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

Comparing libev/Changes (file contents):
Revision 1.226 by root, Sun Feb 20 02:56:55 2011 UTC vs.
Revision 1.361 by root, Wed Jan 22 13:33:44 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_prefork hint? 3TODO: for next ABI/API change, consider moving EV__IOFDSSET into io->fd instead and provide a getter.
4TODO: document EV_TSTAMP_T
4 5
6TODO: maybe allow events = 0 for io watchers
7 - the 4.31 timerfd code wrongly changes the priority of the signal
8 fd watcher, which is usually harmless unless signal fds are
9 also used (found via cpan tester service).
10 - the documentation wrongly claimed that user may modify fd and events
11 members in io watchers when the wqatcher was stopped
12 (found by b_jonas).
13 - new ev_io_modify mutator which changes only the events member,
14 which can be faster. also added ev::io::set (int events) method
15 to ev++.h.
16 - officially allow a zero events mask for io watchers. this should
17 work with older libev versions as well but was not officially
18 allowed before.
19 - do not wake up every minute when timerfd is used to detect timejumps.
20 - support a lot more "uncommon" compile time configurations,
21 such as ev_embed enabled but ev_timer disabled.
22 - use a start/stop wrapper class to reduce code duplication in
23 ev++.h and make it needlessly more c++-y.
24 - the linux aio backend is no longer compiled in by default.
25 - update to libecb version 0x00010008.
26
274.31 Fri Dec 20 21:58:29 CET 2019
28 - handle backends with minimum wait time a bit better by not
29 waiting in the presence of already-expired timers
30 (behaviour reported by Felipe Gasper).
31 - new feature: use timerfd to detect timejumps quickly,
32 can be disabled with the new EVFLAG_NOTIMERFD loop flag.
33 - document EV_USE_SIGNALFD feature macro.
34
354.30 (EV only)
36 - change non-autoconf test for __kernel_rwf_t by testing
37 LINUX_VERSION_CODE, the most direct test I could find.
38 - fix a bug in the io_uring backend that polled the wrong
39 backend fd, causing it to not work in many cases.
40
414.29 (EV only)
42 - add io uring autoconf and non-autoconf detection.
43 - disable io_uring when some header files are too old.
44
454.28 (EV only)
46 - linuxaio backend resulted in random memory corruption
47 when loop is forked.
48 - linuxaio backend might have tried to cancel an iocb
49 multiple times (was unable to trigger this).
50 - linuxaio backend now employs a generation counter to
51 avoid handling spurious events from cancelled requests.
52 - io_cancel can return EINTR, deal with it. also, assume
53 io_submit also returns EINTR.
54 - fix some other minor bugs in linuxaio backend.
55 - ev_tstamp type can now be overriden by defining EV_TSTAMP_T.
56 - cleanup: replace expect_true/false and noinline by their
57 libecb counterparts.
58 - move syscall infrastructure from ev_linuxaio.c to ev.c.
59 - prepare io_uring integration.
60 - tweak ev_floor.
61 - epoll, poll, win32 Sleep and other places that use millisecond
62 reslution now all try to round up times.
63 - solaris port backend didn't compile.
64 - abstract time constants into their macros, for more flexibility.
65
664.27 Thu Jun 27 22:43:44 CEST 2019
67 - linux aio backend almost completely rewritten to work around its
68 limitations.
69 - linux aio backend now requires linux 4.19+.
70 - epoll backend now mandatory for linux aio backend.
71 - fail assertions more aggressively on invalid fd's detected
72 in the event loop, do not just silently fd_kill in case of
73 user error.
74 - ev_io_start/ev_io_stop now verify the watcher fd using
75 a syscall when EV_VERIFY is 2 or higher.
76
774.26 (EV only)
78 - update to libecb 0x00010006.
79 - new experimental linux aio backend (linux 4.18+).
80 - removed redundant 0-ptr check in ev_once.
81 - updated/extended ev_set_allocator documentation.
82 - replaced EMPTY2 macro by array_needsize_noinit.
83 - minor code cleanups.
84 - epoll backend now uses epoll_create1 also after fork.
85
864.25 Fri Dec 21 07:49:20 CET 2018
87 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
88 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
89 - move the darwin select workaround higher in ev.c, as newer versions of
90 darwin managed to break their broken select even more.
91 - ANDROID => __ANDROID__ (reported by enh@google.com).
92 - disable epoll_create1 on android because it has broken header files
93 and google is unwilling to fix them (reported by enh@google.com).
