ChangeLog from http://bogomips.org/kgio.git commit cbba2690a75830d68c1db779a1c15ecad3725528 Author: Eric Wong Date: Tue Feb 4 03:06:42 2014 +0000 kgio 2.9.0 - cleanups, bug fixes, minor improvements This adds a new kgio_syssend method for Ruby 1.9+ which behaves like BasicSocket#send, but will not raise exceptions on EAGAIN. Eric Wong (12): test_poll: remove race prone test_poll_EINTR_changed test tryopen: remove RARRAY_PTR usage in init read_write: remove the rest of RARRAY_PTR usage my_writev: stylistic fixes Rakefile: kill raa_update task avoid downsizing casts connect: constify RSTRING-related things set RSTRING_MODIFIED where appropriate for Rubinius split read_write.c into {read,write,writev}.c add kgio_syssend method to wrap send(2) write: correct check for various blocking regions tryopen: additional debug information for bad Errno values Hleb Valoshka (1): Don't use deprecated api commit f991debdcc3cbba721029285e4a77f32d8222aa1 Author: Eric Wong Date: Tue Feb 4 03:01:46 2014 +0000 tryopen: additional debug information for bad Errno values In case users run into the same problem in the future. ref: http://mid.gmane.org/20131227074129.GA6381@dcvr.yhbt.net commit cf4c09c3f280850cd5a018d603e9df08ce54ade2 Author: Eric Wong Date: Tue Feb 4 02:42:51 2014 +0000 write: correct check for various blocking regions This will allow us to support past/future Rubies without rb_thread_io_blocking_region but still have other ways of releasing GVL. commit 070b5aeb9313ef09303dffa7bc72c771f6f86f31 Author: Eric Wong Date: Tue Feb 4 00:51:59 2014 +0000 add kgio_syssend method to wrap send(2) This behaves like kgio_trywrite on GNU/Linux, but allows extra flags to be specified. The main purpose of this is to support use of the MSG_MORE flag on GNU/Linux. commit a876a30e4bfae6e3aa4af2e34f68bf66be5715b4 Author: Eric Wong Date: Thu Nov 21 21:18:45 2013 +0000 split read_write.c into {read,write,writev}.c This helps make our code more manageable, as well as isolating RSTRING_MODIFIED usage to files which actually need it. writev.c can eventually be fixed to avoid modifying RSTRING pointers, but write.c can already benefit from const RSTRING_PTR values right away. commit 8788e248f9978da6708dd1e85d40531bb30d0bb2 Author: Eric Wong Date: Thu Nov 21 19:05:23 2013 +0000 set RSTRING_MODIFIED where appropriate for Rubinius Not yet tested, but it makes sense to do this. commit 7a49e4df0a8f23594a83f21f2fd54955dcb6a0fc Author: Eric Wong Date: Thu Nov 21 18:57:09 2013 +0000 connect: constify RSTRING-related things This should enforce correctness with our non-use of RSTRING_MODIFIED in this file, hopefully allowing Rubinius to optimize better. commit 809b9716a2593c0bfcaa115e7e560615c6357b55 Author: Eric Wong Date: Mon Jan 20 21:16:14 2014 +0000 avoid downsizing casts It's always safe to cast a 32-bit value to a 64-bit value, but compilers can warn going the other way (because VALUE is a pointer). Tested with `ruby -v` ruby 2.2.0dev (2014-01-19 trunk 44646) [x86_64-linux] which enables -Wpointer-to-int-cast commit 008483785d1a5bb801219af8afbb77be18b9bef1 Author: Hleb Valoshka <375gnu@gmail.com> Date: Mon Jan 20 22:36:19 2014 +0300 Don't use deprecated api Signed-off-by: Eric Wong commit a8cb03f58be2c2742f0030a28a9e8454c3d56634 Author: Eric Wong Date: Mon Sep 30 18:22:43 2013 +0000 Rakefile: kill raa_update task RAA is dead commit fc91f1d424af2a94d0fbed4b1f4d0171c8add79f Author: Eric Wong Date: Wed Sep 25 19:37:22 2013 +0000 my_writev: stylistic fixes Rename 'str' to 'ary', as that's the whole point of supporting writev... Also, long lines need to be wrapped at 80 columns commit 13725d97e0c493cc38a4c833c1053216fb5c799b Author: Eric Wong Date: Wed Sep 25 19:26:26 2013 +0000 read_write: remove the rest of RARRAY_PTR usage I have not benchmarked this, but this should not make a difference as far as performance goes. This should also allow better performance of better GCs in Ruby 2.1.0 and Rubinius. commit e0d3b3cbe90b47facf0e67036429502ad8f99c49 Author: Eric Wong Date: Wed Sep 25 19:20:11 2013 +0000 tryopen: remove RARRAY_PTR usage in init This is trivially non-performance-critical, and can only help with advanced GCs in Ruby 2.1.0 and Rubinius commit 98bbc4bd1fda4aa7fddcb57db659bb3e507238e7 Author: Eric Wong Date: Wed Sep 25 17:42:17 2013 +0000 test_poll: remove race prone test_poll_EINTR_changed test We haven't figured out a way to reliably test this w/o races, so lets just remove it for now and trust it works by reading the C code. ref: commit 9bde3ab9a7e6e1776ba43bd0e7a3e9202f1026f6 Author: Eric Wong Date: Wed Sep 11 00:01:21 2013 +0000 kgio 2.8.1 - minor improvements and test fixes Improved error reporting for kgio_accept/kgio_tryaccept. Minor size reduction throughout. There are also several test case fixes for race conditions. Thanks to Hleb Valoshka and the Debian project for all the help with this release! Eric Wong (7): check syscall returns against < 0 instead of == -1 accept: more informative exception on unknown family test_tryopen: skip EACCES test when euid == 0 test/lib_read_write: account for larger-than-normal pipes test_poll: avoid potentially thread-unsafe test test_poll: preserve original trap(:USR1) handler test_poll: be less dependent on signal handler ordering Hleb Valoshka (4): Change prefix of temporary sockets to prevent races Don't dump 20M in case of failure Create own directory for every unix socket in unit tests Close tempfile and unlink it immediately. commit c3be428c452fd9c91d31b5b7d5b697175b6652df Author: Hleb Valoshka <375gnu@gmail.com> Date: Sat Sep 7 23:07:36 2013 +0300 Close tempfile and unlink it immediately. When a Tempfile object is garbage collected, or when the Ruby interpreter exits, its associated temporary file is automatically deleted. This may lead to race condition when Tempfile is used like in these tests. Signed-off-by: Eric Wong commit ff940137dc5098c9d8094e041e9be053cb3e139f Author: Eric Wong Date: Thu Sep 5 23:11:59 2013 +0000 test_poll: be less dependent on signal handler ordering Just spam the thread several times to force the wakeup to happen and not be as dependent on signal handler execution ordering of different Ruby VMs. Tested-by: Hleb Valoshka <375gnu@gmail.com> commit 90340eb8f2c5bf820eabd3c25fb39e45285b1c40 Author: Hleb Valoshka <375gnu@gmail.com> Date: Wed Sep 4 15:07:19 2013 +0300 Create own directory for every unix socket in unit tests [ew: this avoids a TOCTOU issue for multiple test invocations] Signed-off-by: Eric Wong commit 71f656e40de99d240c873ebb23451f76e46e5a20 Author: Eric Wong Date: Tue Sep 3 22:47:20 2013 +0000 test_poll: preserve original trap(:USR1) handler The existing SIGUSR1 handler may not always be "DEFAULT", so restore the original one. commit 377028d998f32c9bf8926a275019445103328c3d Author: Eric Wong Date: Tue Sep 3 22:45:26 2013 +0000 test_poll: avoid potentially thread-unsafe test Modifying pollset in a different pollset is thread-unsafe, so just do that inside the signal handler as that should fire before restarting poll(). commit 259347346169d4a2ab5b33719fe0985edc64bd3c Author: Eric Wong Date: Tue Sep 3 22:43:04 2013 +0000 test/lib_read_write: account for larger-than-normal pipes Linux pipe buffers may use larger pages and increase the capacity of pipes. Thus the 10000 write attempt is not sufficient; just infinite loop until we hit :wait_writable. commit dae148a9d6de7501acf8a68218b3e9d2c5ff6ff7 Author: Eric Wong Date: Tue Sep 3 19:35:46 2013 +0000 test_tryopen: skip EACCES test when euid == 0 This fails when the test is run as root (which may be the case of some Ruby installations) or fakeroot (which is the case of Debian build systems). commit 5c99abe6108d8d6a52c07b969afbc556c59b33be Author: Hleb Valoshka <375GNU@Gmail.COM> Date: Sat Aug 24 17:31:42 2013 +0300 Don't dump 20M in case of failure Signed-off-by: Eric Wong commit fc88b648ca5c2ba6c845c357192d7e36f9dedc5c Author: Hleb Valoshka <375GNU@Gmail.COM> Date: Sat Aug 24 16:50:17 2013 +0300 Change prefix of temporary sockets to prevent races Note: this is an incomplete fix and the race can still happen. A proper fix would involve using a temporary directory for each test and placing the Unix socket in there. Signed-off-by: Eric Wong commit 30c12374f8d23f0e635946fc38546820216457bf Author: Eric Wong Date: Fri Aug 16 02:00:21 2013 +0000 accept: more informative exception on unknown family This may help us diagnose issues on strange OSes. commit 9f30805bc4ff65ad4ca0be9dfb26b1a9bdc70c51 Author: Eric Wong Date: Fri Aug 16 01:52:15 2013 +0000 check syscall returns against < 0 instead of == -1 This may help us avoid errors in case of C library bugs, and also results in smaller code: $ ~/linux/scripts/bloat-o-meter before.so after.so add/remove: 0/0 grow/shrink: 0/9 up/down: 0/-66 (-66) function old new delta s_trywrite 160 159 -1 kgio_write 160 159 -1 kgio_trywrite 160 159 -1 my_recv 616 610 -6 my_peek 616 610 -6 stream_connect 456 448 -8 my_socket 222 213 -9 my_writev 1703 1687 -16 write_check 427 409 -18 commit 8be51237720fd18cb45188f29c717bbac0ca1964 Author: Eric Wong Date: Fri Jan 18 10:25:20 2013 +0000 kgio 2.8.0 - TCP Fast Open, writev/trywritev TCP Fast Open in Linux 3.7 and later is now supported in the client via Kgio::Socket#kgio_fastopen. This release also adds the kgio_writev and kgio_trywritev methods, thanks to funny-falcon commit 3e555a62c75406d15199fd7bdb287704e5738352 Author: Eric Wong Date: Fri Jan 18 10:50:28 2013 +0000 rename fastopen => kgio_fastopen in Kgio::Socket In the unlikely case the Ruby Socket class implements its own "fastopen" method, we will avoid conflicting. commit c751f42f5f6a5e54a399df472015ab6d2ffc3f7a Author: Eric Wong Date: Sun Dec 30 11:43:10 2012 +0000 accept: do not set blocking if non-blocking is set This is prone to race conditions in multiprocess situations where one process is relying on non-blocking operation while another (likely newer process) relies on blocking operation. Since the blocking process can always fall back to calling rb_io_wait_readable(), use that instead and give up some scalability for higher reliability. Those interested in avoiding thundering herds will have to stop/start their processes using blocking sockets (and tolerate some downtime). commit c63ad2b2e0e25f0765605e8ba2d7038b5e28d878 Author: Eric Wong Date: Thu Dec 27 02:16:26 2012 +0000 fastopen: fix argument order in RDoc example Oops :x commit 5f696156e097a1e66cb0c5c2a7cf62b38fd97605 Author: Eric Wong Date: Thu Dec 27 01:29:01 2012 +0000 read_write: remove unused variable commit f61cef65b8a8816160c622324b4f1aad55034e4a Author: Eric Wong Date: Thu Dec 27 01:16:56 2012 +0000 implement TCP Fast Open support (client + server) Server support just requires exposing one constant for setsockopt: Kgio::TCP_FASTOPEN Client support implements a new Kgio::Socket#fastopen method. This new method wraps the the sendto() syscall. With TCP Fast Open, the sendto() syscall is overloaded for stream sockets to implement the functionality of both connect() + write() Since it only makes sense to use _blocking_ I/O for sendto(), TFO clients are only supported in Ruby implementations with native threads. commit 7a3fc55424338ad458cc719d4cb3c4e28802d5cb Author: Eric Wong Date: Thu Dec 13 00:02:50 2012 +0000 Kgio::Socket.new retains compatibility with Socket.new This allows us to create an unconnected socket, just like the normal Socket class it inherits from. commit 48fc432a3b9dfd2b0435f0975556d4a321a5239b Author: Eric Wong Date: Wed Dec 12 21:40:50 2012 +0000 connect: factor out tcp_getaddr() function This will be reused for TCP fast open support. commit 9ddd17b0e296eb279f05d418da6ad46319bcf0b5 Author: Eric Wong Date: Wed Dec 12 21:21:29 2012 +0000 connect: split out my_socket() function This makes the retry logic for mismatched libc headers/kernel versions easier to understand and follow. commit 8b4df8ece93ddc4e2fb685905461c1ed27b22295 Author: Eric Wong Date: Wed Nov 21 23:16:00 2012 +0000 tryopen: include errno.h header just in case errno.h is not guaranteed to be included in existing headers, so we need to #include it to ensure errno and friends are usable. Thanks to stuart on the kgio mailing list for noticing ref: <062571308.133355.1353536890665.JavaMail.sas1@172.29.251.236> commit f020550fc802f299fdcdec695ac80d53ef3d24d9 Author: Eric Wong Date: Mon Jul 2 04:21:40 2012 +0000 test workaround for platforms with unreliable signals Ruby may not respond well to signals on all platforms, especially not after fork()-ing in the face of a running pthread (timer thread on 1.9.2). SIGKILL bypasses Ruby (and all userspace) signal handling on Debian GNU/kFreeBSD. commit 488a148d8b172e152e3450062b172ba516ab84b3 Author: Eric Wong Date: Mon Jul 2 04:20:20 2012 +0000 test/lib_read_write: wait for readability before tryread On FreeBSD, writing to a loopback TCP socket does not guarantee immediate readability on the other end. Tested on Debian GNU/kFreeBSD 6.0 commit c79babfd175aa7b4be9d4d1a10a64c17b93730a0 Author: Eric Wong Date: Mon Jul 2 03:16:09 2012 +0000 test_poll: skip signal torture on Debian GNU/kfreebsd This cascades test failures on a platform with questionable signal/fork handling. Tested on: Debian GNU/kFreeBSD 6.0 commit ff27e74a49bf6746ffe74cfc865430221f0bafe0 Author: Sokolov Yura 'funny-falcon Date: Fri Jun 1 13:42:58 2012 +0400 add `#kgio_writev` and `#kgio_trywritev` Add methods for using writev(2) syscall for sending array of string in a single syscall. This is more efficient than concatenating strings on Ruby side or sending them one by one. `#kgio_trywritev` returns array of strings which are not sent to the socket. If there were objects other than string, they could be converted using `#to_s` method, but this is not strictly applied, cause `#kgio_*writev` tries to write at most `sysconf(_SC_IOV_MAX)` items at once (for Linux its value is 1024). First string of returned array could be part of string from array, so that you should assume it is not in consistent state. `#kgio_writev` semantic differs a bit from `#kgio_write` in term of buffers mutability: currently `#kgio_write` tries to follow string changes made concurrently, but `#kgio_writev` works with array's lightweight copy. Signed-off-by: Eric Wong commit fa52cc5d0ef7d04b844868e08e2e7ec3c9e3396e Author: Eric Wong Date: Wed May 30 12:31:19 2012 -0700 tryopen: avoid ambiguous name for subst function Define rb_thread_blocking_region as a macro for MRI 1.8 to prevent confusing output in tools such as valgrind/gdb. commit a72e6cd0dd3038ae2a1b5ef94780143f5ab041c0 Author: Sokolov Yura 'funny-falcon Date: Wed May 30 17:56:55 2012 +0400 use rb_str_subseq for tail string on write Use rb_str_subseq for taking string's tail. rb_str_subseq do not allocate additional memory in this case. And although it prevents from collecting original string, it seems that tests wins both in performance and in memory usage. Use fallback to rb_str_substr on ruby1.8 Signed-off-by: Eric Wong commit 021eaddbfb41d82c0082657f60021bad52b3a6dc Author: Sokolov Yura 'funny-falcon Date: Wed May 30 17:56:54 2012 +0400 Fix UnixClientReadServerWrite test class name Signed-off-by: Eric Wong commit ab3fa8e85c02227985c56261d4898339f85d2b20 Author: Eric Wong Date: Fri Mar 23 21:10:22 2012 +0000 kgio 2.7.4 - small fixes and cleanups Fix build for platforms lacking both TCP_CORK _and_ TCP_NOPUSH There are many test case fixes and cleanups, too. commit 49f28a5257d20a7f4b0aa790424ca207287aa7b6 Author: Eric Wong Date: Sat Mar 24 00:13:40 2012 +0000 test: more workaround for FreeBSD 9.0 Followup-to: e26358413c9d87e1ce8f6cda5cf0b8dd53979ed2 commit e26358413c9d87e1ce8f6cda5cf0b8dd53979ed2 Author: Eric Wong Date: Tue Mar 20 22:00:47 2012 +0000 test/lib_read_write: test workarounds for TCP in FreeBSD 9.0 Under load, TCP sockets may not