkey(n)) == LUA_TDEADKEY && iscollectable(key) &&\n gcvalue(gkey(n)) == gcvalue(key))) {\n i = cast_int(n - gnode(t, 0)); /* key index in hash table */\n /* hash elements are numbered after array ones */\n return i + t->sizearray;\n }\n else n = gnext(n);\n } while (n);\n luaG_runerror(L, \"invalid key to \" LUA_QL(\"next\")); /* key not found */\n return 0; /* to avoid warnings */\n }\n}\n\n\nint luaH_next (lua_State *L, Table *t, StkId key) {\n int i = findindex(L, t, key); /* find original element */\n for (i++; i < t->sizearray; i++) { /* try first array part */\n if (!ttisnil(&t->array[i])) { /* a non-nil value? */\n setnvalue(key, cast_num(i+1));\n setobj2s(L, key+1, &t->array[i]);\n return 1;\n }\n }\n for (i -= t->sizearray; i < sizenode(t); i++) { /* then hash part */\n if (!ttisnil(gval(gnode(t, i)))) { /* a non-nil value? */\n setobj2s(L, key, key2tval(gnode(t, i)));\n setobj2s(L, key+1, gval(gnode(t, i)));\n return 1;\n }\n }\n return 0; /* no more elements */\n}\n\n\n/*\n** {=============================================================\n** Rehash\n** ==============================================================\n*/\n\n\nstatic int computesizes (int nums[], int *narray) {\n int i;\n int twotoi; /* 2^i */\n int a = 0; /* number of elements smaller than 2^i */\n int na = 0; /* number of elements to go to array part */\n int n = 0; /* optimal size for array part */\n for (i = 0, twotoi = 1; twotoi/2 < *narray; i++, twotoi *= 2) {\n if (nums[i] > 0) {\n a += nums[i];\n if (a > twotoi/2) { /* more than half elements present? */\n n = twotoi; /* optimal size (till now) */\n na = a; /* all elements smaller than n will go to array part */\n }\n }\n if (a == *narray) break; /* all elements already counted */\n }\n *narray = n;\n lua_assert(*narray/2 <= na && na <= *narray);\n return na;\n}\n\n\nstatic int countint (const TValue *key, int *nums) {\n int k = arrayindex(key);\n if (0 < k && k <= MAXASIZE) { /* is `key' an appropriate array index? */\n nums[ceillog2(k)]++; /* count as such */\n return 1;\n }\n else\n return 0;\n}\n\n\nstatic int numusearray (const Table *t, int *nums) {\n int lg;\n int ttlg; /* 2^lg */\n int ause = 0; /* summation of `nums' */\n int i = 1; /* count to traverse all array keys */\n for (lg=0, ttlg=1; lg<=MAXBITS; lg++, ttlg*=2) { /* for each slice */\n int lc = 0; /* counter */\n int lim = ttlg;\n if (lim > t->sizearray) {\n lim = t->sizearray; /* adjust upper limit */\n if (i > lim)\n break; /* no more elements to count */\n }\n /* count elements in range (2^(lg-1), 2^lg] */\n for (; i <= lim; i++) {\n if (!ttisnil(&t->array[i-1]))\n lc++;\n }\n nums[lg] += lc;\n ause += lc;\n }\n return ause;\n}\n\n\nstatic int numusehash (const Table *t, int *nums, int *pnasize) {\n int totaluse = 0; /* total number of elements */\n int ause = 0; /* summation of `nums' */\n int i = sizenode(t);\n while (i--) {\n Node *n = &t->node[i];\n if (!ttisnil(gval(n))) {\n ause += countint(key2tval(n), nums);\n totaluse++;\n }\n }\n *pnasize += ause;\n return totaluse;\n}\n\n\nstatic void setarrayvector (lua_State *L, Table *t, int size) {\n int i;\n luaM_reallocvector(L, t->array, t->sizearray, size, TValue);\n for (i=t->sizearray; iarray[i]);\n t->sizearray = size;\n}\n\n\nstatic void setnodevector (lua_State *L, Table *t, int size) {\n int lsize;\n if (size == 0) { /* no elements to hash part? */\n t->node = cast(Node *, dummynode); /* use common `dummynode' */\n lsize = 0;\n }\n else {\n int i;\n lsize = ceillog2(size);\n if (lsize > MAXBITS)\n luaG_runerror(L, \"table overflow\");\n size = twoto(lsize);\n t->node = luaM_newvector(L, size, Node);\n for (i=0; ilsizenode = cast_byte(lsize);\n t->lastfree = gnode(t, size); /* all positions are free */\n}\n\n\nstatic void resize (lua_State *L, Table *t, int nasize, int nhsize) {\n int i;\n int oldasize = t->sizearray;\n int oldhsize = t->lsizenode;\n Node *nold = t->node; /* save old hash ... */\n if (nasize > oldasize) /* array part must grow? */\n setarrayvector(L, t, nasize);\n /* create new hash part with appropriate size */\n setnodevector(L, t, nhsize); \n if (nasize < oldasize) { /* array part must shrink? */\n t->sizearray = nasize;\n /* re-insert elements from vanishing slice */\n for (i=nasize; iarray[i]))\n setobjt2t(L, luaH_setnum(L, t, i+1), &t->array[i]);\n }\n /* shrink array */\n luaM_reallocvector(L, t->array, oldasize, nasize, TValue);\n }\n /* re-insert elements from hash part */\n for (i = twoto(oldhsize) - 1; i >= 0; i--) {\n Node *old = nold+i;\n if (!ttisnil(gval(old)))\n setobjt2t(L, luaH_set(L, t, key2tval(old)), gval(old));\n }\n if (nold != dummynode)\n luaM_freearray(L, nold, twoto(oldhsize), Node); /* free old array */\n}\n\n\nvoid luaH_resizearray (lua_State *L, Table *t, int nasize) {\n int nsize = (t->node == dummynode) ? 0 : sizenode(t);\n resize(L, t, nasize, nsize);\n}\n\n\nstatic void rehash (lua_State *L, Table *t, const TValue *ek) {\n int nasize, na;\n int nums[MAXBITS+1]; /* nums[i] = number of keys between 2^(i-1) and 2^i */\n int i;\n int totaluse;\n for (i=0; i<=MAXBITS; i++) nums[i] = 0; /* reset counts */\n nasize = numusearray(t, nums); /* count keys in array part */\n totaluse = nasize; /* all those keys are integer keys */\n totaluse += numusehash(t, nums, &nasize); /* count keys in hash part */\n /* count extra key */\n nasize += countint(ek, nums);\n totaluse++;\n /* compute new size for array part */\n na = computesizes(nums, &nasize);\n /* resize the table to new computed sizes */\n resize(L, t, nasize, totaluse - na);\n}\n\n\n\n/*\n** }=============================================================\n*/\n\n\nTable *luaH_new (lua_State *L, int narray, int nhash) {\n Table *t = luaM_new(L, Table);\n luaC_link(L, obj2gco(t), LUA_TTABLE);\n t->metatable = NULL;\n t->flags = cast_byte(~0);\n /* temporary values (kept only if some malloc fails) */\n t->array = NULL;\n t->sizearray = 0;\n t->lsizenode = 0;\n t->node = cast(Node *, dummynode);\n setarrayvector(L, t, narray);\n setnodevector(L, t, nhash);\n return t;\n}\n\n\nvoid luaH_free (lua_State *L, Table *t) {\n if (t->node != dummynode)\n luaM_freearray(L, t->node, sizenode(t), Node);\n luaM_freearray(L, t->array, t->sizearray, TValue);\n luaM_free(L, t);\n}\n\n\nstatic Node *getfreepos (Table *t) {\n while (t->lastfree-- > t->node) {\n if (ttisnil(gkey(t->lastfree)))\n return t->lastfree;\n }\n return NULL; /* could not find a free place */\n}\n\n\n\n/*\n** inserts a new key into a hash table; first, check whether key's main \n** position is free. If not, check whether colliding node is in its main \n** position or not: if it is not, move colliding node to an empty place and \n** put new key in its main position; otherwise (colliding node is in its main \n** position), new key goes to an empty position. \n*/\nstatic TValue *newkey (lua_State *L, Table *t, const TValue *key) {\n Node *mp = mainposition(t, key);\n if (!ttisnil(gval(mp)) || mp == dummynode) {\n Node *othern;\n Node *n = getfreepos(t); /* get a free place */\n if (n == NULL) { /* cannot find a free place? */\n rehash(L, t, key); /* grow table */\n return luaH_set(L, t, key); /* re-insert key into grown table */\n }\n lua_assert(n != dummynode);\n othern = mainposition(t, key2tval(mp));\n if (othern != mp) { /* is colliding node out of its main position? */\n /* yes; move colliding node into free position */\n while (gnext(othern) != mp) othern = gnext(othern); /* find previous */\n gnext(othern) = n; /* redo the chain with `n' in place of `mp' */\n *n = *mp; /* copy colliding node into free pos. (mp->next also goes) */\n gnext(mp) = NULL; /* now `mp' is free */\n setnilvalue(gval(mp));\n }\n else { /* colliding node is in its own main position */\n /* new node will go into free position */\n gnext(n) = gnext(mp); /* chain new position */\n gnext(mp) = n;\n mp = n;\n }\n }\n gkey(mp)->value = key->value; gkey(mp)->tt = key->tt;\n luaC_barriert(L, t, key);\n lua_assert(ttisnil(gval(mp)));\n return gval(mp);\n}\n\n\n/*\n** search function for integers\n*/\nconst TValue *luaH_getnum (Table *t, int key) {\n /* (1 <= key && key <= t->sizearray) */\n if (cast(unsigned int, key-1) < cast(unsigned int, t->sizearray))\n return &t->array[key-1];\n else {\n lua_Number nk = cast_num(key);\n Node *n = hashnum(t, nk);\n do { /* check whether `key' is somewhere in the chain */\n if (ttisnumber(gkey(n)) && luai_numeq(nvalue(gkey(n)), nk))\n return gval(n); /* that's it */\n else n = gnext(n);\n } while (n);\n return luaO_nilobject;\n }\n}\n\n\n/*\n** search function for strings\n*/\nconst TValue *luaH_getstr (Table *t, TString *key) {\n Node *n = hashstr(t, key);\n do { /* check whether `key' is somewhere in the chain */\n if (ttisstring(gkey(n)) && rawtsvalue(gkey(n)) == key)\n return gval(n); /* that's it */\n else n = gnext(n);\n } while (n);\n return luaO_nilobject;\n}\n\n\n/*\n** main search function\n*/\nconst TValue *luaH_get (Table *t, const TValue *key) {\n switch (ttype(key)) {\n case LUA_TNIL: return luaO_nilobject;\n case LUA_TSTRING: return luaH_getstr(t, rawtsvalue(key));\n case LUA_TNUMBER: {\n int k;\n lua_Number n = nvalue(key);\n lua_number2int(k, n);\n if (luai_numeq(cast_num(k), nvalue(key))) /* index is int? */\n return luaH_getnum(t, k); /* use specialized version */\n /* else go through */\n }\n default: {\n Node *n = mainposition(t, key);\n do { /* check whether `key' is somewhere in the chain */\n if (luaO_rawequalObj(key2tval(n), key))\n return gval(n); /* that's it */\n else n = gnext(n);\n } while (n);\n return luaO_nilobject;\n }\n }\n}\n\n\nTValue *luaH_set (lua_State *L, Table *t, const TValue *key) {\n const TValue *p = luaH_get(t, key);\n t->flags = 0;\n if (p != luaO_nilobject)\n return cast(TValue *, p);\n else {\n if (ttisnil(key)) luaG_runerror(L, \"table index is nil\");\n else if (ttisnumber(key) && luai_numisnan(nvalue(key)))\n luaG_runerror(L, \"table index is NaN\");\n return newkey(L, t, key);\n }\n}\n\n\nTValue *luaH_setnum (lua_State *L, Table *t, int key) {\n const TValue *p = luaH_getnum(t, key);\n if (p != luaO_nilobject)\n return cast(TValue *, p);\n else {\n TValue k;\n setnvalue(&k, cast_num(key));\n return newkey(L, t, &k);\n }\n}\n\n\nTValue *luaH_setstr (lua_State *L, Table *t, TString *key) {\n const TValue *p = luaH_getstr(t, key);\n if (p != luaO_nilobject)\n return cast(TValue *, p);\n else {\n TValue k;\n setsvalue(L, &k, key);\n return newkey(L, t, &k);\n }\n}\n\n\nstatic int unbound_search (Table *t, unsigned int j) {\n unsigned int i = j; /* i is zero or a present index */\n j++;\n /* find `i' and `j' such that i is present and j is not */\n while (!ttisnil(luaH_getnum(t, j))) {\n i = j;\n j *= 2;\n if (j > cast(unsigned int, MAX_INT)) { /* overflow? */\n /* table was built with bad purposes: resort to linear search */\n i = 1;\n while (!ttisnil(luaH_getnum(t, i))) i++;\n return i - 1;\n }\n }\n /* now do a binary search between them */\n while (j - i > 1) {\n unsigned int m = (i+j)/2;\n if (ttisnil(luaH_getnum(t, m))) j = m;\n else i = m;\n }\n return i;\n}\n\n\n/*\n** Try to find a boundary in table `t'. A `boundary' is an integer index\n** such that t[i] is non-nil and t[i+1] is nil (and 0 if t[1] is nil).\n*/\nint luaH_getn (Table *t) {\n unsigned int j = t->sizearray;\n if (j > 0 && ttisnil(&t->array[j - 1])) {\n /* there is a boundary in the array part: (binary) search for it */\n unsigned int i = 0;\n while (j - i > 1) {\n unsigned int m = (i+j)/2;\n if (ttisnil(&t->array[m - 1])) j = m;\n else i = m;\n }\n return i;\n }\n /* else must find a boundary in hash part */\n else if (t->node == dummynode) /* hash part is empty? */\n return j; /* that is easy... */\n else return unbound_search(t, j);\n}\n\n\n\n#if defined(LUA_DEBUG)\n\nNode *luaH_mainposition (const Table *t, const TValue *key) {\n return mainposition(t, key);\n}\n\nint luaH_isdummy (Node *n) { return n == dummynode; }\n\n#endif\n" }, { "path": "deps/lua/src/ltable.h", "content": "/*\n** $Id: ltable.h,v 2.10.1.1 2007/12/27 13:02:25 roberto Exp $\n** Lua tables (hash)\n** See Copyright Notice in lua.h\n*/\n\n#ifndef ltable_h\n#define ltable_h\n\n#include \"lobject.h\"\n\n\n#define gnode(t,i)\t(&(t)->node[i])\n#define gkey(n)\t\t(&(n)->i_key.nk)\n#define gval(n)\t\t(&(n)->i_val)\n#define gnext(n)\t((n)->i_key.nk.next)\n\n#define key2tval(n)\t(&(n)->i_key.tvk)\n\n\nLUAI_FUNC const TValue *luaH_getnum (Table *t, int key);\nLUAI_FUNC TValue *luaH_setnum (lua_State *L, Table *t, int key);\nLUAI_FUNC const TValue *luaH_getstr (Table *t, TString *key);\nLUAI_FUNC TValue *luaH_setstr (lua_State *L, Table *t, TString *key);\nLUAI_FUNC const TValue *luaH_get (Table *t, const TValue *key);\nLUAI_FUNC TValue *luaH_set (lua_State *L, Table *t, const TValue *key);\nLUAI_FUNC Table *luaH_new (lua_State *L, int narray, int lnhash);\nLUAI_FUNC void luaH_resizearray (lua_State *L, Table *t, int nasize);\nLUAI_FUNC void luaH_free (lua_State *L, Table *t);\nLUAI_FUNC int luaH_next (lua_State *L, Table *t, StkId key);\nLUAI_FUNC int luaH_getn (Table *t);\n\n\n#if defined(LUA_DEBUG)\nLUAI_FUNC Node *luaH_mainposition (const Table *t, const TValue *key);\nLUAI_FUNC int luaH_isdummy (Node *n);\n#endif\n\n\n#endif\n" }, { "path": "deps/lua/src/ltablib.c", "content": "/*\n** $Id: ltablib.c,v 1.38.1.3 2008/02/14 16:46:58 roberto Exp $\n** Library for Table Manipulation\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define ltablib_c\n#define LUA_LIB\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n#define aux_getn(L,n)\t(luaL_checktype(L, n, LUA_TTABLE), luaL_getn(L, n))\n\n\nstatic int foreachi (lua_State *L) {\n int i;\n int n = aux_getn(L, 1);\n luaL_checktype(L, 2, LUA_TFUNCTION);\n for (i=1; i <= n; i++) {\n lua_pushvalue(L, 2); /* function */\n lua_pushinteger(L, i); /* 1st argument */\n lua_rawgeti(L, 1, i); /* 2nd argument */\n lua_call(L, 2, 1);\n if (!lua_isnil(L, -1))\n return 1;\n lua_pop(L, 1); /* remove nil result */\n }\n return 0;\n}\n\n\nstatic int foreach (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n luaL_checktype(L, 2, LUA_TFUNCTION);\n lua_pushnil(L); /* first key */\n while (lua_next(L, 1)) {\n lua_pushvalue(L, 2); /* function */\n lua_pushvalue(L, -3); /* key */\n lua_pushvalue(L, -3); /* value */\n lua_call(L, 2, 1);\n if (!lua_isnil(L, -1))\n return 1;\n lua_pop(L, 2); /* remove value and result */\n }\n return 0;\n}\n\n\nstatic int maxn (lua_State *L) {\n lua_Number max = 0;\n luaL_checktype(L, 1, LUA_TTABLE);\n lua_pushnil(L); /* first key */\n while (lua_next(L, 1)) {\n lua_pop(L, 1); /* remove value */\n if (lua_type(L, -1) == LUA_TNUMBER) {\n lua_Number v = lua_tonumber(L, -1);\n if (v > max) max = v;\n }\n }\n lua_pushnumber(L, max);\n return 1;\n}\n\n\nstatic int getn (lua_State *L) {\n lua_pushinteger(L, aux_getn(L, 1));\n return 1;\n}\n\n\nstatic int setn (lua_State *L) {\n luaL_checktype(L, 1, LUA_TTABLE);\n#ifndef luaL_setn\n luaL_setn(L, 1, luaL_checkint(L, 2));\n#else\n luaL_error(L, LUA_QL(\"setn\") \" is obsolete\");\n#endif\n lua_pushvalue(L, 1);\n return 1;\n}\n\n\nstatic int tinsert (lua_State *L) {\n int e = aux_getn(L, 1) + 1; /* first empty element */\n int pos; /* where to insert new element */\n switch (lua_gettop(L)) {\n case 2: { /* called with only 2 arguments */\n pos = e; /* insert new element at the end */\n break;\n }\n case 3: {\n int i;\n pos = luaL_checkint(L, 2); /* 2nd argument is the position */\n if (pos > e) e = pos; /* `grow' array if necessary */\n for (i = e; i > pos; i--) { /* move up elements */\n lua_rawgeti(L, 1, i-1);\n lua_rawseti(L, 1, i); /* t[i] = t[i-1] */\n }\n break;\n }\n default: {\n return luaL_error(L, \"wrong number of arguments to \" LUA_QL(\"insert\"));\n }\n }\n luaL_setn(L, 1, e); /* new size */\n lua_rawseti(L, 1, pos); /* t[pos] = v */\n return 0;\n}\n\n\nstatic int tremove (lua_State *L) {\n int e = aux_getn(L, 1);\n int pos = luaL_optint(L, 2, e);\n if (!(1 <= pos && pos <= e)) /* position is outside bounds? */\n return 0; /* nothing to remove */\n luaL_setn(L, 1, e - 1); /* t.n = n-1 */\n lua_rawgeti(L, 1, pos); /* result = t[pos] */\n for ( ;pos= P */\n while (lua_rawgeti(L, 1, ++i), sort_comp(L, -1, -2)) {\n if (i>u) luaL_error(L, \"invalid order function for sorting\");\n lua_pop(L, 1); /* remove a[i] */\n }\n /* repeat --j until a[j] <= P */\n while (lua_rawgeti(L, 1, --j), sort_comp(L, -3, -1)) {\n if (j\n\n#define ltm_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"lobject.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n\n\n\nconst char *const luaT_typenames[] = {\n \"nil\", \"boolean\", \"userdata\", \"number\",\n \"string\", \"table\", \"function\", \"userdata\", \"thread\",\n \"proto\", \"upval\"\n};\n\n\nvoid luaT_init (lua_State *L) {\n static const char *const luaT_eventname[] = { /* ORDER TM */\n \"__index\", \"__newindex\",\n \"__gc\", \"__mode\", \"__eq\",\n \"__add\", \"__sub\", \"__mul\", \"__div\", \"__mod\",\n \"__pow\", \"__unm\", \"__len\", \"__lt\", \"__le\",\n \"__concat\", \"__call\"\n };\n int i;\n for (i=0; itmname[i] = luaS_new(L, luaT_eventname[i]);\n luaS_fix(G(L)->tmname[i]); /* never collect these names */\n }\n}\n\n\n/*\n** function to be used with macro \"fasttm\": optimized for absence of\n** tag methods\n*/\nconst TValue *luaT_gettm (Table *events, TMS event, TString *ename) {\n const TValue *tm = luaH_getstr(events, ename);\n lua_assert(event <= TM_EQ);\n if (ttisnil(tm)) { /* no tag method? */\n events->flags |= cast_byte(1u<metatable;\n break;\n case LUA_TUSERDATA:\n mt = uvalue(o)->metatable;\n break;\n default:\n mt = G(L)->mt[ttype(o)];\n }\n return (mt ? luaH_getstr(mt, G(L)->tmname[event]) : luaO_nilobject);\n}\n\n" }, { "path": "deps/lua/src/ltm.h", "content": "/*\n** $Id: ltm.h,v 2.6.1.1 2007/12/27 13:02:25 roberto Exp $\n** Tag methods\n** See Copyright Notice in lua.h\n*/\n\n#ifndef ltm_h\n#define ltm_h\n\n\n#include \"lobject.h\"\n\n\n/*\n* WARNING: if you change the order of this enumeration,\n* grep \"ORDER TM\"\n*/\ntypedef enum {\n TM_INDEX,\n TM_NEWINDEX,\n TM_GC,\n TM_MODE,\n TM_EQ, /* last tag method with `fast' access */\n TM_ADD,\n TM_SUB,\n TM_MUL,\n TM_DIV,\n TM_MOD,\n TM_POW,\n TM_UNM,\n TM_LEN,\n TM_LT,\n TM_LE,\n TM_CONCAT,\n TM_CALL,\n TM_N\t\t/* number of elements in the enum */\n} TMS;\n\n\n\n#define gfasttm(g,et,e) ((et) == NULL ? NULL : \\\n ((et)->flags & (1u<<(e))) ? NULL : luaT_gettm(et, e, (g)->tmname[e]))\n\n#define fasttm(l,et,e)\tgfasttm(G(l), et, e)\n\nLUAI_DATA const char *const luaT_typenames[];\n\n\nLUAI_FUNC const TValue *luaT_gettm (Table *events, TMS event, TString *ename);\nLUAI_FUNC const TValue *luaT_gettmbyobj (lua_State *L, const TValue *o,\n TMS event);\nLUAI_FUNC void luaT_init (lua_State *L);\n\n#endif\n" }, { "path": "deps/lua/src/lua.c", "content": "/*\n** $Id: lua.c,v 1.160.1.2 2007/12/28 15:32:23 roberto Exp $\n** Lua stand-alone interpreter\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n#include \n\n#define lua_c\n\n#include \"lua.h\"\n\n#include \"lauxlib.h\"\n#include \"lualib.h\"\n\n\n\nstatic lua_State *globalL = NULL;\n\nstatic const char *progname = LUA_PROGNAME;\n\n\n\nstatic void lstop (lua_State *L, lua_Debug *ar) {\n (void)ar; /* unused arg. */\n lua_sethook(L, NULL, 0, 0);\n luaL_error(L, \"interrupted!\");\n}\n\n\nstatic void laction (int i) {\n signal(i, SIG_DFL); /* if another SIGINT happens before lstop,\n terminate process (default action) */\n lua_sethook(globalL, lstop, LUA_MASKCALL | LUA_MASKRET | LUA_MASKCOUNT, 1);\n}\n\n\nstatic void print_usage (void) {\n fprintf(stderr,\n \"usage: %s [options] [script [args]].\\n\"\n \"Available options are:\\n\"\n \" -e stat execute string \" LUA_QL(\"stat\") \"\\n\"\n \" -l name require library \" LUA_QL(\"name\") \"\\n\"\n \" -i enter interactive mode after executing \" LUA_QL(\"script\") \"\\n\"\n \" -v show version information\\n\"\n \" -- stop handling options\\n\"\n \" - execute stdin and stop handling options\\n\"\n ,\n progname);\n fflush(stderr);\n}\n\n\nstatic void l_message (const char *pname, const char *msg) {\n if (pname) fprintf(stderr, \"%s: \", pname);\n fprintf(stderr, \"%s\\n\", msg);\n fflush(stderr);\n}\n\n\nstatic int report (lua_State *L, int status) {\n if (status && !lua_isnil(L, -1)) {\n const char *msg = lua_tostring(L, -1);\n if (msg == NULL) msg = \"(error object is not a string)\";\n l_message(progname, msg);\n lua_pop(L, 1);\n }\n return status;\n}\n\n\nstatic int traceback (lua_State *L) {\n if (!lua_isstring(L, 1)) /* 'message' not a string? */\n return 1; /* keep it intact */\n lua_getfield(L, LUA_GLOBALSINDEX, \"debug\");\n if (!lua_istable(L, -1)) {\n lua_pop(L, 1);\n return 1;\n }\n lua_getfield(L, -1, \"traceback\");\n if (!lua_isfunction(L, -1)) {\n lua_pop(L, 2);\n return 1;\n }\n lua_pushvalue(L, 1); /* pass error message */\n lua_pushinteger(L, 2); /* skip this function and traceback */\n lua_call(L, 2, 1); /* call debug.traceback */\n return 1;\n}\n\n\nstatic int docall (lua_State *L, int narg, int clear) {\n int status;\n int base = lua_gettop(L) - narg; /* function index */\n lua_pushcfunction(L, traceback); /* push traceback function */\n lua_insert(L, base); /* put it under chunk and args */\n signal(SIGINT, laction);\n status = lua_pcall(L, narg, (clear ? 0 : LUA_MULTRET), base);\n signal(SIGINT, SIG_DFL);\n lua_remove(L, base); /* remove traceback function */\n /* force a complete garbage collection in case of errors */\n if (status != 0) lua_gc(L, LUA_GCCOLLECT, 0);\n return status;\n}\n\n\nstatic void print_version (void) {\n l_message(NULL, LUA_RELEASE \" \" LUA_COPYRIGHT);\n}\n\n\nstatic int getargs (lua_State *L, char **argv, int n) {\n int narg;\n int i;\n int argc = 0;\n while (argv[argc]) argc++; /* count total number of arguments */\n narg = argc - (n + 1); /* number of arguments to the script */\n luaL_checkstack(L, narg + 3, \"too many arguments to script\");\n for (i=n+1; i < argc; i++)\n lua_pushstring(L, argv[i]);\n lua_createtable(L, narg, n + 1);\n for (i=0; i < argc; i++) {\n lua_pushstring(L, argv[i]);\n lua_rawseti(L, -2, i - n);\n }\n return narg;\n}\n\n\nstatic int dofile (lua_State *L, const char *name) {\n int status = luaL_loadfile(L, name) || docall(L, 0, 1);\n return report(L, status);\n}\n\n\nstatic int dostring (lua_State *L, const char *s, const char *name) {\n int status = luaL_loadbuffer(L, s, strlen(s), name) || docall(L, 0, 1);\n return report(L, status);\n}\n\n\nstatic int dolibrary (lua_State *L, const char *name) {\n lua_getglobal(L, \"require\");\n lua_pushstring(L, name);\n return report(L, docall(L, 1, 1));\n}\n\n\nstatic const char *get_prompt (lua_State *L, int firstline) {\n const char *p;\n lua_getfield(L, LUA_GLOBALSINDEX, firstline ? \"_PROMPT\" : \"_PROMPT2\");\n p = lua_tostring(L, -1);\n if (p == NULL) p = (firstline ? LUA_PROMPT : LUA_PROMPT2);\n lua_pop(L, 1); /* remove global */\n return p;\n}\n\n\nstatic int incomplete (lua_State *L, int status) {\n if (status == LUA_ERRSYNTAX) {\n size_t lmsg;\n const char *msg = lua_tolstring(L, -1, &lmsg);\n const char *tp = msg + lmsg - (sizeof(LUA_QL(\"\")) - 1);\n if (strstr(msg, LUA_QL(\"\")) == tp) {\n lua_pop(L, 1);\n return 1;\n }\n }\n return 0; /* else... */\n}\n\n\nstatic int pushline (lua_State *L, int firstline) {\n char buffer[LUA_MAXINPUT];\n char *b = buffer;\n size_t l;\n const char *prmt = get_prompt(L, firstline);\n if (lua_readline(L, b, prmt) == 0)\n return 0; /* no input */\n l = strlen(b);\n if (l > 0 && b[l-1] == '\\n') /* line ends with newline? */\n b[l-1] = '\\0'; /* remove it */\n if (firstline && b[0] == '=') /* first line starts with `=' ? */\n lua_pushfstring(L, \"return %s\", b+1); /* change it to `return' */\n else\n lua_pushstring(L, b);\n lua_freeline(L, b);\n return 1;\n}\n\n\nstatic int loadline (lua_State *L) {\n int status;\n lua_settop(L, 0);\n if (!pushline(L, 1))\n return -1; /* no input */\n for (;;) { /* repeat until gets a complete line */\n status = luaL_loadbuffer(L, lua_tostring(L, 1), lua_strlen(L, 1), \"=stdin\");\n if (!incomplete(L, status)) break; /* cannot try to add lines? */\n if (!pushline(L, 0)) /* no more input? */\n return -1;\n lua_pushliteral(L, \"\\n\"); /* add a new line... */\n lua_insert(L, -2); /* ...between the two lines */\n lua_concat(L, 3); /* join them */\n }\n lua_saveline(L, 1);\n lua_remove(L, 1); /* remove line */\n return status;\n}\n\n\nstatic void dotty (lua_State *L) {\n int status;\n const char *oldprogname = progname;\n progname = NULL;\n while ((status = loadline(L)) != -1) {\n if (status == 0) status = docall(L, 0, 0);\n report(L, status);\n if (status == 0 && lua_gettop(L) > 0) { /* any result to print? */\n lua_getglobal(L, \"print\");\n lua_insert(L, 1);\n if (lua_pcall(L, lua_gettop(L)-1, 0, 0) != 0)\n l_message(progname, lua_pushfstring(L,\n \"error calling \" LUA_QL(\"print\") \" (%s)\",\n lua_tostring(L, -1)));\n }\n }\n lua_settop(L, 0); /* clear stack */\n fputs(\"\\n\", stdout);\n fflush(stdout);\n progname = oldprogname;\n}\n\n\nstatic int handle_script (lua_State *L, char **argv, int n) {\n int status;\n const char *fname;\n int narg = getargs(L, argv, n); /* collect arguments */\n lua_setglobal(L, \"arg\");\n fname = argv[n];\n if (strcmp(fname, \"-\") == 0 && strcmp(argv[n-1], \"--\") != 0) \n fname = NULL; /* stdin */\n status = luaL_loadfile(L, fname);\n lua_insert(L, -(narg+1));\n if (status == 0)\n status = docall(L, narg, 0);\n else\n lua_pop(L, narg); \n return report(L, status);\n}\n\n\n/* check that argument has no extra characters at the end */\n#define notail(x)\t{if ((x)[2] != '\\0') return -1;}\n\n\nstatic int collectargs (char **argv, int *pi, int *pv, int *pe) {\n int i;\n for (i = 1; argv[i] != NULL; i++) {\n if (argv[i][0] != '-') /* not an option? */\n return i;\n switch (argv[i][1]) { /* option */\n case '-':\n notail(argv[i]);\n return (argv[i+1] != NULL ? i+1 : 0);\n case '\\0':\n return i;\n case 'i':\n notail(argv[i]);\n *pi = 1; /* go through */\n case 'v':\n notail(argv[i]);\n *pv = 1;\n break;\n case 'e':\n *pe = 1; /* go through */\n case 'l':\n if (argv[i][2] == '\\0') {\n i++;\n if (argv[i] == NULL) return -1;\n }\n break;\n default: return -1; /* invalid option */\n }\n }\n return 0;\n}\n\n\nstatic int runargs (lua_State *L, char **argv, int n) {\n int i;\n for (i = 1; i < n; i++) {\n if (argv[i] == NULL) continue;\n lua_assert(argv[i][0] == '-');\n switch (argv[i][1]) { /* option */\n case 'e': {\n const char *chunk = argv[i] + 2;\n if (*chunk == '\\0') chunk = argv[++i];\n lua_assert(chunk != NULL);\n if (dostring(L, chunk, \"=(command line)\") != 0)\n return 1;\n break;\n }\n case 'l': {\n const char *filename = argv[i] + 2;\n if (*filename == '\\0') filename = argv[++i];\n lua_assert(filename != NULL);\n if (dolibrary(L, filename))\n return 1; /* stop if file fails */\n break;\n }\n default: break;\n }\n }\n return 0;\n}\n\n\nstatic int handle_luainit (lua_State *L) {\n const char *init = getenv(LUA_INIT);\n if (init == NULL) return 0; /* status OK */\n else if (init[0] == '@')\n return dofile(L, init+1);\n else\n return dostring(L, init, \"=\" LUA_INIT);\n}\n\n\nstruct Smain {\n int argc;\n char **argv;\n int status;\n};\n\n\nstatic int pmain (lua_State *L) {\n struct Smain *s = (struct Smain *)lua_touserdata(L, 1);\n char **argv = s->argv;\n int script;\n int has_i = 0, has_v = 0, has_e = 0;\n globalL = L;\n if (argv[0] && argv[0][0]) progname = argv[0];\n lua_gc(L, LUA_GCSTOP, 0); /* stop collector during initialization */\n luaL_openlibs(L); /* open libraries */\n lua_gc(L, LUA_GCRESTART, 0);\n s->status = handle_luainit(L);\n if (s->status != 0) return 0;\n script = collectargs(argv, &has_i, &has_v, &has_e);\n if (script < 0) { /* invalid args? */\n print_usage();\n s->status = 1;\n return 0;\n }\n if (has_v) print_version();\n s->status = runargs(L, argv, (script > 0) ? script : s->argc);\n if (s->status != 0) return 0;\n if (script)\n s->status = handle_script(L, argv, script);\n if (s->status != 0) return 0;\n if (has_i)\n dotty(L);\n else if (script == 0 && !has_e && !has_v) {\n if (lua_stdin_is_tty()) {\n print_version();\n dotty(L);\n }\n else dofile(L, NULL); /* executes stdin as a file */\n }\n return 0;\n}\n\n\nint main (int argc, char **argv) {\n int status;\n struct Smain s;\n lua_State *L = lua_open(); /* create state */\n if (L == NULL) {\n l_message(argv[0], \"cannot create state: not enough memory\");\n return EXIT_FAILURE;\n }\n s.argc = argc;\n s.argv = argv;\n status = lua_cpcall(L, &pmain, &s);\n report(L, status);\n lua_close(L);\n return (status || s.status) ? EXIT_FAILURE : EXIT_SUCCESS;\n}\n\n" }, { "path": "deps/lua/src/lua.h", "content": "/*\n** $Id: lua.h,v 1.218.1.7 2012/01/13 20:36:20 roberto Exp $\n** Lua - An Extensible Extension Language\n** Lua.org, PUC-Rio, Brazil (http://www.lua.org)\n** See Copyright Notice at the end of this file\n*/\n\n\n#ifndef lua_h\n#define lua_h\n\n#include \n#include \n\n\n#include \"luaconf.h\"\n\n\n#define LUA_VERSION\t\"Lua 5.1\"\n#define LUA_RELEASE\t\"Lua 5.1.5\"\n#define LUA_VERSION_NUM\t501\n#define LUA_COPYRIGHT\t\"Copyright (C) 1994-2012 Lua.org, PUC-Rio\"\n#define LUA_AUTHORS \t\"R. Ierusalimschy, L. H. de Figueiredo & W. Celes\"\n\n\n/* mark for precompiled code (`Lua') */\n#define\tLUA_SIGNATURE\t\"\\033Lua\"\n\n/* option for multiple returns in `lua_pcall' and `lua_call' */\n#define LUA_MULTRET\t(-1)\n\n\n/*\n** pseudo-indices\n*/\n#define LUA_REGISTRYINDEX\t(-10000)\n#define LUA_ENVIRONINDEX\t(-10001)\n#define LUA_GLOBALSINDEX\t(-10002)\n#define lua_upvalueindex(i)\t(LUA_GLOBALSINDEX-(i))\n\n\n/* thread status; 0 is OK */\n#define LUA_YIELD\t1\n#define LUA_ERRRUN\t2\n#define LUA_ERRSYNTAX\t3\n#define LUA_ERRMEM\t4\n#define LUA_ERRERR\t5\n\n\ntypedef struct lua_State lua_State;\n\ntypedef int (*lua_CFunction) (lua_State *L);\n\n\n/*\n** functions that read/write blocks when loading/dumping Lua chunks\n*/\ntypedef const char * (*lua_Reader) (lua_State *L, void *ud, size_t *sz);\n\ntypedef int (*lua_Writer) (lua_State *L, const void* p, size_t sz, void* ud);\n\n\n/*\n** prototype for memory-allocation functions\n*/\ntypedef void * (*lua_Alloc) (void *ud, void *ptr, size_t osize, size_t nsize);\n\n\n/*\n** basic types\n*/\n#define LUA_TNONE\t\t(-1)\n\n#define LUA_TNIL\t\t0\n#define LUA_TBOOLEAN\t\t1\n#define LUA_TLIGHTUSERDATA\t2\n#define LUA_TNUMBER\t\t3\n#define LUA_TSTRING\t\t4\n#define LUA_TTABLE\t\t5\n#define LUA_TFUNCTION\t\t6\n#define LUA_TUSERDATA\t\t7\n#define LUA_TTHREAD\t\t8\n\n\n\n/* minimum Lua stack available to a C function */\n#define LUA_MINSTACK\t20\n\n\n/*\n** generic extra include file\n*/\n#if defined(LUA_USER_H)\n#include LUA_USER_H\n#endif\n\n\n/* type of numbers in Lua */\ntypedef LUA_NUMBER lua_Number;\n\n\n/* type for integer functions */\ntypedef LUA_INTEGER lua_Integer;\n\n\n\n/*\n** state manipulation\n*/\nLUA_API lua_State *(lua_newstate) (lua_Alloc f, void *ud);\nLUA_API void (lua_close) (lua_State *L);\nLUA_API lua_State *(lua_newthread) (lua_State *L);\n\nLUA_API lua_CFunction (lua_atpanic) (lua_State *L, lua_CFunction panicf);\n\n\n/*\n** basic stack manipulation\n*/\nLUA_API int (lua_gettop) (lua_State *L);\nLUA_API void (lua_settop) (lua_State *L, int idx);\nLUA_API void (lua_pushvalue) (lua_State *L, int idx);\nLUA_API void (lua_remove) (lua_State *L, int idx);\nLUA_API void (lua_insert) (lua_State *L, int idx);\nLUA_API void (lua_replace) (lua_State *L, int idx);\nLUA_API int (lua_checkstack) (lua_State *L, int sz);\n\nLUA_API void (lua_xmove) (lua_State *from, lua_State *to, int n);\n\n\n/*\n** access functions (stack -> C)\n*/\n\nLUA_API int (lua_isnumber) (lua_State *L, int idx);\nLUA_API int (lua_isstring) (lua_State *L, int idx);\nLUA_API int (lua_iscfunction) (lua_State *L, int idx);\nLUA_API int (lua_isuserdata) (lua_State *L, int idx);\nLUA_API int (lua_type) (lua_State *L, int idx);\nLUA_API const char *(lua_typename) (lua_State *L, int tp);\n\nLUA_API int (lua_equal) (lua_State *L, int idx1, int idx2);\nLUA_API int (lua_rawequal) (lua_State *L, int idx1, int idx2);\nLUA_API int (lua_lessthan) (lua_State *L, int idx1, int idx2);\n\nLUA_API lua_Number (lua_tonumber) (lua_State *L, int idx);\nLUA_API lua_Integer (lua_tointeger) (lua_State *L, int idx);\nLUA_API int (lua_toboolean) (lua_State *L, int idx);\nLUA_API const char *(lua_tolstring) (lua_State *L, int idx, size_t *len);\nLUA_API size_t (lua_objlen) (lua_State *L, int idx);\nLUA_API lua_CFunction (lua_tocfunction) (lua_State *L, int idx);\nLUA_API void\t *(lua_touserdata) (lua_State *L, int idx);\nLUA_API lua_State *(lua_tothread) (lua_State *L, int idx);\nLUA_API const void *(lua_topointer) (lua_State *L, int idx);\n\n\n/*\n** push functions (C -> stack)\n*/\nLUA_API void (lua_pushnil) (lua_State *L);\nLUA_API void (lua_pushnumber) (lua_State *L, lua_Number n);\nLUA_API void (lua_pushinteger) (lua_State *L, lua_Integer n);\nLUA_API void (lua_pushlstring) (lua_State *L, const char *s, size_t l);\nLUA_API void (lua_pushstring) (lua_State *L, const char *s);\nLUA_API const char *(lua_pushvfstring) (lua_State *L, const char *fmt,\n va_list argp);\nLUA_API const char *(lua_pushfstring) (lua_State *L, const char *fmt, ...);\nLUA_API void (lua_pushcclosure) (lua_State *L, lua_CFunction fn, int n);\nLUA_API void (lua_pushboolean) (lua_State *L, int b);\nLUA_API void (lua_pushlightuserdata) (lua_State *L, void *p);\nLUA_API int (lua_pushthread) (lua_State *L);\n\n\n/*\n** get functions (Lua -> stack)\n*/\nLUA_API void (lua_gettable) (lua_State *L, int idx);\nLUA_API void (lua_getfield) (lua_State *L, int idx, const char *k);\nLUA_API void (lua_rawget) (lua_State *L, int idx);\nLUA_API void (lua_rawgeti) (lua_State *L, int idx, int n);\nLUA_API void (lua_createtable) (lua_State *L, int narr, int nrec);\nLUA_API void *(lua_newuserdata) (lua_State *L, size_t sz);\nLUA_API int (lua_getmetatable) (lua_State *L, int objindex);\nLUA_API void (lua_getfenv) (lua_State *L, int idx);\n\n\n/*\n** set functions (stack -> Lua)\n*/\nLUA_API void (lua_settable) (lua_State *L, int idx);\nLUA_API void (lua_setfield) (lua_State *L, int idx, const char *k);\nLUA_API void (lua_rawset) (lua_State *L, int idx);\nLUA_API void (lua_rawseti) (lua_State *L, int idx, int n);\nLUA_API int (lua_setmetatable) (lua_State *L, int objindex);\nLUA_API int (lua_setfenv) (lua_State *L, int idx);\n\n\n/*\n** `load' and `call' functions (load and run Lua code)\n*/\nLUA_API void (lua_call) (lua_State *L, int nargs, int nresults);\nLUA_API int (lua_pcall) (lua_State *L, int nargs, int nresults, int errfunc);\nLUA_API int (lua_cpcall) (lua_State *L, lua_CFunction func, void *ud);\nLUA_API int (lua_load) (lua_State *L, lua_Reader reader, void *dt,\n const char *chunkname);\n\nLUA_API int (lua_dump) (lua_State *L, lua_Writer writer, void *data);\n\n\n/*\n** coroutine functions\n*/\nLUA_API int (lua_yield) (lua_State *L, int nresults);\nLUA_API int (lua_resume) (lua_State *L, int narg);\nLUA_API int (lua_status) (lua_State *L);\n\n/*\n** garbage-collection function and options\n*/\n\n#define LUA_GCSTOP\t\t0\n#define LUA_GCRESTART\t\t1\n#define LUA_GCCOLLECT\t\t2\n#define LUA_GCCOUNT\t\t3\n#define LUA_GCCOUNTB\t\t4\n#define LUA_GCSTEP\t\t5\n#define LUA_GCSETPAUSE\t\t6\n#define LUA_GCSETSTEPMUL\t7\n\nLUA_API int (lua_gc) (lua_State *L, int what, int data);\n\n\n/*\n** miscellaneous functions\n*/\n\nLUA_API int (lua_error) (lua_State *L);\n\nLUA_API int (lua_next) (lua_State *L, int idx);\n\nLUA_API void (lua_concat) (lua_State *L, int n);\n\nLUA_API lua_Alloc (lua_getallocf) (lua_State *L, void **ud);\nLUA_API void lua_setallocf (lua_State *L, lua_Alloc f, void *ud);\n\n\n\n/* \n** ===============================================================\n** some useful macros\n** ===============================================================\n*/\n\n#define lua_pop(L,n)\t\tlua_settop(L, -(n)-1)\n\n#define lua_newtable(L)\t\tlua_createtable(L, 0, 0)\n\n#define lua_register(L,n,f) (lua_pushcfunction(L, (f)), lua_setglobal(L, (n)))\n\n#define lua_pushcfunction(L,f)\tlua_pushcclosure(L, (f), 0)\n\n#define lua_strlen(L,i)\t\tlua_objlen(L, (i))\n\n#define lua_isfunction(L,n)\t(lua_type(L, (n)) == LUA_TFUNCTION)\n#define lua_istable(L,n)\t(lua_type(L, (n)) == LUA_TTABLE)\n#define lua_islightuserdata(L,n)\t(lua_type(L, (n)) == LUA_TLIGHTUSERDATA)\n#define lua_isnil(L,n)\t\t(lua_type(L, (n)) == LUA_TNIL)\n#define lua_isboolean(L,n)\t(lua_type(L, (n)) == LUA_TBOOLEAN)\n#define lua_isthread(L,n)\t(lua_type(L, (n)) == LUA_TTHREAD)\n#define lua_isnone(L,n)\t\t(lua_type(L, (n)) == LUA_TNONE)\n#define lua_isnoneornil(L, n)\t(lua_type(L, (n)) <= 0)\n\n#define lua_pushliteral(L, s)\t\\\n\tlua_pushlstring(L, \"\" s, (sizeof(s)/sizeof(char))-1)\n\n#define lua_setglobal(L,s)\tlua_setfield(L, LUA_GLOBALSINDEX, (s))\n#define lua_getglobal(L,s)\tlua_getfield(L, LUA_GLOBALSINDEX, (s))\n\n#define lua_tostring(L,i)\tlua_tolstring(L, (i), NULL)\n\n\n\n/*\n** compatibility macros and functions\n*/\n\n#define lua_open()\tluaL_newstate()\n\n#define lua_getregistry(L)\tlua_pushvalue(L, LUA_REGISTRYINDEX)\n\n#define lua_getgccount(L)\tlua_gc(L, LUA_GCCOUNT, 0)\n\n#define lua_Chunkreader\t\tlua_Reader\n#define lua_Chunkwriter\t\tlua_Writer\n\n\n/* hack */\nLUA_API void lua_setlevel\t(lua_State *from, lua_State *to);\n\n\n/*\n** {======================================================================\n** Debug API\n** =======================================================================\n*/\n\n\n/*\n** Event codes\n*/\n#define LUA_HOOKCALL\t0\n#define LUA_HOOKRET\t1\n#define LUA_HOOKLINE\t2\n#define LUA_HOOKCOUNT\t3\n#define LUA_HOOKTAILRET 4\n\n\n/*\n** Event masks\n*/\n#define LUA_MASKCALL\t(1 << LUA_HOOKCALL)\n#define LUA_MASKRET\t(1 << LUA_HOOKRET)\n#define LUA_MASKLINE\t(1 << LUA_HOOKLINE)\n#define LUA_MASKCOUNT\t(1 << LUA_HOOKCOUNT)\n\ntypedef struct lua_Debug lua_Debug; /* activation record */\n\n\n/* Functions to be called by the debuger in specific events */\ntypedef void (*lua_Hook) (lua_State *L, lua_Debug *ar);\n\n\nLUA_API int lua_getstack (lua_State *L, int level, lua_Debug *ar);\nLUA_API int lua_getinfo (lua_State *L, const char *what, lua_Debug *ar);\nLUA_API const char *lua_getlocal (lua_State *L, const lua_Debug *ar, int n);\nLUA_API const char *lua_setlocal (lua_State *L, const lua_Debug *ar, int n);\nLUA_API const char *lua_getupvalue (lua_State *L, int funcindex, int n);\nLUA_API const char *lua_setupvalue (lua_State *L, int funcindex, int n);\n\nLUA_API int lua_sethook (lua_State *L, lua_Hook func, int mask, int count);\nLUA_API lua_Hook lua_gethook (lua_State *L);\nLUA_API int lua_gethookmask (lua_State *L);\nLUA_API int lua_gethookcount (lua_State *L);\n\n\nstruct lua_Debug {\n int event;\n const char *name;\t/* (n) */\n const char *namewhat;\t/* (n) `global', `local', `field', `method' */\n const char *what;\t/* (S) `Lua', `C', `main', `tail' */\n const char *source;\t/* (S) */\n int currentline;\t/* (l) */\n int nups;\t\t/* (u) number of upvalues */\n int linedefined;\t/* (S) */\n int lastlinedefined;\t/* (S) */\n char short_src[LUA_IDSIZE]; /* (S) */\n /* private part */\n int i_ci; /* active function */\n};\n\n/* }====================================================================== */\n\n\n/******************************************************************************\n* Copyright (C) 1994-2012 Lua.org, PUC-Rio. All rights reserved.\n*\n* Permission is hereby granted, free of charge, to any person obtaining\n* a copy of this software and associated documentation files (the\n* \"Software\"), to deal in the Software without restriction, including\n* without limitation the rights to use, copy, modify, merge, publish,\n* distribute, sublicense, and/or sell copies of the Software, and to\n* permit persons to whom the Software is furnished to do so, subject to\n* the following conditions:\n*\n* The above copyright notice and this permission notice shall be\n* included in all copies or substantial portions of the Software.\n*\n* THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n******************************************************************************/\n\n\n#endif\n" }, { "path": "deps/lua/src/lua_cjson.c", "content": "#define VERSION \"1.0.3\"\n\n/* CJSON - JSON support for Lua\n *\n * Copyright (c) 2010-2011 Mark Pulford \n *\n * Permission is hereby granted, free of charge, to any person obtaining\n * a copy of this software and associated documentation files (the\n * \"Software\"), to deal in the Software without restriction, including\n * without limitation the rights to use, copy, modify, merge, publish,\n * distribute, sublicense, and/or sell copies of the Software, and to\n * permit persons to whom the Software is furnished to do so, subject to\n * the following conditions:\n *\n * The above copyright notice and this permission notice shall be\n * included in all copies or substantial portions of the Software.\n *\n * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n */\n\n/* Caveats:\n * - JSON \"null\" values are represented as lightuserdata since Lua\n * tables cannot contain \"nil\". Compare with cjson.null.\n * - Invalid UTF-8 characters are not detected and will be passed\n * untouched. If required, UTF-8 error checking should be done\n * outside this library.\n * - Javascript comments are not part of the JSON spec, and are not\n * currently supported.\n *\n * Note: Decoding is slower than encoding. Lua spends significant\n * time (30%) managing tables when parsing JSON since it is\n * difficult to know object/array sizes ahead of time.\n */\n\n#include \n#include \n#include \n#include \"lua.h\"\n#include \"lauxlib.h\"\n\n#include \"strbuf.h\"\n\n#ifdef MISSING_ISINF\n#define isinf(x) (!isnan(x) && isnan((x) - (x)))\n#endif\n\n#define DEFAULT_SPARSE_CONVERT 0\n#define DEFAULT_SPARSE_RATIO 2\n#define DEFAULT_SPARSE_SAFE 10\n#define DEFAULT_MAX_DEPTH 20\n#define DEFAULT_ENCODE_REFUSE_BADNUM 1\n#define DEFAULT_DECODE_REFUSE_BADNUM 0\n#define DEFAULT_ENCODE_KEEP_BUFFER 1\n\ntypedef enum {\n T_OBJ_BEGIN,\n T_OBJ_END,\n T_ARR_BEGIN,\n T_ARR_END,\n T_STRING,\n T_NUMBER,\n T_BOOLEAN,\n T_NULL,\n T_COLON,\n T_COMMA,\n T_END,\n T_WHITESPACE,\n T_ERROR,\n T_UNKNOWN\n} json_token_type_t;\n\nstatic const char *json_token_type_name[] = {\n \"T_OBJ_BEGIN\",\n \"T_OBJ_END\",\n \"T_ARR_BEGIN\",\n \"T_ARR_END\",\n \"T_STRING\",\n \"T_NUMBER\",\n \"T_BOOLEAN\",\n \"T_NULL\",\n \"T_COLON\",\n \"T_COMMA\",\n \"T_END\",\n \"T_WHITESPACE\",\n \"T_ERROR\",\n \"T_UNKNOWN\",\n NULL\n};\n\ntypedef struct {\n json_token_type_t ch2token[256];\n char escape2char[256]; /* Decoding */\n#if 0\n char escapes[35][8]; /* Pre-generated escape string buffer */\n char *char2escape[256]; /* Encoding */\n#endif\n strbuf_t encode_buf;\n char number_fmt[8]; /* \"%.XXg\\0\" */\n int current_depth;\n\n int encode_sparse_convert;\n int encode_sparse_ratio;\n int encode_sparse_safe;\n int encode_max_depth;\n int encode_refuse_badnum;\n int decode_refuse_badnum;\n int encode_keep_buffer;\n int encode_number_precision;\n} json_config_t;\n\ntypedef struct {\n const char *data;\n int index;\n strbuf_t *tmp; /* Temporary storage for strings */\n json_config_t *cfg;\n} json_parse_t;\n\ntypedef struct {\n json_token_type_t type;\n int index;\n union {\n const char *string;\n double number;\n int boolean;\n } value;\n int string_len;\n} json_token_t;\n\nstatic const char *char2escape[256] = {\n \"\\\\u0000\", \"\\\\u0001\", \"\\\\u0002\", \"\\\\u0003\",\n \"\\\\u0004\", \"\\\\u0005\", \"\\\\u0006\", \"\\\\u0007\",\n \"\\\\b\", \"\\\\t\", \"\\\\n\", \"\\\\u000b\",\n \"\\\\f\", \"\\\\r\", \"\\\\u000e\", \"\\\\u000f\",\n \"\\\\u0010\", \"\\\\u0011\", \"\\\\u0012\", \"\\\\u0013\",\n \"\\\\u0014\", \"\\\\u0015\", \"\\\\u0016\", \"\\\\u0017\",\n \"\\\\u0018\", \"\\\\u0019\", \"\\\\u001a\", \"\\\\u001b\",\n \"\\\\u001c\", \"\\\\u001d\", \"\\\\u001e\", \"\\\\u001f\",\n NULL, NULL, \"\\\\\\\"\", NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, \"\\\\/\",\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, \"\\\\\\\\\", NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, \"\\\\u007f\",\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,\n};\n\nstatic int json_config_key;\n\n/* ===== CONFIGURATION ===== */\n\nstatic json_config_t *json_fetch_config(lua_State *l)\n{\n json_config_t *cfg;\n\n lua_pushlightuserdata(l, &json_config_key);\n lua_gettable(l, LUA_REGISTRYINDEX);\n cfg = lua_touserdata(l, -1);\n if (!cfg)\n luaL_error(l, \"BUG: Unable to fetch CJSON configuration\");\n\n lua_pop(l, 1);\n\n return cfg;\n}\n\nstatic void json_verify_arg_count(lua_State *l, int args)\n{\n luaL_argcheck(l, lua_gettop(l) <= args, args + 1,\n \"found too many arguments\");\n}\n\n/* Configures handling of extremely sparse arrays:\n * convert: Convert extremely sparse arrays into objects? Otherwise error.\n * ratio: 0: always allow sparse; 1: never allow sparse; >1: use ratio\n * safe: Always use an array when the max index <= safe */\nstatic int json_cfg_encode_sparse_array(lua_State *l)\n{\n json_config_t *cfg;\n int val;\n\n json_verify_arg_count(l, 3);\n cfg = json_fetch_config(l);\n\n switch (lua_gettop(l)) {\n case 3:\n val = luaL_checkinteger(l, 3);\n luaL_argcheck(l, val >= 0, 3, \"expected integer >= 0\");\n cfg->encode_sparse_safe = val;\n case 2:\n val = luaL_checkinteger(l, 2);\n luaL_argcheck(l, val >= 0, 2, \"expected integer >= 0\");\n cfg->encode_sparse_ratio = val;\n case 1:\n luaL_argcheck(l, lua_isboolean(l, 1), 1, \"expected boolean\");\n cfg->encode_sparse_convert = lua_toboolean(l, 1);\n }\n\n lua_pushboolean(l, cfg->encode_sparse_convert);\n lua_pushinteger(l, cfg->encode_sparse_ratio);\n lua_pushinteger(l, cfg->encode_sparse_safe);\n\n return 3;\n}\n\n/* Configures the maximum number of nested arrays/objects allowed when\n * encoding */\nstatic int json_cfg_encode_max_depth(lua_State *l)\n{\n json_config_t *cfg;\n int depth;\n\n json_verify_arg_count(l, 1);\n cfg = json_fetch_config(l);\n\n if (lua_gettop(l)) {\n depth = luaL_checkinteger(l, 1);\n luaL_argcheck(l, depth > 0, 1, \"expected positive integer\");\n cfg->encode_max_depth = depth;\n }\n\n lua_pushinteger(l, cfg->encode_max_depth);\n\n return 1;\n}\n\nstatic void json_set_number_precision(json_config_t *cfg, int prec)\n{\n cfg->encode_number_precision = prec;\n sprintf(cfg->number_fmt, \"%%.%dg\", prec);\n}\n\n/* Configures number precision when converting doubles to text */\nstatic int json_cfg_encode_number_precision(lua_State *l)\n{\n json_config_t *cfg;\n int precision;\n\n json_verify_arg_count(l, 1);\n cfg = json_fetch_config(l);\n\n if (lua_gettop(l)) {\n precision = luaL_checkinteger(l, 1);\n luaL_argcheck(l, 1 <= precision && precision <= 14, 1,\n \"expected integer between 1 and 14\");\n json_set_number_precision(cfg, precision);\n }\n\n lua_pushinteger(l, cfg->encode_number_precision);\n\n return 1;\n}\n\n/* Configures JSON encoding buffer persistence */\nstatic int json_cfg_encode_keep_buffer(lua_State *l)\n{\n json_config_t *cfg;\n\n json_verify_arg_count(l, 1);\n cfg = json_fetch_config(l);\n\n if (lua_gettop(l)) {\n luaL_checktype(l, 1, LUA_TBOOLEAN);\n cfg->encode_keep_buffer = lua_toboolean(l, 1);\n }\n\n lua_pushboolean(l, cfg->encode_keep_buffer);\n\n return 1;\n}\n\n/* On argument: decode enum and set config variables\n * **options must point to a NULL terminated array of 4 enums\n * Returns: current enum value */\nstatic void json_enum_option(lua_State *l, const char **options,\n int *opt1, int *opt2)\n{\n int setting;\n\n if (lua_gettop(l)) {\n if (lua_isboolean(l, 1))\n setting = lua_toboolean(l, 1) * 3;\n else\n setting = luaL_checkoption(l, 1, NULL, options);\n\n *opt1 = setting & 1 ? 1 : 0;\n *opt2 = setting & 2 ? 1 : 0;\n } else {\n setting = *opt1 | (*opt2 << 1);\n }\n\n if (setting)\n lua_pushstring(l, options[setting]);\n else\n lua_pushboolean(l, 0);\n}\n\n\n/* When enabled, rejects: NaN, Infinity, hexidecimal numbers */\nstatic int json_cfg_refuse_invalid_numbers(lua_State *l)\n{\n static const char *options_enc_dec[] = { \"none\", \"encode\", \"decode\",\n \"both\", NULL };\n json_config_t *cfg;\n\n json_verify_arg_count(l, 1);\n cfg = json_fetch_config(l);\n\n json_enum_option(l, options_enc_dec,\n &cfg->encode_refuse_badnum,\n &cfg->decode_refuse_badnum);\n\n return 1;\n}\n\nstatic int json_destroy_config(lua_State *l)\n{\n json_config_t *cfg;\n\n cfg = lua_touserdata(l, 1);\n if (cfg)\n strbuf_free(&cfg->encode_buf);\n cfg = NULL;\n\n return 0;\n}\n\nstatic void json_create_config(lua_State *l)\n{\n json_config_t *cfg;\n int i;\n\n cfg = lua_newuserdata(l, sizeof(*cfg));\n\n /* Create GC method to clean up strbuf */\n lua_newtable(l);\n lua_pushcfunction(l, json_destroy_config);\n lua_setfield(l, -2, \"__gc\");\n lua_setmetatable(l, -2);\n\n strbuf_init(&cfg->encode_buf, 0);\n\n cfg->encode_sparse_convert = DEFAULT_SPARSE_CONVERT;\n cfg->encode_sparse_ratio = DEFAULT_SPARSE_RATIO;\n cfg->encode_sparse_safe = DEFAULT_SPARSE_SAFE;\n cfg->encode_max_depth = DEFAULT_MAX_DEPTH;\n cfg->encode_refuse_badnum = DEFAULT_ENCODE_REFUSE_BADNUM;\n cfg->decode_refuse_badnum = DEFAULT_DECODE_REFUSE_BADNUM;\n cfg->encode_keep_buffer = DEFAULT_ENCODE_KEEP_BUFFER;\n json_set_number_precision(cfg, 14);\n\n /* Decoding init */\n\n /* Tag all characters as an error */\n for (i = 0; i < 256; i++)\n cfg->ch2token[i] = T_ERROR;\n\n /* Set tokens that require no further processing */\n cfg->ch2token['{'] = T_OBJ_BEGIN;\n cfg->ch2token['}'] = T_OBJ_END;\n cfg->ch2token['['] = T_ARR_BEGIN;\n cfg->ch2token[']'] = T_ARR_END;\n cfg->ch2token[','] = T_COMMA;\n cfg->ch2token[':'] = T_COLON;\n cfg->ch2token['\\0'] = T_END;\n cfg->ch2token[' '] = T_WHITESPACE;\n cfg->ch2token['\\t'] = T_WHITESPACE;\n cfg->ch2token['\\n'] = T_WHITESPACE;\n cfg->ch2token['\\r'] = T_WHITESPACE;\n\n /* Update characters that require further processing */\n cfg->ch2token['f'] = T_UNKNOWN; /* false? */\n cfg->ch2token['i'] = T_UNKNOWN; /* inf, ininity? */\n cfg->ch2token['I'] = T_UNKNOWN;\n cfg->ch2token['n'] = T_UNKNOWN; /* null, nan? */\n cfg->ch2token['N'] = T_UNKNOWN;\n cfg->ch2token['t'] = T_UNKNOWN; /* true? */\n cfg->ch2token['\"'] = T_UNKNOWN; /* string? */\n cfg->ch2token['+'] = T_UNKNOWN; /* number? */\n cfg->ch2token['-'] = T_UNKNOWN;\n for (i = 0; i < 10; i++)\n cfg->ch2token['0' + i] = T_UNKNOWN;\n\n /* Lookup table for parsing escape characters */\n for (i = 0; i < 256; i++)\n cfg->escape2char[i] = 0; /* String error */\n cfg->escape2char['\"'] = '\"';\n cfg->escape2char['\\\\'] = '\\\\';\n cfg->escape2char['/'] = '/';\n cfg->escape2char['b'] = '\\b';\n cfg->escape2char['t'] = '\\t';\n cfg->escape2char['n'] = '\\n';\n cfg->escape2char['f'] = '\\f';\n cfg->escape2char['r'] = '\\r';\n cfg->escape2char['u'] = 'u'; /* Unicode parsing required */\n\n\n#if 0\n /* Initialise separate storage for pre-generated escape codes.\n * Escapes 0-31 map directly, 34, 92, 127 follow afterwards to\n * save memory. */\n for (i = 0 ; i < 32; i++)\n sprintf(cfg->escapes[i], \"\\\\u%04x\", i);\n strcpy(cfg->escapes[8], \"\\b\"); /* Override simpler escapes */\n strcpy(cfg->escapes[9], \"\\t\");\n strcpy(cfg->escapes[10], \"\\n\");\n strcpy(cfg->escapes[12], \"\\f\");\n strcpy(cfg->escapes[13], \"\\r\");\n strcpy(cfg->escapes[32], \"\\\\\\\"\"); /* chr(34) */\n strcpy(cfg->escapes[33], \"\\\\\\\\\"); /* chr(92) */\n sprintf(cfg->escapes[34], \"\\\\u%04x\", 127); /* char(127) */\n\n /* Initialise encoding escape lookup table */\n for (i = 0; i < 32; i++)\n cfg->char2escape[i] = cfg->escapes[i];\n for (i = 32; i < 256; i++)\n cfg->char2escape[i] = NULL;\n cfg->char2escape[34] = cfg->escapes[32];\n cfg->char2escape[92] = cfg->escapes[33];\n cfg->char2escape[127] = cfg->escapes[34];\n#endif\n}\n\n/* ===== ENCODING ===== */\n\nstatic void json_encode_exception(lua_State *l, json_config_t *cfg, int lindex,\n const char *reason)\n{\n if (!cfg->encode_keep_buffer)\n strbuf_free(&cfg->encode_buf);\n luaL_error(l, \"Cannot serialise %s: %s\",\n lua_typename(l, lua_type(l, lindex)), reason);\n}\n\n/* json_append_string args:\n * - lua_State\n * - JSON strbuf\n * - String (Lua stack index)\n *\n * Returns nothing. Doesn't remove string from Lua stack */\nstatic void json_append_string(lua_State *l, strbuf_t *json, int lindex)\n{\n const char *escstr;\n int i;\n const char *str;\n size_t len;\n\n str = lua_tolstring(l, lindex, &len);\n\n /* Worst case is len * 6 (all unicode escapes).\n * This buffer is reused constantly for small strings\n * If there are any excess pages, they won't be hit anyway.\n * This gains ~5% speedup. */\n strbuf_ensure_empty_length(json, len * 6 + 2);\n\n strbuf_append_char_unsafe(json, '\\\"');\n for (i = 0; i < len; i++) {\n escstr = char2escape[(unsigned char)str[i]];\n if (escstr)\n strbuf_append_string(json, escstr);\n else\n strbuf_append_char_unsafe(json, str[i]);\n }\n strbuf_append_char_unsafe(json, '\\\"');\n}\n\n/* Find the size of the array on the top of the Lua stack\n * -1 object (not a pure array)\n * >=0 elements in array\n */\nstatic int lua_array_length(lua_State *l, json_config_t *cfg)\n{\n double k;\n int max;\n int items;\n\n max = 0;\n items = 0;\n\n lua_pushnil(l);\n /* table, startkey */\n while (lua_next(l, -2) != 0) {\n /* table, key, value */\n if (lua_type(l, -2) == LUA_TNUMBER &&\n (k = lua_tonumber(l, -2))) {\n /* Integer >= 1 ? */\n if (floor(k) == k && k >= 1) {\n if (k > max)\n max = k;\n items++;\n lua_pop(l, 1);\n continue;\n }\n }\n\n /* Must not be an array (non integer key) */\n lua_pop(l, 2);\n return -1;\n }\n\n /* Encode excessively sparse arrays as objects (if enabled) */\n if (cfg->encode_sparse_ratio > 0 &&\n max > items * cfg->encode_sparse_ratio &&\n max > cfg->encode_sparse_safe) {\n if (!cfg->encode_sparse_convert)\n json_encode_exception(l, cfg, -1, \"excessively sparse array\");\n\n return -1;\n }\n\n return max;\n}\n\nstatic void json_encode_descend(lua_State *l, json_config_t *cfg)\n{\n cfg->current_depth++;\n\n if (cfg->current_depth > cfg->encode_max_depth) {\n if (!cfg->encode_keep_buffer)\n strbuf_free(&cfg->encode_buf);\n luaL_error(l, \"Cannot serialise, excessive nesting (%d)\",\n cfg->current_depth);\n }\n}\n\nstatic void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json);\n\n/* json_append_array args:\n * - lua_State\n * - JSON strbuf\n * - Size of passwd Lua array (top of stack) */\nstatic void json_append_array(lua_State *l, json_config_t *cfg, strbuf_t *json,\n int array_length)\n{\n int comma, i;\n\n json_encode_descend(l, cfg);\n\n strbuf_append_char(json, '[');\n\n comma = 0;\n for (i = 1; i <= array_length; i++) {\n if (comma)\n strbuf_append_char(json, ',');\n else\n comma = 1;\n\n lua_rawgeti(l, -1, i);\n json_append_data(l, cfg, json);\n lua_pop(l, 1);\n }\n\n strbuf_append_char(json, ']');\n\n cfg->current_depth--;\n}\n\nstatic void json_append_number(lua_State *l, strbuf_t *json, int index,\n json_config_t *cfg)\n{\n double num = lua_tonumber(l, index);\n\n if (cfg->encode_refuse_badnum && (isinf(num) || isnan(num)))\n json_encode_exception(l, cfg, index, \"must not be NaN or Inf\");\n\n /* Lowest double printed with %.14g is 21 characters long:\n * -1.7976931348623e+308\n *\n * Use 32 to include the \\0, and a few extra just in case..\n */\n strbuf_append_fmt(json, 32, cfg->number_fmt, num);\n}\n\nstatic void json_append_object(lua_State *l, json_config_t *cfg,\n strbuf_t *json)\n{\n int comma, keytype;\n\n json_encode_descend(l, cfg);\n\n /* Object */\n strbuf_append_char(json, '{');\n\n lua_pushnil(l);\n /* table, startkey */\n comma = 0;\n while (lua_next(l, -2) != 0) {\n if (comma)\n strbuf_append_char(json, ',');\n else\n comma = 1;\n\n /* table, key, value */\n keytype = lua_type(l, -2);\n if (keytype == LUA_TNUMBER) {\n strbuf_append_char(json, '\"');\n json_append_number(l, json, -2, cfg);\n strbuf_append_mem(json, \"\\\":\", 2);\n } else if (keytype == LUA_TSTRING) {\n json_append_string(l, json, -2);\n strbuf_append_char(json, ':');\n } else {\n json_encode_exception(l, cfg, -2,\n \"table key must be a number or string\");\n /* never returns */\n }\n\n /* table, key, value */\n json_append_data(l, cfg, json);\n lua_pop(l, 1);\n /* table, key */\n }\n\n strbuf_append_char(json, '}');\n\n cfg->current_depth--;\n}\n\n/* Serialise Lua data into JSON string. */\nstatic void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json)\n{\n int len;\n\n switch (lua_type(l, -1)) {\n case LUA_TSTRING:\n json_append_string(l, json, -1);\n break;\n case LUA_TNUMBER:\n json_append_number(l, json, -1, cfg);\n break;\n case LUA_TBOOLEAN:\n if (lua_toboolean(l, -1))\n strbuf_append_mem(json, \"true\", 4);\n else\n strbuf_append_mem(json, \"false\", 5);\n break;\n case LUA_TTABLE:\n len = lua_array_length(l, cfg);\n if (len > 0)\n json_append_array(l, cfg, json, len);\n else\n json_append_object(l, cfg, json);\n break;\n case LUA_TNIL:\n strbuf_append_mem(json, \"null\", 4);\n break;\n case LUA_TLIGHTUSERDATA:\n if (lua_touserdata(l, -1) == NULL) {\n strbuf_append_mem(json, \"null\", 4);\n break;\n }\n default:\n /* Remaining types (LUA_TFUNCTION, LUA_TUSERDATA, LUA_TTHREAD,\n * and LUA_TLIGHTUSERDATA) cannot be serialised */\n json_encode_exception(l, cfg, -1, \"type not supported\");\n /* never returns */\n }\n}\n\nstatic int json_encode(lua_State *l)\n{\n json_config_t *cfg;\n char *json;\n int len;\n\n /* Can't use json_verify_arg_count() since we need to ensure\n * there is only 1 argument */\n luaL_argcheck(l, lua_gettop(l) == 1, 1, \"expected 1 argument\");\n\n cfg = json_fetch_config(l);\n cfg->current_depth = 0;\n\n /* Reset the persistent buffer if it exists.\n * Otherwise allocate a new buffer. */\n if (strbuf_allocated(&cfg->encode_buf))\n strbuf_reset(&cfg->encode_buf);\n else\n strbuf_init(&cfg->encode_buf, 0);\n\n json_append_data(l, cfg, &cfg->encode_buf);\n json = strbuf_string(&cfg->encode_buf, &len);\n\n lua_pushlstring(l, json, len);\n\n if (!cfg->encode_keep_buffer)\n strbuf_free(&cfg->encode_buf);\n\n return 1;\n}\n\n/* ===== DECODING ===== */\n\nstatic void json_process_value(lua_State *l, json_parse_t *json,\n json_token_t *token);\n\nstatic int hexdigit2int(char hex)\n{\n if ('0' <= hex && hex <= '9')\n return hex - '0';\n\n /* Force lowercase */\n hex |= 0x20;\n if ('a' <= hex && hex <= 'f')\n return 10 + hex - 'a';\n\n return -1;\n}\n\nstatic int decode_hex4(const char *hex)\n{\n int digit[4];\n int i;\n\n /* Convert ASCII hex digit to numeric digit\n * Note: this returns an error for invalid hex digits, including\n * NULL */\n for (i = 0; i < 4; i++) {\n digit[i] = hexdigit2int(hex[i]);\n if (digit[i] < 0) {\n return -1;\n }\n }\n\n return (digit[0] << 12) +\n (digit[1] << 8) +\n (digit[2] << 4) +\n digit[3];\n}\n\n/* Converts a Unicode codepoint to UTF-8.\n * Returns UTF-8 string length, and up to 4 bytes in *utf8 */\nstatic int codepoint_to_utf8(char *utf8, int codepoint)\n{\n /* 0xxxxxxx */\n if (codepoint <= 0x7F) {\n utf8[0] = codepoint;\n return 1;\n }\n\n /* 110xxxxx 10xxxxxx */\n if (codepoint <= 0x7FF) {\n utf8[0] = (codepoint >> 6) | 0xC0;\n utf8[1] = (codepoint & 0x3F) | 0x80;\n return 2;\n }\n\n /* 1110xxxx 10xxxxxx 10xxxxxx */\n if (codepoint <= 0xFFFF) {\n utf8[0] = (codepoint >> 12) | 0xE0;\n utf8[1] = ((codepoint >> 6) & 0x3F) | 0x80;\n utf8[2] = (codepoint & 0x3F) | 0x80;\n return 3;\n }\n\n /* 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */\n if (codepoint <= 0x1FFFFF) {\n utf8[0] = (codepoint >> 18) | 0xF0;\n utf8[1] = ((codepoint >> 12) & 0x3F) | 0x80;\n utf8[2] = ((codepoint >> 6) & 0x3F) | 0x80;\n utf8[3] = (codepoint & 0x3F) | 0x80;\n return 4;\n }\n\n return 0;\n}\n\n\n/* Called when index pointing to beginning of UTF-16 code escape: \\uXXXX\n * \\u is guaranteed to exist, but the remaining hex characters may be\n * missing.\n * Translate to UTF-8 and append to temporary token string.\n * Must advance index to the next character to be processed.\n * Returns: 0 success\n * -1 error\n */\nstatic int json_append_unicode_escape(json_parse_t *json)\n{\n char utf8[4]; /* Surrogate pairs require 4 UTF-8 bytes */\n int codepoint;\n int surrogate_low;\n int len;\n int escape_len = 6;\n\n /* Fetch UTF-16 code unit */\n codepoint = decode_hex4(&json->data[json->index + 2]);\n if (codepoint < 0)\n return -1;\n\n /* UTF-16 surrogate pairs take the following 2 byte form:\n * 11011 x yyyyyyyyyy\n * When x = 0: y is the high 10 bits of the codepoint\n * x = 1: y is the low 10 bits of the codepoint\n *\n * Check for a surrogate pair (high or low) */\n if ((codepoint & 0xF800) == 0xD800) {\n /* Error if the 1st surrogate is not high */\n if (codepoint & 0x400)\n return -1;\n\n /* Ensure the next code is a unicode escape */\n if (json->data[json->index + escape_len] != '\\\\' ||\n json->data[json->index + escape_len + 1] != 'u') {\n return -1;\n }\n\n /* Fetch the next codepoint */\n surrogate_low = decode_hex4(&json->data[json->index + 2 + escape_len]);\n if (surrogate_low < 0)\n return -1;\n\n /* Error if the 2nd code is not a low surrogate */\n if ((surrogate_low & 0xFC00) != 0xDC00)\n return -1;\n\n /* Calculate Unicode codepoint */\n codepoint = (codepoint & 0x3FF) << 10;\n surrogate_low &= 0x3FF;\n codepoint = (codepoint | surrogate_low) + 0x10000;\n escape_len = 12;\n }\n\n /* Convert codepoint to UTF-8 */\n len = codepoint_to_utf8(utf8, codepoint);\n if (!len)\n return -1;\n\n /* Append bytes and advance parse index */\n strbuf_append_mem_unsafe(json->tmp, utf8, len);\n json->index += escape_len;\n\n return 0;\n}\n\nstatic void json_set_token_error(json_token_t *token, json_parse_t *json,\n const char *errtype)\n{\n token->type = T_ERROR;\n token->index = json->index;\n token->value.string = errtype;\n}\n\nstatic void json_next_string_token(json_parse_t *json, json_token_t *token)\n{\n char *escape2char = json->cfg->escape2char;\n char ch;\n\n /* Caller must ensure a string is next */\n assert(json->data[json->index] == '\"');\n\n /* Skip \" */\n json->index++;\n\n /* json->tmp is the temporary strbuf used to accumulate the\n * decoded string value. */\n strbuf_reset(json->tmp);\n while ((ch = json->data[json->index]) != '\"') {\n if (!ch) {\n /* Premature end of the string */\n json_set_token_error(token, json, \"unexpected end of string\");\n return;\n }\n\n /* Handle escapes */\n if (ch == '\\\\') {\n /* Fetch escape character */\n ch = json->data[json->index + 1];\n\n /* Translate escape code and append to tmp string */\n ch = escape2char[(unsigned char)ch];\n if (ch == 'u') {\n if (json_append_unicode_escape(json) == 0)\n continue;\n\n json_set_token_error(token, json,\n \"invalid unicode escape code\");\n return;\n }\n if (!ch) {\n json_set_token_error(token, json, \"invalid escape code\");\n return;\n }\n\n /* Skip '\\' */\n json->index++;\n }\n /* Append normal character or translated single character\n * Unicode escapes are handled above */\n strbuf_append_char_unsafe(json->tmp, ch);\n json->index++;\n }\n json->index++; /* Eat final quote (\") */\n\n strbuf_ensure_null(json->tmp);\n\n token->type = T_STRING;\n token->value.string = strbuf_string(json->tmp, &token->string_len);\n}\n\n/* JSON numbers should take the following form:\n * -?(0|[1-9]|[1-9][0-9]+)(.[0-9]+)?([eE][-+]?[0-9]+)?\n *\n * json_next_number_token() uses strtod() which allows other forms:\n * - numbers starting with '+'\n * - NaN, -NaN, infinity, -infinity\n * - hexidecimal numbers\n * - numbers with leading zeros\n *\n * json_is_invalid_number() detects \"numbers\" which may pass strtod()'s\n * error checking, but should not be allowed with strict JSON.\n *\n * json_is_invalid_number() may pass numbers which cause strtod()\n * to generate an error.\n */\nstatic int json_is_invalid_number(json_parse_t *json)\n{\n int i = json->index;\n\n /* Reject numbers starting with + */\n if (json->data[i] == '+')\n return 1;\n\n /* Skip minus sign if it exists */\n if (json->data[i] == '-')\n i++;\n\n /* Reject numbers starting with 0x, or leading zeros */\n if (json->data[i] == '0') {\n int ch2 = json->data[i + 1];\n\n if ((ch2 | 0x20) == 'x' || /* Hex */\n ('0' <= ch2 && ch2 <= '9')) /* Leading zero */\n return 1;\n\n return 0;\n } else if (json->data[i] <= '9') {\n return 0; /* Ordinary number */\n }\n\n\n /* Reject inf/nan */\n if (!strncasecmp(&json->data[i], \"inf\", 3))\n return 1;\n if (!strncasecmp(&json->data[i], \"nan\", 3))\n return 1;\n\n /* Pass all other numbers which may still be invalid, but\n * strtod() will catch them. */\n return 0;\n}\n\nstatic void json_next_number_token(json_parse_t *json, json_token_t *token)\n{\n const char *startptr;\n char *endptr;\n\n token->type = T_NUMBER;\n startptr = &json->data[json->index];\n token->value.number = strtod(&json->data[json->index], &endptr);\n if (startptr == endptr)\n json_set_token_error(token, json, \"invalid number\");\n else\n json->index += endptr - startptr; /* Skip the processed number */\n\n return;\n}\n\n/* Fills in the token struct.\n * T_STRING will return a pointer to the json_parse_t temporary string\n * T_ERROR will leave the json->index pointer at the error.\n */\nstatic void json_next_token(json_parse_t *json, json_token_t *token)\n{\n json_token_type_t *ch2token = json->cfg->ch2token;\n int ch;\n\n /* Eat whitespace. FIXME: UGLY */\n token->type = ch2token[(unsigned char)json->data[json->index]];\n while (token->type == T_WHITESPACE)\n token->type = ch2token[(unsigned char)json->data[++json->index]];\n\n token->index = json->index;\n\n /* Don't advance the pointer for an error or the end */\n if (token->type == T_ERROR) {\n json_set_token_error(token, json, \"invalid token\");\n return;\n }\n\n if (token->type == T_END) {\n return;\n }\n\n /* Found a known single character token, advance index and return */\n if (token->type != T_UNKNOWN) {\n json->index++;\n return;\n }\n\n /* Process characters which triggered T_UNKNOWN */\n ch = json->data[json->index];\n\n /* Must use strncmp() to match the front of the JSON string.\n * JSON identifier must be lowercase.\n * When strict_numbers if disabled, either case is allowed for\n * Infinity/NaN (since we are no longer following the spec..) */\n if (ch == '\"') {\n json_next_string_token(json, token);\n return;\n } else if (ch == '-' || ('0' <= ch && ch <= '9')) {\n if (json->cfg->decode_refuse_badnum && json_is_invalid_number(json)) {\n json_set_token_error(token, json, \"invalid number\");\n return;\n }\n json_next_number_token(json, token);\n return;\n } else if (!strncmp(&json->data[json->index], \"true\", 4)) {\n token->type = T_BOOLEAN;\n token->value.boolean = 1;\n json->index += 4;\n return;\n } else if (!strncmp(&json->data[json->index], \"false\", 5)) {\n token->type = T_BOOLEAN;\n token->value.boolean = 0;\n json->index += 5;\n return;\n } else if (!strncmp(&json->data[json->index], \"null\", 4)) {\n token->type = T_NULL;\n json->index += 4;\n return;\n } else if (!json->cfg->decode_refuse_badnum &&\n json_is_invalid_number(json)) {\n /* When refuse_badnum is disabled, only attempt to process\n * numbers we know are invalid JSON (Inf, NaN, hex)\n * This is required to generate an appropriate token error,\n * otherwise all bad tokens will register as \"invalid number\"\n */\n json_next_number_token(json, token);\n return;\n }\n\n /* Token starts with t/f/n but isn't recognised above. */\n json_set_token_error(token, json, \"invalid token\");\n}\n\n/* This function does not return.\n * DO NOT CALL WITH DYNAMIC MEMORY ALLOCATED.\n * The only supported exception is the temporary parser string\n * json->tmp struct.\n * json and token should exist on the stack somewhere.\n * luaL_error() will long_jmp and release the stack */\nstatic void json_throw_parse_error(lua_State *l, json_parse_t *json,\n const char *exp, json_token_t *token)\n{\n const char *found;\n\n strbuf_free(json->tmp);\n\n if (token->type == T_ERROR)\n found = token->value.string;\n else\n found = json_token_type_name[token->type];\n\n /* Note: token->index is 0 based, display starting from 1 */\n luaL_error(l, \"Expected %s but found %s at character %d\",\n exp, found, token->index + 1);\n}\n\nstatic void json_decode_checkstack(lua_State *l, json_parse_t *json, int n)\n{\n if (lua_checkstack(l, n))\n return;\n\n strbuf_free(json->tmp);\n luaL_error(l, \"Too many nested data structures\");\n}\n\nstatic void json_parse_object_context(lua_State *l, json_parse_t *json)\n{\n json_token_t token;\n\n /* 3 slots required:\n * .., table, key, value */\n json_decode_checkstack(l, json, 3);\n\n lua_newtable(l);\n\n json_next_token(json, &token);\n\n /* Handle empty objects */\n if (token.type == T_OBJ_END) {\n return;\n }\n\n while (1) {\n if (token.type != T_STRING)\n json_throw_parse_error(l, json, \"object key string\", &token);\n\n /* Push key */\n lua_pushlstring(l, token.value.string, token.string_len);\n\n json_next_token(json, &token);\n if (token.type != T_COLON)\n json_throw_parse_error(l, json, \"colon\", &token);\n\n /* Fetch value */\n json_next_token(json, &token);\n json_process_value(l, json, &token);\n\n /* Set key = value */\n lua_rawset(l, -3);\n\n json_next_token(json, &token);\n\n if (token.type == T_OBJ_END)\n return;\n\n if (token.type != T_COMMA)\n json_throw_parse_error(l, json, \"comma or object end\", &token);\n\n json_next_token(json, &token);\n }\n}\n\n/* Handle the array context */\nstatic void json_parse_array_context(lua_State *l, json_parse_t *json)\n{\n json_token_t token;\n int i;\n\n /* 2 slots required:\n * .., table, value */\n json_decode_checkstack(l, json, 2);\n\n lua_newtable(l);\n\n json_next_token(json, &token);\n\n /* Handle empty arrays */\n if (token.type == T_ARR_END)\n return;\n\n for (i = 1; ; i++) {\n json_process_value(l, json, &token);\n lua_rawseti(l, -2, i); /* arr[i] = value */\n\n json_next_token(json, &token);\n\n if (token.type == T_ARR_END)\n return;\n\n if (token.type != T_COMMA)\n json_throw_parse_error(l, json, \"comma or array end\", &token);\n\n json_next_token(json, &token);\n }\n}\n\n/* Handle the \"value\" context */\nstatic void json_process_value(lua_State *l, json_parse_t *json,\n json_token_t *token)\n{\n switch (token->type) {\n case T_STRING:\n lua_pushlstring(l, token->value.string, token->string_len);\n break;;\n case T_NUMBER:\n lua_pushnumber(l, token->value.number);\n break;;\n case T_BOOLEAN:\n lua_pushboolean(l, token->value.boolean);\n break;;\n case T_OBJ_BEGIN:\n json_parse_object_context(l, json);\n break;;\n case T_ARR_BEGIN:\n json_parse_array_context(l, json);\n break;;\n case T_NULL:\n /* In Lua, setting \"t[k] = nil\" will delete k from the table.\n * Hence a NULL pointer lightuserdata object is used instead */\n lua_pushlightuserdata(l, NULL);\n break;;\n default:\n json_throw_parse_error(l, json, \"value\", token);\n }\n}\n\n/* json_text must be null terminated string */\nstatic void lua_json_decode(lua_State *l, const char *json_text, int json_len)\n{\n json_parse_t json;\n json_token_t token;\n\n json.cfg = json_fetch_config(l);\n json.data = json_text;\n json.index = 0;\n\n /* Ensure the temporary buffer can hold the entire string.\n * This means we no longer need to do length checks since the decoded\n * string must be smaller than the entire json string */\n json.tmp = strbuf_new(json_len);\n\n json_next_token(&json, &token);\n json_process_value(l, &json, &token);\n\n /* Ensure there is no more input left */\n json_next_token(&json, &token);\n\n if (token.type != T_END)\n json_throw_parse_error(l, &json, \"the end\", &token);\n\n strbuf_free(json.tmp);\n}\n\nstatic int json_decode(lua_State *l)\n{\n const char *json;\n size_t len;\n\n json_verify_arg_count(l, 1);\n\n json = luaL_checklstring(l, 1, &len);\n\n /* Detect Unicode other than UTF-8 (see RFC 4627, Sec 3)\n *\n * CJSON can support any simple data type, hence only the first\n * character is guaranteed to be ASCII (at worst: '\"'). This is\n * still enough to detect whether the wrong encoding is in use. */\n if (len >= 2 && (!json[0] || !json[1]))\n luaL_error(l, \"JSON parser does not support UTF-16 or UTF-32\");\n\n lua_json_decode(l, json, len);\n\n return 1;\n}\n\n/* ===== INITIALISATION ===== */\n\nint luaopen_cjson(lua_State *l)\n{\n luaL_Reg reg[] = {\n { \"encode\", json_encode },\n { \"decode\", json_decode },\n { \"encode_sparse_array\", json_cfg_encode_sparse_array },\n { \"encode_max_depth\", json_cfg_encode_max_depth },\n { \"encode_number_precision\", json_cfg_encode_number_precision },\n { \"encode_keep_buffer\", json_cfg_encode_keep_buffer },\n { \"refuse_invalid_numbers\", json_cfg_refuse_invalid_numbers },\n { NULL, NULL }\n };\n\n /* Use json_fetch_config as a pointer.\n * It's faster than using a config string, and more unique */\n lua_pushlightuserdata(l, &json_config_key);\n json_create_config(l);\n lua_settable(l, LUA_REGISTRYINDEX);\n\n luaL_register(l, \"cjson\", reg);\n\n /* Set cjson.null */\n lua_pushlightuserdata(l, NULL);\n lua_setfield(l, -2, \"null\");\n\n /* Set cjson.version */\n lua_pushliteral(l, VERSION);\n lua_setfield(l, -2, \"version\");\n\n /* Return cjson table */\n return 1;\n}\n\n/* vi:ai et sw=4 ts=4:\n */\n" }, { "path": "deps/lua/src/lua_cmsgpack.c", "content": "#include \n#include \n#include \n#include \n#include \n\n#include \"lua.h\"\n#include \"lauxlib.h\"\n\n#define LUACMSGPACK_VERSION \"lua-cmsgpack 0.3.0\"\n#define LUACMSGPACK_COPYRIGHT \"Copyright (C) 2012, Salvatore Sanfilippo\"\n#define LUACMSGPACK_DESCRIPTION \"MessagePack C implementation for Lua\"\n\n#define LUACMSGPACK_MAX_NESTING 16 /* Max tables nesting. */\n\n/* ==============================================================================\n * MessagePack implementation and bindings for Lua 5.1.\n * Copyright(C) 2012 Salvatore Sanfilippo \n *\n * http://github.com/antirez/lua-cmsgpack\n *\n * For MessagePack specification check the following web site:\n * http://wiki.msgpack.org/display/MSGPACK/Format+specification\n *\n * See Copyright Notice at the end of this file.\n *\n * CHANGELOG:\n * 19-Feb-2012 (ver 0.1.0): Initial release.\n * 20-Feb-2012 (ver 0.2.0): Tables encoding improved.\n * 20-Feb-2012 (ver 0.2.1): Minor bug fixing.\n * 20-Feb-2012 (ver 0.3.0): Module renamed lua-cmsgpack (was lua-msgpack).\n * ============================================================================ */\n\n/* --------------------------- Endian conversion --------------------------------\n * We use it only for floats and doubles, all the other conversions are performed\n * in an endian independent fashion. So the only thing we need is a function\n * that swaps a binary string if the arch is little endian (and left it untouched\n * otherwise). */\n\n/* Reverse memory bytes if arch is little endian. Given the conceptual\n * simplicity of the Lua build system we prefer to check for endianess at runtime.\n * The performance difference should be acceptable. */\nstatic void memrevifle(void *ptr, size_t len) {\n unsigned char *p = ptr, *e = p+len-1, aux;\n int test = 1;\n unsigned char *testp = (unsigned char*) &test;\n\n if (testp[0] == 0) return; /* Big endian, nothign to do. */\n len /= 2;\n while(len--) {\n aux = *p;\n *p = *e;\n *e = aux;\n p++;\n e--;\n }\n}\n\n/* ----------------------------- String buffer ----------------------------------\n * This is a simple implementation of string buffers. The only opereation\n * supported is creating empty buffers and appending bytes to it.\n * The string buffer uses 2x preallocation on every realloc for O(N) append\n * behavior. */\n\ntypedef struct mp_buf {\n unsigned char *b;\n size_t len, free;\n} mp_buf;\n\nstatic mp_buf *mp_buf_new(void) {\n mp_buf *buf = malloc(sizeof(*buf));\n \n buf->b = NULL;\n buf->len = buf->free = 0;\n return buf;\n}\n\nvoid mp_buf_append(mp_buf *buf, const unsigned char *s, size_t len) {\n if (buf->free < len) {\n size_t newlen = buf->len+len;\n\n buf->b = realloc(buf->b,newlen*2);\n buf->free = newlen;\n }\n memcpy(buf->b+buf->len,s,len);\n buf->len += len;\n buf->free -= len;\n}\n\nvoid mp_buf_free(mp_buf *buf) {\n free(buf->b);\n free(buf);\n}\n\n/* ------------------------------ String cursor ----------------------------------\n * This simple data structure is used for parsing. Basically you create a cursor\n * using a string pointer and a length, then it is possible to access the\n * current string position with cursor->p, check the remaining length\n * in cursor->left, and finally consume more string using\n * mp_cur_consume(cursor,len), to advance 'p' and subtract 'left'.\n * An additional field cursor->error is set to zero on initialization and can\n * be used to report errors. */\n\n#define MP_CUR_ERROR_NONE 0\n#define MP_CUR_ERROR_EOF 1 /* Not enough data to complete the opereation. */\n#define MP_CUR_ERROR_BADFMT 2 /* Bad data format */\n\ntypedef struct mp_cur {\n const unsigned char *p;\n size_t left;\n int err;\n} mp_cur;\n\nstatic mp_cur *mp_cur_new(const unsigned char *s, size_t len) {\n mp_cur *cursor = malloc(sizeof(*cursor));\n\n cursor->p = s;\n cursor->left = len;\n cursor->err = MP_CUR_ERROR_NONE;\n return cursor;\n}\n\nstatic void mp_cur_free(mp_cur *cursor) {\n free(cursor);\n}\n\n#define mp_cur_consume(_c,_len) do { _c->p += _len; _c->left -= _len; } while(0)\n\n/* When there is not enough room we set an error in the cursor and return, this\n * is very common across the code so we have a macro to make the code look\n * a bit simpler. */\n#define mp_cur_need(_c,_len) do { \\\n if (_c->left < _len) { \\\n _c->err = MP_CUR_ERROR_EOF; \\\n return; \\\n } \\\n} while(0)\n\n/* --------------------------- Low level MP encoding -------------------------- */\n\nstatic void mp_encode_bytes(mp_buf *buf, const unsigned char *s, size_t len) {\n unsigned char hdr[5];\n int hdrlen;\n\n if (len < 32) {\n hdr[0] = 0xa0 | (len&0xff); /* fix raw */\n hdrlen = 1;\n } else if (len <= 0xffff) {\n hdr[0] = 0xda;\n hdr[1] = (len&0xff00)>>8;\n hdr[2] = len&0xff;\n hdrlen = 3;\n } else {\n hdr[0] = 0xdb;\n hdr[1] = (len&0xff000000)>>24;\n hdr[2] = (len&0xff0000)>>16;\n hdr[3] = (len&0xff00)>>8;\n hdr[4] = len&0xff;\n hdrlen = 5;\n }\n mp_buf_append(buf,hdr,hdrlen);\n mp_buf_append(buf,s,len);\n}\n\n/* we assume IEEE 754 internal format for single and double precision floats. */\nstatic void mp_encode_double(mp_buf *buf, double d) {\n unsigned char b[9];\n float f = d;\n\n assert(sizeof(f) == 4 && sizeof(d) == 8);\n if (d == (double)f) {\n b[0] = 0xca; /* float IEEE 754 */\n memcpy(b+1,&f,4);\n memrevifle(b+1,4);\n mp_buf_append(buf,b,5);\n } else if (sizeof(d) == 8) {\n b[0] = 0xcb; /* double IEEE 754 */\n memcpy(b+1,&d,8);\n memrevifle(b+1,8);\n mp_buf_append(buf,b,9);\n }\n}\n\nstatic void mp_encode_int(mp_buf *buf, int64_t n) {\n unsigned char b[9];\n int enclen;\n\n if (n >= 0) {\n if (n <= 127) {\n b[0] = n & 0x7f; /* positive fixnum */\n enclen = 1;\n } else if (n <= 0xff) {\n b[0] = 0xcc; /* uint 8 */\n b[1] = n & 0xff;\n enclen = 2;\n } else if (n <= 0xffff) {\n b[0] = 0xcd; /* uint 16 */\n b[1] = (n & 0xff00) >> 8;\n b[2] = n & 0xff;\n enclen = 3;\n } else if (n <= 0xffffffffLL) {\n b[0] = 0xce; /* uint 32 */\n b[1] = (n & 0xff000000) >> 24;\n b[2] = (n & 0xff0000) >> 16;\n b[3] = (n & 0xff00) >> 8;\n b[4] = n & 0xff;\n enclen = 5;\n } else {\n b[0] = 0xcf; /* uint 64 */\n b[1] = (n & 0xff00000000000000LL) >> 56;\n b[2] = (n & 0xff000000000000LL) >> 48;\n b[3] = (n & 0xff0000000000LL) >> 40;\n b[4] = (n & 0xff00000000LL) >> 32;\n b[5] = (n & 0xff000000) >> 24;\n b[6] = (n & 0xff0000) >> 16;\n b[7] = (n & 0xff00) >> 8;\n b[8] = n & 0xff;\n enclen = 9;\n }\n } else {\n if (n >= -32) {\n b[0] = ((char)n); /* negative fixnum */\n enclen = 1;\n } else if (n >= -128) {\n b[0] = 0xd0; /* int 8 */\n b[1] = n & 0xff;\n enclen = 2;\n } else if (n >= -32768) {\n b[0] = 0xd1; /* int 16 */\n b[1] = (n & 0xff00) >> 8;\n b[2] = n & 0xff;\n enclen = 3;\n } else if (n >= -2147483648LL) {\n b[0] = 0xd2; /* int 32 */\n b[1] = (n & 0xff000000) >> 24;\n b[2] = (n & 0xff0000) >> 16;\n b[3] = (n & 0xff00) >> 8;\n b[4] = n & 0xff;\n enclen = 5;\n } else {\n b[0] = 0xd3; /* int 64 */\n b[1] = (n & 0xff00000000000000LL) >> 56;\n b[2] = (n & 0xff000000000000LL) >> 48;\n b[3] = (n & 0xff0000000000LL) >> 40;\n b[4] = (n & 0xff00000000LL) >> 32;\n b[5] = (n & 0xff000000) >> 24;\n b[6] = (n & 0xff0000) >> 16;\n b[7] = (n & 0xff00) >> 8;\n b[8] = n & 0xff;\n enclen = 9;\n }\n }\n mp_buf_append(buf,b,enclen);\n}\n\nstatic void mp_encode_array(mp_buf *buf, int64_t n) {\n unsigned char b[5];\n int enclen;\n\n if (n <= 15) {\n b[0] = 0x90 | (n & 0xf); /* fix array */\n enclen = 1;\n } else if (n <= 65535) {\n b[0] = 0xdc; /* array 16 */\n b[1] = (n & 0xff00) >> 8;\n b[2] = n & 0xff;\n enclen = 3;\n } else {\n b[0] = 0xdd; /* array 32 */\n b[1] = (n & 0xff000000) >> 24;\n b[2] = (n & 0xff0000) >> 16;\n b[3] = (n & 0xff00) >> 8;\n b[4] = n & 0xff;\n enclen = 5;\n }\n mp_buf_append(buf,b,enclen);\n}\n\nstatic void mp_encode_map(mp_buf *buf, int64_t n) {\n unsigned char b[5];\n int enclen;\n\n if (n <= 15) {\n b[0] = 0x80 | (n & 0xf); /* fix map */\n enclen = 1;\n } else if (n <= 65535) {\n b[0] = 0xde; /* map 16 */\n b[1] = (n & 0xff00) >> 8;\n b[2] = n & 0xff;\n enclen = 3;\n } else {\n b[0] = 0xdf; /* map 32 */\n b[1] = (n & 0xff000000) >> 24;\n b[2] = (n & 0xff0000) >> 16;\n b[3] = (n & 0xff00) >> 8;\n b[4] = n & 0xff;\n enclen = 5;\n }\n mp_buf_append(buf,b,enclen);\n}\n\n/* ----------------------------- Lua types encoding --------------------------- */\n\nstatic void mp_encode_lua_string(lua_State *L, mp_buf *buf) {\n size_t len;\n const char *s;\n\n s = lua_tolstring(L,-1,&len);\n mp_encode_bytes(buf,(const unsigned char*)s,len);\n}\n\nstatic void mp_encode_lua_bool(lua_State *L, mp_buf *buf) {\n unsigned char b = lua_toboolean(L,-1) ? 0xc3 : 0xc2;\n mp_buf_append(buf,&b,1);\n}\n\nstatic void mp_encode_lua_number(lua_State *L, mp_buf *buf) {\n lua_Number n = lua_tonumber(L,-1);\n\n if (floor(n) != n) {\n mp_encode_double(buf,(double)n);\n } else {\n mp_encode_int(buf,(int64_t)n);\n }\n}\n\nstatic void mp_encode_lua_type(lua_State *L, mp_buf *buf, int level);\n\n/* Convert a lua table into a message pack list. */\nstatic void mp_encode_lua_table_as_array(lua_State *L, mp_buf *buf, int level) {\n size_t len = lua_objlen(L,-1), j;\n\n mp_encode_array(buf,len);\n for (j = 1; j <= len; j++) {\n lua_pushnumber(L,j);\n lua_gettable(L,-2);\n mp_encode_lua_type(L,buf,level+1);\n }\n}\n\n/* Convert a lua table into a message pack key-value map. */\nstatic void mp_encode_lua_table_as_map(lua_State *L, mp_buf *buf, int level) {\n size_t len = 0;\n\n /* First step: count keys into table. No other way to do it with the\n * Lua API, we need to iterate a first time. Note that an alternative\n * would be to do a single run, and then hack the buffer to insert the\n * map opcodes for message pack. Too hachish for this lib. */\n lua_pushnil(L);\n while(lua_next(L,-2)) {\n lua_pop(L,1); /* remove value, keep key for next iteration. */\n len++;\n }\n\n /* Step two: actually encoding of the map. */\n mp_encode_map(buf,len);\n lua_pushnil(L);\n while(lua_next(L,-2)) {\n /* Stack: ... key value */\n lua_pushvalue(L,-2); /* Stack: ... key value key */\n mp_encode_lua_type(L,buf,level+1); /* encode key */\n mp_encode_lua_type(L,buf,level+1); /* encode val */\n }\n}\n\n/* Returns true if the Lua table on top of the stack is exclusively composed\n * of keys from numerical keys from 1 up to N, with N being the total number\n * of elements, without any hole in the middle. */\nstatic int table_is_an_array(lua_State *L) {\n long count = 0, max = 0, idx = 0;\n lua_Number n;\n\n lua_pushnil(L);\n while(lua_next(L,-2)) {\n /* Stack: ... key value */\n lua_pop(L,1); /* Stack: ... key */\n if (lua_type(L,-1) != LUA_TNUMBER) goto not_array;\n n = lua_tonumber(L,-1);\n idx = n;\n if (idx != n || idx < 1) goto not_array;\n count++;\n max = idx;\n }\n /* We have the total number of elements in \"count\". Also we have\n * the max index encountered in \"idx\". We can't reach this code\n * if there are indexes <= 0. If you also note that there can not be\n * repeated keys into a table, you have that if idx==count you are sure\n * that there are all the keys form 1 to count (both included). */\n return idx == count;\n\nnot_array:\n lua_pop(L,1);\n return 0;\n}\n\n/* If the length operator returns non-zero, that is, there is at least\n * an object at key '1', we serialize to message pack list. Otherwise\n * we use a map. */\nstatic void mp_encode_lua_table(lua_State *L, mp_buf *buf, int level) {\n if (table_is_an_array(L))\n mp_encode_lua_table_as_array(L,buf,level);\n else\n mp_encode_lua_table_as_map(L,buf,level);\n}\n\nstatic void mp_encode_lua_null(lua_State *L, mp_buf *buf) {\n unsigned char b[1];\n\n b[0] = 0xc0;\n mp_buf_append(buf,b,1);\n}\n\nstatic void mp_encode_lua_type(lua_State *L, mp_buf *buf, int level) {\n int t = lua_type(L,-1);\n\n /* Limit the encoding of nested tables to a specfiied maximum depth, so that\n * we survive when called against circular references in tables. */\n if (t == LUA_TTABLE && level == LUACMSGPACK_MAX_NESTING) t = LUA_TNIL;\n switch(t) {\n case LUA_TSTRING: mp_encode_lua_string(L,buf); break;\n case LUA_TBOOLEAN: mp_encode_lua_bool(L,buf); break;\n case LUA_TNUMBER: mp_encode_lua_number(L,buf); break;\n case LUA_TTABLE: mp_encode_lua_table(L,buf,level); break;\n default: mp_encode_lua_null(L,buf); break;\n }\n lua_pop(L,1);\n}\n\nstatic int mp_pack(lua_State *L) {\n mp_buf *buf = mp_buf_new();\n\n mp_encode_lua_type(L,buf,0);\n lua_pushlstring(L,(char*)buf->b,buf->len);\n mp_buf_free(buf);\n return 1;\n}\n\n/* --------------------------------- Decoding --------------------------------- */\n\nvoid mp_decode_to_lua_type(lua_State *L, mp_cur *c);\n\nvoid mp_decode_to_lua_array(lua_State *L, mp_cur *c, size_t len) {\n int index = 1;\n\n lua_newtable(L);\n while(len--) {\n lua_pushnumber(L,index++);\n mp_decode_to_lua_type(L,c);\n if (c->err) return;\n lua_settable(L,-3);\n }\n}\n\nvoid mp_decode_to_lua_hash(lua_State *L, mp_cur *c, size_t len) {\n lua_newtable(L);\n while(len--) {\n mp_decode_to_lua_type(L,c); /* key */\n if (c->err) return;\n mp_decode_to_lua_type(L,c); /* value */\n if (c->err) return;\n lua_settable(L,-3);\n }\n}\n\n/* Decode a Message Pack raw object pointed by the string cursor 'c' to\n * a Lua type, that is left as the only result on the stack. */\nvoid mp_decode_to_lua_type(lua_State *L, mp_cur *c) {\n mp_cur_need(c,1);\n switch(c->p[0]) {\n case 0xcc: /* uint 8 */\n mp_cur_need(c,2);\n lua_pushnumber(L,c->p[1]);\n mp_cur_consume(c,2);\n break;\n case 0xd0: /* int 8 */\n mp_cur_need(c,2);\n lua_pushnumber(L,(char)c->p[1]);\n mp_cur_consume(c,2);\n break;\n case 0xcd: /* uint 16 */\n mp_cur_need(c,3);\n lua_pushnumber(L,\n (c->p[1] << 8) |\n c->p[2]);\n mp_cur_consume(c,3);\n break;\n case 0xd1: /* int 16 */\n mp_cur_need(c,3);\n lua_pushnumber(L,(int16_t)\n (c->p[1] << 8) |\n c->p[2]);\n mp_cur_consume(c,3);\n break;\n case 0xce: /* uint 32 */\n mp_cur_need(c,5);\n lua_pushnumber(L,\n ((uint32_t)c->p[1] << 24) |\n ((uint32_t)c->p[2] << 16) |\n ((uint32_t)c->p[3] << 8) |\n (uint32_t)c->p[4]);\n mp_cur_consume(c,5);\n break;\n case 0xd2: /* int 32 */\n mp_cur_need(c,5);\n lua_pushnumber(L,\n ((int32_t)c->p[1] << 24) |\n ((int32_t)c->p[2] << 16) |\n ((int32_t)c->p[3] << 8) |\n (int32_t)c->p[4]);\n mp_cur_consume(c,5);\n break;\n case 0xcf: /* uint 64 */\n mp_cur_need(c,9);\n lua_pushnumber(L,\n ((uint64_t)c->p[1] << 56) |\n ((uint64_t)c->p[2] << 48) |\n ((uint64_t)c->p[3] << 40) |\n ((uint64_t)c->p[4] << 32) |\n ((uint64_t)c->p[5] << 24) |\n ((uint64_t)c->p[6] << 16) |\n ((uint64_t)c->p[7] << 8) |\n (uint64_t)c->p[8]);\n mp_cur_consume(c,9);\n break;\n case 0xd3: /* int 64 */\n mp_cur_need(c,9);\n lua_pushnumber(L,\n ((int64_t)c->p[1] << 56) |\n ((int64_t)c->p[2] << 48) |\n ((int64_t)c->p[3] << 40) |\n ((int64_t)c->p[4] << 32) |\n ((int64_t)c->p[5] << 24) |\n ((int64_t)c->p[6] << 16) |\n ((int64_t)c->p[7] << 8) |\n (int64_t)c->p[8]);\n mp_cur_consume(c,9);\n break;\n case 0xc0: /* nil */\n lua_pushnil(L);\n mp_cur_consume(c,1);\n break;\n case 0xc3: /* true */\n lua_pushboolean(L,1);\n mp_cur_consume(c,1);\n break;\n case 0xc2: /* false */\n lua_pushboolean(L,0);\n mp_cur_consume(c,1);\n break;\n case 0xca: /* float */\n mp_cur_need(c,5);\n assert(sizeof(float) == 4);\n {\n float f;\n memcpy(&f,c->p+1,4);\n memrevifle(&f,4);\n lua_pushnumber(L,f);\n mp_cur_consume(c,5);\n }\n break;\n case 0xcb: /* double */\n mp_cur_need(c,9);\n assert(sizeof(double) == 8);\n {\n double d;\n memcpy(&d,c->p+1,8);\n memrevifle(&d,8);\n lua_pushnumber(L,d);\n mp_cur_consume(c,9);\n }\n break;\n case 0xda: /* raw 16 */\n mp_cur_need(c,3);\n {\n size_t l = (c->p[1] << 8) | c->p[2];\n mp_cur_need(c,3+l);\n lua_pushlstring(L,(char*)c->p+3,l);\n mp_cur_consume(c,3+l);\n }\n break;\n case 0xdb: /* raw 32 */\n mp_cur_need(c,5);\n {\n size_t l = (c->p[1] << 24) |\n (c->p[2] << 16) |\n (c->p[3] << 8) |\n c->p[4];\n mp_cur_need(c,5+l);\n lua_pushlstring(L,(char*)c->p+5,l);\n mp_cur_consume(c,5+l);\n }\n break;\n case 0xdc: /* array 16 */\n mp_cur_need(c,3);\n {\n size_t l = (c->p[1] << 8) | c->p[2];\n mp_cur_consume(c,3);\n mp_decode_to_lua_array(L,c,l);\n }\n break;\n case 0xdd: /* array 32 */\n mp_cur_need(c,5);\n {\n size_t l = (c->p[1] << 24) |\n (c->p[2] << 16) |\n (c->p[3] << 8) |\n c->p[4];\n mp_cur_consume(c,5);\n mp_decode_to_lua_array(L,c,l);\n }\n break;\n case 0xde: /* map 16 */\n mp_cur_need(c,3);\n {\n size_t l = (c->p[1] << 8) | c->p[2];\n mp_cur_consume(c,3);\n mp_decode_to_lua_hash(L,c,l);\n }\n break;\n case 0xdf: /* map 32 */\n mp_cur_need(c,5);\n {\n size_t l = (c->p[1] << 24) |\n (c->p[2] << 16) |\n (c->p[3] << 8) |\n c->p[4];\n mp_cur_consume(c,5);\n mp_decode_to_lua_hash(L,c,l);\n }\n break;\n default: /* types that can't be idenitified by first byte value. */\n if ((c->p[0] & 0x80) == 0) { /* positive fixnum */\n lua_pushnumber(L,c->p[0]);\n mp_cur_consume(c,1);\n } else if ((c->p[0] & 0xe0) == 0xe0) { /* negative fixnum */\n lua_pushnumber(L,(signed char)c->p[0]);\n mp_cur_consume(c,1);\n } else if ((c->p[0] & 0xe0) == 0xa0) { /* fix raw */\n size_t l = c->p[0] & 0x1f;\n mp_cur_need(c,1+l);\n lua_pushlstring(L,(char*)c->p+1,l);\n mp_cur_consume(c,1+l);\n } else if ((c->p[0] & 0xf0) == 0x90) { /* fix map */\n size_t l = c->p[0] & 0xf;\n mp_cur_consume(c,1);\n mp_decode_to_lua_array(L,c,l);\n } else if ((c->p[0] & 0xf0) == 0x80) { /* fix map */\n size_t l = c->p[0] & 0xf;\n mp_cur_consume(c,1);\n mp_decode_to_lua_hash(L,c,l);\n } else {\n c->err = MP_CUR_ERROR_BADFMT;\n }\n }\n}\n\nstatic int mp_unpack(lua_State *L) {\n size_t len;\n const unsigned char *s;\n mp_cur *c;\n\n if (!lua_isstring(L,-1)) {\n lua_pushstring(L,\"MessagePack decoding needs a string as input.