94 - avoid a minor compilation warning on win32.
95 - c++: remove deprecated dynamic throw() specifications.
96 - c++: improve the (unsupported) bad_loop exception class.
97 - backport perl ev_periodic example to C, untested.
98 - update libecb, biggets change is to include a memory fence
99 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
100 - minor autoconf/automake modernisation.
101
1024.24 Wed Dec 28 05:19:55 CET 2016
103 - bump version to 4.24, as the release tarball inexplicably
104 didn't have the right version in ev.h, even though the cvs-tagged
105 version did have the right one (reported by Ales Teska).
106
1074.23 Wed Nov 16 18:23:41 CET 2016
108 - move some declarations at the beginning to help certain retarded
109 microsoft compilers, even though their documentation claims
110 otherwise (reported by Ruslan Osmanov).
111
1124.22 Sun Dec 20 22:11:50 CET 2015
113 - when epoll detects unremovable fds in the fd set, rebuild
114 only the epoll descriptor, not the signal pipe, to avoid
115 SIGPIPE in ev_async_send. This doesn't solve it on fork,
116 so document what needs to be done in ev_loop_fork
117 (analyzed by Benjamin Mahler).
118 - remove superfluous sys/timeb.h include on win32
119 (analyzed by Jason Madden).
120 - updated libecb.
121
1224.20 Sat Jun 20 13:01:43 CEST 2015
123 - prefer noexcept over throw () with C++ 11.
124 - update ecb.h due to incompatibilities with c11.
125 - fix a potential aliasing issue when reading and writing
126 watcher callbacks.
127
1284.19 Thu Sep 25 08:18:25 CEST 2014
129 - ev.h wasn't valid C++ anymore, which tripped compilers other than
130 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
131 C++ doesn't support typedefs for function pointers fully, so the affected
132 declarations have to spell out the types each time.
133 - when not using autoconf, tighten the check for clock_gettime and related
134 functionality.
135
1364.18 Fri Sep 5 17:55:26 CEST 2014
137 - events on files were not always generated properly with the
138 epoll backend (testcase by Assaf Inbal).
139 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
140 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
141 - use a reasonable fallback for EV_NSIG instead of erroring out
142 when we can't detect the signal set size.
143 - in the absence of autoconf, do not use the clock syscall
144 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
145 doing clock_gettime in userspace).
146 - ensure extern "C" function pointers are used for externally-visible
147 loop callbacks (not watcher callbacks yet).
148 - (ecb) work around memory barriers and volatile apparently both being
149 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
150
1514.15 Fri Mar 1 12:04:50 CET 2013
152 - destroying a non-default loop would stop the global waitpid
153 watcher (Denis Bilenko).
154 - queueing pending watchers of higher priority from a watcher now invokes
155 them in a timely fashion (reported by Denis Bilenko).
156 - add throw() to all libev functions that cannot throw exceptions, for
157 further code size decrease when compiling for C++.
158 - add throw () to callbacks that must not throw exceptions (allocator,
159 syserr, loop acquire/release, periodic reschedule cbs).
160 - fix event_base_loop return code, add event_get_callback, event_base_new,
161 event_base_get_method calls to improve libevent 1.x emulation and add
162 some libevent 2.x functionality (based on a patch by Jeff Davey).
163 - add more memory fences to fix a bug reported by Jeff Davey. Better
164 be overfenced than underprotected.
165 - ev_run now returns a boolean status (true meaning watchers are
166 still active).
167 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
168 libev's EV_ERROR (reported by 191919).
169 - (ecb) add memory fence support for xlC (Darin McBride).
170 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
171 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
172 - work around some kernels losing file descriptors by leaking
173 the kqueue descriptor in the child.
174 - work around linux inotify not reporting IN_ATTRIB changes for directories
175 in many cases.
176 - include sys/syscall.h instead of plain syscall.h.
177 - check for io watcher loops in ev_verify, check for the most
178 common reported usage bug in ev_io_start.
179 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
180 - always use WSASend/WSARecv directly on windows, hoping that this
181 works in all cases (unlike read/write/send/recv...).
182 - try to detect signals around a fork faster (test program by
183 Denis Bilenko).
184 - work around recent glibc versions that leak memory in realloc.
185 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
186 the watcher base set (loop) method.
187 - rewrite the async/signal pipe logic to always keep a valid fd, which
188 simplifies (and hopefully correctifies :) the race checking
189 on fork, at the cost of one extra fd.
190 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
191 inotify-supporting filesystems.
192 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
193 versions ignore it before
194 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
195 - add some untested android support.
196 - enum expressions must be of type int (reported by Juan Pablo L).
197
1984.11 Sat Feb 4 19:52:39 CET 2012
199 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
200 was documented already, but not implemented in the repeating case.
201 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
202 - fix a race where the workaround against the epoll fork bugs
203 caused signals to not be handled anymore.
204 - correct backend_fudge for most backends, and implement a windows
205 specific workaround to avoid looping because we call both
206 select and Sleep, both with different time resolutions.
207 - document range and guarantees of ev_sleep.
208 - document reasonable ranges for periodics interval and offset.
209 - rename backend_fudge to backend_mintime to avoid future confusion :)
5 - change the default periodic reschedule function to hopefully be more 210 - change the default periodic reschedule function to hopefully be more
6 exact and correct even in corner cases or in the far future. 211 exact and correct even in corner cases or in the far future.
7 - document reasonable ranges for interval and offset.
8 - do not rely on -lm anymore: use it when available but use our 212 - do not rely on -lm anymore: use it when available but use our
9 own floor () if it is missing. This should make it easier to embed. 213 own floor () if it is missing. This should make it easier to embed,
214 as no external libraries are required.
215 - strategically import macros from libecb and mark rarely-used functions
216 as cache-cold (saving almost 2k code size on typical amd64 setups).
217 - add Symbols.ev and Symbols.event files, that were missing.
218 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
219 - fix #3 "be smart about timeouts" to not "deadlock" when
220 timeout == now, also improve the section overall.
221 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
222 - support new EV_API_STATIC mode to make all libev symbols
223 static.
224 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
225 were empty.
10 226
114.04 Wed Feb 16 09:01:51 CET 2011 2274.04 Wed Feb 16 09:01:51 CET 2011
12 - fix two problems in the native win32 backend, where reuse of fd's 228 - fix two problems in the native win32 backend, where reuse of fd's
13 with different underlying handles caused handles not to be removed 229 with different underlying handles caused handles not to be removed
14 or added to the select set (analyzed and tested by Bert Belder). 230 or added to the select set (analyzed and tested by Bert Belder).
… …
100 catch race conditions, but the users have spoken. 316 catch race conditions, but the users have spoken.
101 - point out the unspecified signal mask in the documentation, and 317 - point out the unspecified signal mask in the documentation, and
102 that this is a race condition regardless of EV_SIGNALFD. 318 that this is a race condition regardless of EV_SIGNALFD.
103 - backport inotify code to C89. 319 - backport inotify code to C89.
104 - inotify file descriptors could leak into child processes. 320 - inotify file descriptors could leak into child processes.
105 - ev_stat watchers could keep an errornous extra ref on the loop, 321 - ev_stat watchers could keep an erroneous extra ref on the loop,
106 preventing exit when unregistering all watchers (testcases 322 preventing exit when unregistering all watchers (testcases
107 provided by ry@tinyclouds.org). 323 provided by ry@tinyclouds.org).
108 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration 324 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
109 symbols to make it easier for apps to do their own fd management. 325 symbols to make it easier for apps to do their own fd management.
110 - support EV_IDLE_ENABLE being disabled in ev++.h 326 - support EV_IDLE_ENABLE being disabled in ev++.h
… …
168 - correctly zero the idx values for arrays, so destroying and 384 - correctly zero the idx values for arrays, so destroying and
169 reinitialising the default loop actually works (patch by 385 reinitialising the default loop actually works (patch by
170 Malek Hadj-Ali). 386 Malek Hadj-Ali).
171 - implement ev_suspend and ev_resume. 387 - implement ev_suspend and ev_resume.
172 - new EV_CUSTOM revents flag for use by applications. 388 - new EV_CUSTOM revents flag for use by applications.
173 - add documentation section about priorites. 389 - add documentation section about priorities.
174 - add a glossary to the dcoumentation. 390 - add a glossary to the documentation.
175 - extend the ev_fork description slightly. 391 - extend the ev_fork description slightly.
176 - optimize a jump out of call_pending. 392 - optimize a jump out of call_pending.
177 393
1783.53 Sun Feb 15 02:38:20 CET 2009 3943.53 Sun Feb 15 02:38:20 CET 2009
179 - fix a bug in event pipe creation on win32 that would cause a 395 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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