register as readable right away after the writer finishes. This can be expected for implementations where loopback TCP is a closer simulation of non-local TCP traffic. These test failures were noticed under FreeBSD 9.0. commit 2cab4f2fa642241dbcaf8881d39bd275a59dc67b Author: Eric Wong Date: Fri Mar 23 11:56:04 2012 -0700 test_tryopen: fix horribly-named test for EACCES We can't actually test for EPERM without changing permissions/ownership, and we can't do that without root... commit ce62ddbef053ad31af2f3ec6fcb7d2488859383d Author: Eric Wong Date: Thu Mar 22 06:18:41 2012 +0000 test/lib_read_write: increase test reliability IO#readpartial may not drain the socket buffers enough for kgio_write to succeed on some platforms. So use IO#read for read-in-full behavior. commit 74b9f78e11b915439555290dc3bdd4331303561c Author: Eric Wong Date: Mon Mar 19 06:05:06 2012 +0000 test/*: remove assert_nothing_raised It makes test failures hard to track down, tests will already fail if exceptions are thrown and we'll get nice backtraces. commit 85ae255f73b9b81ae3d17e6420dbb95a29dbe8b7 Author: Eric Wong Date: Mon Mar 19 05:53:03 2012 +0000 fix for non-Linux, non-TCP_NOPUSH platforms We don't need to care for TCP_NOPUSH in read_write.c, it's entirely in autopush.c and no-op on platforms without TCP_CORK/TCP_NOPUSH. TCP_CORK/TCP_NOPUSH are non-POSIX, so it's entirely possible some Free systems will lack them. Reported-by: Edho Arief commit 5ea4cdd1275c0f862bf9dcd1d344dc57d70e5392 Author: Eric Wong Date: Thu Mar 15 07:04:33 2012 +0000 kgio 2.7.3 - compatibility fixes Fixed build and autopush support under Debian GNU/kFreeBSD. Test case fixes for timing-sensitive tests. commit 2c2befb1caa47fe3bf2e6d31dd0733956d178c87 Author: Eric Wong Date: Wed Mar 7 10:14:17 2012 +0000 HACKING: add instructions for running tests Unlike most Rubyists, I prefer GNU make to Rake. commit 92dca5e5554e056f892fcb6cae20693b39b4044b Author: Eric Wong Date: Wed Mar 7 08:37:11 2012 +0000 test_poll: workaround for timing-sensitive test on slow hosts poll(2) may return successfully before it gets interrupted by the signal. Found and fix confirmed by 375gnu on the kgio mailing list. ref: commit 3a847e231d494829077a300912588f499c0bc2af Author: Eric Wong Date: Mon Mar 5 23:19:30 2012 +0000 test: increase delta range for timing-sensitive test This appears to be needed for Debian GNU/kFreeBSD under KVM. commit 1129029ab1bf886979a66a69b04d244dba8b63cf Author: Eric Wong Date: Mon Mar 5 23:19:29 2012 +0000 autopush: fix/enable under Debian GNU/kFreeBSD It seems autopush support in our autopush code has always been broken outside of Linux-based systems, as we never marked the socket as having pending data. commit 56cce133d979c22bbef80fdba1881d8f40876e2f Author: Eric Wong Date: Mon Mar 5 22:37:20 2012 +0000 accept4: require SOCK_NONBLOCK/SOCK_CLOEXEC macros The check for the accept4() function actually succeeds on a stock installation of Debian GNU/kFreeBSD 6.0, but the eglibc headers fail to define the necessary flags. commit 3e7bd918153cd09dd3bdd4e6963e173ae050ae68 Author: Eric Wong Date: Sun Jan 8 03:38:17 2012 +0000 kgio 2.7.2 - for older, and older Rubies Fix a missing #include for Ruby 1.8.5 users. No need to upgrade to this (nor 2.7.1) if you're on a modern version of Ruby. commit d9fe99171c191c55240d756fbc498d82e419f13a Author: Eric Wong Date: Sun Jan 8 03:35:30 2012 +0000 tryopen: remember to include ancient_ruby.h for 1.8.5 Apparently the old Ruby 1.8.6 installation lying around isn't old enough. commit 0dc3909d2ac0f711f038c7a387a9a1da2d5fcb62 Author: Eric Wong Date: Sun Jan 8 01:58:44 2012 +0000 kgio 2.7.1 - compatibility with older Rubies This release fixes some compatibility issues with people stuck on older versions of Ruby/RubyGems. * define RARRAY_PTR/RARRAY_LEN macros for Ruby 1.8.6 * test/test_autopush: skip strace tests if not available * gemspec: disable development dependencies for old systems commit fd1ea5a73155f577f7b77b18c2d6d23af287c123 Author: Eric Wong Date: Sun Jan 8 01:52:52 2012 +0000 gemspec: disable development dependencies for old systems "Enterprise" users are sometimes stuck on older Rubies/RubyGems and this is still required for them. commit e9b90724d8a2ee8c405acdf112adfabcb5cc3159 Author: Eric Wong Date: Sun Jan 8 01:50:21 2012 +0000 test/test_autopush: skip strace tests if not available No need to completely fail on a test. commit 77775e385fe92d1309c65c585b4643712c58e5ba Author: Eric Wong Date: Sat Jan 7 11:15:37 2012 +0000 define RARRAY_PTR/RARRAY_LEN macros for Ruby 1.8.6 Apparently Ruby 1.8.6 is still in use... commit 86f703bfd872536902e7f5293acea3ed0ba0f495 Author: Eric Wong Date: Tue Dec 13 06:09:48 2011 +0000 kgio 2.7.0 - minor updates When running under Ruby trunk/2.0.0dev, all IO objects created by kgio will be close-on-exec by default to match the (future) 2.0.0 behavior. accept()ed sockets in kgio have always been close-on-exec by default.. Singleton Kgio.accept_* methods are deprecated as the kgio_accept/kgio_tryaccept methods all take an additional flags argument. There are various, test, documentation, and error message improvements. commit a40c1d4b30253c68d7997324bcebceb68018bc37 Author: Eric Wong Date: Wed Nov 16 18:42:47 2011 -0800 doc: update documentation regarding accept method flags There's no reason for SOCK_NONBLOCK with Ruby, and SOCK_CLOEXEC has always been on-by-default with accept() wrappers. commit be3672501ecde716dae723e887d4a9e4d731240c Author: Eric Wong Date: Wed Nov 16 18:37:37 2011 -0800 connect,tryopen: set close-on-exec flag for new fds on Ruby 2.0+ All IO objects created by Kgio will have FD_CLOEXEC descriptor flag set on it when run under Ruby 2.0.0dev. This matches the upcoming behavior of Ruby 2.0.0dev for IO objects in the core and standard library. This change does not affect users on Ruby 1.9.3 and earlier. accept()-ed sockets in kgio have _always_ had FD_CLOEXEC set by default. commit 48dc3c5a1943801311567e72a8e69fcb0cd8cf8d Author: Eric Wong Date: Tue Nov 15 13:30:26 2011 -0800 tests: remove tests for IO#nonblock? after accept There's no point in testing a Ruby implementation detail and these tests fail under OpenBSD where the accept()-ed socket inherits the O_NONBLOCK flag from the parent. commit a1a648fe905808ffa902c44ba7626e3b3eeda627 Author: Eric Wong Date: Tue Nov 15 13:20:25 2011 -0800 accept: deprecate singleton Kgio.accept_* methods The kgio_accept and kgio_tryaccept methods now take an additional flags argument, so there's no reason to set global flags anywhere. commit 246cfe96f12ce06a5b504873789ada2efd288885 Author: Eric Wong Date: Tue Nov 15 13:18:32 2011 -0800 accept: always set O_NONBLOCK for accept()-ed sockets in 1.8 This is mostly an implementation detail, but it's already true on OpenBSD (and maybe other BSDs), and also requires no additional syscalls on newer Linux systems. commit 4cef568a9d06033c295c1f4920918c6fed36a24d Author: Eric Wong Date: Mon Aug 29 17:33:43 2011 -0700 test_poll: test for closing a polled IO in sighandler This needs to work similarly to IO.select. commit bb37d358b3326a03a69f65e12c775bb9861b3ad5 Author: Eric Wong Date: Wed Aug 24 17:36:46 2011 -0700 .wrongdoc.yml: add public/private email addresses We want feedback! commit 71eefc0f191d2dde969e6f05f65ec1e80a305e11 Author: Eric Wong Date: Wed Aug 24 17:34:55 2011 -0700 wait.c: fix misspelling in rdoc eye kan spel! commit 7d70f24ac40c984a91f3709b4bd277aaa781746e Author: Eric Wong Date: Tue Aug 9 12:35:18 2011 -0700 LICENSE: s/GNU C Library/kgio/ This is not glibc and I'm not Ulrich Drepper. commit 4ed70d1c840172631e4347e0d0b86a7149f4c8bb Author: Eric Wong Date: Tue Aug 9 12:34:25 2011 -0700 TODO: update SSL/TLS support status :) Monkeys! commit b42164253fa0f4c7c5749eab651c64ddd6cc906a Author: Eric Wong Date: Thu Jul 21 15:57:10 2011 -0700 use rb_update_max_fd() under MRI 1.9.3+ This helps exec() and similar methods close descriptors on shutdown. commit d064ac9334be079d0e830bc2361c065cbaa52a64 Author: Eric Wong Date: Fri Jul 15 11:12:09 2011 -0700 connect: more descriptive error for TCP port We want more descriptive error messages and don't want crazy stuff like floats. commit dcaa9f9be83bfd59503033ae8f8eeca79c68c9df Author: Eric Wong Date: Thu Jul 14 18:54:37 2011 -0700 kgio 2.6.0 - minor feature update and cleanups We now export SOCK_NONBLOCK, SOCK_CLOEXEC constants in the Kgio namespace to make kgio_tryaccept/kgio_accept easier-to-use. There are also some minor internal cleanups. commit 8baed92a7c02adcc913bcc7760b77a240b529c63 Author: Eric Wong Date: Thu Jul 14 17:32:10 2011 -0700 add prototype for rb_thread_blocking_io_region() It's no in the public headers, but Ruby 1.9.3 will have it and it's still superior to rb_thread_blocking_region() even though it's not ideal. commit 823f41d0f86dda497c166a839d8215275f5d48a0 Author: Eric Wong Date: Thu Jul 14 17:29:57 2011 -0700 missing_accept4: don't use FD_CLOEXEC for SOCK_CLOEXEC emulation It's too confusing and may break binary compatibility if the system is later upgraded. commit 142beefba460685fea5b6646e1ba629f9ee207b3 Author: Eric Wong Date: Thu Jul 14 17:29:05 2011 -0700 accept.c: fix RDoc for new Kgio::SOCK_* constants We don't want people using the compatibility constants since they're actually broken on systems with real accept4(). commit c22f2b5ebccbca8e04aa22821964f67c4a81c675 Author: Eric Wong Date: Wed Jul 6 19:11:47 2011 -0700 export SOCK_NONBLOCK, SOCK_CLOEXEC constants in Kgio It's more reliable than relying on IO::NONBLOCK and Fcntl::FD_CLOEXEC constants. The existing constants are not guaranteed to be equivalent to what accept4() takes even though the current Linux implementation does it this way. commit 8e1a53f99a752d8ccba324560a9e52bf6e80680d Author: Eric Wong Date: Wed Jul 6 18:32:39 2011 -0700 accept.c: reinstate errno after calling my_fileno() my_fileno() may change errno on some Rubies commit e720827b48c3318627d1471ad4ec90d6166520fd Author: Eric Wong Date: Fri Jun 17 19:01:06 2011 -0700 doc: update documentation for kgio_addr attribute We support IPv6, not just IPv4 commit ce08273658b14dc3d53273b514e46b6e65882bf6 Author: Eric Wong Date: Fri Jun 17 18:35:38 2011 -0700 doc: call-seq for kgio_wait_writable It's there for kgio_wait_readable commit 72ab71667b2a9c27d1eda73aacb8b86187f317d0 Author: Eric Wong Date: Fri Jun 17 18:10:44 2011 -0700 ancient_ruby.h: avoid symbol clobbering In case the toolchain can't test the feature properly or Ruby is upgraded and the symbol is added. commit 079f9227dce2e69bd3460c783e12fa05c687b7d3 Author: Eric Wong Date: Wed Jun 15 22:16:15 2011 -0700 fix misc compiler warnings Some installations of Ruby clobbered my usual CFLAGS=-Wall commit 8fe21f6758bb877efacce1fa6573e72625252585 Author: Eric Wong Date: Wed Jun 15 08:30:03 2011 +0000 make timed kgio_wait_* implementation safer IO.select can handle fd >= 1024 safely in some Rubies while fd_set may not. We could use rb_thread_fd_select(), but rb_wait_for_single_fd() is available now so the former is not worth the maintenance hassle. commit 9159f70862e3e6a76d821c4a70bc68a603793a49 Author: Eric Wong Date: Wed Jun 15 02:39:13 2011 +0000 Kgio::File includes Kgio::PipeMethods module Kgio::File may be used to open FIFOs, so non-blocking I/O is still useful in that context. commit 6d6f704e29d7bf1e95f1c9c60cbab82ec3d430d2 Author: Eric Wong Date: Wed Jun 15 02:34:49 2011 +0000 Kgio::File.tryopen runs GC on ENOMEM It is possible but unlikely to get ENOMEM on open(2), so try to GC away some files. commit e693b871567119345c2c567bfa2ad46e210d655b Author: Eric Wong Date: Wed Jun 15 02:27:06 2011 +0000 doc: add rdoc for Kgio::File.tryopen New feature in 2.5, we now have 100% documentation again. commit 5550222b389c2971ee98bdd62c749ce228efda06 Author: Eric Wong Date: Wed Jun 15 02:12:31 2011 +0000 kgio_wait_*able: documentation for optional timeout New features are better if they're documentated. commit 3fe0ad91d7a81a84ecc9e75ba8f5162bad30b2ac Author: Eric Wong Date: Wed Jun 15 02:06:32 2011 +0000 doc: use librelist.org instead of librelist.com Non-profit TLD is better sounding for a Free Software project. commit 2e20f8b1a1b74ae5442e44de70196a5e3121c642 Author: Eric Wong Date: Wed Jun 15 02:04:00 2011 +0000 GIT-VERSION-GEN: bump version for new API commit aa9594b0d38b012729a46fd519fcc369600f4c3e Merge: bbf9a3b fdde0a2 Author: Eric Wong Date: Wed Jun 15 02:05:43 2011 +0000 Merge branch '2.4-stable' * 2.4-stable: kgio 2.4.2 - OpenSolaris build fix extconf: -lnsl and -lsocket checks for OpenSolaris commit fdde0a27c51134d7a6a9cd1d66d93d1ac6640940 Author: Eric Wong Date: Tue Jun 14 18:40:07 2011 +0000 kgio 2.4.2 - OpenSolaris build fix * adds -lnsl and -lsocket checks for OpenSolaris commit 1a7eed4c69abb7bafd3e3dc2acd13e243995e98e Author: Eric Wong Date: Mon Jun 13 14:26:10 2011 -0700 extconf: -lnsl and -lsocket checks for OpenSolaris Reported via private email. (cherry picked from commit d224563823accca63fd871260e3f0dad6758c8d4) commit bbf9a3bc0ca2d91705e27ad8bfb5c0ed9651a2ef Author: Eric Wong Date: Tue Jun 14 08:41:32 2011 +0000 add timeout to kgio_wait_{read,writ}able io/wait doesn't have an IO#wait_writable method, yet[1] and IO#wait checks FIONREAD which makes it unsuitable for certain descriptors. This method uses the new rb_wait_for_single_fd() function in Ruby 1.9.r. This internally uses ppoll() under Linux, meaning it performs the same regardless of the FD used. [1] http://redmine.ruby-lang.org/issues/4647 [2] http://redmine.ruby-lang.org/issues/4849 commit d224563823accca63fd871260e3f0dad6758c8d4 Author: Eric Wong Date: Mon Jun 13 14:26:10 2011 -0700 extconf: -lnsl and -lsocket checks for OpenSolaris Reported via private email. commit 83c6584be53d6863e647067ba385b42ed5347cdb Author: Eric Wong Date: Mon May 23 11:28:29 2011 -0700 README: clarify that we only work on Unix-like systems We won't waste time with inferior, non-Free platforms. commit 2ade0059c67a0ae4fa1b416c500169e3ac66bfff Author: Eric Wong Date: Fri May 20 20:00:15 2011 -0700 pkg.mk: update to the latest version Fixes locale issues with grep and adds check-warnings commit 8184059bf16d73f0a386ddbf68c4949d1dec3bdf Merge: a5357ad f656c49 Author: Eric Wong Date: Fri May 20 19:54:48 2011 -0700 Merge branch '2.4-stable' * 2.4-stable: kgio 2.4.1 - Kgio.poll avoids EINTR, really Kgio.poll: ensure EINTR never gets raised commit f656c49f77d896cbbb1e684d826472c09dcc2253 Author: Eric Wong Date: Fri May 20 19:53:03 2011 -0700 kgio 2.4.1 - Kgio.poll avoids EINTR, really This release fixes a race condition that could allow Errno::EINTR to be raised even though the 2.4.0 release was supposed to stop that. Nobody uses Kgio.poll, really, so this shouldn't be an issue for real code, yet. commit f809a87f70f0937a87b5d3a83704847daceef4dd Author: Eric Wong Date: Fri May 20 19:50:34 2011 -0700 Kgio.poll: ensure EINTR never gets raised Retry on a zero timeout if we get interrupted even if the timeout expired. This is also what IO.select does in Ruby itself. commit a5357ad014d2eacc99dd7ee46040686cbf1d871c Author: Eric Wong Date: Fri May 20 19:50:34 2011 -0700 Kgio.poll: ensure EINTR never gets raised Retry on a zero timeout if we get interrupted even if the timeout expired. This is also what IO.select does in Ruby itself. commit 605765ded31c784727077dfca573092ba725f717 Author: Eric Wong Date: Fri May 13 17:57:10 2011 -0700 Kgio.tryopen => Kgio::File.tryopen This will allow users to subclass Kgio::File and override certain behavior (e.g. overriding "#each"). commit c8bd876fb5086e5b79299869b4c29f1f7f020b4d Author: Eric Wong Date: Fri May 13 17:48:56 2011 -0700 return Kgio::File for Kgio.tryopen This also allows us to return/override #to_path and #path if necessary, but so far everything works with MRI 1.8, MRI 1.9, and Rubinius. commit 6cefcff5889cceaa001f76f4be1a1c5e513b241d Author: Eric Wong Date: Fri May 13 15:55:50 2011 -0700 add Kgio.tryopen method For the case where a file is readable, it's faster to just call open() instead of stat()-ing and then calling open(). open() failures are still relatively cheap, too, they're still more expensive than a failed stat() but cheaper than raising an exception. commit ab732113e13f1690fd2c1a18d1c66beb7864d847 Author: Eric Wong Date: Thu May 5 15:58:17 2011 -0700 kgio 2.4.0 - portability fixes and more == All supported platforms (*nix + MRI 1.8+, Rubinius) * OpenBSD (and possibly other *BSD) fixes, thanks to Jeremy Evans. * kgio_accept and kgio_tryaccept now take an optional second argument for flags (like the accept4() flags argument). == Ruby 1.9-only things * Kgio.poll no longer raises Errno::EINTR to match IO.select. == Ruby 1.9 trunk things * close() on an active FD in a different thread is better handled/detected. * copy-on-write for strings is properly triggered commit 43190caf48309fa6aa5423d2d1ae5c320ad07fb5 Author: Eric Wong Date: Thu May 5 15:38:00 2011 -0700 test_cross_thread_close: disable on RUBY_ENGINE != "ruby" These aren't well-defined semantics... commit 9c98bbd79f152fd72b257f9c37ca185587e56225 Author: Eric Wong Date: Thu May 5 13:24:36 2011 -0700 read_write: call rb_str_modify() before rb_str_resize() This is needed under Ruby trunk if the string is not actually resized. commit c821ebeb851807840f74c4cb4f1a10e691bf222a Author: Eric Wong Date: Thu May 5 13:11:53 2011 -0700 poll: deal with pollset changes on EINTR This allows callers to modify the pollset, interrupt the polling thread, and then poll with the modified pollset. This is also important for dealing with closed IO objects that may have been invalidated due to GC, too. commit f589a9ae18216e1220bea8f905f33051e88f1ae7 Author: Eric Wong Date: Thu May 5 12:30:45 2011 -0700 test_tcp_connect: disable wait_writable_set test on non-Linux OpenBSD seems to make connect() return success immediately even on a non-blocking socket, so it's hard to test for. Thanks to Jeremy Evans to reporting the issue under OpenBSD. ref: http://mid.gmane.org/20110505181846.GB9693@jeremyevans.local commit 9900efecb60635ad97b5c00c76eb60252839b1c1 Author: Jeremy Evans Date: Thu May 5 11:18:46 2011 -0700 fix kgio_peek for !KGIO_NOPUSH systems Fix kgio_peek by adding an empty kgio_autopush_recv if !KGIO_NOPUSH. ref: http://mid.gmane.org/20110505181846.GB9693@jeremyevans.local Acked-by: Eric Wong commit 577cf3056d9f3088145aea51bbc09a0c90a7695e Author: Jeremy Evans Date: Thu May 5 11:49:01 2011 -0700 connect: zero out hints argument for getaddrinfo Some systems like OpenBSD are stricter about irrelevant fields than GNU/Linux. [ew: commit message] ref: http://mid.gmane.org/20110505181846.GB9693@jeremyevans.local Acked-by: Eric Wong commit 537e4c341137a45330e28376e8f29da7df44808f Author: Eric Wong Date: Wed May 4 19:26:38 2011 -0700 Kgio.poll restarts on interrupt This changes our semantics, but it's unlikely anybody relies on EINTR right now... commit f2a3998e0c0f63ad14acf5ccc0141fc6cdce24e3 Author: Eric Wong Date: Wed Apr 27 11:44:51 2011 -0700 extconf: remove unnecessary dir_config statement I didn't know this about mkmf at the time... commit 3033f2e8f178c0f150cfd3e2a070570154a27430 Author: Eric Wong Date: Thu Apr 21 20:08:01 2011 +0000 doc: improve kgio_accept/kgio_tryaccept docs Documenting the new flags options and also improving style of the existing class overrides. commit b885cf9a2ef0864dcebb9bba7b1fcf3eb08f9ae8 Author: Eric Wong Date: Thu Apr 21 20:06:08 2011 +0000 test_autopush: attempting to fix a timing test... Not fun on slow systems, unfortunately... commit be46333541acd72bde3544a3e86f6ead0bb364d0 Author: Eric Wong Date: Wed Apr 20 18:36:09 2011 -0700 doc: fix trywrite call-seq commit 3b48d6c823a7da34c0b37d8eb3c11964c4a3ba89 Author: Eric Wong Date: Thu Apr 14 02:31:50 2011 +0000 test_autopush: use assert_in_delta for test Some systems are slower than others... commit f36f7584b766cb5d558fcfa94ea639e6090bcb54 Author: Eric Wong Date: Mon Apr 11 18:21:09 2011 +0000 gemspec: bump wrongdoc version commit 212f6fdfe51e2167d362686bb2622ce04e6e5de5 Author: Eric Wong Date: Mon Apr 11 18:20:27 2011 +0000 gemspec: remove unnecessary require_paths RubyGems handles it already commit fefd652d6cc5825bebbe164a360c4a06b1399dcb Author: Eric Wong Date: Sun Apr 10 07:27:40 2011 +0000 add test for cross thread close Ruby 1.9.3dev has better support for it commit 983e8ff915118d00b683109df0834733b485086d Author: Eric Wong Date: Sun Apr 10 07:04:41 2011 +0000 read_write: detect closed fd on EINTR Another thread may have killed the descriptor. commit 01949865e2715b487e098dbc60ba3ae7a01a3b54 Author: Eric Wong Date: Sun Apr 10 06:07:55 2011 +0000 accept: better detect closed files on EINTR/EAGAIN Another thread may have killed the acceptor object. commit 59782a15d0be87130934cbecb34ed639be68b44a Author: Eric Wong Date: Sun Mar 20 16:27:55 2011 -0700 *accept methods can take flags argument, like accept4() This allows applications to not rely on global accept4_flags commit 0ef079617b7d71cc26574247918c4a3e18454b21 Author: Eric Wong Date: Sat Mar 19 18:29:15 2011 -0700 accept: prepare optional flags argument to accept() wrappers Don't force people to rely on global flags, there may be blocking parts of an otherwise non-blocking program... commit f8b8b3f73d238d4c29368b4a0f768c5afb03d43d Author: Eric Wong Date: Thu Mar 17 14:06:45 2011 -0700 HACKING: updates for wrongdoc vs rdoc commit c79b5f9037ce69fb3ebce470a14af505aa1c8f5a Author: Eric Wong Date: Tue Mar 15 12:00:52 2011 +0000 kgio 2.3.3 - minor fixes We no longer over-allocate memory for Kgio.poll (1.9.x-only). Under Ruby 1.9.3dev, we also use rb_thread_io_blocking_region to properly deal with cross-thread IO#close. commit 82a3e7d24e3cd51f15df593590986b7c5d0834aa Author: Eric Wong Date: Mon Mar 14 19:35:01 2011 +0000 poll: fix over-allocation for poll(2) Oops, fortunately we allocated too much and not too little. commit 6299ef55e20454eaca0b9860ac4a9bd6ddd143a7 Author: Eric Wong Date: Mon Mar 14 19:35:01 2011 +0000 use rb_thread_blocking_io_region if possible It's in Ruby 1.9.3dev and tracks waiting FDs on blocked threads. commit 59ed57abf542b89babf595e5508cba42ceb9fd47 Author: Eric Wong Date: Fri Feb 25 16:48:11 2011 +0000 doc: more consistent references for kgio_wait_*able methods Much nicer this way... commit 88ae3fb48de345a3a102ac4d17bb71e8a4691230 Author: Eric Wong Date: Tue Feb 15 08:55:25 2011 -0800 kgio 2.3.2 - OpenBSD build fix Thanks to Jeremy Evans, this release fixes the build under OpenBSD. commit 8556c2ccf64840a080a928312d8feed2834d4d29 Author: Jeremy Evans Date: Mon Feb 14 17:03:53 2011 -0800 Fix build on OpenBSD OpenBSD's getnameinfo(3) requires the sys/types.h header file, and including it should not cause a problem for other OSes. Acked-by: Eric Wong commit 6d19ebb3a917b566830f8d33e95b1eea2e99658d Author: Eric Wong Date: Sun Feb 13 16:50:13 2011 -0800 kgio 2.3.1 - compatibility fix * connect.c: disable AI_NUMERICSERV It's not needed since we already verify the service is a numeric port. AI_NUMERICSERV is not available in older glibc (<2.3.4) and probably other old systems. commit b30aa658d7b4d946427119b23cfc264cc7172149 Author: Eric Wong Date: Sun Feb 13 16:33:29 2011 -0800 connect.c: disable AI_NUMERICSERV It's not needed since we already verify the service is a numeric port. AI_NUMERICSERV is not available in older glibc (<2.3.4) and probably other old systems. commit f74bcf37ed31611d14aba2d4b4518c6a3dea0f40 Author: Eric Wong Date: Wed Feb 9 10:10:36 2011 +0000 kgio 2.3.0 - MSG_PEEK and poll(2) support recv() with MSG_PEEK for sockets is added with the try* interface. SocketMethods#kgio_trypeek and SocketMethods#kgio_peek or Kgio.trypeek for non-Kgio-enabled sockets. For Ruby 1.9 only: poll(2) is exposed via the Kgio.poll singleton method and should provide an alternative for IO.select users. Both of these new features should work well on modern Unix-like operating systems. commit 00d1bc89ed3811701195bf5ddb400a0a0067126f Author: Eric Wong Date: Tue Feb 8 14:11:45 2011 -0800 move poll support checks to kgio.h We may use poll elsewhere... commit 257b090af54e6a1cecd44325f8664c4c682a6740 Author: Eric Wong Date: Tue Feb 8 13:56:39 2011 -0800 pkg.mk: update to the latest * Fixes Ruby 1.9.3dev deprecation warnings * Fixes some documentation dependency issues * Allows RUBY_TEST_OPTS to be passed to unit tests commit d8616b605ad4d83b69e2679e1c210e476cc18e00 Author: Eric Wong Date: Tue Feb 8 13:50:55 2011 -0800 doc: fully RDoc all methods and classes Whee! commit 3f6bffb4c4297df48a69d146243fbe5ba8040cb5 Author: Eric Wong Date: Tue Feb 8 13:50:01 2011 -0800 doc: fix accept -> tryaccept copy+paste error Oops commit 823e8978b38dcfb642e5059e879af4209ebba0f1 Author: Eric Wong Date: Tue Feb 8 13:36:56 2011 -0800 README: fix download link/ref commit 5c480aee3067006b5da6d45b7de41d8401b70848 Author: Eric Wong Date: Tue Feb 8 10:41:51 2011 +0000 preliminary poll(2) support It's a nice alternative to IO.select for higher-numbered file descriptors, especially sparse ones. Our interface also generates less garbage than IO.select does. commit 47653194bf6ad53b9f5fca1b266c30855df5ebbd Author: Eric Wong Date: Sun Feb 6 21:56:35 2011 +0000 add support for recv() with MSG_PEEK Kgio.trypeek, kgio_trypeek and kgio_peek methods are added for using with sockets. commit 75a7da2bd757617995f5492df1205e4a3459618b Author: Eric Wong Date: Thu Feb 3 18:46:21 2011 -0800 kgio 2.2.0 - kinder, gentler I/O for the Internets * sockets accept()ed by a TCP_NOPUSH/TCP_CORK listener automatically flush on kgio_*read calls if there is pending data. "Kgio.autopush = false" disables this globally, and Kgio::Socket also get "kgio_autopush=" to enable/disable on a per-object individual basis. * ECONNRESET exceptions get empty backtraces for kgio_*read. There's nothing a programmer can do about these, so there's no point in going through the expensive backtrace generation process. * Kgio.try* singleton methods added for working with non-Kgio enhanced objects. No more needing to use Object#extend and blowing away your method cache to make existing I/O objects kinder and gentler. * IPv6 support should be complete, systems without a native getaddrinfo(3) are now unsupported (and will remain so unless somebody complains). There should be no other backwards-incompatible changes other than requiring getaddrinfo(3) and friends for IPv6 support. commit c8fb5aa33262a455997ff6a57659a8d125f36d66 Author: Eric Wong Date: Thu Feb 3 14:28:11 2011 -0800 add SocketMethods#kgio_addr! This refreshes (or sets) the @kgio_addr ivar for sockets that didn't go through kgio_accept or kgio_tryaccept. commit cff0dd2f73acc73f721b2a589af9e37baedd2489 Author: Eric Wong Date: Wed Feb 2 13:56:31 2011 -0800 fix typos in ipv6 test case Oops, RTFE :P commit 17abe6ce8f01810022b948c71de0026b4ac89597 Author: Eric Wong Date: Wed Feb 2 21:33:28 2011 +0000 add proper IPv6 support No extra #ifdefs, we just won't support old systems without getaddrinfo() and friends anymore. I doubt anybody still has them... commit 879f2f0ee9133f34ec3e24141bdb4936e3408d3a Author: Eric Wong Date: Tue Feb 1 14:00:07 2011 -0800 avoid re-interning if GCC is not used (or under 1.8) Needless calls to rb_intern are wasteful in even semi-frequently used code. commit 499f158c74b7c455dca08fc30be88cb699ee24c6 Author: Eric Wong Date: Tue Feb 1 13:58:20 2011 -0800 kgio_*read: empty backtrace for ECONNRESET There's nothing a programmer can do about ECONNRESET so just make the exception cheaper to raise so it can still be logged. commit 36f69750cd69cbf892580da04be6675e23d92f6f Author: Eric Wong Date: Mon Jan 31 18:39:31 2011 -0800 add singleton methods for non-Kgio objects This allows people to more easily use Kgio in existing apps. commit 37e50a9a5fcd45242373379c0dc61ebf8ff609af Author: Eric Wong Date: Mon Jan 31 18:27:24 2011 -0800 autopush: enable accessors for client sockets Might as well allow clients to efficiently handle TCP_CORK/TCP_NOPUSH, too. commit d4773fc63a847119004c17a1b8803a815f99d98a Author: Eric Wong Date: Mon Jan 31 17:34:07 2011 -0800 autopush: enable this by default TCP_CORK (and presuably TCP_NOPUSH) aren't remotely useful in Rainbows! without this and there's almost no overhead for MRI, either. commit 8a1fc65c88dee174940735bb46074c72ac47ce61 Author: Eric Wong Date: Mon Jan 31 17:05:48 2011 -0800 autopush: optimize away ivar usage under MRI We know that all versions of MRI have a small RFile structure that is allocated in the same object slots as other Ruby types and also zeroed on allocation. This optimization enables us to fall back to using ivars in case MRI changes or if we're used on other Rubies. commit 6479b6d3934b8930910e0057f516aa019dd7a8c7 Author: Eric Wong Date: Mon Jan 31 15:18:33 2011 -0800 autopush: enable for TCP_NOPUSH under FreeBSD Hopefully it works for people who use TCP_NOPUSH... commit 15744a90cda72e9007914cd2a78b0b2949193479 Author: Eric Wong Date: Mon Jan 31 13:58:53 2011 -0800 autopush: simplify implementation and just use ivars Duh... commit 313d2bb8d37dbc5602e464def90b3e7fa9f60924 Author: Eric Wong Date: Mon Jan 31 13:03:02 2011 -0800 rename nopush_smart to autopush This is probably a better name for it, libautocork is a nice name even though we won't use it directly. commit 910f6f3df099c04fcd55bd6b20785cce69cb36ae Author: Eric Wong Date: Thu Jan 27 19:43:39 2011 -0800 preliminary implementation of "smart_nopush" It only supports TCP_CORK under Linux right now. We use a very basic strategy to use TCP_CORK semantics optimally in most TCP servers: On corked sockets, we will uncork on recv() if there was a previous send(). Otherwise we do not fiddle with TCP_CORK at all. Under Linux, we can rely on TCP_CORK being inherited in an accept()-ed client socket so we can avoid syscalls for each accept()-ed client if we already know the accept() socket corks. This module does NOTHING for client TCP sockets, we only deal with accept()-ed sockets right now. commit ec91ac3d8c8d9236ba0cd01794c9c4a3ee3f7eeb Author: Eric Wong Date: Thu Jan 27 14:11:16 2011 -0800 revamp packaging makefile, update URLs More common code that's still GNU make is better for my sanity. Also, bogomips.org went on a URL diet recently. commit f6c79438ed195bb706903d104cce850bfbfbac41 Author: Eric Wong Date: Tue Jan 18 15:52:25 2011 -0800 add tests for empty writes, too There could be some platforms that dislike it... commit 9c81cc3fd8d2b3dce68d69d8e0c56a4c5d89ebf0 Author: Eric Wong Date: Thu Jan 13 14:46:02 2011 -0800 Makefile: remove non-existent target reference Oops commit fb8104e1f2a5d1cdcb99a19b6a4bdabf0b1c2643 Author: Eric Wong Date: Sat Dec 25 18:06:47 2010 -0800 kgio 2.1.1 - one small Rubinius fix We now avoid errno side-effects in kgio_wait_*able methods. This affects Rubinius, but may affect other Ruby platforms (particularly those that use stdio) as well. commit 6ab4331f8137e949ab57f014f96ff3918a315044 Author: Eric Wong Date: Sat Dec 25 18:02:16 2010 -0800 avoid errno side-effects in kgio_wait_*able Retrieving the file descriptor may have side-effects on certain Ruby implementations (e.g. Rubinius), so ensure our errno is preserved before calling rb_io_wait_*able(). commit 3a0323b642ee054319a5e64ffe28e089bbd013e4 Author: Eric Wong Date: Sun Dec 26 01:08:58 2010 +0000 gemspec: point folks to the public mailing list It's more useful that way. commit 9d5c9e6c9975cb5c10e7384aed9ed22ae0ee57c8 Author: Eric Wong Date: Sun Dec 26 01:02:00 2010 +0000 kgio 2.1.0 - accept improvements and fixes kgio_accept and kgio_tryaccept now take an optional argument to override the default Kgio::Socket class that is returned. These methods also fall back to using regular accept() if kgio was built on a system with accept4() and later run on a system without accept4(). commit dcd5eff7dd5d5861b67667f48424979be9bcabc8 Author: Eric Wong Date: Sun Dec 26 00:21:25 2010 +0000 quiet down some harmless compiler warnings Less noise means we'll notice real bugs sooner. commit 5280f35f131d88f90afffff0e10f7900530728aa Author: Eric Wong Date: Sat Dec 25 23:00:05 2010 +0000 accept4: fall back to regular accept() on ENOSYS kgio may occasionally be built on a system with accept4() and then deployed on one without it. Handle this case gracefully since it unfortunately happens on production systems. commit b859c4a12905cbd71d19cde2aaa9f88ec0374cc5 Author: Eric Wong Date: Sat Dec 25 22:44:53 2010 +0000 accept methods may take an optional argument This is preferred as we no longer have to rely on a global constant. commit ef069ece624906b3946248421620d8458bcef605 Author: Eric Wong Date: Fri Dec 24 09:21:19 2010 +0000 Rakefile: fix RAA license Oops, we were never Ruby licensed. commit 472240687caf3f113a3ff408729f8205c475d7d5 Author: Eric Wong Date: Fri Dec 24 09:20:40 2010 +0000 doc: use wrongdoc for documentation wrongdoc factors out a bunch of common code from this project into its own and removes JavaScript from RDoc to boot. commit 64bbc95d2192fb621b763c1c4d1ae32940c1a5ac Author: Eric Wong Date: Wed Dec 22 12:12:18 2010 -0800 fix errors in RDoc Noticed-by: Iñaki Baz Castillo commit f093312ad1ed336363f352991b6b99d96f7aed1d Author: Eric Wong Date: Thu Nov 18 17:16:53 2010 -0800 kgio 2.0.0 - major internal API changes (no code changes from 2.0.0pre1) This release should make Kgio easier and more consistent to use across a variety of libraries/applications. The global Kgio.wait_*able(=) accessor methods are gone in favor of having default kgio_wait_readable and kgio_wait_writable methods added to all Kgio-using classes. Sub-classes may (and are encouraged to) redefine these if needed. Eric Wong (7): expand Kgio::*#kgio_read! documentation prefer symbolic names for waiting read/writability EOFError message matches Ruby's README: Gemcutter => RubyGems.org update documentation with mailing list info add default kgio_wait_*able methods switch entirely to kgio_wait_*able methods commit edfa7e60de5556b6abc9febe6a21e12dadbafd0b Author: Eric Wong Date: Thu Nov 18 15:42:27 2010 -0800 Rakefile: list prerelease tags as well Since we do prerelease nowadays before real ones. commit d78a2075bdb0a30bf0064d2857011c330cc0d09e Author: Eric Wong Date: Thu Nov 18 15:38:12 2010 -0800 move website to bogomips.org This project is useful enough for others and to stand alone without needing to be associated with Unicorn. commit 28070c522aff233eadb7e167f8d4e8122cd0bb47 Author: Eric Wong Date: Thu Nov 18 15:15:40 2010 -0800 kgio 2.0.0pre1 - major internal API changes This release should make Kgio easier and more consistent to use across a variety of libraries/applications. The global Kgio.wait_*able(=) accessor methods are gone in favor of having default kgio_wait_readable and kgio_wait_writable methods added to all Kgio-using classes. Sub-classes may (and are encouraged to) redefine these if needed. Eric Wong (7): expand Kgio::*#kgio_read! documentation prefer symbolic names for waiting read/writability EOFError message matches Ruby's README: Gemcutter => RubyGems.org update documentation with mailing list info add default kgio_wait_*able methods switch entirely to kgio_wait_*able methods commit c69955e64648ab6a3471a54f7885a320428682f9 Author: Eric Wong Date: Thu Nov 18 14:37:05 2010 -0800 switch entirely to kgio_wait_*able methods This removes the global Kgio.wait_*able accesors and requires each class to define (or fall back to) the Kgio::DefaultWaiters methods. commit f1b497e601ed2acb54f75dc989d0a5ec7afebca0 Author: Eric Wong Date: Thu Nov 18 13:38:32 2010 -0800 add default kgio_wait_*able methods It makes it easier for people to use certain overrides without killing other methods. This is the first step in fixing problems people were having with dalli 0.11.1+ while running Unicorn. commit 827ad6b4fba768a5cac8fb4e83fbbf61cf7a3194 Author: Eric Wong Date: Mon Nov 15 10:33:55 2010 -0800 update documentation with mailing list info We're a real project, apparently, so it can have its own mailing list. commit fd88eae588c1e715dcaf3a1a000391cc13481e02 Author: Eric Wong Date: Mon Nov 15 10:22:49 2010 -0800 README: Gemcutter => RubyGems.org That's the new name for it and it's official commit 8615a3f9554df0fd7f7f088cd49cf1e3be49de9f Author: Eric Wong Date: Fri Nov 12 20:25:50 2010 -0800 EOFError message matches Ruby's This makes messages appear less different than Ruby when using kgio_read! Requested-by: Mike Perham commit 2772ed8bfe108b66b7493bc5cb0c40ddeb1ca57d Author: Eric Wong Date: Fri Nov 5 09:01:08 2010 +0800 prefer symbolic names for waiting read/writability There's no point in using constants that point to symbols instead of just the symbols themselves. commit bf3b507791403811bece9dff915ca10757bca519 Author: Eric Wong Date: Thu Oct 28 21:02:31 2010 +0000 expand Kgio::*#kgio_read! documentation If the author can forget why it was written, so can the rest of the world. commit f4d08a07a02393cca5ddd1277acc4f95c83307ff Author: Eric Wong Date: Fri Oct 8 14:55:16 2010 -0700 kgio 1.3.1 - fix zero-length reads kgio_read and kgio_tryread will now return an empty string when a length of zero is specified instead of nil (which would signal an EOF). This emulates the behavior of IO#read, IO#readpartial, IO#sysread, IO#read_nonblock in core Ruby for consistency. commit d225ede82d820d045bd7cfb826f444cf6601577c Author: Eric Wong Date: Fri Oct 8 02:53:38 2010 -0700 return empty string on length=0 This matches behavior of all the core Ruby methods. commit e4599227f0da0f652cbcb52838e631d7384dcd0d Author: Eric Wong Date: Thu Oct 7 20:02:40 2010 -0700 kgio 1.3.0 - bug and usability fixes * make Kgio::WaitWritable and Kgio::WaitReadable symbols * trywrite: fix stupid off-by-one error causing corrupt writes on retries commit f5fc35221d37141b0f72278c7b969211410e94c0 Author: Eric Wong Date: Thu Oct 7 20:00:09 2010 -0700 tests: don't trust what I think I know about Ruby case/when and === didn't actually work as I expected them to. commit 2152188f41bf2a5067e84a4404b48b2282a9dd55 Author: Eric Wong Date: Thu Oct 7 19:56:57 2010 -0700 trywrite: fix stupid off-by-one error causing corrupt writes Oops! commit c448ad898ecb7f354a32a320294da4727fc9af52 Author: Eric Wong Date: Thu Oct 7 19:55:49 2010 -0700 make WaitWritable and WaitReadable symbols This makes them easier to compare with === when used in case/when statements in Ruby commit 49f0b98c69f1f0bf637953d0bfc96b764f00ab9b Author: Eric Wong Date: Thu Oct 7 07:15:49 2010 +0000 kgio 1.2.1 - doc and *BSD workarounds This fixes our accept4() wrapper which did not work as expected on some *BSD-based systems due to fcntl(fd, F_GETFL) returning false information. Linux 2.6+ users are unnaffected, including those without accept4(). Also some RDoc fixes. commit 03344bb763f5269afe7fafd56a47270719c7ef9e Author: Eric Wong Date: Thu Oct 7 07:14:07 2010 +0000 doc: fix RDoc generation Oops, completely broken by the splitting of the code. commit 637317eb479525dca543eda7a8977410bc43b832 Author: Eric Wong Date: Wed Oct 6 14:08:35 2010 -0700 accept4: workaround (P)OS X bug w/O_NONBLOCK Apparently fcntl(fd, F_GETFL) can return falsely return the O_NONBLOCK flag without actually having it set in the kernel. This is totally broken on the part of the OS. commit ca76c75f8a24d0cd6828fe16ca3790a277b35f8d Author: Eric Wong Date: Wed Oct 6 14:06:27 2010 -0700 build: pick on on modified extension files We build more than one file nowadays. commit 65f96b7750616bc210397c16eea40961e578a788 Author: Eric Wong Date: Wed Oct 6 11:51:04 2010 -0700 doc: fix typo in Kgio.accept_cloexec= doc oops... commit 414dd17f1009c571e2d7657721271756e3d4dd8e Author: Eric Wong Date: Tue Oct 5 16:09:40 2010 -0700 