\");\n lua_error(L);\n }\n\n s = (const unsigned char*) lua_tolstring(L,-1,&len);\n c = mp_cur_new(s,len);\n mp_decode_to_lua_type(L,c);\n \n if (c->err == MP_CUR_ERROR_EOF) {\n mp_cur_free(c);\n lua_pushstring(L,\"Missing bytes in input.\");\n lua_error(L);\n } else if (c->err == MP_CUR_ERROR_BADFMT) {\n mp_cur_free(c);\n lua_pushstring(L,\"Bad data format in input.\");\n lua_error(L);\n } else if (c->left != 0) {\n mp_cur_free(c);\n lua_pushstring(L,\"Extra bytes in input.\");\n lua_error(L);\n }\n mp_cur_free(c);\n return 1;\n}\n\n/* ---------------------------------------------------------------------------- */\n\nstatic const struct luaL_reg thislib[] = {\n {\"pack\", mp_pack},\n {\"unpack\", mp_unpack},\n {NULL, NULL}\n};\n\nLUALIB_API int luaopen_cmsgpack (lua_State *L) {\n luaL_register(L, \"cmsgpack\", thislib);\n\n lua_pushliteral(L, LUACMSGPACK_VERSION);\n lua_setfield(L, -2, \"_VERSION\");\n lua_pushliteral(L, LUACMSGPACK_COPYRIGHT);\n lua_setfield(L, -2, \"_COPYRIGHT\");\n lua_pushliteral(L, LUACMSGPACK_DESCRIPTION);\n lua_setfield(L, -2, \"_DESCRIPTION\"); \n return 1;\n}\n\n/******************************************************************************\n* Copyright (C) 2012 Salvatore Sanfilippo. All rights reserved.\n*\n* Permission is hereby granted, free of charge, to any person obtaining\n* a copy of this software and associated documentation files (the\n* \"Software\"), to deal in the Software without restriction, including\n* without limitation the rights to use, copy, modify, merge, publish,\n* distribute, sublicense, and/or sell copies of the Software, and to\n* permit persons to whom the Software is furnished to do so, subject to\n* the following conditions:\n*\n* The above copyright notice and this permission notice shall be\n* included in all copies or substantial portions of the Software.\n*\n* THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n******************************************************************************/\n" }, { "path": "deps/lua/src/lua_struct.c", "content": "/*\n** {======================================================\n** Library for packing/unpacking structures.\n** $Id: struct.c,v 1.4 2012/07/04 18:54:29 roberto Exp $\n** See Copyright Notice at the end of this file\n** =======================================================\n*/\n/*\n** Valid formats:\n** > - big endian\n** < - little endian\n** ![num] - alignment\n** x - pading\n** b/B - signed/unsigned byte\n** h/H - signed/unsigned short\n** l/L - signed/unsigned long\n** T - size_t\n** i/In - signed/unsigned integer with size `n' (default is size of int)\n** cn - sequence of `n' chars (from/to a string); when packing, n==0 means\n the whole string; when unpacking, n==0 means use the previous\n read number as the string length\n** s - zero-terminated string\n** f - float\n** d - double\n** ' ' - ignored\n*/\n\n\n#include \n#include \n#include \n#include \n#include \n\n\n#include \"lua.h\"\n#include \"lauxlib.h\"\n\n\n#if (LUA_VERSION_NUM >= 502)\n\n#define luaL_register(L,n,f)\tluaL_newlib(L,f)\n\n#endif\n\n\n/* basic integer type */\n#if !defined(STRUCT_INT)\n#define STRUCT_INT\tlong\n#endif\n\ntypedef STRUCT_INT Inttype;\n\n/* corresponding unsigned version */\ntypedef unsigned STRUCT_INT Uinttype;\n\n\n/* maximum size (in bytes) for integral types */\n#define MAXINTSIZE\t32\n\n/* is 'x' a power of 2? */\n#define isp2(x)\t\t((x) > 0 && ((x) & ((x) - 1)) == 0)\n\n/* dummy structure to get alignment requirements */\nstruct cD {\n char c;\n double d;\n};\n\n\n#define PADDING\t\t(sizeof(struct cD) - sizeof(double))\n#define MAXALIGN \t(PADDING > sizeof(int) ? PADDING : sizeof(int))\n\n\n/* endian options */\n#define BIG\t0\n#define LITTLE\t1\n\n\nstatic union {\n int dummy;\n char endian;\n} const native = {1};\n\n\ntypedef struct Header {\n int endian;\n int align;\n} Header;\n\n\nstatic int getnum (const char **fmt, int df) {\n if (!isdigit(**fmt)) /* no number? */\n return df; /* return default value */\n else {\n int a = 0;\n do {\n a = a*10 + *((*fmt)++) - '0';\n } while (isdigit(**fmt));\n return a;\n }\n}\n\n\n#define defaultoptions(h)\t((h)->endian = native.endian, (h)->align = 1)\n\n\n\nstatic size_t optsize (lua_State *L, char opt, const char **fmt) {\n switch (opt) {\n case 'B': case 'b': return sizeof(char);\n case 'H': case 'h': return sizeof(short);\n case 'L': case 'l': return sizeof(long);\n case 'T': return sizeof(size_t);\n case 'f': return sizeof(float);\n case 'd': return sizeof(double);\n case 'x': return 1;\n case 'c': return getnum(fmt, 1);\n case 'i': case 'I': {\n int sz = getnum(fmt, sizeof(int));\n if (sz > MAXINTSIZE)\n luaL_error(L, \"integral size %d is larger than limit of %d\",\n sz, MAXINTSIZE);\n return sz;\n }\n default: return 0; /* other cases do not need alignment */\n }\n}\n\n\n/*\n** return number of bytes needed to align an element of size 'size'\n** at current position 'len'\n*/\nstatic int gettoalign (size_t len, Header *h, int opt, size_t size) {\n if (size == 0 || opt == 'c') return 0;\n if (size > (size_t)h->align)\n size = h->align; /* respect max. alignment */\n return (size - (len & (size - 1))) & (size - 1);\n}\n\n\n/*\n** options to control endianess and alignment\n*/\nstatic void controloptions (lua_State *L, int opt, const char **fmt,\n Header *h) {\n switch (opt) {\n case ' ': return; /* ignore white spaces */\n case '>': h->endian = BIG; return;\n case '<': h->endian = LITTLE; return;\n case '!': {\n int a = getnum(fmt, MAXALIGN);\n if (!isp2(a))\n luaL_error(L, \"alignment %d is not a power of 2\", a);\n h->align = a;\n return;\n }\n default: {\n const char *msg = lua_pushfstring(L, \"invalid format option '%c'\", opt);\n luaL_argerror(L, 1, msg);\n }\n }\n}\n\n\nstatic void putinteger (lua_State *L, luaL_Buffer *b, int arg, int endian,\n int size) {\n lua_Number n = luaL_checknumber(L, arg);\n Uinttype value;\n char buff[MAXINTSIZE];\n if (n < 0)\n value = (Uinttype)(Inttype)n;\n else\n value = (Uinttype)n;\n if (endian == LITTLE) {\n int i;\n for (i = 0; i < size; i++) {\n buff[i] = (value & 0xff);\n value >>= 8;\n }\n }\n else {\n int i;\n for (i = size - 1; i >= 0; i--) {\n buff[i] = (value & 0xff);\n value >>= 8;\n }\n }\n luaL_addlstring(b, buff, size);\n}\n\n\nstatic void correctbytes (char *b, int size, int endian) {\n if (endian != native.endian) {\n int i = 0;\n while (i < --size) {\n char temp = b[i];\n b[i++] = b[size];\n b[size] = temp;\n }\n }\n}\n\n\nstatic int b_pack (lua_State *L) {\n luaL_Buffer b;\n const char *fmt = luaL_checkstring(L, 1);\n Header h;\n int arg = 2;\n size_t totalsize = 0;\n defaultoptions(&h);\n lua_pushnil(L); /* mark to separate arguments from string buffer */\n luaL_buffinit(L, &b);\n while (*fmt != '\\0') {\n int opt = *fmt++;\n size_t size = optsize(L, opt, &fmt);\n int toalign = gettoalign(totalsize, &h, opt, size);\n totalsize += toalign;\n while (toalign-- > 0) luaL_addchar(&b, '\\0');\n switch (opt) {\n case 'b': case 'B': case 'h': case 'H':\n case 'l': case 'L': case 'T': case 'i': case 'I': { /* integer types */\n putinteger(L, &b, arg++, h.endian, size);\n break;\n }\n case 'x': {\n luaL_addchar(&b, '\\0');\n break;\n }\n case 'f': {\n float f = (float)luaL_checknumber(L, arg++);\n correctbytes((char *)&f, size, h.endian);\n luaL_addlstring(&b, (char *)&f, size);\n break;\n }\n case 'd': {\n double d = luaL_checknumber(L, arg++);\n correctbytes((char *)&d, size, h.endian);\n luaL_addlstring(&b, (char *)&d, size);\n break;\n }\n case 'c': case 's': {\n size_t l;\n const char *s = luaL_checklstring(L, arg++, &l);\n if (size == 0) size = l;\n luaL_argcheck(L, l >= (size_t)size, arg, \"string too short\");\n luaL_addlstring(&b, s, size);\n if (opt == 's') {\n luaL_addchar(&b, '\\0'); /* add zero at the end */\n size++;\n }\n break;\n }\n default: controloptions(L, opt, &fmt, &h);\n }\n totalsize += size;\n }\n luaL_pushresult(&b);\n return 1;\n}\n\n\nstatic lua_Number getinteger (const char *buff, int endian,\n int issigned, int size) {\n Uinttype l = 0;\n int i;\n if (endian == BIG) {\n for (i = 0; i < size; i++) {\n l <<= 8;\n l |= (Uinttype)(unsigned char)buff[i];\n }\n }\n else {\n for (i = size - 1; i >= 0; i--) {\n l <<= 8;\n l |= (Uinttype)(unsigned char)buff[i];\n }\n }\n if (!issigned)\n return (lua_Number)l;\n else { /* signed format */\n Uinttype mask = (Uinttype)(~((Uinttype)0)) << (size*8 - 1);\n if (l & mask) /* negative value? */\n l |= mask; /* signal extension */\n return (lua_Number)(Inttype)l;\n }\n}\n\n\nstatic int b_unpack (lua_State *L) {\n Header h;\n const char *fmt = luaL_checkstring(L, 1);\n size_t ld;\n const char *data = luaL_checklstring(L, 2, &ld);\n size_t pos = luaL_optinteger(L, 3, 1) - 1;\n defaultoptions(&h);\n lua_settop(L, 2);\n while (*fmt) {\n int opt = *fmt++;\n size_t size = optsize(L, opt, &fmt);\n pos += gettoalign(pos, &h, opt, size);\n luaL_argcheck(L, pos+size <= ld, 2, \"data string too short\");\n luaL_checkstack(L, 1, \"too many results\");\n switch (opt) {\n case 'b': case 'B': case 'h': case 'H':\n case 'l': case 'L': case 'T': case 'i': case 'I': { /* integer types */\n int issigned = islower(opt);\n lua_Number res = getinteger(data+pos, h.endian, issigned, size);\n lua_pushnumber(L, res);\n break;\n }\n case 'x': {\n break;\n }\n case 'f': {\n float f;\n memcpy(&f, data+pos, size);\n correctbytes((char *)&f, sizeof(f), h.endian);\n lua_pushnumber(L, f);\n break;\n }\n case 'd': {\n double d;\n memcpy(&d, data+pos, size);\n correctbytes((char *)&d, sizeof(d), h.endian);\n lua_pushnumber(L, d);\n break;\n }\n case 'c': {\n if (size == 0) {\n if (!lua_isnumber(L, -1))\n luaL_error(L, \"format `c0' needs a previous size\");\n size = lua_tonumber(L, -1);\n lua_pop(L, 1);\n luaL_argcheck(L, pos+size <= ld, 2, \"data string too short\");\n }\n lua_pushlstring(L, data+pos, size);\n break;\n }\n case 's': {\n const char *e = (const char *)memchr(data+pos, '\\0', ld - pos);\n if (e == NULL)\n luaL_error(L, \"unfinished string in data\");\n size = (e - (data+pos)) + 1;\n lua_pushlstring(L, data+pos, size - 1);\n break;\n }\n default: controloptions(L, opt, &fmt, &h);\n }\n pos += size;\n }\n lua_pushinteger(L, pos + 1);\n return lua_gettop(L) - 2;\n}\n\n\nstatic int b_size (lua_State *L) {\n Header h;\n const char *fmt = luaL_checkstring(L, 1);\n size_t pos = 0;\n defaultoptions(&h);\n while (*fmt) {\n int opt = *fmt++;\n size_t size = optsize(L, opt, &fmt);\n pos += gettoalign(pos, &h, opt, size);\n if (opt == 's')\n luaL_argerror(L, 1, \"option 's' has no fixed size\");\n else if (opt == 'c' && size == 0)\n luaL_argerror(L, 1, \"option 'c0' has no fixed size\");\n if (!isalnum(opt))\n controloptions(L, opt, &fmt, &h);\n pos += size;\n }\n lua_pushinteger(L, pos);\n return 1;\n}\n\n/* }====================================================== */\n\n\n\nstatic const struct luaL_Reg thislib[] = {\n {\"pack\", b_pack},\n {\"unpack\", b_unpack},\n {\"size\", b_size},\n {NULL, NULL}\n};\n\n\nLUALIB_API int luaopen_struct (lua_State *L);\n\nLUALIB_API int luaopen_struct (lua_State *L) {\n luaL_register(L, \"struct\", thislib);\n return 1;\n}\n\n\n/******************************************************************************\n* Copyright (C) 2010-2012 Lua.org, PUC-Rio. All rights reserved.\n*\n* Permission is hereby granted, free of charge, to any person obtaining\n* a copy of this software and associated documentation files (the\n* \"Software\"), to deal in the Software without restriction, including\n* without limitation the rights to use, copy, modify, merge, publish,\n* distribute, sublicense, and/or sell copies of the Software, and to\n* permit persons to whom the Software is furnished to do so, subject to\n* the following conditions:\n*\n* The above copyright notice and this permission notice shall be\n* included in all copies or substantial portions of the Software.\n*\n* THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n******************************************************************************/\n\n" }, { "path": "deps/lua/src/luac.c", "content": "/*\n** $Id: luac.c,v 1.54 2006/06/02 17:37:11 lhf Exp $\n** Lua compiler (saves bytecodes to files; also list bytecodes)\n** See Copyright Notice in lua.h\n*/\n\n#include \n#include \n#include \n#include \n\n#define luac_c\n#define LUA_CORE\n\n#include \"lua.h\"\n#include \"lauxlib.h\"\n\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lstring.h\"\n#include \"lundump.h\"\n\n#define PROGNAME\t\"luac\"\t\t/* default program name */\n#define\tOUTPUT\t\tPROGNAME \".out\"\t/* default output file */\n\nstatic int listing=0;\t\t\t/* list bytecodes? */\nstatic int dumping=1;\t\t\t/* dump bytecodes? */\nstatic int stripping=0;\t\t\t/* strip debug information? */\nstatic char Output[]={ OUTPUT };\t/* default output file name */\nstatic const char* output=Output;\t/* actual output file name */\nstatic const char* progname=PROGNAME;\t/* actual program name */\n\nstatic void fatal(const char* message)\n{\n fprintf(stderr,\"%s: %s\\n\",progname,message);\n exit(EXIT_FAILURE);\n}\n\nstatic void cannot(const char* what)\n{\n fprintf(stderr,\"%s: cannot %s %s: %s\\n\",progname,what,output,strerror(errno));\n exit(EXIT_FAILURE);\n}\n\nstatic void usage(const char* message)\n{\n if (*message=='-')\n fprintf(stderr,\"%s: unrecognized option \" LUA_QS \"\\n\",progname,message);\n else\n fprintf(stderr,\"%s: %s\\n\",progname,message);\n fprintf(stderr,\n \"usage: %s [options] [filenames].\\n\"\n \"Available options are:\\n\"\n \" - process stdin\\n\"\n \" -l list\\n\"\n \" -o name output to file \" LUA_QL(\"name\") \" (default is \\\"%s\\\")\\n\"\n \" -p parse only\\n\"\n \" -s strip debug information\\n\"\n \" -v show version information\\n\"\n \" -- stop handling options\\n\",\n progname,Output);\n exit(EXIT_FAILURE);\n}\n\n#define\tIS(s)\t(strcmp(argv[i],s)==0)\n\nstatic int doargs(int argc, char* argv[])\n{\n int i;\n int version=0;\n if (argv[0]!=NULL && *argv[0]!=0) progname=argv[0];\n for (i=1; itop+(i))->l.p)\n\nstatic const Proto* combine(lua_State* L, int n)\n{\n if (n==1)\n return toproto(L,-1);\n else\n {\n int i,pc;\n Proto* f=luaF_newproto(L);\n setptvalue2s(L,L->top,f); incr_top(L);\n f->source=luaS_newliteral(L,\"=(\" PROGNAME \")\");\n f->maxstacksize=1;\n pc=2*n+1;\n f->code=luaM_newvector(L,pc,Instruction);\n f->sizecode=pc;\n f->p=luaM_newvector(L,n,Proto*);\n f->sizep=n;\n pc=0;\n for (i=0; ip[i]=toproto(L,i-n-1);\n f->code[pc++]=CREATE_ABx(OP_CLOSURE,0,i);\n f->code[pc++]=CREATE_ABC(OP_CALL,0,1,1);\n }\n f->code[pc++]=CREATE_ABC(OP_RETURN,0,1,0);\n return f;\n }\n}\n\nstatic int writer(lua_State* L, const void* p, size_t size, void* u)\n{\n UNUSED(L);\n return (fwrite(p,size,1,(FILE*)u)!=1) && (size!=0);\n}\n\nstruct Smain {\n int argc;\n char** argv;\n};\n\nstatic int pmain(lua_State* L)\n{\n struct Smain* s = (struct Smain*)lua_touserdata(L, 1);\n int argc=s->argc;\n char** argv=s->argv;\n const Proto* f;\n int i;\n if (!lua_checkstack(L,argc)) fatal(\"too many input files\");\n for (i=0; i1);\n if (dumping)\n {\n FILE* D= (output==NULL) ? stdout : fopen(output,\"wb\");\n if (D==NULL) cannot(\"open\");\n lua_lock(L);\n luaU_dump(L,f,writer,D,stripping);\n lua_unlock(L);\n if (ferror(D)) cannot(\"write\");\n if (fclose(D)) cannot(\"close\");\n }\n return 0;\n}\n\nint main(int argc, char* argv[])\n{\n lua_State* L;\n struct Smain s;\n int i=doargs(argc,argv);\n argc-=i; argv+=i;\n if (argc<=0) usage(\"no input files given\");\n L=lua_open();\n if (L==NULL) fatal(\"not enough memory for state\");\n s.argc=argc;\n s.argv=argv;\n if (lua_cpcall(L,pmain,&s)!=0) fatal(lua_tostring(L,-1));\n lua_close(L);\n return EXIT_SUCCESS;\n}\n" }, { "path": "deps/lua/src/luaconf.h", "content": "/*\n** $Id: luaconf.h,v 1.82.1.7 2008/02/11 16:25:08 roberto Exp $\n** Configuration file for Lua\n** See Copyright Notice in lua.h\n*/\n\n\n#ifndef lconfig_h\n#define lconfig_h\n\n#include \n#include \n\n\n/*\n** ==================================================================\n** Search for \"@@\" to find all configurable definitions.\n** ===================================================================\n*/\n\n\n/*\n@@ LUA_ANSI controls the use of non-ansi features.\n** CHANGE it (define it) if you want Lua to avoid the use of any\n** non-ansi feature or library.\n*/\n#if defined(__STRICT_ANSI__)\n#define LUA_ANSI\n#endif\n\n\n#if !defined(LUA_ANSI) && defined(_WIN32)\n#define LUA_WIN\n#endif\n\n#if defined(LUA_USE_LINUX)\n#define LUA_USE_POSIX\n#define LUA_USE_DLOPEN\t\t/* needs an extra library: -ldl */\n#define LUA_USE_READLINE\t/* needs some extra libraries */\n#endif\n\n#if defined(LUA_USE_MACOSX)\n#define LUA_USE_POSIX\n#define LUA_DL_DYLD\t\t/* does not need extra library */\n#endif\n\n\n\n/*\n@@ LUA_USE_POSIX includes all functionallity listed as X/Open System\n@* Interfaces Extension (XSI).\n** CHANGE it (define it) if your system is XSI compatible.\n*/\n#if defined(LUA_USE_POSIX)\n#define LUA_USE_MKSTEMP\n#define LUA_USE_ISATTY\n#define LUA_USE_POPEN\n#define LUA_USE_ULONGJMP\n#endif\n\n\n/*\n@@ LUA_PATH and LUA_CPATH are the names of the environment variables that\n@* Lua check to set its paths.\n@@ LUA_INIT is the name of the environment variable that Lua\n@* checks for initialization code.\n** CHANGE them if you want different names.\n*/\n#define LUA_PATH \"LUA_PATH\"\n#define LUA_CPATH \"LUA_CPATH\"\n#define LUA_INIT\t\"LUA_INIT\"\n\n\n/*\n@@ LUA_PATH_DEFAULT is the default path that Lua uses to look for\n@* Lua libraries.\n@@ LUA_CPATH_DEFAULT is the default path that Lua uses to look for\n@* C libraries.\n** CHANGE them if your machine has a non-conventional directory\n** hierarchy or if you want to install your libraries in\n** non-conventional directories.\n*/\n#if defined(_WIN32)\n/*\n** In Windows, any exclamation mark ('!') in the path is replaced by the\n** path of the directory of the executable file of the current process.\n*/\n#define LUA_LDIR\t\"!\\\\lua\\\\\"\n#define LUA_CDIR\t\"!\\\\\"\n#define LUA_PATH_DEFAULT \\\n\t\t\".\\\\?.lua;\" LUA_LDIR\"?.lua;\" LUA_LDIR\"?\\\\init.lua;\" \\\n\t\t LUA_CDIR\"?.lua;\" LUA_CDIR\"?\\\\init.lua\"\n#define LUA_CPATH_DEFAULT \\\n\t\".\\\\?.dll;\" LUA_CDIR\"?.dll;\" LUA_CDIR\"loadall.dll\"\n\n#else\n#define LUA_ROOT\t\"/usr/local/\"\n#define LUA_LDIR\tLUA_ROOT \"share/lua/5.1/\"\n#define LUA_CDIR\tLUA_ROOT \"lib/lua/5.1/\"\n#define LUA_PATH_DEFAULT \\\n\t\t\"./?.lua;\" LUA_LDIR\"?.lua;\" LUA_LDIR\"?/init.lua;\" \\\n\t\t LUA_CDIR\"?.lua;\" LUA_CDIR\"?/init.lua\"\n#define LUA_CPATH_DEFAULT \\\n\t\"./?.so;\" LUA_CDIR\"?.so;\" LUA_CDIR\"loadall.so\"\n#endif\n\n\n/*\n@@ LUA_DIRSEP is the directory separator (for submodules).\n** CHANGE it if your machine does not use \"/\" as the directory separator\n** and is not Windows. (On Windows Lua automatically uses \"\\\".)\n*/\n#if defined(_WIN32)\n#define LUA_DIRSEP\t\"\\\\\"\n#else\n#define LUA_DIRSEP\t\"/\"\n#endif\n\n\n/*\n@@ LUA_PATHSEP is the character that separates templates in a path.\n@@ LUA_PATH_MARK is the string that marks the substitution points in a\n@* template.\n@@ LUA_EXECDIR in a Windows path is replaced by the executable's\n@* directory.\n@@ LUA_IGMARK is a mark to ignore all before it when bulding the\n@* luaopen_ function name.\n** CHANGE them if for some reason your system cannot use those\n** characters. (E.g., if one of those characters is a common character\n** in file/directory names.) Probably you do not need to change them.\n*/\n#define LUA_PATHSEP\t\";\"\n#define LUA_PATH_MARK\t\"?\"\n#define LUA_EXECDIR\t\"!\"\n#define LUA_IGMARK\t\"-\"\n\n\n/*\n@@ LUA_INTEGER is the integral type used by lua_pushinteger/lua_tointeger.\n** CHANGE that if ptrdiff_t is not adequate on your machine. (On most\n** machines, ptrdiff_t gives a good choice between int or long.)\n*/\n#define LUA_INTEGER\tptrdiff_t\n\n\n/*\n@@ LUA_API is a mark for all core API functions.\n@@ LUALIB_API is a mark for all standard library functions.\n** CHANGE them if you need to define those functions in some special way.\n** For instance, if you want to create one Windows DLL with the core and\n** the libraries, you may want to use the following definition (define\n** LUA_BUILD_AS_DLL to get it).\n*/\n#if defined(LUA_BUILD_AS_DLL)\n\n#if defined(LUA_CORE) || defined(LUA_LIB)\n#define LUA_API __declspec(dllexport)\n#else\n#define LUA_API __declspec(dllimport)\n#endif\n\n#else\n\n#define LUA_API\t\textern\n\n#endif\n\n/* more often than not the libs go together with the core */\n#define LUALIB_API\tLUA_API\n\n\n/*\n@@ LUAI_FUNC is a mark for all extern functions that are not to be\n@* exported to outside modules.\n@@ LUAI_DATA is a mark for all extern (const) variables that are not to\n@* be exported to outside modules.\n** CHANGE them if you need to mark them in some special way. Elf/gcc\n** (versions 3.2 and later) mark them as \"hidden\" to optimize access\n** when Lua is compiled as a shared library.\n*/\n#if defined(luaall_c)\n#define LUAI_FUNC\tstatic\n#define LUAI_DATA\t/* empty */\n\n#elif defined(__GNUC__) && ((__GNUC__*100 + __GNUC_MINOR__) >= 302) && \\\n defined(__ELF__)\n#define LUAI_FUNC\t__attribute__((visibility(\"hidden\"))) extern\n#define LUAI_DATA\tLUAI_FUNC\n\n#else\n#define LUAI_FUNC\textern\n#define LUAI_DATA\textern\n#endif\n\n\n\n/*\n@@ LUA_QL describes how error messages quote program elements.\n** CHANGE it if you want a different appearance.\n*/\n#define LUA_QL(x)\t\"'\" x \"'\"\n#define LUA_QS\t\tLUA_QL(\"%s\")\n\n\n/*\n@@ LUA_IDSIZE gives the maximum size for the description of the source\n@* of a function in debug information.\n** CHANGE it if you want a different size.\n*/\n#define LUA_IDSIZE\t60\n\n\n/*\n** {==================================================================\n** Stand-alone configuration\n** ===================================================================\n*/\n\n#if defined(lua_c) || defined(luaall_c)\n\n/*\n@@ lua_stdin_is_tty detects whether the standard input is a 'tty' (that\n@* is, whether we're running lua interactively).\n** CHANGE it if you have a better definition for non-POSIX/non-Windows\n** systems.\n*/\n#if defined(LUA_USE_ISATTY)\n#include \n#define lua_stdin_is_tty()\tisatty(0)\n#elif defined(LUA_WIN)\n#include \n#include \n#define lua_stdin_is_tty()\t_isatty(_fileno(stdin))\n#else\n#define lua_stdin_is_tty()\t1 /* assume stdin is a tty */\n#endif\n\n\n/*\n@@ LUA_PROMPT is the default prompt used by stand-alone Lua.\n@@ LUA_PROMPT2 is the default continuation prompt used by stand-alone Lua.\n** CHANGE them if you want different prompts. (You can also change the\n** prompts dynamically, assigning to globals _PROMPT/_PROMPT2.)\n*/\n#define LUA_PROMPT\t\t\"> \"\n#define LUA_PROMPT2\t\t\">> \"\n\n\n/*\n@@ LUA_PROGNAME is the default name for the stand-alone Lua program.\n** CHANGE it if your stand-alone interpreter has a different name and\n** your system is not able to detect that name automatically.\n*/\n#define LUA_PROGNAME\t\t\"lua\"\n\n\n/*\n@@ LUA_MAXINPUT is the maximum length for an input line in the\n@* stand-alone interpreter.\n** CHANGE it if you need longer lines.\n*/\n#define LUA_MAXINPUT\t512\n\n\n/*\n@@ lua_readline defines how to show a prompt and then read a line from\n@* the standard input.\n@@ lua_saveline defines how to \"save\" a read line in a \"history\".\n@@ lua_freeline defines how to free a line read by lua_readline.\n** CHANGE them if you want to improve this functionality (e.g., by using\n** GNU readline and history facilities).\n*/\n#if defined(LUA_USE_READLINE)\n#include \n#include \n#include \n#define lua_readline(L,b,p)\t((void)L, ((b)=readline(p)) != NULL)\n#define lua_saveline(L,idx) \\\n\tif (lua_strlen(L,idx) > 0) /* non-empty line? */ \\\n\t add_history(lua_tostring(L, idx)); /* add it to history */\n#define lua_freeline(L,b)\t((void)L, free(b))\n#else\n#define lua_readline(L,b,p)\t\\\n\t((void)L, fputs(p, stdout), fflush(stdout), /* show prompt */ \\\n\tfgets(b, LUA_MAXINPUT, stdin) != NULL) /* get line */\n#define lua_saveline(L,idx)\t{ (void)L; (void)idx; }\n#define lua_freeline(L,b)\t{ (void)L; (void)b; }\n#endif\n\n#endif\n\n/* }================================================================== */\n\n\n/*\n@@ LUAI_GCPAUSE defines the default pause between garbage-collector cycles\n@* as a percentage.\n** CHANGE it if you want the GC to run faster or slower (higher values\n** mean larger pauses which mean slower collection.) You can also change\n** this value dynamically.\n*/\n#define LUAI_GCPAUSE\t200 /* 200% (wait memory to double before next GC) */\n\n\n/*\n@@ LUAI_GCMUL defines the default speed of garbage collection relative to\n@* memory allocation as a percentage.\n** CHANGE it if you want to change the granularity of the garbage\n** collection. (Higher values mean coarser collections. 0 represents\n** infinity, where each step performs a full collection.) You can also\n** change this value dynamically.\n*/\n#define LUAI_GCMUL\t200 /* GC runs 'twice the speed' of memory allocation */\n\n\n\n/*\n@@ LUA_COMPAT_GETN controls compatibility with old getn behavior.\n** CHANGE it (define it) if you want exact compatibility with the\n** behavior of setn/getn in Lua 5.0.\n*/\n#undef LUA_COMPAT_GETN\n\n/*\n@@ LUA_COMPAT_LOADLIB controls compatibility about global loadlib.\n** CHANGE it to undefined as soon as you do not need a global 'loadlib'\n** function (the function is still available as 'package.loadlib').\n*/\n#undef LUA_COMPAT_LOADLIB\n\n/*\n@@ LUA_COMPAT_VARARG controls compatibility with old vararg feature.\n** CHANGE it to undefined as soon as your programs use only '...' to\n** access vararg parameters (instead of the old 'arg' table).\n*/\n#define LUA_COMPAT_VARARG\n\n/*\n@@ LUA_COMPAT_MOD controls compatibility with old math.mod function.\n** CHANGE it to undefined as soon as your programs use 'math.fmod' or\n** the new '%' operator instead of 'math.mod'.\n*/\n#define LUA_COMPAT_MOD\n\n/*\n@@ LUA_COMPAT_LSTR controls compatibility with old long string nesting\n@* facility.\n** CHANGE it to 2 if you want the old behaviour, or undefine it to turn\n** off the advisory error when nesting [[...]].\n*/\n#define LUA_COMPAT_LSTR\t\t1\n\n/*\n@@ LUA_COMPAT_GFIND controls compatibility with old 'string.gfind' name.\n** CHANGE it to undefined as soon as you rename 'string.gfind' to\n** 'string.gmatch'.\n*/\n#define LUA_COMPAT_GFIND\n\n/*\n@@ LUA_COMPAT_OPENLIB controls compatibility with old 'luaL_openlib'\n@* behavior.\n** CHANGE it to undefined as soon as you replace to 'luaL_register'\n** your uses of 'luaL_openlib'\n*/\n#define LUA_COMPAT_OPENLIB\n\n\n\n/*\n@@ luai_apicheck is the assert macro used by the Lua-C API.\n** CHANGE luai_apicheck if you want Lua to perform some checks in the\n** parameters it gets from API calls. This may slow down the interpreter\n** a bit, but may be quite useful when debugging C code that interfaces\n** with Lua. A useful redefinition is to use assert.h.\n*/\n#if defined(LUA_USE_APICHECK)\n#include \n#define luai_apicheck(L,o)\t{ (void)L; assert(o); }\n#else\n#define luai_apicheck(L,o)\t{ (void)L; }\n#endif\n\n\n/*\n@@ LUAI_BITSINT defines the number of bits in an int.\n** CHANGE here if Lua cannot automatically detect the number of bits of\n** your machine. Probably you do not need to change this.\n*/\n/* avoid overflows in comparison */\n#if INT_MAX-20 < 32760\n#define LUAI_BITSINT\t16\n#elif INT_MAX > 2147483640L\n/* int has at least 32 bits */\n#define LUAI_BITSINT\t32\n#else\n#error \"you must define LUA_BITSINT with number of bits in an integer\"\n#endif\n\n\n/*\n@@ LUAI_UINT32 is an unsigned integer with at least 32 bits.\n@@ LUAI_INT32 is an signed integer with at least 32 bits.\n@@ LUAI_UMEM is an unsigned integer big enough to count the total\n@* memory used by Lua.\n@@ LUAI_MEM is a signed integer big enough to count the total memory\n@* used by Lua.\n** CHANGE here if for some weird reason the default definitions are not\n** good enough for your machine. (The definitions in the 'else'\n** part always works, but may waste space on machines with 64-bit\n** longs.) Probably you do not need to change this.\n*/\n#if LUAI_BITSINT >= 32\n#define LUAI_UINT32\tunsigned int\n#define LUAI_INT32\tint\n#define LUAI_MAXINT32\tINT_MAX\n#define LUAI_UMEM\tsize_t\n#define LUAI_MEM\tptrdiff_t\n#else\n/* 16-bit ints */\n#define LUAI_UINT32\tunsigned long\n#define LUAI_INT32\tlong\n#define LUAI_MAXINT32\tLONG_MAX\n#define LUAI_UMEM\tunsigned long\n#define LUAI_MEM\tlong\n#endif\n\n\n/*\n@@ LUAI_MAXCALLS limits the number of nested calls.\n** CHANGE it if you need really deep recursive calls. This limit is\n** arbitrary; its only purpose is to stop infinite recursion before\n** exhausting memory.\n*/\n#define LUAI_MAXCALLS\t20000\n\n\n/*\n@@ LUAI_MAXCSTACK limits the number of Lua stack slots that a C function\n@* can use.\n** CHANGE it if you need lots of (Lua) stack space for your C\n** functions. This limit is arbitrary; its only purpose is to stop C\n** functions to consume unlimited stack space. (must be smaller than\n** -LUA_REGISTRYINDEX)\n*/\n#define LUAI_MAXCSTACK\t8000\n\n\n\n/*\n** {==================================================================\n** CHANGE (to smaller values) the following definitions if your system\n** has a small C stack. (Or you may want to change them to larger\n** values if your system has a large C stack and these limits are\n** too rigid for you.) Some of these constants control the size of\n** stack-allocated arrays used by the compiler or the interpreter, while\n** others limit the maximum number of recursive calls that the compiler\n** or the interpreter can perform. Values too large may cause a C stack\n** overflow for some forms of deep constructs.\n** ===================================================================\n*/\n\n\n/*\n@@ LUAI_MAXCCALLS is the maximum depth for nested C calls (short) and\n@* syntactical nested non-terminals in a program.\n*/\n#define LUAI_MAXCCALLS\t\t200\n\n\n/*\n@@ LUAI_MAXVARS is the maximum number of local variables per function\n@* (must be smaller than 250).\n*/\n#define LUAI_MAXVARS\t\t200\n\n\n/*\n@@ LUAI_MAXUPVALUES is the maximum number of upvalues per function\n@* (must be smaller than 250).\n*/\n#define LUAI_MAXUPVALUES\t60\n\n\n/*\n@@ LUAL_BUFFERSIZE is the buffer size used by the lauxlib buffer system.\n*/\n#define LUAL_BUFFERSIZE\t\tBUFSIZ\n\n/* }================================================================== */\n\n\n\n\n/*\n** {==================================================================\n@@ LUA_NUMBER is the type of numbers in Lua.\n** CHANGE the following definitions only if you want to build Lua\n** with a number type different from double. You may also need to\n** change lua_number2int & lua_number2integer.\n** ===================================================================\n*/\n\n#define LUA_NUMBER_DOUBLE\n#define LUA_NUMBER\tdouble\n\n/*\n@@ LUAI_UACNUMBER is the result of an 'usual argument conversion'\n@* over a number.\n*/\n#define LUAI_UACNUMBER\tdouble\n\n\n/*\n@@ LUA_NUMBER_SCAN is the format for reading numbers.\n@@ LUA_NUMBER_FMT is the format for writing numbers.\n@@ lua_number2str converts a number to a string.\n@@ LUAI_MAXNUMBER2STR is maximum size of previous conversion.\n@@ lua_str2number converts a string to a number.\n*/\n#define LUA_NUMBER_SCAN\t\t\"%lf\"\n#define LUA_NUMBER_FMT\t\t\"%.14g\"\n#define lua_number2str(s,n)\tsprintf((s), LUA_NUMBER_FMT, (n))\n#define LUAI_MAXNUMBER2STR\t32 /* 16 digits, sign, point, and \\0 */\n#define lua_str2number(s,p)\tstrtod((s), (p))\n\n\n/*\n@@ The luai_num* macros define the primitive operations over numbers.\n*/\n#if defined(LUA_CORE)\n#include \n#define luai_numadd(a,b)\t((a)+(b))\n#define luai_numsub(a,b)\t((a)-(b))\n#define luai_nummul(a,b)\t((a)*(b))\n#define luai_numdiv(a,b)\t((a)/(b))\n#define luai_nummod(a,b)\t((a) - floor((a)/(b))*(b))\n#define luai_numpow(a,b)\t(pow(a,b))\n#define luai_numunm(a)\t\t(-(a))\n#define luai_numeq(a,b)\t\t((a)==(b))\n#define luai_numlt(a,b)\t\t((a)<(b))\n#define luai_numle(a,b)\t\t((a)<=(b))\n#define luai_numisnan(a)\t(!luai_numeq((a), (a)))\n#endif\n\n\n/*\n@@ lua_number2int is a macro to convert lua_Number to int.\n@@ lua_number2integer is a macro to convert lua_Number to lua_Integer.\n** CHANGE them if you know a faster way to convert a lua_Number to\n** int (with any rounding method and without throwing errors) in your\n** system. In Pentium machines, a naive typecast from double to int\n** in C is extremely slow, so any alternative is worth trying.\n*/\n\n/* On a Pentium, resort to a trick */\n#if defined(LUA_NUMBER_DOUBLE) && !defined(LUA_ANSI) && !defined(__SSE2__) && \\\n (defined(__i386) || defined (_M_IX86) || defined(__i386__))\n\n/* On a Microsoft compiler, use assembler */\n#if defined(_MSC_VER)\n\n#define lua_number2int(i,d) __asm fld d __asm fistp i\n#define lua_number2integer(i,n)\t\tlua_number2int(i, n)\n\n/* the next trick should work on any Pentium, but sometimes clashes\n with a DirectX idiosyncrasy */\n#else\n\nunion luai_Cast { double l_d; long l_l; };\n#define lua_number2int(i,d) \\\n { volatile union luai_Cast u; u.l_d = (d) + 6755399441055744.0; (i) = u.l_l; }\n#define lua_number2integer(i,n)\t\tlua_number2int(i, n)\n\n#endif\n\n\n/* this option always works, but may be slow */\n#else\n#define lua_number2int(i,d)\t((i)=(int)(d))\n#define lua_number2integer(i,d)\t((i)=(lua_Integer)(d))\n\n#endif\n\n/* }================================================================== */\n\n\n/*\n@@ LUAI_USER_ALIGNMENT_T is a type that requires maximum alignment.\n** CHANGE it if your system requires alignments larger than double. (For\n** instance, if your system supports long doubles and they must be\n** aligned in 16-byte boundaries, then you should add long double in the\n** union.) Probably you do not need to change this.\n*/\n#define LUAI_USER_ALIGNMENT_T\tunion { double u; void *s; long l; }\n\n\n/*\n@@ LUAI_THROW/LUAI_TRY define how Lua does exception handling.\n** CHANGE them if you prefer to use longjmp/setjmp even with C++\n** or if want/don't to use _longjmp/_setjmp instead of regular\n** longjmp/setjmp. By default, Lua handles errors with exceptions when\n** compiling as C++ code, with _longjmp/_setjmp when asked to use them,\n** and with longjmp/setjmp otherwise.\n*/\n#if defined(__cplusplus)\n/* C++ exceptions */\n#define LUAI_THROW(L,c)\tthrow(c)\n#define LUAI_TRY(L,c,a)\ttry { a } catch(...) \\\n\t{ if ((c)->status == 0) (c)->status = -1; }\n#define luai_jmpbuf\tint /* dummy variable */\n\n#elif defined(LUA_USE_ULONGJMP)\n/* in Unix, try _longjmp/_setjmp (more efficient) */\n#define LUAI_THROW(L,c)\t_longjmp((c)->b, 1)\n#define LUAI_TRY(L,c,a)\tif (_setjmp((c)->b) == 0) { a }\n#define luai_jmpbuf\tjmp_buf\n\n#else\n/* default handling with long jumps */\n#define LUAI_THROW(L,c)\tlongjmp((c)->b, 1)\n#define LUAI_TRY(L,c,a)\tif (setjmp((c)->b) == 0) { a }\n#define luai_jmpbuf\tjmp_buf\n\n#endif\n\n\n/*\n@@ LUA_MAXCAPTURES is the maximum number of captures that a pattern\n@* can do during pattern-matching.\n** CHANGE it if you need more captures. This limit is arbitrary.\n*/\n#define LUA_MAXCAPTURES\t\t32\n\n\n/*\n@@ lua_tmpnam is the function that the OS library uses to create a\n@* temporary name.\n@@ LUA_TMPNAMBUFSIZE is the maximum size of a name created by lua_tmpnam.\n** CHANGE them if you have an alternative to tmpnam (which is considered\n** insecure) or if you want the original tmpnam anyway. By default, Lua\n** uses tmpnam except when POSIX is available, where it uses mkstemp.\n*/\n#if defined(loslib_c) || defined(luaall_c)\n\n#if defined(LUA_USE_MKSTEMP)\n#include \n#define LUA_TMPNAMBUFSIZE\t32\n#define lua_tmpnam(b,e)\t{ \\\n\tstrcpy(b, \"/tmp/lua_XXXXXX\"); \\\n\te = mkstemp(b); \\\n\tif (e != -1) close(e); \\\n\te = (e == -1); }\n\n#else\n#define LUA_TMPNAMBUFSIZE\tL_tmpnam\n#define lua_tmpnam(b,e)\t\t{ e = (tmpnam(b) == NULL); }\n#endif\n\n#endif\n\n\n/*\n@@ lua_popen spawns a new process connected to the current one through\n@* the file streams.\n** CHANGE it if you have a way to implement it in your system.\n*/\n#if defined(LUA_USE_POPEN)\n\n#define lua_popen(L,c,m)\t((void)L, fflush(NULL), popen(c,m))\n#define lua_pclose(L,file)\t((void)L, (pclose(file) != -1))\n\n#elif defined(LUA_WIN)\n\n#define lua_popen(L,c,m)\t((void)L, _popen(c,m))\n#define lua_pclose(L,file)\t((void)L, (_pclose(file) != -1))\n\n#else\n\n#define lua_popen(L,c,m)\t((void)((void)c, m), \\\n\t\tluaL_error(L, LUA_QL(\"popen\") \" not supported\"), (FILE*)0)\n#define lua_pclose(L,file)\t\t((void)((void)L, file), 0)\n\n#endif\n\n/*\n@@ LUA_DL_* define which dynamic-library system Lua should use.\n** CHANGE here if Lua has problems choosing the appropriate\n** dynamic-library system for your platform (either Windows' DLL, Mac's\n** dyld, or Unix's dlopen). If your system is some kind of Unix, there\n** is a good chance that it has dlopen, so LUA_DL_DLOPEN will work for\n** it. To use dlopen you also need to adapt the src/Makefile (probably\n** adding -ldl to the linker options), so Lua does not select it\n** automatically. (When you change the makefile to add -ldl, you must\n** also add -DLUA_USE_DLOPEN.)\n** If you do not want any kind of dynamic library, undefine all these\n** options.\n** By default, _WIN32 gets LUA_DL_DLL and MAC OS X gets LUA_DL_DYLD.\n*/\n#if defined(LUA_USE_DLOPEN)\n#define LUA_DL_DLOPEN\n#endif\n\n#if defined(LUA_WIN)\n#define LUA_DL_DLL\n#endif\n\n\n/*\n@@ LUAI_EXTRASPACE allows you to add user-specific data in a lua_State\n@* (the data goes just *before* the lua_State pointer).\n** CHANGE (define) this if you really need that. This value must be\n** a multiple of the maximum alignment required for your machine.\n*/\n#define LUAI_EXTRASPACE\t\t0\n\n\n/*\n@@ luai_userstate* allow user-specific actions on threads.\n** CHANGE them if you defined LUAI_EXTRASPACE and need to do something\n** extra when a thread is created/deleted/resumed/yielded.\n*/\n#define luai_userstateopen(L)\t\t((void)L)\n#define luai_userstateclose(L)\t\t((void)L)\n#define luai_userstatethread(L,L1)\t((void)L)\n#define luai_userstatefree(L)\t\t((void)L)\n#define luai_userstateresume(L,n)\t((void)L)\n#define luai_userstateyield(L,n)\t((void)L)\n\n\n/*\n@@ LUA_INTFRMLEN is the length modifier for integer conversions\n@* in 'string.format'.\n@@ LUA_INTFRM_T is the integer type correspoding to the previous length\n@* modifier.\n** CHANGE them if your system supports long long or does not support long.\n*/\n\n#if defined(LUA_USELONGLONG)\n\n#define LUA_INTFRMLEN\t\t\"ll\"\n#define LUA_INTFRM_T\t\tlong long\n\n#else\n\n#define LUA_INTFRMLEN\t\t\"l\"\n#define LUA_INTFRM_T\t\tlong\n\n#endif\n\n\n\n/* =================================================================== */\n\n/*\n** Local configuration. You can use this space to add your redefinitions\n** without modifying the main part of the file.\n*/\n\n\n\n#endif\n\n" }, { "path": "deps/lua/src/lualib.h", "content": "/*\n** $Id: lualib.h,v 1.36.1.1 2007/12/27 13:02:25 roberto Exp $\n** Lua standard libraries\n** See Copyright Notice in lua.h\n*/\n\n\n#ifndef lualib_h\n#define lualib_h\n\n#include \"lua.h\"\n\n\n/* Key to file-handle type */\n#define LUA_FILEHANDLE\t\t\"FILE*\"\n\n\n#define LUA_COLIBNAME\t\"coroutine\"\nLUALIB_API int (luaopen_base) (lua_State *L);\n\n#define LUA_TABLIBNAME\t\"table\"\nLUALIB_API int (luaopen_table) (lua_State *L);\n\n#define LUA_IOLIBNAME\t\"io\"\nLUALIB_API int (luaopen_io) (lua_State *L);\n\n#define LUA_OSLIBNAME\t\"os\"\nLUALIB_API int (luaopen_os) (lua_State *L);\n\n#define LUA_STRLIBNAME\t\"string\"\nLUALIB_API int (luaopen_string) (lua_State *L);\n\n#define LUA_MATHLIBNAME\t\"math\"\nLUALIB_API int (luaopen_math) (lua_State *L);\n\n#define LUA_DBLIBNAME\t\"debug\"\nLUALIB_API int (luaopen_debug) (lua_State *L);\n\n#define LUA_LOADLIBNAME\t\"package\"\nLUALIB_API int (luaopen_package) (lua_State *L);\n\n\n/* open all previous libraries */\nLUALIB_API void (luaL_openlibs) (lua_State *L); \n\n\n\n#ifndef lua_assert\n#define lua_assert(x)\t((void)0)\n#endif\n\n\n#endif\n" }, { "path": "deps/lua/src/lundump.c", "content": "/*\n** $Id: lundump.c,v 2.7.1.4 2008/04/04 19:51:41 roberto Exp $\n** load precompiled Lua chunks\n** See Copyright Notice in lua.h\n*/\n\n#include \n\n#define lundump_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lmem.h\"\n#include \"lobject.h\"\n#include \"lstring.h\"\n#include \"lundump.h\"\n#include \"lzio.h\"\n\ntypedef struct {\n lua_State* L;\n ZIO* Z;\n Mbuffer* b;\n const char* name;\n} LoadState;\n\n#ifdef LUAC_TRUST_BINARIES\n#define IF(c,s)\n#define error(S,s)\n#else\n#define IF(c,s)\t\tif (c) error(S,s)\n\nstatic void error(LoadState* S, const char* why)\n{\n luaO_pushfstring(S->L,\"%s: %s in precompiled chunk\",S->name,why);\n luaD_throw(S->L,LUA_ERRSYNTAX);\n}\n#endif\n\n#define LoadMem(S,b,n,size)\tLoadBlock(S,b,(n)*(size))\n#define\tLoadByte(S)\t\t(lu_byte)LoadChar(S)\n#define LoadVar(S,x)\t\tLoadMem(S,&x,1,sizeof(x))\n#define LoadVector(S,b,n,size)\tLoadMem(S,b,n,size)\n\nstatic void LoadBlock(LoadState* S, void* b, size_t size)\n{\n size_t r=luaZ_read(S->Z,b,size);\n IF (r!=0, \"unexpected end\");\n}\n\nstatic int LoadChar(LoadState* S)\n{\n char x;\n LoadVar(S,x);\n return x;\n}\n\nstatic int LoadInt(LoadState* S)\n{\n int x;\n LoadVar(S,x);\n IF (x<0, \"bad integer\");\n return x;\n}\n\nstatic lua_Number LoadNumber(LoadState* S)\n{\n lua_Number x;\n LoadVar(S,x);\n return x;\n}\n\nstatic TString* LoadString(LoadState* S)\n{\n size_t size;\n LoadVar(S,size);\n if (size==0)\n return NULL;\n else\n {\n char* s=luaZ_openspace(S->L,S->b,size);\n LoadBlock(S,s,size);\n return luaS_newlstr(S->L,s,size-1);\t\t/* remove trailing '\\0' */\n }\n}\n\nstatic void LoadCode(LoadState* S, Proto* f)\n{\n int n=LoadInt(S);\n f->code=luaM_newvector(S->L,n,Instruction);\n f->sizecode=n;\n LoadVector(S,f->code,n,sizeof(Instruction));\n}\n\nstatic Proto* LoadFunction(LoadState* S, TString* p);\n\nstatic void LoadConstants(LoadState* S, Proto* f)\n{\n int i,n;\n n=LoadInt(S);\n f->k=luaM_newvector(S->L,n,TValue);\n f->sizek=n;\n for (i=0; ik[i]);\n for (i=0; ik[i];\n int t=LoadChar(S);\n switch (t)\n {\n case LUA_TNIL:\n \tsetnilvalue(o);\n\tbreak;\n case LUA_TBOOLEAN:\n \tsetbvalue(o,LoadChar(S)!=0);\n\tbreak;\n case LUA_TNUMBER:\n\tsetnvalue(o,LoadNumber(S));\n\tbreak;\n case LUA_TSTRING:\n\tsetsvalue2n(S->L,o,LoadString(S));\n\tbreak;\n default:\n\terror(S,\"bad constant\");\n\tbreak;\n }\n }\n n=LoadInt(S);\n f->p=luaM_newvector(S->L,n,Proto*);\n f->sizep=n;\n for (i=0; ip[i]=NULL;\n for (i=0; ip[i]=LoadFunction(S,f->source);\n}\n\nstatic void LoadDebug(LoadState* S, Proto* f)\n{\n int i,n;\n n=LoadInt(S);\n f->lineinfo=luaM_newvector(S->L,n,int);\n f->sizelineinfo=n;\n LoadVector(S,f->lineinfo,n,sizeof(int));\n n=LoadInt(S);\n f->locvars=luaM_newvector(S->L,n,LocVar);\n f->sizelocvars=n;\n for (i=0; ilocvars[i].varname=NULL;\n for (i=0; ilocvars[i].varname=LoadString(S);\n f->locvars[i].startpc=LoadInt(S);\n f->locvars[i].endpc=LoadInt(S);\n }\n n=LoadInt(S);\n f->upvalues=luaM_newvector(S->L,n,TString*);\n f->sizeupvalues=n;\n for (i=0; iupvalues[i]=NULL;\n for (i=0; iupvalues[i]=LoadString(S);\n}\n\nstatic Proto* LoadFunction(LoadState* S, TString* p)\n{\n Proto* f;\n if (++S->L->nCcalls > LUAI_MAXCCALLS) error(S,\"code too deep\");\n f=luaF_newproto(S->L);\n setptvalue2s(S->L,S->L->top,f); incr_top(S->L);\n f->source=LoadString(S); if (f->source==NULL) f->source=p;\n f->linedefined=LoadInt(S);\n f->lastlinedefined=LoadInt(S);\n f->nups=LoadByte(S);\n f->numparams=LoadByte(S);\n f->is_vararg=LoadByte(S);\n f->maxstacksize=LoadByte(S);\n LoadCode(S,f);\n LoadConstants(S,f);\n LoadDebug(S,f);\n IF (!luaG_checkcode(f), \"bad code\");\n S->L->top--;\n S->L->nCcalls--;\n return f;\n}\n\nstatic void LoadHeader(LoadState* S)\n{\n char h[LUAC_HEADERSIZE];\n char s[LUAC_HEADERSIZE];\n luaU_header(h);\n LoadBlock(S,s,LUAC_HEADERSIZE);\n IF (memcmp(h,s,LUAC_HEADERSIZE)!=0, \"bad header\");\n}\n\n/*\n** load precompiled chunk\n*/\nProto* luaU_undump (lua_State* L, ZIO* Z, Mbuffer* buff, const char* name)\n{\n LoadState S;\n if (*name=='@' || *name=='=')\n S.name=name+1;\n else if (*name==LUA_SIGNATURE[0])\n S.name=\"binary string\";\n else\n S.name=name;\n S.L=L;\n S.Z=Z;\n S.b=buff;\n LoadHeader(&S);\n return LoadFunction(&S,luaS_newliteral(L,\"=?\"));\n}\n\n/*\n* make header\n*/\nvoid luaU_header (char* h)\n{\n int x=1;\n memcpy(h,LUA_SIGNATURE,sizeof(LUA_SIGNATURE)-1);\n h+=sizeof(LUA_SIGNATURE)-1;\n *h++=(char)LUAC_VERSION;\n *h++=(char)LUAC_FORMAT;\n *h++=(char)*(char*)&x;\t\t\t\t/* endianness */\n *h++=(char)sizeof(int);\n *h++=(char)sizeof(size_t);\n *h++=(char)sizeof(Instruction);\n *h++=(char)sizeof(lua_Number);\n *h++=(char)(((lua_Number)0.5)==0);\t\t/* is lua_Number integral? */\n}\n" }, { "path": "deps/lua/src/lundump.h", "content": "/*\n** $Id: lundump.h,v 1.37.1.1 2007/12/27 13:02:25 roberto Exp $\n** load precompiled Lua chunks\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lundump_h\n#define lundump_h\n\n#include \"lobject.h\"\n#include \"lzio.h\"\n\n/* load one chunk; from lundump.c */\nLUAI_FUNC Proto* luaU_undump (lua_State* L, ZIO* Z, Mbuffer* buff, const char* name);\n\n/* make header; from lundump.c */\nLUAI_FUNC void luaU_header (char* h);\n\n/* dump one chunk; from ldump.c */\nLUAI_FUNC int luaU_dump (lua_State* L, const Proto* f, lua_Writer w, void* data, int strip);\n\n#ifdef luac_c\n/* print one chunk; from print.c */\nLUAI_FUNC void luaU_print (const Proto* f, int full);\n#endif\n\n/* for header of binary files -- this is Lua 5.1 */\n#define LUAC_VERSION\t\t0x51\n\n/* for header of binary files -- this is the official format */\n#define LUAC_FORMAT\t\t0\n\n/* size of header of binary files */\n#define LUAC_HEADERSIZE\t\t12\n\n#endif\n" }, { "path": "deps/lua/src/lvm.c", "content": "/*\n** $Id: lvm.c,v 2.63.1.5 2011/08/17 20:43:11 roberto Exp $\n** Lua virtual machine\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n#include \n#include \n\n#define lvm_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"ldebug.h\"\n#include \"ldo.h\"\n#include \"lfunc.h\"\n#include \"lgc.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lstate.h\"\n#include \"lstring.h\"\n#include \"ltable.h\"\n#include \"ltm.h\"\n#include \"lvm.h\"\n\n\n\n/* limit for table tag-method chains (to avoid loops) */\n#define MAXTAGLOOP\t100\n\n\nconst TValue *luaV_tonumber (const TValue *obj, TValue *n) {\n lua_Number num;\n if (ttisnumber(obj)) return obj;\n if (ttisstring(obj) && luaO_str2d(svalue(obj), &num)) {\n setnvalue(n, num);\n return n;\n }\n else\n return NULL;\n}\n\n\nint luaV_tostring (lua_State *L, StkId obj) {\n if (!ttisnumber(obj))\n return 0;\n else {\n char s[LUAI_MAXNUMBER2STR];\n lua_Number n = nvalue(obj);\n lua_number2str(s, n);\n setsvalue2s(L, obj, luaS_new(L, s));\n return 1;\n }\n}\n\n\nstatic void traceexec (lua_State *L, const Instruction *pc) {\n lu_byte mask = L->hookmask;\n const Instruction *oldpc = L->savedpc;\n L->savedpc = pc;\n if ((mask & LUA_MASKCOUNT) && L->hookcount == 0) {\n resethookcount(L);\n luaD_callhook(L, LUA_HOOKCOUNT, -1);\n }\n if (mask & LUA_MASKLINE) {\n Proto *p = ci_func(L->ci)->l.p;\n int npc = pcRel(pc, p);\n int newline = getline(p, npc);\n /* call linehook when enter a new function, when jump back (loop),\n or when enter a new line */\n if (npc == 0 || pc <= oldpc || newline != getline(p, pcRel(oldpc, p)))\n luaD_callhook(L, LUA_HOOKLINE, newline);\n }\n}\n\n\nstatic void callTMres (lua_State *L, StkId res, const TValue *f,\n const TValue *p1, const TValue *p2) {\n ptrdiff_t result = savestack(L, res);\n setobj2s(L, L->top, f); /* push function */\n setobj2s(L, L->top+1, p1); /* 1st argument */\n setobj2s(L, L->top+2, p2); /* 2nd argument */\n luaD_checkstack(L, 3);\n L->top += 3;\n luaD_call(L, L->top - 3, 1);\n res = restorestack(L, result);\n L->top--;\n setobjs2s(L, res, L->top);\n}\n\n\n\nstatic void callTM (lua_State *L, const TValue *f, const TValue *p1,\n const TValue *p2, const TValue *p3) {\n setobj2s(L, L->top, f); /* push function */\n setobj2s(L, L->top+1, p1); /* 1st argument */\n setobj2s(L, L->top+2, p2); /* 2nd argument */\n setobj2s(L, L->top+3, p3); /* 3th argument */\n luaD_checkstack(L, 4);\n L->top += 4;\n luaD_call(L, L->top - 4, 0);\n}\n\n\nvoid luaV_gettable (lua_State *L, const TValue *t, TValue *key, StkId val) {\n int loop;\n for (loop = 0; loop < MAXTAGLOOP; loop++) {\n const TValue *tm;\n if (ttistable(t)) { /* `t' is a table? */\n Table *h = hvalue(t);\n const TValue *res = luaH_get(h, key); /* do a primitive get */\n if (!ttisnil(res) || /* result is no nil? */\n (tm = fasttm(L, h->metatable, TM_INDEX)) == NULL) { /* or no TM? */\n setobj2s(L, val, res);\n return;\n }\n /* else will try the tag method */\n }\n else if (ttisnil(tm = luaT_gettmbyobj(L, t, TM_INDEX)))\n luaG_typeerror(L, t, \"index\");\n if (ttisfunction(tm)) {\n callTMres(L, val, tm, t, key);\n return;\n }\n t = tm; /* else repeat with `tm' */ \n }\n luaG_runerror(L, \"loop in gettable\");\n}\n\n\nvoid luaV_settable (lua_State *L, const TValue *t, TValue *key, StkId val) {\n int loop;\n TValue temp;\n for (loop = 0; loop < MAXTAGLOOP; loop++) {\n const TValue *tm;\n if (ttistable(t)) { /* `t' is a table? */\n Table *h = hvalue(t);\n TValue *oldval = luaH_set(L, h, key); /* do a primitive set */\n if (!ttisnil(oldval) || /* result is no nil? */\n (tm = fasttm(L, h->metatable, TM_NEWINDEX)) == NULL) { /* or no TM? */\n setobj2t(L, oldval, val);\n h->flags = 0;\n luaC_barriert(L, h, val);\n return;\n }\n /* else will try the tag method */\n }\n else if (ttisnil(tm = luaT_gettmbyobj(L, t, TM_NEWINDEX)))\n luaG_typeerror(L, t, \"index\");\n if (ttisfunction(tm)) {\n callTM(L, tm, t, key, val);\n return;\n }\n /* else repeat with `tm' */\n setobj(L, &temp, tm); /* avoid pointing inside table (may rehash) */\n t = &temp;\n }\n luaG_runerror(L, \"loop in settable\");\n}\n\n\nstatic int call_binTM (lua_State *L, const TValue *p1, const TValue *p2,\n StkId res, TMS event) {\n const TValue *tm = luaT_gettmbyobj(L, p1, event); /* try first operand */\n if (ttisnil(tm))\n tm = luaT_gettmbyobj(L, p2, event); /* try second operand */\n if (ttisnil(tm)) return 0;\n callTMres(L, res, tm, p1, p2);\n return 1;\n}\n\n\nstatic const TValue *get_compTM (lua_State *L, Table *mt1, Table *mt2,\n TMS event) {\n const TValue *tm1 = fasttm(L, mt1, event);\n const TValue *tm2;\n if (tm1 == NULL) return NULL; /* no metamethod */\n if (mt1 == mt2) return tm1; /* same metatables => same metamethods */\n tm2 = fasttm(L, mt2, event);\n if (tm2 == NULL) return NULL; /* no metamethod */\n if (luaO_rawequalObj(tm1, tm2)) /* same metamethods? */\n return tm1;\n return NULL;\n}\n\n\nstatic int call_orderTM (lua_State *L, const TValue *p1, const TValue *p2,\n TMS event) {\n const TValue *tm1 = luaT_gettmbyobj(L, p1, event);\n const TValue *tm2;\n if (ttisnil(tm1)) return -1; /* no metamethod? */\n tm2 = luaT_gettmbyobj(L, p2, event);\n if (!luaO_rawequalObj(tm1, tm2)) /* different metamethods? */\n return -1;\n callTMres(L, L->top, tm1, p1, p2);\n return !l_isfalse(L->top);\n}\n\n\nstatic int l_strcmp (const TString *ls, const TString *rs) {\n const char *l = getstr(ls);\n size_t ll = ls->tsv.len;\n const char *r = getstr(rs);\n size_t lr = rs->tsv.len;\n for (;;) {\n int temp = strcoll(l, r);\n if (temp != 0) return temp;\n else { /* strings are equal up to a `\\0' */\n size_t len = strlen(l); /* index of first `\\0' in both strings */\n if (len == lr) /* r is finished? */\n return (len == ll) ? 0 : 1;\n else if (len == ll) /* l is finished? */\n return -1; /* l is smaller than r (because r is not finished) */\n /* both strings longer than `len'; go on comparing (after the `\\0') */\n len++;\n l += len; ll -= len; r += len; lr -= len;\n }\n }\n}\n\n\nint luaV_lessthan (lua_State *L, const TValue *l, const TValue *r) {\n int res;\n if (ttype(l) != ttype(r))\n return luaG_ordererror(L, l, r);\n else if (ttisnumber(l))\n return luai_numlt(nvalue(l), nvalue(r));\n else if (ttisstring(l))\n return l_strcmp(rawtsvalue(l), rawtsvalue(r)) < 0;\n else if ((res = call_orderTM(L, l, r, TM_LT)) != -1)\n return res;\n return luaG_ordererror(L, l, r);\n}\n\n\nstatic int lessequal (lua_State *L, const TValue *l, const TValue *r) {\n int res;\n if (ttype(l) != ttype(r))\n return luaG_ordererror(L, l, r);\n else if (ttisnumber(l))\n return luai_numle(nvalue(l), nvalue(r));\n else if (ttisstring(l))\n return l_strcmp(rawtsvalue(l), rawtsvalue(r)) <= 0;\n else if ((res = call_orderTM(L, l, r, TM_LE)) != -1) /* first try `le' */\n return res;\n else if ((res = call_orderTM(L, r, l, TM_LT)) != -1) /* else try `lt' */\n return !res;\n return luaG_ordererror(L, l, r);\n}\n\n\nint luaV_equalval (lua_State *L, const TValue *t1, const TValue *t2) {\n const TValue *tm;\n lua_assert(ttype(t1) == ttype(t2));\n switch (ttype(t1)) {\n case LUA_TNIL: return 1;\n case LUA_TNUMBER: return luai_numeq(nvalue(t1), nvalue(t2));\n case LUA_TBOOLEAN: return bvalue(t1) == bvalue(t2); /* true must be 1 !! */\n case LUA_TLIGHTUSERDATA: return pvalue(t1) == pvalue(t2);\n case LUA_TUSERDATA: {\n if (uvalue(t1) == uvalue(t2)) return 1;\n tm = get_compTM(L, uvalue(t1)->metatable, uvalue(t2)->metatable,\n TM_EQ);\n break; /* will try TM */\n }\n case LUA_TTABLE: {\n if (hvalue(t1) == hvalue(t2)) return 1;\n tm = get_compTM(L, hvalue(t1)->metatable, hvalue(t2)->metatable, TM_EQ);\n break; /* will try TM */\n }\n default: return gcvalue(t1) == gcvalue(t2);\n }\n if (tm == NULL) return 0; /* no TM? */\n callTMres(L, L->top, tm, t1, t2); /* call TM */\n return !l_isfalse(L->top);\n}\n\n\nvoid luaV_concat (lua_State *L, int total, int last) {\n do {\n StkId top = L->base + last + 1;\n int n = 2; /* number of elements handled in this pass (at least 2) */\n if (!(ttisstring(top-2) || ttisnumber(top-2)) || !tostring(L, top-1)) {\n if (!call_binTM(L, top-2, top-1, top-2, TM_CONCAT))\n luaG_concaterror(L, top-2, top-1);\n } else if (tsvalue(top-1)->len == 0) /* second op is empty? */\n (void)tostring(L, top - 2); /* result is first op (as string) */\n else {\n /* at least two string values; get as many as possible */\n size_t tl = tsvalue(top-1)->len;\n char *buffer;\n int i;\n /* collect total length */\n for (n = 1; n < total && tostring(L, top-n-1); n++) {\n size_t l = tsvalue(top-n-1)->len;\n if (l >= MAX_SIZET - tl) luaG_runerror(L, \"string length overflow\");\n tl += l;\n }\n buffer = luaZ_openspace(L, &G(L)->buff, tl);\n tl = 0;\n for (i=n; i>0; i--) { /* concat all strings */\n size_t l = tsvalue(top-i)->len;\n memcpy(buffer+tl, svalue(top-i), l);\n tl += l;\n }\n setsvalue2s(L, top-n, luaS_newlstr(L, buffer, tl));\n }\n total -= n-1; /* got `n' strings to create 1 new */\n last -= n-1;\n } while (total > 1); /* repeat until only 1 result left */\n}\n\n\nstatic void Arith (lua_State *L, StkId ra, const TValue *rb,\n const TValue *rc, TMS op) {\n TValue tempb, tempc;\n const TValue *b, *c;\n if ((b = luaV_tonumber(rb, &tempb)) != NULL &&\n (c = luaV_tonumber(rc, &tempc)) != NULL) {\n lua_Number nb = nvalue(b), nc = nvalue(c);\n switch (op) {\n case TM_ADD: setnvalue(ra, luai_numadd(nb, nc)); break;\n case TM_SUB: setnvalue(ra, luai_numsub(nb, nc)); break;\n case TM_MUL: setnvalue(ra, luai_nummul(nb, nc)); break;\n case TM_DIV: setnvalue(ra, luai_numdiv(nb, nc)); break;\n case TM_MOD: setnvalue(ra, luai_nummod(nb, nc)); break;\n case TM_POW: setnvalue(ra, luai_numpow(nb, nc)); break;\n case TM_UNM: setnvalue(ra, luai_numunm(nb)); break;\n default: lua_assert(0); break;\n }\n }\n else if (!call_binTM(L, rb, rc, ra, op))\n luaG_aritherror(L, rb, rc);\n}\n\n\n\n/*\n** some macros for common tasks in `luaV_execute'\n*/\n\n#define runtime_check(L, c)\t{ if (!(c)) break; }\n\n#define RA(i)\t(base+GETARG_A(i))\n/* to be used after possible stack reallocation */\n#define RB(i)\tcheck_exp(getBMode(GET_OPCODE(i)) == OpArgR, base+GETARG_B(i))\n#define RC(i)\tcheck_exp(getCMode(GET_OPCODE(i)) == OpArgR, base+GETARG_C(i))\n#define RKB(i)\tcheck_exp(getBMode(GET_OPCODE(i)) == OpArgK, \\\n\tISK(GETARG_B(i)) ? k+INDEXK(GETARG_B(i)) : base+GETARG_B(i))\n#define RKC(i)\tcheck_exp(getCMode(GET_OPCODE(i)) == OpArgK, \\\n\tISK(GETARG_C(i)) ? k+INDEXK(GETARG_C(i)) : base+GETARG_C(i))\n#define KBx(i)\tcheck_exp(getBMode(GET_OPCODE(i)) == OpArgK, k+GETARG_Bx(i))\n\n\n#define dojump(L,pc,i)\t{(pc) += (i); luai_threadyield(L);}\n\n\n#define Protect(x)\t{ L->savedpc = pc; {x;}; base = L->base; }\n\n\n#define arith_op(op,tm) { \\\n TValue *rb = RKB(i); \\\n TValue *rc = RKC(i); \\\n if (ttisnumber(rb) && ttisnumber(rc)) { \\\n lua_Number nb = nvalue(rb), nc = nvalue(rc); \\\n setnvalue(ra, op(nb, nc)); \\\n } \\\n else \\\n Protect(Arith(L, ra, rb, rc, tm)); \\\n }\n\n\n\nvoid luaV_execute (lua_State *L, int nexeccalls) {\n LClosure *cl;\n StkId base;\n TValue *k;\n const Instruction *pc;\n reentry: /* entry point */\n lua_assert(isLua(L->ci));\n pc = L->savedpc;\n cl = &clvalue(L->ci->func)->l;\n base = L->base;\n k = cl->p->k;\n /* main loop of interpreter */\n for (;;) {\n const Instruction i = *pc++;\n StkId ra;\n if ((L->hookmask & (LUA_MASKLINE | LUA_MASKCOUNT)) &&\n (--L->hookcount == 0 || L->hookmask & LUA_MASKLINE)) {\n traceexec(L, pc);\n if (L->status == LUA_YIELD) { /* did hook yield? */\n L->savedpc = pc - 1;\n return;\n }\n base = L->base;\n }\n /* warning!! several calls may realloc the stack and invalidate `ra' */\n ra = RA(i);\n lua_assert(base == L->base && L->base == L->ci->base);\n lua_assert(base <= L->top && L->top <= L->stack + L->stacksize);\n lua_assert(L->top == L->ci->top || luaG_checkopenop(i));\n switch (GET_OPCODE(i)) {\n case OP_MOVE: {\n setobjs2s(L, ra, RB(i));\n continue;\n }\n case OP_LOADK: {\n setobj2s(L, ra, KBx(i));\n continue;\n }\n case OP_LOADBOOL: {\n setbvalue(ra, GETARG_B(i));\n if (GETARG_C(i)) pc++; /* skip next instruction (if C) */\n continue;\n }\n case OP_LOADNIL: {\n TValue *rb = RB(i);\n do {\n setnilvalue(rb--);\n } while (rb >= ra);\n continue;\n }\n case OP_GETUPVAL: {\n int b = GETARG_B(i);\n setobj2s(L, ra, cl->upvals[b]->v);\n continue;\n }\n case OP_GETGLOBAL: {\n TValue g;\n TValue *rb = KBx(i);\n sethvalue(L, &g, cl->env);\n lua_assert(ttisstring(rb));\n Protect(luaV_gettable(L, &g, rb, ra));\n continue;\n }\n case OP_GETTABLE: {\n Protect(luaV_gettable(L, RB(i), RKC(i), ra));\n continue;\n }\n case OP_SETGLOBAL: {\n TValue g;\n sethvalue(L, &g, cl->env);\n lua_assert(ttisstring(KBx(i)));\n Protect(luaV_settable(L, &g, KBx(i), ra));\n continue;\n }\n case OP_SETUPVAL: {\n UpVal *uv = cl->upvals[GETARG_B(i)];\n setobj(L, uv->v, ra);\n luaC_barrier(L, uv, ra);\n continue;\n }\n case OP_SETTABLE: {\n Protect(luaV_settable(L, ra, RKB(i), RKC(i)));\n continue;\n }\n case OP_NEWTABLE: {\n int b = GETARG_B(i);\n int c = GETARG_C(i);\n sethvalue(L, ra, luaH_new(L, luaO_fb2int(b), luaO_fb2int(c)));\n Protect(luaC_checkGC(L));\n continue;\n }\n case OP_SELF: {\n StkId rb = RB(i);\n setobjs2s(L, ra+1, rb);\n Protect(luaV_gettable(L, rb, RKC(i), ra));\n continue;\n }\n case OP_ADD: {\n arith_op(luai_numadd, TM_ADD);\n continue;\n }\n case OP_SUB: {\n arith_op(luai_numsub, TM_SUB);\n continue;\n }\n case OP_MUL: {\n arith_op(luai_nummul, TM_MUL);\n continue;\n }\n case OP_DIV: {\n arith_op(luai_numdiv, TM_DIV);\n continue;\n }\n case OP_MOD: {\n arith_op(luai_nummod, TM_MOD);\n continue;\n }\n case OP_POW: {\n arith_op(luai_numpow, TM_POW);\n continue;\n }\n case OP_UNM: {\n TValue *rb = RB(i);\n if (ttisnumber(rb)) {\n lua_Number nb = nvalue(rb);\n setnvalue(ra, luai_numunm(nb));\n }\n else {\n Protect(Arith(L, ra, rb, rb, TM_UNM));\n }\n continue;\n }\n case OP_NOT: {\n int res = l_isfalse(RB(i)); /* next assignment may change this value */\n setbvalue(ra, res);\n continue;\n }\n case OP_LEN: {\n const TValue *rb = RB(i);\n switch (ttype(rb)) {\n case LUA_TTABLE: {\n setnvalue(ra, cast_num(luaH_getn(hvalue(rb))));\n break;\n }\n case LUA_TSTRING: {\n setnvalue(ra, cast_num(tsvalue(rb)->len));\n break;\n }\n default: { /* try metamethod */\n Protect(\n if (!call_binTM(L, rb, luaO_nilobject, ra, TM_LEN))\n luaG_typeerror(L, rb, \"get length of\");\n )\n }\n }\n continue;\n }\n case OP_CONCAT: {\n int b = GETARG_B(i);\n int c = GETARG_C(i);\n Protect(luaV_concat(L, c-b+1, c); luaC_checkGC(L));\n setobjs2s(L, RA(i), base+b);\n continue;\n }\n case OP_JMP: {\n dojump(L, pc, GETARG_sBx(i));\n continue;\n }\n case OP_EQ: {\n TValue *rb = RKB(i);\n TValue *rc = RKC(i);\n Protect(\n if (equalobj(L, rb, rc) == GETARG_A(i))\n dojump(L, pc, GETARG_sBx(*pc));\n )\n pc++;\n continue;\n }\n case OP_LT: {\n Protect(\n if (luaV_lessthan(L, RKB(i), RKC(i)) == GETARG_A(i))\n dojump(L, pc, GETARG_sBx(*pc));\n )\n pc++;\n continue;\n }\n case OP_LE: {\n Protect(\n if (lessequal(L, RKB(i), RKC(i)) == GETARG_A(i))\n dojump(L, pc, GETARG_sBx(*pc));\n )\n pc++;\n continue;\n }\n case OP_TEST: {\n if (l_isfalse(ra) != GETARG_C(i))\n dojump(L, pc, GETARG_sBx(*pc));\n pc++;\n continue;\n }\n case OP_TESTSET: {\n TValue *rb = RB(i);\n if (l_isfalse(rb) != GETARG_C(i)) {\n setobjs2s(L, ra, rb);\n dojump(L, pc, GETARG_sBx(*pc));\n }\n pc++;\n continue;\n }\n case OP_CALL: {\n int b = GETARG_B(i);\n int nresults = GETARG_C(i) - 1;\n if (b != 0) L->top = ra+b; /* else previous instruction set top */\n L->savedpc = pc;\n switch (luaD_precall(L, ra, nresults)) {\n case PCRLUA: {\n nexeccalls++;\n goto reentry; /* restart luaV_execute over new Lua function */\n }\n case PCRC: {\n /* it was a C function (`precall' called it); adjust results */\n if (nresults >= 0) L->top = L->ci->top;\n base = L->base;\n continue;\n }\n default: {\n return; /* yield */\n }\n }\n }\n case OP_TAILCALL: {\n int b = GETARG_B(i);\n if (b != 0) L->top = ra+b; /* else previous instruction set top */\n L->savedpc = pc;\n lua_assert(GETARG_C(i) - 1 == LUA_MULTRET);\n switch (luaD_precall(L, ra, LUA_MULTRET)) {\n case PCRLUA: {\n /* tail call: put new frame in place of previous one */\n CallInfo *ci = L->ci - 1; /* previous frame */\n int aux;\n StkId func = ci->func;\n StkId pfunc = (ci+1)->func; /* previous function index */\n if (L->openupval) luaF_close(L, ci->base);\n L->base = ci->base = ci->func + ((ci+1)->base - pfunc);\n for (aux = 0; pfunc+aux < L->top; aux++) /* move frame down */\n setobjs2s(L, func+aux, pfunc+aux);\n ci->top = L->top = func+aux; /* correct top */\n lua_assert(L->top == L->base + clvalue(func)->l.p->maxstacksize);\n ci->savedpc = L->savedpc;\n ci->tailcalls++; /* one more call lost */\n L->ci--; /* remove new frame */\n goto reentry;\n }\n case PCRC: { /* it was a C function (`precall' called it) */\n base = L->base;\n continue;\n }\n default: {\n return; /* yield */\n }\n }\n }\n case OP_RETURN: {\n int b = GETARG_B(i);\n if (b != 0) L->top = ra+b-1;\n if (L->openupval) luaF_close(L, base);\n L->savedpc = pc;\n b = luaD_poscall(L, ra);\n if (--nexeccalls == 0) /* was previous function running `here'? */\n return; /* no: return */\n else { /* yes: continue its execution */\n if (b) L->top = L->ci->top;\n lua_assert(isLua(L->ci));\n lua_assert(GET_OPCODE(*((L->ci)->savedpc - 1)) == OP_CALL);\n goto reentry;\n }\n }\n case OP_FORLOOP: {\n lua_Number step = nvalue(ra+2);\n lua_Number idx = luai_numadd(nvalue(ra), step); /* increment index */\n lua_Number limit = nvalue(ra+1);\n if (luai_numlt(0, step) ? luai_numle(idx, limit)\n : luai_numle(limit, idx)) {\n dojump(L, pc, GETARG_sBx(i)); /* jump back */\n setnvalue(ra, idx); /* update internal index... */\n setnvalue(ra+3, idx); /* ...and external index */\n }\n continue;\n }\n case OP_FORPREP: {\n const TValue *init = ra;\n const TValue *plimit = ra+1;\n const TValue *pstep = ra+2;\n L->savedpc = pc; /* next steps may throw errors */\n if (!tonumber(init, ra))\n luaG_runerror(L, LUA_QL(\"for\") \" initial value must be a number\");\n else if (!tonumber(plimit, ra+1))\n luaG_runerror(L, LUA_QL(\"for\") \" limit must be a number\");\n else if (!tonumber(pstep, ra+2))\n luaG_runerror(L, LUA_QL(\"for\") \" step must be a number\");\n setnvalue(ra, luai_numsub(nvalue(ra), nvalue(pstep)));\n dojump(L, pc, GETARG_sBx(i));\n continue;\n }\n case OP_TFORLOOP: {\n StkId cb = ra + 3; /* call base */\n setobjs2s(L, cb+2, ra+2);\n setobjs2s(L, cb+1, ra+1);\n setobjs2s(L, cb, ra);\n L->top = cb+3; /* func. + 2 args (state and index) */\n Protect(luaD_call(L, cb, GETARG_C(i)));\n L->top = L->ci->top;\n cb = RA(i) + 3; /* previous call may change the stack */\n if (!ttisnil(cb)) { /* continue loop? */\n setobjs2s(L, cb-1, cb); /* save control variable */\n dojump(L, pc, GETARG_sBx(*pc)); /* jump back */\n }\n pc++;\n continue;\n }\n case OP_SETLIST: {\n int n = GETARG_B(i);\n int c = GETARG_C(i);\n int last;\n Table *h;\n if (n == 0) {\n n = cast_int(L->top - ra) - 1;\n L->top = L->ci->top;\n }\n if (c == 0) c = cast_int(*pc++);\n runtime_check(L, ttistable(ra));\n h = hvalue(ra);\n last = ((c-1)*LFIELDS_PER_FLUSH) + n;\n if (last > h->sizearray) /* needs more space? */\n luaH_resizearray(L, h, last); /* pre-alloc it at once */\n for (; n > 0; n--) {\n TValue *val = ra+n;\n setobj2t(L, luaH_setnum(L, h, last--), val);\n luaC_barriert(L, h, val);\n }\n continue;\n }\n case OP_CLOSE: {\n luaF_close(L, ra);\n continue;\n }\n case OP_CLOSURE: {\n Proto *p;\n Closure *ncl;\n int nup, j;\n p = cl->p->p[GETARG_Bx(i)];\n nup = p->nups;\n ncl = luaF_newLclosure(L, nup, cl->env);\n ncl->l.p = p;\n for (j=0; jl.upvals[j] = cl->upvals[GETARG_B(*pc)];\n else {\n lua_assert(GET_OPCODE(*pc) == OP_MOVE);\n ncl->l.upvals[j] = luaF_findupval(L, base + GETARG_B(*pc));\n }\n }\n setclvalue(L, ra, ncl);\n Protect(luaC_checkGC(L));\n continue;\n }\n case OP_VARARG: {\n int b = GETARG_B(i) - 1;\n int j;\n CallInfo *ci = L->ci;\n int n = cast_int(ci->base - ci->func) - cl->p->numparams - 1;\n if (b == LUA_MULTRET) {\n Protect(luaD_checkstack(L, n));\n ra = RA(i); /* previous call may change the stack */\n b = n;\n L->top = ra + n;\n }\n for (j = 0; j < b; j++) {\n if (j < n) {\n setobjs2s(L, ra + j, ci->base - n + j);\n }\n else {\n setnilvalue(ra + j);\n }\n }\n continue;\n }\n }\n }\n}\n\n" }, { "path": "deps/lua/src/lvm.h", "content": "/*\n** $Id: lvm.h,v 2.5.1.1 2007/12/27 13:02:25 roberto Exp $\n** Lua virtual machine\n** See Copyright Notice in lua.h\n*/\n\n#ifndef lvm_h\n#define lvm_h\n\n\n#include \"ldo.h\"\n#include \"lobject.h\"\n#include \"ltm.h\"\n\n\n#define tostring(L,o) ((ttype(o) == LUA_TSTRING) || (luaV_tostring(L, o)))\n\n#define tonumber(o,n)\t(ttype(o) == LUA_TNUMBER || \\\n (((o) = luaV_tonumber(o,n)) != NULL))\n\n#define equalobj(L,o1,o2) \\\n\t(ttype(o1) == ttype(o2) && luaV_equalval(L, o1, o2))\n\n\nLUAI_FUNC int luaV_lessthan (lua_State *L, const TValue *l, const TValue *r);\nLUAI_FUNC int luaV_equalval (lua_State *L, const TValue *t1, const TValue *t2);\nLUAI_FUNC const TValue *luaV_tonumber (const TValue *obj, TValue *n);\nLUAI_FUNC int luaV_tostring (lua_State *L, StkId obj);\nLUAI_FUNC void luaV_gettable (lua_State *L, const TValue *t, TValue *key,\n StkId val);\nLUAI_FUNC void luaV_settable (lua_State *L, const TValue *t, TValue *key,\n StkId val);\nLUAI_FUNC void luaV_execute (lua_State *L, int nexeccalls);\nLUAI_FUNC void luaV_concat (lua_State *L, int total, int last);\n\n#endif\n" }, { "path": "deps/lua/src/lzio.c", "content": "/*\n** $Id: lzio.c,v 1.31.1.1 2007/12/27 13:02:25 roberto Exp $\n** a generic input stream interface\n** See Copyright Notice in lua.h\n*/\n\n\n#include \n\n#define lzio_c\n#define LUA_CORE\n\n#include \"lua.h\"\n\n#include \"llimits.h\"\n#include \"lmem.h\"\n#include \"lstate.h\"\n#include \"lzio.h\"\n\n\nint luaZ_fill (ZIO *z) {\n size_t size;\n lua_State *L = z->L;\n const char *buff;\n lua_unlock(L);\n buff = z->reader(L, z->data, &size);\n lua_lock(L);\n if (buff == NULL || size == 0) return EOZ;\n z->n = size - 1;\n z->p = buff;\n return char2int(*(z->p++));\n}\n\n\nint luaZ_lookahead (ZIO *z) {\n if (z->n == 0) {\n if (luaZ_fill(z) == EOZ)\n return EOZ;\n else {\n z->n++; /* luaZ_fill removed first byte; put back it */\n z->p--;\n }\n }\n return char2int(*z->p);\n}\n\n\nvoid luaZ_init (lua_State *L, ZIO *z, lua_Reader reader, void *data) {\n z->L = L;\n z->reader = reader;\n z->data = data;\n z->n = 0;\n z->p = NULL;\n}\n\n\n/* --------------------------------------------------------------- read --- */\nsize_t luaZ_read (ZIO *z, void *b, size_t n) {\n while (n) {\n size_t m;\n if (luaZ_lookahead(z) == EOZ)\n return n; /* return number of missing bytes */\n m = (n <= z->n) ? n : z->n; /* min. between n and z->n */\n memcpy(b, z->p, m);\n z->n -= m;\n z->p += m;\n b = (char *)b + m;\n n -= m;\n }\n return 0;\n}\n\n/* ------------------------------------------------------------------------ */\nchar *luaZ_openspace (lua_State *L, Mbuffer *buff, size_t n) {\n if (n > buff->buffsize) {\n if (n < LUA_MINBUFFER) n = LUA_MINBUFFER;\n luaZ_resizebuffer(L, buff, n);\n }\n return buff->buffer;\n}\n\n\n" }, { "path": "deps/lua/src/lzio.h", "content": "/*\n** $Id: lzio.h,v 1.21.1.1 2007/12/27 13:02:25 roberto Exp $\n** Buffered streams\n** See Copyright Notice in lua.h\n*/\n\n\n#ifndef lzio_h\n#define lzio_h\n\n#include \"lua.h\"\n\n#include \"lmem.h\"\n\n\n#define EOZ\t(-1)\t\t\t/* end of stream */\n\ntypedef struct Zio ZIO;\n\n#define char2int(c)\tcast(int, cast(unsigned char, (c)))\n\n#define zgetc(z) (((z)->n--)>0 ? char2int(*(z)->p++) : luaZ_fill(z))\n\ntypedef struct Mbuffer {\n char *buffer;\n size_t n;\n size_t buffsize;\n} Mbuffer;\n\n#define luaZ_initbuffer(L, buff) ((buff)->buffer = NULL, (buff)->buffsize = 0)\n\n#define luaZ_buffer(buff)\t((buff)->buffer)\n#define luaZ_sizebuffer(buff)\t((buff)->buffsize)\n#define luaZ_bufflen(buff)\t((buff)->n)\n\n#define luaZ_resetbuffer(buff) ((buff)->n = 0)\n\n\n#define luaZ_resizebuffer(L, buff, size) \\\n\t(luaM_reallocvector(L, (buff)->buffer, (buff)->buffsize, size, char), \\\n\t(buff)->buffsize = size)\n\n#define luaZ_freebuffer(L, buff)\tluaZ_resizebuffer(L, buff, 0)\n\n\nLUAI_FUNC char *luaZ_openspace (lua_State *L, Mbuffer *buff, size_t n);\nLUAI_FUNC void luaZ_init (lua_State *L, ZIO *z, lua_Reader reader,\n void *data);\nLUAI_FUNC size_t luaZ_read (ZIO* z, void* b, size_t n);\t/* read next n bytes */\nLUAI_FUNC int luaZ_lookahead (ZIO *z);\n\n\n\n/* --------- Private Part ------------------ */\n\nstruct Zio {\n size_t n;\t\t\t/* bytes still unread */\n const char *p;\t\t/* current position in buffer */\n lua_Reader reader;\n void* data;\t\t\t/* additional data */\n lua_State *L;\t\t\t/* Lua state (for reader) */\n};\n\n\nLUAI_FUNC int luaZ_fill (ZIO *z);\n\n#endif\n" }, { "path": "deps/lua/src/print.c", "content": "/*\n** $Id: print.c,v 1.55a 2006/05/31 13:30:05 lhf Exp $\n** print bytecodes\n** See Copyright Notice in lua.h\n*/\n\n#include \n#include \n\n#define luac_c\n#define LUA_CORE\n\n#include \"ldebug.h\"\n#include \"lobject.h\"\n#include \"lopcodes.h\"\n#include \"lundump.h\"\n\n#define PrintFunction\tluaU_print\n\n#define Sizeof(x)\t((int)sizeof(x))\n#define VOID(p)\t\t((const void*)(p))\n\nstatic void PrintString(const TString* ts)\n{\n const char* s=getstr(ts);\n size_t i,n=ts->tsv.len;\n putchar('\"');\n for (i=0; ik[i];\n switch (ttype(o))\n {\n case LUA_TNIL:\n\tprintf(\"nil\");\n\tbreak;\n case LUA_TBOOLEAN:\n\tprintf(bvalue(o) ? \"true\" : \"false\");\n\tbreak;\n case LUA_TNUMBER:\n\tprintf(LUA_NUMBER_FMT,nvalue(o));\n\tbreak;\n case LUA_TSTRING:\n\tPrintString(rawtsvalue(o));\n\tbreak;\n default:\t\t\t\t/* cannot happen */\n\tprintf(\"? type=%d\",ttype(o));\n\tbreak;\n }\n}\n\nstatic void PrintCode(const Proto* f)\n{\n const Instruction* code=f->code;\n int pc,n=f->sizecode;\n for (pc=0; pc0) printf(\"[%d]\\t\",line); else printf(\"[-]\\t\");\n printf(\"%-9s\\t\",luaP_opnames[o]);\n switch (getOpMode(o))\n {\n case iABC:\n printf(\"%d\",a);\n if (getBMode(o)!=OpArgN) printf(\" %d\",ISK(b) ? (-1-INDEXK(b)) : b);\n if (getCMode(o)!=OpArgN) printf(\" %d\",ISK(c) ? (-1-INDEXK(c)) : c);\n break;\n case iABx:\n if (getBMode(o)==OpArgK) printf(\"%d %d\",a,-1-bx); else printf(\"%d %d\",a,bx);\n break;\n case iAsBx:\n if (o==OP_JMP) printf(\"%d\",sbx); else printf(\"%d %d\",a,sbx);\n break;\n }\n switch (o)\n {\n case OP_LOADK:\n printf(\"\\t; \"); PrintConstant(f,bx);\n break;\n case OP_GETUPVAL:\n case OP_SETUPVAL:\n printf(\"\\t; %s\", (f->sizeupvalues>0) ? getstr(f->upvalues[b]) : \"-\");\n break;\n case OP_GETGLOBAL:\n case OP_SETGLOBAL:\n printf(\"\\t; %s\",svalue(&f->k[bx]));\n break;\n case OP_GETTABLE:\n case OP_SELF:\n if (ISK(c)) { printf(\"\\t; \"); PrintConstant(f,INDEXK(c)); }\n break;\n case OP_SETTABLE:\n case OP_ADD:\n case OP_SUB:\n case OP_MUL:\n case OP_DIV:\n case OP_POW:\n case OP_EQ:\n case OP_LT:\n case OP_LE:\n if (ISK(b) || ISK(c))\n {\n printf(\"\\t; \");\n if (ISK(b)) PrintConstant(f,INDEXK(b)); else printf(\"-\");\n printf(\" \");\n if (ISK(c)) PrintConstant(f,INDEXK(c)); else printf(\"-\");\n }\n break;\n case OP_JMP:\n case OP_FORLOOP:\n case OP_FORPREP:\n printf(\"\\t; to %d\",sbx+pc+2);\n break;\n case OP_CLOSURE:\n printf(\"\\t; %p\",VOID(f->p[bx]));\n break;\n case OP_SETLIST:\n if (c==0) printf(\"\\t; %d\",(int)code[++pc]);\n else printf(\"\\t; %d\",c);\n break;\n default:\n break;\n }\n printf(\"\\n\");\n }\n}\n\n#define SS(x)\t(x==1)?\"\":\"s\"\n#define S(x)\tx,SS(x)\n\nstatic void PrintHeader(const Proto* f)\n{\n const char* s=getstr(f->source);\n if (*s=='@' || *s=='=')\n s++;\n else if (*s==LUA_SIGNATURE[0])\n s=\"(bstring)\";\n else\n s=\"(string)\";\n printf(\"\\n%s <%s:%d,%d> (%d instruction%s, %d bytes at %p)\\n\",\n \t(f->linedefined==0)?\"main\":\"function\",s,\n\tf->linedefined,f->lastlinedefined,\n\tS(f->sizecode),f->sizecode*Sizeof(Instruction),VOID(f));\n printf(\"%d%s param%s, %d slot%s, %d upvalue%s, \",\n\tf->numparams,f->is_vararg?\"+\":\"\",SS(f->numparams),\n\tS(f->maxstacksize),S(f->nups));\n printf(\"%d local%s, %d constant%s, %d function%s\\n\",\n\tS(f->sizelocvars),S(f->sizek),S(f->sizep));\n}\n\nstatic void PrintConstants(const Proto* f)\n{\n int i,n=f->sizek;\n printf(\"constants (%d) for %p:\\n\",n,VOID(f));\n for (i=0; isizelocvars;\n printf(\"locals (%d) for %p:\\n\",n,VOID(f));\n for (i=0; ilocvars[i].varname),f->locvars[i].startpc+1,f->locvars[i].endpc+1);\n }\n}\n\nstatic void PrintUpvalues(const Proto* f)\n{\n int i,n=f->sizeupvalues;\n printf(\"upvalues (%d) for %p:\\n\",n,VOID(f));\n if (f->upvalues==NULL) return;\n for (i=0; iupvalues[i]));\n }\n}\n\nvoid PrintFunction(const Proto* f, int full)\n{\n int i,n=f->sizep;\n PrintHeader(f);\n PrintCode(f);\n if (full)\n {\n PrintConstants(f);\n PrintLocals(f);\n PrintUpvalues(f);\n }\n for (i=0; ip[i],full);\n}\n" }, { "path": "deps/lua/src/strbuf.c", "content": "/* strbuf - string buffer routines\n *\n * Copyright (c) 2010-2011 Mark Pulford \n *\n * Permission is hereby granted, free of charge, to any person obtaining\n * a copy of this software and associated documentation files (the\n * \"Software\"), to deal in the Software without restriction, including\n * without limitation the rights to use, copy, modify, merge, publish,\n * distribute, sublicense, and/or sell copies of the Software, and to\n * permit persons to whom the Software is furnished to do so, subject to\n * the following conditions:\n *\n * The above copyright notice and this permission notice shall be\n * included in all copies or substantial portions of the Software.\n *\n * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n */\n\n#include \n#include \n#include \n#include \n\n#include \"strbuf.h\"\n\nvoid die(const char *fmt, ...)\n{\n va_list arg;\n\n va_start(arg, fmt);\n vfprintf(stderr, fmt, arg);\n va_end(arg);\n fprintf(stderr, \"\\n\");\n\n exit(-1);\n}\n\nvoid strbuf_init(strbuf_t *s, int len)\n{\n int size;\n\n if (len <= 0)\n size = STRBUF_DEFAULT_SIZE;\n else\n size = len + 1; /* \\0 terminator */\n\n s->buf = NULL;\n s->size = size;\n s->length = 0;\n s->increment = STRBUF_DEFAULT_INCREMENT;\n s->dynamic = 0;\n s->reallocs = 0;\n s->debug = 0;\n\n s->buf = malloc(size);\n if (!s->buf)\n die(\"Out of memory\");\n\n strbuf_ensure_null(s);\n}\n\nstrbuf_t *strbuf_new(int len)\n{\n strbuf_t *s;\n\n s = malloc(sizeof(strbuf_t));\n if (!s)\n die(\"Out of memory\");\n\n strbuf_init(s, len);\n\n /* Dynamic strbuf allocation / deallocation */\n s->dynamic = 1;\n\n return s;\n}\n\nvoid strbuf_set_increment(strbuf_t *s, int increment)\n{\n /* Increment > 0: Linear buffer growth rate\n * Increment < -1: Exponential buffer growth rate */\n if (increment == 0 || increment == -1)\n die(\"BUG: Invalid string increment\");\n\n s->increment = increment;\n}\n\nstatic inline void debug_stats(strbuf_t *s)\n{\n if (s->debug) {\n fprintf(stderr, \"strbuf(%lx) reallocs: %d, length: %d, size: %d\\n\",\n (long)s, s->reallocs, s->length, s->size);\n }\n}\n\n/* If strbuf_t has not been dynamically allocated, strbuf_free() can\n * be called any number of times strbuf_init() */\nvoid strbuf_free(strbuf_t *s)\n{\n debug_stats(s);\n\n if (s->buf) {\n free(s->buf);\n s->buf = NULL;\n }\n if (s->dynamic)\n free(s);\n}\n\nchar *strbuf_free_to_string(strbuf_t *s, int *len)\n{\n char *buf;\n\n debug_stats(s);\n\n strbuf_ensure_null(s);\n\n buf = s->buf;\n if (len)\n *len = s->length;\n\n if (s->dynamic)\n free(s);\n\n return buf;\n}\n\nstatic int calculate_new_size(strbuf_t *s, int len)\n{\n int reqsize, newsize;\n\n if (len <= 0)\n die(\"BUG: Invalid strbuf length requested\");\n\n /* Ensure there is room for optional NULL termination */\n reqsize = len + 1;\n\n /* If the user has requested to shrink the buffer, do it exactly */\n if (s->size > reqsize)\n return reqsize;\n\n newsize = s->size;\n if (s->increment < 0) {\n /* Exponential sizing */\n while (newsize < reqsize)\n newsize *= -s->increment;\n } else {\n /* Linear sizing */\n newsize = ((newsize + s->increment - 1) / s->increment) * s->increment;\n }\n\n return newsize;\n}\n\n\n/* Ensure strbuf can handle a string length bytes long (ignoring NULL\n * optional termination). */\nvoid strbuf_resize(strbuf_t *s, int len)\n{\n int newsize;\n\n newsize = calculate_new_size(s, len);\n\n if (s->debug > 1) {\n fprintf(stderr, \"strbuf(%lx) resize: %d => %d\\n\",\n (long)s, s->size, newsize);\n }\n\n s->size = newsize;\n s->buf = realloc(s->buf, s->size);\n if (!s->buf)\n die(\"Out of memory\");\n s->reallocs++;\n}\n\nvoid strbuf_append_string(strbuf_t *s, const char *str)\n{\n int space, i;\n\n space = strbuf_empty_length(s);\n\n for (i = 0; str[i]; i++) {\n if (space < 1) {\n strbuf_resize(s, s->length + 1);\n space = strbuf_empty_length(s);\n }\n\n s->buf[s->length] = str[i];\n s->length++;\n space--;\n }\n}\n\n/* strbuf_append_fmt() should only be used when an upper bound\n * is known for the output string. */\nvoid strbuf_append_fmt(strbuf_t *s, int len, const char *fmt, ...)\n{\n va_list arg;\n int fmt_len;\n\n strbuf_ensure_empty_length(s, len);\n\n va_start(arg, fmt);\n fmt_len = vsnprintf(s->buf + s->length, len, fmt, arg);\n va_end(arg);\n\n if (fmt_len < 0)\n die(\"BUG: Unable to convert number\"); /* This should never happen.. */\n\n s->length += fmt_len;\n}\n\n/* strbuf_append_fmt_retry() can be used when the there is no known\n * upper bound for the output string. */\nvoid strbuf_append_fmt_retry(strbuf_t *s, const char *fmt, ...)\n{\n va_list arg;\n int fmt_len, try;\n int empty_len;\n\n /* If the first attempt to append fails, resize the buffer appropriately\n * and try again */\n for (try = 0; ; try++) {\n va_start(arg, fmt);\n /* Append the new formatted string */\n /* fmt_len is the length of the string required, excluding the\n * trailing NULL */\n empty_len = strbuf_empty_length(s);\n /* Add 1 since there is also space to store the terminating NULL. */\n fmt_len = vsnprintf(s->buf + s->length, empty_len + 1, fmt, arg);\n va_end(arg);\n\n if (fmt_len <= empty_len)\n break; /* SUCCESS */\n if (try > 0)\n die(\"BUG: length of formatted string changed\");\n\n strbuf_resize(s, s->length + fmt_len);\n }\n\n s->length += fmt_len;\n}\n\n/* vi:ai et sw=4 ts=4:\n */\n" }, { "path": "deps/lua/src/strbuf.h", "content": "/* strbuf - String buffer routines\n *\n * Copyright (c) 2010-2011 Mark Pulford \n *\n * Permission is hereby granted, free of charge, to any person obtaining\n * a copy of this software and associated documentation files (the\n * \"Software\"), to deal in the Software without restriction, including\n * without limitation the rights to use, copy, modify, merge, publish,\n * distribute, sublicense, and/or sell copies of the Software, and to\n * permit persons to whom the Software is furnished to do so, subject to\n * the following conditions:\n *\n * The above copyright notice and this permission notice shall be\n * included in all copies or substantial portions of the Software.\n *\n * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n */\n\n#include \n#include \n\n/* Size: Total bytes allocated to *buf\n * Length: String length, excluding optional NULL terminator.\n * Increment: Allocation increments when resizing the string buffer.\n * Dynamic: True if created via strbuf_new()\n */\n\ntypedef struct {\n char *buf;\n int size;\n int length;\n int increment;\n int dynamic;\n int reallocs;\n int debug;\n} strbuf_t;\n\n#ifndef STRBUF_DEFAULT_SIZE\n#define STRBUF_DEFAULT_SIZE 1023\n#endif\n#ifndef STRBUF_DEFAULT_INCREMENT\n#define STRBUF_DEFAULT_INCREMENT -2\n#endif\n\n/* Initialise */\nextern strbuf_t *strbuf_new(int len);\nextern void strbuf_init(strbuf_t *s, int len);\nextern void strbuf_set_increment(strbuf_t *s, int increment);\n\n/* Release */\nextern void strbuf_free(strbuf_t *s);\nextern char *strbuf_free_to_string(strbuf_t *s, int *len);\n\n/* Management */\nextern void strbuf_resize(strbuf_t *s, int len);\nstatic int strbuf_empty_length(strbuf_t *s);\nstatic int strbuf_length(strbuf_t *s);\nstatic char *strbuf_string(strbuf_t *s, int *len);\nstatic void strbuf_ensure_empty_length(strbuf_t *s, int len);\n\n/* Update */\nextern void strbuf_append_fmt(strbuf_t *s, int len, const char *fmt, ...);\nextern void strbuf_append_fmt_retry(strbuf_t *s, const char *format, ...);\nstatic void strbuf_append_mem(strbuf_t *s, const char *c, int len);\nextern void strbuf_append_string(strbuf_t *s, const char *str);\nstatic void strbuf_append_char(strbuf_t *s, const char c);\nstatic void strbuf_ensure_null(strbuf_t *s);\n\n/* Reset string for before use */\nstatic inline void strbuf_reset(strbuf_t *s)\n{\n s->length = 0;\n}\n\nstatic inline int strbuf_allocated(strbuf_t *s)\n{\n return s->buf != NULL;\n}\n\n/* Return bytes remaining in the string buffer\n * Ensure there is space for a NULL terminator. */\nstatic inline int strbuf_empty_length(strbuf_t *s)\n{\n return s->size - s->length - 1;\n}\n\nstatic inline void strbuf_ensure_empty_length(strbuf_t *s, int len)\n{\n if (len > strbuf_empty_length(s))\n strbuf_resize(s, s->length + len);\n}\n\nstatic inline int strbuf_length(strbuf_t *s)\n{\n return s->length;\n}\n\nstatic inline void strbuf_append_char(strbuf_t *s, const char c)\n{\n strbuf_ensure_empty_length(s, 1);\n s->buf[s->length++] = c;\n}\n\nstatic inline void strbuf_append_char_unsafe(strbuf_t *s, const char c)\n{\n s->buf[s->length++] = c;\n}\n\nstatic inline void strbuf_append_mem(strbuf_t *s, const char *c, int len)\n{\n strbuf_ensure_empty_length(s, len);\n memcpy(s->buf + s->length, c, len);\n s->length += len;\n}\n\nstatic inline void strbuf_append_mem_unsafe(strbuf_t *s, const char *c, int len)\n{\n memcpy(s->buf + s->length, c, len);\n s->length += len;\n}\n\nstatic inline void strbuf_ensure_null(strbuf_t *s)\n{\n s->buf[s->length] = 0;\n}\n\nstatic inline char *strbuf_string(strbuf_t *s, int *len)\n{\n if (len)\n *len = s->length;\n\n return s->buf;\n}\n\n/* vi:ai et sw=4 ts=4:\n */\n" }, { "path": "deps/lua/test/README", "content": "These are simple tests for Lua. Some of them contain useful code.\nThey are meant to be run to make sure Lua is built correctly and also\nto be read, to see how Lua programs look.\n\nHere is a one-line summary of each program:\n\n bisect.lua\t\tbisection method for solving non-linear equations\n cf.lua\t\ttemperature conversion table (celsius to farenheit)\n echo.lua echo command line arguments\n env.lua environment variables as automatic global variables\n factorial.lua\tfactorial without recursion\n fib.lua\t\tfibonacci function with cache\n fibfor.lua\t\tfibonacci numbers with coroutines and generators\n globals.lua\t\treport global variable usage\n hello.lua\t\tthe first program in every language\n life.lua\t\tConway's Game of Life\n luac.lua\t \tbare-bones luac\n printf.lua\t\tan implementation of printf\n readonly.lua\t\tmake global variables readonly\n sieve.lua\t\tthe sieve of of Eratosthenes programmed with coroutines\n sort.lua\t\ttwo implementations of a sort function\n table.lua\t\tmake table, grouping all data for the same item\n trace-calls.lua\ttrace calls\n trace-globals.lua\ttrace assigments to global variables\n xd.lua\t\thex dump\n\n" }, { "path": "deps/lua/test/bisect.lua", "content": "-- bisection method for solving non-linear equations\n\ndelta=1e-6\t-- tolerance\n\nfunction bisect(f,a,b,fa,fb)\n local c=(a+b)/2\n io.write(n,\" c=\",c,\" a=\",a,\" b=\",b,\"\\n\")\n if c==a or c==b or math.abs(a-b) posted to lua-l\n-- modified to use ANSI terminal escape sequences\n-- modified to use for instead of while\n\nlocal write=io.write\n\nALIVE=\"\"\tDEAD=\"\"\nALIVE=\"O\"\tDEAD=\"-\"\n\nfunction delay() -- NOTE: SYSTEM-DEPENDENT, adjust as necessary\n for i=1,10000 do end\n -- local i=os.clock()+1 while(os.clock() 0 do\n local xm1,x,xp1,xi=self.w-1,self.w,1,self.w\n while xi > 0 do\n local sum = self[ym1][xm1] + self[ym1][x] + self[ym1][xp1] +\n self[y][xm1] + self[y][xp1] +\n self[yp1][xm1] + self[yp1][x] + self[yp1][xp1]\n next[y][x] = ((sum==2) and self[y][x]) or ((sum==3) and 1) or 0\n xm1,x,xp1,xi = x,xp1,xp1+1,xi-1\n end\n ym1,y,yp1,yi = y,yp1,yp1+1,yi-1\n end\nend\n\n-- output the array to screen\nfunction _CELLS:draw()\n local out=\"\" -- accumulate to reduce flicker\n for y=1,self.h do\n for x=1,self.w do\n out=out..(((self[y][x]>0) and ALIVE) or DEAD)\n end\n out=out..\"\\n\"\n end\n write(out)\nend\n\n-- constructor\nfunction CELLS(w,h)\n local c = ARRAY2D(w,h)\n c.spawn = _CELLS.spawn\n c.evolve = _CELLS.evolve\n c.draw = _CELLS.draw\n return c\nend\n\n--\n-- shapes suitable for use with spawn() above\n--\nHEART = { 1,0,1,1,0,1,1,1,1; w=3,h=3 }\nGLIDER = { 0,0,1,1,0,1,0,1,1; w=3,h=3 }\nEXPLODE = { 0,1,0,1,1,1,1,0,1,0,1,0; w=3,h=4 }\nFISH = { 0,1,1,1,1,1,0,0,0,1,0,0,0,0,1,1,0,0,1,0; w=5,h=4 }\nBUTTERFLY = { 1,0,0,0,1,0,1,1,1,0,1,0,0,0,1,1,0,1,0,1,1,0,0,0,1; w=5,h=5 }\n\n-- the main routine\nfunction LIFE(w,h)\n -- create two arrays\n local thisgen = CELLS(w,h)\n local nextgen = CELLS(w,h)\n\n -- create some life\n -- about 1000 generations of fun, then a glider steady-state\n thisgen:spawn(GLIDER,5,4)\n thisgen:spawn(EXPLODE,25,10)\n thisgen:spawn(FISH,4,12)\n\n -- run until break\n local gen=1\n write(\"\\027[2J\")\t-- ANSI clear screen\n while 1 do\n thisgen:evolve(nextgen)\n thisgen,nextgen = nextgen,thisgen\n write(\"\\027[H\")\t-- ANSI home cursor\n thisgen:draw()\n write(\"Life - generation \",gen,\"\\n\")\n gen=gen+1\n if gen>2000 then break end\n --delay()\t\t-- no delay\n end\nend\n\nLIFE(40,20)\n" }, { "path": "deps/lua/test/luac.lua", "content": "-- bare-bones luac in Lua\n-- usage: lua luac.lua file.lua\n\nassert(arg[1]~=nil and arg[2]==nil,\"usage: lua luac.lua file.lua\")\nf=assert(io.open(\"luac.out\",\"wb\"))\nassert(f:write(string.dump(assert(loadfile(arg[1])))))\nassert(f:close())\n" }, { "path": "deps/lua/test/printf.lua", "content": "-- an implementation of printf\n\nfunction printf(...)\n io.write(string.format(...))\nend\n\nprintf(\"Hello %s from %s on %s\\n\",os.getenv\"USER\" or \"there\",_VERSION,os.date())\n" }, { "path": "deps/lua/test/readonly.lua", "content": "-- make global variables readonly\n\nlocal f=function (t,i) error(\"cannot redefine global variable `\"..i..\"'\",2) end\nlocal g={}\nlocal G=getfenv()\nsetmetatable(g,{__index=G,__newindex=f})\nsetfenv(1,g)\n\n-- an example\nrawset(g,\"x\",3)\nx=2\ny=1\t-- cannot redefine `y'\n" }, { "path": "deps/lua/test/sieve.lua", "content": "-- the sieve of of Eratosthenes programmed with coroutines\n-- typical usage: lua -e N=1000 sieve.lua | column\n\n-- generate all the numbers from 2 to n\nfunction gen (n)\n return coroutine.wrap(function ()\n for i=2,n do coroutine.yield(i) end\n end)\nend\n\n-- filter the numbers generated by `g', removing multiples of `p'\nfunction filter (p, g)\n return coroutine.wrap(function ()\n while 1 do\n local n = g()\n if n == nil then return end\n if math.mod(n, p) ~= 0 then coroutine.yield(n) end\n end\n end)\nend\n\nN=N or 1000\t\t-- from command line\nx = gen(N)\t\t-- generate primes up to N\nwhile 1 do\n local n = x()\t\t-- pick a number until done\n if n == nil then break end\n print(n)\t\t-- must be a prime number\n x = filter(n, x)\t-- now remove its multiples\nend\n" }, { "path": "deps/lua/test/sort.lua", "content": "-- two implementations of a sort function\n-- this is an example only. Lua has now a built-in function \"sort\"\n\n-- extracted from Programming Pearls, page 110\nfunction qsort(x,l,u,f)\n if ly end)\n show(\"after reverse selection sort\",x)\n qsort(x,1,n,function (x,y) return x>> \",string.rep(\" \",level))\n if t~=nil and t.currentline>=0 then io.write(t.short_src,\":\",t.currentline,\" \") end\n t=debug.getinfo(2)\n if event==\"call\" then\n level=level+1\n else\n level=level-1 if level<0 then level=0 end\n end\n if t.what==\"main\" then\n if event==\"call\" then\n io.write(\"begin \",t.short_src)\n else\n io.write(\"end \",t.short_src)\n end\n elseif t.what==\"Lua\" then\n-- table.foreach(t,print)\n io.write(event,\" \",t.name or \"(Lua)\",\" <\",t.linedefined,\":\",t.short_src,\">\")\n else\n io.write(event,\" \",t.name or \"(C)\",\" [\",t.what,\"] \")\n end\n io.write(\"\\n\")\nend\n\ndebug.sethook(hook,\"cr\")\nlevel=0\n" }, { "path": "deps/lua/test/trace-globals.lua", "content": "-- trace assigments to global variables\n\ndo\n -- a tostring that quotes strings. note the use of the original tostring.\n local _tostring=tostring\n local tostring=function(a)\n if type(a)==\"string\" then\n return string.format(\"%q\",a)\n else\n return _tostring(a)\n end\n end\n\n local log=function (name,old,new)\n local t=debug.getinfo(3,\"Sl\")\n local line=t.currentline\n io.write(t.short_src)\n if line>=0 then io.write(\":\",line) end\n io.write(\": \",name,\" is now \",tostring(new),\" (was \",tostring(old),\")\",\"\\n\")\n end\n\n local g={}\n local set=function (t,name,value)\n log(name,g[name],value)\n g[name]=value\n end\n setmetatable(getfenv(),{__index=g,__newindex=set})\nend\n\n-- an example\n\na=1\nb=2\na=10\nb=20\nb=nil\nb=200\nprint(a,b,c)\n" }, { "path": "deps/lua/test/xd.lua", "content": "-- hex dump\n-- usage: lua xd.lua < file\n\nlocal offset=0\nwhile true do\n local s=io.read(16)\n if s==nil then return end\n io.write(string.format(\"%08X \",offset))\n string.gsub(s,\"(.)\",\n\tfunction (c) io.write(string.format(\"%02X \",string.byte(c))) end)\n io.write(string.rep(\" \",3*(16-string.len(s))))\n io.write(\" \",string.gsub(s,\"%c\",\".\"),\"\\n\") \n offset=offset+16\nend\n" }, { "path": "redis.conf", "content": "# Redis configuration file example\n\n# Note on units: when memory size is needed, it is possible to specify\n# it in the usual form of 1k 5GB 4M and so forth:\n#\n# 1k => 1000 bytes\n# 1kb => 1024 bytes\n# 1m => 1000000 bytes\n# 1mb => 1024*1024 bytes\n# 1g => 1000000000 bytes\n# 1gb => 1024*1024*1024 bytes\n#\n# units are case insensitive so 1GB 1Gb 1gB are all the same.\n\n################################## INCLUDES ###################################\n\n# Include one or more other config files here. This is useful if you\n# have a standard template that goes to all Redis server but also need\n# to customize a few per-server settings. Include files can include\n# other files, so use this wisely.\n#\n# Notice option \"include\" won't be rewritten by command \"CONFIG REWRITE\"\n# from admin or Redis Sentinel. Since Redis always uses the last processed\n# line as value of a configuration directive, you'd better put includes\n# at the beginning of this file to avoid overwriting config change at runtime.\n#\n# If instead you are interested in using includes to override configuration\n# options, it is better to use include as the last line.\n#\n# include /path/to/local.conf\n# include /path/to/other.conf\n\n################################ GENERAL #####################################\n\n# By default Redis does not run as a daemon. Use 'yes' if you need it.\n# Note that Redis will write a pid file in /var/run/redis.pid when daemonized.\ndaemonize no\n\n# When running daemonized, Redis writes a pid file in /var/run/redis.pid by\n# default. You can specify a custom pid file location here.\npidfile /var/run/redis.pid\n\n# Accept connections on the specified port, default is 6379.\n# If port 0 is specified Redis will not listen on a TCP socket.\nport 6379\n\n# TCP listen() backlog.\n#\n# In high requests-per-second environments you need an high backlog in order\n# to avoid slow clients connections issues. Note that the Linux kernel\n# will silently truncate it to the value of /proc/sys/net/core/somaxconn so\n# make sure to raise both the value of somaxconn and tcp_max_syn_backlog\n# in order to get the desired effect.\ntcp-backlog 511\n\n# By default Redis listens for connections from all the network interfaces\n# available on the server. It is possible to listen to just one or multiple\n# interfaces using the \"bind\" configuration directive, followed by one or\n# more IP addresses.\n#\n# Examples:\n#\n# bind 192.168.1.100 10.0.0.1\n# bind 127.0.0.1\n\n# Specify the path for the Unix socket that will be used to listen for\n# incoming connections. There is no default, so Redis will not listen\n# on a unix socket when not specified.\n#\n# unixsocket /tmp/redis.sock\n# unixsocketperm 755\n\n# Close the connection after a client is idle for N seconds (0 to disable)\ntimeout 0\n\n# TCP keepalive.\n#\n# If non-zero, use SO_KEEPALIVE to send TCP ACKs to clients in absence\n# of communication. This is useful for two reasons:\n#\n# 1) Detect dead peers.\n# 2) Take the connection alive from the point of view of network\n# equipment in the middle.\n#\n# On Linux, the specified value (in seconds) is the period used to send ACKs.\n# Note that to close the connection the double of the time is needed.\n# On other kernels the period depends on the kernel configuration.\n#\n# A reasonable value for this option is 60 seconds.\ntcp-keepalive 0\n\n# Specify the server verbosity level.\n# This can be one of:\n# debug (a lot of information, useful for development/testing)\n# verbose (many rarely useful info, but not a mess like the debug level)\n# notice (moderately verbose, what you want in production probably)\n# warning (only very important / critical messages are logged)\nloglevel notice\n\n# Specify the log file name. Also the empty string can be used to force\n# Redis to log on the standard output. Note that if you use standard\n# output for logging but daemonize, logs will be sent to /dev/null\nlogfile \"\"\n\n# To enable logging to the system logger, just set 'syslog-enabled' to yes,\n# and optionally update the other syslog parameters to suit your needs.\n# syslog-enabled no\n\n# Specify the syslog identity.\n# syslog-ident redis\n\n# Specify the syslog facility. Must be USER or between LOCAL0-LOCAL7.\n# syslog-facility local0\n\n# Set the number of databases. The default database is DB 0, you can select\n# a different one on a per-connection basis using SELECT where\n# dbid is a number between 0 and 'databases'-1\ndatabases 16\n\n################################ SNAPSHOTTING ################################\n#\n# Save the DB on disk:\n#\n# save \n#\n# Will save the DB if both the given number of seconds and the given\n# number of write operations against the DB occurred.\n#\n# In the example below the behaviour will be to save:\n# after 900 sec (15 min) if at least 1 key changed\n# after 300 sec (5 min) if at least 10 keys changed\n# after 60 sec if at least 10000 keys changed\n#\n# Note: you can disable saving at all commenting all the \"save\" lines.\n#\n# It is also possible to remove all the previously configured save\n# points by adding a save directive with a single empty string argument\n# like in the following example:\n#\n# save \"\"\n\nsave 900 1\nsave 300 10\nsave 60 10000\n\n# By default Redis will stop accepting writes if RDB snapshots are enabled\n# (at least one save point) and the latest background save failed.\n# This will make the user aware (in a hard way) that data is not persisting\n# on disk properly, otherwise chances are that no one will notice and some\n# disaster will happen.\n#\n# If the background saving process will start working again Redis will\n# automatically allow writes again.\n#\n# However if you have setup your proper monitoring of the Redis server\n# and persistence, you may want to disable this feature so that Redis will\n# continue to work as usual even if there are problems with disk,\n# permissions, and so forth.\nstop-writes-on-bgsave-error yes\n\n# Compress string objects using LZF when dump .rdb databases?\n# For default that's set to 'yes' as it's almost always a win.\n# If you want to save some CPU in the saving child set it to 'no' but\n# the dataset will likely be bigger if you have compressible values or keys.\nrdbcompression yes\n\n# Since version 5 of RDB a CRC64 checksum is placed at the end of the file.\n# This makes the format more resistant to corruption but there is a performance\n# hit to pay (around 10%) when saving and loading RDB files, so you can disable it\n# for maximum performances.\n#\n# RDB files created with checksum disabled have a checksum of zero that will\n# tell the loading code to skip the check.\nrdbchecksum yes\n\n# The filename where to dump the DB\ndbfilename dump.rdb\n\n# The working directory.\n#\n# The DB will be written inside this directory, with the filename specified\n# above using the 'dbfilename' configuration directive.\n# \n# The Append Only File will also be created inside this directory.\n# \n# Note that you must specify a directory here, not a file name.\ndir ./\n\n################################# REPLICATION #################################\n\n# Master-Slave replication. Use slaveof to make a Redis instance a copy of\n# another Redis server. Note that the configuration is local to the slave\n# so for example it is possible to configure the slave to save the DB with a\n# different interval, or to listen to another port, and so on.\n#\n# slaveof \n\n# If the master is password protected (using the \"requirepass\" configuration\n# directive below) it is possible to tell the slave to authenticate before\n# starting the replication synchronization process, otherwise the master will\n# refuse the slave request.\n#\n# masterauth \n\n# When a slave loses its connection with the master, or when the replication\n# is still in progress, the slave can act in two different ways:\n#\n# 1) if slave-serve-stale-data is set to 'yes' (the default) the slave will\n# still reply to client requests, possibly with out of date data, or the\n# data set may just be empty if this is the first synchronization.\n#\n# 2) if slave-serve-stale-data is set to 'no' the slave will reply with\n# an error \"SYNC with master in progress\" to all the kind of commands\n# but to INFO and SLAVEOF.\n#\nslave-serve-stale-data yes\n\n# You can configure a slave instance to accept writes or not. Writing against\n# a slave instance may be useful to store some ephemeral data (because data\n# written on a slave will be easily deleted after resync with the master) but\n# may also cause problems if clients are writing to it because of a\n# misconfiguration.\n#\n# Since Redis 2.6 by default slaves are read-only.\n#\n# Note: read only slaves are not designed to be exposed to untrusted clients\n# on the internet. It's just a protection layer against misuse of the instance.\n# Still a read only slave exports by default all the administrative commands\n# such as CONFIG, DEBUG, and so forth. To a limited extent you can improve\n# security of read only slaves using 'rename-command' to shadow all the\n# administrative / dangerous commands.\nslave-read-only yes\n\n# Slaves send PINGs to server in a predefined interval. It's possible to change\n# this interval with the repl_ping_slave_period option. The default value is 10\n# seconds.\n#\n# repl-ping-slave-period 10\n\n# The following option sets the replication timeout for:\n#\n# 1) Bulk transfer I/O during SYNC, from the point of view of slave.\n# 2) Master timeout from the point of view of slaves (data, pings).\n# 3) Slave timeout from the point of view of masters (REPLCONF ACK pings).\n#\n# It is important to make sure that this value is greater than the value\n# specified for repl-ping-slave-period otherwise a timeout will be detected\n# every time there is low traffic between the master and the slave.\n#\n# repl-timeout 60\n\n# Disable TCP_NODELAY on the slave socket after SYNC?\n#\n# If you select \"yes\" Redis will use a smaller number of TCP packets and\n# less bandwidth to send data to slaves. But this can add a delay for\n# the data to appear on the slave side, up to 40 milliseconds with\n# Linux kernels using a default configuration.\n#\n# If you select \"no\" the delay for data to appear on the slave side will\n# be reduced but more bandwidth will be used for replication.\n#\n# By default we optimize for low latency, but in very high traffic conditions\n# or when the master and slaves are many hops away, turning this to \"yes\" may\n# be a good idea.\nrepl-disable-tcp-nodelay no\n\n# Set the replication backlog size. The backlog is a buffer that accumulates\n# slave data when slaves are disconnected for some time, so that when a slave\n# wants to reconnect again, often a full resync is not needed, but a partial\n# resync is enough, just passing the portion of data the slave missed while\n# disconnected.\n#\n# The biggest the replication backlog, the longer the time the slave can be\n# disconnected and later be able to perform a partial resynchronization.\n#\n# The backlog is only allocated once there is at least a slave connected.\n#\n# repl-backlog-size 1mb\n\n# After a master has no longer connected slaves for some time, the backlog\n# will be freed. The following option configures the amount of seconds that\n# need to elapse, starting from the time the last slave disconnected, for\n# the backlog buffer to be freed.\n#\n# A value of 0 means to never release the backlog.\n#\n# repl-backlog-ttl 3600\n\n# The slave priority is an integer number published by Redis in the INFO output.\n# It is used by Redis Sentinel in order to select a slave to promote into a\n# master if the master is no longer working correctly.\n#\n# A slave with a low priority number is considered better for promotion, so\n# for instance if there are three slaves with priority 10, 100, 25 Sentinel will\n# pick the one with priority 10, that is the lowest.\n#\n# However a special priority of 0 marks the slave as not able to perform the\n# role of master, so a slave with priority of 0 will never be selected by\n# Redis Sentinel for promotion.\n#\n# By default the priority is 100.\nslave-priority 100\n\n# It is possible for a master to stop accepting writes if there are less than\n# N slaves connected, having a lag less or equal than M seconds.\n#\n# The N slaves need to be in \"online\" state.\n#\n# The lag in seconds, that must be <= the specified value, is calculated from\n# the last ping received from the slave, that is usually sent every second.\n#\n# This option does not GUARANTEES that N replicas will accept the write, but\n# will limit the window of exposure for lost writes in case not enough slaves\n# are available, to the specified number of seconds.\n#\n# For example to require at least 3 slaves with a lag <= 10 seconds use:\n#\n# min-slaves-to-write 3\n# min-slaves-max-lag 10\n#\n# Setting one or the other to 0 disables the feature.\n#\n# By default min-slaves-to-write is set to 0 (feature disabled) and\n# min-slaves-max-lag is set to 10.\n\n################################## SECURITY ###################################\n\n# Require clients to issue AUTH before processing any other\n# commands. This might be useful in environments in which you do not trust\n# others with access to the host running redis-server.\n#\n# This should stay commented out for backward compatibility and because most\n# people do not need auth (e.g. they run their own servers).\n# \n# Warning: since Redis is pretty fast an outside user can try up to\n# 150k passwords per second against a good box. This means that you should\n# use a very strong password otherwise it will be very easy to break.\n#\n# requirepass foobared\n\n# Command renaming.\n#\n# It is possible to change the name of dangerous commands in a shared\n# environment. For instance the CONFIG command may be renamed into something\n# hard to guess so that it will still be available for internal-use tools\n# but not available for general clients.\n#\n# Example:\n#\n# rename-command CONFIG b840fc02d524045429941cc15f59e41cb7be6c52\n#\n# It is also possible to completely kill a command by renaming it into\n# an empty string:\n#\n# rename-command CONFIG \"\"\n#\n# Please note that changing the name of commands that are logged into the\n# AOF file or transmitted to slaves may cause problems.\n\n################################### LIMITS ####################################\n\n# Set the max number of connected clients at the same time. By default\n# this limit is set to 10000 clients, however if the Redis server is not\n# able to configure the process file limit to allow for the specified limit\n# the max number of allowed clients is set to the current file limit\n# minus 32 (as Redis reserves a few file descriptors for internal uses).\n#\n# Once the limit is reached Redis will close all the new connections sending\n# an error 'max number of clients reached'.\n#\n# maxclients 10000\n\n# Don't use more memory than the specified amount of bytes.\n# When the memory limit is reached Redis will try to remove keys\n# according to the eviction policy selected (see maxmemory-policy).\n#\n# If Redis can't remove keys according to the policy, or if the policy is\n# set to 'noeviction', Redis will start to reply with errors to commands\n# that would use more memory, like SET, LPUSH, and so on, and will continue\n# to reply to read-only commands like GET.\n#\n# This option is usually useful when using Redis as an LRU cache, or to set\n# a hard memory limit for an instance (using the 'noeviction' policy).\n#\n# WARNING: If you have slaves attached to an instance with maxmemory on,\n# the size of the output buffers needed to feed the slaves are subtracted\n# from the used memory count, so that network problems / resyncs will\n# not trigger a loop where keys are evicted, and in turn the output\n# buffer of slaves is full with DELs of keys evicted triggering the deletion\n# of more keys, and so forth until the database is completely emptied.\n#\n# In short... if you have slaves attached it is suggested that you set a lower\n# limit for maxmemory so that there is some free RAM on the system for slave\n# output buffers (but this is not needed if the policy is 'noeviction').\n#\n# maxmemory \n\n# MAXMEMORY POLICY: how Redis will select what to remove when maxmemory\n# is reached. You can select among five behaviors:\n# \n# volatile-lru -> remove the key with an expire set using an LRU algorithm\n# allkeys-lru -> remove any key accordingly to the LRU algorithm\n# volatile-random -> remove a random key with an expire set\n# allkeys-random -> remove a random key, any key\n# volatile-ttl -> remove the key with the nearest expire time (minor TTL)\n# noeviction -> don't expire at all, just return an error on write operations\n# \n# Note: with any of the above policies, Redis will return an error on write\n# operations, when there are not suitable keys for eviction.\n#\n# At the date of writing this commands are: set setnx setex append\n# incr decr rpush lpush rpushx lpushx linsert lset rpoplpush sadd\n# sinter sinterstore sunion sunionstore sdiff sdiffstore zadd zincrby\n# zunionstore zinterstore hset hsetnx hmset hincrby incrby decrby\n# getset mset msetnx exec sort\n#\n# The default is:\n#\n# maxmemory-policy noeviction\n\n# LRU and minimal TTL algorithms are not precise algorithms but approximated\n# algorithms (in order to save memory), so you can tune it for speed or\n# accuracy. For default Redis will check five keys and pick the one that was\n# used less recently, you can change the sample size using the following\n# configuration directive.\n#\n# The default of 5 produces good enough results. 10 Approximates very closely\n# true LRU but costs a bit more CPU. 3 is very fast but not very accurate.\n#\n# maxmemory-samples 5\n\n############################## APPEND ONLY MODE ###############################\n\n# By default Redis asynchronously dumps the dataset on disk. This mode is\n# good enough in many applications, but an issue with the Redis process or\n# a power outage may result into a few minutes of writes lost (depending on\n# the configured save points).\n#\n# The Append Only File is an alternative persistence mode that provides\n# much better durability. For instance using the default data fsync policy\n# (see later in the config file) Redis can lose just one second of writes in a\n# dramatic event like a server power outage, or a single write if something\n# wrong with the Redis process itself happens, but the operating system is\n# still running correctly.\n#\n# AOF and RDB persistence can be enabled at the same time without problems.\n# If the AOF is enabled on startup Redis will load the AOF, that is the file\n# with the better durability guarantees.\n#\n# Please check http://redis.io/topics/persistence for more information.\n\nappendonly no\n\n# The name of the append only file (default: \"appendonly.aof\")\n\nappendfilename \"appendonly.aof\"\n\n# The fsync() call tells the Operating System to actually write data on disk\n# instead to wait for more data in the output buffer. Some OS will really flush \n# data on disk, some other OS will just try to do it ASAP.\n#\n# Redis supports three different modes:\n#\n# no: don't fsync, just let the OS flush the data when it wants. Faster.\n# always: fsync after every write to the append only log . Slow, Safest.\n# everysec: fsync only one time every second. Compromise.\n#\n# The default is \"everysec\", as that's usually the right compromise between\n# speed and data safety. It's up to you to understand if you can relax this to\n# \"no\" that will let the operating system flush the output buffer when\n# it wants, for better performances (but if you can live with the idea of\n# some data loss consider the default persistence mode that's snapshotting),\n# or on the contrary, use \"always\" that's very slow but a bit safer than\n# everysec.\n#\n# More details please check the following article:\n# http://antirez.com/post/redis-persistence-demystified.html\n#\n# If unsure, use \"everysec\".\n\n# appendfsync always\nappendfsync everysec\n# appendfsync no\n\n# When the AOF fsync policy is set to always or everysec, and a background\n# saving process (a background save or AOF log background rewriting) is\n# performing a lot of I/O against the disk, in some Linux configurations\n# Redis may block too long on the fsync() call. Note that there is no fix for\n# this currently, as even performing fsync in a different thread will block\n# our synchronous write(2) call.\n#\n# In order to mitigate this problem it's possible to use the following option\n# that will prevent fsync() from being called in the main process while a\n# BGSAVE or BGREWRITEAOF is in progress.\n#\n# This means that while another child is saving, the durability of Redis is\n# the same as \"appendfsync none\". In practical terms, this means that it is\n# possible to lose up to 30 seconds of log in the worst scenario (with the\n# default Linux settings).\n# \n# If you have latency problems turn this to \"yes\". Otherwise leave it as\n# \"no\" that is the safest pick from the point of view of durability.\n\nno-appendfsync-on-rewrite no\n\n# Automatic rewrite of the append only file.\n# Redis is able to automatically rewrite the log file implicitly calling\n# BGREWRITEAOF when the AOF log size grows by the specified percentage.\n# \n# This is how it works: Redis remembers the size of the AOF file after the\n# latest rewrite (if no rewrite has happened since the restart, the size of\n# the AOF at startup is used).\n#\n# This base size is compared to the current size. If the current size is\n# bigger than the specified percentage, the rewrite is triggered. Also\n# you need to specify a minimal size for the AOF file to be rewritten, this\n# is useful to avoid rewriting the AOF file even if the percentage increase\n# is reached but it is still pretty small.\n#\n# Specify a percentage of zero in order to disable the automatic AOF\n# rewrite feature.\n\nauto-aof-rewrite-percentage 100\nauto-aof-rewrite-min-size 64mb\n\n################################ LUA SCRIPTING ###############################\n\n# Max execution time of a Lua script in milliseconds.\n#\n# If the maximum execution time is reached Redis will log that a script is\n# still in execution after the maximum allowed time and will start to\n# reply to queries with an error.\n#\n# When a long running script exceed the maximum execution time only the\n# SCRIPT KILL and SHUTDOWN NOSAVE commands are available. The first can be\n# used to stop a script that did not yet called write commands. The second\n# is the only way to shut down the server in the case a write commands was\n# already issue by the script but the user don't want to wait for the natural\n# termination of the script.\n#\n# Set it to 0 or a negative value for unlimited execution without warnings.\nlua-time-limit 5000\n\n################################ REDIS CLUSTER ###############################\n#\n# Normal Redis instances can't be part of a Redis Cluster; only nodes that are\n# started as cluster nodes can. In order to start a Redis instance as a\n# cluster node enable the cluster support uncommenting the following:\n#\n# cluster-enabled yes\n\n# Every cluster node has a cluster configuration file. This file is not\n# intended to be edited by hand. It is created and updated by Redis nodes.\n# Every Redis Cluster node requires a different cluster configuration file.\n# Make sure that instances running in the same system does not have\n# overlapping cluster configuration file names.\n#\n# cluster-config-file nodes-6379.conf\n\n# Cluster node timeout is the amount of milliseconds a node must be unreachable \n# for it to be considered in failure state.\n# Most other internal time limits are multiple of the node timeout.\n#\n# cluster-node-timeout 15000\n\n# Cluster slaves are able to migrate to orphaned masters, that are masters\n# that are left without working slaves. This improves the cluster ability\n# to resist to failures as otherwise an orphaned master can't be failed over\n# in case of failure if it has no working slaves.\n#\n# Slaves migrate to orphaned masters only if there are still at least a\n# given number of other working slaves for their old master. This number\n# is the \"migration barrier\". A migration barrier of 1 means that a slave\n# will migrate only if there is at least 1 other working slave for its master\n# and so forth. It usually reflects the number of slaves you want for every\n# master in your cluster.\n#\n# Default is 1 (slaves migrate only if their masters remain with at least\n# one slave). To disable migration just set it to a very large value.\n# A value of 0 can be set but is useful only for debugging and dangerous\n# in production.\n#\n# cluster-migration-barrier 1\n\n# In order to setup your cluster make sure to read the documentation\n# available at http://redis.io web site.\n\n################################## SLOW LOG ###################################\n\n# The Redis Slow Log is a system to log queries that exceeded a specified\n# execution time. The execution time does not include the I/O operations\n# like talking with the client, sending the reply and so forth,\n# but just the time needed to actually execute the command (this is the only\n# stage of command execution where the thread is blocked and can not serve\n# other requests in the meantime).\n# \n# You can configure the slow log with two parameters: one tells Redis\n# what is the execution time, in microseconds, to exceed in order for the\n# command to get logged, and the other parameter is the length of the\n# slow log. When a new command is logged the oldest one is removed from the\n# queue of logged commands.\n\n# The following time is expressed in microseconds, so 1000000 is equivalent\n# to one second. Note that a negative number disables the slow log, while\n# a value of zero forces the logging of every command.\nslowlog-log-slower-than 10000\n\n# There is no limit to this length. Just be aware that it will consume memory.\n# You can reclaim memory used by the slow log with SLOWLOG RESET.\nslowlog-max-len 128\n\n############################# Event notification ##############################\n\n# Redis can notify Pub/Sub clients about events happening in the key space.\n# This feature is documented at http://redis.io/topics/keyspace-events\n# \n# For instance if keyspace events notification is enabled, and a client\n# performs a DEL operation on key \"foo\" stored in the Database 0, two\n# messages will be published via Pub/Sub:\n#\n# PUBLISH __keyspace@0__:foo del\n# PUBLISH __keyevent@0__:del foo\n#\n# It is possible to select the events that Redis will notify among a set\n# of classes. Every class is identified by a single character:\n#\n# K Keyspace events, published with __keyspace@__ prefix.\n# E Keyevent events, published with __keyevent@__ prefix.\n# g Generic commands (non-type specific) like DEL, EXPIRE, RENAME, ...\n# $ String commands\n# l List commands\n# s Set commands\n# h Hash commands\n# z Sorted set commands\n# x Expired events (events generated every time a key expires)\n# e Evicted events (events generated when a key is evicted for maxmemory)\n# A Alias for g$lshzxe, so that the \"AKE\" string means all the events.\n#\n# The \"notify-keyspace-events\" takes as argument a string that is composed\n# by zero or multiple characters. The empty string means that notifications\n# are disabled at all.\n#\n# Example: to enable list and generic events, from the point of view of the\n# event name, use:\n#\n# notify-keyspace-events Elg\n#\n# Example 2: to get the stream of the expired keys subscribing to channel\n# name __keyevent@0__:expired use:\n#\n# notify-keyspace-events Ex\n#\n# By default all notifications are disabled because most users don't need\n# this feature and the feature has some overhead. Note that if you don't\n# specify at least one of K or E, no events will be delivered.\nnotify-keyspace-events \"\"\n\n############################### ADVANCED CONFIG ###############################\n\n# Hashes are encoded using a memory efficient data structure when they have a\n# small number of entries, and the biggest entry does not exceed a given\n# threshold. These thresholds can be configured using the following directives.\nhash-max-ziplist-entries 512\nhash-max-ziplist-value 64\n\n# Similarly to hashes, small lists are also encoded in a special way in order\n# to save a lot of space. The special representation is only used when\n# you are under the following limits:\nlist-max-ziplist-entries 512\nlist-max-ziplist-value 64\n\n# Sets have a special encoding in just one case: when a set is composed\n# of just strings that happens to be integers in radix 10 in the range\n# of 64 bit signed integers.\n# The following configuration setting sets the limit in the size of the\n# set in order to use this special memory saving encoding.\nset-max-intset-entries 512\n\n# Similarly to hashes and lists, sorted sets are also specially encoded in\n# order to save a lot of space. This encoding is only used when the length and\n# elements of a sorted set are below the following limits:\nzset-max-ziplist-entries 128\nzset-max-ziplist-value 64\n\n# HyperLogLog sparse representation bytes limit. The limit includes the\n# 16 bytes header. When an HyperLogLog using the sparse representation crosses\n# this limit, it is converted into the dense representation.\n#\n# A value greater than 16000 is totally useless, since at that point the\n# dense representation is more memory efficient.\n# \n# The suggested value is ~ 3000 in order to have the benefits of\n# the space efficient encoding without slowing down too much PFADD,\n# which is O(N) with the sparse encoding. The value can be raised to\n# ~ 10000 when CPU is not a concern, but space is, and the data set is\n# composed of many HyperLogLogs with cardinality in the 0 - 15000 range.\nhll-sparse-max-bytes 3000\n\n# Active rehashing uses 1 millisecond every 100 milliseconds of CPU time in\n# order to help rehashing the main Redis hash table (the one mapping top-level\n# keys to values). The hash table implementation Redis uses (see dict.c)\n# performs a lazy rehashing: the more operation you run into a hash table\n# that is rehashing, the more rehashing \"steps\" are performed, so if the\n# server is idle the rehashing is never complete and some more memory is used\n# by the hash table.\n# \n# The default is to use this millisecond 10 times every second in order to\n# active rehashing the main dictionaries, freeing memory when possible.\n#\n# If unsure:\n# use \"activerehashing no\" if you have hard latency requirements and it is\n# not a good thing in your environment that Redis can reply form time to time\n# to queries with 2 milliseconds delay.\n#\n# use \"activerehashing yes\" if you don't have such hard requirements but\n# want to free memory asap when possible.\nactiverehashing yes\n\n# The client output buffer limits can be used to force disconnection of clients\n# that are not reading data from the server fast enough for some reason (a\n# common reason is that a Pub/Sub client can't consume messages as fast as the\n# publisher can produce them).\n#\n# The limit can be set differently for the three different classes of clients:\n#\n# normal -> normal clients\n# slave -> slave clients and MONITOR clients\n# pubsub -> clients subscribed to at least one pubsub channel or pattern\n#\n# The syntax of every client-output-buffer-limit directive is the following:\n#\n# client-output-buffer-limit \n#\n# A client is immediately disconnected once the hard limit is reached, or if\n# the soft limit is reached and remains reached for the specified number of\n# seconds (continuously).\n# So for instance if the hard limit is 32 megabytes and the soft limit is\n# 16 megabytes / 10 seconds, the client will get disconnected immediately\n# if the size of the output buffers reach 32 megabytes, but will also get\n# disconnected if the client reaches 16 megabytes and continuously overcomes\n# the limit for 10 seconds.\n#\n# By default normal clients are not limited because they don't receive data\n# without asking (in a push way), but just after a request, so only\n# asynchronous clients may create a scenario where data is requested faster\n# than it can read.\n#\n# Instead there is a default limit for pubsub and slave clients, since\n# subscribers and slaves receive data in a push fashion.\n#\n# Both the hard or the soft limit can be disabled by setting them to zero.\nclient-output-buffer-limit normal 0 0 0\nclient-output-buffer-limit slave 256mb 64mb 60\nclient-output-buffer-limit pubsub 32mb 8mb 60\n\n# Redis calls an internal function to perform many background tasks, like\n# closing connections of clients in timeout, purging expired keys that are\n# never requested, and so forth.\n#\n# Not all tasks are performed with the same frequency, but Redis checks for\n# tasks to perform accordingly to the specified \"hz\" value.\n#\n# By default \"hz\" is set to 10. Raising the value will use more CPU when\n# Redis is idle, but at the same time will make Redis more responsive when\n# there are many keys expiring at the same time, and timeouts may be\n# handled with more precision.\n#\n# The range is between 1 and 500, however a value over 100 is usually not\n# a good idea. Most users should use the default of 10 and raise this up to\n# 100 only in environments where very low latency is required.\nhz 10\n\n# When a child rewrites the AOF file, if the following option is enabled\n# the file will be fsync-ed every 32 MB of data generated. This is useful\n# in order to commit the file to the disk more incrementally and avoid\n# big latency spikes.\naof-rewrite-incremental-fsync yes\n\n" }, { "path": "runtest", "content": "#!/bin/sh\nTCL_VERSIONS=\"8.5 8.6\"\nTCLSH=\"\"\n\nfor VERSION in $TCL_VERSIONS; do\n\tTCL=`which tclsh$VERSION 2>/dev/null` && TCLSH=$TCL\ndone\n\nif [ -z $TCLSH ]\nthen\n echo \"You need tcl 8.5 or newer in order to run the Redis test\"\n exit 1\nfi\n$TCLSH tests/test_helper.tcl $*\n" }, { "path": "runtest-cluster", "content": "#!/bin/sh\nTCL_VERSIONS=\"8.5 8.6\"\nTCLSH=\"\"\n\nfor VERSION in $TCL_VERSIONS; do\n\tTCL=`which tclsh$VERSION 2>/dev/null` && TCLSH=$TCL\ndone\n\nif [ -z $TCLSH ]\nthen\n echo \"You need tcl 8.5 or newer in order to run the Redis Sentinel test\"\n exit 1\nfi\n$TCLSH tests/cluster/run.tcl $*\n" }, { "path": "runtest-sentinel", "content": "#!/bin/sh\nTCL_VERSIONS=\"8.5 8.6\"\nTCLSH=\"\"\n\nfor VERSION in $TCL_VERSIONS; do\n\tTCL=`which tclsh$VERSION 2>/dev/null` && TCLSH=$TCL\ndone\n\nif [ -z $TCLSH ]\nthen\n echo \"You need tcl 8.5 or newer in order to run the Redis Sentinel test\"\n exit 1\nfi\n$TCLSH tests/sentinel/run.tcl $*\n" }, { "path": "sentinel.conf", "content": "# Example sentinel.conf\n\n# port \n# The port that this sentinel instance will run on\nport 26379\n\n# dir \n# Every long running process should have a well-defined working directory.\n# For Redis Sentinel to chdir to /tmp at startup is the simplest thing\n# for the process to don't interferer with administrative tasks such as\n# unmounting filesystems.\ndir /tmp\n\n# sentinel monitor \n#\n# Tells Sentinel to monitor this master, and to consider it in O_DOWN\n# (Objectively Down) state only if at least sentinels agree.\n#\n# Note that whatever is the ODOWN quorum, a Sentinel will require to\n# be elected by the majority of the known Sentinels in order to\n# start a failover, so no failover can be performed in minority.\n#\n# Note: master name should not include special characters or spaces.\n# The valid charset is A-z 0-9 and the three characters \".-_\".\nsentinel monitor mymaster 127.0.0.1 6379 2\n\n# sentinel auth-pass \n#\n# Set the password to use to authenticate with the master and slaves.\n# Useful if there is a password set in the Redis instances to monitor.\n#\n# Note that the master password is also used for slaves, so it is not\n# possible to set a different password in masters and slaves instances\n# if you want to be able to monitor these instances with Sentinel.\n#\n# However you can have Redis instances without the authentication enabled\n# mixed with Redis instances requiring the authentication (as long as the\n# password set is the same for all the instances requiring the password) as\n# the AUTH command will have no effect in Redis instances with authentication\n# switched off.\n#\n# Example:\n#\n# sentinel auth-pass mymaster MySUPER--secret-0123passw0rd\n\n# sentinel down-after-milliseconds \n#\n# Number of milliseconds the master (or any attached slave or sentinel) should\n# be unreachable (as in, not acceptable reply to PING, continuously, for the\n# specified period) in order to consider it in S_DOWN state (Subjectively\n# Down).\n#\n# Default is 30 seconds.\nsentinel down-after-milliseconds mymaster 30000\n\n# sentinel parallel-syncs \n#\n# How many slaves we can reconfigure to point to the new slave simultaneously\n# during the failover. Use a low number if you use the slaves to serve query\n# to avoid that all the slaves will be unreachable at about the same\n# time while performing the synchronization with the master.\nsentinel parallel-syncs mymaster 1\n\n# sentinel failover-timeout \n#\n# Specifies the failover timeout in milliseconds. It is used in many ways:\n#\n# - The time needed to re-start a failover after a previous failover was\n# already tried against the same master by a given Sentinel, is two\n# times the failover timeout.\n#\n# - The time needed for a slave replicating to a wrong master according\n# to a Sentinel current configuration, to be forced to replicate\n# with the right master, is exactly the failover timeout (counting since\n# the moment a Sentinel detected the misconfiguration).\n#\n# - The time needed to cancel a failover that is already in progress but\n# did not produced any configuration change (SLAVEOF NO ONE yet not\n# acknowledged by the promoted slave).\n#\n# - The maximum time a failover in progress waits for all the slaves to be\n# reconfigured as slaves of the new master. However even after this time\n# the slaves will be reconfigured by the Sentinels anyway, but not with\n# the exact parallel-syncs progression as specified.\n#\n# Default is 3 minutes.\nsentinel failover-timeout mymaster 180000\n\n# SCRIPTS EXECUTION\n#\n# sentinel notification-script and sentinel reconfig-script are used in order\n# to configure scripts that are called to notify the system administrator\n# or to reconfigure clients after a failover. The scripts are executed\n# with the following rules for error handling:\n#\n# If script exits with \"1\" the execution is retried later (up to a maximum\n# number of times currently set to 10).\n#\n# If script exits with \"2\" (or an higher value) the script execution is\n# not retried.\n#\n# If script terminates because it receives a signal the behavior is the same\n# as exit code 1.\n#\n# A script has a maximum running time of 60 seconds. After this limit is\n# reached the script is terminated with a SIGKILL and the execution retried.\n\n# NOTIFICATION SCRIPT\n#\n# sentinel notification-script \n# \n# Call the specified notification script for any sentinel event that is\n# generated in the WARNING level (for instance -sdown, -odown, and so forth).\n# This script should notify the system administrator via email, SMS, or any\n# other messaging system, that there is something wrong with the monitored\n# Redis systems.\n#\n# The script is called with just two arguments: the first is the event type\n# and the second the event description.\n#\n# The script must exist and be executable in order for sentinel to start if\n# this option is provided.\n#\n# Example:\n#\n# sentinel notification-script mymaster /var/redis/notify.sh\n\n# CLIENTS RECONFIGURATION SCRIPT\n#\n# sentinel client-reconfig-script \n#\n# When the master changed because of a failover a script can be called in\n# order to perform application-specific tasks to notify the clients that the\n# configuration has changed and the master is at a different address.\n# \n# The following arguments are passed to the script:\n#\n# \n#\n# is currently always \"failover\"\n# is either \"leader\" or \"observer\"\n# \n# The arguments from-ip, from-port, to-ip, to-port are used to communicate\n# the old address of the master and the new address of the elected slave\n# (now a master).\n#\n# This script should be resistant to multiple invocations.\n#\n# Example:\n#\n# sentinel client-reconfig-script mymaster /var/redis/reconfig.sh\n\n\n" }, { "path": "src/.gitignore", "content": "*.gcda\n*.gcno\n*.gcov\nredis.info\nlcov-html\n" }, { "path": "src/Makefile", "content": "# Redis Makefile\n# Copyright (C) 2009 Salvatore Sanfilippo \n# This file is released under the BSD license, see the COPYING file\n#\n# The Makefile composes the final FINAL_CFLAGS and FINAL_LDFLAGS using\n# what is needed for Redis plus the standard CFLAGS and LDFLAGS passed.\n# However when building the dependencies (Jemalloc, Lua, Hiredis, ...)\n# CFLAGS and LDFLAGS are propagated to the dependencies, so to pass\n# flags only to be used when compiling / linking Redis itself REDIS_CFLAGS\n# and REDIS_LDFLAGS are used instead (this is the case of 'make gcov').\n#\n# Dependencies are stored in the Makefile.dep file. To rebuild this file\n# Just use 'make dep', but this is only needed by developers.\n\nrelease_hdr := $(shell sh -c './mkreleasehdr.sh')\nuname_S := $(shell sh -c 'uname -s 2>/dev/null || echo not')\nOPTIMIZATION?=-O2\nDEPENDENCY_TARGETS=hiredis linenoise lua\n\n# Default settings\nSTD=-std=c99 -pedantic\nWARN=-Wall\nOPT=$(OPTIMIZATION)\n\nPREFIX?=/usr/local\nINSTALL_BIN=$(PREFIX)/bin\nINSTALL=install\n\n# Default allocator\nifeq ($(uname_S),Linux)\n\tMALLOC=jemalloc\nelse\n\tMALLOC=libc\nendif\n\n# Backwards compatibility for selecting an allocator\nifeq ($(USE_TCMALLOC),yes)\n\tMALLOC=tcmalloc\nendif\n\nifeq ($(USE_TCMALLOC_MINIMAL),yes)\n\tMALLOC=tcmalloc_minimal\nendif\n\nifeq ($(USE_JEMALLOC),yes)\n\tMALLOC=jemalloc\nendif\n\n# Override default settings if possible\n-include .make-settings\n\nFINAL_CFLAGS=$(STD) $(WARN) $(OPT) $(DEBUG) $(CFLAGS) $(REDIS_CFLAGS)\nFINAL_LDFLAGS=$(LDFLAGS) $(REDIS_LDFLAGS) $(DEBUG)\nFINAL_LIBS=-lm\nDEBUG=-g -ggdb\n\nifeq ($(uname_S),SunOS)\n\t# SunOS\n\tINSTALL=cp -pf\n\tFINAL_CFLAGS+= -D__EXTENSIONS__ -D_XPG6\n\tFINAL_LIBS+= -ldl -lnsl -lsocket -lpthread\nelse\nifeq ($(uname_S),Darwin)\n\t# Darwin (nothing to do)\nelse\n\t# All the other OSes (notably Linux)\n\tFINAL_LDFLAGS+= -rdynamic\n\tFINAL_LIBS+= -pthread\nendif\nendif\n\n# Include paths to dependencies\nFINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src\n\nifeq ($(MALLOC),tcmalloc)\n\tFINAL_CFLAGS+= -DUSE_TCMALLOC\n\tFINAL_LIBS+= -ltcmalloc\nendif\n\nifeq ($(MALLOC),tcmalloc_minimal)\n\tFINAL_CFLAGS+= -DUSE_TCMALLOC\n\tFINAL_LIBS+= -ltcmalloc_minimal\nendif\n\nifeq ($(MALLOC),jemalloc)\n\tDEPENDENCY_TARGETS+= jemalloc\n\tFINAL_CFLAGS+= -DUSE_JEMALLOC -I../deps/jemalloc/include\n\tFINAL_LIBS+= ../deps/jemalloc/lib/libjemalloc.a -ldl\nendif\n\nREDIS_CC=$(QUIET_CC)$(CC) $(FINAL_CFLAGS)\nREDIS_LD=$(QUIET_LINK)$(CC) $(FINAL_LDFLAGS)\nREDIS_INSTALL=$(QUIET_INSTALL)$(INSTALL)\n\nCCCOLOR=\"\\033[34m\"\nLINKCOLOR=\"\\033[34;1m\"\nSRCCOLOR=\"\\033[33m\"\nBINCOLOR=\"\\033[37;1m\"\nMAKECOLOR=\"\\033[32;1m\"\nENDCOLOR=\"\\033[0m\"\n\nifndef V\nQUIET_CC = @printf ' %b %b\\n' $(CCCOLOR)CC$(ENDCOLOR) $(SRCCOLOR)$@$(ENDCOLOR) 1>&2;\nQUIET_LINK = @printf ' %b %b\\n' $(LINKCOLOR)LINK$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR) 1>&2;\nQUIET_INSTALL = @printf ' %b %b\\n' $(LINKCOLOR)INSTALL$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR) 1>&2;\nendif\n\nREDIS_SERVER_NAME=redis-server\nREDIS_SENTINEL_NAME=redis-sentinel\nREDIS_SERVER_OBJ=adlist.o ae.o anet.o dict.o redis.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o\nREDIS_CLI_NAME=redis-cli\nREDIS_CLI_OBJ=anet.o sds.o adlist.o redis-cli.o zmalloc.o release.o anet.o ae.o crc64.o\nREDIS_BENCHMARK_NAME=redis-benchmark\nREDIS_BENCHMARK_OBJ=ae.o anet.o redis-benchmark.o sds.o adlist.o zmalloc.o redis-benchmark.o\nREDIS_CHECK_DUMP_NAME=redis-check-dump\nREDIS_CHECK_DUMP_OBJ=redis-check-dump.o lzf_c.o lzf_d.o crc64.o\nREDIS_CHECK_AOF_NAME=redis-check-aof\nREDIS_CHECK_AOF_OBJ=redis-check-aof.o\n\nall: $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_DUMP_NAME) $(REDIS_CHECK_AOF_NAME)\n\t@echo \"\"\n\t@echo \"Hint: To run 'make test' is a good idea ;)\"\n\t@echo \"\"\n\n.PHONY: all\n\n# Deps (use make dep to generate this)\ninclude Makefile.dep\n\ndep:\n\t$(REDIS_CC) -MM *.c > Makefile.dep\n\n.PHONY: dep\n\npersist-settings: distclean\n\techo STD=$(STD) >> .make-settings\n\techo WARN=$(WARN) >> .make-settings\n\techo OPT=$(OPT) >> .make-settings\n\techo MALLOC=$(MALLOC) >> .make-settings\n\techo CFLAGS=$(CFLAGS) >> .make-settings\n\techo LDFLAGS=$(LDFLAGS) >> .make-settings\n\techo REDIS_CFLAGS=$(REDIS_CFLAGS) >> .make-settings\n\techo REDIS_LDFLAGS=$(REDIS_LDFLAGS) >> .make-settings\n\techo PREV_FINAL_CFLAGS=$(FINAL_CFLAGS) >> .make-settings\n\techo PREV_FINAL_LDFLAGS=$(FINAL_LDFLAGS) >> .make-settings\n\t-(cd ../deps && $(MAKE) $(DEPENDENCY_TARGETS))\n\n.PHONY: persist-settings\n\n# Prerequisites target\n.make-prerequisites:\n\t@touch $@\n\n# Clean everything, persist settings and build dependencies if anything changed\nifneq ($(strip $(PREV_FINAL_CFLAGS)), $(strip $(FINAL_CFLAGS)))\n.make-prerequisites: persist-settings\nendif\n\nifneq ($(strip $(PREV_FINAL_LDFLAGS)), $(strip $(FINAL_LDFLAGS)))\n.make-prerequisites: persist-settings\nendif\n\n# redis-server\n$(REDIS_SERVER_NAME): $(REDIS_SERVER_OBJ)\n\t$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a ../deps/lua/src/liblua.a $(FINAL_LIBS)\n\n# redis-sentinel\n$(REDIS_SENTINEL_NAME): $(REDIS_SERVER_NAME)\n\t$(REDIS_INSTALL) $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME)\n\n# redis-cli\n$(REDIS_CLI_NAME): $(REDIS_CLI_OBJ)\n\t$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a ../deps/linenoise/linenoise.o $(FINAL_LIBS)\n\n# redis-benchmark\n$(REDIS_BENCHMARK_NAME): $(REDIS_BENCHMARK_OBJ)\n\t$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a $(FINAL_LIBS)\n\n# redis-check-dump\n$(REDIS_CHECK_DUMP_NAME): $(REDIS_CHECK_DUMP_OBJ)\n\t$(REDIS_LD) -o $@ $^ $(FINAL_LIBS)\n\n# redis-check-aof\n$(REDIS_CHECK_AOF_NAME): $(REDIS_CHECK_AOF_OBJ)\n\t$(REDIS_LD) -o $@ $^ $(FINAL_LIBS)\n\n# Because the jemalloc.h header is generated as a part of the jemalloc build,\n# building it should complete before building any other object. Instead of\n# depending on a single artifact, build all dependencies first.\n%.o: %.c .make-prerequisites\n\t$(REDIS_CC) -c $<\n\nclean:\n\trm -rf $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_DUMP_NAME) $(REDIS_CHECK_AOF_NAME) *.o *.gcda *.gcno *.gcov redis.info lcov-html\n\n.PHONY: clean\n\ndistclean: clean\n\t-(cd ../deps && $(MAKE) distclean)\n\t-(rm -f .make-*)\n\n.PHONY: distclean\n\ntest: $(REDIS_SERVER_NAME) $(REDIS_CHECK_AOF_NAME)\n\t@(cd ..; ./runtest)\n\ntest-sentinel: $(REDIS_SENTINEL_NAME)\n\t@(cd ..; ./runtest-sentinel)\n\ncheck: test\n\nlcov:\n\t$(MAKE) gcov\n\t@(set -e; cd ..; ./runtest --clients 1)\n\t@geninfo -o redis.info .\n\t@genhtml --legend -o lcov-html redis.info\n\n.PHONY: lcov\n\nbench: $(REDIS_BENCHMARK_NAME)\n\t./$(REDIS_BENCHMARK_NAME)\n\n32bit:\n\t@echo \"\"\n\t@echo \"WARNING: if it fails under Linux you probably need to install libc6-dev-i386\"\n\t@echo \"\"\n\t$(MAKE) CFLAGS=\"-m32\" LDFLAGS=\"-m32\"\n\ngcov:\n\t$(MAKE) REDIS_CFLAGS=\"-fprofile-arcs -ftest-coverage -DCOVERAGE_TEST\" REDIS_LDFLAGS=\"-fprofile-arcs -ftest-coverage\"\n\nnoopt:\n\t$(MAKE) OPTIMIZATION=\"-O0\"\n\nvalgrind:\n\t$(MAKE) OPTIMIZATION=\"-O0\" MALLOC=\"libc\"\n\nsrc/help.h:\n\t@../utils/generate-command-help.rb > help.h\n\ninstall: all\n\t@mkdir -p $(INSTALL_BIN)\n\t$(REDIS_INSTALL) $(REDIS_SERVER_NAME) $(INSTALL_BIN)\n\t$(REDIS_INSTALL) $(REDIS_BENCHMARK_NAME) $(INSTALL_BIN)\n\t$(REDIS_INSTALL) $(REDIS_CLI_NAME) $(INSTALL_BIN)\n\t$(REDIS_INSTALL) $(REDIS_CHECK_DUMP_NAME) $(INSTALL_BIN)\n\t$(REDIS_INSTALL) $(REDIS_CHECK_AOF_NAME) $(INSTALL_BIN)\n" }, { "path": "src/Makefile.dep", "content": "adlist.o: adlist.c adlist.h zmalloc.h\nae.o: ae.c ae.h zmalloc.h config.h ae_kqueue.c ae_epoll.c ae_select.c ae_evport.c\nae_epoll.o: ae_epoll.c\nae_evport.o: ae_evport.c\nae_kqueue.o: ae_kqueue.c\nae_select.o: ae_select.c\nanet.o: anet.c fmacros.h anet.h\naof.o: aof.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h bio.h\nbio.o: bio.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h bio.h\nbitops.o: bitops.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nblocked.o: blocked.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\ncluster.o: cluster.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h cluster.h endianconv.h\nconfig.o: config.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h cluster.h\ncrc16.o: crc16.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\ncrc64.o: crc64.c\ndb.o: db.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h cluster.h\ndebug.o: debug.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h sha1.h crc64.h bio.h\ndict.o: dict.c fmacros.h dict.h zmalloc.h redisassert.h\nendianconv.o: endianconv.c\nhyperloglog.o: hyperloglog.c redis.h fmacros.h config.h \\\n ../deps/lua/src/lua.h ../deps/lua/src/luaconf.h ae.h sds.h dict.h \\\n adlist.h zmalloc.h anet.h ziplist.h intset.h version.h util.h rdb.h \\\n rio.h\nintset.o: intset.c intset.h zmalloc.h endianconv.h config.h\nlzf_c.o: lzf_c.c lzfP.h\nlzf_d.o: lzf_d.c lzfP.h\nmemtest.o: memtest.c config.h\nmulti.o: multi.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nnetworking.o: networking.c redis.h fmacros.h config.h \\\n ../deps/lua/src/lua.h ../deps/lua/src/luaconf.h ae.h sds.h dict.h \\\n adlist.h zmalloc.h anet.h ziplist.h intset.h version.h util.h rdb.h \\\n rio.h\nnotify.o: notify.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nobject.o: object.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\npqsort.o: pqsort.c\npubsub.o: pubsub.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nrand.o: rand.c\nrdb.o: rdb.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h lzf.h zipmap.h \\\n endianconv.h\nredis-benchmark.o: redis-benchmark.c fmacros.h ae.h \\\n ../deps/hiredis/hiredis.h sds.h adlist.h zmalloc.h\nredis-check-aof.o: redis-check-aof.c fmacros.h config.h\nredis-check-dump.o: redis-check-dump.c lzf.h crc64.h\nredis-cli.o: redis-cli.c fmacros.h version.h ../deps/hiredis/hiredis.h \\\n sds.h zmalloc.h ../deps/linenoise/linenoise.h help.h anet.h ae.h\nredis.o: redis.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h cluster.h slowlog.h \\\n bio.h asciilogo.h\nrelease.o: release.c release.h version.h crc64.h\nreplication.o: replication.c redis.h fmacros.h config.h \\\n ../deps/lua/src/lua.h ../deps/lua/src/luaconf.h ae.h sds.h dict.h \\\n adlist.h zmalloc.h anet.h ziplist.h intset.h version.h util.h rdb.h \\\n rio.h\nrio.o: rio.c fmacros.h rio.h sds.h util.h crc64.h config.h redis.h \\\n ../deps/lua/src/lua.h ../deps/lua/src/luaconf.h ae.h dict.h adlist.h \\\n zmalloc.h anet.h ziplist.h intset.h version.h rdb.h\nscripting.o: scripting.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h sha1.h rand.h \\\n ../deps/lua/src/lauxlib.h ../deps/lua/src/lua.h ../deps/lua/src/lualib.h\nsds.o: sds.c sds.h zmalloc.h\nsentinel.o: sentinel.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h \\\n ../deps/hiredis/hiredis.h ../deps/hiredis/async.h \\\n ../deps/hiredis/hiredis.h\nsetproctitle.o: setproctitle.c\nsha1.o: sha1.c sha1.h config.h\nslowlog.o: slowlog.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h slowlog.h\nsort.o: sort.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h pqsort.h\nsyncio.o: syncio.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nt_hash.o: t_hash.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nt_list.o: t_list.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nt_set.o: t_set.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nt_string.o: t_string.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nt_zset.o: t_zset.c redis.h fmacros.h config.h ../deps/lua/src/lua.h \\\n ../deps/lua/src/luaconf.h ae.h sds.h dict.h adlist.h zmalloc.h anet.h \\\n ziplist.h intset.h version.h util.h rdb.h rio.h\nutil.o: util.c fmacros.h util.h sds.h\nziplist.o: ziplist.c zmalloc.h util.h sds.h ziplist.h endianconv.h \\\n config.h redisassert.h\nzipmap.o: zipmap.c zmalloc.h endianconv.h config.h\nzmalloc.o: zmalloc.c config.h zmalloc.h\n" }, { "path": "src/adlist.c", "content": "/* adlist.c - A generic doubly linked list implementation\n *\n * Copyright (c) 2006-2010, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \n#include \"adlist.h\"\n#include \"zmalloc.h\"\n\n/* Create a new list. The created list can be freed with\n * AlFreeList(), but private value of every node need to be freed\n * by the user before to call AlFreeList().\n *\n * On error, NULL is returned. Otherwise the pointer to the new list. */\n/*\n * 创建一个新的链表\n *\n * 创建成功返回链表，失败返回 NULL 。\n *\n * T = O(1)\n */\nlist *listCreate(void)\n{\n struct list *list;\n\n // 分配内存\n if ((list = zmalloc(sizeof(*list))) == NULL)\n return NULL;\n\n // 初始化属性\n list->head = list->tail = NULL;\n list->len = 0;\n list->dup = NULL;\n list->free = NULL;\n list->match = NULL;\n\n return list;\n}\n\n/* Free the whole list.\n *\n * This function can't fail. */\n/*\n * 释放整个链表，以及链表中所有节点\n *\n * T = O(N)\n */\nvoid listRelease(list *list)\n{\n unsigned long len;\n listNode *current, *next;\n\n // 指向头指针\n current = list->head;\n // 遍历整个链表\n len = list->len;\n while(len--) {\n next = current->next;\n\n // 如果有设置值释放函数，那么调用它\n if (list->free) list->free(current->value);\n\n // 释放节点结构\n zfree(current);\n\n current = next;\n }\n\n // 释放链表结构\n zfree(list);\n}\n\n/* Add a new node to the list, to head, contaning the specified 'value'\n * pointer as value.\n *\n * On error, NULL is returned and no operation is performed (i.e. the\n * list remains unaltered).\n * On success the 'list' pointer you pass to the function is returned. */\n/*\n * 将一个包含有给定值指针 value 的新节点添加到链表的表头\n *\n * 如果为新节点分配内存出错，那么不执行任何动作，仅返回 NULL\n *\n * 如果执行成功，返回传入的链表指针\n *\n * T = O(1)\n */\nlist *listAddNodeHead(list *list, void *value)\n{\n listNode *node;\n\n // 为节点分配内存\n if ((node = zmalloc(sizeof(*node))) == NULL)\n return NULL;\n\n // 保存值指针\n node->value = value;\n\n // 添加节点到空链表\n if (list->len == 0) {\n list->head = list->tail = node;\n node->prev = node->next = NULL;\n // 添加节点到非空链表\n } else {\n node->prev = NULL;\n node->next = list->head;\n list->head->prev = node;\n list->head = node;\n }\n\n // 更新链表节点数\n list->len++;\n\n return list;\n}\n\n/* Add a new node to the list, to tail, containing the specified 'value'\n * pointer as value.\n *\n * On error, NULL is returned and no operation is performed (i.e. the\n * list remains unaltered).\n * On success the 'list' pointer you pass to the function is returned. */\n/*\n * 将一个包含有给定值指针 value 的新节点添加到链表的表尾\n *\n * 如果为新节点分配内存出错，那么不执行任何动作，仅返回 NULL\n *\n * 如果执行成功，返回传入的链表指针\n *\n * T = O(1)\n */\nlist *listAddNodeTail(list *list, void *value)\n{\n listNode *node;\n\n // 为新节点分配内存\n if ((node = zmalloc(sizeof(*node))) == NULL)\n return NULL;\n\n // 保存值指针\n node->value = value;\n\n // 目标链表为空\n if (list->len == 0) {\n list->head = list->tail = node;\n node->prev = node->next = NULL;\n // 目标链表非空\n } else {\n node->prev = list->tail;\n node->next = NULL;\n list->tail->next = node;\n list->tail = node;\n }\n\n // 更新链表节点数\n list->len++;\n\n return list;\n}\n\n/*\n * 创建一个包含值 value 的新节点，并将它插入到 old_node 的之前或之后\n *\n * 如果 after 为 0 ，将新节点插入到 old_node 之前。\n * 如果 after 为 1 ，将新节点插入到 old_node 之后。\n *\n * T = O(1)\n */\nlist *listInsertNode(list *list, listNode *old_node, void *value, int after) {\n listNode *node;\n\n // 创建新节点\n if ((node = zmalloc(sizeof(*node))) == NULL)\n return NULL;\n\n // 保存值\n node->value = value;\n\n // 将新节点添加到给定节点之后\n if (after) {\n node->prev = old_node;\n node->next = old_node->next;\n // 给定节点是原表尾节点\n if (list->tail == old_node) {\n list->tail = node;\n }\n // 将新节点添加到给定节点之前\n } else {\n node->next = old_node;\n node->prev = old_node->prev;\n // 给定节点是原表头节点\n if (list->head == old_node) {\n list->head = node;\n }\n }\n\n // 更新新节点的前置指针\n if (node->prev != NULL) {\n node->prev->next = node;\n }\n // 更新新节点的后置指针\n if (node->next != NULL) {\n node->next->prev = node;\n }\n\n // 更新链表节点数\n list->len++;\n\n return list;\n}\n\n/* Remove the specified node from the specified list.\n * It's up to the caller to free the private value of the node.\n *\n * This function can't fail. */\n/*\n * 从链表 list 中删除给定节点 node \n * \n * 对节点私有值(private value of the node)的释放工作由调用者进行。\n *\n * T = O(1)\n */\nvoid listDelNode(list *list, listNode *node)\n{\n // 调整前置节点的指针\n if (node->prev)\n node->prev->next = node->next;\n else\n list->head = node->next;\n\n // 调整后置节点的指针\n if (node->next)\n node->next->prev = node->prev;\n else\n list->tail = node->prev;\n\n // 释放值\n if (list->free) list->free(node->value);\n\n // 释放节点\n zfree(node);\n\n // 链表数减一\n list->len--;\n}\n\n/* Returns a list iterator 'iter'. After the initialization every\n * call to listNext() will return the next element of the list.\n *\n * This function can't fail. */\n/*\n * 为给定链表创建一个迭代器，\n * 之后每次对这个迭代器调用 listNext 都返回被迭代到的链表节点\n *\n * direction 参数决定了迭代器的迭代方向：\n * AL_START_HEAD ：从表头向表尾迭代\n * AL_START_TAIL ：从表尾想表头迭代\n *\n * T = O(1)\n */\nlistIter *listGetIterator(list *list, int direction)\n{\n // 为迭代器分配内存\n listIter *iter;\n if ((iter = zmalloc(sizeof(*iter))) == NULL) return NULL;\n\n // 根据迭代方向，设置迭代器的起始节点\n if (direction == AL_START_HEAD)\n iter->next = list->head;\n else\n iter->next = list->tail;\n\n // 记录迭代方向\n iter->direction = direction;\n\n return iter;\n}\n\n/* Release the iterator memory */\n/*\n * 释放迭代器\n *\n * T = O(1)\n */\nvoid listReleaseIterator(listIter *iter) {\n zfree(iter);\n}\n\n/* Create an iterator in the list private iterator structure */\n/*\n * 将迭代器的方向设置为 AL_START_HEAD ，\n * 并将迭代指针重新指向表头节点。\n *\n * T = O(1)\n */\nvoid listRewind(list *list, listIter *li) {\n li->next = list->head;\n li->direction = AL_START_HEAD;\n}\n\n/*\n * 将迭代器的方向设置为 AL_START_TAIL ，\n * 并将迭代指针重新指向表尾节点。\n *\n * T = O(1)\n */\nvoid listRewindTail(list *list, listIter *li) {\n li->next = list->tail;\n li->direction = AL_START_TAIL;\n}\n\n/* Return the next element of an iterator.\n * It's valid to remove the currently returned element using\n * listDelNode(), but not to remove other elements.\n *\n * The function returns a pointer to the next element of the list,\n * or NULL if there are no more elements, so the classical usage patter\n * is:\n *\n * iter = listGetIterator(list,);\n * while ((node = listNext(iter)) != NULL) {\n * doSomethingWith(listNodeValue(node));\n * }\n *\n * */\n/*\n * 返回迭代器当前所指向的节点。\n *\n * 删除当前节点是允许的，但不能修改链表里的其他节点。\n *\n * 函数要么返回一个节点，要么返回 NULL ，常见的用法是：\n *\n * iter = listGetIterator(list,);\n * while ((node = listNext(iter)) != NULL) {\n * doSomethingWith(listNodeValue(node));\n * }\n *\n * T = O(1)\n */\nlistNode *listNext(listIter *iter)\n{\n listNode *current = iter->next;\n\n if (current != NULL) {\n // 根据方向选择下一个节点\n if (iter->direction == AL_START_HEAD)\n // 保存下一个节点，防止当前节点被删除而造成指针丢失\n iter->next = current->next;\n else\n // 保存下一个节点，防止当前节点被删除而造成指针丢失\n iter->next = current->prev;\n }\n\n return current;\n}\n\n/* Duplicate the whole list. On out of memory NULL is returned.\n * On success a copy of the original list is returned.\n *\n * The 'Dup' method set with listSetDupMethod() function is used\n * to copy the node value. Otherwise the same pointer value of\n * the original node is used as value of the copied node.\n *\n * The original list both on success or error is never modified. */\n/*\n * 复制整个链表。\n *\n * 复制成功返回输入链表的副本，\n * 如果因为内存不足而造成复制失败，返回 NULL 。\n *\n * 如果链表有设置值复制函数 dup ，那么对值的复制将使用复制函数进行，\n * 否则，新节点将和旧节点共享同一个指针。\n *\n * 无论复制是成功还是失败，输入节点都不会修改。\n *\n * T = O(N)\n */\nlist *listDup(list *orig)\n{\n list *copy;\n listIter *iter;\n listNode *node;\n\n // 创建新链表\n if ((copy = listCreate()) == NULL)\n return NULL;\n\n // 设置节点值处理函数\n copy->dup = orig->dup;\n copy->free = orig->free;\n copy->match = orig->match;\n\n // 迭代整个输入链表\n iter = listGetIterator(orig, AL_START_HEAD);\n while((node = listNext(iter)) != NULL) {\n void *value;\n\n // 复制节点值到新节点\n if (copy->dup) {\n value = copy->dup(node->value);\n if (value == NULL) {\n listRelease(copy);\n listReleaseIterator(iter);\n return NULL;\n }\n } else\n value = node->value;\n\n // 将节点添加到链表\n if (listAddNodeTail(copy, value) == NULL) {\n listRelease(copy);\n listReleaseIterator(iter);\n return NULL;\n }\n }\n\n // 释放迭代器\n listReleaseIterator(iter);\n\n // 返回副本\n return copy;\n}\n\n/* Search the list for a node matching a given key.\n * The match is performed using the 'match' method\n * set with listSetMatchMethod(). If no 'match' method\n * is set, the 'value' pointer of every node is directly\n * compared with the 'key' pointer.\n *\n * On success the first matching node pointer is returned\n * (search starts from head). If no matching node exists\n * NULL is returned. */\n/* \n * 查找链表 list 中值和 key 匹配的节点。\n * \n * 对比操作由链表的 match 函数负责进行，\n * 如果没有设置 match 函数，\n * 那么直接通过对比值的指针来决定是否匹配。\n *\n * 如果匹配成功，那么第一个匹配的节点会被返回。\n * 如果没有匹配任何节点，那么返回 NULL 。\n *\n * T = O(N)\n */\nlistNode *listSearchKey(list *list, void *key)\n{\n listIter *iter;\n listNode *node;\n\n // 迭代整个链表\n iter = listGetIterator(list, AL_START_HEAD);\n while((node = listNext(iter)) != NULL) {\n \n // 对比\n if (list->match) {\n if (list->match(node->value, key)) {\n listReleaseIterator(iter);\n // 找到\n return node;\n }\n } else {\n if (key == node->value) {\n listReleaseIterator(iter);\n // 找到\n return node;\n }\n }\n }\n \n listReleaseIterator(iter);\n\n // 未找到\n return NULL;\n}\n\n/* Return the element at the specified zero-based index\n * where 0 is the head, 1 is the element next to head\n * and so on. Negative integers are used in order to count\n * from the tail, -1 is the last element, -2 the penultimate\n * and so on. If the index is out of range NULL is returned. */\n/*\n * 返回链表在给定索引上的值。\n *\n * 索引以 0 为起始，也可以是负数， -1 表示链表最后一个节点，诸如此类。\n *\n * 如果索引超出范围（out of range），返回 NULL 。\n *\n * T = O(N)\n */\nlistNode *listIndex(list *list, long index) {\n listNode *n;\n\n // 如果索引为负数，从表尾开始查找\n if (index < 0) {\n index = (-index)-1;\n n = list->tail;\n while(index-- && n) n = n->prev;\n // 如果索引为正数，从表头开始查找\n } else {\n n = list->head;\n while(index-- && n) n = n->next;\n }\n\n return n;\n}\n\n/* Rotate the list removing the tail node and inserting it to the head. */\n/*\n * 取出链表的表尾节点，并将它移动到表头，成为新的表头节点。\n *\n * T = O(1)\n */\nvoid listRotate(list *list) {\n listNode *tail = list->tail;\n\n if (listLength(list) <= 1) return;\n\n /* Detach current tail */\n // 取出表尾节点\n list->tail = tail->prev;\n list->tail->next = NULL;\n\n /* Move it as head */\n // 插入到表头\n list->head->prev = tail;\n tail->prev = NULL;\n tail->next = list->head;\n list->head = tail;\n}\n" }, { "path": "src/adlist.h", "content": "/* adlist.h - A generic doubly linked list implementation\n *\n * Copyright (c) 2006-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __ADLIST_H__\n#define __ADLIST_H__\n\n/* Node, List, and Iterator are the only data structures used currently. */\n\n/*\n * 双端链表节点\n */\ntypedef struct listNode {\n\n // 前置节点\n struct listNode *prev;\n\n // 后置节点\n struct listNode *next;\n\n // 节点的值\n void *value;\n\n} listNode;\n\n/*\n * 双端链表迭代器\n */\ntypedef struct listIter {\n\n // 当前迭代到的节点\n listNode *next;\n\n // 迭代的方向\n int direction;\n\n} listIter;\n\n/*\n * 双端链表结构\n */\ntypedef struct list {\n\n // 表头节点\n listNode *head;\n\n // 表尾节点\n listNode *tail;\n\n // 节点值复制函数\n void *(*dup)(void *ptr);\n\n // 节点值释放函数\n void (*free)(void *ptr);\n\n // 节点值对比函数\n int (*match)(void *ptr, void *key);\n\n // 链表所包含的节点数量\n unsigned long len;\n\n} list;\n\n/* Functions implemented as macros */\n// 返回给定链表所包含的节点数量\n// T = O(1)\n#define listLength(l) ((l)->len)\n// 返回给定链表的表头节点\n// T = O(1)\n#define listFirst(l) ((l)->head)\n// 返回给定链表的表尾节点\n// T = O(1)\n#define listLast(l) ((l)->tail)\n// 返回给定节点的前置节点\n// T = O(1)\n#define listPrevNode(n) ((n)->prev)\n// 返回给定节点的后置节点\n// T = O(1)\n#define listNextNode(n) ((n)->next)\n// 返回给定节点的值\n// T = O(1)\n#define listNodeValue(n) ((n)->value)\n\n// 将链表 l 的值复制函数设置为 m\n// T = O(1)\n#define listSetDupMethod(l,m) ((l)->dup = (m))\n// 将链表 l 的值释放函数设置为 m\n// T = O(1)\n#define listSetFreeMethod(l,m) ((l)->free = (m))\n// 将链表的对比函数设置为 m\n// T = O(1)\n#define listSetMatchMethod(l,m) ((l)->match = (m))\n\n// 返回给定链表的值复制函数\n// T = O(1)\n#define listGetDupMethod(l) ((l)->dup)\n// 返回给定链表的值释放函数\n// T = O(1)\n#define listGetFree(l) ((l)->free)\n// 返回给定链表的值对比函数\n// T = O(1)\n#define listGetMatchMethod(l) ((l)->match)\n\n/* Prototypes */\nlist *listCreate(void);\nvoid listRelease(list *list);\nlist *listAddNodeHead(list *list, void *value);\nlist *listAddNodeTail(list *list, void *value);\nlist *listInsertNode(list *list, listNode *old_node, void *value, int after);\nvoid listDelNode(list *list, listNode *node);\nlistIter *listGetIterator(list *list, int direction);\nlistNode *listNext(listIter *iter);\nvoid listReleaseIterator(listIter *iter);\nlist *listDup(list *orig);\nlistNode *listSearchKey(list *list, void *key);\nlistNode *listIndex(list *list, long index);\nvoid listRewind(list *list, listIter *li);\nvoid listRewindTail(list *list, listIter *li);\nvoid listRotate(list *list);\n\n/* Directions for iterators \n *\n * 迭代器进行迭代的方向\n */\n// 从表头向表尾进行迭代\n#define AL_START_HEAD 0\n// 从表尾到表头进行迭代\n#define AL_START_TAIL 1\n\n#endif /* __ADLIST_H__ */\n" }, { "path": "src/ae.c", "content": "/* A simple event-driven programming library. Originally I wrote this code\n * for the Jim's event-loop (Jim is a Tcl interpreter) but later translated\n * it in form of a library for easy reuse.\n *\n * Copyright (c) 2006-2010, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"ae.h\"\n#include \"zmalloc.h\"\n#include \"config.h\"\n\n/* Include the best multiplexing layer supported by this system.\n * The following should be ordered by performances, descending. */\n#ifdef HAVE_EVPORT\n#include \"ae_evport.c\"\n#else\n #ifdef HAVE_EPOLL\n #include \"ae_epoll.c\"\n #else\n #ifdef HAVE_KQUEUE\n #include \"ae_kqueue.c\"\n #else\n #include \"ae_select.c\"\n #endif\n #endif\n#endif\n\n/*\n * 初始化事件处理器状态\n */\naeEventLoop *aeCreateEventLoop(int setsize) {\n aeEventLoop *eventLoop;\n int i;\n\n // 创建事件状态结构\n if ((eventLoop = zmalloc(sizeof(*eventLoop))) == NULL) goto err;\n\n // 初始化文件事件结构和已就绪文件事件结构数组\n eventLoop->events = zmalloc(sizeof(aeFileEvent)*setsize);\n eventLoop->fired = zmalloc(sizeof(aeFiredEvent)*setsize);\n if (eventLoop->events == NULL || eventLoop->fired == NULL) goto err;\n // 设置数组大小\n eventLoop->setsize = setsize;\n // 初始化执行最近一次执行时间\n eventLoop->lastTime = time(NULL);\n\n // 初始化时间事件结构\n eventLoop->timeEventHead = NULL;\n eventLoop->timeEventNextId = 0;\n\n eventLoop->stop = 0;\n eventLoop->maxfd = -1;\n eventLoop->beforesleep = NULL;\n if (aeApiCreate(eventLoop) == -1) goto err;\n\n /* Events with mask == AE_NONE are not set. So let's initialize the\n * vector with it. */\n // 初始化监听事件\n for (i = 0; i < setsize; i++)\n eventLoop->events[i].mask = AE_NONE;\n\n // 返回事件循环\n return eventLoop;\n\nerr:\n if (eventLoop) {\n zfree(eventLoop->events);\n zfree(eventLoop->fired);\n zfree(eventLoop);\n }\n return NULL;\n}\n\n/* Return the current set size. */\n// 返回当前事件槽大小\nint aeGetSetSize(aeEventLoop *eventLoop) {\n return eventLoop->setsize;\n}\n\n/* Resize the maximum set size of the event loop.\n *\n * 调整事件槽的大小\n *\n * If the requested set size is smaller than the current set size, but\n * there is already a file descriptor in use that is >= the requested\n * set size minus one, AE_ERR is returned and the operation is not\n * performed at all.\n *\n * 如果尝试调整大小为 setsize ，但是有 >= setsize 的文件描述符存在\n * 那么返回 AE_ERR ，不进行任何动作。\n *\n * Otherwise AE_OK is returned and the operation is successful. \n *\n * 否则，执行大小调整操作，并返回 AE_OK 。\n */\nint aeResizeSetSize(aeEventLoop *eventLoop, int setsize) {\n int i;\n\n if (setsize == eventLoop->setsize) return AE_OK;\n if (eventLoop->maxfd >= setsize) return AE_ERR;\n if (aeApiResize(eventLoop,setsize) == -1) return AE_ERR;\n\n eventLoop->events = zrealloc(eventLoop->events,sizeof(aeFileEvent)*setsize);\n eventLoop->fired = zrealloc(eventLoop->fired,sizeof(aeFiredEvent)*setsize);\n eventLoop->setsize = setsize;\n\n /* Make sure that if we created new slots, they are initialized with\n * an AE_NONE mask. */\n for (i = eventLoop->maxfd+1; i < setsize; i++)\n eventLoop->events[i].mask = AE_NONE;\n return AE_OK;\n}\n\n/*\n * 删除事件处理器\n */\nvoid aeDeleteEventLoop(aeEventLoop *eventLoop) {\n aeApiFree(eventLoop);\n zfree(eventLoop->events);\n zfree(eventLoop->fired);\n zfree(eventLoop);\n}\n\n/*\n * 停止事件处理器\n */\nvoid aeStop(aeEventLoop *eventLoop) {\n eventLoop->stop = 1;\n}\n\n/*\n * 根据 mask 参数的值，监听 fd 文件的状态，\n * 当 fd 可用时，执行 proc 函数\n */\nint aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask,\n aeFileProc *proc, void *clientData)\n{\n if (fd >= eventLoop->setsize) {\n errno = ERANGE;\n return AE_ERR;\n }\n\n if (fd >= eventLoop->setsize) return AE_ERR;\n\n // 取出文件事件结构\n aeFileEvent *fe = &eventLoop->events[fd];\n\n // 监听指定 fd 的指定事件\n if (aeApiAddEvent(eventLoop, fd, mask) == -1)\n return AE_ERR;\n\n // 设置文件事件类型，以及事件的处理器\n fe->mask |= mask;\n if (mask & AE_READABLE) fe->rfileProc = proc;\n if (mask & AE_WRITABLE) fe->wfileProc = proc;\n\n // 私有数据\n fe->clientData = clientData;\n\n // 如果有需要，更新事件处理器的最大 fd\n if (fd > eventLoop->maxfd)\n eventLoop->maxfd = fd;\n\n return AE_OK;\n}\n\n/*\n * 将 fd 从 mask 指定的监听队列中删除\n */\nvoid aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask)\n{\n if (fd >= eventLoop->setsize) return;\n\n // 取出文件事件结构\n aeFileEvent *fe = &eventLoop->events[fd];\n\n // 未设置监听的事件类型，直接返回\n if (fe->mask == AE_NONE) return;\n\n // 计算新掩码\n fe->mask = fe->mask & (~mask);\n if (fd == eventLoop->maxfd && fe->mask == AE_NONE) {\n /* Update the max fd */\n int j;\n\n for (j = eventLoop->maxfd-1; j >= 0; j--)\n if (eventLoop->events[j].mask != AE_NONE) break;\n eventLoop->maxfd = j;\n }\n\n // 取消对给定 fd 的给定事件的监视\n aeApiDelEvent(eventLoop, fd, mask);\n}\n\n/*\n * 获取给定 fd 正在监听的事件类型\n */\nint aeGetFileEvents(aeEventLoop *eventLoop, int fd) {\n if (fd >= eventLoop->setsize) return 0;\n aeFileEvent *fe = &eventLoop->events[fd];\n\n return fe->mask;\n}\n\n/*\n * 取出当前时间的秒和毫秒，\n * 并分别将它们保存到 seconds 和 milliseconds 参数中\n */\nstatic void aeGetTime(long *seconds, long *milliseconds)\n{\n struct timeval tv;\n\n gettimeofday(&tv, NULL);\n *seconds = tv.tv_sec;\n *milliseconds = tv.tv_usec/1000;\n}\n\n/*\n * 在当前时间上加上 milliseconds 毫秒，\n * 并且将加上之后的秒数和毫秒数分别保存在 sec 和 ms 指针中。\n */\nstatic void aeAddMillisecondsToNow(long long milliseconds, long *sec, long *ms) {\n long cur_sec, cur_ms, when_sec, when_ms;\n\n // 获取当前时间\n aeGetTime(&cur_sec, &cur_ms);\n\n // 计算增加 milliseconds 之后的秒数和毫秒数\n when_sec = cur_sec + milliseconds/1000;\n when_ms = cur_ms + milliseconds%1000;\n\n // 进位：\n // 如果 when_ms 大于等于 1000\n // 那么将 when_sec 增大一秒\n if (when_ms >= 1000) {\n when_sec ++;\n when_ms -= 1000;\n }\n\n // 保存到指针中\n *sec = when_sec;\n *ms = when_ms;\n}\n\n/*\n * 创建时间事件\n */\nlong long aeCreateTimeEvent(aeEventLoop *eventLoop, long long milliseconds,\n aeTimeProc *proc, void *clientData,\n aeEventFinalizerProc *finalizerProc)\n{\n // 更新时间计数器\n long long id = eventLoop->timeEventNextId++;\n\n // 创建时间事件结构\n aeTimeEvent *te;\n\n te = zmalloc(sizeof(*te));\n if (te == NULL) return AE_ERR;\n\n // 设置 ID\n te->id = id;\n\n // 设定处理事件的时间\n aeAddMillisecondsToNow(milliseconds,&te->when_sec,&te->when_ms);\n // 设置事件处理器\n te->timeProc = proc;\n te->finalizerProc = finalizerProc;\n // 设置私有数据\n te->clientData = clientData;\n\n // 将新事件放入表头\n te->next = eventLoop->timeEventHead;\n eventLoop->timeEventHead = te;\n\n return id;\n}\n\n/*\n * 删除给定 id 的时间事件\n */\nint aeDeleteTimeEvent(aeEventLoop *eventLoop, long long id)\n{\n aeTimeEvent *te, *prev = NULL;\n\n // 遍历链表\n te = eventLoop->timeEventHead;\n while(te) {\n\n // 发现目标事件，删除\n if (te->id == id) {\n\n if (prev == NULL)\n eventLoop->timeEventHead = te->next;\n else\n prev->next = te->next;\n\n // 执行清理处理器\n if (te->finalizerProc)\n te->finalizerProc(eventLoop, te->clientData);\n\n // 释放时间事件\n zfree(te);\n\n return AE_OK;\n }\n prev = te;\n te = te->next;\n }\n\n return AE_ERR; /* NO event with the specified ID found */\n}\n\n/* Search the first timer to fire.\n * This operation is useful to know how many time the select can be\n * put in sleep without to delay any event.\n * If there are no timers NULL is returned.\n *\n * Note that's O(N) since time events are unsorted.\n * Possible optimizations (not needed by Redis so far, but...):\n * 1) Insert the event in order, so that the nearest is just the head.\n * Much better but still insertion or deletion of timers is O(N).\n * 2) Use a skiplist to have this operation as O(1) and insertion as O(log(N)).\n */\n// 寻找里目前时间最近的时间事件\n// 因为链表是乱序的，所以查找复杂度为 O（N）\nstatic aeTimeEvent *aeSearchNearestTimer(aeEventLoop *eventLoop)\n{\n aeTimeEvent *te = eventLoop->timeEventHead;\n aeTimeEvent *nearest = NULL;\n\n while(te) {\n if (!nearest || te->when_sec < nearest->when_sec ||\n (te->when_sec == nearest->when_sec &&\n te->when_ms < nearest->when_ms))\n nearest = te;\n te = te->next;\n }\n return nearest;\n}\n\n/* Process time events\n *\n * 处理所有已到达的时间事件\n */\nstatic int processTimeEvents(aeEventLoop *eventLoop) {\n int processed = 0;\n aeTimeEvent *te;\n long long maxId;\n time_t now = time(NULL);\n\n /* If the system clock is moved to the future, and then set back to the\n * right value, time events may be delayed in a random way. Often this\n * means that scheduled operations will not be performed soon enough.\n *\n * Here we try to detect system clock skews, and force all the time\n * events to be processed ASAP when this happens: the idea is that\n * processing events earlier is less dangerous than delaying them\n * indefinitely, and practice suggests it is. */\n // 通过重置事件的运行时间，\n // 防止因时间穿插（skew）而造成的事件处理混乱\n if (now < eventLoop->lastTime) {\n te = eventLoop->timeEventHead;\n while(te) {\n te->when_sec = 0;\n te = te->next;\n }\n }\n // 更新最后一次处理时间事件的时间\n eventLoop->lastTime = now;\n\n // 遍历链表\n // 执行那些已经到达的事件\n te = eventLoop->timeEventHead;\n maxId = eventLoop->timeEventNextId-1;\n while(te) {\n long now_sec, now_ms;\n long long id;\n\n // 跳过无效事件\n if (te->id > maxId) {\n te = te->next;\n continue;\n }\n \n // 获取当前时间\n aeGetTime(&now_sec, &now_ms);\n\n // 如果当前时间等于或等于事件的执行时间，那么说明事件已到达，执行这个事件\n if (now_sec > te->when_sec ||\n (now_sec == te->when_sec && now_ms >= te->when_ms))\n {\n int retval;\n\n id = te->id;\n // 执行事件处理器，并获取返回值\n retval = te->timeProc(eventLoop, id, te->clientData);\n processed++;\n /* After an event is processed our time event list may\n * no longer be the same, so we restart from head.\n * Still we make sure to don't process events registered\n * by event handlers itself in order to don't loop forever.\n * To do so we saved the max ID we want to handle.\n *\n * FUTURE OPTIMIZATIONS:\n * Note that this is NOT great algorithmically. Redis uses\n * a single time event so it's not a problem but the right\n * way to do this is to add the new elements on head, and\n * to flag deleted elements in a special way for later\n * deletion (putting references to the nodes to delete into\n * another linked list). */\n\n // 记录是否有需要循环执行这个事件时间\n if (retval != AE_NOMORE) {\n // 是的， retval 毫秒之后继续执行这个时间事件\n aeAddMillisecondsToNow(retval,&te->when_sec,&te->when_ms);\n } else {\n // 不，将这个事件删除\n aeDeleteTimeEvent(eventLoop, id);\n }\n\n // 因为执行事件之后，事件列表可能已经被改变了\n // 因此需要将 te 放回表头，继续开始执行事件\n te = eventLoop->timeEventHead;\n } else {\n te = te->next;\n }\n }\n return processed;\n}\n\n/* Process every pending time event, then every pending file event\n * (that may be registered by time event callbacks just processed).\n *\n * 处理所有已到达的时间事件，以及所有已就绪的文件事件。\n *\n * Without special flags the function sleeps until some file event\n * fires, or when the next time event occurs (if any).\n *\n * 如果不传入特殊 flags 的话，那么函数睡眠直到文件事件就绪，\n * 或者下个时间事件到达（如果有的话）。\n *\n * If flags is 0, the function does nothing and returns.\n * 如果 flags 为 0 ，那么函数不作动作，直接返回。\n *\n * if flags has AE_ALL_EVENTS set, all the kind of events are processed.\n * 如果 flags 包含 AE_ALL_EVENTS ，所有类型的事件都会被处理。\n *\n * if flags has AE_FILE_EVENTS set, file events are processed.\n * 如果 flags 包含 AE_FILE_EVENTS ，那么处理文件事件。\n *\n * if flags has AE_TIME_EVENTS set, time events are processed.\n * 如果 flags 包含 AE_TIME_EVENTS ，那么处理时间事件。\n *\n * if flags has AE_DONT_WAIT set the function returns ASAP until all\n * the events that's possible to process without to wait are processed.\n * 如果 flags 包含 AE_DONT_WAIT ，\n * 那么函数在处理完所有不许阻塞的事件之后，即刻返回。\n *\n * The function returns the number of events processed. \n * 函数的返回值为已处理事件的数量\n */\nint aeProcessEvents(aeEventLoop *eventLoop, int flags)\n{\n int processed = 0, numevents;\n\n /* Nothing to do? return ASAP */\n if (!(flags & AE_TIME_EVENTS) && !(flags & AE_FILE_EVENTS)) return 0;\n\n /* Note that we want call select() even if there are no\n * file events to process as long as we want to process time\n * events, in order to sleep until the next time event is ready\n * to fire. */\n if (eventLoop->maxfd != -1 ||\n ((flags & AE_TIME_EVENTS) && !(flags & AE_DONT_WAIT))) {\n int j;\n aeTimeEvent *shortest = NULL;\n struct timeval tv, *tvp;\n\n // 获取最近的时间事件\n if (flags & AE_TIME_EVENTS && !(flags & AE_DONT_WAIT))\n shortest = aeSearchNearestTimer(eventLoop);\n if (shortest) {\n // 如果时间事件存在的话\n // 那么根据最近可执行时间事件和现在时间的时间差来决定文件事件的阻塞时间\n long now_sec, now_ms;\n\n /* Calculate the time missing for the nearest\n * timer to fire. */\n // 计算距今最近的时间事件还要多久才能达到\n // 并将该时间距保存在 tv 结构中\n aeGetTime(&now_sec, &now_ms);\n tvp = &tv;\n tvp->tv_sec = shortest->when_sec - now_sec;\n if (shortest->when_ms < now_ms) {\n tvp->tv_usec = ((shortest->when_ms+1000) - now_ms)*1000;\n tvp->tv_sec --;\n } else {\n tvp->tv_usec = (shortest->when_ms - now_ms)*1000;\n }\n\n // 时间差小于 0 ，说明事件已经可以执行了，将秒和毫秒设为 0 （不阻塞）\n if (tvp->tv_sec < 0) tvp->tv_sec = 0;\n if (tvp->tv_usec < 0) tvp->tv_usec = 0;\n } else {\n \n // 执行到这一步，说明没有时间事件\n // 那么根据 AE_DONT_WAIT 是否设置来决定是否阻塞，以及阻塞的时间长度\n\n /* If we have to check for events but need to return\n * ASAP because of AE_DONT_WAIT we need to set the timeout\n * to zero */\n if (flags & AE_DONT_WAIT) {\n // 设置文件事件不阻塞\n tv.tv_sec = tv.tv_usec = 0;\n tvp = &tv;\n } else {\n /* Otherwise we can block */\n // 文件事件可以阻塞直到有事件到达为止\n tvp = NULL; /* wait forever */\n }\n }\n\n // 处理文件事件，阻塞时间由 tvp 决定\n numevents = aeApiPoll(eventLoop, tvp);\n for (j = 0; j < numevents; j++) {\n // 从已就绪数组中获取事件\n aeFileEvent *fe = &eventLoop->events[eventLoop->fired[j].fd];\n\n int mask = eventLoop->fired[j].mask;\n int fd = eventLoop->fired[j].fd;\n int rfired = 0;\n\n /* note the fe->mask & mask & ... code: maybe an already processed\n * event removed an element that fired and we still didn't\n * processed, so we check if the event is still valid. */\n // 读事件\n if (fe->mask & mask & AE_READABLE) {\n // rfired 确保读/写事件只能执行其中一个\n rfired = 1;\n fe->rfileProc(eventLoop,fd,fe->clientData,mask);\n }\n // 写事件\n if (fe->mask & mask & AE_WRITABLE) {\n if (!rfired || fe->wfileProc != fe->rfileProc)\n fe->wfileProc(eventLoop,fd,fe->clientData,mask);\n }\n\n processed++;\n }\n }\n\n /* Check time events */\n // 执行时间事件\n if (flags & AE_TIME_EVENTS)\n processed += processTimeEvents(eventLoop);\n\n return processed; /* return the number of processed file/time events */\n}\n\n/* Wait for milliseconds until the given file descriptor becomes\n * writable/readable/exception \n *\n * 在给定毫秒内等待，直到 fd 变成可写、可读或异常\n */\nint aeWait(int fd, int mask, long long milliseconds) {\n struct pollfd pfd;\n int retmask = 0, retval;\n\n memset(&pfd, 0, sizeof(pfd));\n pfd.fd = fd;\n if (mask & AE_READABLE) pfd.events |= POLLIN;\n if (mask & AE_WRITABLE) pfd.events |= POLLOUT;\n\n if ((retval = poll(&pfd, 1, milliseconds))== 1) {\n if (pfd.revents & POLLIN) retmask |= AE_READABLE;\n if (pfd.revents & POLLOUT) retmask |= AE_WRITABLE;\n\tif (pfd.revents & POLLERR) retmask |= AE_WRITABLE;\n if (pfd.revents & POLLHUP) retmask |= AE_WRITABLE;\n return retmask;\n } else {\n return retval;\n }\n}\n\n/*\n * 事件处理器的主循环\n */\nvoid aeMain(aeEventLoop *eventLoop) {\n\n eventLoop->stop = 0;\n\n while (!eventLoop->stop) {\n\n // 如果有需要在事件处理前执行的函数，那么运行它\n if (eventLoop->beforesleep != NULL)\n eventLoop->beforesleep(eventLoop);\n\n // 开始处理事件\n aeProcessEvents(eventLoop, AE_ALL_EVENTS);\n }\n}\n\n/*\n * 返回所使用的多路复用库的名称\n */\nchar *aeGetApiName(void) {\n return aeApiName();\n}\n\n/*\n * 设置处理事件前需要被执行的函数\n */\nvoid aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep) {\n eventLoop->beforesleep = beforesleep;\n}\n" }, { "path": "src/ae.h", "content": "/* A simple event-driven programming library. Originally I wrote this code\n * for the Jim's event-loop (Jim is a Tcl interpreter) but later translated\n * it in form of a library for easy reuse.\n *\n * Copyright (c) 2006-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __AE_H__\n#define __AE_H__\n\n/*\n * 事件执行状态\n */\n// 成功\n#define AE_OK 0\n// 出错\n#define AE_ERR -1\n\n/*\n * 文件事件状态\n */\n// 未设置\n#define AE_NONE 0\n// 可读\n#define AE_READABLE 1\n// 可写\n#define AE_WRITABLE 2\n\n/*\n * 时间处理器的执行 flags\n */\n// 文件事件\n#define AE_FILE_EVENTS 1\n// 时间事件\n#define AE_TIME_EVENTS 2\n// 所有事件\n#define AE_ALL_EVENTS (AE_FILE_EVENTS|AE_TIME_EVENTS)\n// 不阻塞，也不进行等待\n#define AE_DONT_WAIT 4\n\n/*\n * 决定时间事件是否要持续执行的 flag\n */\n#define AE_NOMORE -1\n\n/* Macros */\n#define AE_NOTUSED(V) ((void) V)\n\n/*\n * 事件处理器状态\n */\nstruct aeEventLoop;\n\n/* Types and data structures \n *\n * 事件接口\n */\ntypedef void aeFileProc(struct aeEventLoop *eventLoop, int fd, void *clientData, int mask);\ntypedef int aeTimeProc(struct aeEventLoop *eventLoop, long long id, void *clientData);\ntypedef void aeEventFinalizerProc(struct aeEventLoop *eventLoop, void *clientData);\ntypedef void aeBeforeSleepProc(struct aeEventLoop *eventLoop);\n\n/* File event structure\n *\n * 文件事件结构\n */\ntypedef struct aeFileEvent {\n\n // 监听事件类型掩码，\n // 值可以是 AE_READABLE 或 AE_WRITABLE ，\n // 或者 AE_READABLE | AE_WRITABLE\n int mask; /* one of AE_(READABLE|WRITABLE) */\n\n // 读事件处理器\n aeFileProc *rfileProc;\n\n // 写事件处理器\n aeFileProc *wfileProc;\n\n // 多路复用库的私有数据\n void *clientData;\n\n} aeFileEvent;\n\n/* Time event structure\n *\n * 时间事件结构\n */\ntypedef struct aeTimeEvent {\n\n // 时间事件的唯一标识符\n long long id; /* time event identifier. */\n\n // 事件的到达时间\n long when_sec; /* seconds */\n long when_ms; /* milliseconds */\n\n // 事件处理函数\n aeTimeProc *timeProc;\n\n // 事件释放函数\n aeEventFinalizerProc *finalizerProc;\n\n // 多路复用库的私有数据\n void *clientData;\n\n // 指向下个时间事件结构，形成链表\n struct aeTimeEvent *next;\n\n} aeTimeEvent;\n\n/* A fired event\n *\n * 已就绪事件\n */\ntypedef struct aeFiredEvent {\n\n // 已就绪文件描述符\n int fd;\n\n // 事件类型掩码，\n // 值可以是 AE_READABLE 或 AE_WRITABLE\n // 或者是两者的或\n int mask;\n\n} aeFiredEvent;\n\n/* State of an event based program \n *\n * 事件处理器的状态\n */\ntypedef struct aeEventLoop {\n\n // 目前已注册的最大描述符\n int maxfd; /* highest file descriptor currently registered */\n\n // 目前已追踪的最大描述符\n int setsize; /* max number of file descriptors tracked */\n\n // 用于生成时间事件 id\n long long timeEventNextId;\n\n // 最后一次执行时间事件的时间\n time_t lastTime; /* Used to detect system clock skew */\n\n // 已注册的文件事件\n aeFileEvent *events; /* Registered events */\n\n // 已就绪的文件事件\n aeFiredEvent *fired; /* Fired events */\n\n // 时间事件\n aeTimeEvent *timeEventHead;\n\n // 事件处理器的开关\n int stop;\n\n // 多路复用库的私有数据\n void *apidata; /* This is used for polling API specific data */\n\n // 在处理事件前要执行的函数\n aeBeforeSleepProc *beforesleep;\n\n} aeEventLoop;\n\n/* Prototypes */\naeEventLoop *aeCreateEventLoop(int setsize);\nvoid aeDeleteEventLoop(aeEventLoop *eventLoop);\nvoid aeStop(aeEventLoop *eventLoop);\nint aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask,\n aeFileProc *proc, void *clientData);\nvoid aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask);\nint aeGetFileEvents(aeEventLoop *eventLoop, int fd);\nlong long aeCreateTimeEvent(aeEventLoop *eventLoop, long long milliseconds,\n aeTimeProc *proc, void *clientData,\n aeEventFinalizerProc *finalizerProc);\nint aeDeleteTimeEvent(aeEventLoop *eventLoop, long long id);\nint aeProcessEvents(aeEventLoop *eventLoop, int flags);\nint aeWait(int fd, int mask, long long milliseconds);\nvoid aeMain(aeEventLoop *eventLoop);\nchar *aeGetApiName(void);\nvoid aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep);\nint aeGetSetSize(aeEventLoop *eventLoop);\nint aeResizeSetSize(aeEventLoop *eventLoop, int setsize);\n\n#endif\n" }, { "path": "src/ae_epoll.c", "content": "/* Linux epoll(2) based ae.c module\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \n\n/*\n * 事件状态\n */\ntypedef struct aeApiState {\n\n // epoll_event 实例描述符\n int epfd;\n\n // 事件槽\n struct epoll_event *events;\n\n} aeApiState;\n\n/*\n * 创建一个新的 epoll 实例，并将它赋值给 eventLoop\n */\nstatic int aeApiCreate(aeEventLoop *eventLoop) {\n\n aeApiState *state = zmalloc(sizeof(aeApiState));\n\n if (!state) return -1;\n\n // 初始化事件槽空间\n state->events = zmalloc(sizeof(struct epoll_event)*eventLoop->setsize);\n if (!state->events) {\n zfree(state);\n return -1;\n }\n\n // 创建 epoll 实例\n state->epfd = epoll_create(1024); /* 1024 is just a hint for the kernel */\n if (state->epfd == -1) {\n zfree(state->events);\n zfree(state);\n return -1;\n }\n\n // 赋值给 eventLoop\n eventLoop->apidata = state;\n return 0;\n}\n\n/*\n * 调整事件槽大小\n */\nstatic int aeApiResize(aeEventLoop *eventLoop, int setsize) {\n aeApiState *state = eventLoop->apidata;\n\n state->events = zrealloc(state->events, sizeof(struct epoll_event)*setsize);\n return 0;\n}\n\n/*\n * 释放 epoll 实例和事件槽\n */\nstatic void aeApiFree(aeEventLoop *eventLoop) {\n aeApiState *state = eventLoop->apidata;\n\n close(state->epfd);\n zfree(state->events);\n zfree(state);\n}\n\n/*\n * 关联给定事件到 fd\n */\nstatic int aeApiAddEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n struct epoll_event ee;\n\n /* If the fd was already monitored for some event, we need a MOD\n * operation. Otherwise we need an ADD operation. \n *\n * 如果 fd 没有关联任何事件，那么这是一个 ADD 操作。\n *\n * 如果已经关联了某个/某些事件，那么这是一个 MOD 操作。\n */\n int op = eventLoop->events[fd].mask == AE_NONE ?\n EPOLL_CTL_ADD : EPOLL_CTL_MOD;\n\n // 注册事件到 epoll\n ee.events = 0;\n mask |= eventLoop->events[fd].mask; /* Merge old events */\n if (mask & AE_READABLE) ee.events |= EPOLLIN;\n if (mask & AE_WRITABLE) ee.events |= EPOLLOUT;\n ee.data.u64 = 0; /* avoid valgrind warning */\n ee.data.fd = fd;\n\n if (epoll_ctl(state->epfd,op,fd,&ee) == -1) return -1;\n\n return 0;\n}\n\n/*\n * 从 fd 中删除给定事件\n */\nstatic void aeApiDelEvent(aeEventLoop *eventLoop, int fd, int delmask) {\n aeApiState *state = eventLoop->apidata;\n struct epoll_event ee;\n\n int mask = eventLoop->events[fd].mask & (~delmask);\n\n ee.events = 0;\n if (mask & AE_READABLE) ee.events |= EPOLLIN;\n if (mask & AE_WRITABLE) ee.events |= EPOLLOUT;\n ee.data.u64 = 0; /* avoid valgrind warning */\n ee.data.fd = fd;\n if (mask != AE_NONE) {\n epoll_ctl(state->epfd,EPOLL_CTL_MOD,fd,&ee);\n } else {\n /* Note, Kernel < 2.6.9 requires a non null event pointer even for\n * EPOLL_CTL_DEL. */\n epoll_ctl(state->epfd,EPOLL_CTL_DEL,fd,&ee);\n }\n}\n\n/*\n * 获取可执行事件\n */\nstatic int aeApiPoll(aeEventLoop *eventLoop, struct timeval *tvp) {\n aeApiState *state = eventLoop->apidata;\n int retval, numevents = 0;\n\n // 等待时间\n retval = epoll_wait(state->epfd,state->events,eventLoop->setsize,\n tvp ? (tvp->tv_sec*1000 + tvp->tv_usec/1000) : -1);\n\n // 有至少一个事件就绪？\n if (retval > 0) {\n int j;\n\n // 为已就绪事件设置相应的模式\n // 并加入到 eventLoop 的 fired 数组中\n numevents = retval;\n for (j = 0; j < numevents; j++) {\n int mask = 0;\n struct epoll_event *e = state->events+j;\n\n if (e->events & EPOLLIN) mask |= AE_READABLE;\n if (e->events & EPOLLOUT) mask |= AE_WRITABLE;\n if (e->events & EPOLLERR) mask |= AE_WRITABLE;\n if (e->events & EPOLLHUP) mask |= AE_WRITABLE;\n\n eventLoop->fired[j].fd = e->data.fd;\n eventLoop->fired[j].mask = mask;\n }\n }\n \n // 返回已就绪事件个数\n return numevents;\n}\n\n/*\n * 返回当前正在使用的 poll 库的名字\n */\nstatic char *aeApiName(void) {\n return \"epoll\";\n}\n" }, { "path": "src/ae_evport.c", "content": "/* ae.c module for illumos event ports.\n *\n * Copyright (c) 2012, Joyent, Inc. All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \n#include \n#include \n#include \n\n#include \n#include \n\n#include \n\nstatic int evport_debug = 0;\n\n/*\n * This file implements the ae API using event ports, present on Solaris-based\n * systems since Solaris 10. Using the event port interface, we associate file\n * descriptors with the port. Each association also includes the set of poll(2)\n * events that the consumer is interested in (e.g., POLLIN and POLLOUT).\n *\n * There's one tricky piece to this implementation: when we return events via\n * aeApiPoll, the corresponding file descriptors become dissociated from the\n * port. This is necessary because poll events are level-triggered, so if the\n * fd didn't become dissociated, it would immediately fire another event since\n * the underlying state hasn't changed yet. We must re-associate the file\n * descriptor, but only after we know that our caller has actually read from it.\n * The ae API does not tell us exactly when that happens, but we do know that\n * it must happen by the time aeApiPoll is called again. Our solution is to\n * keep track of the last fds returned by aeApiPoll and re-associate them next\n * time aeApiPoll is invoked.\n *\n * To summarize, in this module, each fd association is EITHER (a) represented\n * only via the in-kernel association OR (b) represented by pending_fds and\n * pending_masks. (b) is only true for the last fds we returned from aeApiPoll,\n * and only until we enter aeApiPoll again (at which point we restore the\n * in-kernel association).\n */\n#define MAX_EVENT_BATCHSZ 512\n\ntypedef struct aeApiState {\n int portfd; /* event port */\n int npending; /* # of pending fds */\n int pending_fds[MAX_EVENT_BATCHSZ]; /* pending fds */\n int pending_masks[MAX_EVENT_BATCHSZ]; /* pending fds' masks */\n} aeApiState;\n\nstatic int aeApiCreate(aeEventLoop *eventLoop) {\n int i;\n aeApiState *state = zmalloc(sizeof(aeApiState));\n if (!state) return -1;\n\n state->portfd = port_create();\n if (state->portfd == -1) {\n zfree(state);\n return -1;\n }\n\n state->npending = 0;\n\n for (i = 0; i < MAX_EVENT_BATCHSZ; i++) {\n state->pending_fds[i] = -1;\n state->pending_masks[i] = AE_NONE;\n }\n\n eventLoop->apidata = state;\n return 0;\n}\n\nstatic int aeApiResize(aeEventLoop *eventLoop, int setsize) {\n /* Nothing to resize here. */\n return 0;\n}\n\nstatic void aeApiFree(aeEventLoop *eventLoop) {\n aeApiState *state = eventLoop->apidata;\n\n close(state->portfd);\n zfree(state);\n}\n\nstatic int aeApiLookupPending(aeApiState *state, int fd) {\n int i;\n\n for (i = 0; i < state->npending; i++) {\n if (state->pending_fds[i] == fd)\n return (i);\n }\n\n return (-1);\n}\n\n/*\n * Helper function to invoke port_associate for the given fd and mask.\n */\nstatic int aeApiAssociate(const char *where, int portfd, int fd, int mask) {\n int events = 0;\n int rv, err;\n\n if (mask & AE_READABLE)\n events |= POLLIN;\n if (mask & AE_WRITABLE)\n events |= POLLOUT;\n\n if (evport_debug)\n fprintf(stderr, \"%s: port_associate(%d, 0x%x) = \", where, fd, events);\n\n rv = port_associate(portfd, PORT_SOURCE_FD, fd, events,\n (void *)(uintptr_t)mask);\n err = errno;\n\n if (evport_debug)\n fprintf(stderr, \"%d (%s)\\n\", rv, rv == 0 ? \"no error\" : strerror(err));\n\n if (rv == -1) {\n fprintf(stderr, \"%s: port_associate: %s\\n\", where, strerror(err));\n\n if (err == EAGAIN)\n fprintf(stderr, \"aeApiAssociate: event port limit exceeded.\");\n }\n\n return rv;\n}\n\nstatic int aeApiAddEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n int fullmask, pfd;\n\n if (evport_debug)\n fprintf(stderr, \"aeApiAddEvent: fd %d mask 0x%x\\n\", fd, mask);\n\n /*\n * Since port_associate's \"events\" argument replaces any existing events, we\n * must be sure to include whatever events are already associated when\n * we call port_associate() again.\n */\n fullmask = mask | eventLoop->events[fd].mask;\n pfd = aeApiLookupPending(state, fd);\n\n if (pfd != -1) {\n /*\n * This fd was recently returned from aeApiPoll. It should be safe to\n * assume that the consumer has processed that poll event, but we play\n * it safer by simply updating pending_mask. The fd will be\n * re-associated as usual when aeApiPoll is called again.\n */\n if (evport_debug)\n fprintf(stderr, \"aeApiAddEvent: adding to pending fd %d\\n\", fd);\n state->pending_masks[pfd] |= fullmask;\n return 0;\n }\n\n return (aeApiAssociate(\"aeApiAddEvent\", state->portfd, fd, fullmask));\n}\n\nstatic void aeApiDelEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n int fullmask, pfd;\n\n if (evport_debug)\n fprintf(stderr, \"del fd %d mask 0x%x\\n\", fd, mask);\n\n pfd = aeApiLookupPending(state, fd);\n\n if (pfd != -1) {\n if (evport_debug)\n fprintf(stderr, \"deleting event from pending fd %d\\n\", fd);\n\n /*\n * This fd was just returned from aeApiPoll, so it's not currently\n * associated with the port. All we need to do is update\n * pending_mask appropriately.\n */\n state->pending_masks[pfd] &= ~mask;\n\n if (state->pending_masks[pfd] == AE_NONE)\n state->pending_fds[pfd] = -1;\n\n return;\n }\n\n /*\n * The fd is currently associated with the port. Like with the add case\n * above, we must look at the full mask for the file descriptor before\n * updating that association. We don't have a good way of knowing what the\n * events are without looking into the eventLoop state directly. We rely on\n * the fact that our caller has already updated the mask in the eventLoop.\n */\n\n fullmask = eventLoop->events[fd].mask;\n if (fullmask == AE_NONE) {\n /*\n * We're removing *all* events, so use port_dissociate to remove the\n * association completely. Failure here indicates a bug.\n */\n if (evport_debug)\n fprintf(stderr, \"aeApiDelEvent: port_dissociate(%d)\\n\", fd);\n\n if (port_dissociate(state->portfd, PORT_SOURCE_FD, fd) != 0) {\n perror(\"aeApiDelEvent: port_dissociate\");\n abort(); /* will not return */\n }\n } else if (aeApiAssociate(\"aeApiDelEvent\", state->portfd, fd,\n fullmask) != 0) {\n /*\n * ENOMEM is a potentially transient condition, but the kernel won't\n * generally return it unless things are really bad. EAGAIN indicates\n * we've reached an resource limit, for which it doesn't make sense to\n * retry (counter-intuitively). All other errors indicate a bug. In any\n * of these cases, the best we can do is to abort.\n */\n abort(); /* will not return */\n }\n}\n\nstatic int aeApiPoll(aeEventLoop *eventLoop, struct timeval *tvp) {\n aeApiState *state = eventLoop->apidata;\n struct timespec timeout, *tsp;\n int mask, i;\n uint_t nevents;\n port_event_t event[MAX_EVENT_BATCHSZ];\n\n /*\n * If we've returned fd events before, we must re-associate them with the\n * port now, before calling port_get(). See the block comment at the top of\n * this file for an explanation of why.\n */\n for (i = 0; i < state->npending; i++) {\n if (state->pending_fds[i] == -1)\n /* This fd has since been deleted. */\n continue;\n\n if (aeApiAssociate(\"aeApiPoll\", state->portfd,\n state->pending_fds[i], state->pending_masks[i]) != 0) {\n /* See aeApiDelEvent for why this case is fatal. */\n abort();\n }\n\n state->pending_masks[i] = AE_NONE;\n state->pending_fds[i] = -1;\n }\n\n state->npending = 0;\n\n if (tvp != NULL) {\n timeout.tv_sec = tvp->tv_sec;\n timeout.tv_nsec = tvp->tv_usec * 1000;\n tsp = &timeout;\n } else {\n tsp = NULL;\n }\n\n /*\n * port_getn can return with errno == ETIME having returned some events (!).\n * So if we get ETIME, we check nevents, too.\n */\n nevents = 1;\n if (port_getn(state->portfd, event, MAX_EVENT_BATCHSZ, &nevents,\n tsp) == -1 && (errno != ETIME || nevents == 0)) {\n if (errno == ETIME || errno == EINTR)\n return 0;\n\n /* Any other error indicates a bug. */\n perror(\"aeApiPoll: port_get\");\n abort();\n }\n\n state->npending = nevents;\n\n for (i = 0; i < nevents; i++) {\n mask = 0;\n if (event[i].portev_events & POLLIN)\n mask |= AE_READABLE;\n if (event[i].portev_events & POLLOUT)\n mask |= AE_WRITABLE;\n\n eventLoop->fired[i].fd = event[i].portev_object;\n eventLoop->fired[i].mask = mask;\n\n if (evport_debug)\n fprintf(stderr, \"aeApiPoll: fd %d mask 0x%x\\n\",\n (int)event[i].portev_object, mask);\n\n state->pending_fds[i] = event[i].portev_object;\n state->pending_masks[i] = (uintptr_t)event[i].portev_user;\n }\n\n return nevents;\n}\n\nstatic char *aeApiName(void) {\n return \"evport\";\n}\n" }, { "path": "src/ae_kqueue.c", "content": "/* Kqueue(2)-based ae.c module\n *\n * Copyright (C) 2009 Harish Mallipeddi - harish.mallipeddi@gmail.com\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \n#include \n#include \n\ntypedef struct aeApiState {\n int kqfd;\n struct kevent *events;\n} aeApiState;\n\nstatic int aeApiCreate(aeEventLoop *eventLoop) {\n aeApiState *state = zmalloc(sizeof(aeApiState));\n\n if (!state) return -1;\n state->events = zmalloc(sizeof(struct kevent)*eventLoop->setsize);\n if (!state->events) {\n zfree(state);\n return -1;\n }\n state->kqfd = kqueue();\n if (state->kqfd == -1) {\n zfree(state->events);\n zfree(state);\n return -1;\n }\n eventLoop->apidata = state;\n return 0; \n}\n\nstatic int aeApiResize(aeEventLoop *eventLoop, int setsize) {\n aeApiState *state = eventLoop->apidata;\n\n state->events = zrealloc(state->events, sizeof(struct kevent)*setsize);\n return 0;\n}\n\nstatic void aeApiFree(aeEventLoop *eventLoop) {\n aeApiState *state = eventLoop->apidata;\n\n close(state->kqfd);\n zfree(state->events);\n zfree(state);\n}\n\nstatic int aeApiAddEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n struct kevent ke;\n \n if (mask & AE_READABLE) {\n EV_SET(&ke, fd, EVFILT_READ, EV_ADD, 0, 0, NULL);\n if (kevent(state->kqfd, &ke, 1, NULL, 0, NULL) == -1) return -1;\n }\n if (mask & AE_WRITABLE) {\n EV_SET(&ke, fd, EVFILT_WRITE, EV_ADD, 0, 0, NULL);\n if (kevent(state->kqfd, &ke, 1, NULL, 0, NULL) == -1) return -1;\n }\n return 0;\n}\n\nstatic void aeApiDelEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n struct kevent ke;\n\n if (mask & AE_READABLE) {\n EV_SET(&ke, fd, EVFILT_READ, EV_DELETE, 0, 0, NULL);\n kevent(state->kqfd, &ke, 1, NULL, 0, NULL);\n }\n if (mask & AE_WRITABLE) {\n EV_SET(&ke, fd, EVFILT_WRITE, EV_DELETE, 0, 0, NULL);\n kevent(state->kqfd, &ke, 1, NULL, 0, NULL);\n }\n}\n\nstatic int aeApiPoll(aeEventLoop *eventLoop, struct timeval *tvp) {\n aeApiState *state = eventLoop->apidata;\n int retval, numevents = 0;\n\n if (tvp != NULL) {\n struct timespec timeout;\n timeout.tv_sec = tvp->tv_sec;\n timeout.tv_nsec = tvp->tv_usec * 1000;\n retval = kevent(state->kqfd, NULL, 0, state->events, eventLoop->setsize,\n &timeout);\n } else {\n retval = kevent(state->kqfd, NULL, 0, state->events, eventLoop->setsize,\n NULL);\n }\n\n if (retval > 0) {\n int j;\n \n numevents = retval;\n for(j = 0; j < numevents; j++) {\n int mask = 0;\n struct kevent *e = state->events+j;\n \n if (e->filter == EVFILT_READ) mask |= AE_READABLE;\n if (e->filter == EVFILT_WRITE) mask |= AE_WRITABLE;\n eventLoop->fired[j].fd = e->ident; \n eventLoop->fired[j].mask = mask; \n }\n }\n return numevents;\n}\n\nstatic char *aeApiName(void) {\n return \"kqueue\";\n}\n" }, { "path": "src/ae_select.c", "content": "/* Select()-based ae.c module.\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \n\ntypedef struct aeApiState {\n fd_set rfds, wfds;\n /* We need to have a copy of the fd sets as it's not safe to reuse\n * FD sets after select(). */\n fd_set _rfds, _wfds;\n} aeApiState;\n\nstatic int aeApiCreate(aeEventLoop *eventLoop) {\n aeApiState *state = zmalloc(sizeof(aeApiState));\n\n if (!state) return -1;\n FD_ZERO(&state->rfds);\n FD_ZERO(&state->wfds);\n eventLoop->apidata = state;\n return 0;\n}\n\nstatic int aeApiResize(aeEventLoop *eventLoop, int setsize) {\n /* Just ensure we have enough room in the fd_set type. */\n if (setsize >= FD_SETSIZE) return -1;\n return 0;\n}\n\nstatic void aeApiFree(aeEventLoop *eventLoop) {\n zfree(eventLoop->apidata);\n}\n\nstatic int aeApiAddEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n\n if (mask & AE_READABLE) FD_SET(fd,&state->rfds);\n if (mask & AE_WRITABLE) FD_SET(fd,&state->wfds);\n return 0;\n}\n\nstatic void aeApiDelEvent(aeEventLoop *eventLoop, int fd, int mask) {\n aeApiState *state = eventLoop->apidata;\n\n if (mask & AE_READABLE) FD_CLR(fd,&state->rfds);\n if (mask & AE_WRITABLE) FD_CLR(fd,&state->wfds);\n}\n\nstatic int aeApiPoll(aeEventLoop *eventLoop, struct timeval *tvp) {\n aeApiState *state = eventLoop->apidata;\n int retval, j, numevents = 0;\n\n memcpy(&state->_rfds,&state->rfds,sizeof(fd_set));\n memcpy(&state->_wfds,&state->wfds,sizeof(fd_set));\n\n retval = select(eventLoop->maxfd+1,\n &state->_rfds,&state->_wfds,NULL,tvp);\n if (retval > 0) {\n for (j = 0; j <= eventLoop->maxfd; j++) {\n int mask = 0;\n aeFileEvent *fe = &eventLoop->events[j];\n\n if (fe->mask == AE_NONE) continue;\n if (fe->mask & AE_READABLE && FD_ISSET(j,&state->_rfds))\n mask |= AE_READABLE;\n if (fe->mask & AE_WRITABLE && FD_ISSET(j,&state->_wfds))\n mask |= AE_WRITABLE;\n eventLoop->fired[numevents].fd = j;\n eventLoop->fired[numevents].mask = mask;\n numevents++;\n }\n }\n return numevents;\n}\n\nstatic char *aeApiName(void) {\n return \"select\";\n}\n" }, { "path": "src/anet.c", "content": "/* anet.c -- Basic TCP socket stuff made a bit less boring\n *\n * Copyright (c) 2006-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"fmacros.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"anet.h\"\n\n/*\n * 打印错误信息\n */\nstatic void anetSetError(char *err, const char *fmt, ...)\n{\n va_list ap;\n\n if (!err) return;\n va_start(ap, fmt);\n vsnprintf(err, ANET_ERR_LEN, fmt, ap);\n va_end(ap);\n}\n\n/*\n * 将 fd 设置为非阻塞模式（O_NONBLOCK）\n */\nint anetNonBlock(char *err, int fd)\n{\n int flags;\n\n /* Set the socket non-blocking.\n * Note that fcntl(2) for F_GETFL and F_SETFL can't be\n * interrupted by a signal. */\n if ((flags = fcntl(fd, F_GETFL)) == -1) {\n anetSetError(err, \"fcntl(F_GETFL): %s\", strerror(errno));\n return ANET_ERR;\n }\n if (fcntl(fd, F_SETFL, flags | O_NONBLOCK) == -1) {\n anetSetError(err, \"fcntl(F_SETFL,O_NONBLOCK): %s\", strerror(errno));\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\n/* Set TCP keep alive option to detect dead peers. The interval option\n * is only used for Linux as we are using Linux-specific APIs to set\n * the probe send time, interval, and count.\n *\n * 修改 TCP 连接的 keep alive 选项\n */\nint anetKeepAlive(char *err, int fd, int interval)\n{\n int val = 1;\n\n if (setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &val, sizeof(val)) == -1)\n {\n anetSetError(err, \"setsockopt SO_KEEPALIVE: %s\", strerror(errno));\n return ANET_ERR;\n }\n\n#ifdef __linux__\n /* Default settings are more or less garbage, with the keepalive time\n * set to 7200 by default on Linux. Modify settings to make the feature\n * actually useful. */\n\n /* Send first probe after interval. */\n val = interval;\n if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPIDLE, &val, sizeof(val)) < 0) {\n anetSetError(err, \"setsockopt TCP_KEEPIDLE: %s\\n\", strerror(errno));\n return ANET_ERR;\n }\n\n /* Send next probes after the specified interval. Note that we set the\n * delay as interval / 3, as we send three probes before detecting\n * an error (see the next setsockopt call). */\n val = interval/3;\n if (val == 0) val = 1;\n if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPINTVL, &val, sizeof(val)) < 0) {\n anetSetError(err, \"setsockopt TCP_KEEPINTVL: %s\\n\", strerror(errno));\n return ANET_ERR;\n }\n\n /* Consider the socket in error state after three we send three ACK\n * probes without getting a reply. */\n val = 3;\n if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPCNT, &val, sizeof(val)) < 0) {\n anetSetError(err, \"setsockopt TCP_KEEPCNT: %s\\n\", strerror(errno));\n return ANET_ERR;\n }\n#endif\n\n return ANET_OK;\n}\n\n/*\n * 打开或关闭 Nagle 算法\n */\nstatic int anetSetTcpNoDelay(char *err, int fd, int val)\n{\n if (setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &val, sizeof(val)) == -1)\n {\n anetSetError(err, \"setsockopt TCP_NODELAY: %s\", strerror(errno));\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\n/*\n * 禁用 Nagle 算法\n */\nint anetEnableTcpNoDelay(char *err, int fd)\n{\n return anetSetTcpNoDelay(err, fd, 1);\n}\n\n/*\n * 启用 Nagle 算法\n */\nint anetDisableTcpNoDelay(char *err, int fd) \n{\n return anetSetTcpNoDelay(err, fd, 0);\n}\n\n/*\n * 设置 socket 的最大发送 buffer 字节数\n */\nint anetSetSendBuffer(char *err, int fd, int buffsize)\n{\n if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &buffsize, sizeof(buffsize)) == -1)\n {\n anetSetError(err, \"setsockopt SO_SNDBUF: %s\", strerror(errno));\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\n/*\n * 开启 TCP 的 keep alive 选项\n */\nint anetTcpKeepAlive(char *err, int fd)\n{\n int yes = 1;\n if (setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &yes, sizeof(yes)) == -1) {\n anetSetError(err, \"setsockopt SO_KEEPALIVE: %s\", strerror(errno));\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\n/* anetGenericResolve() is called by anetResolve() and anetResolveIP() to\n * do the actual work. It resolves the hostname \"host\" and set the string\n * representation of the IP address into the buffer pointed by \"ipbuf\".\n *\n * If flags is set to ANET_IP_ONLY the function only resolves hostnames\n * that are actually already IPv4 or IPv6 addresses. This turns the function\n * into a validating / normalizing function. */\n// 解释 host 的地址，并保存到 ipbuf 中\nint anetGenericResolve(char *err, char *host, char *ipbuf, size_t ipbuf_len,\n int flags)\n{\n struct addrinfo hints, *info;\n int rv;\n\n memset(&hints,0,sizeof(hints));\n if (flags & ANET_IP_ONLY) hints.ai_flags = AI_NUMERICHOST;\n hints.ai_family = AF_UNSPEC;\n hints.ai_socktype = SOCK_STREAM; /* specify socktype to avoid dups */\n\n if ((rv = getaddrinfo(host, NULL, &hints, &info)) != 0) {\n anetSetError(err, \"%s\", gai_strerror(rv));\n return ANET_ERR;\n }\n if (info->ai_family == AF_INET) {\n struct sockaddr_in *sa = (struct sockaddr_in *)info->ai_addr;\n inet_ntop(AF_INET, &(sa->sin_addr), ipbuf, ipbuf_len);\n } else {\n struct sockaddr_in6 *sa = (struct sockaddr_in6 *)info->ai_addr;\n inet_ntop(AF_INET6, &(sa->sin6_addr), ipbuf, ipbuf_len);\n }\n\n freeaddrinfo(info);\n return ANET_OK;\n}\n\nint anetResolve(char *err, char *host, char *ipbuf, size_t ipbuf_len) {\n return anetGenericResolve(err,host,ipbuf,ipbuf_len,ANET_NONE);\n}\n\nint anetResolveIP(char *err, char *host, char *ipbuf, size_t ipbuf_len) {\n return anetGenericResolve(err,host,ipbuf,ipbuf_len,ANET_IP_ONLY);\n}\n\n// 设置地址为可重用\nstatic int anetSetReuseAddr(char *err, int fd) {\n int yes = 1;\n /* Make sure connection-intensive things like the redis benckmark\n * will be able to close/open sockets a zillion of times */\n if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) == -1) {\n anetSetError(err, \"setsockopt SO_REUSEADDR: %s\", strerror(errno));\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\n/*\n * 创建并返回 socket\n */\nstatic int anetCreateSocket(char *err, int domain) {\n int s;\n if ((s = socket(domain, SOCK_STREAM, 0)) == -1) {\n anetSetError(err, \"creating socket: %s\", strerror(errno));\n return ANET_ERR;\n }\n\n /* Make sure connection-intensive things like the redis benchmark\n * will be able to close/open sockets a zillion of times */\n if (anetSetReuseAddr(err,s) == ANET_ERR) {\n close(s);\n return ANET_ERR;\n }\n return s;\n}\n\n// 通用连接创建函数，被其他高层函数所调用\n#define ANET_CONNECT_NONE 0\n#define ANET_CONNECT_NONBLOCK 1\nstatic int anetTcpGenericConnect(char *err, char *addr, int port,\n char *source_addr, int flags)\n{\n int s = ANET_ERR, rv;\n char portstr[6]; /* strlen(\"65535\") + 1; */\n struct addrinfo hints, *servinfo, *bservinfo, *p, *b;\n\n snprintf(portstr,sizeof(portstr),\"%d\",port);\n memset(&hints,0,sizeof(hints));\n hints.ai_family = AF_UNSPEC;\n hints.ai_socktype = SOCK_STREAM;\n\n if ((rv = getaddrinfo(addr,portstr,&hints,&servinfo)) != 0) {\n anetSetError(err, \"%s\", gai_strerror(rv));\n return ANET_ERR;\n }\n for (p = servinfo; p != NULL; p = p->ai_next) {\n /* Try to create the socket and to connect it.\n * If we fail in the socket() call, or on connect(), we retry with\n * the next entry in servinfo. */\n if ((s = socket(p->ai_family,p->ai_socktype,p->ai_protocol)) == -1)\n continue;\n if (anetSetReuseAddr(err,s) == ANET_ERR) goto error;\n if (flags & ANET_CONNECT_NONBLOCK && anetNonBlock(err,s) != ANET_OK)\n goto error;\n if (source_addr) {\n int bound = 0;\n /* Using getaddrinfo saves us from self-determining IPv4 vs IPv6 */\n if ((rv = getaddrinfo(source_addr, NULL, &hints, &bservinfo)) != 0) {\n anetSetError(err, \"%s\", gai_strerror(rv));\n goto end;\n }\n for (b = bservinfo; b != NULL; b = b->ai_next) {\n if (bind(s,b->ai_addr,b->ai_addrlen) != -1) {\n bound = 1;\n break;\n }\n }\n if (!bound) {\n anetSetError(err, \"bind: %s\", strerror(errno));\n goto end;\n }\n }\n if (connect(s,p->ai_addr,p->ai_addrlen) == -1) {\n /* If the socket is non-blocking, it is ok for connect() to\n * return an EINPROGRESS error here. */\n if (errno == EINPROGRESS && flags & ANET_CONNECT_NONBLOCK)\n goto end;\n close(s);\n s = ANET_ERR;\n continue;\n }\n\n /* If we ended an iteration of the for loop without errors, we\n * have a connected socket. Let's return to the caller. */\n goto end;\n }\n if (p == NULL)\n anetSetError(err, \"creating socket: %s\", strerror(errno));\n\nerror:\n if (s != ANET_ERR) {\n close(s);\n s = ANET_ERR;\n }\nend:\n freeaddrinfo(servinfo);\n return s;\n}\n\n/*\n * 创建阻塞 TCP 连接\n */\nint anetTcpConnect(char *err, char *addr, int port)\n{\n return anetTcpGenericConnect(err,addr,port,NULL,ANET_CONNECT_NONE);\n}\n\n/*\n * 创建非阻塞 TCP 连接\n */\nint anetTcpNonBlockConnect(char *err, char *addr, int port)\n{\n return anetTcpGenericConnect(err,addr,port,NULL,ANET_CONNECT_NONBLOCK);\n}\n\nint anetTcpNonBlockBindConnect(char *err, char *addr, int port, char *source_addr)\n{\n return anetTcpGenericConnect(err,addr,port,source_addr,ANET_CONNECT_NONBLOCK);\n}\n\nint anetUnixGenericConnect(char *err, char *path, int flags)\n{\n int s;\n struct sockaddr_un sa;\n\n if ((s = anetCreateSocket(err,AF_LOCAL)) == ANET_ERR)\n return ANET_ERR;\n\n sa.sun_family = AF_LOCAL;\n strncpy(sa.sun_path,path,sizeof(sa.sun_path)-1);\n if (flags & ANET_CONNECT_NONBLOCK) {\n if (anetNonBlock(err,s) != ANET_OK)\n return ANET_ERR;\n }\n if (connect(s,(struct sockaddr*)&sa,sizeof(sa)) == -1) {\n if (errno == EINPROGRESS &&\n flags & ANET_CONNECT_NONBLOCK)\n return s;\n\n anetSetError(err, \"connect: %s\", strerror(errno));\n close(s);\n return ANET_ERR;\n }\n return s;\n}\n\n/*\n * 创建阻塞本地连接\n */\nint anetUnixConnect(char *err, char *path)\n{\n return anetUnixGenericConnect(err,path,ANET_CONNECT_NONE);\n}\n\n/*\n * 创建非阻塞本地连接\n */\nint anetUnixNonBlockConnect(char *err, char *path)\n{\n return anetUnixGenericConnect(err,path,ANET_CONNECT_NONBLOCK);\n}\n\n/* Like read(2) but make sure 'count' is read before to return\n * (unless error or EOF condition is encountered) */\n/*\n * 带 short count 处理的读取函数\n */\nint anetRead(int fd, char *buf, int count)\n{\n int nread, totlen = 0;\n while(totlen != count) {\n nread = read(fd,buf,count-totlen);\n if (nread == 0) return totlen;\n if (nread == -1) return -1;\n totlen += nread;\n buf += nread;\n }\n return totlen;\n}\n\n/* Like write(2) but make sure 'count' is read before to return\n * (unless error is encountered) */\n/*\n * 带 short count 处理的写入函数\n */\nint anetWrite(int fd, char *buf, int count)\n{\n int nwritten, totlen = 0;\n while(totlen != count) {\n nwritten = write(fd,buf,count-totlen);\n if (nwritten == 0) return totlen;\n if (nwritten == -1) return -1;\n totlen += nwritten;\n buf += nwritten;\n }\n return totlen;\n}\n\n/*\n * 绑定并创建监听套接字\n */\nstatic int anetListen(char *err, int s, struct sockaddr *sa, socklen_t len, int backlog) {\n if (bind(s,sa,len) == -1) {\n anetSetError(err, \"bind: %s\", strerror(errno));\n close(s);\n return ANET_ERR;\n }\n\n if (listen(s, backlog) == -1) {\n anetSetError(err, \"listen: %s\", strerror(errno));\n close(s);\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\nstatic int anetV6Only(char *err, int s) {\n int yes = 1;\n if (setsockopt(s,IPPROTO_IPV6,IPV6_V6ONLY,&yes,sizeof(yes)) == -1) {\n anetSetError(err, \"setsockopt: %s\", strerror(errno));\n close(s);\n return ANET_ERR;\n }\n return ANET_OK;\n}\n\nstatic int _anetTcpServer(char *err, int port, char *bindaddr, int af, int backlog)\n{\n int s, rv;\n char _port[6]; /* strlen(\"65535\") */\n struct addrinfo hints, *servinfo, *p;\n\n snprintf(_port,6,\"%d\",port);\n memset(&hints,0,sizeof(hints));\n hints.ai_family = af;\n hints.ai_socktype = SOCK_STREAM;\n hints.ai_flags = AI_PASSIVE; /* No effect if bindaddr != NULL */\n\n if ((rv = getaddrinfo(bindaddr,_port,&hints,&servinfo)) != 0) {\n anetSetError(err, \"%s\", gai_strerror(rv));\n return ANET_ERR;\n }\n for (p = servinfo; p != NULL; p = p->ai_next) {\n if ((s = socket(p->ai_family,p->ai_socktype,p->ai_protocol)) == -1)\n continue;\n\n if (af == AF_INET6 && anetV6Only(err,s) == ANET_ERR) goto error;\n if (anetSetReuseAddr(err,s) == ANET_ERR) goto error;\n if (anetListen(err,s,p->ai_addr,p->ai_addrlen,backlog) == ANET_ERR) goto error;\n goto end;\n }\n if (p == NULL) {\n anetSetError(err, \"unable to bind socket\");\n goto error;\n }\n\nerror:\n s = ANET_ERR;\nend:\n freeaddrinfo(servinfo);\n return s;\n}\n\nint anetTcpServer(char *err, int port, char *bindaddr, int backlog)\n{\n return _anetTcpServer(err, port, bindaddr, AF_INET, backlog);\n}\n\nint anetTcp6Server(char *err, int port, char *bindaddr, int backlog)\n{\n return _anetTcpServer(err, port, bindaddr, AF_INET6, backlog);\n}\n\n/*\n * 创建一个本地连接用的服务器监听套接字\n */\nint anetUnixServer(char *err, char *path, mode_t perm, int backlog)\n{\n int s;\n struct sockaddr_un sa;\n\n if ((s = anetCreateSocket(err,AF_LOCAL)) == ANET_ERR)\n return ANET_ERR;\n\n memset(&sa,0,sizeof(sa));\n sa.sun_family = AF_LOCAL;\n strncpy(sa.sun_path,path,sizeof(sa.sun_path)-1);\n if (anetListen(err,s,(struct sockaddr*)&sa,sizeof(sa),backlog) == ANET_ERR)\n return ANET_ERR;\n if (perm)\n chmod(sa.sun_path, perm);\n return s;\n}\n\nstatic int anetGenericAccept(char *err, int s, struct sockaddr *sa, socklen_t *len) {\n int fd;\n while(1) {\n fd = accept(s,sa,len);\n if (fd == -1) {\n if (errno == EINTR)\n continue;\n else {\n anetSetError(err, \"accept: %s\", strerror(errno));\n return ANET_ERR;\n }\n }\n break;\n }\n return fd;\n}\n\n/*\n * TCP 连接 accept 函数\n */\nint anetTcpAccept(char *err, int s, char *ip, size_t ip_len, int *port) {\n int fd;\n struct sockaddr_storage sa;\n socklen_t salen = sizeof(sa);\n if ((fd = anetGenericAccept(err,s,(struct sockaddr*)&sa,&salen)) == -1)\n return ANET_ERR;\n\n if (sa.ss_family == AF_INET) {\n struct sockaddr_in *s = (struct sockaddr_in *)&sa;\n if (ip) inet_ntop(AF_INET,(void*)&(s->sin_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin_port);\n } else {\n struct sockaddr_in6 *s = (struct sockaddr_in6 *)&sa;\n if (ip) inet_ntop(AF_INET6,(void*)&(s->sin6_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin6_port);\n }\n return fd;\n}\n\n/*\n * 本地连接 accept 函数\n */\nint anetUnixAccept(char *err, int s) {\n int fd;\n struct sockaddr_un sa;\n socklen_t salen = sizeof(sa);\n if ((fd = anetGenericAccept(err,s,(struct sockaddr*)&sa,&salen)) == -1)\n return ANET_ERR;\n\n return fd;\n}\n\n/*\n * 获取连接客户端的 IP 和端口号\n */\nint anetPeerToString(int fd, char *ip, size_t ip_len, int *port) {\n struct sockaddr_storage sa;\n socklen_t salen = sizeof(sa);\n\n if (getpeername(fd,(struct sockaddr*)&sa,&salen) == -1) {\n if (port) *port = 0;\n ip[0] = '?';\n ip[1] = '\\0';\n return -1;\n }\n if (sa.ss_family == AF_INET) {\n struct sockaddr_in *s = (struct sockaddr_in *)&sa;\n if (ip) inet_ntop(AF_INET,(void*)&(s->sin_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin_port);\n } else {\n struct sockaddr_in6 *s = (struct sockaddr_in6 *)&sa;\n if (ip) inet_ntop(AF_INET6,(void*)&(s->sin6_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin6_port);\n }\n return 0;\n}\n\n/*\n * 获取服务器本机的 IP 和端口号\n */\nint anetSockName(int fd, char *ip, size_t ip_len, int *port) {\n struct sockaddr_storage sa;\n socklen_t salen = sizeof(sa);\n\n if (getsockname(fd,(struct sockaddr*)&sa,&salen) == -1) {\n if (port) *port = 0;\n ip[0] = '?';\n ip[1] = '\\0';\n return -1;\n }\n if (sa.ss_family == AF_INET) {\n struct sockaddr_in *s = (struct sockaddr_in *)&sa;\n if (ip) inet_ntop(AF_INET,(void*)&(s->sin_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin_port);\n } else {\n struct sockaddr_in6 *s = (struct sockaddr_in6 *)&sa;\n if (ip) inet_ntop(AF_INET6,(void*)&(s->sin6_addr),ip,ip_len);\n if (port) *port = ntohs(s->sin6_port);\n }\n return 0;\n}\n" }, { "path": "src/anet.h", "content": "/* anet.c -- Basic TCP socket stuff made a bit less boring\n *\n * Copyright (c) 2006-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef ANET_H\n#define ANET_H\n\n#define ANET_OK 0\n#define ANET_ERR -1\n#define ANET_ERR_LEN 256\n\n/* Flags used with certain functions. */\n#define ANET_NONE 0\n#define ANET_IP_ONLY (1<<0)\n\n#if defined(__sun)\n#define AF_LOCAL AF_UNIX\n#endif\n\nint anetTcpConnect(char *err, char *addr, int port);\nint anetTcpNonBlockConnect(char *err, char *addr, int port);\nint anetTcpNonBlockBindConnect(char *err, char *addr, int port, char *source_addr);\nint anetUnixConnect(char *err, char *path);\nint anetUnixNonBlockConnect(char *err, char *path);\nint anetRead(int fd, char *buf, int count);\nint anetResolve(char *err, char *host, char *ipbuf, size_t ipbuf_len);\nint anetResolveIP(char *err, char *host, char *ipbuf, size_t ipbuf_len);\nint anetTcpServer(char *err, int port, char *bindaddr, int backlog);\nint anetTcp6Server(char *err, int port, char *bindaddr, int backlog);\nint anetUnixServer(char *err, char *path, mode_t perm, int backlog);\nint anetTcpAccept(char *err, int serversock, char *ip, size_t ip_len, int *port);\nint anetUnixAccept(char *err, int serversock);\nint anetWrite(int fd, char *buf, int count);\nint anetNonBlock(char *err, int fd);\nint anetEnableTcpNoDelay(char *err, int fd);\nint anetDisableTcpNoDelay(char *err, int fd);\nint anetTcpKeepAlive(char *err, int fd);\nint anetPeerToString(int fd, char *ip, size_t ip_len, int *port);\nint anetKeepAlive(char *err, int fd, int interval);\nint anetSockName(int fd, char *ip, size_t ip_len, int *port);\n\n#endif\n" }, { "path": "src/aof.c", "content": "/*\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"redis.h\"\n#include \"bio.h\"\n#include \"rio.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\nvoid aofUpdateCurrentSize(void);\n\n/* ----------------------------------------------------------------------------\n * AOF rewrite buffer implementation.\n *\n * AOF 重写缓存的实现。\n *\n * The following code implement a simple buffer used in order to accumulate\n * changes while the background process is rewriting the AOF file.\n *\n * 以下代码实现了一个简单的缓存，\n * 它可以在 BGREWRITEAOF 执行的过程中，累积所有修改数据集的命令。\n *\n * We only need to append, but can't just use realloc with a large block\n * because 'huge' reallocs are not always handled as one could expect\n * (via remapping of pages at OS level) but may involve copying data.\n *\n * For this reason we use a list of blocks, every block is\n * AOF_RW_BUF_BLOCK_SIZE bytes.\n *\n * 程序需要不断对这个缓存执行 append 操作，\n * 因为分配一个非常大的空间并不总是可能的，\n * 也可能产生大量的复制工作，\n * 所以这里使用多个大小为 AOF_RW_BUF_BLOCK_SIZE 的空间来保存命令。\n *\n * ------------------------------------------------------------------------- */\n\n// 每个缓存块的大小\n#define AOF_RW_BUF_BLOCK_SIZE (1024*1024*10) /* 10 MB per block */\n\ntypedef struct aofrwblock {\n \n // 缓存块已使用字节数和可用字节数\n unsigned long used, free;\n\n // 缓存块\n char buf[AOF_RW_BUF_BLOCK_SIZE];\n\n} aofrwblock;\n\n/* This function free the old AOF rewrite buffer if needed, and initialize\n * a fresh new one. It tests for server.aof_rewrite_buf_blocks equal to NULL\n * so can be used for the first initialization as well. \n *\n * 释放旧的 AOF 重写缓存，并初始化一个新的 AOF 缓存。\n *\n * 这个函数也可以单纯地用于 AOF 重写缓存的初始化。\n */\nvoid aofRewriteBufferReset(void) {\n\n // 释放旧有的缓存（链表）\n if (server.aof_rewrite_buf_blocks)\n listRelease(server.aof_rewrite_buf_blocks);\n\n // 初始化新的缓存（链表）\n server.aof_rewrite_buf_blocks = listCreate();\n listSetFreeMethod(server.aof_rewrite_buf_blocks,zfree);\n}\n\n/* Return the current size of the AOF rerwite buffer. \n *\n * 返回 AOF 重写缓存当前的大小\n */\nunsigned long aofRewriteBufferSize(void) {\n\n // 取出链表中最后的缓存块\n listNode *ln = listLast(server.aof_rewrite_buf_blocks);\n aofrwblock *block = ln ? ln->value : NULL;\n\n // 没有缓存被使用\n if (block == NULL) return 0;\n\n // 总缓存大小 = （缓存块数量-1） * AOF_RW_BUF_BLOCK_SIZE + 最后一个缓存块的大小\n unsigned long size =\n (listLength(server.aof_rewrite_buf_blocks)-1) * AOF_RW_BUF_BLOCK_SIZE;\n size += block->used;\n\n return size;\n}\n\n/* Append data to the AOF rewrite buffer, allocating new blocks if needed. \n *\n * 将字符数组 s 追加到 AOF 缓存的末尾，\n * 如果有需要的话，分配一个新的缓存块。\n */\nvoid aofRewriteBufferAppend(unsigned char *s, unsigned long len) {\n\n // 指向最后一个缓存块\n listNode *ln = listLast(server.aof_rewrite_buf_blocks);\n aofrwblock *block = ln ? ln->value : NULL;\n\n while(len) {\n /* If we already got at least an allocated block, try appending\n * at least some piece into it. \n *\n * 如果已经有至少一个缓存块，那么尝试将内容追加到这个缓存块里面\n */\n if (block) {\n unsigned long thislen = (block->free < len) ? block->free : len;\n if (thislen) { /* The current block is not already full. */\n memcpy(block->buf+block->used, s, thislen);\n block->used += thislen;\n block->free -= thislen;\n s += thislen;\n len -= thislen;\n }\n }\n\n // 如果 block != NULL ，那么这里是创建另一个缓存块买容纳 block 装不下的内容\n // 如果 block == NULL ，那么这里是创建缓存链表的第一个缓存块\n if (len) { /* First block to allocate, or need another block. */\n int numblocks;\n\n // 分配缓存块\n block = zmalloc(sizeof(*block));\n block->free = AOF_RW_BUF_BLOCK_SIZE;\n block->used = 0;\n\n // 链接到链表末尾\n listAddNodeTail(server.aof_rewrite_buf_blocks,block);\n\n /* Log every time we cross more 10 or 100 blocks, respectively\n * as a notice or warning. \n *\n * 每次创建 10 个缓存块就打印一个日志，用作标记或者提醒\n */\n numblocks = listLength(server.aof_rewrite_buf_blocks);\n if (((numblocks+1) % 10) == 0) {\n int level = ((numblocks+1) % 100) == 0 ? REDIS_WARNING :\n REDIS_NOTICE;\n redisLog(level,\"Background AOF buffer size: %lu MB\",\n aofRewriteBufferSize()/(1024*1024));\n }\n }\n }\n}\n\n/* Write the buffer (possibly composed of multiple blocks) into the specified\n * fd. If a short write or any other error happens -1 is returned,\n * otherwise the number of bytes written is returned. \n *\n * 将重写缓存中的所有内容（可能由多个块组成）写入到给定 fd 中。\n *\n * 如果没有 short write 或者其他错误发生，那么返回写入的字节数量，\n * 否则，返回 -1 。\n */\nssize_t aofRewriteBufferWrite(int fd) {\n listNode *ln;\n listIter li;\n ssize_t count = 0;\n\n // 遍历所有缓存块\n listRewind(server.aof_rewrite_buf_blocks,&li);\n while((ln = listNext(&li))) {\n aofrwblock *block = listNodeValue(ln);\n ssize_t nwritten;\n\n if (block->used) {\n\n // 写入缓存块内容到 fd\n nwritten = write(fd,block->buf,block->used);\n if (nwritten != block->used) {\n if (nwritten == 0) errno = EIO;\n return -1;\n }\n\n // 积累写入字节\n count += nwritten;\n }\n }\n\n return count;\n}\n\n/* ----------------------------------------------------------------------------\n * AOF file implementation\n * ------------------------------------------------------------------------- */\n\n/* Starts a background task that performs fsync() against the specified\n * file descriptor (the one of the AOF file) in another thread. \n *\n * 在另一个线程中，对给定的描述符 fd （指向 AOF 文件）执行一个后台 fsync() 操作。\n */\nvoid aof_background_fsync(int fd) {\n bioCreateBackgroundJob(REDIS_BIO_AOF_FSYNC,(void*)(long)fd,NULL,NULL);\n}\n\n/* Called when the user switches from \"appendonly yes\" to \"appendonly no\"\n * at runtime using the CONFIG command. \n *\n * 在用户通过 CONFIG 命令在运行时关闭 AOF 持久化时调用\n */\nvoid stopAppendOnly(void) {\n\n // AOF 必须正在启用，才能调用这个函数\n redisAssert(server.aof_state != REDIS_AOF_OFF);\n\n // 将 AOF 缓存的内容写入并冲洗到 AOF 文件中\n // 参数 1 表示强制模式\n flushAppendOnlyFile(1);\n\n // 冲洗 AOF 文件\n aof_fsync(server.aof_fd);\n\n // 关闭 AOF 文件\n close(server.aof_fd);\n\n // 清空 AOF 状态\n server.aof_fd = -1;\n server.aof_selected_db = -1;\n server.aof_state = REDIS_AOF_OFF;\n\n /* rewrite operation in progress? kill it, wait child exit \n *\n * 如果 BGREWRITEAOF 正在执行，那么杀死它\n * 并等待子进程退出\n */\n if (server.aof_child_pid != -1) {\n int statloc;\n\n redisLog(REDIS_NOTICE,\"Killing running AOF rewrite child: %ld\",\n (long) server.aof_child_pid);\n\n // 杀死子进程\n if (kill(server.aof_child_pid,SIGUSR1) != -1)\n wait3(&statloc,0,NULL);\n\n /* reset the buffer accumulating changes while the child saves \n * 清理未完成的 AOF 重写留下来的缓存和临时文件\n */\n aofRewriteBufferReset();\n aofRemoveTempFile(server.aof_child_pid);\n server.aof_child_pid = -1;\n server.aof_rewrite_time_start = -1;\n }\n}\n\n/* Called when the user switches from \"appendonly no\" to \"appendonly yes\"\n * at runtime using the CONFIG command. \n *\n * 当用户在运行时使用 CONFIG 命令，\n * 从 appendonly no 切换到 appendonly yes 时执行\n */\nint startAppendOnly(void) {\n\n // 将开始时间设为 AOF 最后一次 fsync 时间 \n server.aof_last_fsync = server.unixtime;\n\n // 打开 AOF 文件\n server.aof_fd = open(server.aof_filename,O_WRONLY|O_APPEND|O_CREAT,0644);\n\n redisAssert(server.aof_state == REDIS_AOF_OFF);\n\n // 文件打开失败\n if (server.aof_fd == -1) {\n redisLog(REDIS_WARNING,\"Redis needs to enable the AOF but can't open the append only file: %s\",strerror(errno));\n return REDIS_ERR;\n }\n\n if (rewriteAppendOnlyFileBackground() == REDIS_ERR) {\n // AOF 后台重写失败，关闭 AOF 文件\n close(server.aof_fd);\n redisLog(REDIS_WARNING,\"Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.\");\n return REDIS_ERR;\n }\n\n /* We correctly switched on AOF, now wait for the rerwite to be complete\n * in order to append data on disk. \n *\n * 等待重写执行完毕\n */\n server.aof_state = REDIS_AOF_WAIT_REWRITE;\n\n return REDIS_OK;\n}\n\n/* Write the append only file buffer on disk.\n *\n * 将 AOF 缓存写入到文件中。\n *\n * Since we are required to write the AOF before replying to the client,\n * and the only way the client socket can get a write is entering when the\n * the event loop, we accumulate all the AOF writes in a memory\n * buffer and write it on disk using this function just before entering\n * the event loop again.\n *\n * 因为程序需要在回复客户端之前对 AOF 执行写操作。\n * 而客户端能执行写操作的唯一机会就是在事件 loop 中，\n * 因此，程序将所有 AOF 写累积到缓存中，\n * 并在重新进入事件 loop 之前，将缓存写入到文件中。\n *\n * About the 'force' argument:\n *\n * 关于 force 参数：\n *\n * When the fsync policy is set to 'everysec' we may delay the flush if there\n * is still an fsync() going on in the background thread, since for instance\n * on Linux write(2) will be blocked by the background fsync anyway.\n *\n * 当 fsync 策略为每秒钟保存一次时，如果后台线程仍然有 fsync 在执行，\n * 那么我们可能会延迟执行冲洗（flush）操作，\n * 因为 Linux 上的 write(2) 会被后台的 fsync 阻塞。\n *\n * When this happens we remember that there is some aof buffer to be\n * flushed ASAP, and will try to do that in the serverCron() function.\n *\n * 当这种情况发生时，说明需要尽快冲洗 aof 缓存，\n * 程序会尝试在 serverCron() 函数中对缓存进行冲洗。\n *\n * However if force is set to 1 we'll write regardless of the background\n * fsync. \n *\n * 不过，如果 force 为 1 的话，那么不管后台是否正在 fsync ，\n * 程序都直接进行写入。\n */\n#define AOF_WRITE_LOG_ERROR_RATE 30 /* Seconds between errors logging. */\nvoid flushAppendOnlyFile(int force) {\n ssize_t nwritten;\n int sync_in_progress = 0;\n\n // 缓冲区中没有任何内容，直接返回\n if (sdslen(server.aof_buf) == 0) return;\n\n // 策略为每秒 FSYNC \n if (server.aof_fsync == AOF_FSYNC_EVERYSEC)\n // 是否有 SYNC 正在后台进行？\n sync_in_progress = bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC) != 0;\n\n // 每秒 fsync ，并且强制写入为假\n if (server.aof_fsync == AOF_FSYNC_EVERYSEC && !force) {\n\n /* With this append fsync policy we do background fsyncing.\n *\n * 当 fsync 策略为每秒钟一次时， fsync 在后台执行。\n *\n * If the fsync is still in progress we can try to delay\n * the write for a couple of seconds. \n *\n * 如果后台仍在执行 FSYNC ，那么我们可以延迟写操作一两秒\n * （如果强制执行 write 的话，服务器主线程将阻塞在 write 上面）\n */\n if (sync_in_progress) {\n\n // 有 fsync 正在后台进行 。。。\n\n if (server.aof_flush_postponed_start == 0) {\n /* No previous write postponinig, remember that we are\n * postponing the flush and return. \n *\n * 前面没有推迟过 write 操作，这里将推迟写操作的时间记录下来\n * 然后就返回，不执行 write 或者 fsync\n */\n server.aof_flush_postponed_start = server.unixtime;\n return;\n\n } else if (server.unixtime - server.aof_flush_postponed_start < 2) {\n /* We were already waiting for fsync to finish, but for less\n * than two seconds this is still ok. Postpone again. \n *\n * 如果之前已经因为 fsync 而推迟了 write 操作\n * 但是推迟的时间不超过 2 秒，那么直接返回\n * 不执行 write 或者 fsync\n */\n return;\n\n }\n\n /* Otherwise fall trough, and go write since we can't wait\n * over two seconds. \n *\n * 如果后台还有 fsync 在执行，并且 write 已经推迟 >= 2 秒\n * 那么执行写操作（write 将被阻塞）\n */\n server.aof_delayed_fsync++;\n redisLog(REDIS_NOTICE,\"Asynchronous AOF fsync is taking too long (disk is busy?). Writing the AOF buffer without waiting for fsync to complete, this may slow down Redis.\");\n }\n }\n\n /* If you are following this code path, then we are going to write so\n * set reset the postponed flush sentinel to zero. \n *\n * 执行到这里，程序会对 AOF 文件进行写入。\n *\n * 清零延迟 write 的时间记录\n */\n server.aof_flush_postponed_start = 0;\n\n /* We want to perform a single write. This should be guaranteed atomic\n * at least if the filesystem we are writing is a real physical one.\n *\n * 执行单个 write 操作，如果写入设备是物理的话，那么这个操作应该是原子的\n *\n * While this will save us against the server being killed I don't think\n * there is much to do about the whole server stopping for power problems\n * or alike \n *\n * 当然，如果出现像电源中断这样的不可抗现象，那么 AOF 文件也是可能会出现问题的\n * 这时就要用 redis-check-aof 程序来进行修复。\n */\n nwritten = write(server.aof_fd,server.aof_buf,sdslen(server.aof_buf));\n if (nwritten != (signed)sdslen(server.aof_buf)) {\n\n static time_t last_write_error_log = 0;\n int can_log = 0;\n\n /* Limit logging rate to 1 line per AOF_WRITE_LOG_ERROR_RATE seconds. */\n // 将日志的记录频率限制在每行 AOF_WRITE_LOG_ERROR_RATE 秒\n if ((server.unixtime - last_write_error_log) > AOF_WRITE_LOG_ERROR_RATE) {\n can_log = 1;\n last_write_error_log = server.unixtime;\n }\n\n /* Lof the AOF write error and record the error code. */\n // 如果写入出错，那么尝试将该情况写入到日志里面\n if (nwritten == -1) {\n if (can_log) {\n redisLog(REDIS_WARNING,\"Error writing to the AOF file: %s\",\n strerror(errno));\n server.aof_last_write_errno = errno;\n }\n } else {\n if (can_log) {\n redisLog(REDIS_WARNING,\"Short write while writing to \"\n \"the AOF file: (nwritten=%lld, \"\n \"expected=%lld)\",\n (long long)nwritten,\n (long long)sdslen(server.aof_buf));\n }\n\n // 尝试移除新追加的不完整内容\n if (ftruncate(server.aof_fd, server.aof_current_size) == -1) {\n if (can_log) {\n redisLog(REDIS_WARNING, \"Could not remove short write \"\n \"from the append-only file. Redis may refuse \"\n \"to load the AOF the next time it starts. \"\n \"ftruncate: %s\", strerror(errno));\n }\n } else {\n /* If the ftrunacate() succeeded we can set nwritten to\n * -1 since there is no longer partial data into the AOF. */\n nwritten = -1;\n }\n server.aof_last_write_errno = ENOSPC;\n }\n\n /* Handle the AOF write error. */\n // 处理写入 AOF 文件时出现的错误\n if (server.aof_fsync == AOF_FSYNC_ALWAYS) {\n /* We can't recover when the fsync policy is ALWAYS since the\n * reply for the client is already in the output buffers, and we\n * have the contract with the user that on acknowledged write data\n * is synched on disk. */\n redisLog(REDIS_WARNING,\"Can't recover from AOF write error when the AOF fsync policy is 'always'. Exiting...\");\n exit(1);\n } else {\n /* Recover from failed write leaving data into the buffer. However\n * set an error to stop accepting writes as long as the error\n * condition is not cleared. */\n server.aof_last_write_status = REDIS_ERR;\n\n /* Trim the sds buffer if there was a partial write, and there\n * was no way to undo it with ftruncate(2). */\n if (nwritten > 0) {\n server.aof_current_size += nwritten;\n sdsrange(server.aof_buf,nwritten,-1);\n }\n return; /* We'll try again on the next call... */\n }\n } else {\n /* Successful write(2). If AOF was in error state, restore the\n * OK state and log the event. */\n // 写入成功，更新最后写入状态\n if (server.aof_last_write_status == REDIS_ERR) {\n redisLog(REDIS_WARNING,\n \"AOF write error looks solved, Redis can write again.\");\n server.aof_last_write_status = REDIS_OK;\n }\n }\n\n // 更新写入后的 AOF 文件大小\n server.aof_current_size += nwritten;\n\n /* Re-use AOF buffer when it is small enough. The maximum comes from the\n * arena size of 4k minus some overhead (but is otherwise arbitrary). \n *\n * 如果 AOF 缓存的大小足够小的话，那么重用这个缓存，\n * 否则的话，释放 AOF 缓存。\n */\n if ((sdslen(server.aof_buf)+sdsavail(server.aof_buf)) < 4000) {\n // 清空缓存中的内容，等待重用\n sdsclear(server.aof_buf);\n } else {\n // 释放缓存\n sdsfree(server.aof_buf);\n server.aof_buf = sdsempty();\n }\n\n /* Don't fsync if no-appendfsync-on-rewrite is set to yes and there are\n * children doing I/O in the background. \n *\n * 如果 no-appendfsync-on-rewrite 选项为开启状态，\n * 并且有 BGSAVE 或者 BGREWRITEAOF 正在进行的话，\n * 那么不执行 fsync \n */\n if (server.aof_no_fsync_on_rewrite &&\n (server.aof_child_pid != -1 || server.rdb_child_pid != -1))\n return;\n\n /* Perform the fsync if needed. */\n\n // 总是执行 fsnyc\n if (server.aof_fsync == AOF_FSYNC_ALWAYS) {\n /* aof_fsync is defined as fdatasync() for Linux in order to avoid\n * flushing metadata. */\n aof_fsync(server.aof_fd); /* Let's try to get this data on the disk */\n\n // 更新最后一次执行 fsnyc 的时间\n server.aof_last_fsync = server.unixtime;\n\n // 策略为每秒 fsnyc ，并且距离上次 fsync 已经超过 1 秒\n } else if ((server.aof_fsync == AOF_FSYNC_EVERYSEC &&\n server.unixtime > server.aof_last_fsync)) {\n // 放到后台执行\n if (!sync_in_progress) aof_background_fsync(server.aof_fd);\n // 更新最后一次执行 fsync 的时间\n server.aof_last_fsync = server.unixtime;\n }\n\n // 其实上面无论执行 if 部分还是 else 部分都要更新 fsync 的时间\n // 可以将代码挪到下面来\n // server.aof_last_fsync = server.unixtime;\n}\n\n/*\n * 根据传入的命令和命令参数，将它们还原成协议格式。\n */\nsds catAppendOnlyGenericCommand(sds dst, int argc, robj **argv) {\n char buf[32];\n int len, j;\n robj *o;\n\n // 重建命令的个数，格式为 *\\r\\n\n // 例如 *3\\r\\n\n buf[0] = '*';\n len = 1+ll2string(buf+1,sizeof(buf)-1,argc);\n buf[len++] = '\\r';\n buf[len++] = '\\n';\n dst = sdscatlen(dst,buf,len);\n\n // 重建命令和命令参数，格式为 $\\r\\n\\r\\n\n // 例如 $3\\r\\nSET\\r\\n$3\\r\\nKEY\\r\\n$5\\r\\nVALUE\\r\\n\n for (j = 0; j < argc; j++) {\n o = getDecodedObject(argv[j]);\n\n // 组合 $\\r\\n\n buf[0] = '$';\n len = 1+ll2string(buf+1,sizeof(buf)-1,sdslen(o->ptr));\n buf[len++] = '\\r';\n buf[len++] = '\\n';\n dst = sdscatlen(dst,buf,len);\n\n // 组合 \\r\\n\n dst = sdscatlen(dst,o->ptr,sdslen(o->ptr));\n dst = sdscatlen(dst,\"\\r\\n\",2);\n\n decrRefCount(o);\n }\n\n // 返回重建后的协议内容\n return dst;\n}\n\n/* Create the sds representation of an PEXPIREAT command, using\n * 'seconds' as time to live and 'cmd' to understand what command\n * we are translating into a PEXPIREAT.\n *\n * 创建 PEXPIREAT 命令的 sds 表示，\n * cmd 参数用于指定转换的源指令， seconds 为 TTL （剩余生存时间）。\n *\n * This command is used in order to translate EXPIRE and PEXPIRE commands\n * into PEXPIREAT command so that we retain precision in the append only\n * file, and the time is always absolute and not relative.\n *\n * 这个函数用于将 EXPIRE 、 PEXPIRE 和 EXPIREAT 转换为 PEXPIREAT \n * 从而在保证精确度不变的情况下，将过期时间从相对值转换为绝对值（一个 UNIX 时间戳）。\n *\n * （过期时间必须是绝对值，这样不管 AOF 文件何时被载入，该过期的 key 都会正确地过期。）\n */\nsds catAppendOnlyExpireAtCommand(sds buf, struct redisCommand *cmd, robj *key, robj *seconds) {\n long long when;\n robj *argv[3];\n\n /* Make sure we can use strtol \n *\n * 取出过期值\n */\n seconds = getDecodedObject(seconds);\n when = strtoll(seconds->ptr,NULL,10);\n\n /* Convert argument into milliseconds for EXPIRE, SETEX, EXPIREAT \n *\n * 如果过期值的格式为秒，那么将它转换为毫秒\n */\n if (cmd->proc == expireCommand || cmd->proc == setexCommand ||\n cmd->proc == expireatCommand)\n {\n when *= 1000;\n }\n\n /* Convert into absolute time for EXPIRE, PEXPIRE, SETEX, PSETEX \n *\n * 如果过期值的格式为相对值，那么将它转换为绝对值\n */\n if (cmd->proc == expireCommand || cmd->proc == pexpireCommand ||\n cmd->proc == setexCommand || cmd->proc == psetexCommand)\n {\n when += mstime();\n }\n\n decrRefCount(seconds);\n\n // 构建 PEXPIREAT 命令\n argv[0] = createStringObject(\"PEXPIREAT\",9);\n argv[1] = key;\n argv[2] = createStringObjectFromLongLong(when);\n\n // 追加到 AOF 缓存中\n buf = catAppendOnlyGenericCommand(buf, 3, argv);\n\n decrRefCount(argv[0]);\n decrRefCount(argv[2]);\n\n return buf;\n}\n\n/*\n * 将命令追加到 AOF 文件中，\n * 如果 AOF 重写正在进行，那么也将命令追加到 AOF 重写缓存中。\n */\nvoid feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) {\n sds buf = sdsempty();\n robj *tmpargv[3];\n\n /* The DB this command was targeting is not the same as the last command\n * we appendend. To issue a SELECT command is needed. \n *\n * 使用 SELECT 命令，显式设置数据库，确保之后的命令被设置到正确的数据库\n */\n if (dictid != server.aof_selected_db) {\n char seldb[64];\n\n snprintf(seldb,sizeof(seldb),\"%d\",dictid);\n buf = sdscatprintf(buf,\"*2\\r\\n$6\\r\\nSELECT\\r\\n$%lu\\r\\n%s\\r\\n\",\n (unsigned long)strlen(seldb),seldb);\n\n server.aof_selected_db = dictid;\n }\n\n // EXPIRE 、 PEXPIRE 和 EXPIREAT 命令\n if (cmd->proc == expireCommand || cmd->proc == pexpireCommand ||\n cmd->proc == expireatCommand) {\n /* Translate EXPIRE/PEXPIRE/EXPIREAT into PEXPIREAT \n *\n * 将 EXPIRE 、 PEXPIRE 和 EXPIREAT 都翻译成 PEXPIREAT\n */\n buf = catAppendOnlyExpireAtCommand(buf,cmd,argv[1],argv[2]);\n\n // SETEX 和 PSETEX 命令\n } else if (cmd->proc == setexCommand || cmd->proc == psetexCommand) {\n /* Translate SETEX/PSETEX to SET and PEXPIREAT \n *\n * 将两个命令都翻译成 SET 和 PEXPIREAT\n */\n\n // SET\n tmpargv[0] = createStringObject(\"SET\",3);\n tmpargv[1] = argv[1];\n tmpargv[2] = argv[3];\n buf = catAppendOnlyGenericCommand(buf,3,tmpargv);\n\n // PEXPIREAT\n decrRefCount(tmpargv[0]);\n buf = catAppendOnlyExpireAtCommand(buf,cmd,argv[1],argv[2]);\n\n // 其他命令\n } else {\n /* All the other commands don't need translation or need the\n * same translation already operated in the command vector\n * for the replication itself. */\n buf = catAppendOnlyGenericCommand(buf,argc,argv);\n }\n\n /* Append to the AOF buffer. This will be flushed on disk just before\n * of re-entering the event loop, so before the client will get a\n * positive reply about the operation performed. \n *\n * 将命令追加到 AOF 缓存中，\n * 在重新进入事件循环之前，这些命令会被冲洗到磁盘上，\n * 并向客户端返回一个回复。\n */\n if (server.aof_state == REDIS_AOF_ON)\n server.aof_buf = sdscatlen(server.aof_buf,buf,sdslen(buf));\n\n /* If a background append only file rewriting is in progress we want to\n * accumulate the differences between the child DB and the current one\n * in a buffer, so that when the child process will do its work we\n * can append the differences to the new append only file. \n *\n * 如果 BGREWRITEAOF 正在进行，\n * 那么我们还需要将命令追加到重写缓存中，\n * 从而记录当前正在重写的 AOF 文件和数据库当前状态的差异。\n */\n if (server.aof_child_pid != -1)\n aofRewriteBufferAppend((unsigned char*)buf,sdslen(buf));\n\n // 释放\n sdsfree(buf);\n}\n\n/* ----------------------------------------------------------------------------\n * AOF loading\n * ------------------------------------------------------------------------- */\n\n/* In Redis commands are always executed in the context of a client, so in\n * order to load the append only file we need to create a fake client. \n *\n * Redis 命令必须由客户端执行，\n * 所以 AOF 装载程序需要创建一个无网络连接的客户端来执行 AOF 文件中的命令。\n */\nstruct redisClient *createFakeClient(void) {\n struct redisClient *c = zmalloc(sizeof(*c));\n\n selectDb(c,0);\n\n c->fd = -1;\n c->name = NULL;\n c->querybuf = sdsempty();\n c->querybuf_peak = 0;\n c->argc = 0;\n c->argv = NULL;\n c->bufpos = 0;\n c->flags = 0;\n c->btype = REDIS_BLOCKED_NONE;\n /* We set the fake client as a slave waiting for the synchronization\n * so that Redis will not try to send replies to this client. \n *\n * 将客户端设置为正在等待同步的附属节点，这样客户端就不会发送回复了。\n */\n c->replstate = REDIS_REPL_WAIT_BGSAVE_START;\n c->reply = listCreate();\n c->reply_bytes = 0;\n c->obuf_soft_limit_reached_time = 0;\n c->watched_keys = listCreate();\n c->peerid = NULL;\n listSetFreeMethod(c->reply,decrRefCountVoid);\n listSetDupMethod(c->reply,dupClientReplyValue);\n initClientMultiState(c);\n\n return c;\n}\n\n/*\n * 释放伪客户端\n */\nvoid freeFakeClient(struct redisClient *c) {\n\n // 释放查询缓存\n sdsfree(c->querybuf);\n\n // 释放回复缓存\n listRelease(c->reply);\n\n // 释放监视的键\n listRelease(c->watched_keys);\n\n // 释放事务状态\n freeClientMultiState(c);\n\n zfree(c);\n}\n\n/* Replay the append log file. On error REDIS_OK is returned. On non fatal\n * error (the append only file is zero-length) REDIS_ERR is returned. On\n * fatal error an error message is logged and the program exists.\n *\n * 执行 AOF 文件中的命令。\n *\n * 出错时返回 REDIS_OK 。\n *\n * 出现非执行错误（比如文件长度为 0 ）时返回 REDIS_ERR 。\n *\n * 出现致命错误时打印信息到日志，并且程序退出。\n */\nint loadAppendOnlyFile(char *filename) {\n\n // 为客户端\n struct redisClient *fakeClient;\n\n // 打开 AOF 文件\n FILE *fp = fopen(filename,\"r\");\n\n struct redis_stat sb;\n int old_aof_state = server.aof_state;\n long loops = 0;\n\n // 检查文件的正确性\n if (fp && redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) {\n server.aof_current_size = 0;\n fclose(fp);\n return REDIS_ERR;\n }\n\n // 检查文件是否正常打开\n if (fp == NULL) {\n redisLog(REDIS_WARNING,\"Fatal error: can't open the append log file for reading: %s\",strerror(errno));\n exit(1);\n }\n\n /* Temporarily disable AOF, to prevent EXEC from feeding a MULTI\n * to the same file we're about to read. \n *\n * 暂时性地关闭 AOF ，防止在执行 MULTI 时，\n * EXEC 命令被传播到正在打开的 AOF 文件中。\n */\n server.aof_state = REDIS_AOF_OFF;\n\n fakeClient = createFakeClient();\n\n // 设置服务器的状态为：正在载入\n // startLoading 定义于 rdb.c\n startLoading(fp);\n\n while(1) {\n int argc, j;\n unsigned long len;\n robj **argv;\n char buf[128];\n sds argsds;\n struct redisCommand *cmd;\n\n /* Serve the clients from time to time \n *\n * 间隔性地处理客户端发送来的请求\n * 因为服务器正处于载入状态，所以能正常执行的只有 PUBSUB 等模块\n */\n if (!(loops++ % 1000)) {\n loadingProgress(ftello(fp));\n processEventsWhileBlocked();\n }\n\n // 读入文件内容到缓存\n if (fgets(buf,sizeof(buf),fp) == NULL) {\n if (feof(fp))\n // 文件已经读完，跳出\n break;\n else\n goto readerr;\n }\n\n // 确认协议格式，比如 *3\\r\\n\n if (buf[0] != '*') goto fmterr;\n \n // 取出命令参数，比如 *3\\r\\n 中的 3\n argc = atoi(buf+1);\n\n // 至少要有一个参数（被调用的命令）\n if (argc < 1) goto fmterr;\n\n // 从文本中创建字符串对象：包括命令，以及命令参数\n // 例如 $3\\r\\nSET\\r\\n$3\\r\\nKEY\\r\\n$5\\r\\nVALUE\\r\\n\n // 将创建三个包含以下内容的字符串对象：\n // SET 、 KEY 、 VALUE\n argv = zmalloc(sizeof(robj*)*argc);\n for (j = 0; j < argc; j++) {\n if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr;\n\n if (buf[0] != '$') goto fmterr;\n\n // 读取参数值的长度\n len = strtol(buf+1,NULL,10);\n // 读取参数值\n argsds = sdsnewlen(NULL,len);\n if (len && fread(argsds,len,1,fp) == 0) goto fmterr;\n // 为参数创建对象\n argv[j] = createObject(REDIS_STRING,argsds);\n\n if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */\n }\n\n /* Command lookup \n *\n * 查找命令\n */\n cmd = lookupCommand(argv[0]->ptr);\n if (!cmd) {\n redisLog(REDIS_WARNING,\"Unknown command '%s' reading the append only file\", (char*)argv[0]->ptr);\n exit(1);\n }\n\n /* Run the command in the context of a fake client \n *\n * 调用伪客户端，执行命令\n */\n fakeClient->argc = argc;\n fakeClient->argv = argv;\n cmd->proc(fakeClient);\n\n /* The fake client should not have a reply */\n redisAssert(fakeClient->bufpos == 0 && listLength(fakeClient->reply) == 0);\n /* The fake client should never get blocked */\n redisAssert((fakeClient->flags & REDIS_BLOCKED) == 0);\n\n /* Clean up. Command code may have changed argv/argc so we use the\n * argv/argc of the client instead of the local variables. \n *\n * 清理命令和命令参数对象\n */\n for (j = 0; j < fakeClient->argc; j++)\n decrRefCount(fakeClient->argv[j]);\n zfree(fakeClient->argv);\n }\n\n /* This point can only be reached when EOF is reached without errors.\n * If the client is in the middle of a MULTI/EXEC, log error and quit. \n *\n * 如果能执行到这里，说明 AOF 文件的全部内容都可以正确地读取，\n * 但是，还要检查 AOF 是否包含未正确结束的事务\n */\n if (fakeClient->flags & REDIS_MULTI) goto readerr;\n\n // 关闭 AOF 文件\n fclose(fp);\n // 释放伪客户端\n freeFakeClient(fakeClient);\n // 复原 AOF 状态\n server.aof_state = old_aof_state;\n // 停止载入\n stopLoading();\n // 更新服务器状态中， AOF 文件的当前大小\n aofUpdateCurrentSize();\n // 记录前一次重写时的大小\n server.aof_rewrite_base_size = server.aof_current_size;\n \n return REDIS_OK;\n\n// 读入错误\nreaderr:\n // 非预期的末尾，可能是 AOF 文件在写入的中途遭遇了停机\n if (feof(fp)) {\n redisLog(REDIS_WARNING,\"Unexpected end of file reading the append only file\");\n \n // 文件内容出错\n } else {\n redisLog(REDIS_WARNING,\"Unrecoverable error reading the append only file: %s\", strerror(errno));\n }\n exit(1);\n\n// 内容格式错误\nfmterr:\n redisLog(REDIS_WARNING,\"Bad file format reading the append only file: make a backup of your AOF file, then use ./redis-check-aof --fix \");\n exit(1);\n}\n\n/* ----------------------------------------------------------------------------\n * AOF rewrite\n * ------------------------------------------------------------------------- */\n\n/* Delegate writing an object to writing a bulk string or bulk long long.\n * This is not placed in rio.c since that adds the redis.h dependency. \n *\n * 将 obj 所指向的整数对象或字符串对象的值写入到 r 当中。\n */\nint rioWriteBulkObject(rio *r, robj *obj) {\n /* Avoid using getDecodedObject to help copy-on-write (we are often\n * in a child process when this function is called). */\n if (obj->encoding == REDIS_ENCODING_INT) {\n return rioWriteBulkLongLong(r,(long)obj->ptr);\n } else if (sdsEncodedObject(obj)) {\n return rioWriteBulkString(r,obj->ptr,sdslen(obj->ptr));\n } else {\n redisPanic(\"Unknown string encoding\");\n }\n}\n\n/* Emit the commands needed to rebuild a list object.\n * The function returns 0 on error, 1 on success. \n *\n * 将重建列表对象所需的命令写入到 r 。\n *\n * 出错返回 0 ，成功返回 1 。\n *\n * 命令的形式如下： RPUSH item1 item2 ... itemN\n */\nint rewriteListObject(rio *r, robj *key, robj *o) {\n long long count = 0, items = listTypeLength(o);\n\n if (o->encoding == REDIS_ENCODING_ZIPLIST) {\n unsigned char *zl = o->ptr;\n unsigned char *p = ziplistIndex(zl,0);\n unsigned char *vstr;\n unsigned int vlen;\n long long vlong;\n\n // 先构建一个 RPUSH key \n // 然后从 ZIPLIST 中取出最多 REDIS_AOF_REWRITE_ITEMS_PER_CMD 个元素\n // 之后重复第一步，直到 ZIPLIST 为空\n while(ziplistGet(p,&vstr,&vlen,&vlong)) {\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items) == 0) return 0;\n if (rioWriteBulkString(r,\"RPUSH\",5) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n // 取出值\n if (vstr) {\n if (rioWriteBulkString(r,(char*)vstr,vlen) == 0) return 0;\n } else {\n if (rioWriteBulkLongLong(r,vlong) == 0) return 0;\n }\n // 移动指针，并计算被取出元素的数量\n p = ziplistNext(zl,p);\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {\n list *list = o->ptr;\n listNode *ln;\n listIter li;\n\n // 先构建一个 RPUSH key \n // 然后从双端链表中取出最多 REDIS_AOF_REWRITE_ITEMS_PER_CMD 个元素\n // 之后重复第一步，直到链表为空\n listRewind(list,&li);\n while((ln = listNext(&li))) {\n robj *eleobj = listNodeValue(ln);\n\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items) == 0) return 0;\n if (rioWriteBulkString(r,\"RPUSH\",5) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n\n // 取出值\n if (rioWriteBulkObject(r,eleobj) == 0) return 0;\n\n // 元素计数\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n\n items--;\n }\n } else {\n redisPanic(\"Unknown list encoding\");\n }\n return 1;\n}\n\n/* Emit the commands needed to rebuild a set object.\n * The function returns 0 on error, 1 on success. \n *\n * 将重建集合对象所需的命令写入到 r 。\n *\n * 出错返回 0 ，成功返回 1 。\n *\n * 命令的形式如下： SADD item1 item2 ... itemN\n */\nint rewriteSetObject(rio *r, robj *key, robj *o) {\n long long count = 0, items = setTypeSize(o);\n\n if (o->encoding == REDIS_ENCODING_INTSET) {\n int ii = 0;\n int64_t llval;\n\n while(intsetGet(o->ptr,ii++,&llval)) {\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items) == 0) return 0;\n if (rioWriteBulkString(r,\"SADD\",4) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n if (rioWriteBulkLongLong(r,llval) == 0) return 0;\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n } else if (o->encoding == REDIS_ENCODING_HT) {\n dictIterator *di = dictGetIterator(o->ptr);\n dictEntry *de;\n\n while((de = dictNext(di)) != NULL) {\n robj *eleobj = dictGetKey(de);\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items) == 0) return 0;\n if (rioWriteBulkString(r,\"SADD\",4) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n if (rioWriteBulkObject(r,eleobj) == 0) return 0;\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n dictReleaseIterator(di);\n } else {\n redisPanic(\"Unknown set encoding\");\n }\n return 1;\n}\n\n/* Emit the commands needed to rebuild a sorted set object.\n * The function returns 0 on error, 1 on success. \n *\n * 将重建有序集合对象所需的命令写入到 r 。\n *\n * 出错返回 0 ，成功返回 1 。\n *\n * 命令的形式如下： ZADD score1 member1 score2 member2 ... scoreN memberN\n */\nint rewriteSortedSetObject(rio *r, robj *key, robj *o) {\n long long count = 0, items = zsetLength(o);\n\n if (o->encoding == REDIS_ENCODING_ZIPLIST) {\n unsigned char *zl = o->ptr;\n unsigned char *eptr, *sptr;\n unsigned char *vstr;\n unsigned int vlen;\n long long vll;\n double score;\n\n eptr = ziplistIndex(zl,0);\n redisAssert(eptr != NULL);\n sptr = ziplistNext(zl,eptr);\n redisAssert(sptr != NULL);\n\n while (eptr != NULL) {\n redisAssert(ziplistGet(eptr,&vstr,&vlen,&vll));\n score = zzlGetScore(sptr);\n\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items*2) == 0) return 0;\n if (rioWriteBulkString(r,\"ZADD\",4) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n if (rioWriteBulkDouble(r,score) == 0) return 0;\n if (vstr != NULL) {\n if (rioWriteBulkString(r,(char*)vstr,vlen) == 0) return 0;\n } else {\n if (rioWriteBulkLongLong(r,vll) == 0) return 0;\n }\n zzlNext(zl,&eptr,&sptr);\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n } else if (o->encoding == REDIS_ENCODING_SKIPLIST) {\n zset *zs = o->ptr;\n dictIterator *di = dictGetIterator(zs->dict);\n dictEntry *de;\n\n while((de = dictNext(di)) != NULL) {\n robj *eleobj = dictGetKey(de);\n double *score = dictGetVal(de);\n\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items*2) == 0) return 0;\n if (rioWriteBulkString(r,\"ZADD\",4) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n if (rioWriteBulkDouble(r,*score) == 0) return 0;\n if (rioWriteBulkObject(r,eleobj) == 0) return 0;\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n dictReleaseIterator(di);\n } else {\n redisPanic(\"Unknown sorted zset encoding\");\n }\n return 1;\n}\n\n/* Write either the key or the value of the currently selected item of a hash.\n *\n * 选择写入哈希的 key 或者 value 到 r 中。\n *\n * The 'hi' argument passes a valid Redis hash iterator.\n *\n * hi 为 Redis 哈希迭代器\n *\n * The 'what' filed specifies if to write a key or a value and can be\n * either REDIS_HASH_KEY or REDIS_HASH_VALUE.\n *\n * what 决定了要写入的部分，可以是 REDIS_HASH_KEY 或 REDIS_HASH_VALUE\n *\n * The function returns 0 on error, non-zero on success. \n *\n * 出错返回 0 ，成功返回非 0 。\n */\nstatic int rioWriteHashIteratorCursor(rio *r, hashTypeIterator *hi, int what) {\n\n if (hi->encoding == REDIS_ENCODING_ZIPLIST) {\n unsigned char *vstr = NULL;\n unsigned int vlen = UINT_MAX;\n long long vll = LLONG_MAX;\n\n hashTypeCurrentFromZiplist(hi, what, &vstr, &vlen, &vll);\n if (vstr) {\n return rioWriteBulkString(r, (char*)vstr, vlen);\n } else {\n return rioWriteBulkLongLong(r, vll);\n }\n\n } else if (hi->encoding == REDIS_ENCODING_HT) {\n robj *value;\n\n hashTypeCurrentFromHashTable(hi, what, &value);\n return rioWriteBulkObject(r, value);\n }\n\n redisPanic(\"Unknown hash encoding\");\n return 0;\n}\n\n/* Emit the commands needed to rebuild a hash object.\n * The function returns 0 on error, 1 on success. \n *\n * 将重建哈希对象所需的命令写入到 r 。\n *\n * 出错返回 0 ，成功返回 1 。\n *\n * 命令的形式如下：HMSET field1 value1 field2 value2 ... fieldN valueN\n */\nint rewriteHashObject(rio *r, robj *key, robj *o) {\n hashTypeIterator *hi;\n long long count = 0, items = hashTypeLength(o);\n\n hi = hashTypeInitIterator(o);\n while (hashTypeNext(hi) != REDIS_ERR) {\n if (count == 0) {\n int cmd_items = (items > REDIS_AOF_REWRITE_ITEMS_PER_CMD) ?\n REDIS_AOF_REWRITE_ITEMS_PER_CMD : items;\n\n if (rioWriteBulkCount(r,'*',2+cmd_items*2) == 0) return 0;\n if (rioWriteBulkString(r,\"HMSET\",5) == 0) return 0;\n if (rioWriteBulkObject(r,key) == 0) return 0;\n }\n\n if (rioWriteHashIteratorCursor(r, hi, REDIS_HASH_KEY) == 0) return 0;\n if (rioWriteHashIteratorCursor(r, hi, REDIS_HASH_VALUE) == 0) return 0;\n if (++count == REDIS_AOF_REWRITE_ITEMS_PER_CMD) count = 0;\n items--;\n }\n\n hashTypeReleaseIterator(hi);\n\n return 1;\n}\n\n/* Write a sequence of commands able to fully rebuild the dataset into\n * \"filename\". Used both by REWRITEAOF and BGREWRITEAOF.\n *\n * 将一集足以还原当前数据集的命令写入到 filename 指定的文件中。\n *\n * 这个函数被 REWRITEAOF 和 BGREWRITEAOF 两个命令调用。\n * （REWRITEAOF 似乎已经是一个废弃的命令）\n *\n * In order to minimize the number of commands needed in the rewritten\n * log Redis uses variadic commands when possible, such as RPUSH, SADD\n * and ZADD. However at max REDIS_AOF_REWRITE_ITEMS_PER_CMD items per time\n * are inserted using a single command. \n *\n * 为了最小化重建数据集所需执行的命令数量，\n * Redis 会尽可能地使用接受可变参数数量的命令，比如 RPUSH 、SADD 和 ZADD 等。\n *\n * 不过单个命令每次处理的元素数量不能超过 REDIS_AOF_REWRITE_ITEMS_PER_CMD 。\n */\nint rewriteAppendOnlyFile(char *filename) {\n dictIterator *di = NULL;\n dictEntry *de;\n rio aof;\n FILE *fp;\n char tmpfile[256];\n int j;\n long long now = mstime();\n\n /* Note that we have to use a different temp name here compared to the\n * one used by rewriteAppendOnlyFileBackground() function. \n *\n * 创建临时文件\n *\n * 注意这里创建的文件名和 rewriteAppendOnlyFileBackground() 创建的文件名稍有不同\n */\n snprintf(tmpfile,256,\"temp-rewriteaof-%d.aof\", (int) getpid());\n fp = fopen(tmpfile,\"w\");\n if (!fp) {\n redisLog(REDIS_WARNING, \"Opening the temp file for AOF rewrite in rewriteAppendOnlyFile(): %s\", strerror(errno));\n return REDIS_ERR;\n }\n\n // 初始化文件 io\n rioInitWithFile(&aof,fp);\n\n // 设置每写入 REDIS_AOF_AUTOSYNC_BYTES 字节\n // 就执行一次 FSYNC \n // 防止缓存中积累太多命令内容，造成 I/O 阻塞时间过长\n if (server.aof_rewrite_incremental_fsync)\n rioSetAutoSync(&aof,REDIS_AOF_AUTOSYNC_BYTES);\n\n // 遍历所有数据库\n for (j = 0; j < server.dbnum; j++) {\n\n char selectcmd[] = \"*2\\r\\n$6\\r\\nSELECT\\r\\n\";\n\n redisDb *db = server.db+j;\n\n // 指向键空间\n dict *d = db->dict;\n if (dictSize(d) == 0) continue;\n\n // 创建键空间迭代器\n di = dictGetSafeIterator(d);\n if (!di) {\n fclose(fp);\n return REDIS_ERR;\n }\n\n /* SELECT the new DB \n *\n * 首先写入 SELECT 命令，确保之后的数据会被插入到正确的数据库上\n */\n if (rioWrite(&aof,selectcmd,sizeof(selectcmd)-1) == 0) goto werr;\n if (rioWriteBulkLongLong(&aof,j) == 0) goto werr;\n\n /* Iterate this DB writing every entry \n *\n * 遍历数据库所有键，并通过命令将它们的当前状态（值）记录到新 AOF 文件中\n */\n while((de = dictNext(di)) != NULL) {\n sds keystr;\n robj key, *o;\n long long expiretime;\n\n // 取出键\n keystr = dictGetKey(de);\n\n // 取出值\n o = dictGetVal(de);\n initStaticStringObject(key,keystr);\n\n // 取出过期时间\n expiretime = getExpire(db,&key);\n\n /* If this key is already expired skip it \n *\n * 如果键已经过期，那么跳过它，不保存\n */\n if (expiretime != -1 && expiretime < now) continue;\n\n /* Save the key and associated value \n *\n * 根据值的类型，选择适当的命令来保存值\n */\n if (o->type == REDIS_STRING) {\n /* Emit a SET command */\n char cmd[]=\"*3\\r\\n$3\\r\\nSET\\r\\n\";\n if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;\n /* Key and value */\n if (rioWriteBulkObject(&aof,&key) == 0) goto werr;\n if (rioWriteBulkObject(&aof,o) == 0) goto werr;\n } else if (o->type == REDIS_LIST) {\n if (rewriteListObject(&aof,&key,o) == 0) goto werr;\n } else if (o->type == REDIS_SET) {\n if (rewriteSetObject(&aof,&key,o) == 0) goto werr;\n } else if (o->type == REDIS_ZSET) {\n if (rewriteSortedSetObject(&aof,&key,o) == 0) goto werr;\n } else if (o->type == REDIS_HASH) {\n if (rewriteHashObject(&aof,&key,o) == 0) goto werr;\n } else {\n redisPanic(\"Unknown object type\");\n }\n\n /* Save the expire time \n *\n * 保存键的过期时间\n */\n if (expiretime != -1) {\n char cmd[]=\"*3\\r\\n$9\\r\\nPEXPIREAT\\r\\n\";\n\n // 写入 PEXPIREAT expiretime 命令\n if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;\n if (rioWriteBulkObject(&aof,&key) == 0) goto werr;\n if (rioWriteBulkLongLong(&aof,expiretime) == 0) goto werr;\n }\n }\n\n // 释放迭代器\n dictReleaseIterator(di);\n }\n\n /* Make sure data will not remain on the OS's output buffers */\n // 冲洗并关闭新 AOF 文件\n if (fflush(fp) == EOF) goto werr;\n if (aof_fsync(fileno(fp)) == -1) goto werr;\n if (fclose(fp) == EOF) goto werr;\n\n /* Use RENAME to make sure the DB file is changed atomically only\n * if the generate DB file is ok. \n *\n * 原子地改名，用重写后的新 AOF 文件覆盖旧 AOF 文件\n */\n if (rename(tmpfile,filename) == -1) {\n redisLog(REDIS_WARNING,\"Error moving temp append only file on the final destination: %s\", strerror(errno));\n unlink(tmpfile);\n return REDIS_ERR;\n }\n\n redisLog(REDIS_NOTICE,\"SYNC append only file rewrite performed\");\n\n return REDIS_OK;\n\nwerr:\n fclose(fp);\n unlink(tmpfile);\n redisLog(REDIS_WARNING,\"Write error writing append only file on disk: %s\", strerror(errno));\n if (di) dictReleaseIterator(di);\n return REDIS_ERR;\n}\n\n/* This is how rewriting of the append only file in background works:\n * \n * 以下是后台重写 AOF 文件（BGREWRITEAOF）的工作步骤：\n *\n * 1) The user calls BGREWRITEAOF\n * 用户调用 BGREWRITEAOF\n *\n * 2) Redis calls this function, that forks():\n * Redis 调用这个函数，它执行 fork() ：\n *\n * 2a) the child rewrite the append only file in a temp file.\n * 子进程在临时文件中对 AOF 文件进行重写\n *\n * 2b) the parent accumulates differences in server.aof_rewrite_buf.\n * 父进程将新输入的写命令追加到 server.aof_rewrite_buf 中\n *\n * 3) When the child finished '2a' exists.\n * 当步骤 2a 执行完之后，子进程结束\n *\n * 4) The parent will trap the exit code, if it's OK, will append the\n * data accumulated into server.aof_rewrite_buf into the temp file, and\n * finally will rename(2) the temp file in the actual file name.\n * The the new file is reopened as the new append only file. Profit!\n *\n * 父进程会捕捉子进程的退出信号，\n * 如果子进程的退出状态是 OK 的话，\n * 那么父进程将新输入命令的缓存追加到临时文件，\n * 然后使用 rename(2) 对临时文件改名，用它代替旧的 AOF 文件，\n * 至此，后台 AOF 重写完成。\n */\nint rewriteAppendOnlyFileBackground(void) {\n pid_t childpid;\n long long start;\n\n // 已经有进程在进行 AOF 重写了\n if (server.aof_child_pid != -1) return REDIS_ERR;\n\n // 记录 fork 开始前的时间，计算 fork 耗时用\n start = ustime();\n\n if ((childpid = fork()) == 0) {\n char tmpfile[256];\n\n /* Child */\n\n // 关闭网络连接 fd\n closeListeningSockets(0);\n\n // 为进程设置名字，方便记认\n redisSetProcTitle(\"redis-aof-rewrite\");\n\n // 创建临时文件，并进行 AOF 重写\n snprintf(tmpfile,256,\"temp-rewriteaof-bg-%d.aof\", (int) getpid());\n if (rewriteAppendOnlyFile(tmpfile) == REDIS_OK) {\n size_t private_dirty = zmalloc_get_private_dirty();\n\n if (private_dirty) {\n redisLog(REDIS_NOTICE,\n \"AOF rewrite: %zu MB of memory used by copy-on-write\",\n private_dirty/(1024*1024));\n }\n // 发送重写成功信号\n exitFromChild(0);\n } else {\n // 发送重写失败信号\n exitFromChild(1);\n }\n } else {\n /* Parent */\n // 记录执行 fork 所消耗的时间\n server.stat_fork_time = ustime()-start;\n\n if (childpid == -1) {\n redisLog(REDIS_WARNING,\n \"Can't rewrite append only file in background: fork: %s\",\n strerror(errno));\n return REDIS_ERR;\n }\n\n redisLog(REDIS_NOTICE,\n \"Background append only file rewriting started by pid %d\",childpid);\n\n // 记录 AOF 重写的信息\n server.aof_rewrite_scheduled = 0;\n server.aof_rewrite_time_start = time(NULL);\n server.aof_child_pid = childpid;\n\n // 关闭字典自动 rehash\n updateDictResizePolicy();\n\n /* We set appendseldb to -1 in order to force the next call to the\n * feedAppendOnlyFile() to issue a SELECT command, so the differences\n * accumulated by the parent into server.aof_rewrite_buf will start\n * with a SELECT statement and it will be safe to merge. \n *\n * 将 aof_selected_db 设为 -1 ，\n * 强制让 feedAppendOnlyFile() 下次执行时引发一个 SELECT 命令，\n * 从而确保之后新添加的命令会设置到正确的数据库中\n */\n server.aof_selected_db = -1;\n replicationScriptCacheFlush();\n return REDIS_OK;\n }\n return REDIS_OK; /* unreached */\n}\n\nvoid bgrewriteaofCommand(redisClient *c) {\n\n // 不能重复运行 BGREWRITEAOF\n if (server.aof_child_pid != -1) {\n addReplyError(c,\"Background append only file rewriting already in progress\");\n\n // 如果正在执行 BGSAVE ，那么预定 BGREWRITEAOF\n // 等 BGSAVE 完成之后， BGREWRITEAOF 就会开始执行\n } else if (server.rdb_child_pid != -1) {\n server.aof_rewrite_scheduled = 1;\n addReplyStatus(c,\"Background append only file rewriting scheduled\");\n\n // 执行 BGREWRITEAOF\n } else if (rewriteAppendOnlyFileBackground() == REDIS_OK) {\n addReplyStatus(c,\"Background append only file rewriting started\");\n\n } else {\n addReply(c,shared.err);\n }\n}\n\n/*\n * 删除 AOF 重写所产生的临时文件\n */\nvoid aofRemoveTempFile(pid_t childpid) {\n char tmpfile[256];\n\n snprintf(tmpfile,256,\"temp-rewriteaof-bg-%d.aof\", (int) childpid);\n unlink(tmpfile);\n}\n\n/* Update the server.aof_current_size filed explicitly using stat(2)\n * to check the size of the file. This is useful after a rewrite or after\n * a restart, normally the size is updated just adding the write length\n * to the current length, that is much faster. \n *\n * 将 aof 文件的当前大小记录到服务器状态中。\n *\n * 通常用于 BGREWRITEAOF 执行之后，或者服务器重启之后。\n */\nvoid aofUpdateCurrentSize(void) {\n struct redis_stat sb;\n\n // 读取文件状态\n if (redis_fstat(server.aof_fd,&sb) == -1) {\n redisLog(REDIS_WARNING,\"Unable to obtain the AOF file length. stat: %s\",\n strerror(errno));\n } else {\n // 设置到服务器\n server.aof_current_size = sb.st_size;\n }\n}\n\n/* A background append only file rewriting (BGREWRITEAOF) terminated its work.\n * Handle this. \n *\n * 当子线程完成 AOF 重写时，父进程调用这个函数。\n */\nvoid backgroundRewriteDoneHandler(int exitcode, int bysignal) {\n if (!bysignal && exitcode == 0) {\n int newfd, oldfd;\n char tmpfile[256];\n long long now = ustime();\n\n redisLog(REDIS_NOTICE,\n \"Background AOF rewrite terminated with success\");\n\n /* Flush the differences accumulated by the parent to the\n * rewritten AOF. */\n // 打开保存新 AOF 文件内容的临时文件\n snprintf(tmpfile,256,\"temp-rewriteaof-bg-%d.aof\",\n (int)server.aof_child_pid);\n newfd = open(tmpfile,O_WRONLY|O_APPEND);\n if (newfd == -1) {\n redisLog(REDIS_WARNING,\n \"Unable to open the temporary AOF produced by the child: %s\", strerror(errno));\n goto cleanup;\n }\n\n // 将累积的重写缓存写入到临时文件中\n // 这个函数调用的 write 操作会阻塞主进程\n if (aofRewriteBufferWrite(newfd) == -1) {\n redisLog(REDIS_WARNING,\n \"Error trying to flush the parent diff to the rewritten AOF: %s\", strerror(errno));\n close(newfd);\n goto cleanup;\n }\n\n redisLog(REDIS_NOTICE,\n \"Parent diff successfully flushed to the rewritten AOF (%lu bytes)\", aofRewriteBufferSize());\n\n /* The only remaining thing to do is to rename the temporary file to\n * the configured file and switch the file descriptor used to do AOF\n * writes. We don't want close(2) or rename(2) calls to block the\n * server on old file deletion.\n *\n * 剩下的工作就是将临时文件改名为 AOF 程序指定的文件名，\n * 并将新文件的 fd 设为 AOF 程序的写目标。\n *\n * 不过这里有一个问题 ——\n * 我们不想 close(2) 或者 rename(2) 在删除旧文件时阻塞。\n *\n * There are two possible scenarios:\n *\n * 以下是两个可能的场景：\n *\n * 1) AOF is DISABLED and this was a one time rewrite. The temporary\n * file will be renamed to the configured file. When this file already\n * exists, it will be unlinked, which may block the server.\n *\n * AOF 被关闭，这个是一次单次的写操作。\n * 临时文件会被改名为 AOF 文件。\n * 本来已经存在的 AOF 文件会被 unlink ，这可能会阻塞服务器。\n *\n * 2) AOF is ENABLED and the rewritten AOF will immediately start\n * receiving writes. After the temporary file is renamed to the\n * configured file, the original AOF file descriptor will be closed.\n * Since this will be the last reference to that file, closing it\n * causes the underlying file to be unlinked, which may block the\n * server.\n *\n * AOF 被开启，并且重写后的 AOF 文件会立即被用于接收新的写入命令。\n * 当临时文件被改名为 AOF 文件时，原来的 AOF 文件描述符会被关闭。\n * 因为 Redis 会是最后一个引用这个文件的进程，\n * 所以关闭这个文件会引起 unlink ，这可能会阻塞服务器。\n *\n * To mitigate the blocking effect of the unlink operation (either\n * caused by rename(2) in scenario 1, or by close(2) in scenario 2), we\n * use a background thread to take care of this. First, we\n * make scenario 1 identical to scenario 2 by opening the target file\n * when it exists. The unlink operation after the rename(2) will then\n * be executed upon calling close(2) for its descriptor. Everything to\n * guarantee atomicity for this switch has already happened by then, so\n * we don't care what the outcome or duration of that close operation\n * is, as long as the file descriptor is released again. \n *\n * 为了避免出现阻塞现象，程序会将 close(2) 放到后台线程执行，\n * 这样服务器就可以持续处理请求，不会被中断。\n */\n if (server.aof_fd == -1) {\n /* AOF disabled */\n\n /* Don't care if this fails: oldfd will be -1 and we handle that.\n * One notable case of -1 return is if the old file does\n * not exist. */\n oldfd = open(server.aof_filename,O_RDONLY|O_NONBLOCK);\n } else {\n /* AOF enabled */\n oldfd = -1; /* We'll set this to the current AOF filedes later. */\n }\n\n /* Rename the temporary file. This will not unlink the target file if\n * it exists, because we reference it with \"oldfd\". \n *\n * 对临时文件进行改名，替换现有的 AOF 文件。\n *\n * 旧的 AOF 文件不会在这里被 unlink ，因为 oldfd 引用了它。\n */\n if (rename(tmpfile,server.aof_filename) == -1) {\n redisLog(REDIS_WARNING,\n \"Error trying to rename the temporary AOF file: %s\", strerror(errno));\n close(newfd);\n if (oldfd != -1) close(oldfd);\n goto cleanup;\n }\n\n if (server.aof_fd == -1) {\n /* AOF disabled, we don't need to set the AOF file descriptor\n * to this new file, so we can close it. \n *\n * AOF 被关闭，直接关闭 AOF 文件，\n * 因为关闭 AOF 本来就会引起阻塞，所以这里就算 close 被阻塞也无所谓\n */\n close(newfd);\n } else {\n /* AOF enabled, replace the old fd with the new one. \n *\n * 用新 AOF 文件的 fd 替换原来 AOF 文件的 fd\n */\n oldfd = server.aof_fd;\n server.aof_fd = newfd;\n\n // 因为前面进行了 AOF 重写缓存追加，所以这里立即 fsync 一次\n if (server.aof_fsync == AOF_FSYNC_ALWAYS)\n aof_fsync(newfd);\n else if (server.aof_fsync == AOF_FSYNC_EVERYSEC)\n aof_background_fsync(newfd);\n\n // 强制引发 SELECT\n server.aof_selected_db = -1; /* Make sure SELECT is re-issued */\n\n // 更新 AOF 文件的大小\n aofUpdateCurrentSize();\n\n // 记录前一次重写时的大小\n server.aof_rewrite_base_size = server.aof_current_size;\n\n /* Clear regular AOF buffer since its contents was just written to\n * the new AOF from the background rewrite buffer. \n *\n * 清空 AOF 缓存，因为它的内容已经被写入过了，没用了\n */\n sdsfree(server.aof_buf);\n server.aof_buf = sdsempty();\n }\n\n server.aof_lastbgrewrite_status = REDIS_OK;\n\n redisLog(REDIS_NOTICE, \"Background AOF rewrite finished successfully\");\n\n /* Change state from WAIT_REWRITE to ON if needed \n *\n * 如果是第一次创建 AOF 文件，那么更新 AOF 状态\n */\n if (server.aof_state == REDIS_AOF_WAIT_REWRITE)\n server.aof_state = REDIS_AOF_ON;\n\n /* Asynchronously close the overwritten AOF. \n *\n * 异步关闭旧 AOF 文件\n */\n if (oldfd != -1) bioCreateBackgroundJob(REDIS_BIO_CLOSE_FILE,(void*)(long)oldfd,NULL,NULL);\n\n redisLog(REDIS_VERBOSE,\n \"Background AOF rewrite signal handler took %lldus\", ustime()-now);\n\n // BGREWRITEAOF 重写出错\n } else if (!bysignal && exitcode != 0) {\n server.aof_lastbgrewrite_status = REDIS_ERR;\n\n redisLog(REDIS_WARNING,\n \"Background AOF rewrite terminated with error\");\n\n // 未知错误\n } else {\n server.aof_lastbgrewrite_status = REDIS_ERR;\n\n redisLog(REDIS_WARNING,\n \"Background AOF rewrite terminated by signal %d\", bysignal);\n }\n\ncleanup:\n\n // 清空 AOF 缓冲区\n aofRewriteBufferReset();\n\n // 移除临时文件\n aofRemoveTempFile(server.aof_child_pid);\n\n // 重置默认属性\n server.aof_child_pid = -1;\n server.aof_rewrite_time_last = time(NULL)-server.aof_rewrite_time_start;\n server.aof_rewrite_time_start = -1;\n\n /* Schedule a new rewrite if we are waiting for it to switch the AOF ON. */\n if (server.aof_state == REDIS_AOF_WAIT_REWRITE)\n server.aof_rewrite_scheduled = 1;\n}\n" }, { "path": "src/asciilogo.h", "content": "/*\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\nchar *ascii_logo =\n\" _._ \\n\"\n\" _.-``__ ''-._ \\n\"\n\" _.-`` `. `_. ''-._ Redis %s (%s/%d) %s bit\\n\"\n\" .-`` .-```. ```\\\\/ _.,_ ''-._ \\n\"\n\" ( ' , .-` | `, ) Running in %s mode\\n\"\n\" |`-._`-...-` __...-.``-._|'` _.-'| Port: %d\\n\"\n\" | `-._ `._ / _.-' | PID: %ld\\n\"\n\" `-._ `-._ `-./ _.-' _.-' \\n\"\n\" |`-._`-._ `-.__.-' _.-'_.-'| \\n\"\n\" | `-._`-._ _.-'_.-' | http://redis.io \\n\"\n\" `-._ `-._`-.__.-'_.-' _.-' \\n\"\n\" |`-._`-._ `-.__.-' _.-'_.-'| \\n\"\n\" | `-._`-._ _.-'_.-' | \\n\"\n\" `-._ `-._`-.__.-'_.-' _.-' \\n\"\n\" `-._ `-.__.-' _.-' \\n\"\n\" `-._ _.-' \\n\"\n\" `-.__.-' \\n\\n\";\n" }, { "path": "src/bio.c", "content": "/* Background I/O service for Redis.\n *\n * Redis 的后台 I/O 服务\n *\n * This file implements operations that we need to perform in the background.\n *\n * bio 实现了将工作放在后台执行的功能。\n *\n * Currently there is only a single operation, that is a background close(2)\n * system call. This is needed as when the process is the last owner of a\n * reference to a file closing it means unlinking it, and the deletion of the\n * file is slow, blocking the server.\n *\n * 目前在后台执行的只有 close(2) 操作：\n * 因为当服务器是某个文件的最后一个拥有者时，\n * 关闭一个文件代表 unlinking 它，\n * 并且删除文件非常慢，会阻塞系统，\n * 所以我们将 close(2) 放到后台进行。\n *\n * (译注：现在不止 close(2) ，连 AOF 文件的 fsync 也是放到后台执行的）\n *\n * In the future we'll either continue implementing new things we need or\n * we'll switch to libeio. However there are probably long term uses for this\n * file as we may want to put here Redis specific background tasks (for instance\n * it is not impossible that we'll need a non blocking FLUSHDB/FLUSHALL\n * implementation).\n *\n * 这个后台服务将来可能会增加更多功能，或者切换到 libeio 上面去。\n * 不过我们可能会长期使用这个文件，以便支持一些 Redis 所特有的后台操作。\n * 比如说，将来我们可能需要一个非阻塞的 FLUSHDB 或者 FLUSHALL 也说不定。\n *\n * DESIGN\n * ------\n *\n * The design is trivial, we have a structure representing a job to perform\n * and a different thread and job queue for every job type.\n * Every thread wait for new jobs in its queue, and process every job\n * sequentially.\n *\n * 设计很简单：\n * 用一个结构表示要执行的工作，而每个类型的工作有一个队列和线程，\n * 每个线程都顺序地执行队列中的工作。\n *\n * Jobs of the same type are guaranteed to be processed from the least\n * recently inserted to the most recently inserted (older jobs processed\n * first).\n *\n * 同一类型的工作按 FIFO 的顺序执行。\n *\n * Currently there is no way for the creator of the job to be notified about\n * the completion of the operation, this will only be added when/if needed.\n *\n * 目前还没有办法在任务完成时通知执行者，在有需要的时候，会实现这个功能。\n *\n * ----------------------------------------------------------------------------\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n\n#include \"redis.h\"\n#include \"bio.h\"\n\n// 工作线程，斥互和条件变量\nstatic pthread_t bio_threads[REDIS_BIO_NUM_OPS];\nstatic pthread_mutex_t bio_mutex[REDIS_BIO_NUM_OPS];\nstatic pthread_cond_t bio_condvar[REDIS_BIO_NUM_OPS];\n\n// 存放工作的队列\nstatic list *bio_jobs[REDIS_BIO_NUM_OPS];\n\n/* The following array is used to hold the number of pending jobs for every\n * OP type. This allows us to export the bioPendingJobsOfType() API that is\n * useful when the main thread wants to perform some operation that may involve\n * objects shared with the background thread. The main thread will just wait\n * that there are no longer jobs of this type to be executed before performing\n * the sensible operation. This data is also useful for reporting. */\n// 记录每种类型 job 队列里有多少 job 等待执行\nstatic unsigned long long bio_pending[REDIS_BIO_NUM_OPS];\n\n/* This structure represents a background Job. It is only used locally to this\n * file as the API deos not expose the internals at all.\n *\n * 表示后台任务的数据结构\n *\n * 这个结构只由 API 使用，不会被暴露给外部。\n */\nstruct bio_job {\n\n // 任务创建时的时间\n time_t time; /* Time at which the job was created. */\n\n /* Job specific arguments pointers. If we need to pass more than three\n * arguments we can just pass a pointer to a structure or alike. \n *\n * 任务的参数。参数多于三个时，可以传递数组或者结构\n */\n void *arg1, *arg2, *arg3;\n};\n\nvoid *bioProcessBackgroundJobs(void *arg);\n\n/* Make sure we have enough stack to perform all the things we do in the\n * main thread. \n *\n * 子线程栈大小\n */\n#define REDIS_THREAD_STACK_SIZE (1024*1024*4)\n\n/* Initialize the background system, spawning the thread. \n *\n * 初始化后台任务系统，生成线程\n */\nvoid bioInit(void) {\n pthread_attr_t attr;\n pthread_t thread;\n size_t stacksize;\n int j;\n\n /* Initialization of state vars and objects \n *\n * 初始化 job 队列，以及线程状态\n */\n for (j = 0; j < REDIS_BIO_NUM_OPS; j++) {\n pthread_mutex_init(&bio_mutex[j],NULL);\n pthread_cond_init(&bio_condvar[j],NULL);\n bio_jobs[j] = listCreate();\n bio_pending[j] = 0;\n }\n\n /* Set the stack size as by default it may be small in some system \n *\n * 设置栈大小\n */\n pthread_attr_init(&attr);\n pthread_attr_getstacksize(&attr,&stacksize);\n if (!stacksize) stacksize = 1; /* The world is full of Solaris Fixes */\n while (stacksize < REDIS_THREAD_STACK_SIZE) stacksize *= 2;\n pthread_attr_setstacksize(&attr, stacksize);\n\n /* Ready to spawn our threads. We use the single argument the thread\n * function accepts in order to pass the job ID the thread is\n * responsible of. \n *\n * 创建线程\n */\n for (j = 0; j < REDIS_BIO_NUM_OPS; j++) {\n void *arg = (void*)(unsigned long) j;\n if (pthread_create(&thread,&attr,bioProcessBackgroundJobs,arg) != 0) {\n redisLog(REDIS_WARNING,\"Fatal: Can't initialize Background Jobs.\");\n exit(1);\n }\n bio_threads[j] = thread;\n }\n}\n\n/*\n * 创建后台任务\n */\nvoid bioCreateBackgroundJob(int type, void *arg1, void *arg2, void *arg3) {\n struct bio_job *job = zmalloc(sizeof(*job));\n\n job->time = time(NULL);\n job->arg1 = arg1;\n job->arg2 = arg2;\n job->arg3 = arg3;\n\n pthread_mutex_lock(&bio_mutex[type]);\n\n // 将新工作推入队列\n listAddNodeTail(bio_jobs[type],job);\n bio_pending[type]++;\n\n pthread_cond_signal(&bio_condvar[type]);\n\n pthread_mutex_unlock(&bio_mutex[type]);\n}\n\n/*\n * 处理后台任务\n */\nvoid *bioProcessBackgroundJobs(void *arg) {\n struct bio_job *job;\n unsigned long type = (unsigned long) arg;\n sigset_t sigset;\n\n /* Make the thread killable at any time, so that bioKillThreads()\n * can work reliably. */\n pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL);\n pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);\n\n pthread_mutex_lock(&bio_mutex[type]);\n /* Block SIGALRM so we are sure that only the main thread will\n * receive the watchdog signal. */\n sigemptyset(&sigset);\n sigaddset(&sigset, SIGALRM);\n if (pthread_sigmask(SIG_BLOCK, &sigset, NULL))\n redisLog(REDIS_WARNING,\n \"Warning: can't mask SIGALRM in bio.c thread: %s\", strerror(errno));\n\n while(1) {\n listNode *ln;\n\n /* The loop always starts with the lock hold. */\n if (listLength(bio_jobs[type]) == 0) {\n pthread_cond_wait(&bio_condvar[type],&bio_mutex[type]);\n continue;\n }\n\n /* Pop the job from the queue. \n *\n * 取出（但不删除）队列中的首个任务\n */\n ln = listFirst(bio_jobs[type]);\n job = ln->value;\n\n /* It is now possible to unlock the background system as we know have\n * a stand alone job structure to process.*/\n pthread_mutex_unlock(&bio_mutex[type]);\n\n /* Process the job accordingly to its type. */\n // 执行任务\n if (type == REDIS_BIO_CLOSE_FILE) {\n close((long)job->arg1);\n\n } else if (type == REDIS_BIO_AOF_FSYNC) {\n aof_fsync((long)job->arg1);\n\n } else {\n redisPanic(\"Wrong job type in bioProcessBackgroundJobs().\");\n }\n\n zfree(job);\n\n /* Lock again before reiterating the loop, if there are no longer\n * jobs to process we'll block again in pthread_cond_wait(). */\n pthread_mutex_lock(&bio_mutex[type]);\n // 将执行完成的任务从队列中删除，并减少任务计数器\n listDelNode(bio_jobs[type],ln);\n bio_pending[type]--;\n }\n}\n\n/* Return the number of pending jobs of the specified type. \n *\n * 返回等待中的 type 类型的工作的数量\n */\nunsigned long long bioPendingJobsOfType(int type) {\n unsigned long long val;\n\n pthread_mutex_lock(&bio_mutex[type]);\n val = bio_pending[type];\n pthread_mutex_unlock(&bio_mutex[type]);\n\n return val;\n}\n\n/* Kill the running bio threads in an unclean way. This function should be\n * used only when it's critical to stop the threads for some reason.\n *\n * 不进行清理，直接杀死进程，只在出现严重错误时使用\n *\n * Currently Redis does this only on crash (for instance on SIGSEGV) in order\n * to perform a fast memory check without other threads messing with memory. \n */\nvoid bioKillThreads(void) {\n int err, j;\n\n for (j = 0; j < REDIS_BIO_NUM_OPS; j++) {\n if (pthread_cancel(bio_threads[j]) == 0) {\n if ((err = pthread_join(bio_threads[j],NULL)) != 0) {\n redisLog(REDIS_WARNING,\n \"Bio thread for job type #%d can be joined: %s\",\n j, strerror(err));\n } else {\n redisLog(REDIS_WARNING,\n \"Bio thread for job type #%d terminated\",j);\n }\n }\n }\n}\n" }, { "path": "src/bio.h", "content": "/*\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n/* Exported API */\nvoid bioInit(void);\nvoid bioCreateBackgroundJob(int type, void *arg1, void *arg2, void *arg3);\nunsigned long long bioPendingJobsOfType(int type);\nvoid bioWaitPendingJobsLE(int type, unsigned long long num);\ntime_t bioOlderJobOfType(int type);\nvoid bioKillThreads(void);\n\n/* Background job opcodes */\n#define REDIS_BIO_CLOSE_FILE 0 /* Deferred close(2) syscall. */\n#define REDIS_BIO_AOF_FSYNC 1 /* Deferred AOF fsync. */\n#define REDIS_BIO_NUM_OPS 2\n" }, { "path": "src/bitops.c", "content": "/* Bit operations.\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"redis.h\"\n\n/* -----------------------------------------------------------------------------\n * Helpers and low level bit functions.\n * -------------------------------------------------------------------------- */\n\n/* This helper function used by GETBIT / SETBIT parses the bit offset argument\n * making sure an error is returned if it is negative or if it overflows\n * Redis 512 MB limit for the string value. */\n// 辅佐函数，被 GETBIT 、 SETBIT 所使用\n// 用于检查字符串的大小有否超过 512 MB\nstatic int getBitOffsetFromArgument(redisClient *c, robj *o, size_t *offset) {\n long long loffset;\n char *err = \"bit offset is not an integer or out of range\";\n\n if (getLongLongFromObjectOrReply(c,o,&loffset,err) != REDIS_OK)\n return REDIS_ERR;\n\n /* Limit offset to 512MB in bytes */\n if ((loffset < 0) || ((unsigned long long)loffset >> 3) >= (512*1024*1024))\n {\n addReplyError(c,err);\n return REDIS_ERR;\n }\n\n *offset = (size_t)loffset;\n return REDIS_OK;\n}\n\n/* Count number of bits set in the binary array pointed by 's' and long\n * 'count' bytes. The implementation of this function is required to\n * work with a input string length up to 512 MB. */\n// 计算长度为 count 的二进制数组指针 s 被设置为 1 的位数量\n// 这个函数只能在最大为 512 MB 的字符串上使用\nsize_t redisPopcount(void *s, long count) {\n size_t bits = 0;\n unsigned char *p = s;\n uint32_t *p4;\n // 通过查表来计算，对于 1 字节所能表示的值来说\n // 这些值的二进制表示所带有的 1 的数量\n // 比如整数 3 的二进制表示 0011 ，带有两个 1\n // 正好是查表 bitsinbyte[3] == 2\n static const unsigned char bitsinbyte[256] = {0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,4,5,5,6,5,6,6,7,5,6,6,7,6,7,7,8};\n\n /* Count initial bytes not aligned to 32 bit. */\n while((unsigned long)p & 3 && count) {\n bits += bitsinbyte[*p++];\n count--;\n }\n\n /* Count bits 16 bytes at a time */\n // 每次统计 16 字节\n // 关于这里所使用的优化算法，可以参考：\n // http://yesteapea.wordpress.com/2013/03/03/counting-the-number-of-set-bits-in-an-integer/\n p4 = (uint32_t*)p;\n while(count>=16) {\n uint32_t aux1, aux2, aux3, aux4;\n\n aux1 = *p4++;\n aux2 = *p4++;\n aux3 = *p4++;\n aux4 = *p4++;\n count -= 16;\n\n aux1 = aux1 - ((aux1 >> 1) & 0x55555555);\n aux1 = (aux1 & 0x33333333) + ((aux1 >> 2) & 0x33333333);\n aux2 = aux2 - ((aux2 >> 1) & 0x55555555);\n aux2 = (aux2 & 0x33333333) + ((aux2 >> 2) & 0x33333333);\n aux3 = aux3 - ((aux3 >> 1) & 0x55555555);\n aux3 = (aux3 & 0x33333333) + ((aux3 >> 2) & 0x33333333);\n aux4 = aux4 - ((aux4 >> 1) & 0x55555555);\n aux4 = (aux4 & 0x33333333) + ((aux4 >> 2) & 0x33333333);\n bits += ((((aux1 + (aux1 >> 4)) & 0x0F0F0F0F) * 0x01010101) >> 24) +\n ((((aux2 + (aux2 >> 4)) & 0x0F0F0F0F) * 0x01010101) >> 24) +\n ((((aux3 + (aux3 >> 4)) & 0x0F0F0F0F) * 0x01010101) >> 24) +\n ((((aux4 + (aux4 >> 4)) & 0x0F0F0F0F) * 0x01010101) >> 24);\n }\n\n /* Count the remaining bytes. */\n // 不足 16 字节的，剩下的每个字节通过查表来完成\n p = (unsigned char*)p4;\n while(count--) bits += bitsinbyte[*p++];\n return bits;\n}\n\n/* Return the position of the first bit set to one (if 'bit' is 1) or\n * zero (if 'bit' is 0) in the bitmap starting at 's' and long 'count' bytes.\n *\n * The function is guaranteed to return a value >= 0 if 'bit' is 0 since if\n * no zero bit is found, it returns count*8 assuming the string is zero\n * padded on the right. However if 'bit' is 1 it is possible that there is\n * not a single set bit in the bitmap. In this special case -1 is returned. */\nlong redisBitpos(void *s, long count, int bit) {\n unsigned long *l;\n unsigned char *c;\n unsigned long skipval, word = 0, one;\n long pos = 0; /* Position of bit, to return to the caller. */\n int j;\n\n /* Process whole words first, seeking for first word that is not\n * all ones or all zeros respectively if we are lookig for zeros\n * or ones. This is much faster with large strings having contiguous\n * blocks of 1 or 0 bits compared to the vanilla bit per bit processing.\n *\n * Note that if we start from an address that is not aligned\n * to sizeof(unsigned long) we consume it byte by byte until it is\n * aligned. */\n\n /* Skip initial bits not aligned to sizeof(unsigned long) byte by byte. */\n skipval = bit ? 0 : UCHAR_MAX;\n c = (unsigned char*) s;\n while((unsigned long)c & (sizeof(*l)-1) && count) {\n if (*c != skipval) break;\n c++;\n count--;\n pos += 8;\n }\n\n /* Skip bits with full word step. */\n skipval = bit ? 0 : ULONG_MAX;\n l = (unsigned long*) c;\n while (count >= sizeof(*l)) {\n if (*l != skipval) break;\n l++;\n count -= sizeof(*l);\n pos += sizeof(*l)*8;\n }\n\n /* Load bytes into \"word\" considering the first byte as the most significant\n * (we basically consider it as written in big endian, since we consider the\n * string as a set of bits from left to right, with the first bit at position\n * zero.\n *\n * Note that the loading is designed to work even when the bytes left\n * (count) are less than a full word. We pad it with zero on the right. */\n c = (unsigned char*)l;\n for (j = 0; j < sizeof(*l); j++) {\n word <<= 8;\n if (count) {\n word |= *c;\n c++;\n count--;\n }\n }\n\n /* Special case:\n * If bits in the string are all zero and we are looking for one,\n * return -1 to signal that there is not a single \"1\" in the whole\n * string. This can't happen when we are looking for \"0\" as we assume\n * that the right of the string is zero padded. */\n if (bit == 1 && word == 0) return -1;\n\n /* Last word left, scan bit by bit. The first thing we need is to\n * have a single \"1\" set in the most significant position in an\n * unsigned long. We don't know the size of the long so we use a\n * simple trick. */\n one = ULONG_MAX; /* All bits set to 1.*/\n one >>= 1; /* All bits set to 1 but the MSB. */\n one = ~one; /* All bits set to 0 but the MSB. */\n\n while(one) {\n if (((one & word) != 0) == bit) return pos;\n pos++;\n one >>= 1;\n }\n\n /* If we reached this point, there is a bug in the algorithm, since\n * the case of no match is handled as a special case before. */\n redisPanic(\"End of redisBitpos() reached.\");\n return 0; /* Just to avoid warnings. */\n}\n\n/* -----------------------------------------------------------------------------\n * Bits related string commands: GETBIT, SETBIT, BITCOUNT, BITOP.\n * -------------------------------------------------------------------------- */\n\n#define BITOP_AND 0\n#define BITOP_OR 1\n#define BITOP_XOR 2\n#define BITOP_NOT 3\n\n/* SETBIT key offset bitvalue */\nvoid setbitCommand(redisClient *c) {\n robj *o;\n char *err = \"bit is not an integer or out of range\";\n size_t bitoffset;\n int byte, bit;\n int byteval, bitval;\n long on;\n\n // 获取 offset 参数\n if (getBitOffsetFromArgument(c,c->argv[2],&bitoffset) != REDIS_OK)\n return;\n\n // 获取 value 参数\n if (getLongFromObjectOrReply(c,c->argv[3],&on,err) != REDIS_OK)\n return;\n\n /* Bits can only be set or cleared... */\n // value 参数的值只能是 0 或者 1 ，否则返回错误\n if (on & ~1) {\n addReplyError(c,err);\n return;\n }\n\n // 查找字符串对象\n o = lookupKeyWrite(c->db,c->argv[1]);\n if (o == NULL) {\n\n // 对象不存在，创建一个空字符串对象\n o = createObject(REDIS_STRING,sdsempty());\n\n // 并添加到数据库\n dbAdd(c->db,c->argv[1],o);\n\n } else {\n\n // 对象存在，检查类型是否字符串\n if (checkType(c,o,REDIS_STRING)) return;\n\n o = dbUnshareStringValue(c->db,c->argv[1],o);\n }\n\n /* Grow sds value to the right length if necessary */\n // 计算容纳 offset 参数所指定的偏移量所需的字节数\n // 如果 o 对象的字节不够长的话，就扩展它\n // 长度的计算公式是 bitoffset >> 3 + 1\n // 比如 30 >> 3 + 1 = 4 ，也即是为了设置 offset 30 ，\n // 我们需要创建一个 4 字节（32 位长的 SDS）\n byte = bitoffset >> 3;\n o->ptr = sdsgrowzero(o->ptr,byte+1);\n\n /* Get current values */\n // 将指针定位到要设置的位所在的字节上\n byteval = ((uint8_t*)o->ptr)[byte];\n // 定位到要设置的位上面\n bit = 7 - (bitoffset & 0x7);\n // 记录位现在的值\n bitval = byteval & (1 << bit);\n\n /* Update byte with new bit value and return original value */\n // 更新字节中的位，设置它的值为 on 参数的值\n byteval &= ~(1 << bit);\n byteval |= ((on & 0x1) << bit);\n ((uint8_t*)o->ptr)[byte] = byteval;\n\n // 发送数据库修改通知\n signalModifiedKey(c->db,c->argv[1]);\n notifyKeyspaceEvent(REDIS_NOTIFY_STRING,\"setbit\",c->argv[1],c->db->id);\n server.dirty++;\n\n // 向客户端返回位原来的值\n addReply(c, bitval ? shared.cone : shared.czero);\n}\n\n/* GETBIT key offset */\nvoid getbitCommand(redisClient *c) {\n robj *o;\n char llbuf[32];\n size_t bitoffset;\n size_t byte, bit;\n size_t bitval = 0;\n\n // 读取 offset 参数\n if (getBitOffsetFromArgument(c,c->argv[2],&bitoffset) != REDIS_OK)\n return;\n\n // 查找对象，并进行类型检查\n if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||\n checkType(c,o,REDIS_STRING)) return;\n\n // 计算出 offset 所指定的位所在的字节\n byte = bitoffset >> 3;\n // 计算出位所在的位置\n bit = 7 - (bitoffset & 0x7);\n\n // 取出位\n if (sdsEncodedObject(o)) {\n // 字符串编码，直接取值\n if (byte < sdslen(o->ptr))\n bitval = ((uint8_t*)o->ptr)[byte] & (1 << bit);\n } else {\n // 整数编码，先转换成字符串，再取值\n if (byte < (size_t)ll2string(llbuf,sizeof(llbuf),(long)o->ptr))\n bitval = llbuf[byte] & (1 << bit);\n }\n\n // 返回位\n addReply(c, bitval ? shared.cone : shared.czero);\n}\n\n/* BITOP op_name target_key src_key1 src_key2 src_key3 ... src_keyN */\nvoid bitopCommand(redisClient *c) {\n char *opname = c->argv[1]->ptr;\n robj *o, *targetkey = c->argv[2];\n long op, j, numkeys;\n robj **objects; /* Array of source objects. */\n unsigned char **src; /* Array of source strings pointers. */\n long *len, maxlen = 0; /* Array of length of src strings, and max len. */\n long minlen = 0; /* Min len among the input keys. */\n unsigned char *res = NULL; /* Resulting string. */\n\n /* Parse the operation name. */\n // 读入 op 参数，确定要执行的操作\n if ((opname[0] == 'a' || opname[0] == 'A') && !strcasecmp(opname,\"and\"))\n op = BITOP_AND;\n else if((opname[0] == 'o' || opname[0] == 'O') && !strcasecmp(opname,\"or\"))\n op = BITOP_OR;\n else if((opname[0] == 'x' || opname[0] == 'X') && !strcasecmp(opname,\"xor\"))\n op = BITOP_XOR;\n else if((opname[0] == 'n' || opname[0] == 'N') && !strcasecmp(opname,\"not\"))\n op = BITOP_NOT;\n else {\n addReply(c,shared.syntaxerr);\n return;\n }\n\n /* Sanity check: NOT accepts only a single key argument. */\n // NOT 操作只能接受单个 key 输入\n if (op == BITOP_NOT && c->argc != 4) {\n addReplyError(c,\"BITOP NOT must be called with a single source key.\");\n return;\n }\n\n /* Lookup keys, and store pointers to the string objects into an array. */\n // 查找输入键，并将它们放入一个数组里面\n numkeys = c->argc - 3;\n // 字符串数组，保存 sds 值\n src = zmalloc(sizeof(unsigned char*) * numkeys);\n // 长度数组，保存 sds 的长度\n len = zmalloc(sizeof(long) * numkeys);\n // 对象数组，保存字符串对象\n objects = zmalloc(sizeof(robj*) * numkeys);\n for (j = 0; j < numkeys; j++) {\n\n // 查找对象\n o = lookupKeyRead(c->db,c->argv[j+3]);\n\n /* Handle non-existing keys as empty strings. */\n // 不存在的键被视为空字符串\n if (o == NULL) {\n objects[j] = NULL;\n src[j] = NULL;\n len[j] = 0;\n minlen = 0;\n continue;\n }\n\n /* Return an error if one of the keys is not a string. */\n // 键不是字符串类型，返回错误，放弃执行操作\n if (checkType(c,o,REDIS_STRING)) {\n for (j = j-1; j >= 0; j--) {\n if (objects[j])\n decrRefCount(objects[j]);\n }\n zfree(src);\n zfree(len);\n zfree(objects);\n return;\n }\n\n // 记录对象\n objects[j] = getDecodedObject(o);\n // 记录 sds\n src[j] = objects[j]->ptr;\n // 记录 sds 长度\n len[j] = sdslen(objects[j]->ptr);\n \n // 记录目前最长 sds 的长度\n if (len[j] > maxlen) maxlen = len[j];\n\n // 记录目前最短 sds 的长度\n if (j == 0 || len[j] < minlen) minlen = len[j];\n }\n\n /* Compute the bit operation, if at least one string is not empty. */\n // 如果有至少一个非空字符串，那么执行计算\n if (maxlen) {\n\n // 根据最大长度，创建一个 sds ，该 sds 的所有位都被设置为 0\n res = (unsigned char*) sdsnewlen(NULL,maxlen);\n\n unsigned char output, byte;\n long i;\n\n /* Fast path: as far as we have data for all the input bitmaps we\n * can take a fast path that performs much better than the\n * vanilla algorithm. */\n // 在键的数量比较少时，进行优化\n j = 0;\n if (minlen && numkeys <= 16) {\n unsigned long *lp[16];\n unsigned long *lres = (unsigned long*) res;\n\n /* Note: sds pointer is always aligned to 8 byte boundary. */\n memcpy(lp,src,sizeof(unsigned long*)*numkeys);\n memcpy(res,src[0],minlen);\n\n /* Different branches per different operations for speed (sorry). */\n // 当要处理的位大于等于 32 位时\n // 每次载入 4*8 = 32 个位，然后对这些位进行计算，利用缓存，进行加速\n if (op == BITOP_AND) {\n while(minlen >= sizeof(unsigned long)*4) {\n for (i = 1; i < numkeys; i++) {\n lres[0] &= lp[i][0];\n lres[1] &= lp[i][1];\n lres[2] &= lp[i][2];\n lres[3] &= lp[i][3];\n lp[i]+=4;\n }\n lres+=4;\n j += sizeof(unsigned long)*4;\n minlen -= sizeof(unsigned long)*4;\n }\n } else if (op == BITOP_OR) {\n while(minlen >= sizeof(unsigned long)*4) {\n for (i = 1; i < numkeys; i++) {\n lres[0] |= lp[i][0];\n lres[1] |= lp[i][1];\n lres[2] |= lp[i][2];\n lres[3] |= lp[i][3];\n lp[i]+=4;\n }\n lres+=4;\n j += sizeof(unsigned long)*4;\n minlen -= sizeof(unsigned long)*4;\n }\n } else if (op == BITOP_XOR) {\n while(minlen >= sizeof(unsigned long)*4) {\n for (i = 1; i < numkeys; i++) {\n lres[0] ^= lp[i][0];\n lres[1] ^= lp[i][1];\n lres[2] ^= lp[i][2];\n lres[3] ^= lp[i][3];\n lp[i]+=4;\n }\n lres+=4;\n j += sizeof(unsigned long)*4;\n minlen -= sizeof(unsigned long)*4;\n }\n } else if (op == BITOP_NOT) {\n while(minlen >= sizeof(unsigned long)*4) {\n lres[0] = ~lres[0];\n lres[1] = ~lres[1];\n lres[2] = ~lres[2];\n lres[3] = ~lres[3];\n lres+=4;\n j += sizeof(unsigned long)*4;\n minlen -= sizeof(unsigned long)*4;\n }\n }\n }\n\n /* j is set to the next byte to process by the previous loop. */\n // 以正常方式执行位运算\n for (; j < maxlen; j++) {\n output = (len[0] <= j) ? 0 : src[0][j];\n if (op == BITOP_NOT) output = ~output;\n // 遍历所有输入键，对所有输入的 scr[i][j] 字节进行运算\n for (i = 1; i < numkeys; i++) {\n // 如果数组的长度不足，那么相应的字节被假设为 0\n byte = (len[i] <= j) ? 0 : src[i][j];\n switch(op) {\n case BITOP_AND: output &= byte; break;\n case BITOP_OR: output |= byte; break;\n case BITOP_XOR: output ^= byte; break;\n }\n }\n // 保存输出\n res[j] = output;\n }\n }\n\n // 释放资源\n for (j = 0; j < numkeys; j++) {\n if (objects[j])\n decrRefCount(objects[j]);\n }\n zfree(src);\n zfree(len);\n zfree(objects);\n\n /* Store the computed value into the target key */\n if (maxlen) {\n // 保存结果到指定键\n o = createObject(REDIS_STRING,res);\n setKey(c->db,targetkey,o);\n notifyKeyspaceEvent(REDIS_NOTIFY_STRING,\"set\",targetkey,c->db->id);\n decrRefCount(o);\n } else if (dbDelete(c->db,targetkey)) {\n // 输入为空，没有产生结果，仅仅删除指定键\n signalModifiedKey(c->db,targetkey);\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\"del\",targetkey,c->db->id);\n }\n server.dirty++;\n addReplyLongLong(c,maxlen); /* Return the output string length in bytes. */\n}\n\n/* BITCOUNT key [start end] */\nvoid bitcountCommand(redisClient *c) {\n robj *o;\n long start, end, strlen;\n unsigned char *p;\n char llbuf[32];\n\n /* Lookup, check for type, and return 0 for non existing keys. */\n // 查找 key\n if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL ||\n checkType(c,o,REDIS_STRING)) return;\n\n /* Set the 'p' pointer to the string, that can be just a stack allocated\n * array if our string was integer encoded. */\n // 检查对象的编码，并在有需要时转换值的类型\n if (o->encoding == REDIS_ENCODING_INT) {\n p = (unsigned char*) llbuf;\n strlen = ll2string(llbuf,sizeof(llbuf),(long)o->ptr);\n } else {\n p = (unsigned char*) o->ptr;\n strlen = sdslen(o->ptr);\n }\n\n /* Parse start/end range if any. */\n // 如果给定了 start 和 end 参数，那么读入它们\n if (c->argc == 4) {\n if (getLongFromObjectOrReply(c,c->argv[2],&start,NULL) != REDIS_OK)\n return;\n if (getLongFromObjectOrReply(c,c->argv[3],&end,NULL) != REDIS_OK)\n return;\n /* Convert negative indexes */\n if (start < 0) start = strlen+start;\n if (end < 0) end = strlen+end;\n if (start < 0) start = 0;\n if (end < 0) end = 0;\n if (end >= strlen) end = strlen-1;\n } else if (c->argc == 2) {\n /* The whole string. */\n start = 0;\n end = strlen-1;\n } else {\n /* Syntax error. */\n addReply(c,shared.syntaxerr);\n return;\n }\n\n /* Precondition: end >= 0 && end < strlen, so the only condition where\n * zero can be returned is: start > end. */\n if (start > end) {\n // 超出范围的 case ，略过\n addReply(c,shared.czero);\n } else {\n long bytes = end-start+1;\n\n // 遍历数组，统计值为 1 的位的数量\n addReplyLongLong(c,redisPopcount(p+start,bytes));\n }\n}\n\n/* BITPOS key bit [start [end]] */\nvoid bitposCommand(redisClient *c) {\n robj *o;\n long bit, start, end, strlen;\n unsigned char *p;\n char llbuf[32];\n int end_given = 0;\n\n /* Parse the bit argument to understand what we are looking for, set\n * or clear bits. */\n if (getLongFromObjectOrReply(c,c->argv[2],&bit,NULL) != REDIS_OK)\n return;\n if (bit != 0 && bit != 1) {\n addReplyError(c, \"The bit argument must be 1 or 0.\");\n return;\n }\n\n /* If the key does not exist, from our point of view it is an infinite\n * array of 0 bits. If the user is looking for the fist clear bit return 0,\n * If the user is looking for the first set bit, return -1. */\n if ((o = lookupKeyRead(c->db,c->argv[1])) == NULL) {\n addReplyLongLong(c, bit ? -1 : 0);\n return;\n }\n if (checkType(c,o,REDIS_STRING)) return;\n\n /* Set the 'p' pointer to the string, that can be just a stack allocated\n * array if our string was integer encoded. */\n if (o->encoding == REDIS_ENCODING_INT) {\n p = (unsigned char*) llbuf;\n strlen = ll2string(llbuf,sizeof(llbuf),(long)o->ptr);\n } else {\n p = (unsigned char*) o->ptr;\n strlen = sdslen(o->ptr);\n }\n\n /* Parse start/end range if any. */\n if (c->argc == 4 || c->argc == 5) {\n if (getLongFromObjectOrReply(c,c->argv[3],&start,NULL) != REDIS_OK)\n return;\n if (c->argc == 5) {\n if (getLongFromObjectOrReply(c,c->argv[4],&end,NULL) != REDIS_OK)\n return;\n end_given = 1;\n } else {\n end = strlen-1;\n }\n /* Convert negative indexes */\n if (start < 0) start = strlen+start;\n if (end < 0) end = strlen+end;\n if (start < 0) start = 0;\n if (end < 0) end = 0;\n if (end >= strlen) end = strlen-1;\n } else if (c->argc == 3) {\n /* The whole string. */\n start = 0;\n end = strlen-1;\n } else {\n /* Syntax error. */\n addReply(c,shared.syntaxerr);\n return;\n }\n\n /* For empty ranges (start > end) we return -1 as an empty range does\n * not contain a 0 nor a 1. */\n if (start > end) {\n addReplyLongLong(c, -1);\n } else {\n long bytes = end-start+1;\n long pos = redisBitpos(p+start,bytes,bit);\n\n /* If we are looking for clear bits, and the user specified an exact\n * range with start-end, we can't consider the right of the range as\n * zero padded (as we do when no explicit end is given).\n *\n * So if redisBitpos() returns the first bit outside the range,\n * we return -1 to the caller, to mean, in the specified range there\n * is not a single \"0\" bit. */\n if (end_given && bit == 0 && pos == bytes*8) {\n addReplyLongLong(c,-1);\n return;\n }\n if (pos != -1) pos += start*8; /* Adjust for the bytes we skipped. */\n addReplyLongLong(c,pos);\n }\n}\n" }, { "path": "src/blocked.c", "content": "/* blocked.c - generic support for blocking operations like BLPOP & WAIT.\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n *\n * ---------------------------------------------------------------------------\n *\n * API:\n *\n * getTimeoutFromObjectOrReply() is just an utility function to parse a\n * timeout argument since blocking operations usually require a timeout.\n *\n * blockClient() set the REDIS_BLOCKED flag in the client, and set the\n * specified block type 'btype' filed to one of REDIS_BLOCKED_* macros.\n *\n * unblockClient() unblocks the client doing the following:\n * 1) It calls the btype-specific function to cleanup the state.\n * 2) It unblocks the client by unsetting the REDIS_BLOCKED flag.\n * 3) It puts the client into a list of just unblocked clients that are\n * processed ASAP in the beforeSleep() event loop callback, so that\n * if there is some query buffer to process, we do it. This is also\n * required because otherwise there is no 'readable' event fired, we\n * already read the pending commands. We also set the REDIS_UNBLOCKED\n * flag to remember the client is in the unblocked_clients list.\n *\n * processUnblockedClients() is called inside the beforeSleep() function\n * to process the query buffer from unblocked clients and remove the clients\n * from the blocked_clients queue.\n *\n * replyToBlockedClientTimedOut() is called by the cron function when\n * a client blocked reaches the specified timeout (if the timeout is set\n * to 0, no timeout is processed).\n * It usually just needs to send a reply to the client.\n *\n * When implementing a new type of blocking opeation, the implementation\n * should modify unblockClient() and replyToBlockedClientTimedOut() in order\n * to handle the btype-specific behavior of this two functions.\n */\n\n#include \"redis.h\"\n\n/* Get a timeout value from an object and store it into 'timeout'.\n * The final timeout is always stored as milliseconds as a time where the\n * timeout will expire, however the parsing is performed according to\n * the 'unit' that can be seconds or milliseconds.\n *\n * Note that if the timeout is zero (usually from the point of view of\n * commands API this means no timeout) the value stored into 'timeout'\n * is zero. */\n// 根据输入参数，取出最大等待时间\nint getTimeoutFromObjectOrReply(redisClient *c, robj *object, mstime_t *timeout, int unit) {\n long long tval;\n\n if (getLongLongFromObjectOrReply(c,object,&tval,\n \"timeout is not an integer or out of range\") != REDIS_OK)\n return REDIS_ERR;\n\n if (tval < 0) {\n addReplyError(c,\"timeout is negative\");\n return REDIS_ERR;\n }\n\n if (tval > 0) {\n if (unit == UNIT_SECONDS) tval *= 1000;\n tval += mstime();\n }\n *timeout = tval;\n\n return REDIS_OK;\n}\n\n/* Block a client for the specific operation type. Once the REDIS_BLOCKED\n * flag is set client query buffer is not longer processed, but accumulated,\n * and will be processed when the client is unblocked. */\n// 对给定的客户端进行阻塞\nvoid blockClient(redisClient *c, int btype) {\n c->flags |= REDIS_BLOCKED;\n c->btype = btype;\n server.bpop_blocked_clients++;\n}\n\n/* This function is called in the beforeSleep() function of the event loop\n * in order to process the pending input buffer of clients that were\n * unblocked after a blocking operation. */\n// 取消所有在 unblocked_clients 链表中的客户端的阻塞状态\nvoid processUnblockedClients(void) {\n listNode *ln;\n redisClient *c;\n\n while (listLength(server.unblocked_clients)) {\n ln = listFirst(server.unblocked_clients);\n redisAssert(ln != NULL);\n c = ln->value;\n listDelNode(server.unblocked_clients,ln);\n c->flags &= ~REDIS_UNBLOCKED;\n c->btype = REDIS_BLOCKED_NONE;\n\n /* Process remaining data in the input buffer. */\n if (c->querybuf && sdslen(c->querybuf) > 0) {\n server.current_client = c;\n processInputBuffer(c);\n server.current_client = NULL;\n }\n }\n}\n\n/* Unblock a client calling the right function depending on the kind\n * of operation the client is blocking for. */\n// 取消给定的客户端的阻塞状态\nvoid unblockClient(redisClient *c) {\n if (c->btype == REDIS_BLOCKED_LIST) {\n unblockClientWaitingData(c);\n } else if (c->btype == REDIS_BLOCKED_WAIT) {\n unblockClientWaitingReplicas(c);\n } else {\n redisPanic(\"Unknown btype in unblockClient().\");\n }\n /* Clear the flags, and put the client in the unblocked list so that\n * we'll process new commands in its query buffer ASAP. */\n c->flags &= ~REDIS_BLOCKED;\n c->flags |= REDIS_UNBLOCKED;\n c->btype = REDIS_BLOCKED_NONE;\n server.bpop_blocked_clients--;\n listAddNodeTail(server.unblocked_clients,c);\n}\n\n/* This function gets called when a blocked client timed out in order to\n * send it a reply of some kind. */\n// 等待超时，向被阻塞的客户端返回通知\nvoid replyToBlockedClientTimedOut(redisClient *c) {\n if (c->btype == REDIS_BLOCKED_LIST) {\n addReply(c,shared.nullmultibulk);\n } else if (c->btype == REDIS_BLOCKED_WAIT) {\n addReplyLongLong(c,replicationCountAcksByOffset(c->bpop.reploffset));\n } else {\n redisPanic(\"Unknown btype in replyToBlockedClientTimedOut().\");\n }\n}\n\n" }, { "path": "src/cluster.c", "content": "/* Redis Cluster implementation.\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"redis.h\"\n#include \"cluster.h\"\n#include \"endianconv.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n/* A global reference to myself is handy to make code more clear.\n * Myself always points to server.cluster->myself, that is, the clusterNode\n * that represents this node. */\n// 为了方便起见，维持一个 myself 全局变量，让它总是指向 cluster->myself 。\nclusterNode *myself = NULL;\n\nclusterNode *createClusterNode(char *nodename, int flags);\nint clusterAddNode(clusterNode *node);\nvoid clusterAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);\nvoid clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask);\nvoid clusterSendPing(clusterLink *link, int type);\nvoid clusterSendFail(char *nodename);\nvoid clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request);\nvoid clusterUpdateState(void);\nint clusterNodeGetSlotBit(clusterNode *n, int slot);\nsds clusterGenNodesDescription(int filter);\nclusterNode *clusterLookupNode(char *name);\nint clusterNodeAddSlave(clusterNode *master, clusterNode *slave);\nint clusterAddSlot(clusterNode *n, int slot);\nint clusterDelSlot(int slot);\nint clusterDelNodeSlots(clusterNode *node);\nint clusterNodeSetSlotBit(clusterNode *n, int slot);\nvoid clusterSetMaster(clusterNode *n);\nvoid clusterHandleSlaveFailover(void);\nvoid clusterHandleSlaveMigration(int max_slaves);\nint bitmapTestBit(unsigned char *bitmap, int pos);\nvoid clusterDoBeforeSleep(int flags);\nvoid clusterSendUpdate(clusterLink *link, clusterNode *node);\nvoid resetManualFailover(void);\nvoid clusterCloseAllSlots(void);\nvoid clusterSetNodeAsMaster(clusterNode *n);\nvoid clusterDelNode(clusterNode *delnode);\n\n/* -----------------------------------------------------------------------------\n * Initialization\n * -------------------------------------------------------------------------- */\n\n/* Return the greatest configEpoch found in the cluster. */\nuint64_t clusterGetMaxEpoch(void) {\n uint64_t max = 0;\n dictIterator *di;\n dictEntry *de;\n\n // 选出节点中的最大纪元\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n if (node->configEpoch > max) max = node->configEpoch;\n }\n dictReleaseIterator(di);\n if (max < server.cluster->currentEpoch) max = server.cluster->currentEpoch;\n return max;\n}\n\n// 载入集群配置\n/* Load the cluster config from 'filename'.\n *\n * If the file does not exist or is zero-length (this may happen because\n * when we lock the nodes.conf file, we create a zero-length one for the\n * sake of locking if it does not already exist), REDIS_ERR is returned.\n * If the configuration was loaded from the file, REDIS_OK is returned. */\nint clusterLoadConfig(char *filename) {\n FILE *fp = fopen(filename,\"r\");\n struct stat sb;\n char *line;\n int maxline, j;\n\n if (fp == NULL) {\n if (errno == ENOENT) {\n return REDIS_ERR;\n } else {\n redisLog(REDIS_WARNING,\n \"Loading the cluster node config from %s: %s\",\n filename, strerror(errno));\n exit(1);\n }\n }\n\n /* Check if the file is zero-length: if so return REDIS_ERR to signal\n * we have to write the config. */\n if (fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) {\n fclose(fp);\n return REDIS_ERR;\n }\n\n /* Parse the file. Note that single liens of the cluster config file can\n * be really long as they include all the hash slots of the node.\n * 集群配置文件中的行可能会非常长，\n * 因为它会在行里面记录所有哈希槽的节点。\n *\n * This means in the worst possible case, half of the Redis slots will be\n * present in a single line, possibly in importing or migrating state, so\n * together with the node ID of the sender/receiver.\n *\n * 在最坏情况下，一个行可能保存了半数的哈希槽数据，\n * 并且可能带有导入或导出状态，以及发送者和接受者的 ID 。\n *\n * To simplify we allocate 1024+REDIS_CLUSTER_SLOTS*128 bytes per line. \n *\n * 为了简单起见，我们为每行分配 1024+REDIS_CLUSTER_SLOTS*128 字节的空间\n */\n maxline = 1024+REDIS_CLUSTER_SLOTS*128;\n line = zmalloc(maxline);\n while(fgets(line,maxline,fp) != NULL) {\n int argc;\n sds *argv;\n clusterNode *n, *master;\n char *p, *s;\n\n /* Skip blank lines, they can be created either by users manually\n * editing nodes.conf or by the config writing process if stopped\n * before the truncate() call. */\n if (line[0] == '\\n') continue;\n\n /* Split the line into arguments for processing. */\n argv = sdssplitargs(line,&argc);\n if (argv == NULL) goto fmterr;\n\n /* Handle the special \"vars\" line. Don't pretend it is the last\n * line even if it actually is when generated by Redis. */\n if (strcasecmp(argv[0],\"vars\") == 0) {\n for (j = 1; j < argc; j += 2) {\n if (strcasecmp(argv[j],\"currentEpoch\") == 0) {\n server.cluster->currentEpoch =\n strtoull(argv[j+1],NULL,10);\n } else if (strcasecmp(argv[j],\"lastVoteEpoch\") == 0) {\n server.cluster->lastVoteEpoch =\n strtoull(argv[j+1],NULL,10);\n } else {\n redisLog(REDIS_WARNING,\n \"Skipping unknown cluster config variable '%s'\",\n argv[j]);\n }\n }\n continue;\n }\n\n /* Create this node if it does not exist */\n // 检查节点是否已经存在\n n = clusterLookupNode(argv[0]);\n if (!n) {\n // 未存在则创建这个节点\n n = createClusterNode(argv[0],0);\n clusterAddNode(n);\n }\n /* Address and port */\n // 设置节点的 ip 和 port\n if ((p = strchr(argv[1],':')) == NULL) goto fmterr;\n *p = '\\0';\n memcpy(n->ip,argv[1],strlen(argv[1])+1);\n n->port = atoi(p+1);\n\n /* Parse flags */\n // 分析节点的 flag\n p = s = argv[2];\n while(p) {\n p = strchr(s,',');\n if (p) *p = '\\0';\n // 这是节点本身\n if (!strcasecmp(s,\"myself\")) {\n redisAssert(server.cluster->myself == NULL);\n myself = server.cluster->myself = n;\n n->flags |= REDIS_NODE_MYSELF;\n // 这是一个主节点\n } else if (!strcasecmp(s,\"master\")) {\n n->flags |= REDIS_NODE_MASTER;\n // 这是一个从节点\n } else if (!strcasecmp(s,\"slave\")) {\n n->flags |= REDIS_NODE_SLAVE;\n // 这是一个疑似下线节点\n } else if (!strcasecmp(s,\"fail?\")) {\n n->flags |= REDIS_NODE_PFAIL;\n // 这是一个已下线节点\n } else if (!strcasecmp(s,\"fail\")) {\n n->flags |= REDIS_NODE_FAIL;\n n->fail_time = mstime();\n // 等待向节点发送 PING\n } else if (!strcasecmp(s,\"handshake\")) {\n n->flags |= REDIS_NODE_HANDSHAKE;\n // 尚未获得这个节点的地址\n } else if (!strcasecmp(s,\"noaddr\")) {\n n->flags |= REDIS_NODE_NOADDR;\n // 无 flag\n } else if (!strcasecmp(s,\"noflags\")) {\n /* nothing to do */\n } else {\n redisPanic(\"Unknown flag in redis cluster config file\");\n }\n if (p) s = p+1;\n }\n\n /* Get master if any. Set the master and populate master's\n * slave list. */\n // 如果有主节点的话，那么设置主节点\n if (argv[3][0] != '-') {\n master = clusterLookupNode(argv[3]);\n // 如果主节点不存在，那么添加它\n if (!master) {\n master = createClusterNode(argv[3],0);\n clusterAddNode(master);\n }\n // 设置主节点\n n->slaveof = master;\n // 将节点 n 加入到主节点 master 的从节点名单中\n clusterNodeAddSlave(master,n);\n }\n\n /* Set ping sent / pong received timestamps */\n // 设置最近一次发送 PING 命令以及接收 PING 命令回复的时间戳\n if (atoi(argv[4])) n->ping_sent = mstime();\n if (atoi(argv[5])) n->pong_received = mstime();\n\n /* Set configEpoch for this node. */\n // 设置配置纪元\n n->configEpoch = strtoull(argv[6],NULL,10);\n\n /* Populate hash slots served by this instance. */\n // 取出节点服务的槽\n for (j = 8; j < argc; j++) {\n int start, stop;\n\n // 正在导入或导出槽\n if (argv[j][0] == '[') {\n /* Here we handle migrating / importing slots */\n int slot;\n char direction;\n clusterNode *cn;\n\n p = strchr(argv[j],'-');\n redisAssert(p != NULL);\n *p = '\\0';\n // 导入 or 导出？\n direction = p[1]; /* Either '>' or '<' */\n // 槽\n slot = atoi(argv[j]+1);\n p += 3;\n // 目标节点\n cn = clusterLookupNode(p);\n // 如果目标不存在，那么创建\n if (!cn) {\n cn = createClusterNode(p,0);\n clusterAddNode(cn);\n }\n // 根据方向，设定本节点要导入或者导出的槽的目标\n if (direction == '>') {\n server.cluster->migrating_slots_to[slot] = cn;\n } else {\n server.cluster->importing_slots_from[slot] = cn;\n }\n continue;\n\n // 没有导入或导出，这是一个区间范围的槽\n // 比如 0 - 10086\n } else if ((p = strchr(argv[j],'-')) != NULL) {\n *p = '\\0';\n start = atoi(argv[j]);\n stop = atoi(p+1);\n\n // 没有导入或导出，这是单一个槽\n // 比如 10086\n } else {\n start = stop = atoi(argv[j]);\n }\n\n // 将槽载入节点\n while(start <= stop) clusterAddSlot(n, start++);\n }\n\n sdsfreesplitres(argv,argc);\n }\n zfree(line);\n fclose(fp);\n\n /* Config sanity check */\n redisAssert(server.cluster->myself != NULL);\n redisLog(REDIS_NOTICE,\"Node configuration loaded, I'm %.40s\", myself->name);\n\n /* Something that should never happen: currentEpoch smaller than\n * the max epoch found in the nodes configuration. However we handle this\n * as some form of protection against manual editing of critical files. */\n if (clusterGetMaxEpoch() > server.cluster->currentEpoch) {\n server.cluster->currentEpoch = clusterGetMaxEpoch();\n }\n return REDIS_OK;\n\nfmterr:\n redisLog(REDIS_WARNING,\n \"Unrecoverable error: corrupted cluster config file.\");\n fclose(fp);\n exit(1);\n}\n\n/* Cluster node configuration is exactly the same as CLUSTER NODES output.\n *\n * This function writes the node config and returns 0, on error -1\n * is returned.\n *\n * Note: we need to write the file in an atomic way from the point of view\n * of the POSIX filesystem semantics, so that if the server is stopped\n * or crashes during the write, we'll end with either the old file or the\n * new one. Since we have the full payload to write available we can use\n * a single write to write the whole file. If the pre-existing file was\n * bigger we pad our payload with newlines that are anyway ignored and truncate\n * the file afterward. */\n// 写入 nodes.conf 文件\nint clusterSaveConfig(int do_fsync) {\n sds ci;\n size_t content_size;\n struct stat sb;\n int fd;\n\n server.cluster->todo_before_sleep &= ~CLUSTER_TODO_SAVE_CONFIG;\n\n /* Get the nodes description and concatenate our \"vars\" directive to\n * save currentEpoch and lastVoteEpoch. */\n ci = clusterGenNodesDescription(REDIS_NODE_HANDSHAKE);\n ci = sdscatprintf(ci,\"vars currentEpoch %llu lastVoteEpoch %llu\\n\",\n (unsigned long long) server.cluster->currentEpoch,\n (unsigned long long) server.cluster->lastVoteEpoch);\n content_size = sdslen(ci);\n\n if ((fd = open(server.cluster_configfile,O_WRONLY|O_CREAT,0644))\n == -1) goto err;\n\n /* Pad the new payload if the existing file length is greater. */\n if (fstat(fd,&sb) != -1) {\n if (sb.st_size > content_size) {\n ci = sdsgrowzero(ci,sb.st_size);\n memset(ci+content_size,'\\n',sb.st_size-content_size);\n }\n }\n if (write(fd,ci,sdslen(ci)) != (ssize_t)sdslen(ci)) goto err;\n if (do_fsync) {\n server.cluster->todo_before_sleep &= ~CLUSTER_TODO_FSYNC_CONFIG;\n fsync(fd);\n }\n\n /* Truncate the file if needed to remove the final \\n padding that\n * is just garbage. */\n if (content_size != sdslen(ci) && ftruncate(fd,content_size) == -1) {\n /* ftruncate() failing is not a critical error. */\n }\n close(fd);\n sdsfree(ci);\n return 0;\n\nerr:\n if (fd != -1) close(fd);\n sdsfree(ci);\n return -1;\n}\n\n// 尝试写入 nodes.conf 文件，失败则退出\nvoid clusterSaveConfigOrDie(int do_fsync) {\n if (clusterSaveConfig(do_fsync) == -1) {\n redisLog(REDIS_WARNING,\"Fatal: can't update cluster config file.\");\n exit(1);\n }\n}\n\n/* Lock the cluster config using flock(), and leaks the file descritor used to\n * acquire the lock so that the file will be locked forever.\n *\n * This works because we always update nodes.conf with a new version\n * in-place, reopening the file, and writing to it in place (later adjusting\n * the length with ftruncate()).\n *\n * On success REDIS_OK is returned, otherwise an error is logged and\n * the function returns REDIS_ERR to signal a lock was not acquired. */\nint clusterLockConfig(char *filename) {\n /* To lock it, we need to open the file in a way it is created if\n * it does not exist, otherwise there is a race condition with other\n * processes. */\n int fd = open(filename,O_WRONLY|O_CREAT,0644);\n if (fd == -1) {\n redisLog(REDIS_WARNING,\n \"Can't open %s in order to acquire a lock: %s\",\n filename, strerror(errno));\n return REDIS_ERR;\n }\n\n if (flock(fd,LOCK_EX|LOCK_NB) == -1) {\n if (errno == EWOULDBLOCK) {\n redisLog(REDIS_WARNING,\n \"Sorry, the cluster configuration file %s is already used \"\n \"by a different Redis Cluster node. Please make sure that \"\n \"different nodes use different cluster configuration \"\n \"files.\", filename);\n } else {\n redisLog(REDIS_WARNING,\n \"Impossible to lock %s: %s\", filename, strerror(errno));\n }\n close(fd);\n return REDIS_ERR;\n }\n /* Lock acquired: leak the 'fd' by not closing it, so that we'll retain the\n * lock to the file as long as the process exists. */\n return REDIS_OK;\n}\n\n// 初始化集群\nvoid clusterInit(void) {\n int saveconf = 0;\n\n // 初始化配置\n server.cluster = zmalloc(sizeof(clusterState));\n server.cluster->myself = NULL;\n server.cluster->currentEpoch = 0;\n server.cluster->state = REDIS_CLUSTER_FAIL;\n server.cluster->size = 1;\n server.cluster->todo_before_sleep = 0;\n server.cluster->nodes = dictCreate(&clusterNodesDictType,NULL);\n server.cluster->nodes_black_list =\n dictCreate(&clusterNodesBlackListDictType,NULL);\n server.cluster->failover_auth_time = 0;\n server.cluster->failover_auth_count = 0;\n server.cluster->failover_auth_rank = 0;\n server.cluster->failover_auth_epoch = 0;\n server.cluster->lastVoteEpoch = 0;\n server.cluster->stats_bus_messages_sent = 0;\n server.cluster->stats_bus_messages_received = 0;\n memset(server.cluster->slots,0, sizeof(server.cluster->slots));\n clusterCloseAllSlots();\n\n /* Lock the cluster config file to make sure every node uses\n * its own nodes.conf. */\n if (clusterLockConfig(server.cluster_configfile) == REDIS_ERR)\n exit(1);\n\n /* Load or create a new nodes configuration. */\n if (clusterLoadConfig(server.cluster_configfile) == REDIS_ERR) {\n /* No configuration found. We will just use the random name provided\n * by the createClusterNode() function. */\n myself = server.cluster->myself =\n createClusterNode(NULL,REDIS_NODE_MYSELF|REDIS_NODE_MASTER);\n redisLog(REDIS_NOTICE,\"No cluster configuration found, I'm %.40s\",\n myself->name);\n clusterAddNode(myself);\n saveconf = 1;\n }\n\n // 保存 nodes.conf 文件\n if (saveconf) clusterSaveConfigOrDie(1);\n\n /* We need a listening TCP port for our cluster messaging needs. */\n // 监听 TCP 端口\n server.cfd_count = 0;\n\n /* Port sanity check II\n * The other handshake port check is triggered too late to stop\n * us from trying to use a too-high cluster port number. */\n if (server.port > (65535-REDIS_CLUSTER_PORT_INCR)) {\n redisLog(REDIS_WARNING, \"Redis port number too high. \"\n \"Cluster communication port is 10,000 port \"\n \"numbers higher than your Redis port. \"\n \"Your Redis port number must be \"\n \"lower than 55535.\");\n exit(1);\n }\n\n if (listenToPort(server.port+REDIS_CLUSTER_PORT_INCR,\n server.cfd,&server.cfd_count) == REDIS_ERR)\n {\n exit(1);\n } else {\n int j;\n\n for (j = 0; j < server.cfd_count; j++) {\n // 关联监听事件处理器\n if (aeCreateFileEvent(server.el, server.cfd[j], AE_READABLE,\n clusterAcceptHandler, NULL) == AE_ERR)\n redisPanic(\"Unrecoverable error creating Redis Cluster \"\n \"file event.\");\n }\n }\n\n /* The slots -> keys map is a sorted set. Init it. */\n // slots -> keys 映射是一个有序集合\n server.cluster->slots_to_keys = zslCreate();\n resetManualFailover();\n}\n\n/* Reset a node performing a soft or hard reset:\n *\n * 1) All other nodes are forget.\n * 2) All the assigned / open slots are released.\n * 3) If the node is a slave, it turns into a master.\n * 5) Only for hard reset: a new Node ID is generated.\n * 6) Only for hard reset: currentEpoch and configEpoch are set to 0.\n * 7) The new configuration is saved and the cluster state updated. */\nvoid clusterReset(int hard) {\n dictIterator *di;\n dictEntry *de;\n int j;\n\n /* Turn into master. */\n if (nodeIsSlave(myself)) {\n clusterSetNodeAsMaster(myself);\n replicationUnsetMaster();\n }\n\n /* Close slots, reset manual failover state. */\n clusterCloseAllSlots();\n resetManualFailover();\n\n /* Unassign all the slots. */\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) clusterDelSlot(j);\n\n /* Forget all the nodes, but myself. */\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n if (node == myself) continue;\n clusterDelNode(node);\n }\n dictReleaseIterator(di);\n\n /* Hard reset only: set epochs to 0, change node ID. */\n if (hard) {\n sds oldname;\n\n server.cluster->currentEpoch = 0;\n server.cluster->lastVoteEpoch = 0;\n myself->configEpoch = 0;\n\n /* To change the Node ID we need to remove the old name from the\n * nodes table, change the ID, and re-add back with new name. */\n oldname = sdsnewlen(myself->name, REDIS_CLUSTER_NAMELEN);\n dictDelete(server.cluster->nodes,oldname);\n sdsfree(oldname);\n getRandomHexChars(myself->name, REDIS_CLUSTER_NAMELEN);\n clusterAddNode(myself);\n }\n\n /* Make sure to persist the new config and update the state. */\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_FSYNC_CONFIG);\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER communication link\n * -------------------------------------------------------------------------- */\n\n// 创建节点连接\nclusterLink *createClusterLink(clusterNode *node) {\n clusterLink *link = zmalloc(sizeof(*link));\n link->ctime = mstime();\n link->sndbuf = sdsempty();\n link->rcvbuf = sdsempty();\n link->node = node;\n link->fd = -1;\n return link;\n}\n\n/* Free a cluster link, but does not free the associated node of course.\n * This function will just make sure that the original node associated\n * with this link will have the 'link' field set to NULL. */\n// 将给定的连接清空\n// 并将包含这个连接的节点的 link 属性设为 NULL\nvoid freeClusterLink(clusterLink *link) {\n\n // 删除事件处理器\n if (link->fd != -1) {\n aeDeleteFileEvent(server.el, link->fd, AE_WRITABLE);\n aeDeleteFileEvent(server.el, link->fd, AE_READABLE);\n }\n\n // 释放输入缓冲区和输出缓冲区\n sdsfree(link->sndbuf);\n sdsfree(link->rcvbuf);\n\n // 将节点的 link 属性设为 NULL\n if (link->node)\n link->node->link = NULL;\n\n // 关闭连接\n close(link->fd);\n\n // 释放连接结构\n zfree(link);\n}\n\n// 监听事件处理器\n#define MAX_CLUSTER_ACCEPTS_PER_CALL 1000\nvoid clusterAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) {\n int cport, cfd;\n int max = MAX_CLUSTER_ACCEPTS_PER_CALL;\n char cip[REDIS_IP_STR_LEN];\n clusterLink *link;\n REDIS_NOTUSED(el);\n REDIS_NOTUSED(mask);\n REDIS_NOTUSED(privdata);\n\n /* If the server is starting up, don't accept cluster connections:\n * UPDATE messages may interact with the database content. */\n if (server.masterhost == NULL && server.loading) return;\n\n while(max--) {\n cfd = anetTcpAccept(server.neterr, fd, cip, sizeof(cip), &cport);\n if (cfd == ANET_ERR) {\n if (errno != EWOULDBLOCK)\n redisLog(REDIS_VERBOSE,\n \"Accepting cluster node: %s\", server.neterr);\n return;\n }\n anetNonBlock(NULL,cfd);\n anetEnableTcpNoDelay(NULL,cfd);\n\n /* Use non-blocking I/O for cluster messages. */\n redisLog(REDIS_VERBOSE,\"Accepted cluster node %s:%d\", cip, cport);\n /* Create a link object we use to handle the connection.\n * It gets passed to the readable handler when data is available.\n * Initiallly the link->node pointer is set to NULL as we don't know\n * which node is, but the right node is references once we know the\n * node identity. */\n link = createClusterLink(NULL);\n link->fd = cfd;\n aeCreateFileEvent(server.el,cfd,AE_READABLE,clusterReadHandler,link);\n }\n}\n\n/* -----------------------------------------------------------------------------\n * Key space handling\n * -------------------------------------------------------------------------- */\n\n/* We have 16384 hash slots. The hash slot of a given key is obtained\n * as the least significant 14 bits of the crc16 of the key.\n *\n * However if the key contains the {...} pattern, only the part between\n * { and } is hashed. This may be useful in the future to force certain\n * keys to be in the same node (assuming no resharding is in progress). */\n// 计算给定键应该被分配到那个槽\nunsigned int keyHashSlot(char *key, int keylen) {\n int s, e; /* start-end indexes of { and } */\n\n for (s = 0; s < keylen; s++)\n if (key[s] == '{') break;\n\n /* No '{' ? Hash the whole key. This is the base case. */\n if (s == keylen) return crc16(key,keylen) & 0x3FFF;\n\n /* '{' found? Check if we have the corresponding '}'. */\n for (e = s+1; e < keylen; e++)\n if (key[e] == '}') break;\n\n /* No '}' or nothing betweeen {} ? Hash the whole key. */\n if (e == keylen || e == s+1) return crc16(key,keylen) & 0x3FFF;\n\n /* If we are here there is both a { and a } on its right. Hash\n * what is in the middle between { and }. */\n return crc16(key+s+1,e-s-1) & 0x3FFF;\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER node API\n * -------------------------------------------------------------------------- */\n\n/* Create a new cluster node, with the specified flags.\n *\n * 创建一个带有指定 flag 的集群节点。\n *\n * If \"nodename\" is NULL this is considered a first handshake and a random\n * node name is assigned to this node (it will be fixed later when we'll\n * receive the first pong).\n *\n * 如果 nodename 参数为 NULL ，那么表示我们尚未向节点发送 PING ，\n * 集群会为节点设置一个随机的命令，\n * 这个命令在之后接收到节点的 PONG 回复之后就会被更新。\n *\n * The node is created and returned to the user, but it is not automatically\n * added to the nodes hash table. \n *\n * 函数会返回被创建的节点，但不会自动将它添加到当前节点的节点哈希表中\n * （nodes hash table）。\n */\nclusterNode *createClusterNode(char *nodename, int flags) {\n clusterNode *node = zmalloc(sizeof(*node));\n\n // 设置名字\n if (nodename)\n memcpy(node->name, nodename, REDIS_CLUSTER_NAMELEN);\n else\n getRandomHexChars(node->name, REDIS_CLUSTER_NAMELEN);\n\n // 初始化属性\n node->ctime = mstime();\n node->configEpoch = 0;\n node->flags = flags;\n memset(node->slots,0,sizeof(node->slots));\n node->numslots = 0;\n node->numslaves = 0;\n node->slaves = NULL;\n node->slaveof = NULL;\n node->ping_sent = node->pong_received = 0;\n node->fail_time = 0;\n node->link = NULL;\n memset(node->ip,0,sizeof(node->ip));\n node->port = 0;\n node->fail_reports = listCreate();\n node->voted_time = 0;\n node->repl_offset_time = 0;\n node->repl_offset = 0;\n listSetFreeMethod(node->fail_reports,zfree);\n\n return node;\n}\n\n/* This function is called every time we get a failure report from a node.\n *\n * 这个函数会在当前节点接到某个节点的下线报告时调用。\n *\n * The side effect is to populate the fail_reports list (or to update\n * the timestamp of an existing report).\n *\n * 函数的作用就是将下线节点的下线报告添加到 fail_reports 列表，\n * 如果这个下线节点的下线报告已经存在，\n * 那么更新该报告的时间戳。\n *\n * 'failing' is the node that is in failure state according to the\n * 'sender' node.\n *\n * failing 参数指向下线节点，而 sender 参数则指向报告 failing 已下线的节点。\n *\n * The function returns 0 if it just updates a timestamp of an existing\n * failure report from the same sender. 1 is returned if a new failure\n * report is created. \n *\n * 函数返回 0 表示对已存在的报告进行了更新，\n * 返回 1 则表示创建了一条新的下线报告。\n */\nint clusterNodeAddFailureReport(clusterNode *failing, clusterNode *sender) {\n\n // 指向保存下线报告的链表\n list *l = failing->fail_reports;\n\n listNode *ln;\n listIter li;\n clusterNodeFailReport *fr;\n\n /* If a failure report from the same sender already exists, just update\n * the timestamp. */\n // 查找 sender 节点的下线报告是否已经存在\n listRewind(l,&li);\n while ((ln = listNext(&li)) != NULL) {\n fr = ln->value;\n // 如果存在的话，那么只更新该报告的时间戳\n if (fr->node == sender) {\n fr->time = mstime();\n return 0;\n }\n }\n\n /* Otherwise create a new report. */\n // 否则的话，就创建一个新的报告\n fr = zmalloc(sizeof(*fr));\n fr->node = sender;\n fr->time = mstime();\n\n // 将报告添加到列表\n listAddNodeTail(l,fr);\n\n return 1;\n}\n\n/* Remove failure reports that are too old, where too old means reasonably\n * older than the global node timeout. Note that anyway for a node to be\n * flagged as FAIL we need to have a local PFAIL state that is at least\n * older than the global node timeout, so we don't just trust the number\n * of failure reports from other nodes. \n *\n * 移除对 node 节点的过期的下线报告，\n * 多长时间为过期是根据 node timeout 选项的值来决定的。\n *\n * 注意，\n * 要将一个节点标记为 FAIL 状态，\n * 当前节点将 node 标记为 PFAIL 状态的时间至少应该超过 node timeout ，\n * 所以报告 node 已下线的节点数量并不是当前节点将 node 标记为 FAIL 的唯一条件。\n */\nvoid clusterNodeCleanupFailureReports(clusterNode *node) {\n\n // 指向该节点的所有下线报告\n list *l = node->fail_reports;\n\n listNode *ln;\n listIter li;\n clusterNodeFailReport *fr;\n\n // 下线报告的最大保质期（超过这个时间的报告会被删除）\n mstime_t maxtime = server.cluster_node_timeout *\n REDIS_CLUSTER_FAIL_REPORT_VALIDITY_MULT;\n mstime_t now = mstime();\n\n // 遍历所有下线报告\n listRewind(l,&li);\n while ((ln = listNext(&li)) != NULL) {\n fr = ln->value;\n // 删除过期报告\n if (now - fr->time > maxtime) listDelNode(l,ln);\n }\n}\n\n/* Remove the failing report for 'node' if it was previously considered\n * failing by 'sender'. This function is called when a node informs us via\n * gossip that a node is OK from its point of view (no FAIL or PFAIL flags).\n *\n * 从 node 节点的下线报告中移除 sender 对 node 的下线报告。\n *\n * 这个函数在以下情况使用：当前节点认为 node 已下线（FAIL 或者 PFAIL），\n * 但 sender 却向当前节点发来报告，说它认为 node 节点没有下线，\n * 那么当前节点就要移除 sender 对 node 的下线报告 \n * —— 如果 sender 曾经报告过 node 下线的话。\n *\n * Note that this function is called relatively often as it gets called even\n * when there are no nodes failing, and is O(N), however when the cluster is\n * fine the failure reports list is empty so the function runs in constant\n * time.\n *\n * 即使在节点没有下线的情况下，这个函数也会被调用，并且调用的次数还比较频繁。\n * 在一般情况下，这个函数的复杂度为 O(N) ，\n * 不过在不存在下线报告的情况下，这个函数的复杂度仅为常数时间。\n *\n * The function returns 1 if the failure report was found and removed.\n * Otherwise 0 is returned. \n *\n * 函数返回 1 表示下线报告已经被成功移除，\n * 0 表示 sender 没有发送过 node 的下线报告，删除失败。\n */\nint clusterNodeDelFailureReport(clusterNode *node, clusterNode *sender) {\n list *l = node->fail_reports;\n listNode *ln;\n listIter li;\n clusterNodeFailReport *fr;\n\n /* Search for a failure report from this sender. */\n // 查找 sender 对 node 的下线报告\n listRewind(l,&li);\n while ((ln = listNext(&li)) != NULL) {\n fr = ln->value;\n if (fr->node == sender) break;\n }\n // sender 没有报告过 node 下线，直接返回\n if (!ln) return 0; /* No failure report from this sender. */\n\n /* Remove the failure report. */\n // 删除 sender 对 node 的下线报告\n listDelNode(l,ln);\n // 删除对 node 的下线报告中，过期的报告\n clusterNodeCleanupFailureReports(node);\n\n return 1;\n}\n\n/* Return the number of external nodes that believe 'node' is failing,\n * not including this node, that may have a PFAIL or FAIL state for this\n * node as well. \n *\n * 计算不包括本节点在内的，\n * 将 node 标记为 PFAIL 或者 FAIL 的节点的数量。\n */\nint clusterNodeFailureReportsCount(clusterNode *node) {\n\n // 移除过期的下线报告\n clusterNodeCleanupFailureReports(node);\n\n // 统计下线报告的数量\n return listLength(node->fail_reports);\n}\n\n// 移除主节点 master 的从节点 slave\nint clusterNodeRemoveSlave(clusterNode *master, clusterNode *slave) {\n int j;\n\n // 在 slaves 数组中找到从节点 slave 所属的主节点，\n // 将主节点中的 slave 信息移除\n for (j = 0; j < master->numslaves; j++) {\n if (master->slaves[j] == slave) {\n memmove(master->slaves+j,master->slaves+(j+1),\n (master->numslaves-1)-j);\n master->numslaves--;\n return REDIS_OK;\n }\n }\n return REDIS_ERR;\n}\n\n// 将 slave 加入到 master 的从节点名单中\nint clusterNodeAddSlave(clusterNode *master, clusterNode *slave) {\n int j;\n\n /* If it's already a slave, don't add it again. */\n // 如果 slave 已经存在，那么不做操作\n for (j = 0; j < master->numslaves; j++)\n if (master->slaves[j] == slave) return REDIS_ERR;\n\n // 将 slave 添加到 slaves 数组里面\n master->slaves = zrealloc(master->slaves,\n sizeof(clusterNode*)*(master->numslaves+1));\n master->slaves[master->numslaves] = slave;\n master->numslaves++;\n\n return REDIS_OK;\n}\n\n// 重置给定节点的从节点名单\nvoid clusterNodeResetSlaves(clusterNode *n) {\n zfree(n->slaves);\n n->numslaves = 0;\n n->slaves = NULL;\n}\n\nint clusterCountNonFailingSlaves(clusterNode *n) {\n int j, okslaves = 0;\n\n for (j = 0; j < n->numslaves; j++)\n if (!nodeFailed(n->slaves[j])) okslaves++;\n return okslaves;\n}\n\n// 释放节点\nvoid freeClusterNode(clusterNode *n) {\n sds nodename;\n\n nodename = sdsnewlen(n->name, REDIS_CLUSTER_NAMELEN);\n\n // 从 nodes 表中删除节点\n redisAssert(dictDelete(server.cluster->nodes,nodename) == DICT_OK);\n sdsfree(nodename);\n\n // 移除从节点\n if (n->slaveof) clusterNodeRemoveSlave(n->slaveof, n);\n\n // 释放连接\n if (n->link) freeClusterLink(n->link);\n \n // 释放失败报告\n listRelease(n->fail_reports);\n\n // 释放节点结构\n zfree(n);\n}\n\n/* Add a node to the nodes hash table */\n// 将给定 node 添加到节点表里面\nint clusterAddNode(clusterNode *node) {\n int retval;\n // 将 node 添加到当前节点的 nodes 表中\n // 这样接下来当前节点就会创建连向 node 的节点\n retval = dictAdd(server.cluster->nodes,\n sdsnewlen(node->name,REDIS_CLUSTER_NAMELEN), node);\n return (retval == DICT_OK) ? REDIS_OK : REDIS_ERR;\n}\n\n/* Remove a node from the cluster:\n *\n * 从集群中移除一个节点：\n *\n * 1) Mark all the nodes handled by it as unassigned.\n * 将所有由该节点负责的槽全部设置为未分配\n * 2) Remove all the failure reports sent by this node.\n * 移除所有由这个节点发送的下线报告（failure report）\n * 3) Free the node, that will in turn remove it from the hash table\n * and from the list of slaves of its master, if it is a slave node.\n * 释放这个节点，\n * 清除它在各个节点的 nodes 表中的数据，\n * 如果它是一个从节点的话，\n * 还要在它的主节点的 slaves 表中清除关于这个节点的数据。\n */\nvoid clusterDelNode(clusterNode *delnode) {\n int j;\n dictIterator *di;\n dictEntry *de;\n\n /* 1) Mark slots as unassigned. */\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n // 取消向该节点接收槽的计划\n if (server.cluster->importing_slots_from[j] == delnode)\n server.cluster->importing_slots_from[j] = NULL;\n // 取消向该节点移交槽的计划\n if (server.cluster->migrating_slots_to[j] == delnode)\n server.cluster->migrating_slots_to[j] = NULL;\n // 将所有由该节点负责的槽设置为未分配\n if (server.cluster->slots[j] == delnode)\n clusterDelSlot(j);\n }\n\n /* 2) Remove failure reports. */\n // 移除所有由该节点发送的下线报告\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n if (node == delnode) continue;\n clusterNodeDelFailureReport(node,delnode);\n }\n dictReleaseIterator(di);\n\n /* 3) Remove this node from its master's slaves if needed. */\n // 将节点从它的主节点的从节点列表中移除\n if (nodeIsSlave(delnode) && delnode->slaveof)\n clusterNodeRemoveSlave(delnode->slaveof,delnode);\n\n /* 4) Free the node, unlinking it from the cluster. */\n // 释放节点\n freeClusterNode(delnode);\n}\n\n/* Node lookup by name */\n// 根据名字，查找给定的节点\nclusterNode *clusterLookupNode(char *name) {\n sds s = sdsnewlen(name, REDIS_CLUSTER_NAMELEN);\n dictEntry *de;\n\n de = dictFind(server.cluster->nodes,s);\n sdsfree(s);\n if (de == NULL) return NULL;\n return dictGetVal(de);\n}\n\n/* This is only used after the handshake. When we connect a given IP/PORT\n * as a result of CLUSTER MEET we don't have the node name yet, so we\n * pick a random one, and will fix it when we receive the PONG request using\n * this function. */\n// 在第一次向节点发送 CLUSTER MEET 命令的时候\n// 因为发送命令的节点还不知道目标节点的名字\n// 所以它会给目标节点分配一个随机的名字\n// 当目标节点向发送节点返回 PONG 回复时\n// 发送节点就知道了目标节点的 IP 和 port\n// 这时发送节点就可以通过调用这个函数\n// 为目标节点改名\nvoid clusterRenameNode(clusterNode *node, char *newname) {\n int retval;\n sds s = sdsnewlen(node->name, REDIS_CLUSTER_NAMELEN);\n\n redisLog(REDIS_DEBUG,\"Renaming node %.40s into %.40s\",\n node->name, newname);\n retval = dictDelete(server.cluster->nodes, s);\n sdsfree(s);\n redisAssert(retval == DICT_OK);\n memcpy(node->name, newname, REDIS_CLUSTER_NAMELEN);\n clusterAddNode(node);\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER nodes blacklist\n *\n * 集群节点黑名单\n *\n * The nodes blacklist is just a way to ensure that a given node with a given\n * Node ID is not readded before some time elapsed (this time is specified\n * in seconds in REDIS_CLUSTER_BLACKLIST_TTL).\n *\n * 黑名单用于禁止一个给定的节点在 REDIS_CLUSTER_BLACKLIST_TTL 指定的时间内，\n * 被重新添加到集群中。\n *\n * This is useful when we want to remove a node from the cluster completely:\n * when CLUSTER FORGET is called, it also puts the node into the blacklist so\n * that even if we receive gossip messages from other nodes that still remember\n * about the node we want to remove, we don't re-add it before some time.\n *\n * 当我们需要从集群中彻底移除一个节点时，就需要用到黑名单：\n * 在执行 CLUSTER FORGET 命令时，节点会被添加进黑名单里面，\n * 这样即使我们从仍然记得被移除节点的其他节点那里收到关于被移除节点的消息，\n * 我们也不会重新将被移除节点添加至集群。\n *\n * Currently the REDIS_CLUSTER_BLACKLIST_TTL is set to 1 minute, this means\n * that redis-trib has 60 seconds to send CLUSTER FORGET messages to nodes\n * in the cluster without dealing with the problem of other nodes re-adding\n * back the node to nodes we already sent the FORGET command to.\n *\n * REDIS_CLUSTER_BLACKLIST_TTL 当前的值为 1 分钟，\n * 这意味着 redis-trib 有 60 秒的时间，可以向集群中的所有节点发送 CLUSTER FORGET\n * 命令，而不必担心有其他节点会将被 CLUSTER FORGET 移除的节点重新添加到集群里面。\n *\n * The data structure used is a hash table with an sds string representing\n * the node ID as key, and the time when it is ok to re-add the node as\n * value.\n *\n * 黑名单的底层实现是一个字典，\n * 字典的键为 SDS 表示的节点 id ，字典的值为可以重新添加节点的时间戳。\n * -------------------------------------------------------------------------- */\n\n#define REDIS_CLUSTER_BLACKLIST_TTL 60 /* 1 minute. */\n\n\n/* Before of the addNode() or Exists() operations we always remove expired\n * entries from the black list. This is an O(N) operation but it is not a\n * problem since add / exists operations are called very infrequently and\n * the hash table is supposed to contain very little elements at max.\n *\n * 在执行 addNode() 操作或者 Exists() 操作之前，\n * 我们总是会先执行这个函数，移除黑名单中的过期节点。\n *\n * 这个函数的复杂度为 O(N) ，不过它不会对效率产生影响，\n * 因为这个函数执行的次数并不频繁，并且字典的链表里面包含的节点数量也非常少。\n *\n * However without the cleanup during long uptimes and with some automated\n * node add/removal procedures, entries could accumulate. \n *\n * 定期清理过期节点是为了防止字典中的节点堆积过多。\n */\nvoid clusterBlacklistCleanup(void) {\n dictIterator *di;\n dictEntry *de;\n\n // 遍历黑名单中的所有节点\n di = dictGetSafeIterator(server.cluster->nodes_black_list);\n while((de = dictNext(di)) != NULL) {\n int64_t expire = dictGetUnsignedIntegerVal(de);\n\n // 删除过期节点\n if (expire < server.unixtime)\n dictDelete(server.cluster->nodes_black_list,dictGetKey(de));\n }\n dictReleaseIterator(di);\n}\n\n/* Cleanup the blacklist and add a new node ID to the black list. */\n// 清除黑名单中的过期节点，然后将新的节点添加到黑名单中\nvoid clusterBlacklistAddNode(clusterNode *node) {\n dictEntry *de;\n sds id = sdsnewlen(node->name,REDIS_CLUSTER_NAMELEN);\n\n // 先清理过期名单\n clusterBlacklistCleanup();\n\n // 添加节点\n if (dictAdd(server.cluster->nodes_black_list,id,NULL) == DICT_OK) {\n /* If the key was added, duplicate the sds string representation of\n * the key for the next lookup. We'll free it at the end. */\n id = sdsdup(id);\n }\n // 设置过期时间\n de = dictFind(server.cluster->nodes_black_list,id);\n dictSetUnsignedIntegerVal(de,time(NULL)+REDIS_CLUSTER_BLACKLIST_TTL);\n sdsfree(id);\n}\n\n/* Return non-zero if the specified node ID exists in the blacklist.\n * You don't need to pass an sds string here, any pointer to 40 bytes\n * will work. */\n// 检查给定 id 所指定的节点是否存在于黑名单中。\n// nodeid 参数不必是一个 SDS 值，只要一个 40 字节长的字符串即可\nint clusterBlacklistExists(char *nodeid) {\n\n // 构建 SDS 表示的节点名\n sds id = sdsnewlen(nodeid,REDIS_CLUSTER_NAMELEN);\n int retval;\n\n // 清除过期黑名单\n clusterBlacklistCleanup();\n\n // 检查节点是否存在\n retval = dictFind(server.cluster->nodes_black_list,id) != NULL;\n sdsfree(id);\n\n return retval;\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER messages exchange - PING/PONG and gossip\n * -------------------------------------------------------------------------- */\n\n/* This function checks if a given node should be marked as FAIL.\n * It happens if the following conditions are met:\n *\n * 此函数用于判断是否需要将 node 标记为 FAIL 。\n *\n * 将 node 标记为 FAIL 需要满足以下两个条件：\n *\n * 1) We received enough failure reports from other master nodes via gossip.\n * Enough means that the majority of the masters signaled the node is\n * down recently.\n * 有半数以上的主节点将 node 标记为 PFAIL 状态。\n * 2) We believe this node is in PFAIL state.\n * 当前节点也将 node 标记为 PFAIL 状态。\n *\n * If a failure is detected we also inform the whole cluster about this\n * event trying to force every other node to set the FAIL flag for the node.\n *\n * 如果确认 node 已经进入了 FAIL 状态，\n * 那么节点还会向其他节点发送 FAIL 消息，让其他节点也将 node 标记为 FAIL 。\n *\n * Note that the form of agreement used here is weak, as we collect the majority\n * of masters state during some time, and even if we force agreement by\n * propagating the FAIL message, because of partitions we may not reach every\n * node. However:\n *\n * 注意，集群判断一个 node 进入 FAIL 所需的条件是弱（weak）的，\n * 因为节点们对 node 的状态报告并不是实时的，而是有一段时间间隔\n * （这段时间内 node 的状态可能已经发生了改变），\n * 并且尽管当前节点会向其他节点发送 FAIL 消息，\n * 但因为网络分裂（network partition）的问题，\n * 有一部分节点可能还是会不知道将 node 标记为 FAIL 。\n *\n * 不过：\n *\n * 1) Either we reach the majority and eventually the FAIL state will propagate\n * to all the cluster.\n * 只要我们成功将 node 标记为 FAIL ，\n * 那么这个 FAIL 状态最终（eventually）总会传播至整个集群的所有节点。\n * 2) Or there is no majority so no slave promotion will be authorized and the\n * FAIL flag will be cleared after some time.\n * 又或者，因为没有半数的节点支持，当前节点不能将 node 标记为 FAIL ，\n * 所以对 FAIL 节点的故障转移将无法进行， FAIL 标识可能会在之后被移除。\n * \n */\nvoid markNodeAsFailingIfNeeded(clusterNode *node) {\n int failures;\n\n // 标记为 FAIL 所需的节点数量，需要超过集群节点数量的一半\n int needed_quorum = (server.cluster->size / 2) + 1;\n\n if (!nodeTimedOut(node)) return; /* We can reach it. */\n if (nodeFailed(node)) return; /* Already FAILing. */\n\n // 统计将 node 标记为 PFAIL 或者 FAIL 的节点数量（不包括当前节点）\n failures = clusterNodeFailureReportsCount(node);\n\n /* Also count myself as a voter if I'm a master. */\n // 如果当前节点是主节点，那么将当前节点也算在 failures 之内\n if (nodeIsMaster(myself)) failures++;\n // 报告下线节点的数量不足节点总数的一半，不能将节点判断为 FAIL ，返回\n if (failures < needed_quorum) return; /* No weak agreement from masters. */\n\n redisLog(REDIS_NOTICE,\n \"Marking node %.40s as failing (quorum reached).\", node->name);\n\n /* Mark the node as failing. */\n // 将 node 标记为 FAIL\n node->flags &= ~REDIS_NODE_PFAIL;\n node->flags |= REDIS_NODE_FAIL;\n node->fail_time = mstime();\n\n /* Broadcast the failing node name to everybody, forcing all the other\n * reachable nodes to flag the node as FAIL. */\n // 如果当前节点是主节点的话，那么向其他节点发送报告 node 的 FAIL 信息\n // 让其他节点也将 node 标记为 FAIL\n if (nodeIsMaster(myself)) clusterSendFail(node->name);\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n}\n\n/* This function is called only if a node is marked as FAIL, but we are able\n * to reach it again. It checks if there are the conditions to undo the FAIL\n * state. \n *\n * 这个函数在当前节点接收到一个被标记为 FAIL 的节点那里收到消息时使用，\n * 它可以检查是否应该将节点的 FAIL 状态移除。\n */\nvoid clearNodeFailureIfNeeded(clusterNode *node) {\n mstime_t now = mstime();\n\n redisAssert(nodeFailed(node));\n\n /* For slaves we always clear the FAIL flag if we can contact the\n * node again. */\n // 如果 FAIL 的是从节点，那么当前节点会直接移除该节点的 FAIL\n if (nodeIsSlave(node) || node->numslots == 0) {\n redisLog(REDIS_NOTICE,\n \"Clear FAIL state for node %.40s: %s is reachable again.\",\n node->name,\n nodeIsSlave(node) ? \"slave\" : \"master without slots\");\n // 移除\n node->flags &= ~REDIS_NODE_FAIL;\n\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n }\n\n /* If it is a master and...\n *\n * 如果 FAIL 的是一个主节点，并且：\n *\n * 1) The FAIL state is old enough.\n * 节点被标记为 FAIL 状态已经有一段时间了\n *\n * 2) It is yet serving slots from our point of view (not failed over).\n * 从当前节点的视角来看，这个节点还有负责处理的槽\n *\n * Apparently no one is going to fix these slots, clear the FAIL flag. \n *\n * 那么说明 FAIL 节点仍然有槽没有迁移完，那么当前节点移除该节点的 FAIL 标识。\n */\n if (nodeIsMaster(node) && node->numslots > 0 &&\n (now - node->fail_time) >\n (server.cluster_node_timeout * REDIS_CLUSTER_FAIL_UNDO_TIME_MULT))\n {\n redisLog(REDIS_NOTICE,\n \"Clear FAIL state for node %.40s: is reachable again and nobody is serving its slots after some time.\",\n node->name);\n\n // 撤销 FAIL 状态\n node->flags &= ~REDIS_NODE_FAIL;\n\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n }\n}\n\n/* Return true if we already have a node in HANDSHAKE state matching the\n * specified ip address and port number. This function is used in order to\n * avoid adding a new handshake node for the same address multiple times. \n *\n * 如果当前节点已经向 ip 和 port 所指定的节点进行了握手，\n * 那么返回 1 。\n *\n * 这个函数用于防止对同一个节点进行多次握手。\n */\nint clusterHandshakeInProgress(char *ip, int port) {\n dictIterator *di;\n dictEntry *de;\n\n // 遍历所有已知节点\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n // 跳过非握手状态的节点，之后剩下的都是正在握手的节点\n if (!nodeInHandshake(node)) continue;\n\n // 给定 ip 和 port 的节点正在进行握手\n if (!strcasecmp(node->ip,ip) && node->port == port) break;\n }\n dictReleaseIterator(di);\n\n // 检查节点是否正在握手\n return de != NULL;\n}\n\n/* Start an handshake with the specified address if there is not one\n * already in progress. Returns non-zero if the handshake was actually\n * started. On error zero is returned and errno is set to one of the\n * following values:\n *\n * 如果还没有与指定的地址进行过握手，那么进行握手。\n * 返回 1 表示握手已经开始，\n * 返回 0 并将 errno 设置为以下值来表示意外情况：\n *\n * EAGAIN - There is already an handshake in progress for this address.\n * 已经有握手在进行中了。\n * EINVAL - IP or port are not valid. \n * ip 或者 port 参数不合法。\n */\nint clusterStartHandshake(char *ip, int port) {\n clusterNode *n;\n char norm_ip[REDIS_IP_STR_LEN];\n struct sockaddr_storage sa;\n\n /* IP sanity check */\n // ip 合法性检查\n if (inet_pton(AF_INET,ip,\n &(((struct sockaddr_in *)&sa)->sin_addr)))\n {\n sa.ss_family = AF_INET;\n } else if (inet_pton(AF_INET6,ip,\n &(((struct sockaddr_in6 *)&sa)->sin6_addr)))\n {\n sa.ss_family = AF_INET6;\n } else {\n errno = EINVAL;\n return 0;\n }\n\n /* Port sanity check */\n // port 合法性检查\n if (port <= 0 || port > (65535-REDIS_CLUSTER_PORT_INCR)) {\n errno = EINVAL;\n return 0;\n }\n\n /* Set norm_ip as the normalized string representation of the node\n * IP address. */\n if (sa.ss_family == AF_INET)\n inet_ntop(AF_INET,\n (void*)&(((struct sockaddr_in *)&sa)->sin_addr),\n norm_ip,REDIS_IP_STR_LEN);\n else\n inet_ntop(AF_INET6,\n (void*)&(((struct sockaddr_in6 *)&sa)->sin6_addr),\n norm_ip,REDIS_IP_STR_LEN);\n\n // 检查节点是否已经发送握手请求，如果是的话，那么直接返回，防止出现重复握手\n if (clusterHandshakeInProgress(norm_ip,port)) {\n errno = EAGAIN;\n return 0;\n }\n\n /* Add the node with a random address (NULL as first argument to\n * createClusterNode()). Everything will be fixed during the\n * handskake. */\n // 对给定地址的节点设置一个随机名字\n // 当 HANDSHAKE 完成时，当前节点会取得给定地址节点的真正名字\n // 到时会用真名替换随机名\n n = createClusterNode(NULL,REDIS_NODE_HANDSHAKE|REDIS_NODE_MEET);\n memcpy(n->ip,norm_ip,sizeof(n->ip));\n n->port = port;\n\n // 将节点添加到集群当中\n clusterAddNode(n);\n\n return 1;\n}\n\n/* Process the gossip section of PING or PONG packets.\n *\n * 解释 MEET 、 PING 或 PONG 消息中和 gossip 协议有关的信息。\n *\n * Note that this function assumes that the packet is already sanity-checked\n * by the caller, not in the content of the gossip section, but in the\n * length. \n *\n * 注意，这个函数假设调用者已经根据消息的长度，对消息进行过合法性检查。\n */\nvoid clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) {\n\n // 记录这条消息中包含了多少个节点的信息\n uint16_t count = ntohs(hdr->count);\n\n // 指向第一个节点的信息\n clusterMsgDataGossip *g = (clusterMsgDataGossip*) hdr->data.ping.gossip;\n\n // 取出发送者\n clusterNode *sender = link->node ? link->node : clusterLookupNode(hdr->sender);\n\n // 遍历所有节点的信息\n while(count--) {\n sds ci = sdsempty();\n\n // 分析节点的 flag\n uint16_t flags = ntohs(g->flags);\n\n // 信息节点\n clusterNode *node;\n\n // 取出节点的 flag\n if (flags == 0) ci = sdscat(ci,\"noflags,\");\n if (flags & REDIS_NODE_MYSELF) ci = sdscat(ci,\"myself,\");\n if (flags & REDIS_NODE_MASTER) ci = sdscat(ci,\"master,\");\n if (flags & REDIS_NODE_SLAVE) ci = sdscat(ci,\"slave,\");\n if (flags & REDIS_NODE_PFAIL) ci = sdscat(ci,\"fail?,\");\n if (flags & REDIS_NODE_FAIL) ci = sdscat(ci,\"fail,\");\n if (flags & REDIS_NODE_HANDSHAKE) ci = sdscat(ci,\"handshake,\");\n if (flags & REDIS_NODE_NOADDR) ci = sdscat(ci,\"noaddr,\");\n if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';\n\n redisLog(REDIS_DEBUG,\"GOSSIP %.40s %s:%d %s\",\n g->nodename,\n g->ip,\n ntohs(g->port),\n ci);\n sdsfree(ci);\n\n /* Update our state accordingly to the gossip sections */\n // 使用消息中的信息对节点进行更新\n node = clusterLookupNode(g->nodename);\n // 节点已经存在于当前节点\n if (node) {\n /* We already know this node.\n Handle failure reports, only when the sender is a master. */\n // 如果 sender 是一个主节点，那么我们需要处理下线报告\n if (sender && nodeIsMaster(sender) && node != myself) {\n // 节点处于 FAIL 或者 PFAIL 状态\n if (flags & (REDIS_NODE_FAIL|REDIS_NODE_PFAIL)) {\n\n // 添加 sender 对 node 的下线报告\n if (clusterNodeAddFailureReport(node,sender)) {\n redisLog(REDIS_VERBOSE,\n \"Node %.40s reported node %.40s as not reachable.\",\n sender->name, node->name);\n }\n\n // 尝试将 node 标记为 FAIL\n markNodeAsFailingIfNeeded(node);\n\n // 节点处于正常状态\n } else {\n\n // 如果 sender 曾经发送过对 node 的下线报告\n // 那么清除该报告\n if (clusterNodeDelFailureReport(node,sender)) {\n redisLog(REDIS_VERBOSE,\n \"Node %.40s reported node %.40s is back online.\",\n sender->name, node->name);\n }\n }\n }\n\n /* If we already know this node, but it is not reachable, and\n * we see a different address in the gossip section, start an\n * handshake with the (possibly) new address: this will result\n * into a node address update if the handshake will be\n * successful. */\n // 如果节点之前处于 PFAIL 或者 FAIL 状态\n // 并且该节点的 IP 或者端口号已经发生变化\n // 那么可能是节点换了新地址，尝试对它进行握手\n if (node->flags & (REDIS_NODE_FAIL|REDIS_NODE_PFAIL) &&\n (strcasecmp(node->ip,g->ip) || node->port != ntohs(g->port)))\n {\n clusterStartHandshake(g->ip,ntohs(g->port));\n }\n\n // 当前节点不认识 node\n } else {\n /* If it's not in NOADDR state and we don't have it, we\n * start a handshake process against this IP/PORT pairs.\n *\n * 如果 node 不在 NOADDR 状态，并且当前节点不认识 node \n * 那么向 node 发送 HANDSHAKE 消息。\n *\n * Note that we require that the sender of this gossip message\n * is a well known node in our cluster, otherwise we risk\n * joining another cluster.\n *\n * 注意，当前节点必须保证 sender 是本集群的节点，\n * 否则我们将有加入了另一个集群的风险。\n */\n if (sender &&\n !(flags & REDIS_NODE_NOADDR) &&\n !clusterBlacklistExists(g->nodename))\n {\n clusterStartHandshake(g->ip,ntohs(g->port));\n }\n }\n\n /* Next node */\n // 处理下个节点的信息\n g++;\n }\n}\n\n/* IP -> string conversion. 'buf' is supposed to at least be 46 bytes. */\n// 将 ip 转换为字符串\nvoid nodeIp2String(char *buf, clusterLink *link) {\n anetPeerToString(link->fd, buf, REDIS_IP_STR_LEN, NULL);\n}\n\n/* Update the node address to the IP address that can be extracted\n * from link->fd, and at the specified port.\n *\n * 更新节点的地址， IP 和端口可以从 link->fd 获得。\n *\n * Also disconnect the node link so that we'll connect again to the new\n * address.\n *\n * 并且断开当前的节点连接，并根据新地址创建新连接。\n *\n * If the ip/port pair are already correct no operation is performed at\n * all.\n *\n * 如果 ip 和端口和现在的连接相同，那么不执行任何动作。\n *\n * The function returns 0 if the node address is still the same,\n * otherwise 1 is returned. \n *\n * 函数返回 0 表示地址不变，地址已被更新则返回 1 。\n */\nint nodeUpdateAddressIfNeeded(clusterNode *node, clusterLink *link, int port) {\n char ip[REDIS_IP_STR_LEN];\n\n /* We don't proceed if the link is the same as the sender link, as this\n * function is designed to see if the node link is consistent with the\n * symmetric link that is used to receive PINGs from the node.\n *\n * As a side effect this function never frees the passed 'link', so\n * it is safe to call during packet processing. */\n // 连接不变，直接返回\n if (link == node->link) return 0;\n\n // 获取字符串格式的 ip 地址\n nodeIp2String(ip,link);\n // 获取端口号\n if (node->port == port && strcmp(ip,node->ip) == 0) return 0;\n\n /* IP / port is different, update it. */\n memcpy(node->ip,ip,sizeof(ip));\n node->port = port;\n\n // 释放旧连接（新连接会在之后自动创建）\n if (node->link) freeClusterLink(node->link);\n\n redisLog(REDIS_WARNING,\"Address updated for node %.40s, now %s:%d\",\n node->name, node->ip, node->port);\n\n /* Check if this is our master and we have to change the\n * replication target as well. */\n // 如果连接来自当前节点（从节点）的主节点，那么根据新地址设置复制对象\n if (nodeIsSlave(myself) && myself->slaveof == node)\n replicationSetMaster(node->ip, node->port);\n return 1;\n}\n\n/* Reconfigure the specified node 'n' as a master. This function is called when\n * a node that we believed to be a slave is now acting as master in order to\n * update the state of the node. \n *\n * 将节点 n 设置为主节点。\n */\nvoid clusterSetNodeAsMaster(clusterNode *n) {\n\n // 已经是主节点了。\n if (nodeIsMaster(n)) return;\n\n // 移除 slaveof\n if (n->slaveof) clusterNodeRemoveSlave(n->slaveof,n);\n\n // 关闭 SLAVE 标识\n n->flags &= ~REDIS_NODE_SLAVE;\n\n // 打开 MASTER 标识\n n->flags |= REDIS_NODE_MASTER;\n\n // 清零 slaveof 属性\n n->slaveof = NULL;\n\n /* Update config and state. */\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n}\n\n/* This function is called when we receive a master configuration via a\n * PING, PONG or UPDATE packet. What we receive is a node, a configEpoch of the\n * node, and the set of slots claimed under this configEpoch.\n *\n * 这个函数在节点通过 PING 、 PONG 、 UPDATE 消息接收到一个 master 的配置时调用，\n * 函数以一个节点，节点的 configEpoch ，\n * 以及节点在 configEpoch 纪元下的槽配置作为参数。\n *\n * What we do is to rebind the slots with newer configuration compared to our\n * local configuration, and if needed, we turn ourself into a replica of the\n * node (see the function comments for more info).\n *\n * 这个函数要做的就是在 slots 参数的新配置和本节点的当前配置进行对比，\n * 并更新本节点对槽的布局，\n * 如果有需要的话，函数还会将本节点转换为 sender 的从节点，\n * 更多信息请参考函数中的注释。\n *\n * The 'sender' is the node for which we received a configuration update.\n * Sometimes it is not actaully the \"Sender\" of the information, like in the case\n * we receive the info via an UPDATE packet. \n *\n * 根据情况， sender 参数可以是消息的发送者，也可以是消息发送者的主节点。\n */\nvoid clusterUpdateSlotsConfigWith(clusterNode *sender, uint64_t senderConfigEpoch, unsigned char *slots) {\n int j;\n clusterNode *curmaster, *newmaster = NULL;\n /* The dirty slots list is a list of slots for which we lose the ownership\n * while having still keys inside. This usually happens after a failover\n * or after a manual cluster reconfiguration operated by the admin.\n *\n * If the update message is not able to demote a master to slave (in this\n * case we'll resync with the master updating the whole key space), we\n * need to delete all the keys in the slots we lost ownership. */\n uint16_t dirty_slots[REDIS_CLUSTER_SLOTS];\n int dirty_slots_count = 0;\n\n /* Here we set curmaster to this node or the node this node\n * replicates to if it's a slave. In the for loop we are\n * interested to check if slots are taken away from curmaster. */\n // 1）如果当前节点是主节点，那么将 curmaster 设置为当前节点\n // 2）如果当前节点是从节点，那么将 curmaster 设置为当前节点正在复制的主节点\n // 稍后在 for 循环中我们将使用 curmaster 检查与当前节点有关的槽是否发生了变动\n curmaster = nodeIsMaster(myself) ? myself : myself->slaveof;\n\n if (sender == myself) {\n redisLog(REDIS_WARNING,\"Discarding UPDATE message about myself.\");\n return;\n }\n\n // 更新槽布局\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n\n // 如果 slots 中的槽 j 已经被指派，那么执行以下代码\n if (bitmapTestBit(slots,j)) {\n /* The slot is already bound to the sender of this message. */\n if (server.cluster->slots[j] == sender) continue;\n\n /* The slot is in importing state, it should be modified only\n * manually via redis-trib (example: a resharding is in progress\n * and the migrating side slot was already closed and is advertising\n * a new config. We still want the slot to be closed manually). */\n if (server.cluster->importing_slots_from[j]) continue;\n\n /* We rebind the slot to the new node claiming it if:\n * 1) The slot was unassigned or the new node claims it with a\n * greater configEpoch.\n * 2) We are not currently importing the slot. */\n if (server.cluster->slots[j] == NULL ||\n server.cluster->slots[j]->configEpoch < senderConfigEpoch)\n {\n /* Was this slot mine, and still contains keys? Mark it as\n * a dirty slot. */\n if (server.cluster->slots[j] == myself &&\n countKeysInSlot(j) &&\n sender != myself)\n {\n dirty_slots[dirty_slots_count] = j;\n dirty_slots_count++;\n }\n\n // 负责槽 j 的原节点是当前节点的主节点？\n // 如果是的话，说明故障转移发生了，将当前节点的复制对象设置为新的主节点\n if (server.cluster->slots[j] == curmaster)\n newmaster = sender;\n\n // 将槽 j 设为未指派\n clusterDelSlot(j);\n\n // 将槽 j 指派给 sender\n clusterAddSlot(sender,j);\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_FSYNC_CONFIG);\n }\n }\n }\n\n /* If at least one slot was reassigned from a node to another node\n * with a greater configEpoch, it is possible that:\n *\n * 如果当前节点（或者当前节点的主节点）有至少一个槽被指派到了 sender\n * 并且 sender 的 configEpoch 比当前节点的纪元要大，\n * 那么可能发生了：\n *\n * 1) We are a master left without slots. This means that we were\n * failed over and we should turn into a replica of the new\n * master.\n * 当前节点是一个不再处理任何槽的主节点，\n * 这时应该将当前节点设置为新主节点的从节点。\n * 2) We are a slave and our master is left without slots. We need\n * to replicate to the new slots owner. \n * 当前节点是一个从节点，\n * 并且当前节点的主节点已经不再处理任何槽，\n * 这时应该将当前节点设置为新主节点的从节点。\n */\n if (newmaster && curmaster->numslots == 0) {\n redisLog(REDIS_WARNING,\n \"Configuration change detected. Reconfiguring myself \"\n \"as a replica of %.40s\", sender->name);\n // 将 sender 设置为当前节点的主节点\n clusterSetMaster(sender);\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_FSYNC_CONFIG);\n } else if (dirty_slots_count) {\n /* If we are here, we received an update message which removed\n * ownership for certain slots we still have keys about, but still\n * we are serving some slots, so this master node was not demoted to\n * a slave.\n *\n * In order to maintain a consistent state between keys and slots\n * we need to remove all the keys from the slots we lost. */\n for (j = 0; j < dirty_slots_count; j++)\n delKeysInSlot(dirty_slots[j]);\n }\n}\n\n/* This function is called when this node is a master, and we receive from\n * another master a configuration epoch that is equal to our configuration\n * epoch.\n *\n * BACKGROUND\n *\n * It is not possible that different slaves get the same config\n * epoch during a failover election, because the slaves need to get voted\n * by a majority. However when we perform a manual resharding of the cluster\n * the node will assign a configuration epoch to itself without to ask\n * for agreement. Usually resharding happens when the cluster is working well\n * and is supervised by the sysadmin, however it is possible for a failover\n * to happen exactly while the node we are resharding a slot to assigns itself\n * a new configuration epoch, but before it is able to propagate it.\n *\n * So technically it is possible in this condition that two nodes end with\n * the same configuration epoch.\n *\n * Another possibility is that there are bugs in the implementation causing\n * this to happen.\n *\n * Moreover when a new cluster is created, all the nodes start with the same\n * configEpoch. This collision resolution code allows nodes to automatically\n * end with a different configEpoch at startup automatically.\n *\n * In all the cases, we want a mechanism that resolves this issue automatically\n * as a safeguard. The same configuration epoch for masters serving different\n * set of slots is not harmful, but it is if the nodes end serving the same\n * slots for some reason (manual errors or software bugs) without a proper\n * failover procedure.\n *\n * In general we want a system that eventually always ends with different\n * masters having different configuration epochs whatever happened, since\n * nothign is worse than a split-brain condition in a distributed system.\n *\n * BEHAVIOR\n *\n * When this function gets called, what happens is that if this node\n * has the lexicographically smaller Node ID compared to the other node\n * with the conflicting epoch (the 'sender' node), it will assign itself\n * the greatest configuration epoch currently detected among nodes plus 1.\n *\n * This means that even if there are multiple nodes colliding, the node\n * with the greatest Node ID never moves forward, so eventually all the nodes\n * end with a different configuration epoch.\n */\nvoid clusterHandleConfigEpochCollision(clusterNode *sender) {\n /* Prerequisites: nodes have the same configEpoch and are both masters. */\n if (sender->configEpoch != myself->configEpoch ||\n !nodeIsMaster(sender) || !nodeIsMaster(myself)) return;\n /* Don't act if the colliding node has a smaller Node ID. */\n if (memcmp(sender->name,myself->name,REDIS_CLUSTER_NAMELEN) <= 0) return;\n /* Get the next ID available at the best of this node knowledge. */\n server.cluster->currentEpoch++;\n myself->configEpoch = server.cluster->currentEpoch;\n clusterSaveConfigOrDie(1);\n redisLog(REDIS_VERBOSE,\n \"WARNING: configEpoch collision with node %.40s.\"\n \" Updating my configEpoch to %llu\",\n sender->name,\n (unsigned long long) myself->configEpoch);\n}\n\n/* When this function is called, there is a packet to process starting\n * at node->rcvbuf. Releasing the buffer is up to the caller, so this\n * function should just handle the higher level stuff of processing the\n * packet, modifying the cluster state if needed.\n *\n * 当这个函数被调用时，说明 node->rcvbuf 中有一条待处理的信息。\n * 信息处理完毕之后的释放工作由调用者处理，所以这个函数只需负责处理信息就可以了。\n *\n * The function returns 1 if the link is still valid after the packet\n * was processed, otherwise 0 if the link was freed since the packet\n * processing lead to some inconsistency error (for instance a PONG\n * received from the wrong sender ID). \n *\n * 如果函数返回 1 ，那么说明处理信息时没有遇到问题，连接依然可用。\n * 如果函数返回 0 ，那么说明信息处理时遇到了不一致问题\n * （比如接收到的 PONG 是发送自不正确的发送者 ID 的），连接已经被释放。\n */\nint clusterProcessPacket(clusterLink *link) {\n\n // 指向消息头\n clusterMsg *hdr = (clusterMsg*) link->rcvbuf;\n\n // 消息的长度\n uint32_t totlen = ntohl(hdr->totlen);\n\n // 消息的类型\n uint16_t type = ntohs(hdr->type);\n\n // 消息发送者的标识\n uint16_t flags = ntohs(hdr->flags);\n\n uint64_t senderCurrentEpoch = 0, senderConfigEpoch = 0;\n\n clusterNode *sender;\n\n // 更新接受消息计数器\n server.cluster->stats_bus_messages_received++;\n\n redisLog(REDIS_DEBUG,\"--- Processing packet of type %d, %lu bytes\",\n type, (unsigned long) totlen);\n\n /* Perform sanity checks */\n // 合法性检查\n if (totlen < 16) return 1; /* At least signature, version, totlen, count. */\n if (ntohs(hdr->ver) != 0) return 1; /* Can't handle versions other than 0.*/\n if (totlen > sdslen(link->rcvbuf)) return 1;\n if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG ||\n type == CLUSTERMSG_TYPE_MEET)\n {\n uint16_t count = ntohs(hdr->count);\n uint32_t explen; /* expected length of this packet */\n\n explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n explen += (sizeof(clusterMsgDataGossip)*count);\n if (totlen != explen) return 1;\n } else if (type == CLUSTERMSG_TYPE_FAIL) {\n uint32_t explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n\n explen += sizeof(clusterMsgDataFail);\n if (totlen != explen) return 1;\n } else if (type == CLUSTERMSG_TYPE_PUBLISH) {\n uint32_t explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n\n explen += sizeof(clusterMsgDataPublish) +\n ntohl(hdr->data.publish.msg.channel_len) +\n ntohl(hdr->data.publish.msg.message_len);\n if (totlen != explen) return 1;\n } else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST ||\n type == CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK ||\n type == CLUSTERMSG_TYPE_MFSTART)\n {\n uint32_t explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n\n if (totlen != explen) return 1;\n } else if (type == CLUSTERMSG_TYPE_UPDATE) {\n uint32_t explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n\n explen += sizeof(clusterMsgDataUpdate);\n if (totlen != explen) return 1;\n }\n\n /* Check if the sender is a known node. */\n // 查找发送者节点\n sender = clusterLookupNode(hdr->sender);\n // 节点存在，并且不是 HANDSHAKE 节点\n // 那么个更新节点的配置纪元信息\n if (sender && !nodeInHandshake(sender)) {\n /* Update our curretEpoch if we see a newer epoch in the cluster. */\n senderCurrentEpoch = ntohu64(hdr->currentEpoch);\n senderConfigEpoch = ntohu64(hdr->configEpoch);\n if (senderCurrentEpoch > server.cluster->currentEpoch)\n server.cluster->currentEpoch = senderCurrentEpoch;\n /* Update the sender configEpoch if it is publishing a newer one. */\n if (senderConfigEpoch > sender->configEpoch) {\n sender->configEpoch = senderConfigEpoch;\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_FSYNC_CONFIG);\n }\n /* Update the replication offset info for this node. */\n sender->repl_offset = ntohu64(hdr->offset);\n sender->repl_offset_time = mstime();\n /* If we are a slave performing a manual failover and our master\n * sent its offset while already paused, populate the MF state. */\n if (server.cluster->mf_end &&\n nodeIsSlave(myself) &&\n myself->slaveof == sender &&\n hdr->mflags[0] & CLUSTERMSG_FLAG0_PAUSED &&\n server.cluster->mf_master_offset == 0)\n {\n server.cluster->mf_master_offset = sender->repl_offset;\n redisLog(REDIS_WARNING,\n \"Received replication offset for paused \"\n \"master manual failover: %lld\",\n server.cluster->mf_master_offset);\n }\n }\n\n /* Process packets by type. */\n // 根据消息的类型，处理节点\n\n // 这是一条 PING 消息或者 MEET 消息\n if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_MEET) {\n redisLog(REDIS_DEBUG,\"Ping packet received: %p\", (void*)link->node);\n\n /* Add this node if it is new for us and the msg type is MEET.\n *\n * 如果当前节点是第一次遇见这个节点，并且对方发来的是 MEET 信息，\n * 那么将这个节点添加到集群的节点列表里面。\n *\n * In this stage we don't try to add the node with the right\n * flags, slaveof pointer, and so forth, as this details will be\n * resolved when we'll receive PONGs from the node. \n *\n * 节点目前的 flag 、 slaveof 等属性的值都是未设置的，\n * 等当前节点向对方发送 PING 命令之后，\n * 这些信息可以从对方回复的 PONG 信息中取得。\n */\n if (!sender && type == CLUSTERMSG_TYPE_MEET) {\n clusterNode *node;\n\n // 创建 HANDSHAKE 状态的新节点\n node = createClusterNode(NULL,REDIS_NODE_HANDSHAKE);\n\n // 设置 IP 和端口\n nodeIp2String(node->ip,link);\n node->port = ntohs(hdr->port);\n\n // 将新节点添加到集群\n clusterAddNode(node);\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG);\n }\n\n /* Get info from the gossip section */\n // 分析并取出消息中的 gossip 节点信息\n clusterProcessGossipSection(hdr,link);\n\n /* Anyway reply with a PONG */\n // 向目标节点返回一个 PONG\n clusterSendPing(link,CLUSTERMSG_TYPE_PONG);\n }\n\n /* PING or PONG: process config information. */\n // 这是一条 PING 、 PONG 或者 MEET 消息\n if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG ||\n type == CLUSTERMSG_TYPE_MEET)\n {\n redisLog(REDIS_DEBUG,\"%s packet received: %p\",\n type == CLUSTERMSG_TYPE_PING ? \"ping\" : \"pong\",\n (void*)link->node);\n\n // 连接的 clusterNode 结构存在\n if (link->node) {\n // 节点处于 HANDSHAKE 状态\n if (nodeInHandshake(link->node)) {\n /* If we already have this node, try to change the\n * IP/port of the node with the new one. */\n if (sender) {\n redisLog(REDIS_VERBOSE,\n \"Handshake: we already know node %.40s, \"\n \"updating the address if needed.\", sender->name);\n // 如果有需要的话，更新节点的地址\n if (nodeUpdateAddressIfNeeded(sender,link,ntohs(hdr->port)))\n {\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n }\n /* Free this node as we alrady have it. This will\n * cause the link to be freed as well. */\n // 释放节点\n freeClusterNode(link->node);\n return 0;\n }\n\n /* First thing to do is replacing the random name with the\n * right node name if this was a handshake stage. */\n // 用节点的真名替换在 HANDSHAKE 时创建的随机名字\n clusterRenameNode(link->node, hdr->sender);\n redisLog(REDIS_DEBUG,\"Handshake with node %.40s completed.\",\n link->node->name);\n\n // 关闭 HANDSHAKE 状态\n link->node->flags &= ~REDIS_NODE_HANDSHAKE;\n\n // 设置节点的角色\n link->node->flags |= flags&(REDIS_NODE_MASTER|REDIS_NODE_SLAVE);\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG);\n\n // 节点已存在，但它的 id 和当前节点保存的 id 不同\n } else if (memcmp(link->node->name,hdr->sender,\n REDIS_CLUSTER_NAMELEN) != 0)\n {\n /* If the reply has a non matching node ID we\n * disconnect this node and set it as not having an associated\n * address. */\n // 那么将这个节点设为 NOADDR \n // 并断开连接\n redisLog(REDIS_DEBUG,\"PONG contains mismatching sender ID\");\n link->node->flags |= REDIS_NODE_NOADDR;\n link->node->ip[0] = '\\0';\n link->node->port = 0;\n\n // 断开连接\n freeClusterLink(link);\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG);\n return 0;\n }\n }\n\n /* Update the node address if it changed. */\n // 如果发送的消息为 PING \n // 并且发送者不在 HANDSHAKE 状态\n // 那么更新发送者的信息\n if (sender && type == CLUSTERMSG_TYPE_PING &&\n !nodeInHandshake(sender) &&\n nodeUpdateAddressIfNeeded(sender,link,ntohs(hdr->port)))\n {\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n }\n\n /* Update our info about the node */\n // 如果这是一条 PONG 消息，那么更新我们关于 node 节点的认识\n if (link->node && type == CLUSTERMSG_TYPE_PONG) {\n\n // 最后一次接到该节点的 PONG 的时间\n link->node->pong_received = mstime();\n\n // 清零最近一次等待 PING 命令的时间\n link->node->ping_sent = 0;\n\n /* The PFAIL condition can be reversed without external\n * help if it is momentary (that is, if it does not\n * turn into a FAIL state).\n *\n * 接到节点的 PONG 回复，我们可以移除节点的 PFAIL 状态。\n *\n * The FAIL condition is also reversible under specific\n * conditions detected by clearNodeFailureIfNeeded(). \n *\n * 如果节点的状态为 FAIL ，\n * 那么是否撤销该状态要根据 clearNodeFailureIfNeeded() 函数来决定。\n */\n if (nodeTimedOut(link->node)) {\n // 撤销 PFAIL\n link->node->flags &= ~REDIS_NODE_PFAIL;\n\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n } else if (nodeFailed(link->node)) {\n // 看是否可以撤销 FAIL\n clearNodeFailureIfNeeded(link->node);\n }\n }\n\n /* Check for role switch: slave -> master or master -> slave. */\n // 检测节点的身份信息，并在需要时进行更新\n if (sender) {\n\n // 发送消息的节点的 slaveof 为 REDIS_NODE_NULL_NAME\n // 那么 sender 就是一个主节点\n if (!memcmp(hdr->slaveof,REDIS_NODE_NULL_NAME,\n sizeof(hdr->slaveof)))\n {\n /* Node is a master. */\n // 设置 sender 为主节点\n clusterSetNodeAsMaster(sender);\n\n // sender 的 slaveof 不为空，那么这是一个从节点\n } else {\n\n /* Node is a slave. */\n // 取出 sender 的主节点\n clusterNode *master = clusterLookupNode(hdr->slaveof);\n\n // sender 由主节点变成了从节点，重新配置 sender\n if (nodeIsMaster(sender)) {\n /* Master turned into a slave! Reconfigure the node. */\n\n // 删除所有由该节点负责的槽\n clusterDelNodeSlots(sender);\n\n // 更新标识\n sender->flags &= ~REDIS_NODE_MASTER;\n sender->flags |= REDIS_NODE_SLAVE;\n\n /* Remove the list of slaves from the node. */\n // 移除 sender 的从节点名单\n if (sender->numslaves) clusterNodeResetSlaves(sender);\n\n /* Update config and state. */\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n }\n\n /* Master node changed for this slave? */\n\n // 检查 sender 的主节点是否变更\n if (master && sender->slaveof != master) {\n // 如果 sender 之前的主节点不是现在的主节点\n // 那么在旧主节点的从节点列表中移除 sender\n if (sender->slaveof)\n clusterNodeRemoveSlave(sender->slaveof,sender);\n\n // 并在新主节点的从节点列表中添加 sender\n clusterNodeAddSlave(master,sender);\n\n // 更新 sender 的主节点\n sender->slaveof = master;\n\n /* Update config. */\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG);\n }\n }\n }\n\n /* Update our info about served slots.\n *\n * 更新当前节点对 sender 所处理槽的认识。\n *\n * Note: this MUST happen after we update the master/slave state\n * so that REDIS_NODE_MASTER flag will be set. \n *\n * 这部分的更新 *必须* 在更新 sender 的主/从节点信息之后，\n * 因为这里需要用到 REDIS_NODE_MASTER 标识。\n */\n\n /* Many checks are only needed if the set of served slots this\n * instance claims is different compared to the set of slots we have\n * for it. Check this ASAP to avoid other computational expansive\n * checks later. */\n clusterNode *sender_master = NULL; /* Sender or its master if slave. */\n int dirty_slots = 0; /* Sender claimed slots don't match my view? */\n\n if (sender) {\n sender_master = nodeIsMaster(sender) ? sender : sender->slaveof;\n if (sender_master) {\n dirty_slots = memcmp(sender_master->slots,\n hdr->myslots,sizeof(hdr->myslots)) != 0;\n }\n }\n\n /* 1) If the sender of the message is a master, and we detected that\n * the set of slots it claims changed, scan the slots to see if we\n * need to update our configuration. */\n // 如果 sender 是主节点，并且 sender 的槽布局出现了变动\n // 那么检查当前节点对 sender 的槽布局设置，看是否需要进行更新\n if (sender && nodeIsMaster(sender) && dirty_slots)\n clusterUpdateSlotsConfigWith(sender,senderConfigEpoch,hdr->myslots);\n\n /* 2) We also check for the reverse condition, that is, the sender\n * claims to serve slots we know are served by a master with a\n * greater configEpoch. If this happens we inform the sender.\n *\n * 检测和条件 1 的相反条件，也即是，\n * sender 处理的槽的配置纪元比当前节点已知的某个节点的配置纪元要低，\n * 如果是这样的话，通知 sender 。\n *\n * This is useful because sometimes after a partition heals, a\n * reappearing master may be the last one to claim a given set of\n * hash slots, but with a configuration that other instances know to\n * be deprecated. Example:\n *\n * 这种情况可能会出现在网络分裂中，\n * 一个重新上线的主节点可能会带有已经过时的槽布局。\n *\n * 比如说：\n *\n * A and B are master and slave for slots 1,2,3.\n * A 负责槽 1 、 2 、 3 ，而 B 是 A 的从节点。\n *\n * A is partitioned away, B gets promoted.\n * A 从网络中分裂出去，B 被提升为主节点。\n *\n * B is partitioned away, and A returns available.\n * B 从网络中分裂出去， A 重新上线（但是它所使用的槽布局是旧的）。\n *\n * Usually B would PING A publishing its set of served slots and its\n * configEpoch, but because of the partition B can't inform A of the\n * new configuration, so other nodes that have an updated table must\n * do it. In this way A will stop to act as a master (or can try to\n * failover if there are the conditions to win the election).\n *\n * 在正常情况下， B 应该向 A 发送 PING 消息，告知 A ，自己（B）已经接替了\n * 槽 1、 2、 3 ，并且带有更更的配置纪元，但因为网络分裂的缘故，\n * 节点 B 没办法通知节点 A ，\n * 所以通知节点 A 它带有的槽布局已经更新的工作就交给其他知道 B 带有更高配置纪元的节点来做。\n * 当 A 接到其他节点关于节点 B 的消息时，\n * 节点 A 就会停止自己的主节点工作，又或者重新进行故障转移。\n */\n if (sender && dirty_slots) {\n int j;\n\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n\n // 检测 slots 中的槽 j 是否已经被指派\n if (bitmapTestBit(hdr->myslots,j)) {\n\n // 当前节点认为槽 j 由 sender 负责处理，\n // 或者当前节点认为该槽未指派，那么跳过该槽\n if (server.cluster->slots[j] == sender ||\n server.cluster->slots[j] == NULL) continue;\n\n // 当前节点槽 j 的配置纪元比 sender 的配置纪元要大\n if (server.cluster->slots[j]->configEpoch >\n senderConfigEpoch)\n {\n redisLog(REDIS_VERBOSE,\n \"Node %.40s has old slots configuration, sending \"\n \"an UPDATE message about %.40s\",\n sender->name, server.cluster->slots[j]->name);\n\n // 向 sender 发送关于槽 j 的更新信息\n clusterSendUpdate(sender->link,\n server.cluster->slots[j]);\n\n /* TODO: instead of exiting the loop send every other\n * UPDATE packet for other nodes that are the new owner\n * of sender's slots. */\n break;\n }\n }\n }\n }\n\n /* If our config epoch collides with the sender's try to fix\n * the problem. */\n if (sender &&\n nodeIsMaster(myself) && nodeIsMaster(sender) &&\n senderConfigEpoch == myself->configEpoch)\n {\n clusterHandleConfigEpochCollision(sender);\n }\n\n /* Get info from the gossip section */\n // 分析并提取出消息 gossip 协议部分的信息\n clusterProcessGossipSection(hdr,link);\n\n // 这是一条 FAIL 消息： sender 告知当前节点，某个节点已经进入 FAIL 状态。\n } else if (type == CLUSTERMSG_TYPE_FAIL) {\n clusterNode *failing;\n\n if (sender) {\n\n // 获取下线节点的消息\n failing = clusterLookupNode(hdr->data.fail.about.nodename);\n // 下线的节点既不是当前节点，也没有处于 FAIL 状态\n if (failing &&\n !(failing->flags & (REDIS_NODE_FAIL|REDIS_NODE_MYSELF)))\n {\n redisLog(REDIS_NOTICE,\n \"FAIL message received from %.40s about %.40s\",\n hdr->sender, hdr->data.fail.about.nodename);\n\n // 打开 FAIL 状态\n failing->flags |= REDIS_NODE_FAIL;\n failing->fail_time = mstime();\n // 关闭 PFAIL 状态\n failing->flags &= ~REDIS_NODE_PFAIL;\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE);\n }\n } else {\n redisLog(REDIS_NOTICE,\n \"Ignoring FAIL message from unknonw node %.40s about %.40s\",\n hdr->sender, hdr->data.fail.about.nodename);\n }\n\n // 这是一条 PUBLISH 消息\n } else if (type == CLUSTERMSG_TYPE_PUBLISH) {\n robj *channel, *message;\n uint32_t channel_len, message_len;\n\n /* Don't bother creating useless objects if there are no\n * Pub/Sub subscribers. */\n // 只在有订阅者时创建消息对象\n if (dictSize(server.pubsub_channels) ||\n listLength(server.pubsub_patterns))\n {\n // 频道长度\n channel_len = ntohl(hdr->data.publish.msg.channel_len);\n\n // 消息长度\n message_len = ntohl(hdr->data.publish.msg.message_len);\n\n // 频道\n channel = createStringObject(\n (char*)hdr->data.publish.msg.bulk_data,channel_len);\n\n // 消息\n message = createStringObject(\n (char*)hdr->data.publish.msg.bulk_data+channel_len,\n message_len);\n // 发送消息\n pubsubPublishMessage(channel,message);\n\n decrRefCount(channel);\n decrRefCount(message);\n }\n\n // 这是一条请求获得故障迁移授权的消息： sender 请求当前节点为它进行故障转移投票\n } else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST) {\n if (!sender) return 1; /* We don't know that node. */\n // 如果条件允许的话，向 sender 投票，支持它进行故障转移\n clusterSendFailoverAuthIfNeeded(sender,hdr);\n\n // 这是一条故障迁移投票信息： sender 支持当前节点执行故障转移操作\n } else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK) {\n if (!sender) return 1; /* We don't know that node. */\n\n /* We consider this vote only if the sender is a master serving\n * a non zero number of slots, and its currentEpoch is greater or\n * equal to epoch where this node started the election. */\n // 只有正在处理至少一个槽的主节点的投票会被视为是有效投票\n // 只有符合以下条件， sender 的投票才算有效：\n // 1） sender 是主节点\n // 2） sender 正在处理至少一个槽\n // 3） sender 的配置纪元大于等于当前节点的配置纪元\n if (nodeIsMaster(sender) && sender->numslots > 0 &&\n senderCurrentEpoch >= server.cluster->failover_auth_epoch)\n {\n // 增加支持票数\n server.cluster->failover_auth_count++;\n\n /* Maybe we reached a quorum here, set a flag to make sure\n * we check ASAP. */\n clusterDoBeforeSleep(CLUSTER_TODO_HANDLE_FAILOVER);\n }\n\n } else if (type == CLUSTERMSG_TYPE_MFSTART) {\n /* This message is acceptable only if I'm a master and the sender\n * is one of my slaves. */\n if (!sender || sender->slaveof != myself) return 1;\n /* Manual failover requested from slaves. Initialize the state\n * accordingly. */\n resetManualFailover();\n server.cluster->mf_end = mstime() + REDIS_CLUSTER_MF_TIMEOUT;\n server.cluster->mf_slave = sender;\n pauseClients(mstime()+(REDIS_CLUSTER_MF_TIMEOUT*2));\n redisLog(REDIS_WARNING,\"Manual failover requested by slave %.40s.\",\n sender->name);\n } else if (type == CLUSTERMSG_TYPE_UPDATE) {\n clusterNode *n; /* The node the update is about. */\n uint64_t reportedConfigEpoch =\n ntohu64(hdr->data.update.nodecfg.configEpoch);\n\n if (!sender) return 1; /* We don't know the sender. */\n\n // 获取需要更新的节点\n n = clusterLookupNode(hdr->data.update.nodecfg.nodename);\n if (!n) return 1; /* We don't know the reported node. */\n\n // 消息的纪元并不大于节点 n 所处的配置纪元\n // 无须更新\n if (n->configEpoch >= reportedConfigEpoch) return 1; /* Nothing new. */\n\n /* If in our current config the node is a slave, set it as a master. */\n // 如果节点 n 为从节点，但它的槽配置更新了\n // 那么说明这个节点已经变为主节点，将它设置为主节点\n if (nodeIsSlave(n)) clusterSetNodeAsMaster(n);\n\n /* Update the node's configEpoch. */\n n->configEpoch = reportedConfigEpoch;\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_FSYNC_CONFIG);\n\n /* Check the bitmap of served slots and udpate our\n * config accordingly. */\n // 将消息中对 n 的槽布局与当前节点对 n 的槽布局进行对比\n // 在有需要时更新当前节点对 n 的槽布局的认识\n clusterUpdateSlotsConfigWith(n,reportedConfigEpoch,\n hdr->data.update.nodecfg.slots);\n } else {\n redisLog(REDIS_WARNING,\"Received unknown packet type: %d\", type);\n }\n return 1;\n}\n\n/* This function is called when we detect the link with this node is lost.\n\n 这个函数在发现节点的连接已经丢失时使用。\n\n We set the node as no longer connected. The Cluster Cron will detect\n this connection and will try to get it connected again.\n\n 我们将节点的状态设置为断开状态，Cluster Cron 会根据该状态尝试重新连接节点。\n\n Instead if the node is a temporary node used to accept a query, we\n completely free the node on error. \n\n 如果连接是一个临时连接的话，那么它就会被永久释放，不再进行重连。\n\n */\nvoid handleLinkIOError(clusterLink *link) {\n freeClusterLink(link);\n}\n\n/* Send data. This is handled using a trivial send buffer that gets\n * consumed by write(). We don't try to optimize this for speed too much\n * as this is a very low traffic channel. \n *\n * 写事件处理器，用于向集群节点发送信息。\n */\nvoid clusterWriteHandler(aeEventLoop *el, int fd, void *privdata, int mask) {\n clusterLink *link = (clusterLink*) privdata;\n ssize_t nwritten;\n REDIS_NOTUSED(el);\n REDIS_NOTUSED(mask);\n\n // 写入信息\n nwritten = write(fd, link->sndbuf, sdslen(link->sndbuf));\n\n // 写入错误\n if (nwritten <= 0) {\n redisLog(REDIS_DEBUG,\"I/O error writing to node link: %s\",\n strerror(errno));\n handleLinkIOError(link);\n return;\n }\n\n // 删除已写入的部分\n sdsrange(link->sndbuf,nwritten,-1);\n\n // 如果所有当前节点输出缓冲区里面的所有内容都已经写入完毕\n // （缓冲区为空）\n // 那么删除写事件处理器\n if (sdslen(link->sndbuf) == 0)\n aeDeleteFileEvent(server.el, link->fd, AE_WRITABLE);\n}\n\n/* Read data. Try to read the first field of the header first to check the\n * full length of the packet. When a whole packet is in memory this function\n * will call the function to process the packet. And so forth. */\n// 读事件处理器\n// 首先读入内容的头，以判断读入内容的长度\n// 如果内容是一个 whole packet ，那么调用函数来处理这个 packet 。\nvoid clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) {\n char buf[sizeof(clusterMsg)];\n ssize_t nread;\n clusterMsg *hdr;\n clusterLink *link = (clusterLink*) privdata;\n int readlen, rcvbuflen;\n REDIS_NOTUSED(el);\n REDIS_NOTUSED(mask);\n\n // 尽可能地多读数据\n while(1) { /* Read as long as there is data to read. */\n\n // 检查输入缓冲区的长度\n rcvbuflen = sdslen(link->rcvbuf);\n // 头信息（8 字节）未读入完\n if (rcvbuflen < 8) {\n /* First, obtain the first 8 bytes to get the full message\n * length. */\n readlen = 8 - rcvbuflen;\n // 已读入完整的信息\n } else {\n /* Finally read the full message. */\n hdr = (clusterMsg*) link->rcvbuf;\n if (rcvbuflen == 8) {\n /* Perform some sanity check on the message signature\n * and length. */\n if (memcmp(hdr->sig,\"RCmb\",4) != 0 ||\n ntohl(hdr->totlen) < CLUSTERMSG_MIN_LEN)\n {\n redisLog(REDIS_WARNING,\n \"Bad message length or signature received \"\n \"from Cluster bus.\");\n handleLinkIOError(link);\n return;\n }\n }\n // 记录已读入内容长度\n readlen = ntohl(hdr->totlen) - rcvbuflen;\n if (readlen > sizeof(buf)) readlen = sizeof(buf);\n }\n\n // 读入内容\n nread = read(fd,buf,readlen);\n\n // 没有内容可读\n if (nread == -1 && errno == EAGAIN) return; /* No more data ready. */\n\n // 处理读入错误\n if (nread <= 0) {\n /* I/O error... */\n redisLog(REDIS_DEBUG,\"I/O error reading from node link: %s\",\n (nread == 0) ? \"connection closed\" : strerror(errno));\n handleLinkIOError(link);\n return;\n } else {\n /* Read data and recast the pointer to the new buffer. */\n // 将读入的内容追加进输入缓冲区里面\n link->rcvbuf = sdscatlen(link->rcvbuf,buf,nread);\n hdr = (clusterMsg*) link->rcvbuf;\n rcvbuflen += nread;\n }\n\n /* Total length obtained? Process this packet. */\n // 检查已读入内容的长度，看是否整条信息已经被读入了\n if (rcvbuflen >= 8 && rcvbuflen == ntohl(hdr->totlen)) {\n // 如果是的话，执行处理信息的函数\n if (clusterProcessPacket(link)) {\n sdsfree(link->rcvbuf);\n link->rcvbuf = sdsempty();\n } else {\n return; /* Link no longer valid. */\n }\n }\n }\n}\n\n/* Put stuff into the send buffer.\n *\n * 发送信息\n *\n * It is guaranteed that this function will never have as a side effect\n * the link to be invalidated, so it is safe to call this function\n * from event handlers that will do stuff with the same link later. \n *\n * 因为发送不会对连接本身造成不良的副作用，\n * 所以可以在发送信息的处理器上做一些针对连接本身的动作。\n */\nvoid clusterSendMessage(clusterLink *link, unsigned char *msg, size_t msglen) {\n\n // 安装写事件处理器\n if (sdslen(link->sndbuf) == 0 && msglen != 0)\n aeCreateFileEvent(server.el,link->fd,AE_WRITABLE,\n clusterWriteHandler,link);\n\n // 将信息追加到输出缓冲区\n link->sndbuf = sdscatlen(link->sndbuf, msg, msglen);\n\n // 增一发送信息计数\n server.cluster->stats_bus_messages_sent++;\n}\n\n/* Send a message to all the nodes that are part of the cluster having\n * a connected link.\n *\n * 向节点连接的所有其他节点发送信息。\n *\n * It is guaranteed that this function will never have as a side effect\n * some node->link to be invalidated, so it is safe to call this function\n * from event handlers that will do stuff with node links later. */\nvoid clusterBroadcastMessage(void *buf, size_t len) {\n dictIterator *di;\n dictEntry *de;\n\n // 遍历所有已知节点\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n // 不向未连接节点发送信息\n if (!node->link) continue;\n\n // 不向节点自身或者 HANDSHAKE 状态的节点发送信息\n if (node->flags & (REDIS_NODE_MYSELF|REDIS_NODE_HANDSHAKE))\n continue;\n\n // 发送信息\n clusterSendMessage(node->link,buf,len);\n }\n dictReleaseIterator(di);\n}\n\n/* Build the message header */\n// 构建信息\nvoid clusterBuildMessageHdr(clusterMsg *hdr, int type) {\n int totlen = 0;\n uint64_t offset;\n clusterNode *master;\n\n /* If this node is a master, we send its slots bitmap and configEpoch.\n *\n * 如果这是一个主节点，那么发送该节点的槽 bitmap 和配置纪元。\n *\n * If this node is a slave we send the master's information instead (the\n * node is flagged as slave so the receiver knows that it is NOT really\n * in charge for this slots.\n * 如果这是一个从节点，\n * 那么发送这个节点的主节点的槽 bitmap 和配置纪元。\n *\n * 因为接收信息的节点通过标识可以知道这个节点是一个从节点，\n * 所以接收信息的节点不会将从节点错认作是主节点。\n */\n master = (nodeIsSlave(myself) && myself->slaveof) ?\n myself->slaveof : myself;\n\n // 清零信息头\n memset(hdr,0,sizeof(*hdr));\n\n hdr->sig[0] = 'R';\n hdr->sig[1] = 'C';\n hdr->sig[2] = 'm';\n hdr->sig[3] = 'b';\n\n // 设置信息类型\n hdr->type = htons(type);\n\n // 设置信息发送者\n memcpy(hdr->sender,myself->name,REDIS_CLUSTER_NAMELEN);\n\n // 设置当前节点负责的槽\n memcpy(hdr->myslots,master->slots,sizeof(hdr->myslots));\n\n // 清零 slaveof 域\n memset(hdr->slaveof,0,REDIS_CLUSTER_NAMELEN);\n\n // 如果节点是从节点的话，那么设置 slaveof 域\n if (myself->slaveof != NULL)\n memcpy(hdr->slaveof,myself->slaveof->name, REDIS_CLUSTER_NAMELEN);\n\n // 设置端口号\n hdr->port = htons(server.port);\n\n // 设置标识\n hdr->flags = htons(myself->flags);\n\n // 设置状态\n hdr->state = server.cluster->state;\n\n /* Set the currentEpoch and configEpochs. */\n // 设置集群当前配置纪元\n hdr->currentEpoch = htonu64(server.cluster->currentEpoch);\n // 设置主节点当前配置纪元\n hdr->configEpoch = htonu64(master->configEpoch);\n\n /* Set the replication offset. */\n // 设置复制偏移量\n if (nodeIsSlave(myself))\n offset = replicationGetSlaveOffset();\n else\n offset = server.master_repl_offset;\n hdr->offset = htonu64(offset);\n\n /* Set the message flags. */\n if (nodeIsMaster(myself) && server.cluster->mf_end)\n hdr->mflags[0] |= CLUSTERMSG_FLAG0_PAUSED;\n\n /* Compute the message length for certain messages. For other messages\n * this is up to the caller. */\n // 计算信息的长度\n if (type == CLUSTERMSG_TYPE_FAIL) {\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n totlen += sizeof(clusterMsgDataFail);\n } else if (type == CLUSTERMSG_TYPE_UPDATE) {\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n totlen += sizeof(clusterMsgDataUpdate);\n }\n\n // 设置信息的长度\n hdr->totlen = htonl(totlen);\n /* For PING, PONG, and MEET, fixing the totlen field is up to the caller. */\n}\n\n/* Send a PING or PONG packet to the specified node, making sure to add enough\n * gossip informations. */\n// 向指定节点发送一条 MEET 、 PING 或者 PONG 消息\nvoid clusterSendPing(clusterLink *link, int type) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n int gossipcount = 0, totlen;\n /* freshnodes is the number of nodes we can still use to populate the\n * gossip section of the ping packet. Basically we start with the nodes\n * we have in memory minus two (ourself and the node we are sending the\n * message to). Every time we add a node we decrement the counter, so when\n * it will drop to <= zero we know there is no more gossip info we can\n * send. */\n // freshnodes 是用于发送 gossip 信息的计数器\n // 每次发送一条信息时，程序将 freshnodes 的值减一\n // 当 freshnodes 的数值小于等于 0 时，程序停止发送 gossip 信息\n // freshnodes 的数量是节点目前的 nodes 表中的节点数量减去 2 \n // 这里的 2 指两个节点，一个是 myself 节点（也即是发送信息的这个节点）\n // 另一个是接受 gossip 信息的节点\n int freshnodes = dictSize(server.cluster->nodes)-2;\n\n // 如果发送的信息是 PING ，那么更新最后一次发送 PING 命令的时间戳\n if (link->node && type == CLUSTERMSG_TYPE_PING)\n link->node->ping_sent = mstime();\n\n // 将当前节点的信息（比如名字、地址、端口号、负责处理的槽）记录到消息里面\n clusterBuildMessageHdr(hdr,type);\n\n /* Populate the gossip fields */\n // 从当前节点已知的节点中随机选出两个节点\n // 并通过这条消息捎带给目标节点，从而实现 gossip 协议\n\n // 每个节点有 freshnodes 次发送 gossip 信息的机会\n // 每次向目标节点发送 2 个被选中节点的 gossip 信息（gossipcount 计数）\n while(freshnodes > 0 && gossipcount < 3) {\n // 从 nodes 字典中随机选出一个节点（被选中节点）\n dictEntry *de = dictGetRandomKey(server.cluster->nodes);\n clusterNode *this = dictGetVal(de);\n\n clusterMsgDataGossip *gossip;\n int j;\n\n /* In the gossip section don't include:\n * 以下节点不能作为被选中节点：\n * 1) Myself.\n * 节点本身。\n * 2) Nodes in HANDSHAKE state.\n * 处于 HANDSHAKE 状态的节点。\n * 3) Nodes with the NOADDR flag set.\n * 带有 NOADDR 标识的节点\n * 4) Disconnected nodes if they don't have configured slots.\n * 因为不处理任何槽而被断开连接的节点 \n */\n if (this == myself ||\n this->flags & (REDIS_NODE_HANDSHAKE|REDIS_NODE_NOADDR) ||\n (this->link == NULL && this->numslots == 0))\n {\n freshnodes--; /* otherwise we may loop forever. */\n continue;\n }\n\n /* Check if we already added this node */\n // 检查被选中节点是否已经在 hdr->data.ping.gossip 数组里面\n // 如果是的话说明这个节点之前已经被选中了\n // 不要再选中它（否则就会出现重复）\n for (j = 0; j < gossipcount; j++) {\n if (memcmp(hdr->data.ping.gossip[j].nodename,this->name,\n REDIS_CLUSTER_NAMELEN) == 0) break;\n }\n if (j != gossipcount) continue;\n\n /* Add it */\n\n // 这个被选中节点有效，计数器减一\n freshnodes--;\n\n // 指向 gossip 信息结构\n gossip = &(hdr->data.ping.gossip[gossipcount]);\n\n // 将被选中节点的名字记录到 gossip 信息\n memcpy(gossip->nodename,this->name,REDIS_CLUSTER_NAMELEN);\n // 将被选中节点的 PING 命令发送时间戳记录到 gossip 信息\n gossip->ping_sent = htonl(this->ping_sent);\n // 将被选中节点的 PING 命令回复的时间戳记录到 gossip 信息\n gossip->pong_received = htonl(this->pong_received);\n // 将被选中节点的 IP 记录到 gossip 信息\n memcpy(gossip->ip,this->ip,sizeof(this->ip));\n // 将被选中节点的端口号记录到 gossip 信息\n gossip->port = htons(this->port);\n // 将被选中节点的标识值记录到 gossip 信息\n gossip->flags = htons(this->flags);\n\n // 这个被选中节点有效，计数器增一\n gossipcount++;\n }\n\n // 计算信息长度\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n totlen += (sizeof(clusterMsgDataGossip)*gossipcount);\n // 将被选中节点的数量（gossip 信息中包含了多少个节点的信息）\n // 记录在 count 属性里面\n hdr->count = htons(gossipcount);\n // 将信息的长度记录到信息里面\n hdr->totlen = htonl(totlen);\n\n // 发送信息\n clusterSendMessage(link,buf,totlen);\n}\n\n/* Send a PONG packet to every connected node that's not in handshake state\n * and for which we have a valid link.\n *\n * 向所有未在 HANDSHAKE 状态，并且连接正常的节点发送 PONG 回复。\n *\n * In Redis Cluster pongs are not used just for failure detection, but also\n * to carry important configuration information. So broadcasting a pong is\n * useful when something changes in the configuration and we want to make\n * the cluster aware ASAP (for instance after a slave promotion). *\n * 在集群中， PONG 不仅可以用来检测节点状态，\n * 还可以携带一些重要的信息。\n *\n * 因此广播 PONG 回复在配置发生变化（比如从节点转变为主节点），\n * 并且当前节点想让其他节点尽快知悉这一变化的时候，\n * 就会广播 PONG 回复。\n *\n * The 'target' argument specifies the receiving instances using the\n * defines below:\n *\n * CLUSTER_BROADCAST_ALL -> All known instances.\n * CLUSTER_BROADCAST_LOCAL_SLAVES -> All slaves in my master-slaves ring.\n */\n#define CLUSTER_BROADCAST_ALL 0\n#define CLUSTER_BROADCAST_LOCAL_SLAVES 1\nvoid clusterBroadcastPong(int target) {\n dictIterator *di;\n dictEntry *de;\n\n // 遍历所有节点\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n // 不向未建立连接的节点发送\n if (!node->link) continue;\n if (node == myself || nodeInHandshake(node)) continue;\n if (target == CLUSTER_BROADCAST_LOCAL_SLAVES) {\n int local_slave =\n nodeIsSlave(node) && node->slaveof &&\n (node->slaveof == myself || node->slaveof == myself->slaveof);\n if (!local_slave) continue;\n }\n // 发送 PONG 信息\n clusterSendPing(node->link,CLUSTERMSG_TYPE_PONG);\n }\n dictReleaseIterator(di);\n}\n\n/* Send a PUBLISH message.\n *\n * 发送一条 PUBLISH 消息。\n *\n * If link is NULL, then the message is broadcasted to the whole cluster. \n *\n * 如果 link 参数为 NULL ，那么将消息广播给整个集群。\n */\nvoid clusterSendPublish(clusterLink *link, robj *channel, robj *message) {\n unsigned char buf[sizeof(clusterMsg)], *payload;\n clusterMsg *hdr = (clusterMsg*) buf;\n uint32_t totlen;\n uint32_t channel_len, message_len;\n\n // 频道\n channel = getDecodedObject(channel);\n\n // 消息\n message = getDecodedObject(message);\n\n // 频道和消息的长度\n channel_len = sdslen(channel->ptr);\n message_len = sdslen(message->ptr);\n\n // 构建消息\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_PUBLISH);\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n totlen += sizeof(clusterMsgDataPublish) + channel_len + message_len;\n\n hdr->data.publish.msg.channel_len = htonl(channel_len);\n hdr->data.publish.msg.message_len = htonl(message_len);\n hdr->totlen = htonl(totlen);\n\n /* Try to use the local buffer if possible */\n if (totlen < sizeof(buf)) {\n payload = buf;\n } else {\n payload = zmalloc(totlen);\n memcpy(payload,hdr,sizeof(*hdr));\n hdr = (clusterMsg*) payload;\n }\n\n // 保存频道和消息到消息结构中\n memcpy(hdr->data.publish.msg.bulk_data,channel->ptr,sdslen(channel->ptr));\n memcpy(hdr->data.publish.msg.bulk_data+sdslen(channel->ptr),\n message->ptr,sdslen(message->ptr));\n\n // 选择发送到节点还是广播至整个集群\n if (link)\n clusterSendMessage(link,payload,totlen);\n else\n clusterBroadcastMessage(payload,totlen);\n\n decrRefCount(channel);\n decrRefCount(message);\n if (payload != buf) zfree(payload);\n}\n\n/* Send a FAIL message to all the nodes we are able to contact.\n *\n * 向当前节点已知的所有节点发送 FAIL 信息。\n *\n * The FAIL message is sent when we detect that a node is failing\n * (REDIS_NODE_PFAIL) and we also receive a gossip confirmation of this:\n * we switch the node state to REDIS_NODE_FAIL and ask all the other\n * nodes to do the same ASAP. \n *\n * 如果当前节点将 node 标记为 PFAIL 状态，\n * 并且通过 gossip 协议，\n * 从足够数量的节点那些得到了 node 已经下线的支持， \n * 那么当前节点会将 node 标记为 FAIL ，\n * 并执行这个函数，向其他 node 发送 FAIL 消息，\n * 要求它们也将 node 标记为 FAIL 。\n */\nvoid clusterSendFail(char *nodename) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n\n // 创建下线消息\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_FAIL);\n\n // 记录命令\n memcpy(hdr->data.fail.about.nodename,nodename,REDIS_CLUSTER_NAMELEN);\n\n // 广播消息\n clusterBroadcastMessage(buf,ntohl(hdr->totlen));\n}\n\n/* Send an UPDATE message to the specified link carrying the specified 'node'\n * slots configuration. The node name, slots bitmap, and configEpoch info\n * are included. \n *\n * 向连接 link 发送包含给定 node 槽配置的 UPDATE 消息，\n * 包括节点名称，槽位图，以及配置纪元。\n */\nvoid clusterSendUpdate(clusterLink *link, clusterNode *node) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n\n if (link == NULL) return;\n\n // 创建消息\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_UPDATE);\n\n // 设置节点名\n memcpy(hdr->data.update.nodecfg.nodename,node->name,REDIS_CLUSTER_NAMELEN);\n\n // 设置配置纪元\n hdr->data.update.nodecfg.configEpoch = htonu64(node->configEpoch);\n\n // 更新节点的槽位图\n memcpy(hdr->data.update.nodecfg.slots,node->slots,sizeof(node->slots));\n\n // 发送信息\n clusterSendMessage(link,buf,ntohl(hdr->totlen));\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER Pub/Sub support\n *\n * For now we do very little, just propagating PUBLISH messages across the whole\n * cluster. In the future we'll try to get smarter and avoiding propagating those\n * messages to hosts without receives for a given channel.\n * -------------------------------------------------------------------------- */\n// 向整个集群的 channel 频道中广播消息 messages\nvoid clusterPropagatePublish(robj *channel, robj *message) {\n clusterSendPublish(NULL, channel, message);\n}\n\n/* -----------------------------------------------------------------------------\n * SLAVE node specific functions\n * -------------------------------------------------------------------------- */\n\n/* This function sends a FAILOVE_AUTH_REQUEST message to every node in order to\n * see if there is the quorum for this slave instance to failover its failing\n * master.\n *\n * 向其他所有节点发送 FAILOVE_AUTH_REQUEST 信息，\n * 看它们是否同意由这个从节点来对下线的主节点进行故障转移。\n *\n * Note that we send the failover request to everybody, master and slave nodes,\n * but only the masters are supposed to reply to our query. \n *\n * 信息会被发送给所有节点，包括主节点和从节点，但只有主节点会回复这条信息。 \n */\nvoid clusterRequestFailoverAuth(void) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n uint32_t totlen;\n\n // 设置信息头（包含当前节点的信息）\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST);\n /* If this is a manual failover, set the CLUSTERMSG_FLAG0_FORCEACK bit\n * in the header to communicate the nodes receiving the message that\n * they should authorized the failover even if the master is working. */\n if (server.cluster->mf_end) hdr->mflags[0] |= CLUSTERMSG_FLAG0_FORCEACK;\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n hdr->totlen = htonl(totlen);\n\n // 发送信息\n clusterBroadcastMessage(buf,totlen);\n}\n\n/* Send a FAILOVER_AUTH_ACK message to the specified node. */\n// 向节点 node 投票，支持它进行故障迁移\nvoid clusterSendFailoverAuth(clusterNode *node) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n uint32_t totlen;\n\n if (!node->link) return;\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK);\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n hdr->totlen = htonl(totlen);\n clusterSendMessage(node->link,buf,totlen);\n}\n\n/* Send a MFSTART message to the specified node. */\n// 向给定的节点发送一条 MFSTART 消息\nvoid clusterSendMFStart(clusterNode *node) {\n unsigned char buf[sizeof(clusterMsg)];\n clusterMsg *hdr = (clusterMsg*) buf;\n uint32_t totlen;\n\n if (!node->link) return;\n clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_MFSTART);\n totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);\n hdr->totlen = htonl(totlen);\n clusterSendMessage(node->link,buf,totlen);\n}\n\n/* Vote for the node asking for our vote if there are the conditions. */\n// 在条件满足的情况下，为请求进行故障转移的节点 node 进行投票，支持它进行故障转移\nvoid clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request) {\n\n // 请求节点的主节点\n clusterNode *master = node->slaveof;\n\n // 请求节点的当前配置纪元\n uint64_t requestCurrentEpoch = ntohu64(request->currentEpoch);\n\n // 请求节点想要获得投票的纪元\n uint64_t requestConfigEpoch = ntohu64(request->configEpoch);\n\n // 请求节点的槽布局\n unsigned char *claimed_slots = request->myslots;\n int force_ack = request->mflags[0] & CLUSTERMSG_FLAG0_FORCEACK;\n int j;\n\n /* IF we are not a master serving at least 1 slot, we don't have the\n * right to vote, as the cluster size in Redis Cluster is the number\n * of masters serving at least one slot, and quorum is the cluster\n * size + 1 */\n\n // 如果节点为从节点，或者是一个没有处理任何槽的主节点，\n // 那么它没有投票权\n if (nodeIsSlave(myself) || myself->numslots == 0) return;\n\n /* Request epoch must be >= our currentEpoch. */\n // 请求的配置纪元必须大于等于当前节点的配置纪元\n if (requestCurrentEpoch < server.cluster->currentEpoch) return;\n\n /* I already voted for this epoch? Return ASAP. */\n // 已经投过票了\n if (server.cluster->lastVoteEpoch == server.cluster->currentEpoch) return;\n\n /* Node must be a slave and its master down.\n * The master can be non failing if the request is flagged\n * with CLUSTERMSG_FLAG0_FORCEACK (manual failover). */\n if (nodeIsMaster(node) || master == NULL ||\n (!nodeFailed(master) && !force_ack)) return;\n\n /* We did not voted for a slave about this master for two\n * times the node timeout. This is not strictly needed for correctness\n * of the algorithm but makes the base case more linear. */\n // 如果之前一段时间已经对请求节点进行过投票，那么不进行投票\n if (mstime() - node->slaveof->voted_time < server.cluster_node_timeout * 2)\n return;\n\n /* The slave requesting the vote must have a configEpoch for the claimed\n * slots that is >= the one of the masters currently serving the same\n * slots in the current configuration. */\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n\n // 跳过未指派节点\n if (bitmapTestBit(claimed_slots, j) == 0) continue;\n\n // 查找是否有某个槽的配置纪元大于节点请求的纪元\n if (server.cluster->slots[j] == NULL ||\n server.cluster->slots[j]->configEpoch <= requestConfigEpoch)\n {\n continue;\n }\n\n // 如果有的话，说明节点请求的纪元已经过期，没有必要进行投票\n /* If we reached this point we found a slot that in our current slots\n * is served by a master with a greater configEpoch than the one claimed\n * by the slave requesting our vote. Refuse to vote for this slave. */\n return;\n }\n\n /* We can vote for this slave. */\n // 为节点投票\n clusterSendFailoverAuth(node);\n // 更新时间值\n server.cluster->lastVoteEpoch = server.cluster->currentEpoch;\n node->slaveof->voted_time = mstime();\n}\n\n/* This function returns the \"rank\" of this instance, a slave, in the context\n * of its master-slaves ring. The rank of the slave is given by the number of\n * other slaves for the same master that have a better replication offset\n * compared to the local one (better means, greater, so they claim more data).\n *\n * A slave with rank 0 is the one with the greatest (most up to date)\n * replication offset, and so forth. Note that because how the rank is computed\n * multiple slaves may have the same rank, in case they have the same offset.\n *\n * The slave rank is used to add a delay to start an election in order to\n * get voted and replace a failing master. Slaves with better replication\n * offsets are more likely to win. */\nint clusterGetSlaveRank(void) {\n long long myoffset;\n int j, rank = 0;\n clusterNode *master;\n\n redisAssert(nodeIsSlave(myself));\n master = myself->slaveof;\n if (master == NULL) return 0; /* Never called by slaves without master. */\n\n myoffset = replicationGetSlaveOffset();\n for (j = 0; j < master->numslaves; j++)\n if (master->slaves[j] != myself &&\n master->slaves[j]->repl_offset > myoffset) rank++;\n return rank;\n}\n\n/* This function is called if we are a slave node and our master serving\n * a non-zero amount of hash slots is in FAIL state.\n *\n * 如果当前节点是一个从节点，并且它正在复制的一个负责非零个槽的主节点处于 FAIL 状态，\n * 那么执行这个函数。\n *\n * The gaol of this function is:\n *\n * 这个函数有三个目标：\n *\n * 1) To check if we are able to perform a failover, is our data updated?\n * 检查是否可以对主节点执行一次故障转移，节点的关于主节点的信息是否准确和最新（updated）？\n * 2) Try to get elected by masters.\n * 选举一个新的主节点\n * 3) Perform the failover informing all the other nodes.\n * 执行故障转移，并通知其他节点\n */\nvoid clusterHandleSlaveFailover(void) {\n mstime_t data_age;\n mstime_t auth_age = mstime() - server.cluster->failover_auth_time;\n int needed_quorum = (server.cluster->size / 2) + 1;\n int manual_failover = server.cluster->mf_end != 0 &&\n server.cluster->mf_can_start;\n int j;\n mstime_t auth_timeout, auth_retry_time;\n\n server.cluster->todo_before_sleep &= ~CLUSTER_TODO_HANDLE_FAILOVER;\n\n /* Compute the failover timeout (the max time we have to send votes\n * and wait for replies), and the failover retry time (the time to wait\n * before waiting again.\n *\n * Timeout is MIN(NODE_TIMEOUT*2,2000) milliseconds.\n * Retry is two times the Timeout.\n */\n auth_timeout = server.cluster_node_timeout*2;\n if (auth_timeout < 2000) auth_timeout = 2000;\n auth_retry_time = auth_timeout*2;\n\n /* Pre conditions to run the function, that must be met both in case\n * of an automatic or manual failover:\n * 1) We are a slave.\n * 2) Our master is flagged as FAIL, or this is a manual failover.\n * 3) It is serving slots. */\n if (nodeIsMaster(myself) ||\n myself->slaveof == NULL ||\n (!nodeFailed(myself->slaveof) && !manual_failover) ||\n myself->slaveof->numslots == 0) return;\n\n /* Set data_age to the number of seconds we are disconnected from\n * the master. */\n // 将 data_age 设置为从节点与主节点的断开秒数\n if (server.repl_state == REDIS_REPL_CONNECTED) {\n data_age = (mstime_t)(server.unixtime - server.master->lastinteraction)\n * 1000;\n } else {\n data_age = (mstime_t)(server.unixtime - server.repl_down_since) * 1000;\n }\n\n /* Remove the node timeout from the data age as it is fine that we are\n * disconnected from our master at least for the time it was down to be\n * flagged as FAIL, that's the baseline. */\n // node timeout 的时间不计入断线时间之内\n if (data_age > server.cluster_node_timeout)\n data_age -= server.cluster_node_timeout;\n\n /* Check if our data is recent enough. For now we just use a fixed\n * constant of ten times the node timeout since the cluster should\n * react much faster to a master down.\n *\n * Check bypassed for manual failovers. */\n // 检查这个从节点的数据是否较新：\n // 目前的检测办法是断线时间不能超过 node timeout 的十倍\n if (data_age >\n ((mstime_t)server.repl_ping_slave_period * 1000) +\n (server.cluster_node_timeout * REDIS_CLUSTER_SLAVE_VALIDITY_MULT))\n {\n if (!manual_failover) return;\n }\n\n /* If the previous failover attempt timedout and the retry time has\n * elapsed, we can setup a new one. */\n if (auth_age > auth_retry_time) {\n server.cluster->failover_auth_time = mstime() +\n 500 + /* Fixed delay of 500 milliseconds, let FAIL msg propagate. */\n random() % 500; /* Random delay between 0 and 500 milliseconds. */\n server.cluster->failover_auth_count = 0;\n server.cluster->failover_auth_sent = 0;\n server.cluster->failover_auth_rank = clusterGetSlaveRank();\n /* We add another delay that is proportional to the slave rank.\n * Specifically 1 second * rank. This way slaves that have a probably\n * less updated replication offset, are penalized. */\n server.cluster->failover_auth_time +=\n server.cluster->failover_auth_rank * 1000;\n /* However if this is a manual failover, no delay is needed. */\n if (server.cluster->mf_end) {\n server.cluster->failover_auth_time = mstime();\n server.cluster->failover_auth_rank = 0;\n }\n redisLog(REDIS_WARNING,\n \"Start of election delayed for %lld milliseconds \"\n \"(rank #%d, offset %lld).\",\n server.cluster->failover_auth_time - mstime(),\n server.cluster->failover_auth_rank,\n replicationGetSlaveOffset());\n /* Now that we have a scheduled election, broadcast our offset\n * to all the other slaves so that they'll updated their offsets\n * if our offset is better. */\n clusterBroadcastPong(CLUSTER_BROADCAST_LOCAL_SLAVES);\n return;\n }\n\n /* It is possible that we received more updated offsets from other\n * slaves for the same master since we computed our election delay.\n * Update the delay if our rank changed.\n *\n * Not performed if this is a manual failover. */\n if (server.cluster->failover_auth_sent == 0 &&\n server.cluster->mf_end == 0)\n {\n int newrank = clusterGetSlaveRank();\n if (newrank > server.cluster->failover_auth_rank) {\n long long added_delay =\n (newrank - server.cluster->failover_auth_rank) * 1000;\n server.cluster->failover_auth_time += added_delay;\n server.cluster->failover_auth_rank = newrank;\n redisLog(REDIS_WARNING,\n \"Slave rank updated to #%d, added %lld milliseconds of delay.\",\n newrank, added_delay);\n }\n }\n\n /* Return ASAP if we can't still start the election. */\n // 如果执行故障转移的时间未到，先返回\n if (mstime() < server.cluster->failover_auth_time) return;\n\n /* Return ASAP if the election is too old to be valid. */\n // 如果距离应该执行故障转移的时间已经过了很久\n // 那么不应该再执行故障转移了（因为可能已经没有需要了）\n // 直接返回\n if (auth_age > auth_timeout) return;\n\n /* Ask for votes if needed. */\n // 向其他节点发送故障转移请求\n if (server.cluster->failover_auth_sent == 0) {\n\n // 增加配置纪元\n server.cluster->currentEpoch++;\n\n // 记录发起故障转移的配置纪元\n server.cluster->failover_auth_epoch = server.cluster->currentEpoch;\n\n redisLog(REDIS_WARNING,\"Starting a failover election for epoch %llu.\",\n (unsigned long long) server.cluster->currentEpoch);\n\n // 向其他所有节点发送信息，看它们是否支持由本节点来对下线主节点进行故障转移\n clusterRequestFailoverAuth();\n\n // 打开标识，表示已发送信息\n server.cluster->failover_auth_sent = 1;\n\n // TODO:\n // 在进入下个事件循环之前，执行：\n // 1）保存配置文件\n // 2）更新节点状态\n // 3）同步配置\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|\n CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_FSYNC_CONFIG);\n return; /* Wait for replies. */\n }\n\n /* Check if we reached the quorum. */\n // 如果当前节点获得了足够多的投票，那么对下线主节点进行故障转移\n if (server.cluster->failover_auth_count >= needed_quorum) {\n // 旧主节点\n clusterNode *oldmaster = myself->slaveof;\n\n redisLog(REDIS_WARNING,\n \"Failover election won: I'm the new master.\");\n\n /* We have the quorum, perform all the steps to correctly promote\n * this slave to a master.\n *\n * 1) Turn this node into a master. \n * 将当前节点的身份由从节点改为主节点\n */\n clusterSetNodeAsMaster(myself);\n // 让从节点取消复制，成为新的主节点\n replicationUnsetMaster();\n\n /* 2) Claim all the slots assigned to our master. */\n // 接收所有主节点负责处理的槽\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n if (clusterNodeGetSlotBit(oldmaster,j)) {\n // 将槽设置为未分配的\n clusterDelSlot(j);\n // 将槽的负责人设置为当前节点\n clusterAddSlot(myself,j);\n }\n }\n\n /* 3) Update my configEpoch to the epoch of the election. */\n // 更新集群配置纪元\n myself->configEpoch = server.cluster->failover_auth_epoch;\n\n /* 4) Update state and save config. */\n // 更新节点状态\n clusterUpdateState();\n // 并保存配置文件\n clusterSaveConfigOrDie(1);\n\n /* 5) Pong all the other nodes so that they can update the state\n * accordingly and detect that we switched to master role. */\n // 向所有节点发送 PONG 信息\n // 让它们可以知道当前节点已经升级为主节点了\n clusterBroadcastPong(CLUSTER_BROADCAST_ALL);\n\n /* 6) If there was a manual failover in progress, clear the state. */\n // 如果有手动故障转移正在执行，那么清理和它有关的状态\n resetManualFailover();\n }\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER slave migration\n *\n * Slave migration is the process that allows a slave of a master that is\n * already covered by at least another slave, to \"migrate\" to a master that\n * is orpaned, that is, left with no working slaves.\n * -------------------------------------------------------------------------- */\n\n/* This function is responsible to decide if this replica should be migrated\n * to a different (orphaned) master. It is called by the clusterCron() function\n * only if:\n *\n * 1) We are a slave node.\n * 2) It was detected that there is at least one orphaned master in\n * the cluster.\n * 3) We are a slave of one of the masters with the greatest number of\n * slaves.\n *\n * This checks are performed by the caller since it requires to iterate\n * the nodes anyway, so we spend time into clusterHandleSlaveMigration()\n * if definitely needed.\n *\n * The fuction is called with a pre-computed max_slaves, that is the max\n * number of working (not in FAIL state) slaves for a single master.\n *\n * Additional conditions for migration are examined inside the function.\n */\nvoid clusterHandleSlaveMigration(int max_slaves) {\n int j, okslaves = 0;\n clusterNode *mymaster = myself->slaveof, *target = NULL, *candidate = NULL;\n dictIterator *di;\n dictEntry *de;\n\n /* Step 1: Don't migrate if the cluster state is not ok. */\n if (server.cluster->state != REDIS_CLUSTER_OK) return;\n\n /* Step 2: Don't migrate if my master will not be left with at least\n * 'migration-barrier' slaves after my migration. */\n if (mymaster == NULL) return;\n for (j = 0; j < mymaster->numslaves; j++)\n if (!nodeFailed(mymaster->slaves[j]) &&\n !nodeTimedOut(mymaster->slaves[j])) okslaves++;\n if (okslaves <= server.cluster_migration_barrier) return;\n\n /* Step 3: Idenitfy a candidate for migration, and check if among the\n * masters with the greatest number of ok slaves, I'm the one with the\n * smaller node ID.\n *\n * Note that this means that eventually a replica migration will occurr\n * since slaves that are reachable again always have their FAIL flag\n * cleared. At the same time this does not mean that there are no\n * race conditions possible (two slaves migrating at the same time), but\n * this is extremely unlikely to happen, and harmless. */\n candidate = myself;\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n int okslaves;\n\n /* Only iterate over working masters. */\n if (nodeIsSlave(node) || nodeFailed(node)) continue;\n okslaves = clusterCountNonFailingSlaves(node);\n\n if (okslaves == 0 && target == NULL && node->numslots > 0)\n target = node;\n\n if (okslaves == max_slaves) {\n for (j = 0; j < node->numslaves; j++) {\n if (memcmp(node->slaves[j]->name,\n candidate->name,\n REDIS_CLUSTER_NAMELEN) < 0)\n {\n candidate = node->slaves[j];\n }\n }\n }\n }\n\n /* Step 4: perform the migration if there is a target, and if I'm the\n * candidate. */\n if (target && candidate == myself) {\n redisLog(REDIS_WARNING,\"Migrating to orphaned master %.40s\",\n target->name);\n clusterSetMaster(target);\n }\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER manual failover\n *\n * This are the important steps performed by slaves during a manual failover:\n * 1) User send CLUSTER FAILOVER command. The failover state is initialized\n * setting mf_end to the millisecond unix time at which we'll abort the\n * attempt.\n * 2) Slave sends a MFSTART message to the master requesting to pause clients\n * for two times the manual failover timeout REDIS_CLUSTER_MF_TIMEOUT.\n * When master is paused for manual failover, it also starts to flag\n * packets with CLUSTERMSG_FLAG0_PAUSED.\n * 3) Slave waits for master to send its replication offset flagged as PAUSED.\n * 4) If slave received the offset from the master, and its offset matches,\n * mf_can_start is set to 1, and clusterHandleSlaveFailover() will perform\n * the failover as usually, with the difference that the vote request\n * will be modified to force masters to vote for a slave that has a\n * working master.\n *\n * From the point of view of the master things are simpler: when a\n * PAUSE_CLIENTS packet is received the master sets mf_end as well and\n * the sender in mf_slave. During the time limit for the manual failover\n * the master will just send PINGs more often to this slave, flagged with\n * the PAUSED flag, so that the slave will set mf_master_offset when receiving\n * a packet from the master with this flag set.\n *\n * The gaol of the manual failover is to perform a fast failover without\n * data loss due to the asynchronous master-slave replication.\n * -------------------------------------------------------------------------- */\n\n/* Reset the manual failover state. This works for both masters and slavesa\n * as all the state about manual failover is cleared.\n *\n * 重置与手动故障转移有关的状态，主节点和从节点都可以使用。\n *\n * The function can be used both to initialize the manual failover state at\n * startup or to abort a manual failover in progress. \n * 这个函数既可以用于在启动集群时进行初始化，\n * 又可以实际地应用在手动故障转移的情况。\n */\nvoid resetManualFailover(void) {\n if (server.cluster->mf_end && clientsArePaused()) {\n server.clients_pause_end_time = 0;\n clientsArePaused(); /* Just use the side effect of the function. */\n }\n server.cluster->mf_end = 0; /* No manual failover in progress. */\n server.cluster->mf_can_start = 0;\n server.cluster->mf_slave = NULL;\n server.cluster->mf_master_offset = 0;\n}\n\n/* If a manual failover timed out, abort it. */\nvoid manualFailoverCheckTimeout(void) {\n if (server.cluster->mf_end && server.cluster->mf_end < mstime()) {\n redisLog(REDIS_WARNING,\"Manual failover timed out.\");\n resetManualFailover();\n }\n}\n\n/* This function is called from the cluster cron function in order to go\n * forward with a manual failover state machine. */\nvoid clusterHandleManualFailover(void) {\n /* Return ASAP if no manual failover is in progress. */\n if (server.cluster->mf_end == 0) return;\n\n /* If mf_can_start is non-zero, the failover was alrady triggered so the\n * next steps are performed by clusterHandleSlaveFailover(). */\n if (server.cluster->mf_can_start) return;\n\n if (server.cluster->mf_master_offset == 0) return; /* Wait for offset... */\n\n if (server.cluster->mf_master_offset == replicationGetSlaveOffset()) {\n /* Our replication offset matches the master replication offset\n * announced after clients were paused. We can start the failover. */\n server.cluster->mf_can_start = 1;\n redisLog(REDIS_WARNING,\n \"All master replication stream processed, \"\n \"manual failover can start.\");\n }\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER cron job\n * -------------------------------------------------------------------------- */\n\n/* This is executed 10 times every second */\n// 集群常规操作函数，默认每秒执行 10 次（每间隔 100 毫秒执行一次）\nvoid clusterCron(void) {\n dictIterator *di;\n dictEntry *de;\n int update_state = 0;\n int orphaned_masters; /* How many masters there are without ok slaves. */\n int max_slaves; /* Max number of ok slaves for a single master. */\n int this_slaves; /* Number of ok slaves for our master (if we are slave). */\n mstime_t min_pong = 0, now = mstime();\n clusterNode *min_pong_node = NULL;\n // 迭代计数器，一个静态变量\n static unsigned long long iteration = 0;\n mstime_t handshake_timeout;\n\n // 记录一次迭代\n iteration++; /* Number of times this function was called so far. */\n\n /* The handshake timeout is the time after which a handshake node that was\n * not turned into a normal node is removed from the nodes. Usually it is\n * just the NODE_TIMEOUT value, but when NODE_TIMEOUT is too small we use\n * the value of 1 second. */\n // 如果一个 handshake 节点没有在 handshake timeout 内\n // 转换成普通节点（normal node），\n // 那么节点会从 nodes 表中移除这个 handshake 节点\n // 一般来说 handshake timeout 的值总是等于 NODE_TIMEOUT\n // 不过如果 NODE_TIMEOUT 太少的话，程序会将值设为 1 秒钟\n handshake_timeout = server.cluster_node_timeout;\n if (handshake_timeout < 1000) handshake_timeout = 1000;\n\n /* Check if we have disconnected nodes and re-establish the connection. */\n // 向集群中的所有断线或者未连接节点发送消息\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n // 跳过当前节点以及没有地址的节点\n if (node->flags & (REDIS_NODE_MYSELF|REDIS_NODE_NOADDR)) continue;\n\n /* A Node in HANDSHAKE state has a limited lifespan equal to the\n * configured node timeout. */\n // 如果 handshake 节点已超时，释放它\n if (nodeInHandshake(node) && now - node->ctime > handshake_timeout) {\n freeClusterNode(node);\n continue;\n }\n\n // 为未创建连接的节点创建连接\n if (node->link == NULL) {\n int fd;\n mstime_t old_ping_sent;\n clusterLink *link;\n\n fd = anetTcpNonBlockBindConnect(server.neterr, node->ip,\n node->port+REDIS_CLUSTER_PORT_INCR,\n server.bindaddr_count ? server.bindaddr[0] : NULL);\n if (fd == -1) {\n redisLog(REDIS_DEBUG, \"Unable to connect to \"\n \"Cluster Node [%s]:%d -> %s\", node->ip,\n node->port+REDIS_CLUSTER_PORT_INCR,\n server.neterr);\n continue;\n }\n link = createClusterLink(node);\n link->fd = fd;\n node->link = link;\n aeCreateFileEvent(server.el,link->fd,AE_READABLE,\n clusterReadHandler,link);\n /* Queue a PING in the new connection ASAP: this is crucial\n * to avoid false positives in failure detection.\n *\n * If the node is flagged as MEET, we send a MEET message instead\n * of a PING one, to force the receiver to add us in its node\n * table. */\n // 向新连接的节点发送 PING 命令，防止节点被识进入下线\n // 如果节点被标记为 MEET ，那么发送 MEET 命令，否则发送 PING 命令\n old_ping_sent = node->ping_sent;\n clusterSendPing(link, node->flags & REDIS_NODE_MEET ?\n CLUSTERMSG_TYPE_MEET : CLUSTERMSG_TYPE_PING);\n\n // 这不是第一次发送 PING 信息，所以可以还原这个时间\n // 等 clusterSendPing() 函数来更新它\n if (old_ping_sent) {\n /* If there was an active ping before the link was\n * disconnected, we want to restore the ping time, otherwise\n * replaced by the clusterSendPing() call. */\n node->ping_sent = old_ping_sent;\n }\n\n /* We can clear the flag after the first packet is sent.\n *\n * 在发送 MEET 信息之后，清除节点的 MEET 标识。\n *\n * If we'll never receive a PONG, we'll never send new packets\n * to this node. Instead after the PONG is received and we\n * are no longer in meet/handshake status, we want to send\n * normal PING packets. \n *\n * 如果当前节点（发送者）没能收到 MEET 信息的回复，\n * 那么它将不再向目标节点发送命令。\n *\n * 如果接收到回复的话，那么节点将不再处于 HANDSHAKE 状态，\n * 并继续向目标节点发送普通 PING 命令。\n */\n node->flags &= ~REDIS_NODE_MEET;\n\n redisLog(REDIS_DEBUG,\"Connecting with Node %.40s at %s:%d\",\n node->name, node->ip, node->port+REDIS_CLUSTER_PORT_INCR);\n }\n }\n dictReleaseIterator(di);\n\n /* Ping some random node 1 time every 10 iterations, so that we usually ping\n * one random node every second. */\n // clusterCron() 每执行 10 次（至少间隔一秒钟），就向一个随机节点发送 gossip 信息\n if (!(iteration % 10)) {\n int j;\n\n /* Check a few random nodes and ping the one with the oldest\n * pong_received time. */\n // 随机 5 个节点，选出其中一个\n for (j = 0; j < 5; j++) {\n\n // 随机在集群中挑选节点\n de = dictGetRandomKey(server.cluster->nodes);\n clusterNode *this = dictGetVal(de);\n\n /* Don't ping nodes disconnected or with a ping currently active. */\n // 不要 PING 连接断开的节点，也不要 PING 最近已经 PING 过的节点\n if (this->link == NULL || this->ping_sent != 0) continue;\n\n if (this->flags & (REDIS_NODE_MYSELF|REDIS_NODE_HANDSHAKE))\n continue;\n\n // 选出 5 个随机节点中最近一次接收 PONG 回复距离现在最旧的节点\n if (min_pong_node == NULL || min_pong > this->pong_received) {\n min_pong_node = this;\n min_pong = this->pong_received;\n }\n }\n\n // 向最久没有收到 PONG 回复的节点发送 PING 命令\n if (min_pong_node) {\n redisLog(REDIS_DEBUG,\"Pinging node %.40s\", min_pong_node->name);\n clusterSendPing(min_pong_node->link, CLUSTERMSG_TYPE_PING);\n }\n }\n\n // 遍历所有节点，检查是否需要将某个节点标记为下线\n /* Iterate nodes to check if we need to flag something as failing.\n * This loop is also responsible to:\n * 1) Check if there are orphaned masters (masters without non failing\n * slaves).\n * 2) Count the max number of non failing slaves for a single master.\n * 3) Count the number of slaves for our master, if we are a slave. */\n orphaned_masters = 0;\n max_slaves = 0;\n this_slaves = 0;\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n now = mstime(); /* Use an updated time at every iteration. */\n mstime_t delay;\n\n // 跳过节点本身、无地址节点、HANDSHAKE 状态的节点\n if (node->flags &\n (REDIS_NODE_MYSELF|REDIS_NODE_NOADDR|REDIS_NODE_HANDSHAKE))\n continue;\n\n /* Orphaned master check, useful only if the current instance\n * is a slave that may migrate to another master. */\n if (nodeIsSlave(myself) && nodeIsMaster(node) && !nodeFailed(node)) {\n int okslaves = clusterCountNonFailingSlaves(node);\n\n if (okslaves == 0 && node->numslots > 0) orphaned_masters++;\n if (okslaves > max_slaves) max_slaves = okslaves;\n if (nodeIsSlave(myself) && myself->slaveof == node)\n this_slaves = okslaves;\n }\n\n /* If we are waiting for the PONG more than half the cluster\n * timeout, reconnect the link: maybe there is a connection\n * issue even if the node is alive. */\n // 如果等到 PONG 到达的时间超过了 node timeout 一半的连接\n // 因为尽管节点依然正常，但连接可能已经出问题了\n if (node->link && /* is connected */\n now - node->link->ctime >\n server.cluster_node_timeout && /* was not already reconnected */\n node->ping_sent && /* we already sent a ping */\n node->pong_received < node->ping_sent && /* still waiting pong */\n /* and we are waiting for the pong more than timeout/2 */\n now - node->ping_sent > server.cluster_node_timeout/2)\n {\n /* Disconnect the link, it will be reconnected automatically. */\n // 释放连接，下次 clusterCron() 会自动重连\n freeClusterLink(node->link);\n }\n\n /* If we have currently no active ping in this instance, and the\n * received PONG is older than half the cluster timeout, send\n * a new ping now, to ensure all the nodes are pinged without\n * a too big delay. */\n // 如果目前没有在 PING 节点\n // 并且已经有 node timeout 一半的时间没有从节点那里收到 PONG 回复\n // 那么向节点发送一个 PING ，确保节点的信息不会太旧\n // （因为一部分节点可能一直没有被随机中）\n if (node->link &&\n node->ping_sent == 0 &&\n (now - node->pong_received) > server.cluster_node_timeout/2)\n {\n clusterSendPing(node->link, CLUSTERMSG_TYPE_PING);\n continue;\n }\n\n /* If we are a master and one of the slaves requested a manual\n * failover, ping it continuously. */\n // 如果这是一个主节点，并且有一个从服务器请求进行手动故障转移\n // 那么向从服务器发送 PING 。\n if (server.cluster->mf_end &&\n nodeIsMaster(myself) &&\n server.cluster->mf_slave == node &&\n node->link)\n {\n clusterSendPing(node->link, CLUSTERMSG_TYPE_PING);\n continue;\n }\n\n /* Check only if we have an active ping for this instance. */\n // 以下代码只在节点发送了 PING 命令的情况下执行\n if (node->ping_sent == 0) continue;\n\n /* Compute the delay of the PONG. Note that if we already received\n * the PONG, then node->ping_sent is zero, so can't reach this\n * code at all. */\n // 计算等待 PONG 回复的时长\n delay = now - node->ping_sent;\n\n // 等待 PONG 回复的时长超过了限制值，将目标节点标记为 PFAIL （疑似下线）\n if (delay > server.cluster_node_timeout) {\n /* Timeout reached. Set the node as possibly failing if it is\n * not already in this state. */\n if (!(node->flags & (REDIS_NODE_PFAIL|REDIS_NODE_FAIL))) {\n redisLog(REDIS_DEBUG,\"*** NODE %.40s possibly failing\",\n node->name);\n // 打开疑似下线标记\n node->flags |= REDIS_NODE_PFAIL;\n update_state = 1;\n }\n }\n }\n dictReleaseIterator(di);\n\n /* If we are a slave node but the replication is still turned off,\n * enable it if we know the address of our master and it appears to\n * be up. */\n // 如果从节点没有在复制主节点，那么对从节点进行设置\n if (nodeIsSlave(myself) &&\n server.masterhost == NULL &&\n myself->slaveof &&\n nodeHasAddr(myself->slaveof))\n {\n replicationSetMaster(myself->slaveof->ip, myself->slaveof->port);\n }\n\n /* Abourt a manual failover if the timeout is reached. */\n manualFailoverCheckTimeout();\n\n if (nodeIsSlave(myself)) {\n clusterHandleManualFailover();\n clusterHandleSlaveFailover();\n /* If there are orphaned slaves, and we are a slave among the masters\n * with the max number of non-failing slaves, consider migrating to\n * the orphaned masters. Note that it does not make sense to try\n * a migration if there is no master with at least *two* working\n * slaves. */\n if (orphaned_masters && max_slaves >= 2 && this_slaves == max_slaves)\n clusterHandleSlaveMigration(max_slaves);\n }\n\n // 更新集群状态\n if (update_state || server.cluster->state == REDIS_CLUSTER_FAIL)\n clusterUpdateState();\n}\n\n/* This function is called before the event handler returns to sleep for\n * events. It is useful to perform operations that must be done ASAP in\n * reaction to events fired but that are not safe to perform inside event\n * handlers, or to perform potentially expansive tasks that we need to do\n * a single time before replying to clients. \n *\n * 在进入下个事件循环时调用。\n * 这个函数做的事都是需要尽快执行，但是不能在执行文件事件期间做的事情。\n */\nvoid clusterBeforeSleep(void) {\n\n /* Handle failover, this is needed when it is likely that there is already\n * the quorum from masters in order to react fast. */\n // 执行故障迁移\n if (server.cluster->todo_before_sleep & CLUSTER_TODO_HANDLE_FAILOVER)\n clusterHandleSlaveFailover();\n\n /* Update the cluster state. */\n // 更新节点的状态\n if (server.cluster->todo_before_sleep & CLUSTER_TODO_UPDATE_STATE)\n clusterUpdateState();\n\n /* Save the config, possibly using fsync. */\n // 保存 nodes.conf 配置文件\n if (server.cluster->todo_before_sleep & CLUSTER_TODO_SAVE_CONFIG) {\n int fsync = server.cluster->todo_before_sleep &\n CLUSTER_TODO_FSYNC_CONFIG;\n clusterSaveConfigOrDie(fsync);\n }\n\n /* Reset our flags (not strictly needed since every single function\n * called for flags set should be able to clear its flag). */\n server.cluster->todo_before_sleep = 0;\n}\n\n// 打开 todo_before_sleep 的指定标识\n// 每个标识代表了节点在结束一个事件循环时要做的工作\nvoid clusterDoBeforeSleep(int flags) {\n server.cluster->todo_before_sleep |= flags;\n}\n\n/* -----------------------------------------------------------------------------\n * Slots management\n * -------------------------------------------------------------------------- */\n\n/* Test bit 'pos' in a generic bitmap. Return 1 if the bit is set,\n * otherwise 0. */\n// 检查位图 bitmap 的 pos 位置是否已经被设置\n// 返回 1 表示已被设置，返回 0 表示未被设置。\nint bitmapTestBit(unsigned char *bitmap, int pos) {\n off_t byte = pos/8;\n int bit = pos&7;\n return (bitmap[byte] & (1<slots,slot);\n bitmapSetBit(n->slots,slot);\n if (!old) n->numslots++;\n return old;\n}\n\n/* Clear the slot bit and return the old value. */\n// 清空槽二进制位，并返回旧值\nint clusterNodeClearSlotBit(clusterNode *n, int slot) {\n int old = bitmapTestBit(n->slots,slot);\n bitmapClearBit(n->slots,slot);\n if (old) n->numslots--;\n return old;\n}\n\n/* Return the slot bit from the cluster node structure. */\n// 返回槽的二进制位的值\nint clusterNodeGetSlotBit(clusterNode *n, int slot) {\n return bitmapTestBit(n->slots,slot);\n}\n\n/* Add the specified slot to the list of slots that node 'n' will\n * serve. Return REDIS_OK if the operation ended with success.\n * If the slot is already assigned to another instance this is considered\n * an error and REDIS_ERR is returned. */\n// 将槽 slot 添加到节点 n 需要处理的槽的列表中\n// 添加成功返回 REDIS_OK ,如果槽已经由这个节点处理了\n// 那么返回 REDIS_ERR 。\nint clusterAddSlot(clusterNode *n, int slot) {\n\n // 槽 slot 已经是节点 n 处理的了\n if (server.cluster->slots[slot]) return REDIS_ERR;\n\n // 设置 bitmap\n clusterNodeSetSlotBit(n,slot);\n\n // 更新集群状态\n server.cluster->slots[slot] = n;\n\n return REDIS_OK;\n}\n\n/* Delete the specified slot marking it as unassigned.\n *\n * 将指定槽标记为未分配（unassigned）。\n *\n * Returns REDIS_OK if the slot was assigned, otherwise if the slot was\n * already unassigned REDIS_ERR is returned. \n *\n * 标记成功返回 REDIS_OK ，\n * 如果槽已经是未分配的，那么返回 REDIS_ERR 。\n */\nint clusterDelSlot(int slot) {\n\n // 获取当前处理槽 slot 的节点 n\n clusterNode *n = server.cluster->slots[slot];\n\n if (!n) return REDIS_ERR;\n\n // 清除位图\n redisAssert(clusterNodeClearSlotBit(n,slot) == 1);\n\n // 清空负责处理槽的节点\n server.cluster->slots[slot] = NULL;\n\n return REDIS_OK;\n}\n\n/* Delete all the slots associated with the specified node.\n * The number of deleted slots is returned. */\n// 删除所有由给定节点处理的槽，并返回被删除槽的数量\nint clusterDelNodeSlots(clusterNode *node) {\n int deleted = 0, j;\n\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n // 如果这个槽由该节点负责，那么删除它\n if (clusterNodeGetSlotBit(node,j)) clusterDelSlot(j);\n deleted++;\n }\n return deleted;\n}\n\n/* Clear the migrating / importing state for all the slots.\n * This is useful at initialization and when turning a master into slave. */\n// 清理所有槽的迁移和导入状态\n// 通常在初始化或者将主节点转为从节点时使用\nvoid clusterCloseAllSlots(void) {\n memset(server.cluster->migrating_slots_to,0,\n sizeof(server.cluster->migrating_slots_to));\n memset(server.cluster->importing_slots_from,0,\n sizeof(server.cluster->importing_slots_from));\n}\n\n/* -----------------------------------------------------------------------------\n * Cluster state evaluation function\n * -------------------------------------------------------------------------- */\n\n/* The following are defines that are only used in the evaluation function\n * and are based on heuristics. Actaully the main point about the rejoin and\n * writable delay is that they should be a few orders of magnitude larger\n * than the network latency. */\n#define REDIS_CLUSTER_MAX_REJOIN_DELAY 5000\n#define REDIS_CLUSTER_MIN_REJOIN_DELAY 500\n#define REDIS_CLUSTER_WRITABLE_DELAY 2000\n\n// 更新节点状态\nvoid clusterUpdateState(void) {\n int j, new_state;\n int unreachable_masters = 0;\n static mstime_t among_minority_time;\n static mstime_t first_call_time = 0;\n\n server.cluster->todo_before_sleep &= ~CLUSTER_TODO_UPDATE_STATE;\n\n /* If this is a master node, wait some time before turning the state\n * into OK, since it is not a good idea to rejoin the cluster as a writable\n * master, after a reboot, without giving the cluster a chance to\n * reconfigure this node. Note that the delay is calculated starting from\n * the first call to this function and not since the server start, in order\n * to don't count the DB loading time. */\n if (first_call_time == 0) first_call_time = mstime();\n if (nodeIsMaster(myself) &&\n mstime() - first_call_time < REDIS_CLUSTER_WRITABLE_DELAY) return;\n\n /* Start assuming the state is OK. We'll turn it into FAIL if there\n * are the right conditions. */\n\n // 先假设节点状态为 OK ，后面再检测节点是否真的下线\n new_state = REDIS_CLUSTER_OK;\n\n /* Check if all the slots are covered. */\n // 检查是否所有槽都已经有某个节点在处理\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n if (server.cluster->slots[j] == NULL ||\n server.cluster->slots[j]->flags & (REDIS_NODE_FAIL))\n {\n new_state = REDIS_CLUSTER_FAIL;\n break;\n }\n }\n\n /* Compute the cluster size, that is the number of master nodes\n * serving at least a single slot.\n *\n * At the same time count the number of unreachable masters with\n * at least one node. */\n // 统计在线并且正在处理至少一个槽的 master 的数量，\n // 以及下线 master 的数量\n {\n dictIterator *di;\n dictEntry *de;\n\n server.cluster->size = 0;\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n if (nodeIsMaster(node) && node->numslots) {\n server.cluster->size++;\n if (node->flags & (REDIS_NODE_FAIL|REDIS_NODE_PFAIL))\n unreachable_masters++;\n }\n }\n dictReleaseIterator(di);\n }\n\n /* If we can't reach at least half the masters, change the cluster state\n * to FAIL, as we are not even able to mark nodes as FAIL in this side\n * of the netsplit because of lack of majority.\n *\n * 如果不能连接到半数以上节点，那么将我们自己的状态设置为 FAIL\n * 因为在少于半数节点的情况下，节点是无法将一个节点判断为 FAIL 的。\n */\n {\n int needed_quorum = (server.cluster->size / 2) + 1;\n\n if (unreachable_masters >= needed_quorum) {\n new_state = REDIS_CLUSTER_FAIL;\n among_minority_time = mstime();\n }\n }\n\n /* Log a state change */\n // 记录状态变更\n if (new_state != server.cluster->state) {\n mstime_t rejoin_delay = server.cluster_node_timeout;\n\n /* If the instance is a master and was partitioned away with the\n * minority, don't let it accept queries for some time after the\n * partition heals, to make sure there is enough time to receive\n * a configuration update. */\n if (rejoin_delay > REDIS_CLUSTER_MAX_REJOIN_DELAY)\n rejoin_delay = REDIS_CLUSTER_MAX_REJOIN_DELAY;\n if (rejoin_delay < REDIS_CLUSTER_MIN_REJOIN_DELAY)\n rejoin_delay = REDIS_CLUSTER_MIN_REJOIN_DELAY;\n\n if (new_state == REDIS_CLUSTER_OK &&\n nodeIsMaster(myself) &&\n mstime() - among_minority_time < rejoin_delay)\n {\n return;\n }\n\n /* Change the state and log the event. */\n redisLog(REDIS_WARNING,\"Cluster state changed: %s\",\n new_state == REDIS_CLUSTER_OK ? \"ok\" : \"fail\");\n\n // 设置新状态\n server.cluster->state = new_state;\n }\n}\n\n/* This function is called after the node startup in order to verify that data\n * loaded from disk is in agreement with the cluster configuration:\n *\n * 1) If we find keys about hash slots we have no responsibility for, the\n * following happens:\n * A) If no other node is in charge according to the current cluster\n * configuration, we add these slots to our node.\n * B) If according to our config other nodes are already in charge for\n * this lots, we set the slots as IMPORTING from our point of view\n * in order to justify we have those slots, and in order to make\n * redis-trib aware of the issue, so that it can try to fix it.\n * 2) If we find data in a DB different than DB0 we return REDIS_ERR to\n * signal the caller it should quit the server with an error message\n * or take other actions.\n *\n * The function always returns REDIS_OK even if it will try to correct\n * the error described in \"1\". However if data is found in DB different\n * from DB0, REDIS_ERR is returned.\n *\n * The function also uses the logging facility in order to warn the user\n * about desynchronizations between the data we have in memory and the\n * cluster configuration. */\n// 检查当前节点的节点配置是否正确，包含的数据是否正确\n// 在启动集群时被调用（看 redis.c ）\nint verifyClusterConfigWithData(void) {\n int j;\n int update_config = 0;\n\n /* If this node is a slave, don't perform the check at all as we\n * completely depend on the replication stream. */\n // 不对从节点进行检查\n if (nodeIsSlave(myself)) return REDIS_OK;\n\n /* Make sure we only have keys in DB0. */\n // 确保只有 0 号数据库有数据\n for (j = 1; j < server.dbnum; j++) {\n if (dictSize(server.db[j].dict)) return REDIS_ERR;\n }\n\n /* Check that all the slots we see populated memory have a corresponding\n * entry in the cluster table. Otherwise fix the table. */\n // 检查槽表是否都有相应的节点，如果不是的话，进行修复\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n if (!countKeysInSlot(j)) continue; /* No keys in this slot. */\n /* Check if we are assigned to this slot or if we are importing it.\n * In both cases check the next slot as the configuration makes\n * sense. */\n // 跳过正在导入的槽\n if (server.cluster->slots[j] == myself ||\n server.cluster->importing_slots_from[j] != NULL) continue;\n\n /* If we are here data and cluster config don't agree, and we have\n * slot 'j' populated even if we are not importing it, nor we are\n * assigned to this slot. Fix this condition. */\n\n update_config++;\n /* Case A: slot is unassigned. Take responsability for it. */\n if (server.cluster->slots[j] == NULL) {\n // 处理未被接受的槽\n redisLog(REDIS_WARNING, \"I've keys about slot %d that is \"\n \"unassigned. Taking responsability \"\n \"for it.\",j);\n clusterAddSlot(myself,j);\n } else {\n // 如果一个槽已经被其他节点接管\n // 那么将槽中的资料发送给对方\n redisLog(REDIS_WARNING, \"I've keys about slot %d that is \"\n \"already assigned to a different node. \"\n \"Setting it in importing state.\",j);\n server.cluster->importing_slots_from[j] = server.cluster->slots[j];\n }\n }\n\n // 保存 nodes.conf 文件\n if (update_config) clusterSaveConfigOrDie(1);\n\n return REDIS_OK;\n}\n\n/* -----------------------------------------------------------------------------\n * SLAVE nodes handling\n * -------------------------------------------------------------------------- */\n\n/* Set the specified node 'n' as master for this node.\n * If this node is currently a master, it is turned into a slave. */\n// 将节点 n 设置为当前节点的主节点\n// 如果当前节点为主节点，那么将它转换为从节点\nvoid clusterSetMaster(clusterNode *n) {\n redisAssert(n != myself);\n redisAssert(myself->numslots == 0);\n\n if (nodeIsMaster(myself)) {\n myself->flags &= ~REDIS_NODE_MASTER;\n myself->flags |= REDIS_NODE_SLAVE;\n clusterCloseAllSlots();\n } else {\n if (myself->slaveof)\n clusterNodeRemoveSlave(myself->slaveof,myself);\n }\n\n // 将 slaveof 属性指向主节点\n myself->slaveof = n;\n\n // 设置主节点的 IP 和地址，开始对它进行复制\n clusterNodeAddSlave(n,myself);\n replicationSetMaster(n->ip, n->port);\n resetManualFailover();\n}\n\n/* -----------------------------------------------------------------------------\n * CLUSTER command\n * -------------------------------------------------------------------------- */\n\n/* Generate a csv-alike representation of the specified cluster node.\n * See clusterGenNodesDescription() top comment for more information.\n *\n * The function returns the string representation as an SDS string. */\n// 生成节点的状态描述信息\nsds clusterGenNodeDescription(clusterNode *node) {\n int j, start;\n sds ci;\n\n /* Node coordinates */\n ci = sdscatprintf(sdsempty(),\"%.40s %s:%d \",\n node->name,\n node->ip,\n node->port);\n\n /* Flags */\n if (node->flags == 0) ci = sdscat(ci,\"noflags,\");\n if (node->flags & REDIS_NODE_MYSELF) ci = sdscat(ci,\"myself,\");\n if (node->flags & REDIS_NODE_MASTER) ci = sdscat(ci,\"master,\");\n if (node->flags & REDIS_NODE_SLAVE) ci = sdscat(ci,\"slave,\");\n if (node->flags & REDIS_NODE_PFAIL) ci = sdscat(ci,\"fail?,\");\n if (node->flags & REDIS_NODE_FAIL) ci = sdscat(ci,\"fail,\");\n if (node->flags & REDIS_NODE_HANDSHAKE) ci =sdscat(ci,\"handshake,\");\n if (node->flags & REDIS_NODE_NOADDR) ci = sdscat(ci,\"noaddr,\");\n if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';\n\n /* Slave of... or just \"-\" */\n if (node->slaveof)\n ci = sdscatprintf(ci,\"%.40s \",node->slaveof->name);\n else\n ci = sdscatprintf(ci,\"- \");\n\n /* Latency from the POV of this node, link status */\n ci = sdscatprintf(ci,\"%lld %lld %llu %s\",\n (long long) node->ping_sent,\n (long long) node->pong_received,\n (unsigned long long) node->configEpoch,\n (node->link || node->flags & REDIS_NODE_MYSELF) ?\n \"connected\" : \"disconnected\");\n\n /* Slots served by this instance */\n start = -1;\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n int bit;\n\n if ((bit = clusterNodeGetSlotBit(node,j)) != 0) {\n if (start == -1) start = j;\n }\n if (start != -1 && (!bit || j == REDIS_CLUSTER_SLOTS-1)) {\n if (bit && j == REDIS_CLUSTER_SLOTS-1) j++;\n\n if (start == j-1) {\n ci = sdscatprintf(ci,\" %d\",start);\n } else {\n ci = sdscatprintf(ci,\" %d-%d\",start,j-1);\n }\n start = -1;\n }\n }\n\n /* Just for MYSELF node we also dump info about slots that\n * we are migrating to other instances or importing from other\n * instances. */\n if (node->flags & REDIS_NODE_MYSELF) {\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n if (server.cluster->migrating_slots_to[j]) {\n ci = sdscatprintf(ci,\" [%d->-%.40s]\",j,\n server.cluster->migrating_slots_to[j]->name);\n } else if (server.cluster->importing_slots_from[j]) {\n ci = sdscatprintf(ci,\" [%d-<-%.40s]\",j,\n server.cluster->importing_slots_from[j]->name);\n }\n }\n }\n return ci;\n}\n\n/* Generate a csv-alike representation of the nodes we are aware of,\n * including the \"myself\" node, and return an SDS string containing the\n * representation (it is up to the caller to free it).\n *\n * 以 csv 格式记录当前节点已知所有节点的信息（包括当前节点自身），\n * 这些信息被保存到一个 sds 里面，并作为函数值返回。\n *\n * All the nodes matching at least one of the node flags specified in\n * \"filter\" are excluded from the output, so using zero as a filter will\n * include all the known nodes in the representation, including nodes in\n * the HANDSHAKE state.\n *\n * filter 参数可以用来指定节点的 flag 标识，\n * 带有被指定标识的节点不会被记录在输出结构中，\n * filter 为 0 表示记录所有节点的信息，包括 HANDSHAKE 状态的节点。\n *\n * The representation obtained using this function is used for the output\n * of the CLUSTER NODES function, and as format for the cluster\n * configuration file (nodes.conf) for a given node. \n *\n * 这个函数生成的结果会被用于 CLUSTER NODES 命令，\n * 以及用于生成 nodes.conf 配置文件。\n */\nsds clusterGenNodesDescription(int filter) {\n sds ci = sdsempty(), ni;\n dictIterator *di;\n dictEntry *de;\n\n // 遍历集群中的所有节点\n di = dictGetSafeIterator(server.cluster->nodes);\n while((de = dictNext(di)) != NULL) {\n clusterNode *node = dictGetVal(de);\n\n // 不打印包含指定 flag 的节点\n if (node->flags & filter) continue;\n\n ni = clusterGenNodeDescription(node);\n ci = sdscatsds(ci,ni);\n sdsfree(ni);\n ci = sdscatlen(ci,\"\\n\",1);\n }\n dictReleaseIterator(di);\n\n return ci;\n}\n\n// 取出一个 slot 数值\nint getSlotOrReply(redisClient *c, robj *o) {\n long long slot;\n\n if (getLongLongFromObject(o,&slot) != REDIS_OK ||\n slot < 0 || slot >= REDIS_CLUSTER_SLOTS)\n {\n addReplyError(c,\"Invalid or out of range slot\");\n return -1;\n }\n return (int) slot;\n}\n\n// CLUSTER 命令的实现\nvoid clusterCommand(redisClient *c) {\n\n // 不能在非集群模式下使用该命令\n if (server.cluster_enabled == 0) {\n addReplyError(c,\"This instance has cluster support disabled\");\n return;\n }\n\n if (!strcasecmp(c->argv[1]->ptr,\"meet\") && c->argc == 4) {\n /* CLUSTER MEET */\n // 将给定地址的节点添加到当前节点所处的集群里面\n\n long long port;\n\n // 检查 port 参数的合法性\n if (getLongLongFromObject(c->argv[3], &port) != REDIS_OK) {\n addReplyErrorFormat(c,\"Invalid TCP port specified: %s\",\n (char*)c->argv[3]->ptr);\n return;\n }\n\n // 尝试与给定地址的节点进行连接\n if (clusterStartHandshake(c->argv[2]->ptr,port) == 0 &&\n errno == EINVAL)\n {\n // 连接失败\n addReplyErrorFormat(c,\"Invalid node address specified: %s:%s\",\n (char*)c->argv[2]->ptr, (char*)c->argv[3]->ptr);\n } else {\n // 连接成功\n addReply(c,shared.ok);\n }\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"nodes\") && c->argc == 2) {\n /* CLUSTER NODES */\n // 列出集群所有节点的信息\n robj *o;\n sds ci = clusterGenNodesDescription(0);\n\n o = createObject(REDIS_STRING,ci);\n addReplyBulk(c,o);\n decrRefCount(o);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"flushslots\") && c->argc == 2) {\n /* CLUSTER FLUSHSLOTS */\n // 删除当前节点的所有槽，让它变为不处理任何槽\n\n // 删除槽必须在数据库为空的情况下进行\n if (dictSize(server.db[0].dict) != 0) {\n addReplyError(c,\"DB must be empty to perform CLUSTER FLUSHSLOTS.\");\n return;\n }\n // 删除所有由该节点处理的槽\n clusterDelNodeSlots(myself);\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n addReply(c,shared.ok);\n\n } else if ((!strcasecmp(c->argv[1]->ptr,\"addslots\") ||\n !strcasecmp(c->argv[1]->ptr,\"delslots\")) && c->argc >= 3)\n {\n /* CLUSTER ADDSLOTS [slot] ... */\n // 将一个或多个 slot 添加到当前节点\n\n /* CLUSTER DELSLOTS [slot] ... */\n // 从当前节点中删除一个或多个 slot\n \n int j, slot;\n\n // 一个数组，记录所有要添加或者删除的槽\n unsigned char *slots = zmalloc(REDIS_CLUSTER_SLOTS);\n\n // 检查这是 delslots 还是 addslots\n int del = !strcasecmp(c->argv[1]->ptr,\"delslots\");\n\n // 将 slots 数组的所有值设置为 0\n memset(slots,0,REDIS_CLUSTER_SLOTS);\n\n /* Check that all the arguments are parsable and that all the\n * slots are not already busy. */\n // 处理所有输入 slot 参数\n for (j = 2; j < c->argc; j++) {\n\n // 获取 slot 数字\n if ((slot = getSlotOrReply(c,c->argv[j])) == -1) {\n zfree(slots);\n return;\n }\n\n // 如果这是 delslots 命令，并且指定槽为未指定，那么返回一个错误\n if (del && server.cluster->slots[slot] == NULL) {\n addReplyErrorFormat(c,\"Slot %d is already unassigned\", slot);\n zfree(slots);\n return;\n // 如果这是 addslots 命令，并且槽已经有节点在负责，那么返回一个错误\n } else if (!del && server.cluster->slots[slot]) {\n addReplyErrorFormat(c,\"Slot %d is already busy\", slot);\n zfree(slots);\n return;\n }\n\n // 如果某个槽指定了一次以上，那么返回一个错误\n if (slots[slot]++ == 1) {\n addReplyErrorFormat(c,\"Slot %d specified multiple times\",\n (int)slot);\n zfree(slots);\n return;\n }\n }\n\n // 处理所有输入 slot\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n if (slots[j]) {\n int retval;\n\n /* If this slot was set as importing we can clear this \n * state as now we are the real owner of the slot. */\n // 如果指定 slot 之前的状态为载入状态，那么现在可以清除这一状态\n // 因为当前节点现在已经是 slot 的负责人了\n if (server.cluster->importing_slots_from[j])\n server.cluster->importing_slots_from[j] = NULL;\n\n // 添加或者删除指定 slot\n retval = del ? clusterDelSlot(j) :\n clusterAddSlot(myself,j);\n redisAssertWithInfo(c,NULL,retval == REDIS_OK);\n }\n }\n zfree(slots);\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n addReply(c,shared.ok);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"setslot\") && c->argc >= 4) {\n /* SETSLOT 10 MIGRATING */\n /* SETSLOT 10 IMPORTING */\n /* SETSLOT 10 STABLE */\n /* SETSLOT 10 NODE */\n int slot;\n clusterNode *n;\n\n // 取出 slot 值\n if ((slot = getSlotOrReply(c,c->argv[2])) == -1) return;\n\n // CLUSTER SETSLOT MIGRATING \n // 将本节点的槽 slot 迁移至 node id 所指定的节点\n if (!strcasecmp(c->argv[3]->ptr,\"migrating\") && c->argc == 5) {\n // 被迁移的槽必须属于本节点\n if (server.cluster->slots[slot] != myself) {\n addReplyErrorFormat(c,\"I'm not the owner of hash slot %u\",slot);\n return;\n }\n\n // 迁移的目标节点必须是本节点已知的\n if ((n = clusterLookupNode(c->argv[4]->ptr)) == NULL) {\n addReplyErrorFormat(c,\"I don't know about node %s\",\n (char*)c->argv[4]->ptr);\n return;\n }\n\n // 为槽设置迁移目标节点\n server.cluster->migrating_slots_to[slot] = n;\n\n // CLUSTER SETSLOT IMPORTING \n // 从节点 node id 中导入槽 slot 到本节点\n } else if (!strcasecmp(c->argv[3]->ptr,\"importing\") && c->argc == 5) {\n\n // 如果 slot 槽本身已经由本节点处理，那么无须进行导入\n if (server.cluster->slots[slot] == myself) {\n addReplyErrorFormat(c,\n \"I'm already the owner of hash slot %u\",slot);\n return;\n }\n\n // node id 指定的节点必须是本节点已知的，这样才能从目标节点导入槽\n if ((n = clusterLookupNode(c->argv[4]->ptr)) == NULL) {\n addReplyErrorFormat(c,\"I don't know about node %s\",\n (char*)c->argv[3]->ptr);\n return;\n }\n\n // 为槽设置导入目标节点\n server.cluster->importing_slots_from[slot] = n;\n\n } else if (!strcasecmp(c->argv[3]->ptr,\"stable\") && c->argc == 4) {\n /* CLUSTER SETSLOT STABLE */\n // 取消对槽 slot 的迁移或者导入\n\n server.cluster->importing_slots_from[slot] = NULL;\n server.cluster->migrating_slots_to[slot] = NULL;\n\n } else if (!strcasecmp(c->argv[3]->ptr,\"node\") && c->argc == 5) {\n /* CLUSTER SETSLOT NODE */\n // 将未指派 slot 指派给 node id 指定的节点\n\n // 查找目标节点\n clusterNode *n = clusterLookupNode(c->argv[4]->ptr);\n\n // 目标节点必须已存在\n if (!n) {\n addReplyErrorFormat(c,\"Unknown node %s\",\n (char*)c->argv[4]->ptr);\n return;\n }\n\n /* If this hash slot was served by 'myself' before to switch\n * make sure there are no longer local keys for this hash slot. */\n // 如果这个槽之前由当前节点负责处理，那么必须保证槽里面没有键存在\n if (server.cluster->slots[slot] == myself && n != myself) {\n if (countKeysInSlot(slot) != 0) {\n addReplyErrorFormat(c,\n \"Can't assign hashslot %d to a different node \"\n \"while I still hold keys for this hash slot.\", slot);\n return;\n }\n }\n /* If this slot is in migrating status but we have no keys\n * for it assigning the slot to another node will clear\n * the migratig status. */\n if (countKeysInSlot(slot) == 0 &&\n server.cluster->migrating_slots_to[slot])\n server.cluster->migrating_slots_to[slot] = NULL;\n\n /* If this node was importing this slot, assigning the slot to\n * itself also clears the importing status. */\n // 撤销本节点对 slot 的导入计划\n if (n == myself &&\n server.cluster->importing_slots_from[slot])\n {\n /* This slot was manually migrated, set this node configEpoch\n * to a new epoch so that the new version can be propagated\n * by the cluster.\n *\n * Note that if this ever results in a collision with another\n * node getting the same configEpoch, for example because a\n * failover happens at the same time we close the slot, the\n * configEpoch collision resolution will fix it assigning\n * a different epoch to each node. */\n uint64_t maxEpoch = clusterGetMaxEpoch();\n\n if (myself->configEpoch == 0 ||\n myself->configEpoch != maxEpoch)\n {\n server.cluster->currentEpoch++;\n myself->configEpoch = server.cluster->currentEpoch;\n clusterDoBeforeSleep(CLUSTER_TODO_FSYNC_CONFIG);\n redisLog(REDIS_WARNING,\n \"configEpoch set to %llu after importing slot %d\",\n (unsigned long long) myself->configEpoch, slot);\n }\n server.cluster->importing_slots_from[slot] = NULL;\n }\n\n // 将槽设置为未指派\n clusterDelSlot(slot);\n\n // 将槽指派给目标节点\n clusterAddSlot(n,slot);\n\n } else {\n addReplyError(c,\n \"Invalid CLUSTER SETSLOT action or number of arguments\");\n return;\n }\n clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|CLUSTER_TODO_UPDATE_STATE);\n addReply(c,shared.ok);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"info\") && c->argc == 2) {\n /* CLUSTER INFO */\n // 打印出集群的当前信息\n\n char *statestr[] = {\"ok\",\"fail\",\"needhelp\"};\n int slots_assigned = 0, slots_ok = 0, slots_pfail = 0, slots_fail = 0;\n int j;\n\n // 统计集群中的已指派节点、已下线节点、疑似下线节点和正常节点的数量\n for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {\n clusterNode *n = server.cluster->slots[j];\n\n // 跳过未指派节点\n if (n == NULL) continue;\n\n // 统计已指派节点的数量\n slots_assigned++;\n if (nodeFailed(n)) {\n slots_fail++;\n } else if (nodeTimedOut(n)) {\n slots_pfail++;\n } else {\n // 正常节点\n slots_ok++;\n }\n }\n\n // 打印信息\n sds info = sdscatprintf(sdsempty(),\n \"cluster_state:%s\\r\\n\"\n \"cluster_slots_assigned:%d\\r\\n\"\n \"cluster_slots_ok:%d\\r\\n\"\n \"cluster_slots_pfail:%d\\r\\n\"\n \"cluster_slots_fail:%d\\r\\n\"\n \"cluster_known_nodes:%lu\\r\\n\"\n \"cluster_size:%d\\r\\n\"\n \"cluster_current_epoch:%llu\\r\\n\"\n \"cluster_stats_messages_sent:%lld\\r\\n\"\n \"cluster_stats_messages_received:%lld\\r\\n\"\n , statestr[server.cluster->state],\n slots_assigned,\n slots_ok,\n slots_pfail,\n slots_fail,\n dictSize(server.cluster->nodes),\n server.cluster->size,\n (unsigned long long) server.cluster->currentEpoch,\n server.cluster->stats_bus_messages_sent,\n server.cluster->stats_bus_messages_received\n );\n addReplySds(c,sdscatprintf(sdsempty(),\"$%lu\\r\\n\",\n (unsigned long)sdslen(info)));\n addReplySds(c,info);\n addReply(c,shared.crlf);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"saveconfig\") && c->argc == 2) {\n // CLUSTER SAVECONFIG 命令\n // 将 nodes.conf 文件保存到磁盘里面\n\n // 保存\n int retval = clusterSaveConfig(1);\n\n // 检查错误\n if (retval == 0)\n addReply(c,shared.ok);\n else\n addReplyErrorFormat(c,\"error saving the cluster node config: %s\",\n strerror(errno));\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"keyslot\") && c->argc == 3) {\n /* CLUSTER KEYSLOT */\n // 返回 key 应该被 hash 到那个槽上\n\n sds key = c->argv[2]->ptr;\n\n addReplyLongLong(c,keyHashSlot(key,sdslen(key)));\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"countkeysinslot\") && c->argc == 3) {\n /* CLUSTER COUNTKEYSINSLOT */\n // 计算指定 slot 上的键数量\n\n long long slot;\n\n // 取出 slot 参数\n if (getLongLongFromObjectOrReply(c,c->argv[2],&slot,NULL) != REDIS_OK)\n return;\n if (slot < 0 || slot >= REDIS_CLUSTER_SLOTS) {\n addReplyError(c,\"Invalid slot\");\n return;\n }\n\n addReplyLongLong(c,countKeysInSlot(slot));\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"getkeysinslot\") && c->argc == 4) {\n /* CLUSTER GETKEYSINSLOT */\n // 打印 count 个属于 slot 槽的键\n\n long long maxkeys, slot;\n unsigned int numkeys, j;\n robj **keys;\n\n // 取出 slot 参数\n if (getLongLongFromObjectOrReply(c,c->argv[2],&slot,NULL) != REDIS_OK)\n return;\n // 取出 count 参数\n if (getLongLongFromObjectOrReply(c,c->argv[3],&maxkeys,NULL)\n != REDIS_OK)\n return;\n // 检查参数的合法性\n if (slot < 0 || slot >= REDIS_CLUSTER_SLOTS || maxkeys < 0) {\n addReplyError(c,\"Invalid slot or number of keys\");\n return;\n }\n\n // 分配一个保存键的数组\n keys = zmalloc(sizeof(robj*)*maxkeys);\n // 将键记录到 keys 数组\n numkeys = getKeysInSlot(slot, keys, maxkeys);\n\n // 打印获得的键\n addReplyMultiBulkLen(c,numkeys);\n for (j = 0; j < numkeys; j++) addReplyBulk(c,keys[j]);\n zfree(keys);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"forget\") && c->argc == 3) {\n /* CLUSTER FORGET */\n // 从集群中删除 NODE_ID 指定的节点\n\n // 查找 NODE_ID 指定的节点\n clusterNode *n = clusterLookupNode(c->argv[2]->ptr);\n\n // 该节点不存在于集群中\n if (!n) {\n addReplyErrorFormat(c,\"Unknown node %s\", (char*)c->argv[2]->ptr);\n return;\n } else if (n == myself) {\n addReplyError(c,\"I tried hard but I can't forget myself...\");\n return;\n } else if (nodeIsSlave(myself) && myself->slaveof == n) {\n addReplyError(c,\"Can't forget my master!\");\n return;\n }\n\n // 将集群添加到黑名单\n clusterBlacklistAddNode(n);\n // 从集群中删除该节点\n clusterDelNode(n);\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_SAVE_CONFIG);\n addReply(c,shared.ok);\n\n } else if (!strcasecmp(c->argv[1]->ptr,\"replicate\") && c->argc == 3) {\n /* CLUSTER REPLICATE */\n // 将当前节点设置为 NODE_ID 指定的节点的从节点（复制品）\n\n // 根据名字查找节点\n clusterNode *n = clusterLookupNode(c->argv[2]->ptr);\n\n /* Lookup the specified node in our table. */\n if (!n) {\n addReplyErrorFormat(c,\"Unknown node %s\", (char*)c->argv[2]->ptr);\n return;\n }\n\n /* I can't replicate myself. */\n // 指定节点是自己，不能进行复制\n if (n == myself) {\n addReplyError(c,\"Can't replicate myself\");\n return;\n }\n\n /* Can't replicate a slave. */\n // 不能复制一个从节点\n if (n->slaveof != NULL) {\n addReplyError(c,\"I can only replicate a master, not a slave.\");\n return;\n }\n\n /* If the instance is currently a master, it should have no assigned\n * slots nor keys to accept to replicate some other node.\n * Slaves can switch to another master without issues. */\n // 节点必须没有被指派任何槽，并且数据库必须为空\n if (nodeIsMaster(myself) &&\n (myself->numslots != 0 || dictSize(server.db[0].dict) != 0)) {\n addReplyError(c,\n \"To set a master the node must be empty and \"\n \"without assigned slots.\");\n return;\n }\n\n /* Set the master. */\n // 将节点 n 设为本节点的主节点\n clusterSetMaster(n);\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);\n addReply(c,shared.ok);\n } else if (!strcasecmp(c->argv[1]->ptr,\"slaves\") && c->argc == 3) {\n /* CLUSTER SLAVES */\n // 打印给定主节点的所有从节点的信息\n\n clusterNode *n = clusterLookupNode(c->argv[2]->ptr);\n int j;\n\n /* Lookup the specified node in our table. */\n if (!n) {\n addReplyErrorFormat(c,\"Unknown node %s\", (char*)c->argv[2]->ptr);\n return;\n }\n\n if (nodeIsSlave(n)) {\n addReplyError(c,\"The specified node is not a master\");\n return;\n }\n\n addReplyMultiBulkLen(c,n->numslaves);\n for (j = 0; j < n->numslaves; j++) {\n sds ni = clusterGenNodeDescription(n->slaves[j]);\n addReplyBulkCString(c,ni);\n sdsfree(ni);\n }\n } else if (!strcasecmp(c->argv[1]->ptr,\"failover\") &&\n (c->argc == 2 || c->argc == 3))\n {\n /* CLUSTER FAILOVER [FORCE] */\n // 执行手动故障转移\n\n int force = 0;\n\n if (c->argc == 3) {\n if (!strcasecmp(c->argv[2]->ptr,\"force\")) {\n force = 1;\n } else {\n addReply(c,shared.syntaxerr);\n return;\n }\n }\n\n // 命令只能发送给从节点\n if (nodeIsMaster(myself)) {\n addReplyError(c,\"You should send CLUSTER FAILOVER to a slave\");\n return;\n } else if (!force &&\n (myself->slaveof == NULL || nodeFailed(myself->slaveof) ||\n myself->slaveof->link == NULL))\n {\n // 如果主节点已下线或者处于失效状态\n // 并且命令没有给定 force 参数，那么命令执行失败\n addReplyError(c,\"Master is down or failed, \"\n \"please use CLUSTER FAILOVER FORCE\");\n return;\n }\n\n // 重置手动故障转移的有关属性\n resetManualFailover();\n // 设定手动故障转移的最大执行时限\n server.cluster->mf_end = mstime() + REDIS_CLUSTER_MF_TIMEOUT;\n\n /* If this is a forced failover, we don't need to talk with our master\n * to agree about the offset. We just failover taking over it without\n * coordination. */\n // 如果这是强制的手动 failover ，那么直接开始 failover ，\n // 无须向其他 master 沟通偏移量。\n if (force) {\n // 如果这是强制的手动故障转移，那么直接开始执行故障转移操作\n server.cluster->mf_can_start = 1;\n } else {\n // 如果不是强制的话，那么需要和主节点比对相互的偏移量是否一致\n clusterSendMFStart(myself->slaveof);\n }\n redisLog(REDIS_WARNING,\"Manual failover user request accepted.\");\n addReply(c,shared.ok);\n } else if (!strcasecmp(c->argv[1]->ptr,\"set-config-epoch\") && c->argc == 3)\n {\n /* CLUSTER SET-CONFIG-EPOCH \n *\n * The user is allowed to set the config epoch only when a node is\n * totally fresh: no config epoch, no other known node, and so forth.\n * This happens at cluster creation time to start with a cluster where\n * every node has a different node ID, without to rely on the conflicts\n * resolution system which is too slow when a big cluster is created. */\n long long epoch;\n\n if (getLongLongFromObjectOrReply(c,c->argv[2],&epoch,NULL) != REDIS_OK)\n return;\n\n if (epoch < 0) {\n addReplyErrorFormat(c,\"Invalid config epoch specified: %lld\",epoch);\n } else if (dictSize(server.cluster->nodes) > 1) {\n addReplyError(c,\"The user can assign a config epoch only when the \"\n \"node does not know any other node.\");\n } else if (myself->configEpoch != 0) {\n addReplyError(c,\"Node config epoch is already non-zero\");\n } else {\n myself->configEpoch = epoch;\n /* No need to fsync the config here since in the unlucky event\n * of a failure to persist the config, the conflict resolution code\n * will assign an unique config to this node. */\n clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|\n CLUSTER_TODO_SAVE_CONFIG);\n addReply(c,shared.ok);\n }\n } else if (!strcasecmp(c->argv[1]->ptr,\"reset\") &&\n (c->argc == 2 || c->argc == 3))\n {\n /* CLUSTER RESET [SOFT|HARD] */\n int hard = 0;\n\n /* Parse soft/hard argument. Default is soft. */\n if (c->argc == 3) {\n if (!strcasecmp(c->argv[2]->ptr,\"hard\")) {\n hard = 1;\n } else if (!strcasecmp(c->argv[2]->ptr,\"soft\")) {\n hard = 0;\n } else {\n addReply(c,shared.syntaxerr);\n return;\n }\n }\n\n /* Slaves can be reset while containing data, but not master nodes\n * that must be empty. */\n if (nodeIsMaster(myself) && dictSize(c->db->dict) != 0) {\n addReplyError(c,\"CLUSTER RESET can't be called with \"\n \"master nodes containing keys\");\n return;\n }\n clusterReset(hard);\n addReply(c,shared.ok);\n } else {\n addReplyError(c,\"Wrong CLUSTER subcommand or number of arguments\");\n }\n}\n\n/* -----------------------------------------------------------------------------\n * DUMP, RESTORE and MIGRATE commands\n * -------------------------------------------------------------------------- */\n\n/* Generates a DUMP-format representation of the object 'o', adding it to the\n * io stream pointed by 'rio'. This function can't fail. \n *\n * 创建对象 o 的一个 DUMP 格式表示，\n * 并将它添加到 rio 指针指向的 io 流当中。\n */\nvoid createDumpPayload(rio *payload, robj *o) {\n unsigned char buf[2];\n uint64_t crc;\n\n /* Serialize the object in a RDB-like format. It consist of an object type\n * byte followed by the serialized object. This is understood by RESTORE. */\n // 将对象序列化为一个 RDB 格式对象\n // 序列化对象以对象类型为首，后跟序列化后的对象\n // 如图\n //\n // |<-- RDB payload -->|\n // 序列化数据\n // +-------------+------+\n // | 1 byte type | obj |\n // +-------------+------+\n rioInitWithBuffer(payload,sdsempty());\n redisAssert(rdbSaveObjectType(payload,o));\n redisAssert(rdbSaveObject(payload,o));\n\n /* Write the footer, this is how it looks like:\n * ----------------+---------------------+---------------+\n * ... RDB payload | 2 bytes RDB version | 8 bytes CRC64 |\n * ----------------+---------------------+---------------+\n * RDB version and CRC are both in little endian.\n */\n\n /* RDB version */\n // 写入 RDB 版本\n buf[0] = REDIS_RDB_VERSION & 0xff;\n buf[1] = (REDIS_RDB_VERSION >> 8) & 0xff;\n payload->io.buffer.ptr = sdscatlen(payload->io.buffer.ptr,buf,2);\n\n /* CRC64 */\n // 写入 CRC 校验和\n crc = crc64(0,(unsigned char*)payload->io.buffer.ptr,\n sdslen(payload->io.buffer.ptr));\n memrev64ifbe(&crc);\n payload->io.buffer.ptr = sdscatlen(payload->io.buffer.ptr,&crc,8);\n\n // 整个数据的结构:\n //\n // | <--- 序列化数据 -->|\n // +-------------+------+---------------------+---------------+\n // | 1 byte type | obj | 2 bytes RDB version | 8 bytes CRC64 |\n // +-------------+------+---------------------+---------------+\n\n}\n\n/* Verify that the RDB version of the dump payload matches the one of this Redis\n * instance and that the checksum is ok.\n *\n * 检查输入的 DUMP 数据中， RDB 版本是否和当前 Redis 实例所使用的 RDB 版本相同，\n * 并检查校验和是否正确。\n *\n * If the DUMP payload looks valid REDIS_OK is returned, otherwise REDIS_ERR\n * is returned. \n *\n * 检查正常返回 REDIS_OK ，否则返回 REDIS_ERR 。\n */\nint verifyDumpPayload(unsigned char *p, size_t len) {\n unsigned char *footer;\n uint16_t rdbver;\n uint64_t crc;\n\n /* At least 2 bytes of RDB version and 8 of CRC64 should be present. */\n // 因为序列化数据至少包含 2 个字节的 RDB 版本\n // 以及 8 个字节的 CRC64 校验和\n // 所以序列化数据不可能少于 10 个字节\n if (len < 10) return REDIS_ERR;\n\n // 指向数据的最后 10 个字节\n footer = p+(len-10);\n\n /* Verify RDB version */\n // 检查序列化数据的版本号，看是否和当前实例使用的版本号一致\n rdbver = (footer[1] << 8) | footer[0];\n if (rdbver != REDIS_RDB_VERSION) return REDIS_ERR;\n\n /* Verify CRC64 */\n // 检查数据的 CRC64 校验和是否正确\n crc = crc64(0,p,len-8);\n memrev64ifbe(&crc);\n return (memcmp(&crc,footer+2,8) == 0) ? REDIS_OK : REDIS_ERR;\n}\n\n/* DUMP keyname\n * DUMP is actually not used by Redis Cluster but it is the obvious\n * complement of RESTORE and can be useful for different applications. */\nvoid dumpCommand(redisClient *c) {\n robj *o, *dumpobj;\n rio payload;\n\n /* Check if the key is here. */\n // 取出给定键的值\n if ((o = lookupKeyRead(c->db,c->argv[1])) == NULL) {\n addReply(c,shared.nullbulk);\n return;\n }\n\n /* Create the DUMP encoded representation. */\n // 创建给定值的一个 DUMP 编码表示\n createDumpPayload(&payload,o);\n\n /* Transfer to the client */\n // 将编码后的键值对数据返回给客户端\n dumpobj = createObject(REDIS_STRING,payload.io.buffer.ptr);\n addReplyBulk(c,dumpobj);\n decrRefCount(dumpobj);\n\n return;\n}\n\n/* RESTORE key ttl serialized-value [REPLACE] */\n// 根据给定的 DUMP 数据，还原出一个键值对数据，并将它保存到数据库里面\nvoid restoreCommand(redisClient *c) {\n long long ttl;\n rio payload;\n int j, type, replace = 0;\n robj *obj;\n\n /* Parse additional options */\n // 是否使用了 REPLACE 选项？\n for (j = 4; j < c->argc; j++) {\n if (!strcasecmp(c->argv[j]->ptr,\"replace\")) {\n replace = 1;\n } else {\n addReply(c,shared.syntaxerr);\n return;\n }\n }\n\n /* Make sure this key does not already exist here... */\n // 如果没有给定 REPLACE 选项，并且键已经存在，那么返回错误\n if (!replace && lookupKeyWrite(c->db,c->argv[1]) != NULL) {\n addReply(c,shared.busykeyerr);\n return;\n }\n\n /* Check if the TTL value makes sense */\n // 取出（可能有的） TTL 值\n if (getLongLongFromObjectOrReply(c,c->argv[2],&ttl,NULL) != REDIS_OK) {\n return;\n } else if (ttl < 0) {\n addReplyError(c,\"Invalid TTL value, must be >= 0\");\n return;\n }\n\n /* Verify RDB version and data checksum. */\n // 检查 RDB 版本和校验和\n if (verifyDumpPayload(c->argv[3]->ptr,sdslen(c->argv[3]->ptr)) == REDIS_ERR)\n {\n addReplyError(c,\"DUMP payload version or checksum are wrong\");\n return;\n }\n\n // 读取 DUMP 数据，并反序列化出键值对的类型和值\n rioInitWithBuffer(&payload,c->argv[3]->ptr);\n if (((type = rdbLoadObjectType(&payload)) == -1) ||\n ((obj = rdbLoadObject(type,&payload)) == NULL))\n {\n addReplyError(c,\"Bad data format\");\n return;\n }\n\n /* Remove the old key if needed. */\n // 如果给定了 REPLACE 选项，那么先删除数据库中已存在的同名键\n if (replace) dbDelete(c->db,c->argv[1]);\n\n /* Create the key and set the TTL if any */\n // 将键值对添加到数据库\n dbAdd(c->db,c->argv[1],obj);\n\n // 如果键带有 TTL 的话，设置键的 TTL\n if (ttl) setExpire(c->db,c->argv[1],mstime()+ttl);\n\n signalModifiedKey(c->db,c->argv[1]);\n\n addReply(c,shared.ok);\n server.dirty++;\n}\n\n/* MIGRATE socket cache implementation.\n *\n * MIGRATE 套接字缓存实现\n *\n * We take a map between host:ip and a TCP socket that we used to connect\n * to this instance in recent time.\n *\n * 保存一个字典，字典的键为 host:ip ，值为最近使用的连接向指定地址的 TCP 套接字。\n *\n * This sockets are closed when the max number we cache is reached, and also\n * in serverCron() when they are around for more than a few seconds. \n *\n * 这个字典在缓存数达到上限时被释放，\n * 并且 serverCron() 也会定期删除字典中的一些过期套接字。\n */\n// 最大缓存数\n#define MIGRATE_SOCKET_CACHE_ITEMS 64 /* max num of items in the cache. */\n// 套接字保质期（超过这个时间的套接字会被删除）\n#define MIGRATE_SOCKET_CACHE_TTL 10 /* close cached socekts after 10 sec. */\n\ntypedef struct migrateCachedSocket {\n\n // 套接字描述符\n int fd;\n\n // 最后一次使用的时间\n time_t last_use_time;\n\n} migrateCachedSocket;\n\n/* Return a TCP scoket connected with the target instance, possibly returning\n * a cached one.\n *\n * 返回一个连接向指定地址的 TCP 套接字，这个套接字可能是一个缓存套接字。\n *\n * This function is responsible of sending errors to the client if a\n * connection can't be established. In this case -1 is returned.\n * Otherwise on success the socket is returned, and the caller should not\n * attempt to free it after usage.\n *\n * 如果连接出错，那么函数返回 -1 。\n * 如果连接正常，那么函数返回 TCP 套接字描述符。\n *\n * If the caller detects an error while using the socket, migrateCloseSocket()\n * should be called so that the connection will be craeted from scratch\n * the next time. \n *\n * 如果调用者在使用这个函数返回的套接字时遇上错误，\n * 那么调用者会使用 migrateCloseSocket() 来关闭出错的套接字，\n * 这样下次要连接相同地址时，服务器就会创建新的套接字来进行连接。\n */\nint migrateGetSocket(redisClient *c, robj *host, robj *port, long timeout) {\n int fd;\n sds name = sdsempty();\n migrateCachedSocket *cs;\n\n /* Check if we have an already cached socket for this ip:port pair. */\n // 根据 ip 和 port 创建地址名字\n name = sdscatlen(name,host->ptr,sdslen(host->ptr));\n name = sdscatlen(name,\":\",1);\n name = sdscatlen(name,port->ptr,sdslen(port->ptr));\n \n // 在套接字缓存中查找套接字是否已经存在\n cs = dictFetchValue(server.migrate_cached_sockets,name);\n // 缓存存在，更新最后一次使用时间，以免它被当作过期套接字而被释放\n if (cs) {\n sdsfree(name);\n cs->last_use_time = server.unixtime;\n return cs->fd;\n }\n\n /* No cached socket, create one. */\n // 没有缓存，创建一个新的缓存\n if (dictSize(server.migrate_cached_sockets) == MIGRATE_SOCKET_CACHE_ITEMS) {\n\n // 如果缓存数已经达到上线，那么在创建套接字之前，先随机删除一个连接\n\n /* Too many items, drop one at random. */\n dictEntry *de = dictGetRandomKey(server.migrate_cached_sockets);\n cs = dictGetVal(de);\n close(cs->fd);\n zfree(cs);\n dictDelete(server.migrate_cached_sockets,dictGetKey(de));\n }\n\n /* Create the socket */\n // 创建连接\n fd = anetTcpNonBlockConnect(server.neterr,c->argv[1]->ptr,\n atoi(c->argv[2]->ptr));\n if (fd == -1) {\n sdsfree(name);\n addReplyErrorFormat(c,\"Can't connect to target node: %s\",\n server.neterr);\n return -1;\n }\n anetEnableTcpNoDelay(server.neterr,fd);\n\n /* Check if it connects within the specified timeout. */\n // 检查连接的超时设置\n if ((aeWait(fd,AE_WRITABLE,timeout) & AE_WRITABLE) == 0) {\n sdsfree(name);\n addReplySds(c,\n sdsnew(\"-IOERR error or timeout connecting to the client\\r\\n\"));\n close(fd);\n return -1;\n }\n\n /* Add to the cache and return it to the caller. */\n // 将连接添加到缓存\n cs = zmalloc(sizeof(*cs));\n cs->fd = fd;\n cs->last_use_time = server.unixtime;\n dictAdd(server.migrate_cached_sockets,name,cs);\n\n return fd;\n}\n\n/* Free a migrate cached connection. */\n// 释放一个缓存连接\nvoid migrateCloseSocket(robj *host, robj *port) {\n sds name = sdsempty();\n migrateCachedSocket *cs;\n\n // 根据 ip 和 port 创建连接的名字\n name = sdscatlen(name,host->ptr,sdslen(host->ptr));\n name = sdscatlen(name,\":\",1);\n name = sdscatlen(name,port->ptr,sdslen(port->ptr));\n // 查找连接\n cs = dictFetchValue(server.migrate_cached_sockets,name);\n if (!cs) {\n sdsfree(name);\n return;\n }\n\n // 关闭连接\n close(cs->fd);\n zfree(cs);\n\n // 从缓存中删除该连接\n dictDelete(server.migrate_cached_sockets,name);\n sdsfree(name);\n}\n\n// 移除过期的连接，由 redis.c/serverCron() 调用\nvoid migrateCloseTimedoutSockets(void) {\n dictIterator *di = dictGetSafeIterator(server.migrate_cached_sockets);\n dictEntry *de;\n\n while((de = dictNext(di)) != NULL) {\n migrateCachedSocket *cs = dictGetVal(de);\n\n // 如果套接字最后一次使用的时间已经超过 MIGRATE_SOCKET_CACHE_TTL \n // 那么表示该套接字过期，释放它！\n if ((server.unixtime - cs->last_use_time) > MIGRATE_SOCKET_CACHE_TTL) {\n close(cs->fd);\n zfree(cs);\n dictDelete(server.migrate_cached_sockets,dictGetKey(de));\n }\n }\n dictReleaseIterator(di);\n}\n\n/* MIGRATE host port key dbid timeout [COPY | REPLACE] */\nvoid migrateCommand(redisClient *c) {\n int fd, copy, replace, j;\n long timeout;\n long dbid;\n long long ttl, expireat;\n robj *o;\n rio cmd, payload;\n int retry_num = 0;\n\ntry_again:\n /* Initialization */\n copy = 0;\n replace = 0;\n ttl = 0;\n\n /* Parse additional options */\n // 读入 COPY 或者 REPLACE 选项\n for (j = 6; j < c->argc; j++) {\n if (!strcasecmp(c->argv[j]->ptr,\"copy\")) {\n copy = 1;\n } else if (!strcasecmp(c->argv[j]->ptr,\"replace\")) {\n replace = 1;\n } else {\n addReply(c,shared.syntaxerr);\n return;\n }\n }\n\n /* Sanity check */\n // 检查输入参数的正确性\n if (getLongFromObjectOrReply(c,c->argv[5],&timeout,NULL) != REDIS_OK)\n return;\n if (getLongFromObjectOrReply(c,c->argv[4],&dbid,NULL) != REDIS_OK)\n return;\n if (timeout <= 0) timeout = 1000;\n\n /* Check if the key is here. If not we reply with success as there is\n * nothing to migrate (for instance the key expired in the meantime), but\n * we include such information in the reply string. */\n // 取出键的值对象\n if ((o = lookupKeyRead(c->db,c->argv[3])) == NULL) {\n addReplySds(c,sdsnew(\"+NOKEY\\r\\n\"));\n return;\n }\n\n /* Connect */\n // 获取套接字连接\n fd = migrateGetSocket(c,c->argv[1],c->argv[2],timeout);\n if (fd == -1) return; /* error sent to the client by migrateGetSocket() */\n\n /* Create RESTORE payload and generate the protocol to call the command. */\n // 创建用于指定数据库的 SELECT 命令，以免键值对被还原到了错误的地方\n rioInitWithBuffer(&cmd,sdsempty());\n redisAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',2));\n redisAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,\"SELECT\",6));\n redisAssertWithInfo(c,NULL,rioWriteBulkLongLong(&cmd,dbid));\n\n // 取出键的过期时间戳\n expireat = getExpire(c->db,c->argv[3]);\n if (expireat != -1) {\n ttl = expireat-mstime();\n if (ttl < 1) ttl = 1;\n }\n redisAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',replace ? 5 : 4));\n\n // 如果运行在集群模式下，那么发送的命令为 RESTORE-ASKING\n // 如果运行在非集群模式下，那么发送的命令为 RESTORE\n if (server.cluster_enabled)\n redisAssertWithInfo(c,NULL,\n rioWriteBulkString(&cmd,\"RESTORE-ASKING\",14));\n else\n redisAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,\"RESTORE\",7));\n\n // 写入键名和过期时间\n redisAssertWithInfo(c,NULL,sdsEncodedObject(c->argv[3]));\n redisAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,c->argv[3]->ptr,\n sdslen(c->argv[3]->ptr)));\n redisAssertWithInfo(c,NULL,rioWriteBulkLongLong(&cmd,ttl));\n\n /* Emit the payload argument, that is the serialized object using\n * the DUMP format. */\n // 将值对象进行序列化\n createDumpPayload(&payload,o);\n // 写入序列化对象\n redisAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,payload.io.buffer.ptr,\n sdslen(payload.io.buffer.ptr)));\n sdsfree(payload.io.buffer.ptr);\n\n /* Add the REPLACE option to the RESTORE command if it was specified\n * as a MIGRATE option. */\n // 是否设置了 REPLACE 命令？\n if (replace)\n // 写入 REPLACE 参数\n redisAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,\"REPLACE\",7));\n\n /* Transfer the query to the other node in 64K chunks. */\n // 以 64 kb 每次的大小向对方发送数据\n errno = 0;\n {\n sds buf = cmd.io.buffer.ptr;\n size_t pos = 0, towrite;\n int nwritten = 0;\n\n while ((towrite = sdslen(buf)-pos) > 0) {\n towrite = (towrite > (64*1024) ? (64*1024) : towrite);\n nwritten = syncWrite(fd,buf+pos,towrite,timeout);\n if (nwritten != (signed)towrite) goto socket_wr_err;\n pos += nwritten;\n }\n }\n\n /* Read back the reply. */\n // 读取命令的回复\n {\n char buf1[1024];\n char buf2[1024];\n\n /* Read the two replies */\n if (syncReadLine(fd, buf1, sizeof(buf1), timeout) <= 0)\n goto socket_rd_err;\n if (syncReadLine(fd, buf2, sizeof(buf2), timeout) <= 0)\n goto socket_rd_err;\n\n // 检查 RESTORE 命令执行是否成功\n\n if (buf1[0] == '-' || buf2[0] == '-') {\n\n // 执行出错。。。\n\n addReplyErrorFormat(c,\"Target instance replied with error: %s\",\n (buf1[0] == '-') ? buf1+1 : buf2+1);\n } else {\n\n // 执行成功。。。\n\n robj *aux;\n\n // 如果没有指定 COPY 选项，那么删除本机数据库中的键\n if (!copy) {\n /* No COPY option: remove the local key, signal the change. */\n dbDelete(c->db,c->argv[3]);\n signalModifiedKey(c->db,c->argv[3]);\n }\n addReply(c,shared.ok);\n server.dirty++;\n\n /* Translate MIGRATE as DEL for replication/AOF. */\n // 如果键被删除了的话，向 AOF 文件和从服务器/节点发送一个 DEL 命令\n aux = createStringObject(\"DEL\",3);\n rewriteClientCommandVector(c,2,aux,c->argv[3]);\n decrRefCount(aux);\n }\n }\n\n sdsfree(cmd.io.buffer.ptr);\n return;\n\nsocket_wr_err:\n sdsfree(cmd.io.buffer.ptr);\n migrateCloseSocket(c->argv[1],c->argv[2]);\n if (errno != ETIMEDOUT && retry_num++ == 0) goto try_again;\n addReplySds(c,\n sdsnew(\"-IOERR error or timeout writing to target instance\\r\\n\"));\n return;\n\nsocket_rd_err:\n sdsfree(cmd.io.buffer.ptr);\n migrateCloseSocket(c->argv[1],c->argv[2]);\n if (errno != ETIMEDOUT && retry_num++ == 0) goto try_again;\n addReplySds(c,\n sdsnew(\"-IOERR error or timeout reading from target node\\r\\n\"));\n return;\n}\n\n/* -----------------------------------------------------------------------------\n * Cluster functions related to serving / redirecting clients\n * -------------------------------------------------------------------------- */\n\n/* The ASKING command is required after a -ASK redirection.\n *\n * 客户端在接到 -ASK 转向之后，需要发送 ASKING 命令。\n *\n * The client should issue ASKING before to actually send the command to\n * the target instance. See the Redis Cluster specification for more\n * information. \n *\n * 客户端应该在向目标节点发送命令之前，向节点发送 ASKING 命令。\n * 具体原因请参考 Redis 集群规范。\n */\nvoid askingCommand(redisClient *c) {\n\n if (server.cluster_enabled == 0) {\n addReplyError(c,\"This instance has cluster support disabled\");\n return;\n }\n\n // 打开客户端的标识\n c->flags |= REDIS_ASKING;\n\n addReply(c,shared.ok);\n}\n\n/* The READONLY command is uesd by clients to enter the read-only mode.\n * In this mode slaves will not redirect clients as long as clients access\n * with read-only commands to keys that are served by the slave's master. */\nvoid readonlyCommand(redisClient *c) {\n if (server.cluster_enabled == 0) {\n addReplyError(c,\"This instance has cluster support disabled\");\n return;\n }\n c->flags |= REDIS_READONLY;\n addReply(c,shared.ok);\n}\n\n/* The READWRITE command just clears the READONLY command state. */\nvoid readwriteCommand(redisClient *c) {\n c->flags &= ~REDIS_READONLY;\n addReply(c,shared.ok);\n}\n\n/* Return the pointer to the cluster node that is able to serve the command.\n * For the function to succeed the command should only target either:\n *\n * 1) A single key (even multiple times like LPOPRPUSH mylist mylist).\n * 2) Multiple keys in the same hash slot, while the slot is stable (no\n * resharding in progress).\n *\n * On success the function returns the node that is able to serve the request.\n * If the node is not 'myself' a redirection must be perfomed. The kind of\n * redirection is specified setting the integer passed by reference\n * 'error_code', which will be set to REDIS_CLUSTER_REDIR_ASK or\n * REDIS_CLUSTER_REDIR_MOVED.\n *\n * When the node is 'myself' 'error_code' is set to REDIS_CLUSTER_REDIR_NONE.\n *\n * If the command fails NULL is returned, and the reason of the failure is\n * provided via 'error_code', which will be set to:\n *\n * REDIS_CLUSTER_REDIR_CROSS_SLOT if the request contains multiple keys that\n * don't belong to the same hash slot.\n *\n * REDIS_CLUSTER_REDIR_UNSTABLE if the request contains mutliple keys\n * belonging to the same slot, but the slot is not stable (in migration or\n * importing state, likely because a resharding is in progress). */\nclusterNode *getNodeByQuery(redisClient *c, struct redisCommand *cmd, robj **argv, int argc, int *hashslot, int *error_code) {\n\n // 初始化为 NULL ，\n // 如果输入命令是无参数命令，那么 n 就会继续为 NULL\n clusterNode *n = NULL;\n\n robj *firstkey = NULL;\n int multiple_keys = 0;\n multiState *ms, _ms;\n multiCmd mc;\n int i, slot = 0, migrating_slot = 0, importing_slot = 0, missing_keys = 0;\n\n /* Set error code optimistically for the base case. */\n if (error_code) *error_code = REDIS_CLUSTER_REDIR_NONE;\n\n /* We handle all the cases as if they were EXEC commands, so we have\n * a common code path for everything */\n // 集群可以执行事务，\n // 但必须确保事务中的所有命令都是针对某个相同的键进行的\n // 这个 if 和接下来的 for 进行的就是这一合法性检测\n if (cmd->proc == execCommand) {\n /* If REDIS_MULTI flag is not set EXEC is just going to return an\n * error. */\n if (!(c->flags & REDIS_MULTI)) return myself;\n ms = &c->mstate;\n } else {\n /* In order to have a single codepath create a fake Multi State\n * structure if the client is not in MULTI/EXEC state, this way\n * we have a single codepath below. */\n ms = &_ms;\n _ms.commands = &mc;\n _ms.count = 1;\n mc.argv = argv;\n mc.argc = argc;\n mc.cmd = cmd;\n }\n\n /* Check that all the keys are in the same hash slot, and obtain this\n * slot and the node associated. */\n for (i = 0; i < ms->count; i++) {\n struct redisCommand *mcmd;\n robj **margv;\n int margc, *keyindex, numkeys, j;\n\n mcmd = ms->commands[i].cmd;\n margc = ms->commands[i].argc;\n margv = ms->commands[i].argv;\n\n // 定位命令的键位置\n keyindex = getKeysFromCommand(mcmd,margv,margc,&numkeys);\n // 遍历命令中的所有键\n for (j = 0; j < numkeys; j++) {\n robj *thiskey = margv[keyindex[j]];\n int thisslot = keyHashSlot((char*)thiskey->ptr,\n sdslen(thiskey->ptr));\n\n if (firstkey == NULL) {\n // 这是事务中第一个被处理的键\n // 获取该键的槽和负责处理该槽的节点\n /* This is the first key we see. Check what is the slot\n * and node. */\n firstkey = thiskey;\n slot = thisslot;\n n = server.cluster->slots[slot];\n redisAssertWithInfo(c,firstkey,n != NULL);\n /* If we are migrating or importing this slot, we need to check\n * if we have all the keys in the request (the only way we\n * can safely serve the request, otherwise we return a TRYAGAIN\n * error). To do so we set the importing/migrating state and\n * increment a counter for every missing key. */\n if (n == myself &&\n server.cluster->migrating_slots_to[slot] != NULL)\n {\n migrating_slot = 1;\n } else if (server.cluster->importing_slots_from[slot] != NULL) {\n importing_slot = 1;\n }\n } else {\n /* If it is not the first key, make sure it is exactly\n * the same key as the first we saw. */\n if (!equalStringObjects(firstkey,thiskey)) {\n if (slot != thisslot) {\n /* Error: multiple keys from different slots. */\n getKeysFreeResult(keyindex);\n if (error_code)\n *error_code = REDIS_CLUSTER_REDIR_CROSS_SLOT;\n return NULL;\n } else {\n /* Flag this request as one with multiple different\n * keys. */\n multiple_keys = 1;\n }\n }\n }\n\n /* Migarting / Improrting slot? Count keys we don't have. */\n if ((migrating_slot || importing_slot) &&\n lookupKeyRead(&server.db[0],thiskey) == NULL)\n {\n missing_keys++;\n }\n }\n getKeysFreeResult(keyindex);\n }\n\n /* No key at all in command? then we can serve the request\n * without redirections or errors. */\n if (n == NULL) return myself;\n\n /* Return the hashslot by reference. */\n if (hashslot) *hashslot = slot;\n\n /* This request is about a slot we are migrating into another instance?\n * Then if we have all the keys. */\n\n /* If we don't have all the keys and we are migrating the slot, send\n * an ASK redirection. */\n if (migrating_slot && missing_keys) {\n if (error_code) *error_code = REDIS_CLUSTER_REDIR_ASK;\n return server.cluster->migrating_slots_to[slot];\n }\n\n /* If we are receiving the slot, and the client correctly flagged the\n * request as \"ASKING\", we can serve the request. However if the request\n * involves multiple keys and we don't have them all, the only option is\n * to send a TRYAGAIN error. */\n if (importing_slot &&\n (c->flags & REDIS_ASKING || cmd->flags & REDIS_CMD_ASKING))\n {\n if (multiple_keys && missing_keys) {\n if (error_code) *error_code = REDIS_CLUSTER_REDIR_UNSTABLE;\n return NULL;\n } else {\n return myself;\n }\n }\n\n /* Handle the read-only client case reading from a slave: if this\n * node is a slave and the request is about an hash slot our master\n * is serving, we can reply without redirection. */\n if (c->flags & REDIS_READONLY &&\n cmd->flags & REDIS_CMD_READONLY &&\n nodeIsSlave(myself) &&\n myself->slaveof == n)\n {\n return myself;\n }\n\n /* Base case: just return the right node. However if this node is not\n * myself, set error_code to MOVED since we need to issue a rediretion. */\n if (n != myself && error_code) *error_code = REDIS_CLUSTER_REDIR_MOVED;\n\n // 返回负责处理槽 slot 的节点 n\n return n;\n}\n" }, { "path": "src/cluster.h", "content": "#ifndef __REDIS_CLUSTER_H\n#define __REDIS_CLUSTER_H\n\n/*-----------------------------------------------------------------------------\n * Redis cluster data structures, defines, exported API.\n *----------------------------------------------------------------------------*/\n\n// 槽数量\n#define REDIS_CLUSTER_SLOTS 16384\n// 集群在线\n#define REDIS_CLUSTER_OK 0 /* Everything looks ok */\n// 集群下线\n#define REDIS_CLUSTER_FAIL 1 /* The cluster can't work */\n// 节点名字的长度\n#define REDIS_CLUSTER_NAMELEN 40 /* sha1 hex length */\n// 集群的实际端口号 = 用户指定的端口号 + REDIS_CLUSTER_PORT_INCR\n#define REDIS_CLUSTER_PORT_INCR 10000 /* Cluster port = baseport + PORT_INCR */\n\n/* The following defines are amunt of time, sometimes expressed as\n * multiplicators of the node timeout value (when ending with MULT). \n *\n * 以下是和时间有关的一些常量，\n * 以 _MULTI 结尾的常量会作为时间值的乘法因子来使用。\n */\n// 默认节点超时时限\n#define REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT 15000\n// 检验下线报告的乘法因子\n#define REDIS_CLUSTER_FAIL_REPORT_VALIDITY_MULT 2 /* Fail report validity. */\n// 撤销主节点 FAIL 状态的乘法因子\n#define REDIS_CLUSTER_FAIL_UNDO_TIME_MULT 2 /* Undo fail if master is back. */\n// 撤销主节点 FAIL 状态的加法因子\n#define REDIS_CLUSTER_FAIL_UNDO_TIME_ADD 10 /* Some additional time. */\n// 在检查从节点数据是否有效时使用的乘法因子\n#define REDIS_CLUSTER_SLAVE_VALIDITY_MULT 10 /* Slave data validity. */\n// 在执行故障转移之前需要等待的秒数，似乎已经废弃\n#define REDIS_CLUSTER_FAILOVER_DELAY 5 /* Seconds */\n// 未使用，似乎已经废弃\n#define REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER 1\n// 在进行手动的故障转移之前，需要等待的超时时间\n#define REDIS_CLUSTER_MF_TIMEOUT 5000 /* Milliseconds to do a manual failover. */\n// 未使用，似乎已经废弃\n#define REDIS_CLUSTER_MF_PAUSE_MULT 2 /* Master pause manual failover mult. */\n\n/* Redirection errors returned by getNodeByQuery(). */\n/* 由 getNodeByQuery() 函数返回的转向错误。 */\n// 节点可以处理这个命令\n#define REDIS_CLUSTER_REDIR_NONE 0 /* Node can serve the request. */\n// 键在其他槽\n#define REDIS_CLUSTER_REDIR_CROSS_SLOT 1 /* Keys in different slots. */\n// 键所处的槽正在进行 reshard\n#define REDIS_CLUSTER_REDIR_UNSTABLE 2 /* Keys in slot resharding. */\n// 需要进行 ASK 转向\n#define REDIS_CLUSTER_REDIR_ASK 3 /* -ASK redirection required. */\n// 需要进行 MOVED 转向\n#define REDIS_CLUSTER_REDIR_MOVED 4 /* -MOVED redirection required. */\n\n// 前置定义，防止编译错误\nstruct clusterNode;\n\n\n/* clusterLink encapsulates everything needed to talk with a remote node. */\n// clusterLink 包含了与其他节点进行通讯所需的全部信息\ntypedef struct clusterLink {\n\n // 连接的创建时间\n mstime_t ctime; /* Link creation time */\n\n // TCP 套接字描述符\n int fd; /* TCP socket file descriptor */\n\n // 输出缓冲区，保存着等待发送给其他节点的消息（message）。\n sds sndbuf; /* Packet send buffer */\n\n // 输入缓冲区，保存着从其他节点接收到的消息。\n sds rcvbuf; /* Packet reception buffer */\n\n // 与这个连接相关联的节点，如果没有的话就为 NULL\n struct clusterNode *node; /* Node related to this link if any, or NULL */\n\n} clusterLink;\n\n/* Cluster node flags and macros. */\n// 该节点为主节点\n#define REDIS_NODE_MASTER 1 /* The node is a master */\n// 该节点为从节点\n#define REDIS_NODE_SLAVE 2 /* The node is a slave */\n// 该节点疑似下线，需要对它的状态进行确认\n#define REDIS_NODE_PFAIL 4 /* Failure? Need acknowledge */\n// 该节点已下线\n#define REDIS_NODE_FAIL 8 /* The node is believed to be malfunctioning */\n// 该节点是当前节点自身\n#define REDIS_NODE_MYSELF 16 /* This node is myself */\n// 该节点还未与当前节点完成第一次 PING - PONG 通讯\n#define REDIS_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */\n// 该节点没有地址\n#define REDIS_NODE_NOADDR 64 /* We don't know the address of this node */\n// 当前节点还未与该节点进行过接触\n// 带有这个标识会让当前节点发送 MEET 命令而不是 PING 命令\n#define REDIS_NODE_MEET 128 /* Send a MEET message to this node */\n// 该节点被选中为新的主节点\n#define REDIS_NODE_PROMOTED 256 /* Master was a slave propoted by failover */\n// 空名字（在节点为主节点时，用作消息中的 slaveof 属性的值）\n#define REDIS_NODE_NULL_NAME \"\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\\000\"\n\n// 用于判断节点身份和状态的一系列宏\n#define nodeIsMaster(n) ((n)->flags & REDIS_NODE_MASTER)\n#define nodeIsSlave(n) ((n)->flags & REDIS_NODE_SLAVE)\n#define nodeInHandshake(n) ((n)->flags & REDIS_NODE_HANDSHAKE)\n#define nodeHasAddr(n) (!((n)->flags & REDIS_NODE_NOADDR))\n#define nodeWithoutAddr(n) ((n)->flags & REDIS_NODE_NOADDR)\n#define nodeTimedOut(n) ((n)->flags & REDIS_NODE_PFAIL)\n#define nodeFailed(n) ((n)->flags & REDIS_NODE_FAIL)\n\n/* This structure represent elements of node->fail_reports. */\n// 每个 clusterNodeFailReport 结构保存了一条其他节点对目标节点的下线报告\n// （认为目标节点已经下线）\nstruct clusterNodeFailReport {\n\n // 报告目标节点已经下线的节点\n struct clusterNode *node; /* Node reporting the failure condition. */\n\n // 最后一次从 node 节点收到下线报告的时间\n // 程序使用这个时间戳来检查下线报告是否过期\n mstime_t time; /* Time of the last report from this node. */\n\n} typedef clusterNodeFailReport;\n\n\n// 节点状态\nstruct clusterNode {\n\n // 创建节点的时间\n mstime_t ctime; /* Node object creation time. */\n\n // 节点的名字，由 40 个十六进制字符组成\n // 例如 68eef66df23420a5862208ef5b1a7005b806f2ff\n char name[REDIS_CLUSTER_NAMELEN]; /* Node name, hex string, sha1-size */\n\n // 节点标识\n // 使用各种不同的标识值记录节点的角色（比如主节点或者从节点），\n // 以及节点目前所处的状态（比如在线或者下线）。\n int flags; /* REDIS_NODE_... */\n\n // 节点当前的配置纪元，用于实现故障转移\n uint64_t configEpoch; /* Last configEpoch observed for this node */\n\n // 由这个节点负责处理的槽\n // 一共有 REDIS_CLUSTER_SLOTS / 8 个字节长\n // 每个字节的每个位记录了一个槽的保存状态\n // 位的值为 1 表示槽正由本节点处理，值为 0 则表示槽并非本节点处理\n // 比如 slots[0] 的第一个位保存了槽 0 的保存情况\n // slots[0] 的第二个位保存了槽 1 的保存情况，以此类推\n unsigned char slots[REDIS_CLUSTER_SLOTS/8]; /* slots handled by this node */\n\n // 该节点负责处理的槽数量\n int numslots; /* Number of slots handled by this node */\n\n // 如果本节点是主节点，那么用这个属性记录从节点的数量\n int numslaves; /* Number of slave nodes, if this is a master */\n\n // 指针数组，指向各个从节点\n struct clusterNode **slaves; /* pointers to slave nodes */\n\n // 如果这是一个从节点，那么指向主节点\n struct clusterNode *slaveof; /* pointer to the master node */\n\n // 最后一次发送 PING 命令的时间\n mstime_t ping_sent; /* Unix time we sent latest ping */\n\n // 最后一次接收 PONG 回复的时间戳\n mstime_t pong_received; /* Unix time we received the pong */\n\n // 最后一次被设置为 FAIL 状态的时间\n mstime_t fail_time; /* Unix time when FAIL flag was set */\n\n // 最后一次给某个从节点投票的时间\n mstime_t voted_time; /* Last time we voted for a slave of this master */\n\n // 最后一次从这个节点接收到复制偏移量的时间\n mstime_t repl_offset_time; /* Unix time we received offset for this node */\n\n // 这个节点的复制偏移量\n long long repl_offset; /* Last known repl offset for this node. */\n\n // 节点的 IP 地址\n char ip[REDIS_IP_STR_LEN]; /* Latest known IP address of this node */\n\n // 节点的端口号\n int port; /* Latest known port of this node */\n\n // 保存连接节点所需的有关信息\n clusterLink *link; /* TCP/IP link with this node */\n\n // 一个链表，记录了所有其他节点对该节点的下线报告\n list *fail_reports; /* List of nodes signaling this as failing */\n\n};\ntypedef struct clusterNode clusterNode;\n\n\n// 集群状态，每个节点都保存着一个这样的状态，记录了它们眼中的集群的样子。\n// 另外，虽然这个结构主要用于记录集群的属性，但是为了节约资源，\n// 有些与节点有关的属性，比如 slots_to_keys 、 failover_auth_count \n// 也被放到了这个结构里面。\ntypedef struct clusterState {\n\n // 指向当前节点的指针\n clusterNode *myself; /* This node */\n\n // 集群当前的配置纪元，用于实现故障转移\n uint64_t currentEpoch;\n\n // 集群当前的状态：是在线还是下线\n int state; /* REDIS_CLUSTER_OK, REDIS_CLUSTER_FAIL, ... */\n\n // 集群中至少处理着一个槽的节点的数量。\n int size; /* Num of master nodes with at least one slot */\n\n // 集群节点名单（包括 myself 节点）\n // 字典的键为节点的名字，字典的值为 clusterNode 结构\n dict *nodes; /* Hash table of name -> clusterNode structures */\n\n // 节点黑名单，用于 CLUSTER FORGET 命令\n // 防止被 FORGET 的命令重新被添加到集群里面\n // （不过现在似乎没有在使用的样子，已废弃？还是尚未实现？）\n dict *nodes_black_list; /* Nodes we don't re-add for a few seconds. */\n\n // 记录要从当前节点迁移到目标节点的槽，以及迁移的目标节点\n // migrating_slots_to[i] = NULL 表示槽 i 未被迁移\n // migrating_slots_to[i] = clusterNode_A 表示槽 i 要从本节点迁移至节点 A\n clusterNode *migrating_slots_to[REDIS_CLUSTER_SLOTS];\n\n // 记录要从源节点迁移到本节点的槽，以及进行迁移的源节点\n // importing_slots_from[i] = NULL 表示槽 i 未进行导入\n // importing_slots_from[i] = clusterNode_A 表示正从节点 A 中导入槽 i\n clusterNode *importing_slots_from[REDIS_CLUSTER_SLOTS];\n\n // 负责处理各个槽的节点\n // 例如 slots[i] = clusterNode_A 表示槽 i 由节点 A 处理\n clusterNode *slots[REDIS_CLUSTER_SLOTS];\n\n // 跳跃表，表中以槽作为分值，键作为成员，对槽进行有序排序\n // 当需要对某些槽进行区间（range）操作时，这个跳跃表可以提供方便\n // 具体操作定义在 db.c 里面\n zskiplist *slots_to_keys;\n\n /* The following fields are used to take the slave state on elections. */\n // 以下这些域被用于进行故障转移选举\n\n // 上次执行选举或者下次执行选举的时间\n mstime_t failover_auth_time; /* Time of previous or next election. */\n\n // 节点获得的投票数量\n int failover_auth_count; /* Number of votes received so far. */\n\n // 如果值为 1 ，表示本节点已经向其他节点发送了投票请求\n int failover_auth_sent; /* True if we already asked for votes. */\n\n int failover_auth_rank; /* This slave rank for current auth request. */\n\n uint64_t failover_auth_epoch; /* Epoch of the current election. */\n\n /* Manual failover state in common. */\n /* 共用的手动故障转移状态 */\n\n // 手动故障转移执行的时间限制\n mstime_t mf_end; /* Manual failover time limit (ms unixtime).\n It is zero if there is no MF in progress. */\n /* Manual failover state of master. */\n /* 主服务器的手动故障转移状态 */\n clusterNode *mf_slave; /* Slave performing the manual failover. */\n /* Manual failover state of slave. */\n /* 从服务器的手动故障转移状态 */\n long long mf_master_offset; /* Master offset the slave needs to start MF\n or zero if stil not received. */\n // 指示手动故障转移是否可以开始的标志值\n // 值为非 0 时表示各个主服务器可以开始投票\n int mf_can_start; /* If non-zero signal that the manual failover\n can start requesting masters vote. */\n\n /* The followign fields are uesd by masters to take state on elections. */\n /* 以下这些域由主服务器使用，用于记录选举时的状态 */\n\n // 集群最后一次进行投票的纪元\n uint64_t lastVoteEpoch; /* Epoch of the last vote granted. */\n\n // 在进入下个事件循环之前要做的事情，以各个 flag 来记录\n int todo_before_sleep; /* Things to do in clusterBeforeSleep(). */\n\n // 通过 cluster 连接发送的消息数量\n long long stats_bus_messages_sent; /* Num of msg sent via cluster bus. */\n\n // 通过 cluster 接收到的消息数量\n long long stats_bus_messages_received; /* Num of msg rcvd via cluster bus.*/\n\n} clusterState;\n\n/* clusterState todo_before_sleep flags. */\n// 以下每个 flag 代表了一个服务器在开始下一个事件循环之前\n// 要做的事情\n#define CLUSTER_TODO_HANDLE_FAILOVER (1<<0)\n#define CLUSTER_TODO_UPDATE_STATE (1<<1)\n#define CLUSTER_TODO_SAVE_CONFIG (1<<2)\n#define CLUSTER_TODO_FSYNC_CONFIG (1<<3)\n\n/* Redis cluster messages header */\n\n/* Note that the PING, PONG and MEET messages are actually the same exact\n * kind of packet. PONG is the reply to ping, in the exact format as a PING,\n * while MEET is a special PING that forces the receiver to add the sender\n * as a node (if it is not already in the list). */\n// 注意，PING 、 PONG 和 MEET 实际上是同一种消息。\n// PONG 是对 PING 的回复，它的实际格式也为 PING 消息，\n// 而 MEET 则是一种特殊的 PING 消息，用于强制消息的接收者将消息的发送者添加到集群中\n// （如果节点尚未在节点列表中的话）\n// PING\n#define CLUSTERMSG_TYPE_PING 0 /* Ping */\n// PONG （回复 PING）\n#define CLUSTERMSG_TYPE_PONG 1 /* Pong (reply to Ping) */\n// 请求将某个节点添加到集群中\n#define CLUSTERMSG_TYPE_MEET 2 /* Meet \"let's join\" message */\n// 将某个节点标记为 FAIL\n#define CLUSTERMSG_TYPE_FAIL 3 /* Mark node xxx as failing */\n// 通过发布与订阅功能广播消息\n#define CLUSTERMSG_TYPE_PUBLISH 4 /* Pub/Sub Publish propagation */\n// 请求进行故障转移操作，要求消息的接收者通过投票来支持消息的发送者\n#define CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST 5 /* May I failover? */\n// 消息的接收者同意向消息的发送者投票\n#define CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK 6 /* Yes, you have my vote */\n// 槽布局已经发生变化，消息发送者要求消息接收者进行相应的更新\n#define CLUSTERMSG_TYPE_UPDATE 7 /* Another node slots configuration */\n// 为了进行手动故障转移，暂停各个客户端\n#define CLUSTERMSG_TYPE_MFSTART 8 /* Pause clients for manual failover */\n\n/* Initially we don't know our \"name\", but we'll find it once we connect\n * to the first node, using the getsockname() function. Then we'll use this\n * address for all the next messages. */\n\ntypedef struct {\n\n // 节点的名字\n // 在刚开始的时候，节点的名字会是随机的\n // 当 MEET 信息发送并得到回复之后，集群就会为节点设置正式的名字\n char nodename[REDIS_CLUSTER_NAMELEN];\n\n // 最后一次向该节点发送 PING 消息的时间戳\n uint32_t ping_sent;\n\n // 最后一次从该节点接收到 PONG 消息的时间戳\n uint32_t pong_received;\n\n // 节点的 IP 地址\n char ip[REDIS_IP_STR_LEN]; /* IP address last time it was seen */\n\n // 节点的端口号\n uint16_t port; /* port last time it was seen */\n\n // 节点的标识值\n uint16_t flags;\n\n // 对齐字节，不使用\n uint32_t notused; /* for 64 bit alignment */\n\n} clusterMsgDataGossip;\n\ntypedef struct {\n\n // 下线节点的名字\n char nodename[REDIS_CLUSTER_NAMELEN];\n\n} clusterMsgDataFail;\n\ntypedef struct {\n\n // 频道名长度\n uint32_t channel_len;\n\n // 消息长度\n uint32_t message_len;\n\n // 消息内容，格式为 频道名+消息\n // bulk_data[0:channel_len-1] 为频道名\n // bulk_data[channel_len:channel_len+message_len-1] 为消息\n unsigned char bulk_data[8]; /* defined as 8 just for alignment concerns. */\n\n} clusterMsgDataPublish;\n\ntypedef struct {\n\n // 节点的配置纪元\n uint64_t configEpoch; /* Config epoch of the specified instance. */\n\n // 节点的名字\n char nodename[REDIS_CLUSTER_NAMELEN]; /* Name of the slots owner. */\n\n // 节点的槽布局\n unsigned char slots[REDIS_CLUSTER_SLOTS/8]; /* Slots bitmap. */\n\n} clusterMsgDataUpdate;\n\nunion clusterMsgData {\n\n /* PING, MEET and PONG */\n struct {\n /* Array of N clusterMsgDataGossip structures */\n // 每条消息都包含两个 clusterMsgDataGossip 结构\n clusterMsgDataGossip gossip[1];\n } ping;\n\n /* FAIL */\n struct {\n clusterMsgDataFail about;\n } fail;\n\n /* PUBLISH */\n struct {\n clusterMsgDataPublish msg;\n } publish;\n\n /* UPDATE */\n struct {\n clusterMsgDataUpdate nodecfg;\n } update;\n\n};\n\n// 用来表示集群消息的结构（消息头，header）\ntypedef struct {\n char sig[4]; /* Siganture \"RCmb\" (Redis Cluster message bus). */\n // 消息的长度（包括这个消息头的长度和消息正文的长度）\n uint32_t totlen; /* Total length of this message */\n uint16_t ver; /* Protocol version, currently set to 0. */\n uint16_t notused0; /* 2 bytes not used. */\n\n // 消息的类型\n uint16_t type; /* Message type */\n\n // 消息正文包含的节点信息数量\n // 只在发送 MEET 、 PING 和 PONG 这三种 Gossip 协议消息时使用\n uint16_t count; /* Only used for some kind of messages. */\n\n // 消息发送者的配置纪元\n uint64_t currentEpoch; /* The epoch accordingly to the sending node. */\n\n // 如果消息发送者是一个主节点，那么这里记录的是消息发送者的配置纪元\n // 如果消息发送者是一个从节点，那么这里记录的是消息发送者正在复制的主节点的配置纪元\n uint64_t configEpoch; /* The config epoch if it's a master, or the last\n epoch advertised by its master if it is a\n slave. */\n\n // 节点的复制偏移量\n uint64_t offset; /* Master replication offset if node is a master or\n processed replication offset if node is a slave. */\n\n // 消息发送者的名字（ID）\n char sender[REDIS_CLUSTER_NAMELEN]; /* Name of the sender node */\n\n // 消息发送者目前的槽指派信息\n unsigned char myslots[REDIS_CLUSTER_SLOTS/8];\n\n // 如果消息发送者是一个从节点，那么这里记录的是消息发送者正在复制的主节点的名字\n // 如果消息发送者是一个主节点，那么这里记录的是 REDIS_NODE_NULL_NAME\n // （一个 40 字节长，值全为 0 的字节数组）\n char slaveof[REDIS_CLUSTER_NAMELEN];\n\n char notused1[32]; /* 32 bytes reserved for future usage. */\n\n // 消息发送者的端口号\n uint16_t port; /* Sender TCP base port */\n\n // 消息发送者的标识值\n uint16_t flags; /* Sender node flags */\n\n // 消息发送者所处集群的状态\n unsigned char state; /* Cluster state from the POV of the sender */\n\n // 消息标志\n unsigned char mflags[3]; /* Message flags: CLUSTERMSG_FLAG[012]_... */\n\n // 消息的正文（或者说，内容）\n union clusterMsgData data;\n\n} clusterMsg;\n\n#define CLUSTERMSG_MIN_LEN (sizeof(clusterMsg)-sizeof(union clusterMsgData))\n\n/* Message flags better specify the packet content or are used to\n * provide some information about the node state. */\n#define CLUSTERMSG_FLAG0_PAUSED (1<<0) /* Master paused for manual failover. */\n#define CLUSTERMSG_FLAG0_FORCEACK (1<<1) /* Give ACK to AUTH_REQUEST even if\n master is up. */\n\n/* ---------------------- API exported outside cluster.c -------------------- */\nclusterNode *getNodeByQuery(redisClient *c, struct redisCommand *cmd, robj **argv, int argc, int *hashslot, int *ask);\n\n#endif /* __REDIS_CLUSTER_H */\n" }, { "path": "src/config.c", "content": "/* Configuration file parsing and CONFIG GET/SET commands implementation.\n *\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"redis.h\"\n#include \"cluster.h\"\n\n#include \n#include \n\nstatic struct {\n const char *name;\n const int value;\n} validSyslogFacilities[] = {\n {\"user\", LOG_USER},\n {\"local0\", LOG_LOCAL0},\n {\"local1\", LOG_LOCAL1},\n {\"local2\", LOG_LOCAL2},\n {\"local3\", LOG_LOCAL3},\n {\"local4\", LOG_LOCAL4},\n {\"local5\", LOG_LOCAL5},\n {\"local6\", LOG_LOCAL6},\n {\"local7\", LOG_LOCAL7},\n {NULL, 0}\n};\n\nclientBufferLimitsConfig clientBufferLimitsDefaults[REDIS_CLIENT_LIMIT_NUM_CLASSES] = {\n {0, 0, 0}, /* normal */\n {1024*1024*256, 1024*1024*64, 60}, /* slave */\n {1024*1024*32, 1024*1024*8, 60} /* pubsub */\n};\n\n/*-----------------------------------------------------------------------------\n * Config file parsing\n *----------------------------------------------------------------------------*/\n\nint yesnotoi(char *s) {\n if (!strcasecmp(s,\"yes\")) return 1;\n else if (!strcasecmp(s,\"no\")) return 0;\n else return -1;\n}\n\nvoid appendServerSaveParams(time_t seconds, int changes) {\n server.saveparams = zrealloc(server.saveparams,sizeof(struct saveparam)*(server.saveparamslen+1));\n server.saveparams[server.saveparamslen].seconds = seconds;\n server.saveparams[server.saveparamslen].changes = changes;\n server.saveparamslen++;\n}\n\nvoid resetServerSaveParams() {\n zfree(server.saveparams);\n server.saveparams = NULL;\n server.saveparamslen = 0;\n}\n\nvoid loadServerConfigFromString(char *config) {\n char *err = NULL;\n int linenum = 0, totlines, i;\n int slaveof_linenum = 0;\n sds *lines;\n\n lines = sdssplitlen(config,strlen(config),\"\\n\",1,&totlines);\n\n for (i = 0; i < totlines; i++) {\n sds *argv;\n int argc;\n\n linenum = i+1;\n // 移除字符串的前置空白和后缀空白\n lines[i] = sdstrim(lines[i],\" \\t\\r\\n\");\n\n /* Skip comments and blank lines */\n // 跳过空白行\n if (lines[i][0] == '#' || lines[i][0] == '\\0') continue;\n\n /* Split into arguments */\n // 将字符串分割成多个参数\n argv = sdssplitargs(lines[i],&argc);\n if (argv == NULL) {\n err = \"Unbalanced quotes in configuration line\";\n goto loaderr;\n }\n\n /* Skip this line if the resulting command vector is empty. */\n // 跳过空白参数\n if (argc == 0) {\n sdsfreesplitres(argv,argc);\n continue;\n }\n\n // 将选项名字转换成小写\n // 例如 TIMEOUT 转换成 timeout\n sdstolower(argv[0]);\n\n /* Execute config directives */\n if (!strcasecmp(argv[0],\"timeout\") && argc == 2) {\n server.maxidletime = atoi(argv[1]);\n if (server.maxidletime < 0) {\n err = \"Invalid timeout value\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"tcp-keepalive\") && argc == 2) {\n server.tcpkeepalive = atoi(argv[1]);\n if (server.tcpkeepalive < 0) {\n err = \"Invalid tcp-keepalive value\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"port\") && argc == 2) {\n server.port = atoi(argv[1]);\n if (server.port < 0 || server.port > 65535) {\n err = \"Invalid port\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"tcp-backlog\") && argc == 2) {\n server.tcp_backlog = atoi(argv[1]);\n if (server.tcp_backlog < 0) {\n err = \"Invalid backlog value\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"bind\") && argc >= 2) {\n int j, addresses = argc-1;\n\n if (addresses > REDIS_BINDADDR_MAX) {\n err = \"Too many bind addresses specified\"; goto loaderr;\n }\n for (j = 0; j < addresses; j++)\n server.bindaddr[j] = zstrdup(argv[j+1]);\n server.bindaddr_count = addresses;\n } else if (!strcasecmp(argv[0],\"unixsocket\") && argc == 2) {\n server.unixsocket = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"unixsocketperm\") && argc == 2) {\n errno = 0;\n server.unixsocketperm = (mode_t)strtol(argv[1], NULL, 8);\n if (errno || server.unixsocketperm > 0777) {\n err = \"Invalid socket file permissions\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"save\")) {\n if (argc == 3) {\n int seconds = atoi(argv[1]);\n int changes = atoi(argv[2]);\n if (seconds < 1 || changes < 0) {\n err = \"Invalid save parameters\"; goto loaderr;\n }\n appendServerSaveParams(seconds,changes);\n } else if (argc == 2 && !strcasecmp(argv[1],\"\")) {\n resetServerSaveParams();\n }\n } else if (!strcasecmp(argv[0],\"dir\") && argc == 2) {\n if (chdir(argv[1]) == -1) {\n redisLog(REDIS_WARNING,\"Can't chdir to '%s': %s\",\n argv[1], strerror(errno));\n exit(1);\n }\n } else if (!strcasecmp(argv[0],\"loglevel\") && argc == 2) {\n if (!strcasecmp(argv[1],\"debug\")) server.verbosity = REDIS_DEBUG;\n else if (!strcasecmp(argv[1],\"verbose\")) server.verbosity = REDIS_VERBOSE;\n else if (!strcasecmp(argv[1],\"notice\")) server.verbosity = REDIS_NOTICE;\n else if (!strcasecmp(argv[1],\"warning\")) server.verbosity = REDIS_WARNING;\n else {\n err = \"Invalid log level. Must be one of debug, notice, warning\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"logfile\") && argc == 2) {\n FILE *logfp;\n\n zfree(server.logfile);\n server.logfile = zstrdup(argv[1]);\n if (server.logfile[0] != '\\0') {\n /* Test if we are able to open the file. The server will not\n * be able to abort just for this problem later... */\n logfp = fopen(server.logfile,\"a\");\n if (logfp == NULL) {\n err = sdscatprintf(sdsempty(),\n \"Can't open the log file: %s\", strerror(errno));\n goto loaderr;\n }\n fclose(logfp);\n }\n } else if (!strcasecmp(argv[0],\"syslog-enabled\") && argc == 2) {\n if ((server.syslog_enabled = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"syslog-ident\") && argc == 2) {\n if (server.syslog_ident) zfree(server.syslog_ident);\n server.syslog_ident = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"syslog-facility\") && argc == 2) {\n int i;\n\n for (i = 0; validSyslogFacilities[i].name; i++) {\n if (!strcasecmp(validSyslogFacilities[i].name, argv[1])) {\n server.syslog_facility = validSyslogFacilities[i].value;\n break;\n }\n }\n\n if (!validSyslogFacilities[i].name) {\n err = \"Invalid log facility. Must be one of USER or between LOCAL0-LOCAL7\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"databases\") && argc == 2) {\n server.dbnum = atoi(argv[1]);\n if (server.dbnum < 1) {\n err = \"Invalid number of databases\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"include\") && argc == 2) {\n loadServerConfig(argv[1],NULL);\n } else if (!strcasecmp(argv[0],\"maxclients\") && argc == 2) {\n server.maxclients = atoi(argv[1]);\n if (server.maxclients < 1) {\n err = \"Invalid max clients limit\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"maxmemory\") && argc == 2) {\n server.maxmemory = memtoll(argv[1],NULL);\n } else if (!strcasecmp(argv[0],\"maxmemory-policy\") && argc == 2) {\n if (!strcasecmp(argv[1],\"volatile-lru\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;\n } else if (!strcasecmp(argv[1],\"volatile-random\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_RANDOM;\n } else if (!strcasecmp(argv[1],\"volatile-ttl\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_TTL;\n } else if (!strcasecmp(argv[1],\"allkeys-lru\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_ALLKEYS_LRU;\n } else if (!strcasecmp(argv[1],\"allkeys-random\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_ALLKEYS_RANDOM;\n } else if (!strcasecmp(argv[1],\"noeviction\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;\n } else {\n err = \"Invalid maxmemory policy\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"maxmemory-samples\") && argc == 2) {\n server.maxmemory_samples = atoi(argv[1]);\n if (server.maxmemory_samples <= 0) {\n err = \"maxmemory-samples must be 1 or greater\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"slaveof\") && argc == 3) {\n slaveof_linenum = linenum;\n server.masterhost = sdsnew(argv[1]);\n server.masterport = atoi(argv[2]);\n server.repl_state = REDIS_REPL_CONNECT;\n } else if (!strcasecmp(argv[0],\"repl-ping-slave-period\") && argc == 2) {\n server.repl_ping_slave_period = atoi(argv[1]);\n if (server.repl_ping_slave_period <= 0) {\n err = \"repl-ping-slave-period must be 1 or greater\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"repl-timeout\") && argc == 2) {\n server.repl_timeout = atoi(argv[1]);\n if (server.repl_timeout <= 0) {\n err = \"repl-timeout must be 1 or greater\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"repl-disable-tcp-nodelay\") && argc==2) {\n if ((server.repl_disable_tcp_nodelay = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"repl-backlog-size\") && argc == 2) {\n long long size = memtoll(argv[1],NULL);\n if (size <= 0) {\n err = \"repl-backlog-size must be 1 or greater.\";\n goto loaderr;\n }\n resizeReplicationBacklog(size);\n } else if (!strcasecmp(argv[0],\"repl-backlog-ttl\") && argc == 2) {\n server.repl_backlog_time_limit = atoi(argv[1]);\n if (server.repl_backlog_time_limit < 0) {\n err = \"repl-backlog-ttl can't be negative \";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"masterauth\") && argc == 2) {\n \tserver.masterauth = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"slave-serve-stale-data\") && argc == 2) {\n if ((server.repl_serve_stale_data = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"slave-read-only\") && argc == 2) {\n if ((server.repl_slave_ro = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"rdbcompression\") && argc == 2) {\n if ((server.rdb_compression = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"rdbchecksum\") && argc == 2) {\n if ((server.rdb_checksum = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"activerehashing\") && argc == 2) {\n if ((server.activerehashing = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"daemonize\") && argc == 2) {\n if ((server.daemonize = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"hz\") && argc == 2) {\n server.hz = atoi(argv[1]);\n if (server.hz < REDIS_MIN_HZ) server.hz = REDIS_MIN_HZ;\n if (server.hz > REDIS_MAX_HZ) server.hz = REDIS_MAX_HZ;\n } else if (!strcasecmp(argv[0],\"appendonly\") && argc == 2) {\n int yes;\n\n if ((yes = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n server.aof_state = yes ? REDIS_AOF_ON : REDIS_AOF_OFF;\n } else if (!strcasecmp(argv[0],\"appendfilename\") && argc == 2) {\n if (!pathIsBaseName(argv[1])) {\n err = \"appendfilename can't be a path, just a filename\";\n goto loaderr;\n }\n zfree(server.aof_filename);\n server.aof_filename = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"no-appendfsync-on-rewrite\")\n && argc == 2) {\n if ((server.aof_no_fsync_on_rewrite= yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"appendfsync\") && argc == 2) {\n if (!strcasecmp(argv[1],\"no\")) {\n server.aof_fsync = AOF_FSYNC_NO;\n } else if (!strcasecmp(argv[1],\"always\")) {\n server.aof_fsync = AOF_FSYNC_ALWAYS;\n } else if (!strcasecmp(argv[1],\"everysec\")) {\n server.aof_fsync = AOF_FSYNC_EVERYSEC;\n } else {\n err = \"argument must be 'no', 'always' or 'everysec'\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"auto-aof-rewrite-percentage\") &&\n argc == 2)\n {\n server.aof_rewrite_perc = atoi(argv[1]);\n if (server.aof_rewrite_perc < 0) {\n err = \"Invalid negative percentage for AOF auto rewrite\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"auto-aof-rewrite-min-size\") &&\n argc == 2)\n {\n server.aof_rewrite_min_size = memtoll(argv[1],NULL);\n } else if (!strcasecmp(argv[0],\"aof-rewrite-incremental-fsync\") &&\n argc == 2)\n {\n if ((server.aof_rewrite_incremental_fsync = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"requirepass\") && argc == 2) {\n if (strlen(argv[1]) > REDIS_AUTHPASS_MAX_LEN) {\n err = \"Password is longer than REDIS_AUTHPASS_MAX_LEN\";\n goto loaderr;\n }\n server.requirepass = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"pidfile\") && argc == 2) {\n zfree(server.pidfile);\n server.pidfile = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"dbfilename\") && argc == 2) {\n if (!pathIsBaseName(argv[1])) {\n err = \"dbfilename can't be a path, just a filename\";\n goto loaderr;\n }\n zfree(server.rdb_filename);\n server.rdb_filename = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"hash-max-ziplist-entries\") && argc == 2) {\n server.hash_max_ziplist_entries = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"hash-max-ziplist-value\") && argc == 2) {\n server.hash_max_ziplist_value = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"list-max-ziplist-entries\") && argc == 2){\n server.list_max_ziplist_entries = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"list-max-ziplist-value\") && argc == 2) {\n server.list_max_ziplist_value = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"set-max-intset-entries\") && argc == 2) {\n server.set_max_intset_entries = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"zset-max-ziplist-entries\") && argc == 2) {\n server.zset_max_ziplist_entries = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"zset-max-ziplist-value\") && argc == 2) {\n server.zset_max_ziplist_value = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"hll-sparse-max-bytes\") && argc == 2) {\n server.hll_sparse_max_bytes = memtoll(argv[1], NULL);\n } else if (!strcasecmp(argv[0],\"rename-command\") && argc == 3) {\n struct redisCommand *cmd = lookupCommand(argv[1]);\n int retval;\n\n if (!cmd) {\n err = \"No such command in rename-command\";\n goto loaderr;\n }\n\n /* If the target command name is the empty string we just\n * remove it from the command table. */\n retval = dictDelete(server.commands, argv[1]);\n redisAssert(retval == DICT_OK);\n\n /* Otherwise we re-add the command under a different name. */\n if (sdslen(argv[2]) != 0) {\n sds copy = sdsdup(argv[2]);\n\n retval = dictAdd(server.commands, copy, cmd);\n if (retval != DICT_OK) {\n sdsfree(copy);\n err = \"Target command name already exists\"; goto loaderr;\n }\n }\n } else if (!strcasecmp(argv[0],\"cluster-enabled\") && argc == 2) {\n if ((server.cluster_enabled = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"cluster-config-file\") && argc == 2) {\n zfree(server.cluster_configfile);\n server.cluster_configfile = zstrdup(argv[1]);\n } else if (!strcasecmp(argv[0],\"cluster-node-timeout\") && argc == 2) {\n server.cluster_node_timeout = strtoll(argv[1],NULL,10);\n if (server.cluster_node_timeout <= 0) {\n err = \"cluster node timeout must be 1 or greater\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"cluster-migration-barrier\")\n && argc == 2)\n {\n server.cluster_migration_barrier = atoi(argv[1]);\n if (server.cluster_migration_barrier < 0) {\n err = \"cluster migration barrier must be positive\";\n goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"lua-time-limit\") && argc == 2) {\n server.lua_time_limit = strtoll(argv[1],NULL,10);\n } else if (!strcasecmp(argv[0],\"slowlog-log-slower-than\") &&\n argc == 2)\n {\n server.slowlog_log_slower_than = strtoll(argv[1],NULL,10);\n } else if (!strcasecmp(argv[0],\"slowlog-max-len\") && argc == 2) {\n server.slowlog_max_len = strtoll(argv[1],NULL,10);\n } else if (!strcasecmp(argv[0],\"client-output-buffer-limit\") &&\n argc == 5)\n {\n int class = getClientLimitClassByName(argv[1]);\n unsigned long long hard, soft;\n int soft_seconds;\n\n if (class == -1) {\n err = \"Unrecognized client limit class\";\n goto loaderr;\n }\n hard = memtoll(argv[2],NULL);\n soft = memtoll(argv[3],NULL);\n soft_seconds = atoi(argv[4]);\n if (soft_seconds < 0) {\n err = \"Negative number of seconds in soft limit is invalid\";\n goto loaderr;\n }\n server.client_obuf_limits[class].hard_limit_bytes = hard;\n server.client_obuf_limits[class].soft_limit_bytes = soft;\n server.client_obuf_limits[class].soft_limit_seconds = soft_seconds;\n } else if (!strcasecmp(argv[0],\"stop-writes-on-bgsave-error\") &&\n argc == 2) {\n if ((server.stop_writes_on_bgsave_err = yesnotoi(argv[1])) == -1) {\n err = \"argument must be 'yes' or 'no'\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"slave-priority\") && argc == 2) {\n server.slave_priority = atoi(argv[1]);\n } else if (!strcasecmp(argv[0],\"min-slaves-to-write\") && argc == 2) {\n server.repl_min_slaves_to_write = atoi(argv[1]);\n if (server.repl_min_slaves_to_write < 0) {\n err = \"Invalid value for min-slaves-to-write.\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"min-slaves-max-lag\") && argc == 2) {\n server.repl_min_slaves_max_lag = atoi(argv[1]);\n if (server.repl_min_slaves_max_lag < 0) {\n err = \"Invalid value for min-slaves-max-lag.\"; goto loaderr;\n }\n } else if (!strcasecmp(argv[0],\"notify-keyspace-events\") && argc == 2) {\n int flags = keyspaceEventsStringToFlags(argv[1]);\n\n if (flags == -1) {\n err = \"Invalid event class character. Use 'g$lshzxeA'.\";\n goto loaderr;\n }\n server.notify_keyspace_events = flags;\n } else if (!strcasecmp(argv[0],\"sentinel\")) {\n /* argc == 1 is handled by main() as we need to enter the sentinel\n * mode ASAP. */\n // 如果 SENTINEL 命令不为空，那么执行以下代码\n if (argc != 1) {\n // 如果 SENTINEL 模式未开启，那么出错\n if (!server.sentinel_mode) {\n err = \"sentinel directive while not in sentinel mode\";\n goto loaderr;\n }\n // 载入 SENTINEL 相关选项\n err = sentinelHandleConfiguration(argv+1,argc-1);\n if (err) goto loaderr;\n }\n } else {\n err = \"Bad directive or wrong number of arguments\"; goto loaderr;\n }\n sdsfreesplitres(argv,argc);\n }\n\n /* Sanity checks. */\n if (server.cluster_enabled && server.masterhost) {\n linenum = slaveof_linenum;\n i = linenum-1;\n err = \"slaveof directive not allowed in cluster mode\";\n goto loaderr;\n }\n\n sdsfreesplitres(lines,totlines);\n return;\n\nloaderr:\n fprintf(stderr, \"\\n*** FATAL CONFIG FILE ERROR ***\\n\");\n fprintf(stderr, \"Reading the configuration file, at line %d\\n\", linenum);\n fprintf(stderr, \">>> '%s'\\n\", lines[i]);\n fprintf(stderr, \"%s\\n\", err);\n exit(1);\n}\n\n/* Load the server configuration from the specified filename.\n *\n * 从给定文件中载入服务器配置。\n *\n * The function appends the additional configuration directives stored\n * in the 'options' string to the config file before loading.\n *\n * options 字符串会被追加到文件所载入的内容的后面。\n *\n * Both filename and options can be NULL, in such a case are considered\n * empty. This way loadServerConfig can be used to just load a file or\n * just load a string. \n *\n * filename 和 options 都可以是 NULL ，在这种情况下，\n * 服务器配置文件视为空文件。\n */\nvoid loadServerConfig(char *filename, char *options) {\n sds config = sdsempty();\n char buf[REDIS_CONFIGLINE_MAX+1];\n\n /* Load the file content */\n // 载入文件内容\n if (filename) {\n FILE *fp;\n\n if (filename[0] == '-' && filename[1] == '\\0') {\n fp = stdin;\n } else {\n if ((fp = fopen(filename,\"r\")) == NULL) {\n redisLog(REDIS_WARNING,\n \"Fatal error, can't open config file '%s'\", filename);\n exit(1);\n }\n }\n while(fgets(buf,REDIS_CONFIGLINE_MAX+1,fp) != NULL)\n config = sdscat(config,buf);\n if (fp != stdin) fclose(fp);\n }\n\n /* Append the additional options */\n // 追加 options 字符串到内容的末尾\n if (options) {\n config = sdscat(config,\"\\n\");\n config = sdscat(config,options);\n }\n\n // 根据字符串内容，设置服务器配置\n loadServerConfigFromString(config);\n\n sdsfree(config);\n}\n\n/*-----------------------------------------------------------------------------\n * CONFIG SET implementation\n *----------------------------------------------------------------------------*/\n\nvoid configSetCommand(redisClient *c) {\n robj *o;\n long long ll;\n redisAssertWithInfo(c,c->argv[2],sdsEncodedObject(c->argv[2]));\n redisAssertWithInfo(c,c->argv[3],sdsEncodedObject(c->argv[3]));\n o = c->argv[3];\n\n if (!strcasecmp(c->argv[2]->ptr,\"dbfilename\")) {\n if (!pathIsBaseName(o->ptr)) {\n addReplyError(c, \"dbfilename can't be a path, just a filename\");\n return;\n }\n zfree(server.rdb_filename);\n server.rdb_filename = zstrdup(o->ptr);\n } else if (!strcasecmp(c->argv[2]->ptr,\"requirepass\")) {\n if (sdslen(o->ptr) > REDIS_AUTHPASS_MAX_LEN) goto badfmt;\n zfree(server.requirepass);\n server.requirepass = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL;\n } else if (!strcasecmp(c->argv[2]->ptr,\"masterauth\")) {\n zfree(server.masterauth);\n server.masterauth = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL;\n } else if (!strcasecmp(c->argv[2]->ptr,\"maxmemory\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0) goto badfmt;\n server.maxmemory = ll;\n if (server.maxmemory) {\n if (server.maxmemory < zmalloc_used_memory()) {\n redisLog(REDIS_WARNING,\"WARNING: the new maxmemory value set via CONFIG SET is smaller than the current memory usage. This will result in keys eviction and/or inability to accept new write commands depending on the maxmemory-policy.\");\n }\n freeMemoryIfNeeded();\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"maxclients\")) {\n int orig_value = server.maxclients;\n\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 1) goto badfmt;\n\n /* Try to check if the OS is capable of supporting so many FDs. */\n server.maxclients = ll;\n if (ll > orig_value) {\n adjustOpenFilesLimit();\n if (server.maxclients != ll) {\n addReplyErrorFormat(c,\"The operating system is not able to handle the specified number of clients, try with %d\", server.maxclients);\n server.maxclients = orig_value;\n return;\n }\n if (aeGetSetSize(server.el) <\n server.maxclients + REDIS_EVENTLOOP_FDSET_INCR)\n {\n if (aeResizeSetSize(server.el,\n server.maxclients + REDIS_EVENTLOOP_FDSET_INCR) == AE_ERR)\n {\n addReplyError(c,\"The event loop API used by Redis is not able to handle the specified number of clients\");\n server.maxclients = orig_value;\n return;\n }\n }\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"hz\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.hz = ll;\n if (server.hz < REDIS_MIN_HZ) server.hz = REDIS_MIN_HZ;\n if (server.hz > REDIS_MAX_HZ) server.hz = REDIS_MAX_HZ;\n } else if (!strcasecmp(c->argv[2]->ptr,\"maxmemory-policy\")) {\n if (!strcasecmp(o->ptr,\"volatile-lru\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;\n } else if (!strcasecmp(o->ptr,\"volatile-random\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_RANDOM;\n } else if (!strcasecmp(o->ptr,\"volatile-ttl\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_TTL;\n } else if (!strcasecmp(o->ptr,\"allkeys-lru\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_ALLKEYS_LRU;\n } else if (!strcasecmp(o->ptr,\"allkeys-random\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_ALLKEYS_RANDOM;\n } else if (!strcasecmp(o->ptr,\"noeviction\")) {\n server.maxmemory_policy = REDIS_MAXMEMORY_NO_EVICTION;\n } else {\n goto badfmt;\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"maxmemory-samples\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll <= 0) goto badfmt;\n server.maxmemory_samples = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"timeout\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0 || ll > LONG_MAX) goto badfmt;\n server.maxidletime = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"tcp-keepalive\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0 || ll > INT_MAX) goto badfmt;\n server.tcpkeepalive = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"appendfsync\")) {\n if (!strcasecmp(o->ptr,\"no\")) {\n server.aof_fsync = AOF_FSYNC_NO;\n } else if (!strcasecmp(o->ptr,\"everysec\")) {\n server.aof_fsync = AOF_FSYNC_EVERYSEC;\n } else if (!strcasecmp(o->ptr,\"always\")) {\n server.aof_fsync = AOF_FSYNC_ALWAYS;\n } else {\n goto badfmt;\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"no-appendfsync-on-rewrite\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.aof_no_fsync_on_rewrite = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"appendonly\")) {\n int enable = yesnotoi(o->ptr);\n\n if (enable == -1) goto badfmt;\n if (enable == 0 && server.aof_state != REDIS_AOF_OFF) {\n stopAppendOnly();\n } else if (enable && server.aof_state == REDIS_AOF_OFF) {\n if (startAppendOnly() == REDIS_ERR) {\n addReplyError(c,\n \"Unable to turn on AOF. Check server logs.\");\n return;\n }\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"auto-aof-rewrite-percentage\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.aof_rewrite_perc = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"auto-aof-rewrite-min-size\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.aof_rewrite_min_size = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"aof-rewrite-incremental-fsync\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.aof_rewrite_incremental_fsync = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"save\")) {\n int vlen, j;\n sds *v = sdssplitlen(o->ptr,sdslen(o->ptr),\" \",1,&vlen);\n\n /* Perform sanity check before setting the new config:\n * - Even number of args\n * - Seconds >= 1, changes >= 0 */\n if (vlen & 1) {\n sdsfreesplitres(v,vlen);\n goto badfmt;\n }\n for (j = 0; j < vlen; j++) {\n char *eptr;\n long val;\n\n val = strtoll(v[j], &eptr, 10);\n if (eptr[0] != '\\0' ||\n ((j & 1) == 0 && val < 1) ||\n ((j & 1) == 1 && val < 0)) {\n sdsfreesplitres(v,vlen);\n goto badfmt;\n }\n }\n /* Finally set the new config */\n resetServerSaveParams();\n for (j = 0; j < vlen; j += 2) {\n time_t seconds;\n int changes;\n\n seconds = strtoll(v[j],NULL,10);\n changes = strtoll(v[j+1],NULL,10);\n appendServerSaveParams(seconds, changes);\n }\n sdsfreesplitres(v,vlen);\n } else if (!strcasecmp(c->argv[2]->ptr,\"slave-serve-stale-data\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.repl_serve_stale_data = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"slave-read-only\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.repl_slave_ro = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"dir\")) {\n if (chdir((char*)o->ptr) == -1) {\n addReplyErrorFormat(c,\"Changing directory: %s\", strerror(errno));\n return;\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"hash-max-ziplist-entries\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.hash_max_ziplist_entries = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"hash-max-ziplist-value\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.hash_max_ziplist_value = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"list-max-ziplist-entries\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.list_max_ziplist_entries = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"list-max-ziplist-value\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.list_max_ziplist_value = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"set-max-intset-entries\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.set_max_intset_entries = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"zset-max-ziplist-entries\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.zset_max_ziplist_entries = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"zset-max-ziplist-value\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.zset_max_ziplist_value = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"hll-sparse-max-bytes\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.hll_sparse_max_bytes = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"lua-time-limit\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.lua_time_limit = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"slowlog-log-slower-than\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR) goto badfmt;\n server.slowlog_log_slower_than = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"slowlog-max-len\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.slowlog_max_len = (unsigned)ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"loglevel\")) {\n if (!strcasecmp(o->ptr,\"warning\")) {\n server.verbosity = REDIS_WARNING;\n } else if (!strcasecmp(o->ptr,\"notice\")) {\n server.verbosity = REDIS_NOTICE;\n } else if (!strcasecmp(o->ptr,\"verbose\")) {\n server.verbosity = REDIS_VERBOSE;\n } else if (!strcasecmp(o->ptr,\"debug\")) {\n server.verbosity = REDIS_DEBUG;\n } else {\n goto badfmt;\n }\n } else if (!strcasecmp(c->argv[2]->ptr,\"client-output-buffer-limit\")) {\n int vlen, j;\n sds *v = sdssplitlen(o->ptr,sdslen(o->ptr),\" \",1,&vlen);\n\n /* We need a multiple of 4: */\n if (vlen % 4) {\n sdsfreesplitres(v,vlen);\n goto badfmt;\n }\n\n /* Sanity check of single arguments, so that we either refuse the\n * whole configuration string or accept it all, even if a single\n * error in a single client class is present. */\n for (j = 0; j < vlen; j++) {\n char *eptr;\n long val;\n\n if ((j % 4) == 0) {\n if (getClientLimitClassByName(v[j]) == -1) {\n sdsfreesplitres(v,vlen);\n goto badfmt;\n }\n } else {\n val = strtoll(v[j], &eptr, 10);\n if (eptr[0] != '\\0' || val < 0) {\n sdsfreesplitres(v,vlen);\n goto badfmt;\n }\n }\n }\n /* Finally set the new config */\n for (j = 0; j < vlen; j += 4) {\n int class;\n unsigned long long hard, soft;\n int soft_seconds;\n\n class = getClientLimitClassByName(v[j]);\n hard = strtoll(v[j+1],NULL,10);\n soft = strtoll(v[j+2],NULL,10);\n soft_seconds = strtoll(v[j+3],NULL,10);\n\n server.client_obuf_limits[class].hard_limit_bytes = hard;\n server.client_obuf_limits[class].soft_limit_bytes = soft;\n server.client_obuf_limits[class].soft_limit_seconds = soft_seconds;\n }\n sdsfreesplitres(v,vlen);\n } else if (!strcasecmp(c->argv[2]->ptr,\"stop-writes-on-bgsave-error\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.stop_writes_on_bgsave_err = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"repl-ping-slave-period\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll <= 0) goto badfmt;\n server.repl_ping_slave_period = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"repl-timeout\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll <= 0) goto badfmt;\n server.repl_timeout = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"repl-backlog-size\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll <= 0) goto badfmt;\n resizeReplicationBacklog(ll);\n } else if (!strcasecmp(c->argv[2]->ptr,\"repl-backlog-ttl\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n server.repl_backlog_time_limit = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"watchdog-period\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR || ll < 0) goto badfmt;\n if (ll)\n enableWatchdog(ll);\n else\n disableWatchdog();\n } else if (!strcasecmp(c->argv[2]->ptr,\"rdbcompression\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.rdb_compression = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"notify-keyspace-events\")) {\n int flags = keyspaceEventsStringToFlags(o->ptr);\n\n if (flags == -1) goto badfmt;\n server.notify_keyspace_events = flags;\n } else if (!strcasecmp(c->argv[2]->ptr,\"repl-disable-tcp-nodelay\")) {\n int yn = yesnotoi(o->ptr);\n\n if (yn == -1) goto badfmt;\n server.repl_disable_tcp_nodelay = yn;\n } else if (!strcasecmp(c->argv[2]->ptr,\"slave-priority\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0) goto badfmt;\n server.slave_priority = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"min-slaves-to-write\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0) goto badfmt;\n server.repl_min_slaves_to_write = ll;\n refreshGoodSlavesCount();\n } else if (!strcasecmp(c->argv[2]->ptr,\"min-slaves-max-lag\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0) goto badfmt;\n server.repl_min_slaves_max_lag = ll;\n refreshGoodSlavesCount();\n } else if (!strcasecmp(c->argv[2]->ptr,\"cluster-node-timeout\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll <= 0) goto badfmt;\n server.cluster_node_timeout = ll;\n } else if (!strcasecmp(c->argv[2]->ptr,\"cluster-migration-barrier\")) {\n if (getLongLongFromObject(o,&ll) == REDIS_ERR ||\n ll < 0) goto badfmt;\n server.cluster_migration_barrier = ll;\n } else {\n addReplyErrorFormat(c,\"Unsupported CONFIG parameter: %s\",\n (char*)c->argv[2]->ptr);\n return;\n }\n addReply(c,shared.ok);\n return;\n\nbadfmt: /* Bad format errors */\n addReplyErrorFormat(c,\"Invalid argument '%s' for CONFIG SET '%s'\",\n (char*)o->ptr,\n (char*)c->argv[2]->ptr);\n}\n\n/*-----------------------------------------------------------------------------\n * CONFIG GET implementation\n *----------------------------------------------------------------------------*/\n\n#define config_get_string_field(_name,_var) do { \\\n if (stringmatch(pattern,_name,0)) { \\\n addReplyBulkCString(c,_name); \\\n addReplyBulkCString(c,_var ? _var : \"\"); \\\n matches++; \\\n } \\\n} while(0);\n\n#define config_get_bool_field(_name,_var) do { \\\n if (stringmatch(pattern,_name,0)) { \\\n addReplyBulkCString(c,_name); \\\n addReplyBulkCString(c,_var ? \"yes\" : \"no\"); \\\n matches++; \\\n } \\\n} while(0);\n\n#define config_get_numerical_field(_name,_var) do { \\\n if (stringmatch(pattern,_name,0)) { \\\n ll2string(buf,sizeof(buf),_var); \\\n addReplyBulkCString(c,_name); \\\n addReplyBulkCString(c,buf); \\\n matches++; \\\n } \\\n} while(0);\n\nvoid configGetCommand(redisClient *c) {\n robj *o = c->argv[2];\n void *replylen = addDeferredMultiBulkLength(c);\n char *pattern = o->ptr;\n char buf[128];\n int matches = 0;\n redisAssertWithInfo(c,o,sdsEncodedObject(o));\n\n /* String values */\n config_get_string_field(\"dbfilename\",server.rdb_filename);\n config_get_string_field(\"requirepass\",server.requirepass);\n config_get_string_field(\"masterauth\",server.masterauth);\n config_get_string_field(\"unixsocket\",server.unixsocket);\n config_get_string_field(\"logfile\",server.logfile);\n config_get_string_field(\"pidfile\",server.pidfile);\n\n /* Numerical values */\n config_get_numerical_field(\"maxmemory\",server.maxmemory);\n config_get_numerical_field(\"maxmemory-samples\",server.maxmemory_samples);\n config_get_numerical_field(\"timeout\",server.maxidletime);\n config_get_numerical_field(\"tcp-keepalive\",server.tcpkeepalive);\n config_get_numerical_field(\"auto-aof-rewrite-percentage\",\n server.aof_rewrite_perc);\n config_get_numerical_field(\"auto-aof-rewrite-min-size\",\n server.aof_rewrite_min_size);\n config_get_numerical_field(\"hash-max-ziplist-entries\",\n server.hash_max_ziplist_entries);\n config_get_numerical_field(\"hash-max-ziplist-value\",\n server.hash_max_ziplist_value);\n config_get_numerical_field(\"list-max-ziplist-entries\",\n server.list_max_ziplist_entries);\n config_get_numerical_field(\"list-max-ziplist-value\",\n server.list_max_ziplist_value);\n config_get_numerical_field(\"set-max-intset-entries\",\n server.set_max_intset_entries);\n config_get_numerical_field(\"zset-max-ziplist-entries\",\n server.zset_max_ziplist_entries);\n config_get_numerical_field(\"zset-max-ziplist-value\",\n server.zset_max_ziplist_value);\n config_get_numerical_field(\"hll-sparse-max-bytes\",\n server.hll_sparse_max_bytes);\n config_get_numerical_field(\"lua-time-limit\",server.lua_time_limit);\n config_get_numerical_field(\"slowlog-log-slower-than\",\n server.slowlog_log_slower_than);\n config_get_numerical_field(\"slowlog-max-len\",\n server.slowlog_max_len);\n config_get_numerical_field(\"port\",server.port);\n config_get_numerical_field(\"tcp-backlog\",server.tcp_backlog);\n config_get_numerical_field(\"databases\",server.dbnum);\n config_get_numerical_field(\"repl-ping-slave-period\",server.repl_ping_slave_period);\n config_get_numerical_field(\"repl-timeout\",server.repl_timeout);\n config_get_numerical_field(\"repl-backlog-size\",server.repl_backlog_size);\n config_get_numerical_field(\"repl-backlog-ttl\",server.repl_backlog_time_limit);\n config_get_numerical_field(\"maxclients\",server.maxclients);\n config_get_numerical_field(\"watchdog-period\",server.watchdog_period);\n config_get_numerical_field(\"slave-priority\",server.slave_priority);\n config_get_numerical_field(\"min-slaves-to-write\",server.repl_min_slaves_to_write);\n config_get_numerical_field(\"min-slaves-max-lag\",server.repl_min_slaves_max_lag);\n config_get_numerical_field(\"hz\",server.hz);\n config_get_numerical_field(\"cluster-node-timeout\",server.cluster_node_timeout);\n config_get_numerical_field(\"cluster-migration-barrier\",server.cluster_migration_barrier);\n\n /* Bool (yes/no) values */\n config_get_bool_field(\"no-appendfsync-on-rewrite\",\n server.aof_no_fsync_on_rewrite);\n config_get_bool_field(\"slave-serve-stale-data\",\n server.repl_serve_stale_data);\n config_get_bool_field(\"slave-read-only\",\n server.repl_slave_ro);\n config_get_bool_field(\"stop-writes-on-bgsave-error\",\n server.stop_writes_on_bgsave_err);\n config_get_bool_field(\"daemonize\", server.daemonize);\n config_get_bool_field(\"rdbcompression\", server.rdb_compression);\n config_get_bool_field(\"rdbchecksum\", server.rdb_checksum);\n config_get_bool_field(\"activerehashing\", server.activerehashing);\n config_get_bool_field(\"repl-disable-tcp-nodelay\",\n server.repl_disable_tcp_nodelay);\n config_get_bool_field(\"aof-rewrite-incremental-fsync\",\n server.aof_rewrite_incremental_fsync);\n\n /* Everything we can't handle with macros follows. */\n\n if (stringmatch(pattern,\"appendonly\",0)) {\n addReplyBulkCString(c,\"appendonly\");\n addReplyBulkCString(c,server.aof_state == REDIS_AOF_OFF ? \"no\" : \"yes\");\n matches++;\n }\n if (stringmatch(pattern,\"dir\",0)) {\n char buf[1024];\n\n if (getcwd(buf,sizeof(buf)) == NULL)\n buf[0] = '\\0';\n\n addReplyBulkCString(c,\"dir\");\n addReplyBulkCString(c,buf);\n matches++;\n }\n if (stringmatch(pattern,\"maxmemory-policy\",0)) {\n char *s;\n\n switch(server.maxmemory_policy) {\n case REDIS_MAXMEMORY_VOLATILE_LRU: s = \"volatile-lru\"; break;\n case REDIS_MAXMEMORY_VOLATILE_TTL: s = \"volatile-ttl\"; break;\n case REDIS_MAXMEMORY_VOLATILE_RANDOM: s = \"volatile-random\"; break;\n case REDIS_MAXMEMORY_ALLKEYS_LRU: s = \"allkeys-lru\"; break;\n case REDIS_MAXMEMORY_ALLKEYS_RANDOM: s = \"allkeys-random\"; break;\n case REDIS_MAXMEMORY_NO_EVICTION: s = \"noeviction\"; break;\n default: s = \"unknown\"; break; /* too harmless to panic */\n }\n addReplyBulkCString(c,\"maxmemory-policy\");\n addReplyBulkCString(c,s);\n matches++;\n }\n if (stringmatch(pattern,\"appendfsync\",0)) {\n char *policy;\n\n switch(server.aof_fsync) {\n case AOF_FSYNC_NO: policy = \"no\"; break;\n case AOF_FSYNC_EVERYSEC: policy = \"everysec\"; break;\n case AOF_FSYNC_ALWAYS: policy = \"always\"; break;\n default: policy = \"unknown\"; break; /* too harmless to panic */\n }\n addReplyBulkCString(c,\"appendfsync\");\n addReplyBulkCString(c,policy);\n matches++;\n }\n if (stringmatch(pattern,\"save\",0)) {\n sds buf = sdsempty();\n int j;\n\n for (j = 0; j < server.saveparamslen; j++) {\n buf = sdscatprintf(buf,\"%jd %d\",\n (intmax_t)server.saveparams[j].seconds,\n server.saveparams[j].changes);\n if (j != server.saveparamslen-1)\n buf = sdscatlen(buf,\" \",1);\n }\n addReplyBulkCString(c,\"save\");\n addReplyBulkCString(c,buf);\n sdsfree(buf);\n matches++;\n }\n if (stringmatch(pattern,\"loglevel\",0)) {\n char *s;\n\n switch(server.verbosity) {\n case REDIS_WARNING: s = \"warning\"; break;\n case REDIS_VERBOSE: s = \"verbose\"; break;\n case REDIS_NOTICE: s = \"notice\"; break;\n case REDIS_DEBUG: s = \"debug\"; break;\n default: s = \"unknown\"; break; /* too harmless to panic */\n }\n addReplyBulkCString(c,\"loglevel\");\n addReplyBulkCString(c,s);\n matches++;\n }\n if (stringmatch(pattern,\"client-output-buffer-limit\",0)) {\n sds buf = sdsempty();\n int j;\n\n for (j = 0; j < REDIS_CLIENT_LIMIT_NUM_CLASSES; j++) {\n buf = sdscatprintf(buf,\"%s %llu %llu %ld\",\n getClientLimitClassName(j),\n server.client_obuf_limits[j].hard_limit_bytes,\n server.client_obuf_limits[j].soft_limit_bytes,\n (long) server.client_obuf_limits[j].soft_limit_seconds);\n if (j != REDIS_CLIENT_LIMIT_NUM_CLASSES-1)\n buf = sdscatlen(buf,\" \",1);\n }\n addReplyBulkCString(c,\"client-output-buffer-limit\");\n addReplyBulkCString(c,buf);\n sdsfree(buf);\n matches++;\n }\n if (stringmatch(pattern,\"unixsocketperm\",0)) {\n char buf[32];\n snprintf(buf,sizeof(buf),\"%o\",server.unixsocketperm);\n addReplyBulkCString(c,\"unixsocketperm\");\n addReplyBulkCString(c,buf);\n matches++;\n }\n if (stringmatch(pattern,\"slaveof\",0)) {\n char buf[256];\n\n addReplyBulkCString(c,\"slaveof\");\n if (server.masterhost)\n snprintf(buf,sizeof(buf),\"%s %d\",\n server.masterhost, server.masterport);\n else\n buf[0] = '\\0';\n addReplyBulkCString(c,buf);\n matches++;\n }\n if (stringmatch(pattern,\"notify-keyspace-events\",0)) {\n robj *flagsobj = createObject(REDIS_STRING,\n keyspaceEventsFlagsToString(server.notify_keyspace_events));\n\n addReplyBulkCString(c,\"notify-keyspace-events\");\n addReplyBulk(c,flagsobj);\n decrRefCount(flagsobj);\n matches++;\n }\n if (stringmatch(pattern,\"bind\",0)) {\n sds aux = sdsjoin(server.bindaddr,server.bindaddr_count,\" \");\n\n addReplyBulkCString(c,\"bind\");\n addReplyBulkCString(c,aux);\n sdsfree(aux);\n matches++;\n }\n setDeferredMultiBulkLength(c,replylen,matches*2);\n}\n\n/*-----------------------------------------------------------------------------\n * CONFIG REWRITE implementation\n *----------------------------------------------------------------------------*/\n\n#define REDIS_CONFIG_REWRITE_SIGNATURE \"# Generated by CONFIG REWRITE\"\n\n/* We use the following dictionary type to store where a configuration\n * option is mentioned in the old configuration file, so it's\n * like \"maxmemory\" -> list of line numbers (first line is zero). */\nunsigned int dictSdsCaseHash(const void *key);\nint dictSdsKeyCaseCompare(void *privdata, const void *key1, const void *key2);\nvoid dictSdsDestructor(void *privdata, void *val);\nvoid dictListDestructor(void *privdata, void *val);\n\n/* Sentinel config rewriting is implemented inside sentinel.c by\n * rewriteConfigSentinelOption(). */\nvoid rewriteConfigSentinelOption(struct rewriteConfigState *state);\n\ndictType optionToLineDictType = {\n dictSdsCaseHash, /* hash function */\n NULL, /* key dup */\n NULL, /* val dup */\n dictSdsKeyCaseCompare, /* key compare */\n dictSdsDestructor, /* key destructor */\n dictListDestructor /* val destructor */\n};\n\ndictType optionSetDictType = {\n dictSdsCaseHash, /* hash function */\n NULL, /* key dup */\n NULL, /* val dup */\n dictSdsKeyCaseCompare, /* key compare */\n dictSdsDestructor, /* key destructor */\n NULL /* val destructor */\n};\n\n/* The config rewrite state. */\nstruct rewriteConfigState {\n dict *option_to_line; /* Option -> list of config file lines map */\n dict *rewritten; /* Dictionary of already processed options */\n int numlines; /* Number of lines in current config */\n sds *lines; /* Current lines as an array of sds strings */\n int has_tail; /* True if we already added directives that were\n not present in the original config file. */\n};\n\n/* Append the new line to the current configuration state. */\nvoid rewriteConfigAppendLine(struct rewriteConfigState *state, sds line) {\n state->lines = zrealloc(state->lines, sizeof(char*) * (state->numlines+1));\n state->lines[state->numlines++] = line;\n}\n\n/* Populate the option -> list of line numbers map. */\nvoid rewriteConfigAddLineNumberToOption(struct rewriteConfigState *state, sds option, int linenum) {\n list *l = dictFetchValue(state->option_to_line,option);\n\n if (l == NULL) {\n l = listCreate();\n dictAdd(state->option_to_line,sdsdup(option),l);\n }\n listAddNodeTail(l,(void*)(long)linenum);\n}\n\n/* Add the specified option to the set of processed options.