ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/Changes
(Generate patch)

Comparing libev/Changes (file contents):
Revision 1.250 by root, Sat Feb 4 18:55:50 2012 UTC vs.
Revision 1.356 by root, Tue Jan 21 23:52:35 2020 UTC

1Revision history for libev, a high-performance and full-featured event loop. 1Revision history for libev, a high-performance and full-featured event loop.
2 2
3TODO: ev_loop_wakeup 3TODO: revisit 59.x timer in the light of modern powersaving
4TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone) 4TODO: ^ maybe just increase it when timerfd, modern linux only thing that counts?
5TODO: document EV_TSTAMP_T
6
7TODO: add ev_io_set_events, or ev_io_modify, or ev_io_change, or..., without IOFDSET
8 - the 4.31 timerfd code wrongly changes the priority of the signal
9 fd watcher, which is usually harmless unless signal fds are
10 also used (found via cpan tester service).
11 - do not wake up every minute when timerfd is used to detect timejumps.
12 - support a lot more "uncommon" compile time configurations,
13 such as ev_embed enabled but ev_timer disabled.
14 - the linux aio backend is no longer compiled in by default.
15 - update to libecb version 0x00010008.
16
174.31 Fri Dec 20 21:58:29 CET 2019
18 - handle backends with minimum wait time a bit better by not
19 waiting in the presence of already-expired timers
20 (behaviour reported by Felipe Gasper).
21 - new feature: use timerfd to detect timejumps quickly,
22 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
23 - document EV_USE_SIGNALFD feature macro.
24
254.30 (EV only)
26 - change non-autoconf test for __kernel_rwf_t by testing
27 LINUX_VERSION_CODE, the most direct test I could find.
28 - fix a bug in the io_uring backend that polled the wrong
29 backend fd, causing it to not work in many cases.
30
314.29 (EV only)
32 - add io uring autoconf and non-autoconf detection.
33 - disable io_uring when some header files are too old.
34
354.28 (EV only)
36 - linuxaio backend resulted in random memory corruption
37 when loop is forked.
38 - linuxaio backend might have tried to cancel an iocb
39 multiple times (was unable to trigger this).
40 - linuxaio backend now employs a generation counter to
41 avoid handling spurious events from cancelled requests.
42 - io_cancel can return EINTR, deal with it. also, assume
43 io_submit also returns EINTR.
44 - fix some other minor bugs in linuxaio backend.
45 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
46 - cleanup: replace expect_true/false and noinline by their
47 libecb counterparts.
48 - move syscall infrastructure from ev_linuxaio.c to ev.c.
49 - prepare io_uring integration.
50 - tweak ev_floor.
51 - epoll, poll, win32 Sleep and other places that use millisecond
52 reslution now all try to round up times.
53 - solaris port backend didn't compile.
54 - abstract time constants into their macros, for more flexibility.
55
564.27 Thu Jun 27 22:43:44 CEST 2019
57 - linux aio backend almost completely rewritten to work around its
58 limitations.
59 - linux aio backend now requires linux 4.19+.
60 - epoll backend now mandatory for linux aio backend.
61 - fail assertions more aggressively on invalid fd's detected
62 in the event loop, do not just silently fd_kill in case of
63 user error.
64 - ev_io_start/ev_io_stop now verify the watcher fd using
65 a syscall when EV_VERIFY is 2 or higher.
66
674.26 (EV only)
68 - update to libecb 0x00010006.
69 - new experimental linux aio backend (linux 4.18+).
70 - removed redundant 0-ptr check in ev_once.
71 - updated/extended ev_set_allocator documentation.
72 - replaced EMPTY2 macro by array_needsize_noinit.
73 - minor code cleanups.
74 - epoll backend now uses epoll_create1 also after fork.
75
764.25 Fri Dec 21 07:49:20 CET 2018
77 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
78 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
79 - move the darwin select workaround higher in ev.c, as newer versions of
80 darwin managed to break their broken select even more.
81 - ANDROID => __ANDROID__ (reported by enh@google.com).
82 - disable epoll_create1 on android because it has broken header files
83 and google is unwilling to fix them (reported by enh@google.com).
84 - avoid a minor compilation warning on win32.
85 - c++: remove deprecated dynamic throw() specifications.
86 - c++: improve the (unsupported) bad_loop exception class.
87 - backport perl ev_periodic example to C, untested.
88 - update libecb, biggets change is to include a memory fence
89 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
90 - minor autoconf/automake modernisation.
91
924.24 Wed Dec 28 05:19:55 CET 2016
93 - bump version to 4.24, as the release tarball inexplicably
94 didn't have the right version in ev.h, even though the cvs-tagged
95 version did have the right one (reported by Ales Teska).
96
974.23 Wed Nov 16 18:23:41 CET 2016
98 - move some declarations at the beginning to help certain retarded
99 microsoft compilers, even though their documentation claims
100 otherwise (reported by Ruslan Osmanov).
101
1024.22 Sun Dec 20 22:11:50 CET 2015
103 - when epoll detects unremovable fds in the fd set, rebuild
104 only the epoll descriptor, not the signal pipe, to avoid
105 SIGPIPE in ev_async_send. This doesn't solve it on fork,
106 so document what needs to be done in ev_loop_fork
107 (analyzed by Benjamin Mahler).
108 - remove superfluous sys/timeb.h include on win32
109 (analyzed by Jason Madden).
110 - updated libecb.