kgio 1.2.0 - cleanups and minor improvements The C extension is now split into several files for ease-of-maintenance. Slightly more common, client-triggerable exceptions (EOFError, Errno::EPIPE, Errno::ECONNRESET) are now less expensive as they are generated without backtraces. commit e085bb9600b190692beb5efc85656ebf127ae08c Author: Eric Wong Date: Tue Oct 5 15:45:16 2010 -0700 generate empty backtraces for EPIPE and ECONNRESET Malicious clients may disconnect during big writes to cause EPIPE and ECONNRESET exceptions. Generating backtraces can be expensive with Ruby, so mitigate the DoS vector by lowering the cost of generating an exception. commit b168cc894037620cab82fa82f3ab37a3aab81570 Author: Eric Wong Date: Tue Oct 5 15:26:57 2010 -0700 add kgio_read! methods which may raise EOFError Except EOFError is gently raised to not include a huge backtrace. Large backtraces can be a performance problem on busy servers that malicious clients may exploit to deny service. commit 870ada92db7071c7982913e508ac35b97d6e8761 Author: Eric Wong Date: Tue Oct 5 11:45:02 2010 -0700 GNUmakefile: use portable tar invocation We've been spoiled by GNU tar. commit 2a6115a89d5c95428bd6c3e0bc10e5a3a4c3c3be Author: Eric Wong Date: Wed Sep 29 18:25:58 2010 -0700 refactor and split into separate files Making the code easier to read and navigate. This also frees us from having to use the stupid A4_ prefix for accept4(2) flags since it conflicts with the socket(2) ones. commit 8fe89997453d6c530c3f5e08bc9c1da40a621248 Author: Eric Wong Date: Wed Sep 29 17:13:21 2010 -0700 Make kgio_trywrite more aggressive with retrying Partial writes can be retried until completely denied with EAGAIN. Often times, it is beneficial to retry immediately after a partial write because the kernel may allocate more buffers or the reader can drain the buffers. This helps the caller avoid crossing the Ruby <-> C boundary more than necessary. commit 39c851e595970a2349a8a39878afd94a3324e102 Author: Eric Wong Date: Tue Sep 28 18:16:53 2010 -0700 kgio 1.1.0 - flexible accept methods * an alternate class now be returned by accept/tryaccept by setting "Kgio.accept_class =" commit 911f6ab306aff1e24c9c570eeae33923fa1b99d9 Author: Eric Wong Date: Tue Sep 28 18:04:51 2010 -0700 alternate classes may be returned by accept/tryaccept These can be useful for avoiding wrapper objects and also allows users to more easily try different things without stepping on others' toe^H^H^Hclasses. commit 526b4bd48a20a34ef5959fdc4aa580d5f9199652 Author: Eric Wong Date: Mon Sep 27 19:59:34 2010 -0700 kgio 1.0.1 - compatibility fixes * add compatibility for ancient Rubies (1.8.6) * linux: fix accept4() support for newer Linux commit 20cbc0355104470fb433dd13e87a5d5c7e888ab1 Author: Eric Wong Date: Tue Sep 28 02:56:41 2010 +0000 linux: fix accept4() support for newer Linux Oops :x Tested on Debian sid. commit 24f1d168eb0937f0586c45b266bcd208431f0107 Author: Eric Wong Date: Mon Sep 27 18:06:34 2010 -0700 add compatibility for ancient Rubies This is tested on Ruby 1.8.6-p114, but may work for 1.8.5, too. Ugh, people ought to upgrade. commit e4d204c86e9420023ba3e4d8dbeb6b3fea8d6cf7 Author: Eric Wong Date: Tue Sep 28 00:27:44 2010 +0000 kgio 1.0.0 - initial release Documentation and release infrastructure updates and such... commit 8984b9556a3493570fbb4f747fce712d58f2cdd8 Author: Eric Wong Date: Tue Sep 28 00:07:43 2010 +0000 doc: TODO update commit 2c64a1fc07d3b9a80d112e3b0e2baa7ec29c2f47 Author: Eric Wong Date: Tue Sep 28 00:03:39 2010 +0000 read/write: account for buffer changes during wait It's possible for applications to modify the buffer during reads and writes, so make a best effort to account for those. commit f2ea9918655e8ee0576bee2950d16485031fc361 Author: Eric Wong Date: Mon Sep 27 23:59:59 2010 +0000 tests: fix broken monster trywrite test Oops, use random data so it's easier to detect this. commit 7abc0eb3dd804c2e65660b7dd9c828df0e03b80a Author: Eric Wong Date: Mon Sep 27 15:09:44 2010 -0700 test_tcp*read_write: use blocking kgio_accept in setup Some OSes (FreeBSD 7.0) do not seem to setup connections as quickly. commit 95d2eae6a4da34c504427af6ae0ab4c8c70c0ce5 Author: Eric Wong Date: Mon Sep 27 15:09:43 2010 -0700 set blocking flag before blocking IO#read Some older Rubies may not behave correctly otherwise commit 0806cac89f9d0e169b6c1e4da68c1ad66daa23ae Author: Eric Wong Date: Mon Sep 27 23:16:53 2010 +0000 tess: ensure buffer is cleared on failures No need to leak data. commit 50b86bf23063f3e6c3777b39c9464f73ccfd6ef5 Author: Eric Wong Date: Mon Sep 27 22:55:52 2010 +0000 more documentation Somebody's gotta do it... commit 5123d66fe0b2dad67539a20fe5b91f5b9afd814a Author: Eric Wong Date: Mon Sep 27 16:56:13 2010 +0000 avoid initiating syscalls before rb_io_wait_* Some Ruby implementations (Rubinius) may call lseek and clobber the intended errno when looking up the open file, causing rb_io_wait_* functions to fail. commit 6c818b0b6f76ef733679bcea1024142b4ef3ce00 Author: Eric Wong Date: Mon Sep 27 01:13:30 2010 +0000 add kgio_tryaccept, kgio_accept _really_ blocks We'll stick with the "try" prefix if we're going to be non-blocking. kgio_accept will favor a blocking accept() call where it's possible to release the GVL, allowing it to avoid thundering herd problems. Otherwise it'll use thread-safe blocking under Ruby 1.8. commit f81cb3c05a0eb46ec61ceb295b51ead16e6a0da4 Author: Eric Wong Date: Mon Sep 27 00:57:14 2010 +0000 use SOCK_NONBLOCK for socket(2) if possible This saves us a relatively expensive fcntl() system call. commit 87cf3ce6185b9138032a5af53cecae98f8c93564 Author: Eric Wong Date: Mon Sep 27 00:24:50 2010 +0000 connect: no do not leak descriptors on failure We cannot raise exceptions and expect GC to clean up after us until we've created an actual IO object. commit 6fbde1518578dd1b828efcecaf2caf893bddc110 Author: Eric Wong Date: Mon Sep 27 00:11:43 2010 +0000 "start" singleton methods for non-blocking connect These initiate (but do not wait for) non-blocking connects. commit fdfecc6d815bab8dfc1d8ad6758a66d44ab51e31 Author: Eric Wong Date: Sun Sep 26 07:51:12 2010 +0000 introduce kgio_try* methods Avoid altering behavior based on globals that Kgio.wait_{read,writ}able stored in, since that's too confusing. The non-try variants are closer to the normal IO read/write methods, except they can be more easily plugged into alternate reactors and event frameworks. commit d8ee79e1e5c6e6908009213324db25cf41c583ce Author: Eric Wong Date: Sat Sep 25 17:55:07 2010 +0000 kgio_read returns nil on EOF Just like IO#read commit af03e4471de3d3b91eec16e26e93a84d4a717116 Author: Eric Wong Date: Sat Sep 25 17:47:13 2010 +0000 split out reusable bits into separate headers No point in cluttering up the meat of our code. commit db53263856d864ba6273e6cac73011f699509d71 Author: Eric Wong Date: Sat Sep 25 01:36:13 2010 -0700 only use MSG_DONTWAIT under Linux MSG_DONTWAIT is less consistently implemented/supported on other platforms on stream sockets, so fallback to fcntl() + read()/write() for stream sockets. This also fixes our previously broken support of non-MSG_DONTWAIT systems. commit a82dc40c2a509c4ab692da34b572693f243fbfae Author: Eric Wong Date: Sat Sep 25 01:36:12 2010 -0700 write/send may fail with ECONNRESET Tested on FreeBSD 7.0 commit 0c60192621303f5e4ebd46d43a058de48126bc8a Author: Eric Wong Date: Sat Sep 25 01:36:11 2010 -0700 fix missing netinet/in.h include This is needed for FreeBSD 7.0, at least. commit 0beb82437f4ab0b8422e225080b234361092315e Author: Eric Wong Date: Sat Sep 25 08:15:13 2010 +0000 beef up the test suite We need to test server <-> client interaction more thoroughly since some systems don't implement everything right. commit 460e6b025896dee64b39d194d4c1a536129654de Author: Eric Wong Date: Thu Sep 23 22:56:44 2010 +0000 initial commit + release everything shou^Wmight be working...