\n * This is useful as only unused lines of processed options will be blanked\n * in the config file, while options the rewrite process does not understand\n * remain untouched. */\nvoid rewriteConfigMarkAsProcessed(struct rewriteConfigState *state, char *option) {\n sds opt = sdsnew(option);\n\n if (dictAdd(state->rewritten,opt,NULL) != DICT_OK) sdsfree(opt);\n}\n\n/* Read the old file, split it into lines to populate a newly created\n * config rewrite state, and return it to the caller.\n *\n * If it is impossible to read the old file, NULL is returned.\n * If the old file does not exist at all, an empty state is returned. */\nstruct rewriteConfigState *rewriteConfigReadOldFile(char *path) {\n FILE *fp = fopen(path,\"r\");\n struct rewriteConfigState *state = zmalloc(sizeof(*state));\n char buf[REDIS_CONFIGLINE_MAX+1];\n int linenum = -1;\n\n if (fp == NULL && errno != ENOENT) return NULL;\n\n state->option_to_line = dictCreate(&optionToLineDictType,NULL);\n state->rewritten = dictCreate(&optionSetDictType,NULL);\n state->numlines = 0;\n state->lines = NULL;\n state->has_tail = 0;\n if (fp == NULL) return state;\n\n /* Read the old file line by line, populate the state. */\n while(fgets(buf,REDIS_CONFIGLINE_MAX+1,fp) != NULL) {\n int argc;\n sds *argv;\n sds line = sdstrim(sdsnew(buf),\"\\r\\n\\t \");\n\n linenum++; /* Zero based, so we init at -1 */\n\n /* Handle comments and empty lines. */\n if (line[0] == '#' || line[0] == '\\0') {\n if (!state->has_tail && !strcmp(line,REDIS_CONFIG_REWRITE_SIGNATURE))\n state->has_tail = 1;\n rewriteConfigAppendLine(state,line);\n continue;\n }\n\n /* Not a comment, split into arguments. */\n argv = sdssplitargs(line,&argc);\n if (argv == NULL) {\n /* Apparently the line is unparsable for some reason, for\n * instance it may have unbalanced quotes. Load it as a\n * comment. */\n sds aux = sdsnew(\"# ??? \");\n aux = sdscatsds(aux,line);\n sdsfree(line);\n rewriteConfigAppendLine(state,aux);\n continue;\n }\n\n sdstolower(argv[0]); /* We only want lowercase config directives. */\n\n /* Now we populate the state according to the content of this line.\n * Append the line and populate the option -> line numbers map. */\n rewriteConfigAppendLine(state,line);\n rewriteConfigAddLineNumberToOption(state,argv[0],linenum);\n\n sdsfreesplitres(argv,argc);\n }\n fclose(fp);\n return state;\n}\n\n/* Rewrite the specified configuration option with the new \"line\".\n * It progressively uses lines of the file that were already used for the same\n * configuration option in the old version of the file, removing that line from\n * the map of options -> line numbers.\n *\n * If there are lines associated with a given configuration option and\n * \"force\" is non-zero, the line is appended to the configuration file.\n * Usually \"force\" is true when an option has not its default value, so it\n * must be rewritten even if not present previously.\n * \n * The first time a line is appended into a configuration file, a comment\n * is added to show that starting from that point the config file was generated\n * by CONFIG REWRITE.\n *\n * \"line\" is either used, or freed, so the caller does not need to free it\n * in any way. */\nvoid rewriteConfigRewriteLine(struct rewriteConfigState *state, char *option, sds line, int force) {\n sds o = sdsnew(option);\n list *l = dictFetchValue(state->option_to_line,o);\n\n rewriteConfigMarkAsProcessed(state,option);\n\n if (!l && !force) {\n /* Option not used previously, and we are not forced to use it. */\n sdsfree(line);\n sdsfree(o);\n return;\n }\n\n if (l) {\n listNode *ln = listFirst(l);\n int linenum = (long) ln->value;\n\n /* There are still lines in the old configuration file we can reuse\n * for this option. Replace the line with the new one. */\n listDelNode(l,ln);\n if (listLength(l) == 0) dictDelete(state->option_to_line,o);\n sdsfree(state->lines[linenum]);\n state->lines[linenum] = line;\n } else {\n /* Append a new line. */\n if (!state->has_tail) {\n rewriteConfigAppendLine(state,\n sdsnew(REDIS_CONFIG_REWRITE_SIGNATURE));\n state->has_tail = 1;\n }\n rewriteConfigAppendLine(state,line);\n }\n sdsfree(o);\n}\n\n/* Write the long long 'bytes' value as a string in a way that is parsable\n * inside redis.conf. If possible uses the GB, MB, KB notation. */\nint rewriteConfigFormatMemory(char *buf, size_t len, long long bytes) {\n int gb = 1024*1024*1024;\n int mb = 1024*1024;\n int kb = 1024;\n\n if (bytes && (bytes % gb) == 0) {\n return snprintf(buf,len,\"%lldgb\",bytes/gb);\n } else if (bytes && (bytes % mb) == 0) {\n return snprintf(buf,len,\"%lldmb\",bytes/mb);\n } else if (bytes && (bytes % kb) == 0) {\n return snprintf(buf,len,\"%lldkb\",bytes/kb);\n } else {\n return snprintf(buf,len,\"%lld\",bytes);\n }\n}\n\n/* Rewrite a simple \"option-name \" configuration option. */\nvoid rewriteConfigBytesOption(struct rewriteConfigState *state, char *option, long long value, long long defvalue) {\n char buf[64];\n int force = value != defvalue;\n sds line;\n\n rewriteConfigFormatMemory(buf,sizeof(buf),value);\n line = sdscatprintf(sdsempty(),\"%s %s\",option,buf);\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite a yes/no option. */\nvoid rewriteConfigYesNoOption(struct rewriteConfigState *state, char *option, int value, int defvalue) {\n int force = value != defvalue;\n sds line = sdscatprintf(sdsempty(),\"%s %s\",option,\n value ? \"yes\" : \"no\");\n\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite a string option. */\nvoid rewriteConfigStringOption(struct rewriteConfigState *state, char *option, char *value, char *defvalue) {\n int force = 1;\n sds line;\n\n /* String options set to NULL need to be not present at all in the\n * configuration file to be set to NULL again at the next reboot. */\n if (value == NULL) {\n rewriteConfigMarkAsProcessed(state,option);\n return;\n }\n\n /* Compare the strings as sds strings to have a binary safe comparison. */\n if (defvalue && strcmp(value,defvalue) == 0) force = 0;\n\n line = sdsnew(option);\n line = sdscatlen(line, \" \", 1);\n line = sdscatrepr(line, value, strlen(value));\n\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite a numerical (long long range) option. */\nvoid rewriteConfigNumericalOption(struct rewriteConfigState *state, char *option, long long value, long long defvalue) {\n int force = value != defvalue;\n sds line = sdscatprintf(sdsempty(),\"%s %lld\",option,value);\n\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite a octal option. */\nvoid rewriteConfigOctalOption(struct rewriteConfigState *state, char *option, int value, int defvalue) {\n int force = value != defvalue;\n sds line = sdscatprintf(sdsempty(),\"%s %o\",option,value);\n\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite an enumeration option, after the \"value\" every enum/value pair\n * is specified, terminated by NULL. After NULL the default value is\n * specified. See how the function is used for more information. */\nvoid rewriteConfigEnumOption(struct rewriteConfigState *state, char *option, int value, ...) {\n va_list ap;\n char *enum_name, *matching_name = NULL;\n int enum_val, def_val, force;\n sds line;\n \n va_start(ap, value);\n while(1) {\n enum_name = va_arg(ap,char*);\n enum_val = va_arg(ap,int);\n if (enum_name == NULL) {\n def_val = enum_val;\n break;\n }\n if (value == enum_val) matching_name = enum_name;\n }\n va_end(ap);\n \n force = value != def_val;\n line = sdscatprintf(sdsempty(),\"%s %s\",option,matching_name);\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite the syslog-fability option. */\nvoid rewriteConfigSyslogfacilityOption(struct rewriteConfigState *state) {\n int value = server.syslog_facility, j;\n int force = value != LOG_LOCAL0;\n char *name = NULL, *option = \"syslog-facility\";\n sds line;\n\n for (j = 0; validSyslogFacilities[j].name; j++) {\n if (validSyslogFacilities[j].value == value) {\n name = (char*) validSyslogFacilities[j].name;\n break;\n }\n }\n line = sdscatprintf(sdsempty(),\"%s %s\",option,name);\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite the save option. */\nvoid rewriteConfigSaveOption(struct rewriteConfigState *state) {\n int j;\n sds line;\n\n /* Note that if there are no save parameters at all, all the current\n * config line with \"save\" will be detected as orphaned and deleted,\n * resulting into no RDB persistence as expected. */\n for (j = 0; j < server.saveparamslen; j++) {\n line = sdscatprintf(sdsempty(),\"save %ld %d\",\n (long) server.saveparams[j].seconds, server.saveparams[j].changes);\n rewriteConfigRewriteLine(state,\"save\",line,1);\n }\n /* Mark \"save\" as processed in case server.saveparamslen is zero. */\n rewriteConfigMarkAsProcessed(state,\"save\");\n}\n\n/* Rewrite the dir option, always using absolute paths.*/\nvoid rewriteConfigDirOption(struct rewriteConfigState *state) {\n char cwd[1024];\n\n if (getcwd(cwd,sizeof(cwd)) == NULL) {\n rewriteConfigMarkAsProcessed(state,\"dir\");\n return; /* no rewrite on error. */\n }\n rewriteConfigStringOption(state,\"dir\",cwd,NULL);\n}\n\n/* Rewrite the slaveof option. */\nvoid rewriteConfigSlaveofOption(struct rewriteConfigState *state) {\n char *option = \"slaveof\";\n sds line;\n\n /* If this is a master, we want all the slaveof config options\n * in the file to be removed. Note that if this is a cluster instance\n * we don't want a slaveof directive inside redis.conf. */\n if (server.cluster_enabled || server.masterhost == NULL) {\n rewriteConfigMarkAsProcessed(state,\"slaveof\");\n return;\n }\n line = sdscatprintf(sdsempty(),\"%s %s %d\", option,\n server.masterhost, server.masterport);\n rewriteConfigRewriteLine(state,option,line,1);\n}\n\n/* Rewrite the notify-keyspace-events option. */\nvoid rewriteConfigNotifykeyspaceeventsOption(struct rewriteConfigState *state) {\n int force = server.notify_keyspace_events != 0;\n char *option = \"notify-keyspace-events\";\n sds line, flags;\n\n flags = keyspaceEventsFlagsToString(server.notify_keyspace_events);\n line = sdsnew(option);\n line = sdscatlen(line, \" \", 1);\n line = sdscatrepr(line, flags, sdslen(flags));\n sdsfree(flags);\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Rewrite the client-output-buffer-limit option. */\nvoid rewriteConfigClientoutputbufferlimitOption(struct rewriteConfigState *state) {\n int j;\n char *option = \"client-output-buffer-limit\";\n\n for (j = 0; j < REDIS_CLIENT_LIMIT_NUM_CLASSES; j++) {\n int force = (server.client_obuf_limits[j].hard_limit_bytes !=\n clientBufferLimitsDefaults[j].hard_limit_bytes) ||\n (server.client_obuf_limits[j].soft_limit_bytes !=\n clientBufferLimitsDefaults[j].soft_limit_bytes) ||\n (server.client_obuf_limits[j].soft_limit_seconds !=\n clientBufferLimitsDefaults[j].soft_limit_seconds);\n sds line;\n char hard[64], soft[64];\n\n rewriteConfigFormatMemory(hard,sizeof(hard),\n server.client_obuf_limits[j].hard_limit_bytes);\n rewriteConfigFormatMemory(soft,sizeof(soft),\n server.client_obuf_limits[j].soft_limit_bytes);\n\n line = sdscatprintf(sdsempty(),\"%s %s %s %s %ld\",\n option, getClientLimitClassName(j), hard, soft,\n (long) server.client_obuf_limits[j].soft_limit_seconds);\n rewriteConfigRewriteLine(state,option,line,force);\n }\n}\n\n/* Rewrite the bind option. */\nvoid rewriteConfigBindOption(struct rewriteConfigState *state) {\n int force = 1;\n sds line, addresses;\n char *option = \"bind\";\n\n /* Nothing to rewrite if we don't have bind addresses. */\n if (server.bindaddr_count == 0) {\n rewriteConfigMarkAsProcessed(state,option);\n return;\n }\n\n /* Rewrite as bind ... */\n addresses = sdsjoin(server.bindaddr,server.bindaddr_count,\" \");\n line = sdsnew(option);\n line = sdscatlen(line, \" \", 1);\n line = sdscatsds(line, addresses);\n sdsfree(addresses);\n\n rewriteConfigRewriteLine(state,option,line,force);\n}\n\n/* Glue together the configuration lines in the current configuration\n * rewrite state into a single string, stripping multiple empty lines. */\nsds rewriteConfigGetContentFromState(struct rewriteConfigState *state) {\n sds content = sdsempty();\n int j, was_empty = 0;\n\n for (j = 0; j < state->numlines; j++) {\n /* Every cluster of empty lines is turned into a single empty line. */\n if (sdslen(state->lines[j]) == 0) {\n if (was_empty) continue;\n was_empty = 1;\n } else {\n was_empty = 0;\n }\n content = sdscatsds(content,state->lines[j]);\n content = sdscatlen(content,\"\\n\",1);\n }\n return content;\n}\n\n/* Free the configuration rewrite state. */\nvoid rewriteConfigReleaseState(struct rewriteConfigState *state) {\n sdsfreesplitres(state->lines,state->numlines);\n dictRelease(state->option_to_line);\n dictRelease(state->rewritten);\n zfree(state);\n}\n\n/* At the end of the rewrite process the state contains the remaining\n * map between \"option name\" => \"lines in the original config file\".\n * Lines used by the rewrite process were removed by the function\n * rewriteConfigRewriteLine(), all the other lines are \"orphaned\" and\n * should be replaced by empty lines.\n *\n * This function does just this, iterating all the option names and\n * blanking all the lines still associated. */\nvoid rewriteConfigRemoveOrphaned(struct rewriteConfigState *state) {\n dictIterator *di = dictGetIterator(state->option_to_line);\n dictEntry *de;\n\n while((de = dictNext(di)) != NULL) {\n list *l = dictGetVal(de);\n sds option = dictGetKey(de);\n\n /* Don't blank lines about options the rewrite process\n * don't understand. */\n if (dictFind(state->rewritten,option) == NULL) {\n redisLog(REDIS_DEBUG,\"Not rewritten option: %s\", option);\n continue;\n }\n\n while(listLength(l)) {\n listNode *ln = listFirst(l);\n int linenum = (long) ln->value;\n\n sdsfree(state->lines[linenum]);\n state->lines[linenum] = sdsempty();\n listDelNode(l,ln);\n }\n }\n dictReleaseIterator(di);\n}\n\n/* This function overwrites the old configuration file with the new content.\n *\n * 1) The old file length is obtained.\n * 2) If the new content is smaller, padding is added.\n * 3) A single write(2) call is used to replace the content of the file.\n * 4) Later the file is truncated to the length of the new content.\n *\n * This way we are sure the file is left in a consistent state even if the\n * process is stopped between any of the four operations.\n *\n * The function returns 0 on success, otherwise -1 is returned and errno\n * set accordingly. */\nint rewriteConfigOverwriteFile(char *configfile, sds content) {\n int retval = 0;\n int fd = open(configfile,O_RDWR|O_CREAT,0644);\n int content_size = sdslen(content), padding = 0;\n struct stat sb;\n sds content_padded;\n\n /* 1) Open the old file (or create a new one if it does not\n * exist), get the size. */\n if (fd == -1) return -1; /* errno set by open(). */\n if (fstat(fd,&sb) == -1) {\n close(fd);\n return -1; /* errno set by fstat(). */\n }\n\n /* 2) Pad the content at least match the old file size. */\n content_padded = sdsdup(content);\n if (content_size < sb.st_size) {\n /* If the old file was bigger, pad the content with\n * a newline plus as many \"#\" chars as required. */\n padding = sb.st_size - content_size;\n content_padded = sdsgrowzero(content_padded,sb.st_size);\n content_padded[content_size] = '\\n';\n memset(content_padded+content_size+1,'#',padding-1);\n }\n\n /* 3) Write the new content using a single write(2). */\n if (write(fd,content_padded,strlen(content_padded)) == -1) {\n retval = -1;\n goto cleanup;\n }\n\n /* 4) Truncate the file to the right length if we used padding. */\n if (padding) {\n if (ftruncate(fd,content_size) == -1) {\n /* Non critical error... */\n }\n }\n\ncleanup:\n sdsfree(content_padded);\n close(fd);\n return retval;\n}\n\n/* Rewrite the configuration file at \"path\".\n * If the configuration file already exists, we try at best to retain comments\n * and overall structure.\n *\n * Configuration parameters that are at their default value, unless already\n * explicitly included in the old configuration file, are not rewritten.\n *\n * On error -1 is returned and errno is set accordingly, otherwise 0. */\nint rewriteConfig(char *path) {\n struct rewriteConfigState *state;\n sds newcontent;\n int retval;\n\n /* Step 1: read the old config into our rewrite state. */\n if ((state = rewriteConfigReadOldFile(path)) == NULL) return -1;\n\n /* Step 2: rewrite every single option, replacing or appending it inside\n * the rewrite state. */\n\n rewriteConfigYesNoOption(state,\"daemonize\",server.daemonize,0);\n rewriteConfigStringOption(state,\"pidfile\",server.pidfile,REDIS_DEFAULT_PID_FILE);\n rewriteConfigNumericalOption(state,\"port\",server.port,REDIS_SERVERPORT);\n rewriteConfigNumericalOption(state,\"tcp-backlog\",server.tcp_backlog,REDIS_TCP_BACKLOG);\n rewriteConfigBindOption(state);\n rewriteConfigStringOption(state,\"unixsocket\",server.unixsocket,NULL);\n rewriteConfigOctalOption(state,\"unixsocketperm\",server.unixsocketperm,REDIS_DEFAULT_UNIX_SOCKET_PERM);\n rewriteConfigNumericalOption(state,\"timeout\",server.maxidletime,REDIS_MAXIDLETIME);\n rewriteConfigNumericalOption(state,\"tcp-keepalive\",server.tcpkeepalive,REDIS_DEFAULT_TCP_KEEPALIVE);\n rewriteConfigEnumOption(state,\"loglevel\",server.verbosity,\n \"debug\", REDIS_DEBUG,\n \"verbose\", REDIS_VERBOSE,\n \"notice\", REDIS_NOTICE,\n \"warning\", REDIS_WARNING,\n NULL, REDIS_DEFAULT_VERBOSITY);\n rewriteConfigStringOption(state,\"logfile\",server.logfile,REDIS_DEFAULT_LOGFILE);\n rewriteConfigYesNoOption(state,\"syslog-enabled\",server.syslog_enabled,REDIS_DEFAULT_SYSLOG_ENABLED);\n rewriteConfigStringOption(state,\"syslog-ident\",server.syslog_ident,REDIS_DEFAULT_SYSLOG_IDENT);\n rewriteConfigSyslogfacilityOption(state);\n rewriteConfigSaveOption(state);\n rewriteConfigNumericalOption(state,\"databases\",server.dbnum,REDIS_DEFAULT_DBNUM);\n rewriteConfigYesNoOption(state,\"stop-writes-on-bgsave-error\",server.stop_writes_on_bgsave_err,REDIS_DEFAULT_STOP_WRITES_ON_BGSAVE_ERROR);\n rewriteConfigYesNoOption(state,\"rdbcompression\",server.rdb_compression,REDIS_DEFAULT_RDB_COMPRESSION);\n rewriteConfigYesNoOption(state,\"rdbchecksum\",server.rdb_checksum,REDIS_DEFAULT_RDB_CHECKSUM);\n rewriteConfigStringOption(state,\"dbfilename\",server.rdb_filename,REDIS_DEFAULT_RDB_FILENAME);\n rewriteConfigDirOption(state);\n rewriteConfigSlaveofOption(state);\n rewriteConfigStringOption(state,\"masterauth\",server.masterauth,NULL);\n rewriteConfigYesNoOption(state,\"slave-serve-stale-data\",server.repl_serve_stale_data,REDIS_DEFAULT_SLAVE_SERVE_STALE_DATA);\n rewriteConfigYesNoOption(state,\"slave-read-only\",server.repl_slave_ro,REDIS_DEFAULT_SLAVE_READ_ONLY);\n rewriteConfigNumericalOption(state,\"repl-ping-slave-period\",server.repl_ping_slave_period,REDIS_REPL_PING_SLAVE_PERIOD);\n rewriteConfigNumericalOption(state,\"repl-timeout\",server.repl_timeout,REDIS_REPL_TIMEOUT);\n rewriteConfigBytesOption(state,\"repl-backlog-size\",server.repl_backlog_size,REDIS_DEFAULT_REPL_BACKLOG_SIZE);\n rewriteConfigBytesOption(state,\"repl-backlog-ttl\",server.repl_backlog_time_limit,REDIS_DEFAULT_REPL_BACKLOG_TIME_LIMIT);\n rewriteConfigYesNoOption(state,\"repl-disable-tcp-nodelay\",server.repl_disable_tcp_nodelay,REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY);\n rewriteConfigNumericalOption(state,\"slave-priority\",server.slave_priority,REDIS_DEFAULT_SLAVE_PRIORITY);\n rewriteConfigNumericalOption(state,\"min-slaves-to-write\",server.repl_min_slaves_to_write,REDIS_DEFAULT_MIN_SLAVES_TO_WRITE);\n rewriteConfigNumericalOption(state,\"min-slaves-max-lag\",server.repl_min_slaves_max_lag,REDIS_DEFAULT_MIN_SLAVES_MAX_LAG);\n rewriteConfigStringOption(state,\"requirepass\",server.requirepass,NULL);\n rewriteConfigNumericalOption(state,\"maxclients\",server.maxclients,REDIS_MAX_CLIENTS);\n rewriteConfigBytesOption(state,\"maxmemory\",server.maxmemory,REDIS_DEFAULT_MAXMEMORY);\n rewriteConfigEnumOption(state,\"maxmemory-policy\",server.maxmemory_policy,\n \"volatile-lru\", REDIS_MAXMEMORY_VOLATILE_LRU,\n \"allkeys-lru\", REDIS_MAXMEMORY_ALLKEYS_LRU,\n \"volatile-random\", REDIS_MAXMEMORY_VOLATILE_RANDOM,\n \"allkeys-random\", REDIS_MAXMEMORY_ALLKEYS_RANDOM,\n \"volatile-ttl\", REDIS_MAXMEMORY_VOLATILE_TTL,\n \"noeviction\", REDIS_MAXMEMORY_NO_EVICTION,\n NULL, REDIS_DEFAULT_MAXMEMORY_POLICY);\n rewriteConfigNumericalOption(state,\"maxmemory-samples\",server.maxmemory_samples,REDIS_DEFAULT_MAXMEMORY_SAMPLES);\n rewriteConfigYesNoOption(state,\"appendonly\",server.aof_state != REDIS_AOF_OFF,0);\n rewriteConfigStringOption(state,\"appendfilename\",server.aof_filename,REDIS_DEFAULT_AOF_FILENAME);\n rewriteConfigEnumOption(state,\"appendfsync\",server.aof_fsync,\n \"everysec\", AOF_FSYNC_EVERYSEC,\n \"always\", AOF_FSYNC_ALWAYS,\n \"no\", AOF_FSYNC_NO,\n NULL, REDIS_DEFAULT_AOF_FSYNC);\n rewriteConfigYesNoOption(state,\"no-appendfsync-on-rewrite\",server.aof_no_fsync_on_rewrite,REDIS_DEFAULT_AOF_NO_FSYNC_ON_REWRITE);\n rewriteConfigNumericalOption(state,\"auto-aof-rewrite-percentage\",server.aof_rewrite_perc,REDIS_AOF_REWRITE_PERC);\n rewriteConfigBytesOption(state,\"auto-aof-rewrite-min-size\",server.aof_rewrite_min_size,REDIS_AOF_REWRITE_MIN_SIZE);\n rewriteConfigNumericalOption(state,\"lua-time-limit\",server.lua_time_limit,REDIS_LUA_TIME_LIMIT);\n rewriteConfigYesNoOption(state,\"cluster-enabled\",server.cluster_enabled,0);\n rewriteConfigStringOption(state,\"cluster-config-file\",server.cluster_configfile,REDIS_DEFAULT_CLUSTER_CONFIG_FILE);\n rewriteConfigNumericalOption(state,\"cluster-node-timeout\",server.cluster_node_timeout,REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT);\n rewriteConfigNumericalOption(state,\"cluster-migration-barrier\",server.cluster_migration_barrier,REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER);\n rewriteConfigNumericalOption(state,\"slowlog-log-slower-than\",server.slowlog_log_slower_than,REDIS_SLOWLOG_LOG_SLOWER_THAN);\n rewriteConfigNumericalOption(state,\"slowlog-max-len\",server.slowlog_max_len,REDIS_SLOWLOG_MAX_LEN);\n rewriteConfigNotifykeyspaceeventsOption(state);\n rewriteConfigNumericalOption(state,\"hash-max-ziplist-entries\",server.hash_max_ziplist_entries,REDIS_HASH_MAX_ZIPLIST_ENTRIES);\n rewriteConfigNumericalOption(state,\"hash-max-ziplist-value\",server.hash_max_ziplist_value,REDIS_HASH_MAX_ZIPLIST_VALUE);\n rewriteConfigNumericalOption(state,\"list-max-ziplist-entries\",server.list_max_ziplist_entries,REDIS_LIST_MAX_ZIPLIST_ENTRIES);\n rewriteConfigNumericalOption(state,\"list-max-ziplist-value\",server.list_max_ziplist_value,REDIS_LIST_MAX_ZIPLIST_VALUE);\n rewriteConfigNumericalOption(state,\"set-max-intset-entries\",server.set_max_intset_entries,REDIS_SET_MAX_INTSET_ENTRIES);\n rewriteConfigNumericalOption(state,\"zset-max-ziplist-entries\",server.zset_max_ziplist_entries,REDIS_ZSET_MAX_ZIPLIST_ENTRIES);\n rewriteConfigNumericalOption(state,\"zset-max-ziplist-value\",server.zset_max_ziplist_value,REDIS_ZSET_MAX_ZIPLIST_VALUE);\n rewriteConfigNumericalOption(state,\"hll-sparse-max-bytes\",server.hll_sparse_max_bytes,REDIS_DEFAULT_HLL_SPARSE_MAX_BYTES);\n rewriteConfigYesNoOption(state,\"activerehashing\",server.activerehashing,REDIS_DEFAULT_ACTIVE_REHASHING);\n rewriteConfigClientoutputbufferlimitOption(state);\n rewriteConfigNumericalOption(state,\"hz\",server.hz,REDIS_DEFAULT_HZ);\n rewriteConfigYesNoOption(state,\"aof-rewrite-incremental-fsync\",server.aof_rewrite_incremental_fsync,REDIS_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC);\n if (server.sentinel_mode) rewriteConfigSentinelOption(state);\n\n /* Step 3: remove all the orphaned lines in the old file, that is, lines\n * that were used by a config option and are no longer used, like in case\n * of multiple \"save\" options or duplicated options. */\n rewriteConfigRemoveOrphaned(state);\n\n /* Step 4: generate a new configuration file from the modified state\n * and write it into the original file. */\n newcontent = rewriteConfigGetContentFromState(state);\n retval = rewriteConfigOverwriteFile(server.configfile,newcontent);\n\n sdsfree(newcontent);\n rewriteConfigReleaseState(state);\n return retval;\n}\n\n/*-----------------------------------------------------------------------------\n * CONFIG command entry point\n *----------------------------------------------------------------------------*/\n\nvoid configCommand(redisClient *c) {\n if (!strcasecmp(c->argv[1]->ptr,\"set\")) {\n if (c->argc != 4) goto badarity;\n configSetCommand(c);\n } else if (!strcasecmp(c->argv[1]->ptr,\"get\")) {\n if (c->argc != 3) goto badarity;\n configGetCommand(c);\n } else if (!strcasecmp(c->argv[1]->ptr,\"resetstat\")) {\n if (c->argc != 2) goto badarity;\n resetServerStats();\n resetCommandTableStats();\n addReply(c,shared.ok);\n } else if (!strcasecmp(c->argv[1]->ptr,\"rewrite\")) {\n if (c->argc != 2) goto badarity;\n if (server.configfile == NULL) {\n addReplyError(c,\"The server is running without a config file\");\n return;\n }\n if (rewriteConfig(server.configfile) == -1) {\n redisLog(REDIS_WARNING,\"CONFIG REWRITE failed: %s\", strerror(errno));\n addReplyErrorFormat(c,\"Rewriting config file: %s\", strerror(errno));\n } else {\n redisLog(REDIS_WARNING,\"CONFIG REWRITE executed with success.\");\n addReply(c,shared.ok);\n }\n } else {\n addReplyError(c,\n \"CONFIG subcommand must be one of GET, SET, RESETSTAT, REWRITE\");\n }\n return;\n\nbadarity:\n addReplyErrorFormat(c,\"Wrong number of arguments for CONFIG %s\",\n (char*) c->argv[1]->ptr);\n}\n" }, { "path": "src/config.h", "content": "/*\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#ifndef __CONFIG_H\n#define __CONFIG_H\n\n#ifdef __APPLE__\n#include \n#endif\n\n/* Define redis_fstat to fstat or fstat64() */\n#if defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6)\n#define redis_fstat fstat64\n#define redis_stat stat64\n#else\n#define redis_fstat fstat\n#define redis_stat stat\n#endif\n\n/* Test for proc filesystem */\n#ifdef __linux__\n#define HAVE_PROC_STAT 1\n#define HAVE_PROC_MAPS 1\n#define HAVE_PROC_SMAPS 1\n#endif\n\n/* Test for task_info() */\n#if defined(__APPLE__)\n#define HAVE_TASKINFO 1\n#endif\n\n/* Test for backtrace() */\n#if defined(__APPLE__) || defined(__linux__)\n#define HAVE_BACKTRACE 1\n#endif\n\n/* Test for polling API */\n#ifdef __linux__\n#define HAVE_EPOLL 1\n#endif\n\n#if (defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6)) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined (__NetBSD__)\n#define HAVE_KQUEUE 1\n#endif\n\n#ifdef __sun\n#include \n#ifdef _DTRACE_VERSION\n#define HAVE_EVPORT 1\n#endif\n#endif\n\n/* Define aof_fsync to fdatasync() in Linux and fsync() for all the rest */\n#ifdef __linux__\n#define aof_fsync fdatasync\n#else\n#define aof_fsync fsync\n#endif\n\n/* Define rdb_fsync_range to sync_file_range() on Linux, otherwise we use\n * the plain fsync() call. */\n#ifdef __linux__\n#include \n#include \n#if defined(__GLIBC__) && defined(__GLIBC_PREREQ)\n#if (LINUX_VERSION_CODE >= 0x020611 && __GLIBC_PREREQ(2, 6))\n#define HAVE_SYNC_FILE_RANGE 1\n#endif\n#else\n#if (LINUX_VERSION_CODE >= 0x020611)\n#define HAVE_SYNC_FILE_RANGE 1\n#endif\n#endif\n#endif\n\n#ifdef HAVE_SYNC_FILE_RANGE\n#define rdb_fsync_range(fd,off,size) sync_file_range(fd,off,size,SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE)\n#else\n#define rdb_fsync_range(fd,off,size) fsync(fd)\n#endif\n\n/* Check if we can use setproctitle().\n * BSD systems have support for it, we provide an implementation for\n * Linux and osx. */\n#if (defined __NetBSD__ || defined __FreeBSD__ || defined __OpenBSD__)\n#define USE_SETPROCTITLE\n#endif\n\n#if (defined __linux || defined __APPLE__)\n#define USE_SETPROCTITLE\n#define INIT_SETPROCTITLE_REPLACEMENT\nvoid spt_init(int argc, char *argv[]);\nvoid setproctitle(const char *fmt, ...);\n#endif\n\n/* Byte ordering detection */\n#include /* This will likely define BYTE_ORDER */\n\n#ifndef BYTE_ORDER\n#if (BSD >= 199103)\n# include \n#else\n#if defined(linux) || defined(__linux__)\n# include \n#else\n#define\tLITTLE_ENDIAN\t1234\t/* least-significant byte first (vax, pc) */\n#define\tBIG_ENDIAN\t4321\t/* most-significant byte first (IBM, net) */\n#define\tPDP_ENDIAN\t3412\t/* LSB first in word, MSW first in long (pdp)*/\n\n#if defined(__i386__) || defined(__x86_64__) || defined(__amd64__) || \\\n defined(vax) || defined(ns32000) || defined(sun386) || \\\n defined(MIPSEL) || defined(_MIPSEL) || defined(BIT_ZERO_ON_RIGHT) || \\\n defined(__alpha__) || defined(__alpha)\n#define BYTE_ORDER LITTLE_ENDIAN\n#endif\n\n#if defined(sel) || defined(pyr) || defined(mc68000) || defined(sparc) || \\\n defined(is68k) || defined(tahoe) || defined(ibm032) || defined(ibm370) || \\\n defined(MIPSEB) || defined(_MIPSEB) || defined(_IBMR2) || defined(DGUX) ||\\\n defined(apollo) || defined(__convex__) || defined(_CRAY) || \\\n defined(__hppa) || defined(__hp9000) || \\\n defined(__hp9000s300) || defined(__hp9000s700) || \\\n defined (BIT_ZERO_ON_LEFT) || defined(m68k) || defined(__sparc)\n#define BYTE_ORDER\tBIG_ENDIAN\n#endif\n#endif /* linux */\n#endif /* BSD */\n#endif /* BYTE_ORDER */\n\n/* Sometimes after including an OS-specific header that defines the\n * endianess we end with __BYTE_ORDER but not with BYTE_ORDER that is what\n * the Redis code uses. In this case let's define everything without the\n * underscores. */\n#ifndef BYTE_ORDER\n#ifdef __BYTE_ORDER\n#if defined(__LITTLE_ENDIAN) && defined(__BIG_ENDIAN)\n#ifndef LITTLE_ENDIAN\n#define LITTLE_ENDIAN __LITTLE_ENDIAN\n#endif\n#ifndef BIG_ENDIAN\n#define BIG_ENDIAN __BIG_ENDIAN\n#endif\n#if (__BYTE_ORDER == __LITTLE_ENDIAN)\n#define BYTE_ORDER LITTLE_ENDIAN\n#else\n#define BYTE_ORDER BIG_ENDIAN\n#endif\n#endif\n#endif\n#endif\n\n#if !defined(BYTE_ORDER) || \\\n (BYTE_ORDER != BIG_ENDIAN && BYTE_ORDER != LITTLE_ENDIAN)\n\t/* you must determine what the correct bit order is for\n\t * your compiler - the next line is an intentional error\n\t * which will force your compiles to bomb until you fix\n\t * the above macros.\n\t */\n#error \"Undefined or invalid BYTE_ORDER\"\n#endif\n\n#if (__i386 || __amd64) && __GNUC__\n#define GNUC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)\n#if (GNUC_VERSION >= 40100) || defined(__clang__)\n#define HAVE_ATOMIC\n#endif\n#endif\n\n#endif\n" }, { "path": "src/crc16.c", "content": "#include \"redis.h\"\n\n/* \n * Copyright 2001-2010 Georges Menie (www.menie.org)\n * Copyright 2010-2012 Salvatore Sanfilippo (adapted to Redis coding style)\n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright\n * notice, this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of the University of California, Berkeley nor the\n * names of its contributors may be used to endorse or promote products\n * derived from this software without specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND ANY\n * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\n * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n * DISCLAIMED. IN NO EVENT SHALL THE REGENTS AND CONTRIBUTORS BE LIABLE FOR ANY\n * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES\n * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;\n * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND\n * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS\n * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n */\n\n/* CRC16 implementation according to CCITT standards.\n *\n * Note by @antirez: this is actually the XMODEM CRC 16 algorithm, using the\n * following parameters:\n *\n * Name : \"XMODEM\", also known as \"ZMODEM\", \"CRC-16/ACORN\"\n * Width : 16 bit\n * Poly : 1021 (That is actually x^16 + x^12 + x^5 + 1)\n * Initialization : 0000\n * Reflect Input byte : False\n * Reflect Output CRC : False\n * Xor constant to output CRC : 0000\n * Output for \"123456789\" : 31C3\n */\n\nstatic const uint16_t crc16tab[256]= {\n 0x0000,0x1021,0x2042,0x3063,0x4084,0x50a5,0x60c6,0x70e7,\n 0x8108,0x9129,0xa14a,0xb16b,0xc18c,0xd1ad,0xe1ce,0xf1ef,\n 0x1231,0x0210,0x3273,0x2252,0x52b5,0x4294,0x72f7,0x62d6,\n 0x9339,0x8318,0xb37b,0xa35a,0xd3bd,0xc39c,0xf3ff,0xe3de,\n 0x2462,0x3443,0x0420,0x1401,0x64e6,0x74c7,0x44a4,0x5485,\n 0xa56a,0xb54b,0x8528,0x9509,0xe5ee,0xf5cf,0xc5ac,0xd58d,\n 0x3653,0x2672,0x1611,0x0630,0x76d7,0x66f6,0x5695,0x46b4,\n 0xb75b,0xa77a,0x9719,0x8738,0xf7df,0xe7fe,0xd79d,0xc7bc,\n 0x48c4,0x58e5,0x6886,0x78a7,0x0840,0x1861,0x2802,0x3823,\n 0xc9cc,0xd9ed,0xe98e,0xf9af,0x8948,0x9969,0xa90a,0xb92b,\n 0x5af5,0x4ad4,0x7ab7,0x6a96,0x1a71,0x0a50,0x3a33,0x2a12,\n 0xdbfd,0xcbdc,0xfbbf,0xeb9e,0x9b79,0x8b58,0xbb3b,0xab1a,\n 0x6ca6,0x7c87,0x4ce4,0x5cc5,0x2c22,0x3c03,0x0c60,0x1c41,\n 0xedae,0xfd8f,0xcdec,0xddcd,0xad2a,0xbd0b,0x8d68,0x9d49,\n 0x7e97,0x6eb6,0x5ed5,0x4ef4,0x3e13,0x2e32,0x1e51,0x0e70,\n 0xff9f,0xefbe,0xdfdd,0xcffc,0xbf1b,0xaf3a,0x9f59,0x8f78,\n 0x9188,0x81a9,0xb1ca,0xa1eb,0xd10c,0xc12d,0xf14e,0xe16f,\n 0x1080,0x00a1,0x30c2,0x20e3,0x5004,0x4025,0x7046,0x6067,\n 0x83b9,0x9398,0xa3fb,0xb3da,0xc33d,0xd31c,0xe37f,0xf35e,\n 0x02b1,0x1290,0x22f3,0x32d2,0x4235,0x5214,0x6277,0x7256,\n 0xb5ea,0xa5cb,0x95a8,0x8589,0xf56e,0xe54f,0xd52c,0xc50d,\n 0x34e2,0x24c3,0x14a0,0x0481,0x7466,0x6447,0x5424,0x4405,\n 0xa7db,0xb7fa,0x8799,0x97b8,0xe75f,0xf77e,0xc71d,0xd73c,\n 0x26d3,0x36f2,0x0691,0x16b0,0x6657,0x7676,0x4615,0x5634,\n 0xd94c,0xc96d,0xf90e,0xe92f,0x99c8,0x89e9,0xb98a,0xa9ab,\n 0x5844,0x4865,0x7806,0x6827,0x18c0,0x08e1,0x3882,0x28a3,\n 0xcb7d,0xdb5c,0xeb3f,0xfb1e,0x8bf9,0x9bd8,0xabbb,0xbb9a,\n 0x4a75,0x5a54,0x6a37,0x7a16,0x0af1,0x1ad0,0x2ab3,0x3a92,\n 0xfd2e,0xed0f,0xdd6c,0xcd4d,0xbdaa,0xad8b,0x9de8,0x8dc9,\n 0x7c26,0x6c07,0x5c64,0x4c45,0x3ca2,0x2c83,0x1ce0,0x0cc1,\n 0xef1f,0xff3e,0xcf5d,0xdf7c,0xaf9b,0xbfba,0x8fd9,0x9ff8,\n 0x6e17,0x7e36,0x4e55,0x5e74,0x2e93,0x3eb2,0x0ed1,0x1ef0\n};\n \nuint16_t crc16(const char *buf, int len) {\n int counter;\n uint16_t crc = 0;\n for (counter = 0; counter < len; counter++)\n crc = (crc<<8) ^ crc16tab[((crc>>8) ^ *buf++)&0x00FF];\n return crc;\n}\n" }, { "path": "src/crc64.c", "content": "/* Redis uses the CRC64 variant with \"Jones\" coefficients and init value of 0.\n *\n * Specification of this CRC64 variant follows:\n * Name: crc-64-jones\n * Width: 64 bites\n * Poly: 0xad93d23594c935a9\n * Reflected In: True\n * Xor_In: 0xffffffffffffffff\n * Reflected_Out: True\n * Xor_Out: 0x0\n * Check(\"123456789\"): 0xe9c6d914c4b8d9ca\n *\n * Copyright (c) 2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE. */\n\n#include \n\nstatic const uint64_t crc64_tab[256] = {\n UINT64_C(0x0000000000000000), UINT64_C(0x7ad870c830358979),\n UINT64_C(0xf5b0e190606b12f2), UINT64_C(0x8f689158505e9b8b),\n UINT64_C(0xc038e5739841b68f), UINT64_C(0xbae095bba8743ff6),\n UINT64_C(0x358804e3f82aa47d), UINT64_C(0x4f50742bc81f2d04),\n UINT64_C(0xab28ecb46814fe75), UINT64_C(0xd1f09c7c5821770c),\n UINT64_C(0x5e980d24087fec87), UINT64_C(0x24407dec384a65fe),\n UINT64_C(0x6b1009c7f05548fa), UINT64_C(0x11c8790fc060c183),\n UINT64_C(0x9ea0e857903e5a08), UINT64_C(0xe478989fa00bd371),\n UINT64_C(0x7d08ff3b88be6f81), UINT64_C(0x07d08ff3b88be6f8),\n UINT64_C(0x88b81eabe8d57d73), UINT64_C(0xf2606e63d8e0f40a),\n UINT64_C(0xbd301a4810ffd90e), UINT64_C(0xc7e86a8020ca5077),\n UINT64_C(0x4880fbd87094cbfc), UINT64_C(0x32588b1040a14285),\n UINT64_C(0xd620138fe0aa91f4), UINT64_C(0xacf86347d09f188d),\n UINT64_C(0x2390f21f80c18306), UINT64_C(0x594882d7b0f40a7f),\n UINT64_C(0x1618f6fc78eb277b), UINT64_C(0x6cc0863448deae02),\n UINT64_C(0xe3a8176c18803589), UINT64_C(0x997067a428b5bcf0),\n UINT64_C(0xfa11fe77117cdf02), UINT64_C(0x80c98ebf2149567b),\n UINT64_C(0x0fa11fe77117cdf0), UINT64_C(0x75796f2f41224489),\n UINT64_C(0x3a291b04893d698d), UINT64_C(0x40f16bccb908e0f4),\n UINT64_C(0xcf99fa94e9567b7f), UINT64_C(0xb5418a5cd963f206),\n UINT64_C(0x513912c379682177), UINT64_C(0x2be1620b495da80e),\n UINT64_C(0xa489f35319033385), UINT64_C(0xde51839b2936bafc),\n UINT64_C(0x9101f7b0e12997f8), UINT64_C(0xebd98778d11c1e81),\n UINT64_C(0x64b116208142850a), UINT64_C(0x1e6966e8b1770c73),\n UINT64_C(0x8719014c99c2b083), UINT64_C(0xfdc17184a9f739fa),\n UINT64_C(0x72a9e0dcf9a9a271), UINT64_C(0x08719014c99c2b08),\n UINT64_C(0x4721e43f0183060c), UINT64_C(0x3df994f731b68f75),\n UINT64_C(0xb29105af61e814fe), UINT64_C(0xc849756751dd9d87),\n UINT64_C(0x2c31edf8f1d64ef6), UINT64_C(0x56e99d30c1e3c78f),\n UINT64_C(0xd9810c6891bd5c04), UINT64_C(0xa3597ca0a188d57d),\n UINT64_C(0xec09088b6997f879), UINT64_C(0x96d1784359a27100),\n UINT64_C(0x19b9e91b09fcea8b), UINT64_C(0x636199d339c963f2),\n UINT64_C(0xdf7adabd7a6e2d6f), UINT64_C(0xa5a2aa754a5ba416),\n UINT64_C(0x2aca3b2d1a053f9d), UINT64_C(0x50124be52a30b6e4),\n UINT64_C(0x1f423fcee22f9be0), UINT64_C(0x659a4f06d21a1299),\n UINT64_C(0xeaf2de5e82448912), UINT64_C(0x902aae96b271006b),\n UINT64_C(0x74523609127ad31a), UINT64_C(0x0e8a46c1224f5a63),\n UINT64_C(0x81e2d7997211c1e8), UINT64_C(0xfb3aa75142244891),\n UINT64_C(0xb46ad37a8a3b6595), UINT64_C(0xceb2a3b2ba0eecec),\n UINT64_C(0x41da32eaea507767), UINT64_C(0x3b024222da65fe1e),\n UINT64_C(0xa2722586f2d042ee), UINT64_C(0xd8aa554ec2e5cb97),\n UINT64_C(0x57c2c41692bb501c), UINT64_C(0x2d1ab4dea28ed965),\n UINT64_C(0x624ac0f56a91f461), UINT64_C(0x1892b03d5aa47d18),\n UINT64_C(0x97fa21650afae693), UINT64_C(0xed2251ad3acf6fea),\n UINT64_C(0x095ac9329ac4bc9b), UINT64_C(0x7382b9faaaf135e2),\n UINT64_C(0xfcea28a2faafae69), UINT64_C(0x8632586aca9a2710),\n UINT64_C(0xc9622c4102850a14), UINT64_C(0xb3ba5c8932b0836d),\n UINT64_C(0x3cd2cdd162ee18e6), UINT64_C(0x460abd1952db919f),\n UINT64_C(0x256b24ca6b12f26d), UINT64_C(0x5fb354025b277b14),\n UINT64_C(0xd0dbc55a0b79e09f), UINT64_C(0xaa03b5923b4c69e6),\n UINT64_C(0xe553c1b9f35344e2), UINT64_C(0x9f8bb171c366cd9b),\n UINT64_C(0x10e3202993385610), UINT64_C(0x6a3b50e1a30ddf69),\n UINT64_C(0x8e43c87e03060c18), UINT64_C(0xf49bb8b633338561),\n UINT64_C(0x7bf329ee636d1eea), UINT64_C(0x012b592653589793),\n UINT64_C(0x4e7b2d0d9b47ba97), UINT64_C(0x34a35dc5ab7233ee),\n UINT64_C(0xbbcbcc9dfb2ca865), UINT64_C(0xc113bc55cb19211c),\n UINT64_C(0x5863dbf1e3ac9dec), UINT64_C(0x22bbab39d3991495),\n UINT64_C(0xadd33a6183c78f1e), UINT64_C(0xd70b4aa9b3f20667),\n UINT64_C(0x985b3e827bed2b63), UINT64_C(0xe2834e4a4bd8a21a),\n UINT64_C(0x6debdf121b863991), UINT64_C(0x1733afda2bb3b0e8),\n UINT64_C(0xf34b37458bb86399), UINT64_C(0x8993478dbb8deae0),\n UINT64_C(0x06fbd6d5ebd3716b), UINT64_C(0x7c23a61ddbe6f812),\n UINT64_C(0x3373d23613f9d516), UINT64_C(0x49aba2fe23cc5c6f),\n UINT64_C(0xc6c333a67392c7e4), UINT64_C(0xbc1b436e43a74e9d),\n UINT64_C(0x95ac9329ac4bc9b5), UINT64_C(0xef74e3e19c7e40cc),\n UINT64_C(0x601c72b9cc20db47), UINT64_C(0x1ac40271fc15523e),\n UINT64_C(0x5594765a340a7f3a), UINT64_C(0x2f4c0692043ff643),\n UINT64_C(0xa02497ca54616dc8), UINT64_C(0xdafce7026454e4b1),\n UINT64_C(0x3e847f9dc45f37c0), UINT64_C(0x445c0f55f46abeb9),\n UINT64_C(0xcb349e0da4342532), UINT64_C(0xb1eceec59401ac4b),\n UINT64_C(0xfebc9aee5c1e814f), UINT64_C(0x8464ea266c2b0836),\n UINT64_C(0x0b0c7b7e3c7593bd), UINT64_C(0x71d40bb60c401ac4),\n UINT64_C(0xe8a46c1224f5a634), UINT64_C(0x927c1cda14c02f4d),\n UINT64_C(0x1d148d82449eb4c6), UINT64_C(0x67ccfd4a74ab3dbf),\n UINT64_C(0x289c8961bcb410bb), UINT64_C(0x5244f9a98c8199c2),\n UINT64_C(0xdd2c68f1dcdf0249), UINT64_C(0xa7f41839ecea8b30),\n UINT64_C(0x438c80a64ce15841), UINT64_C(0x3954f06e7cd4d138),\n UINT64_C(0xb63c61362c8a4ab3), UINT64_C(0xcce411fe1cbfc3ca),\n UINT64_C(0x83b465d5d4a0eece), UINT64_C(0xf96c151de49567b7),\n UINT64_C(0x76048445b4cbfc3c), UINT64_C(0x0cdcf48d84fe7545),\n UINT64_C(0x6fbd6d5ebd3716b7), UINT64_C(0x15651d968d029fce),\n UINT64_C(0x9a0d8ccedd5c0445), UINT64_C(0xe0d5fc06ed698d3c),\n UINT64_C(0xaf85882d2576a038), UINT64_C(0xd55df8e515432941),\n UINT64_C(0x5a3569bd451db2ca), UINT64_C(0x20ed197575283bb3),\n UINT64_C(0xc49581ead523e8c2), UINT64_C(0xbe4df122e51661bb),\n UINT64_C(0x3125607ab548fa30), UINT64_C(0x4bfd10b2857d7349),\n UINT64_C(0x04ad64994d625e4d), UINT64_C(0x7e7514517d57d734),\n UINT64_C(0xf11d85092d094cbf), UINT64_C(0x8bc5f5c11d3cc5c6),\n UINT64_C(0x12b5926535897936), UINT64_C(0x686de2ad05bcf04f),\n UINT64_C(0xe70573f555e26bc4), UINT64_C(0x9ddd033d65d7e2bd),\n UINT64_C(0xd28d7716adc8cfb9), UINT64_C(0xa85507de9dfd46c0),\n UINT64_C(0x273d9686cda3dd4b), UINT64_C(0x5de5e64efd965432),\n UINT64_C(0xb99d7ed15d9d8743), UINT64_C(0xc3450e196da80e3a),\n UINT64_C(0x4c2d9f413df695b1), UINT64_C(0x36f5ef890dc31cc8),\n UINT64_C(0x79a59ba2c5dc31cc), UINT64_C(0x037deb6af5e9b8b5),\n UINT64_C(0x8c157a32a5b7233e), UINT64_C(0xf6cd0afa9582aa47),\n UINT64_C(0x4ad64994d625e4da), UINT64_C(0x300e395ce6106da3),\n UINT64_C(0xbf66a804b64ef628), UINT64_C(0xc5bed8cc867b7f51),\n UINT64_C(0x8aeeace74e645255), UINT64_C(0xf036dc2f7e51db2c),\n UINT64_C(0x7f5e4d772e0f40a7), UINT64_C(0x05863dbf1e3ac9de),\n UINT64_C(0xe1fea520be311aaf), UINT64_C(0x9b26d5e88e0493d6),\n UINT64_C(0x144e44b0de5a085d), UINT64_C(0x6e963478ee6f8124),\n UINT64_C(0x21c640532670ac20), UINT64_C(0x5b1e309b16452559),\n UINT64_C(0xd476a1c3461bbed2), UINT64_C(0xaeaed10b762e37ab),\n UINT64_C(0x37deb6af5e9b8b5b), UINT64_C(0x4d06c6676eae0222),\n UINT64_C(0xc26e573f3ef099a9), UINT64_C(0xb8b627f70ec510d0),\n UINT64_C(0xf7e653dcc6da3dd4), UINT64_C(0x8d3e2314f6efb4ad),\n UINT64_C(0x0256b24ca6b12f26), UINT64_C(0x788ec2849684a65f),\n UINT64_C(0x9cf65a1b368f752e), UINT64_C(0xe62e2ad306bafc57),\n UINT64_C(0x6946bb8b56e467dc), UINT64_C(0x139ecb4366d1eea5),\n UINT64_C(0x5ccebf68aecec3a1), UINT64_C(0x2616cfa09efb4ad8),\n UINT64_C(0xa97e5ef8cea5d153), UINT64_C(0xd3a62e30fe90582a),\n UINT64_C(0xb0c7b7e3c7593bd8), UINT64_C(0xca1fc72bf76cb2a1),\n UINT64_C(0x45775673a732292a), UINT64_C(0x3faf26bb9707a053),\n UINT64_C(0x70ff52905f188d57), UINT64_C(0x0a2722586f2d042e),\n UINT64_C(0x854fb3003f739fa5), UINT64_C(0xff97c3c80f4616dc),\n UINT64_C(0x1bef5b57af4dc5ad), UINT64_C(0x61372b9f9f784cd4),\n UINT64_C(0xee5fbac7cf26d75f), UINT64_C(0x9487ca0fff135e26),\n UINT64_C(0xdbd7be24370c7322), UINT64_C(0xa10fceec0739fa5b),\n UINT64_C(0x2e675fb4576761d0), UINT64_C(0x54bf2f7c6752e8a9),\n UINT64_C(0xcdcf48d84fe75459), UINT64_C(0xb71738107fd2dd20),\n UINT64_C(0x387fa9482f8c46ab), UINT64_C(0x42a7d9801fb9cfd2),\n UINT64_C(0x0df7adabd7a6e2d6), UINT64_C(0x772fdd63e7936baf),\n UINT64_C(0xf8474c3bb7cdf024), UINT64_C(0x829f3cf387f8795d),\n UINT64_C(0x66e7a46c27f3aa2c), UINT64_C(0x1c3fd4a417c62355),\n UINT64_C(0x935745fc4798b8de), UINT64_C(0xe98f353477ad31a7),\n UINT64_C(0xa6df411fbfb21ca3), UINT64_C(0xdc0731d78f8795da),\n UINT64_C(0x536fa08fdfd90e51), UINT64_C(0x29b7d047efec8728),\n};\n\nuint64_t crc64(uint64_t crc, const unsigned char *s, uint64_t l) {\n uint64_t j;\n\n for (j = 0; j < l; j++) {\n uint8_t byte = s[j];\n crc = crc64_tab[(uint8_t)crc ^ byte] ^ (crc >> 8);\n }\n return crc;\n}\n\n/* Test main */\n#ifdef TEST_MAIN\n#include \nint main(void) {\n printf(\"e9c6d914c4b8d9ca == %016llx\\n\",\n (unsigned long long) crc64(0,(unsigned char*)\"123456789\",9));\n return 0;\n}\n#endif\n" }, { "path": "src/crc64.h", "content": "#ifndef CRC64_H\n#define CRC64_H\n\n#include \n\nuint64_t crc64(uint64_t crc, const unsigned char *s, uint64_t l);\n\n#endif\n" }, { "path": "src/db.c", "content": "/*\n * Copyright (c) 2009-2012, Salvatore Sanfilippo \n * All rights reserved.\n *\n * Redistribution and use in source and binary forms, with or without\n * modification, are permitted provided that the following conditions are met:\n *\n * * Redistributions of source code must retain the above copyright notice,\n * this list of conditions and the following disclaimer.\n * * Redistributions in binary form must reproduce the above copyright\n * notice, this list of conditions and the following disclaimer in the\n * documentation and/or other materials provided with the distribution.\n * * Neither the name of Redis nor the names of its contributors may be used\n * to endorse or promote products derived from this software without\n * specific prior written permission.\n *\n * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\n * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\n * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\n * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\n * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\n * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\n * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\n * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\n * POSSIBILITY OF SUCH DAMAGE.\n */\n\n#include \"redis.h\"\n#include \"cluster.h\"\n\n#include \n#include \n\nvoid slotToKeyAdd(robj *key);\nvoid slotToKeyDel(robj *key);\nvoid slotToKeyFlush(void);\n\n/*-----------------------------------------------------------------------------\n * C-level DB API\n *----------------------------------------------------------------------------*/\n\n/*\n * 从数据库 db 中取出键 key 的值（对象）\n *\n * 如果 key 的值存在，那么返回该值；否则，返回 NULL 。\n */\nrobj *lookupKey(redisDb *db, robj *key) {\n\n // 查找键空间\n dictEntry *de = dictFind(db->dict,key->ptr);\n\n // 节点存在\n if (de) {\n \n\n // 取出值\n robj *val = dictGetVal(de);\n\n /* Update the access time for the ageing algorithm.\n * Don't do it if we have a saving child, as this will trigger\n * a copy on write madness. */\n // 更新时间信息（只在不存在子进程时执行，防止破坏 copy-on-write 机制）\n if (server.rdb_child_pid == -1 && server.aof_child_pid == -1)\n val->lru = LRU_CLOCK();\n\n // 返回值\n return val;\n } else {\n\n // 节点不存在\n\n return NULL;\n }\n}\n\n/*\n * 为执行读取操作而取出键 key 在数据库 db 中的值。\n *\n * 并根据是否成功找到值，更新服务器的命中/不命中信息。\n *\n * 找到时返回值对象，没找到返回 NULL 。\n */\nrobj *lookupKeyRead(redisDb *db, robj *key) {\n robj *val;\n\n // 检查 key 释放已经过期\n expireIfNeeded(db,key);\n\n // 从数据库中取出键的值\n val = lookupKey(db,key);\n\n // 更新命中/不命中信息\n if (val == NULL)\n server.stat_keyspace_misses++;\n else\n server.stat_keyspace_hits++;\n\n // 返回值\n return val;\n}\n\n/*\n * 为执行写入操作而取出键 key 在数据库 db 中的值。\n *\n * 和 lookupKeyRead 不同，这个函数不会更新服务器的命中/不命中信息。\n *\n * 找到时返回值对象，没找到返回 NULL 。\n */\nrobj *lookupKeyWrite(redisDb *db, robj *key) {\n\n // 删除过期键\n expireIfNeeded(db,key);\n\n // 查找并返回 key 的值对象\n return lookupKey(db,key);\n}\n\n/*\n * 为执行读取操作而从数据库中查找返回 key 的值。\n *\n * 如果 key 存在，那么返回 key 的值对象。\n *\n * 如果 key 不存在，那么向客户端发送 reply 参数中的信息，并返回 NULL 。\n */\nrobj *lookupKeyReadOrReply(redisClient *c, robj *key, robj *reply) {\n\n // 查找\n robj *o = lookupKeyRead(c->db, key);\n\n // 决定是否发送信息\n if (!o) addReply(c,reply);\n\n return o;\n}\n\n/*\n * 为执行写入操作而从数据库中查找返回 key 的值。\n *\n * 如果 key 存在，那么返回 key 的值对象。\n *\n * 如果 key 不存在，那么向客户端发送 reply 参数中的信息，并返回 NULL 。\n */\n\nrobj *lookupKeyWriteOrReply(redisClient *c, robj *key, robj *reply) {\n\n robj *o = lookupKeyWrite(c->db, key);\n\n if (!o) addReply(c,reply);\n\n return o;\n}\n\n/* Add the key to the DB. It's up to the caller to increment the reference\n * counter of the value if needed.\n *\n * 尝试将键值对 key 和 val 添加到数据库中。\n *\n * 调用者负责对 key 和 val 的引用计数进行增加。\n *\n * The program is aborted if the key already exists. \n *\n * 程序在键已经存在时会停止。\n */\nvoid dbAdd(redisDb *db, robj *key, robj *val) {\n\n // 复制键名\n sds copy = sdsdup(key->ptr);\n\n // 尝试添加键值对\n int retval = dictAdd(db->dict, copy, val);\n\n // 如果键已经存在，那么停止\n redisAssertWithInfo(NULL,key,retval == REDIS_OK);\n\n // 如果开启了集群模式，那么将键保存到槽里面\n if (server.cluster_enabled) slotToKeyAdd(key);\n }\n\n/* Overwrite an existing key with a new value. Incrementing the reference\n * count of the new value is up to the caller.\n *\n * 为已存在的键关联一个新值。\n *\n * 调用者负责对新值 val 的引用计数进行增加。\n *\n * This function does not modify the expire time of the existing key.\n *\n * 这个函数不会修改键的过期时间。\n *\n * The program is aborted if the key was not already present. \n *\n * 如果键不存在，那么函数停止。\n */\nvoid dbOverwrite(redisDb *db, robj *key, robj *val) {\n dictEntry *de = dictFind(db->dict,key->ptr);\n \n // 节点必须存在，否则中止\n redisAssertWithInfo(NULL,key,de != NULL);\n\n // 覆写旧值\n dictReplace(db->dict, key->ptr, val);\n}\n\n/* High level Set operation. This function can be used in order to set\n * a key, whatever it was existing or not, to a new object.\n *\n * 高层次的 SET 操作函数。\n *\n * 这个函数可以在不管键 key 是否存在的情况下，将它和 val 关联起来。\n *\n * 1) The ref count of the value object is incremented.\n * 值对象的引用计数会被增加\n *\n * 2) clients WATCHing for the destination key notified.\n * 监视键 key 的客户端会收到键已经被修改的通知\n *\n * 3) The expire time of the key is reset (the key is made persistent). \n * 键的过期时间会被移除（键变为持久的）\n */\nvoid setKey(redisDb *db, robj *key, robj *val) {\n\n // 添加或覆写数据库中的键值对\n if (lookupKeyWrite(db,key) == NULL) {\n dbAdd(db,key,val);\n } else {\n dbOverwrite(db,key,val);\n }\n\n incrRefCount(val);\n\n // 移除键的过期时间\n removeExpire(db,key);\n\n // 发送键修改通知\n signalModifiedKey(db,key);\n}\n\n/*\n * 检查键 key 是否存在于数据库中，存在返回 1 ，不存在返回 0 。\n */\nint dbExists(redisDb *db, robj *key) {\n return dictFind(db->dict,key->ptr) != NULL;\n}\n\n/* Return a random key, in form of a Redis object.\n * If there are no keys, NULL is returned.\n *\n * 随机从数据库中取出一个键，并以字符串对象的方式返回这个键。\n *\n * 如果数据库为空，那么返回 NULL 。\n *\n * The function makes sure to return keys not already expired. \n *\n * 这个函数保证被返回的键都是未过期的。\n */\nrobj *dbRandomKey(redisDb *db) {\n dictEntry *de;\n\n while(1) {\n sds key;\n robj *keyobj;\n\n // 从键空间中随机取出一个键节点\n de = dictGetRandomKey(db->dict);\n\n // 数据库为空\n if (de == NULL) return NULL;\n\n // 取出键\n key = dictGetKey(de);\n // 为键创建一个字符串对象，对象的值为键的名字\n keyobj = createStringObject(key,sdslen(key));\n // 检查键是否带有过期时间\n if (dictFind(db->expires,key)) {\n // 如果键已经过期，那么将它删除，并继续随机下个键\n if (expireIfNeeded(db,keyobj)) {\n decrRefCount(keyobj);\n continue; /* search for another key. This expired. */\n }\n }\n\n // 返回被随机到的键（的名字）\n return keyobj;\n }\n}\n\n/* Delete a key, value, and associated expiration entry if any, from the DB \n *\n * 从数据库中删除给定的键，键的值，以及键的过期时间。\n *\n * 删除成功返回 1 ，因为键不存在而导致删除失败时，返回 0 。\n */\nint dbDelete(redisDb *db, robj *key) {\n\n /* Deleting an entry from the expires dict will not free the sds of\n * the key, because it is shared with the main dictionary. */\n // 删除键的过期时间\n if (dictSize(db->expires) > 0) dictDelete(db->expires,key->ptr);\n\n // 删除键值对\n if (dictDelete(db->dict,key->ptr) == DICT_OK) {\n // 如果开启了集群模式，那么从槽中删除给定的键\n if (server.cluster_enabled) slotToKeyDel(key);\n return 1;\n } else {\n // 键不存在\n return 0;\n }\n}\n\n/* Prepare the string object stored at 'key' to be modified destructively\n * to implement commands like SETBIT or APPEND.\n *\n * An object is usually ready to be modified unless one of the two conditions\n * are true:\n *\n * 1) The object 'o' is shared (refcount > 1), we don't want to affect\n * other users.\n * 2) The object encoding is not \"RAW\".\n *\n * If the object is found in one of the above conditions (or both) by the\n * function, an unshared / not-encoded copy of the string object is stored\n * at 'key' in the specified 'db'. Otherwise the object 'o' itself is\n * returned.\n *\n * USAGE:\n *\n * The object 'o' is what the caller already obtained by looking up 'key'\n * in 'db', the usage pattern looks like this:\n *\n * o = lookupKeyWrite(db,key);\n * if (checkType(c,o,REDIS_STRING)) return;\n * o = dbUnshareStringValue(db,key,o);\n *\n * At this point the caller is ready to modify the object, for example\n * using an sdscat() call to append some data, or anything else.\n */\nrobj *dbUnshareStringValue(redisDb *db, robj *key, robj *o) {\n redisAssert(o->type == REDIS_STRING);\n if (o->refcount != 1 || o->encoding != REDIS_ENCODING_RAW) {\n robj *decoded = getDecodedObject(o);\n o = createRawStringObject(decoded->ptr, sdslen(decoded->ptr));\n decrRefCount(decoded);\n dbOverwrite(db,key,o);\n }\n return o;\n}\n\n/*\n * 清空服务器的所有数据。\n */\nlong long emptyDb(void(callback)(void*)) {\n int j;\n long long removed = 0;\n\n // 清空所有数据库\n for (j = 0; j < server.dbnum; j++) {\n\n // 记录被删除键的数量\n removed += dictSize(server.db[j].dict);\n\n // 删除所有键值对\n dictEmpty(server.db[j].dict,callback);\n // 删除所有键的过期时间\n dictEmpty(server.db[j].expires,callback);\n }\n\n // 如果开启了集群模式，那么还要移除槽记录\n if (server.cluster_enabled) slotToKeyFlush();\n\n // 返回键的数量\n return removed;\n}\n\n/*\n * 将客户端的目标数据库切换为 id 所指定的数据库\n */\nint selectDb(redisClient *c, int id) {\n\n // 确保 id 在正确范围内\n if (id < 0 || id >= server.dbnum)\n return REDIS_ERR;\n\n // 切换数据库（更新指针）\n c->db = &server.db[id];\n\n return REDIS_OK;\n}\n\n/*-----------------------------------------------------------------------------\n * Hooks for key space changes.\n *\n * 键空间改动的钩子。\n *\n * Every time a key in the database is modified the function\n * signalModifiedKey() is called.\n *\n * 每当数据库中的键被改动时， signalModifiedKey() 函数都会被调用。\n *\n * Every time a DB is flushed the function signalFlushDb() is called.\n *\n * 每当一个数据库被清空时， signalFlushDb() 都会被调用。\n *----------------------------------------------------------------------------*/\n\nvoid signalModifiedKey(redisDb *db, robj *key) {\n touchWatchedKey(db,key);\n}\n\nvoid signalFlushedDb(int dbid) {\n touchWatchedKeysOnFlush(dbid);\n}\n\n/*-----------------------------------------------------------------------------\n * Type agnostic commands operating on the key space\n *\n * 与类型无关的数据库操作。\n *----------------------------------------------------------------------------*/\n\n/*\n * 清空客户端指定的数据库\n */\nvoid flushdbCommand(redisClient *c) {\n\n server.dirty += dictSize(c->db->dict);\n\n // 发送通知\n signalFlushedDb(c->db->id);\n\n // 清空指定数据库中的 dict 和 expires 字典\n dictEmpty(c->db->dict,NULL);\n dictEmpty(c->db->expires,NULL);\n\n // 如果开启了集群模式，那么还要移除槽记录\n if (server.cluster_enabled) slotToKeyFlush();\n\n addReply(c,shared.ok);\n}\n\n/*\n * 清空服务器中的所有数据库\n */\nvoid flushallCommand(redisClient *c) {\n\n // 发送通知\n signalFlushedDb(-1);\n\n // 清空所有数据库\n server.dirty += emptyDb(NULL);\n addReply(c,shared.ok);\n\n // 如果正在保存新的 RDB ，那么取消保存操作\n if (server.rdb_child_pid != -1) {\n kill(server.rdb_child_pid,SIGUSR1);\n rdbRemoveTempFile(server.rdb_child_pid);\n }\n\n // 更新 RDB 文件\n if (server.saveparamslen > 0) {\n /* Normally rdbSave() will reset dirty, but we don't want this here\n * as otherwise FLUSHALL will not be replicated nor put into the AOF. */\n // rdbSave() 会清空服务器的 dirty 属性\n // 但为了确保 FLUSHALL 命令会被正常传播，\n // 程序需要保存并在 rdbSave() 调用之后还原服务器的 dirty 属性\n int saved_dirty = server.dirty;\n\n rdbSave(server.rdb_filename);\n\n server.dirty = saved_dirty;\n }\n\n server.dirty++;\n}\n\nvoid delCommand(redisClient *c) {\n int deleted = 0, j;\n\n // 遍历所有输入键\n for (j = 1; j < c->argc; j++) {\n\n // 先删除过期的键\n expireIfNeeded(c->db,c->argv[j]);\n\n // 尝试删除键\n if (dbDelete(c->db,c->argv[j])) {\n\n // 删除键成功，发送通知\n\n signalModifiedKey(c->db,c->argv[j]);\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\n \"del\",c->argv[j],c->db->id);\n\n server.dirty++;\n\n // 成功删除才增加 deleted 计数器的值\n deleted++;\n }\n }\n\n // 返回被删除键的数量\n addReplyLongLong(c,deleted);\n}\n\nvoid existsCommand(redisClient *c) {\n\n // 检查键是否已经过期，如果已过期的话，那么将它删除\n // 这可以避免已过期的键被误认为存在\n expireIfNeeded(c->db,c->argv[1]);\n\n // 在数据库中查找\n if (dbExists(c->db,c->argv[1])) {\n addReply(c, shared.cone);\n } else {\n addReply(c, shared.czero);\n }\n}\n\nvoid selectCommand(redisClient *c) {\n long id;\n\n // 不合法的数据库号码\n if (getLongFromObjectOrReply(c, c->argv[1], &id,\n \"invalid DB index\") != REDIS_OK)\n return;\n\n if (server.cluster_enabled && id != 0) {\n addReplyError(c,\"SELECT is not allowed in cluster mode\");\n return;\n }\n\n // 切换数据库\n if (selectDb(c,id) == REDIS_ERR) {\n addReplyError(c,\"invalid DB index\");\n } else {\n addReply(c,shared.ok);\n }\n}\n\nvoid randomkeyCommand(redisClient *c) {\n robj *key;\n\n // 随机返回键\n if ((key = dbRandomKey(c->db)) == NULL) {\n addReply(c,shared.nullbulk);\n return;\n }\n\n addReplyBulk(c,key);\n decrRefCount(key);\n}\n\nvoid keysCommand(redisClient *c) {\n dictIterator *di;\n dictEntry *de;\n\n // 模式\n sds pattern = c->argv[1]->ptr;\n\n int plen = sdslen(pattern), allkeys;\n unsigned long numkeys = 0;\n void *replylen = addDeferredMultiBulkLength(c);\n\n // 遍历整个数据库，返回（名字）和模式匹配的键\n di = dictGetSafeIterator(c->db->dict);\n allkeys = (pattern[0] == '*' && pattern[1] == '\\0');\n while((de = dictNext(di)) != NULL) {\n sds key = dictGetKey(de);\n robj *keyobj;\n\n // 将键名和模式进行比对\n if (allkeys || stringmatchlen(pattern,plen,key,sdslen(key),0)) {\n\n // 创建一个保存键名字的字符串对象\n keyobj = createStringObject(key,sdslen(key));\n\n // 删除已过期键\n if (expireIfNeeded(c->db,keyobj) == 0) {\n addReplyBulk(c,keyobj);\n numkeys++;\n }\n\n decrRefCount(keyobj);\n }\n }\n dictReleaseIterator(di);\n\n setDeferredMultiBulkLength(c,replylen,numkeys);\n}\n\n/* This callback is used by scanGenericCommand in order to collect elements\n * returned by the dictionary iterator into a list. */\nvoid scanCallback(void *privdata, const dictEntry *de) {\n void **pd = (void**) privdata;\n list *keys = pd[0];\n robj *o = pd[1];\n robj *key, *val = NULL;\n\n if (o == NULL) {\n sds sdskey = dictGetKey(de);\n key = createStringObject(sdskey, sdslen(sdskey));\n } else if (o->type == REDIS_SET) {\n key = dictGetKey(de);\n incrRefCount(key);\n } else if (o->type == REDIS_HASH) {\n key = dictGetKey(de);\n incrRefCount(key);\n val = dictGetVal(de);\n incrRefCount(val);\n } else if (o->type == REDIS_ZSET) {\n key = dictGetKey(de);\n incrRefCount(key);\n val = createStringObjectFromLongDouble(*(double*)dictGetVal(de));\n } else {\n redisPanic(\"Type not handled in SCAN callback.\");\n }\n\n listAddNodeTail(keys, key);\n if (val) listAddNodeTail(keys, val);\n}\n\n/* Try to parse a SCAN cursor stored at object 'o':\n * if the cursor is valid, store it as unsigned integer into *cursor and\n * returns REDIS_OK. Otherwise return REDIS_ERR and send an error to the\n * client. */\nint parseScanCursorOrReply(redisClient *c, robj *o, unsigned long *cursor) {\n char *eptr;\n\n /* Use strtoul() because we need an *unsigned* long, so\n * getLongLongFromObject() does not cover the whole cursor space. */\n errno = 0;\n *cursor = strtoul(o->ptr, &eptr, 10);\n if (isspace(((char*)o->ptr)[0]) || eptr[0] != '\\0' || errno == ERANGE)\n {\n addReplyError(c, \"invalid cursor\");\n return REDIS_ERR;\n }\n return REDIS_OK;\n}\n\n/* This command implements SCAN, HSCAN and SSCAN commands.\n *\n * 这是 SCAN 、 HSCAN 、 SSCAN 命令的实现函数。\n *\n * If object 'o' is passed, then it must be a Hash or Set object, otherwise\n * if 'o' is NULL the command will operate on the dictionary associated with\n * the current database.\n *\n * 如果给定了对象 o ，那么它必须是一个哈希对象或者集合对象，\n * 如果 o 为 NULL 的话，函数将使用当前数据库作为迭代对象。\n *\n * When 'o' is not NULL the function assumes that the first argument in\n * the client arguments vector is a key so it skips it before iterating\n * in order to parse options.\n *\n * 如果参数 o 不为 NULL ，那么说明它是一个键对象，函数将跳过这些键对象，\n * 对给定的命令选项进行分析（parse）。\n *\n * In the case of a Hash object the function returns both the field and value\n * of every element on the Hash. \n *\n * 如果被迭代的是哈希对象，那么函数返回的是键值对。\n */\nvoid scanGenericCommand(redisClient *c, robj *o, unsigned long cursor) {\n int rv;\n int i, j;\n char buf[REDIS_LONGSTR_SIZE];\n list *keys = listCreate();\n listNode *node, *nextnode;\n long count = 10;\n sds pat;\n int patlen, use_pattern = 0;\n dict *ht;\n\n /* Object must be NULL (to iterate keys names), or the type of the object\n * must be Set, Sorted Set, or Hash. */\n // 输入类型检查\n redisAssert(o == NULL || o->type == REDIS_SET || o->type == REDIS_HASH ||\n o->type == REDIS_ZSET);\n\n /* Set i to the first option argument. The previous one is the cursor. */\n // 设置第一个选项参数的索引位置\n // 0 1 2 3 \n // SCAN OPTION SCAN 命令的选项值从索引 2 开始\n // HSCAN OPTION 而其他 *SCAN 命令的选项值从索引 3 开始\n i = (o == NULL) ? 2 : 3; /* Skip the key argument if needed. */\n\n /* Step 1: Parse options. */\n // 分析选项参数\n while (i < c->argc) {\n j = c->argc - i;\n\n // COUNT \n if (!strcasecmp(c->argv[i]->ptr, \"count\") && j >= 2) {\n if (getLongFromObjectOrReply(c, c->argv[i+1], &count, NULL)\n != REDIS_OK)\n {\n goto cleanup;\n }\n\n if (count < 1) {\n addReply(c,shared.syntaxerr);\n goto cleanup;\n }\n\n i += 2;\n\n // MATCH \n } else if (!strcasecmp(c->argv[i]->ptr, \"match\") && j >= 2) {\n pat = c->argv[i+1]->ptr;\n patlen = sdslen(pat);\n\n /* The pattern always matches if it is exactly \"*\", so it is\n * equivalent to disabling it. */\n use_pattern = !(pat[0] == '*' && patlen == 1);\n\n i += 2;\n\n // error\n } else {\n addReply(c,shared.syntaxerr);\n goto cleanup;\n }\n }\n\n /* Step 2: Iterate the collection.\n *\n * Note that if the object is encoded with a ziplist, intset, or any other\n * representation that is not a hash table, we are sure that it is also\n * composed of a small number of elements. So to avoid taking state we\n * just return everything inside the object in a single call, setting the\n * cursor to zero to signal the end of the iteration. */\n // 如果对象的底层实现为 ziplist 、intset 而不是哈希表，\n // 那么这些对象应该只包含了少量元素，\n // 为了保持不让服务器记录迭代状态的设计\n // 我们将 ziplist 或者 intset 里面的所有元素都一次返回给调用者\n // 并向调用者返回游标（cursor） 0\n\n /* Handle the case of a hash table. */\n ht = NULL;\n if (o == NULL) {\n // 迭代目标为数据库\n ht = c->db->dict;\n } else if (o->type == REDIS_SET && o->encoding == REDIS_ENCODING_HT) {\n // 迭代目标为 HT 编码的集合\n ht = o->ptr;\n } else if (o->type == REDIS_HASH && o->encoding == REDIS_ENCODING_HT) {\n // 迭代目标为 HT 编码的哈希\n ht = o->ptr;\n count *= 2; /* We return key / value for this type. */\n } else if (o->type == REDIS_ZSET && o->encoding == REDIS_ENCODING_SKIPLIST) {\n // 迭代目标为 HT 编码的跳跃表\n zset *zs = o->ptr;\n ht = zs->dict;\n count *= 2; /* We return key / value for this type. */\n }\n\n if (ht) {\n void *privdata[2];\n\n /* We pass two pointers to the callback: the list to which it will\n * add new elements, and the object containing the dictionary so that\n * it is possible to fetch more data in a type-dependent way. */\n // 我们向回调函数传入两个指针：\n // 一个是用于记录被迭代元素的列表\n // 另一个是字典对象\n // 从而实现类型无关的数据提取操作\n privdata[0] = keys;\n privdata[1] = o;\n do {\n cursor = dictScan(ht, cursor, scanCallback, privdata);\n } while (cursor && listLength(keys) < count);\n } else if (o->type == REDIS_SET) {\n int pos = 0;\n int64_t ll;\n\n while(intsetGet(o->ptr,pos++,&ll))\n listAddNodeTail(keys,createStringObjectFromLongLong(ll));\n cursor = 0;\n } else if (o->type == REDIS_HASH || o->type == REDIS_ZSET) {\n unsigned char *p = ziplistIndex(o->ptr,0);\n unsigned char *vstr;\n unsigned int vlen;\n long long vll;\n\n while(p) {\n ziplistGet(p,&vstr,&vlen,&vll);\n listAddNodeTail(keys,\n (vstr != NULL) ? createStringObject((char*)vstr,vlen) :\n createStringObjectFromLongLong(vll));\n p = ziplistNext(o->ptr,p);\n }\n cursor = 0;\n } else {\n redisPanic(\"Not handled encoding in SCAN.\");\n }\n\n /* Step 3: Filter elements. */\n node = listFirst(keys);\n while (node) {\n robj *kobj = listNodeValue(node);\n nextnode = listNextNode(node);\n int filter = 0;\n\n /* Filter element if it does not match the pattern. */\n if (!filter && use_pattern) {\n if (sdsEncodedObject(kobj)) {\n if (!stringmatchlen(pat, patlen, kobj->ptr, sdslen(kobj->ptr), 0))\n filter = 1;\n } else {\n char buf[REDIS_LONGSTR_SIZE];\n int len;\n\n redisAssert(kobj->encoding == REDIS_ENCODING_INT);\n len = ll2string(buf,sizeof(buf),(long)kobj->ptr);\n if (!stringmatchlen(pat, patlen, buf, len, 0)) filter = 1;\n }\n }\n\n /* Filter element if it is an expired key. */\n if (!filter && o == NULL && expireIfNeeded(c->db, kobj)) filter = 1;\n\n /* Remove the element and its associted value if needed. */\n if (filter) {\n decrRefCount(kobj);\n listDelNode(keys, node);\n }\n\n /* If this is a hash or a sorted set, we have a flat list of\n * key-value elements, so if this element was filtered, remove the\n * value, or skip it if it was not filtered: we only match keys. */\n if (o && (o->type == REDIS_ZSET || o->type == REDIS_HASH)) {\n node = nextnode;\n nextnode = listNextNode(node);\n if (filter) {\n kobj = listNodeValue(node);\n decrRefCount(kobj);\n listDelNode(keys, node);\n }\n }\n node = nextnode;\n }\n\n /* Step 4: Reply to the client. */\n addReplyMultiBulkLen(c, 2);\n rv = snprintf(buf, sizeof(buf), \"%lu\", cursor);\n redisAssert(rv < sizeof(buf));\n addReplyBulkCBuffer(c, buf, rv);\n\n addReplyMultiBulkLen(c, listLength(keys));\n while ((node = listFirst(keys)) != NULL) {\n robj *kobj = listNodeValue(node);\n addReplyBulk(c, kobj);\n decrRefCount(kobj);\n listDelNode(keys, node);\n }\n\ncleanup:\n listSetFreeMethod(keys,decrRefCountVoid);\n listRelease(keys);\n}\n\n/* The SCAN command completely relies on scanGenericCommand. */\nvoid scanCommand(redisClient *c) {\n unsigned long cursor;\n if (parseScanCursorOrReply(c,c->argv[1],&cursor) == REDIS_ERR) return;\n scanGenericCommand(c,NULL,cursor);\n}\n\nvoid dbsizeCommand(redisClient *c) {\n addReplyLongLong(c,dictSize(c->db->dict));\n}\n\nvoid lastsaveCommand(redisClient *c) {\n addReplyLongLong(c,server.lastsave);\n}\n\nvoid typeCommand(redisClient *c) {\n robj *o;\n char *type;\n\n o = lookupKeyRead(c->db,c->argv[1]);\n\n if (o == NULL) {\n type = \"none\";\n } else {\n switch(o->type) {\n case REDIS_STRING: type = \"string\"; break;\n case REDIS_LIST: type = \"list\"; break;\n case REDIS_SET: type = \"set\"; break;\n case REDIS_ZSET: type = \"zset\"; break;\n case REDIS_HASH: type = \"hash\"; break;\n default: type = \"unknown\"; break;\n }\n }\n\n addReplyStatus(c,type);\n}\n\nvoid shutdownCommand(redisClient *c) {\n int flags = 0;\n\n if (c->argc > 2) {\n addReply(c,shared.syntaxerr);\n return;\n } else if (c->argc == 2) {\n\n // 停机时不进行保存\n if (!strcasecmp(c->argv[1]->ptr,\"nosave\")) {\n flags |= REDIS_SHUTDOWN_NOSAVE;\n\n // 停机时进行保存\n } else if (!strcasecmp(c->argv[1]->ptr,\"save\")) {\n flags |= REDIS_SHUTDOWN_SAVE;\n\n } else {\n addReply(c,shared.syntaxerr);\n return;\n }\n }\n\n /* When SHUTDOWN is called while the server is loading a dataset in\n * memory we need to make sure no attempt is performed to save\n * the dataset on shutdown (otherwise it could overwrite the current DB\n * with half-read data).\n *\n * Also when in Sentinel mode clear the SAVE flag and force NOSAVE. */\n if (server.loading || server.sentinel_mode)\n flags = (flags & ~REDIS_SHUTDOWN_SAVE) | REDIS_SHUTDOWN_NOSAVE;\n\n if (prepareForShutdown(flags) == REDIS_OK) exit(0);\n\n addReplyError(c,\"Errors trying to SHUTDOWN. Check logs.\");\n}\n\nvoid renameGenericCommand(redisClient *c, int nx) {\n robj *o;\n long long expire;\n\n /* To use the same key as src and dst is probably an error */\n // 来源键和目标键不能相同\n if (sdscmp(c->argv[1]->ptr,c->argv[2]->ptr) == 0) {\n addReply(c,shared.sameobjecterr);\n return;\n }\n\n // 取出来源键\n if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr)) == NULL)\n return;\n\n // 增加引用计数，因为后面目标键也会引用这个对象\n // 如果不增加的话，当来源键被删除时，这个值对象也会被删除\n incrRefCount(o);\n\n // 取出来源键的过期时间\n expire = getExpire(c->db,c->argv[1]);\n\n // 检查目标键是否存在\n if (lookupKeyWrite(c->db,c->argv[2]) != NULL) {\n\n // 如果目标键存在，并且执行的是 RENAMENX ，那么直接返回\n if (nx) {\n decrRefCount(o);\n addReply(c,shared.czero);\n return;\n }\n\n // 如果执行的是 RENAME ，那么删除已有的目标键\n /* Overwrite: delete the old key before creating the new one\n * with the same name. */\n dbDelete(c->db,c->argv[2]);\n }\n\n // 将来源键的值对象和目标键进行关联\n dbAdd(c->db,c->argv[2],o);\n\n // 如果有过期时间，那么为目标键设置过期时间\n if (expire != -1) setExpire(c->db,c->argv[2],expire);\n\n // 删除来源键\n dbDelete(c->db,c->argv[1]);\n\n signalModifiedKey(c->db,c->argv[1]);\n signalModifiedKey(c->db,c->argv[2]);\n\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\"rename_from\",\n c->argv[1],c->db->id);\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\"rename_to\",\n c->argv[2],c->db->id);\n\n server.dirty++;\n\n addReply(c,nx ? shared.cone : shared.ok);\n}\n\nvoid renameCommand(redisClient *c) {\n renameGenericCommand(c,0);\n}\n\nvoid renamenxCommand(redisClient *c) {\n renameGenericCommand(c,1);\n}\n\nvoid moveCommand(redisClient *c) {\n robj *o;\n redisDb *src, *dst;\n int srcid;\n\n if (server.cluster_enabled) {\n addReplyError(c,\"MOVE is not allowed in cluster mode\");\n return;\n }\n\n /* Obtain source and target DB pointers */\n // 源数据库\n src = c->db;\n // 源数据库的 id\n srcid = c->db->id;\n\n // 切换到目标数据库\n if (selectDb(c,atoi(c->argv[2]->ptr)) == REDIS_ERR) {\n addReply(c,shared.outofrangeerr);\n return;\n }\n\n // 目标数据库\n dst = c->db;\n\n // 切换回源数据库\n selectDb(c,srcid); /* Back to the source DB */\n\n /* If the user is moving using as target the same\n * DB as the source DB it is probably an error. */\n // 如果源数据库和目标数据库相等，那么返回错误\n if (src == dst) {\n addReply(c,shared.sameobjecterr);\n return;\n }\n\n /* Check if the element exists and get a reference */\n // 取出要移动的对象\n o = lookupKeyWrite(c->db,c->argv[1]);\n if (!o) {\n addReply(c,shared.czero);\n return;\n }\n\n /* Return zero if the key already exists in the target DB */\n // 如果键已经存在于目标数据库，那么返回\n if (lookupKeyWrite(dst,c->argv[1]) != NULL) {\n addReply(c,shared.czero);\n return;\n }\n\n // 将键添加到目标数据库中\n dbAdd(dst,c->argv[1],o);\n // 增加对对象的引用计数，避免接下来在源数据库中删除时 o 被清理\n incrRefCount(o);\n\n /* OK! key moved, free the entry in the source DB */\n // 将键从源数据库中返回\n dbDelete(src,c->argv[1]);\n\n server.dirty++;\n\n addReply(c,shared.cone);\n}\n\n/*-----------------------------------------------------------------------------\n * Expires API\n *----------------------------------------------------------------------------*/\n\n/*\n * 移除键 key 的过期时间\n */\nint removeExpire(redisDb *db, robj *key) {\n /* An expire may only be removed if there is a corresponding entry in the\n * main dict. Otherwise, the key will never be freed. */\n // 确保键带有过期时间\n redisAssertWithInfo(NULL,key,dictFind(db->dict,key->ptr) != NULL);\n\n // 删除过期时间\n return dictDelete(db->expires,key->ptr) == DICT_OK;\n}\n\n/*\n * 将键 key 的过期时间设为 when\n */\nvoid setExpire(redisDb *db, robj *key, long long when) {\n\n dictEntry *kde, *de;\n\n /* Reuse the sds from the main dict in the expire dict */\n // 取出键\n kde = dictFind(db->dict,key->ptr);\n\n redisAssertWithInfo(NULL,key,kde != NULL);\n\n // 根据键取出键的过期时间\n de = dictReplaceRaw(db->expires,dictGetKey(kde));\n\n // 设置键的过期时间\n // 这里是直接使用整数值来保存过期时间，不是用 INT 编码的 String 对象\n dictSetSignedIntegerVal(de,when);\n}\n\n/* Return the expire time of the specified key, or -1 if no expire\n * is associated with this key (i.e. the key is non volatile) \n *\n * 返回给定 key 的过期时间。\n *\n * 如果键没有设置过期时间，那么返回 -1 。\n */\nlong long getExpire(redisDb *db, robj *key) {\n dictEntry *de;\n\n /* No expire? return ASAP */\n // 获取键的过期时间\n // 如果过期时间不存在，那么直接返回\n if (dictSize(db->expires) == 0 ||\n (de = dictFind(db->expires,key->ptr)) == NULL) return -1;\n\n /* The entry was found in the expire dict, this means it should also\n * be present in the main dict (safety check). */\n redisAssertWithInfo(NULL,key,dictFind(db->dict,key->ptr) != NULL);\n\n // 返回过期时间\n return dictGetSignedIntegerVal(de);\n}\n\n/* Propagate expires into slaves and the AOF file.\n * When a key expires in the master, a DEL operation for this key is sent\n * to all the slaves and the AOF file if enabled.\n *\n * 将过期时间传播到附属节点和 AOF 文件。\n *\n * 当一个键在主节点中过期时，\n * 主节点会向所有附属节点和 AOF 文件传播一个显式的 DEL 命令。\n *\n * This way the key expiry is centralized in one place, and since both\n * AOF and the master->slave link guarantee operation ordering, everything\n * will be consistent even if we allow write operations against expiring\n * keys. \n *\n * 这种做法使得对键的过期可以集中在一处处理，\n * 因为 AOF 以及主节点和附属节点之间的链接，都可以保证操作的执行顺序，\n * 所以即使有写操作对过期键执行，所有数据都还是 consistent 的。\n */\nvoid propagateExpire(redisDb *db, robj *key) {\n robj *argv[2];\n\n // 构造一个 DEL key 命令\n argv[0] = shared.del;\n argv[1] = key;\n incrRefCount(argv[0]);\n incrRefCount(argv[1]);\n\n // 传播到 AOF \n if (server.aof_state != REDIS_AOF_OFF)\n feedAppendOnlyFile(server.delCommand,db->id,argv,2);\n\n // 传播到所有附属节点\n replicationFeedSlaves(server.slaves,db->id,argv,2);\n\n decrRefCount(argv[0]);\n decrRefCount(argv[1]);\n}\n\n/*\n * 检查 key 是否已经过期，如果是的话，将它从数据库中删除。\n *\n * 返回 0 表示键没有过期时间，或者键未过期。\n *\n * 返回 1 表示键已经因为过期而被删除了。\n */\nint expireIfNeeded(redisDb *db, robj *key) {\n\n // 取出键的过期时间\n mstime_t when = getExpire(db,key);\n mstime_t now;\n\n // 没有过期时间\n if (when < 0) return 0; /* No expire for this key */\n\n /* Don't expire anything while loading. It will be done later. */\n // 如果服务器正在进行载入，那么不进行任何过期检查\n if (server.loading) return 0;\n\n /* If we are in the context of a Lua script, we claim that time is\n * blocked to when the Lua script started. This way a key can expire\n * only the first time it is accessed and not in the middle of the\n * script execution, making propagation to slaves / AOF consistent.\n * See issue #1525 on Github for more information. */\n now = server.lua_caller ? server.lua_time_start : mstime();\n\n /* If we are running in the context of a slave, return ASAP:\n * the slave key expiration is controlled by the master that will\n * send us synthesized DEL operations for expired keys.\n *\n * Still we try to return the right information to the caller, \n * that is, 0 if we think the key should be still valid, 1 if\n * we think the key is expired at this time. */\n // 当服务器运行在 replication 模式时\n // 附属节点并不主动删除 key\n // 它只返回一个逻辑上正确的返回值\n // 真正的删除操作要等待主节点发来删除命令时才执行\n // 从而保证数据的同步\n if (server.masterhost != NULL) return now > when;\n\n // 运行到这里，表示键带有过期时间，并且服务器为主节点\n\n /* Return when this key has not expired */\n // 如果未过期，返回 0\n if (now <= when) return 0;\n\n /* Delete the key */\n server.stat_expiredkeys++;\n\n // 向 AOF 文件和附属节点传播过期信息\n propagateExpire(db,key);\n\n // 发送事件通知\n notifyKeyspaceEvent(REDIS_NOTIFY_EXPIRED,\n \"expired\",key,db->id);\n\n // 将过期键从数据库中删除\n return dbDelete(db,key);\n}\n\n/*-----------------------------------------------------------------------------\n * Expires Commands\n *----------------------------------------------------------------------------*/\n\n/* This is the generic command implementation for EXPIRE, PEXPIRE, EXPIREAT\n * and PEXPIREAT. Because the commad second argument may be relative or absolute\n * the \"basetime\" argument is used to signal what the base time is (either 0\n * for *AT variants of the command, or the current time for relative expires).\n *\n * 这个函数是 EXPIRE 、 PEXPIRE 、 EXPIREAT 和 PEXPIREAT 命令的底层实现函数。\n *\n * 命令的第二个参数可能是绝对值，也可能是相对值。\n * 当执行 *AT 命令时， basetime 为 0 ，在其他情况下，它保存的就是当前的绝对时间。\n *\n * unit is either UNIT_SECONDS or UNIT_MILLISECONDS, and is only used for\n * the argv[2] parameter. The basetime is always specified in milliseconds. \n *\n * unit 用于指定 argv[2] （传入过期时间）的格式，\n * 它可以是 UNIT_SECONDS 或 UNIT_MILLISECONDS ，\n * basetime 参数则总是毫秒格式的。\n */\nvoid expireGenericCommand(redisClient *c, long long basetime, int unit) {\n robj *key = c->argv[1], *param = c->argv[2];\n long long when; /* unix time in milliseconds when the key will expire. */\n\n // 取出 when 参数\n if (getLongLongFromObjectOrReply(c, param, &when, NULL) != REDIS_OK)\n return;\n\n // 如果传入的过期时间是以秒为单位的，那么将它转换为毫秒\n if (unit == UNIT_SECONDS) when *= 1000;\n when += basetime;\n\n /* No key, return zero. */\n // 取出键\n if (lookupKeyRead(c->db,key) == NULL) {\n addReply(c,shared.czero);\n return;\n }\n\n /* EXPIRE with negative TTL, or EXPIREAT with a timestamp into the past\n * should never be executed as a DEL when load the AOF or in the context\n * of a slave instance.\n *\n * 在载入数据时，或者服务器为附属节点时，\n * 即使 EXPIRE 的 TTL 为负数，或者 EXPIREAT 提供的时间戳已经过期，\n * 服务器也不会主动删除这个键，而是等待主节点发来显式的 DEL 命令。\n *\n * Instead we take the other branch of the IF statement setting an expire\n * (possibly in the past) and wait for an explicit DEL from the master. \n *\n * 程序会继续将（一个可能已经过期的 TTL）设置为键的过期时间，\n * 并且等待主节点发来 DEL 命令。\n */\n if (when <= mstime() && !server.loading && !server.masterhost) {\n\n // when 提供的时间已经过期，服务器为主节点，并且没在载入数据\n\n robj *aux;\n\n redisAssertWithInfo(c,key,dbDelete(c->db,key));\n server.dirty++;\n\n /* Replicate/AOF this as an explicit DEL. */\n // 传播 DEL 命令\n aux = createStringObject(\"DEL\",3);\n\n rewriteClientCommandVector(c,2,aux,key);\n decrRefCount(aux);\n\n signalModifiedKey(c->db,key);\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\"del\",key,c->db->id);\n\n addReply(c, shared.cone);\n\n return;\n } else {\n\n // 设置键的过期时间\n // 如果服务器为附属节点，或者服务器正在载入，\n // 那么这个 when 有可能已经过期的\n setExpire(c->db,key,when);\n\n addReply(c,shared.cone);\n\n signalModifiedKey(c->db,key);\n notifyKeyspaceEvent(REDIS_NOTIFY_GENERIC,\"expire\",key,c->db->id);\n\n server.dirty++;\n\n return;\n }\n}\n\nvoid expireCommand(redisClient *c) {\n expireGenericCommand(c,mstime(),UNIT_SECONDS);\n}\n\nvoid expireatCommand(redisClient *c) {\n expireGenericCommand(c,0,UNIT_SECONDS);\n}\n\nvoid pexpireCommand(redisClient *c) {\n expireGenericCommand(c,mstime(),UNIT_MILLISECONDS);\n}\n\nvoid pexpireatCommand(redisClient *c) {\n expireGenericCommand(c,0,UNIT_MILLISECONDS);\n}\n\n/*\n * 返回键的剩余生存时间。\n *\n * output_ms 指定返回值的格式：\n *\n * - 为 1 时，返回毫秒\n *\n * - 为 0 时，返回秒\n */\nvoid ttlGenericCommand(redisClient *c, int output_ms) {\n long long expire, ttl = -1;\n\n /* If the key does not exist at all, return -2 */\n // 取出键\n if (lookupKeyRead(c->db,c->argv[1]) == NULL) {\n addReplyLongLong(c,-2);\n return;\n }\n\n /* The key exists. Return -1 if it has no expire, or the actual\n * TTL value otherwise. */\n // 取出过期时间\n expire = getExpire(c->db,c->argv[1]);\n\n if (expire != -1) {\n // 计算剩余生存时间\n ttl = expire-mstime();\n if (ttl < 0) ttl = 0;\n }\n\n if (ttl == -1) {\n // 键是持久的\n addReplyLongLong(c,-1);\n } else {\n // 返回 TTL \n // (ttl+500)/1000 计算的是渐近秒数\n addReplyLongLong(c,output_ms ? ttl : ((ttl+500)/1000));\n }\n}\n\nvoid ttlCommand(redisClient *c) {\n ttlGenericCommand(c, 0);\n}\n\nvoid pttlCommand(redisClient *c) {\n ttlGenericCommand(c, 1);\n}\n\nvoid persistCommand(redisClient *c) {\n dictEntry *de;\n\n // 取出键\n de = dictFind(c->db->dict,c->argv[1]->ptr);\n\n if (de == NULL) {\n // 键没有过期时间\n addReply(c,shared.czero);\n\n } else {\n\n // 键带有过期时间，那么将它移除\n if (removeExpire(c->db,c->argv[1])) {\n addReply(c,shared.cone);\n server.dirty++;\n\n // 键已经是持久的了\n } else {\n addReply(c,shared.czero);\n }\n }\n}\n\n/* -----------------------------------------------------------------------------\n * API to get key arguments from commands\n * ---------------------------------------------------------------------------*/\n\n/* The base case is to use the keys position as given in the command table\n * (firstkey, lastkey, step). */\nint *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, int *numkeys) {\n int j, i = 0, last, *keys;\n REDIS_NOTUSED(argv);\n\n if (cmd->firstkey == 0) {\n *numkeys = 0;\n return NULL;\n }\n last = cmd->lastkey;\n if (last < 0) last = argc+last;\n keys = zmalloc(sizeof(int)*((last - cmd->firstkey)+1));\n for (j = cmd->firstkey; j <= last; j += cmd->keystep) {\n redisAssert(j < argc);\n keys[i++] = j;\n }\n *numkeys = i;\n return keys;\n}\n\n/* Return all the arguments that are keys in the command passed via argc / argv.\n *\n * The command returns the positions of all the key arguments inside the array,\n * so the actual return value is an heap allocated array of integers. The\n * length of the array is returned by reference into *numkeys.\n *\n * 'cmd' must be point to the corresponding entry into the redisCommand\n * table, according to the command name in argv[0].\n *\n * This function uses the command table if a command-specific helper function\n * is not required, otherwise it calls the command-specific function. */\nint *getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {\n if (cmd->getkeys_proc) {\n return cmd->getkeys_proc(cmd,argv,argc,numkeys);\n } else {\n return getKeysUsingCommandTable(cmd,argv,argc,numkeys);\n }\n}\n\n/* Free the result of getKeysFromCommand. */\nvoid getKeysFreeResult(int *result) {\n zfree(result);\n}\n\n/* Helper function to extract keys from following commands:\n * ZUNIONSTORE ... \n * ZINTERSTORE ... */\nint *zunionInterGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {\n int i, num, *keys;\n REDIS_NOTUSED(cmd);\n\n num = atoi(argv[2]->ptr);\n /* Sanity check. Don't return any key if the command is going to\n * reply with syntax error. */\n if (num > (argc-3)) {\n *numkeys = 0;\n return NULL;\n }\n\n /* Keys in z{union,inter}store come from two places:\n * argv[1] = storage key,\n * argv[3...n] = keys to intersect */\n keys = zmalloc(sizeof(int)*(num+1));\n\n /* Add all key positions for argv[3...n] to keys[] */\n for (i = 0; i < num; i++) keys[i] = 3+i;\n\n /* Finally add the argv[1] key position (the storage key target). */\n keys[num] = 1;\n *numkeys = num+1; /* Total keys = {union,inter} keys + storage key */\n return keys;\n}\n\n/* Helper function to extract keys from the following commands:\n * EVAL