111
1124.20 Sat Jun 20 13:01:43 CEST 2015
113 - prefer noexcept over throw () with C++ 11.
114 - update ecb.h due to incompatibilities with c11.
115 - fix a potential aliasing issue when reading and writing
116 watcher callbacks.
117
1184.19 Thu Sep 25 08:18:25 CEST 2014
119 - ev.h wasn't valid C++ anymore, which tripped compilers other than
120 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
121 C++ doesn't support typedefs for function pointers fully, so the affected
122 declarations have to spell out the types each time.
123 - when not using autoconf, tighten the check for clock_gettime and related
124 functionality.
125
1264.18 Fri Sep 5 17:55:26 CEST 2014
127 - events on files were not always generated properly with the
128 epoll backend (testcase by Assaf Inbal).
129 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
130 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
131 - use a reasonable fallback for EV_NSIG instead of erroring out
132 when we can't detect the signal set size.
133 - in the absence of autoconf, do not use the clock syscall
134 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
135 doing clock_gettime in userspace).
136 - ensure extern "C" function pointers are used for externally-visible
137 loop callbacks (not watcher callbacks yet).
138 - (ecb) work around memory barriers and volatile apparently both being
139 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
140
1414.15 Fri Mar 1 12:04:50 CET 2013
142 - destroying a non-default loop would stop the global waitpid
143 watcher (Denis Bilenko).
144 - queueing pending watchers of higher priority from a watcher now invokes
145 them in a timely fashion (reported by Denis Bilenko).
146 - add throw() to all libev functions that cannot throw exceptions, for
147 further code size decrease when compiling for C++.
148 - add throw () to callbacks that must not throw exceptions (allocator,
149 syserr, loop acquire/release, periodic reschedule cbs).
150 - fix event_base_loop return code, add event_get_callback, event_base_new,
151 event_base_get_method calls to improve libevent 1.x emulation and add
152 some libevent 2.x functionality (based on a patch by Jeff Davey).
153 - add more memory fences to fix a bug reported by Jeff Davey. Better
154 be overfenced than underprotected.
155 - ev_run now returns a boolean status (true meaning watchers are
156 still active).
157 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
158 libev's EV_ERROR (reported by 191919).
159 - (ecb) add memory fence support for xlC (Darin McBride).
160 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
161 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
162 - work around some kernels losing file descriptors by leaking
163 the kqueue descriptor in the child.
164 - work around linux inotify not reporting IN_ATTRIB changes for directories
165 in many cases.
166 - include sys/syscall.h instead of plain syscall.h.
167 - check for io watcher loops in ev_verify, check for the most
168 common reported usage bug in ev_io_start.
169 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
170 - always use WSASend/WSARecv directly on windows, hoping that this
171 works in all cases (unlike read/write/send/recv...).
172 - try to detect signals around a fork faster (test program by
173 Denis Bilenko).
174 - work around recent glibc versions that leak memory in realloc.
175 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
176 the watcher base set (loop) method.
177 - rewrite the async/signal pipe logic to always keep a valid fd, which
178 simplifies (and hopefully correctifies :) the race checking
179 on fork, at the cost of one extra fd.
180 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
181 inotify-supporting filesystems.
182 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
183 versions ignore it before
184 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
185 - add some untested android support.
186 - enum expressions must be of type int (reported by Juan Pablo L).
5 187
64.11 Sat Feb 4 19:52:39 CET 2012 1884.11 Sat Feb 4 19:52:39 CET 2012
7 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as 189 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
8 was documented already, but not implemented in the repeating case. 190 was documented already, but not implemented in the repeating case.
9 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS. 191 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
… …
27 - fix #3 "be smart about timeouts" to not "deadlock" when 209 - fix #3 "be smart about timeouts" to not "deadlock" when
28 timeout == now, also improve the section overall. 210 timeout == now, also improve the section overall.
29 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom. 211 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
30 - support new EV_API_STATIC mode to make all libev symbols 212 - support new EV_API_STATIC mode to make all libev symbols
31 static. 213 static.
214 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
215 were empty.
32 216
334.04 Wed Feb 16 09:01:51 CET 2011 2174.04 Wed Feb 16 09:01:51 CET 2011
34 - fix two problems in the native win32 backend, where reuse of fd's 218 - fix two problems in the native win32 backend, where reuse of fd's
35 with different underlying handles caused handles not to be removed 219 with different underlying handles caused handles not to be removed
36 or added to the select set (analyzed and tested by Bert Belder). 220 or added to the select set (analyzed and tested by Bert Belder).
… …
191 reinitialising the default loop actually works (patch by 375 reinitialising the default loop actually works (patch by
192 Malek Hadj-Ali). 376 Malek Hadj-Ali).
193 - implement ev_suspend and ev_resume. 377 - implement ev_suspend and ev_resume.
194 - new EV_CUSTOM revents flag for use by applications. 378 - new EV_CUSTOM revents flag for use by applications.
195 - add documentation section about priorities. 379 - add documentation section about priorities.
196 - add a glossary to the dcoumentation. 380 - add a glossary to the documentation.
197 - extend the ev_fork description slightly. 381 - extend the ev_fork description slightly.
198 - optimize a jump out of call_pending. 382 - optimize a jump out of call_pending.
199 383
2003.53 Sun Feb 15 02:38:20 CET 2009 3843.53 Sun Feb 15 02:38:20 CET 2009
201 - fix a bug in event pipe creation on win32 that would cause a 385 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines