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

Comparing libev/Changes (file contents):
Revision 1.102 by root, Sat Mar 14 04:45:39 2009 UTC vs.
Revision 1.327 by root, Sun Jun 23 02:02:24 2019 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
3 - new experimental linux aio backend (linux 4.18+).
4 - removed redundant 0-ptr check in ev_once.
5 - updated/extended ev_set_allocator documentation.
6 - replaced EMPTY2 macro by array_needsize_noinit.
7 - minor code cleanups.
8 - epoll backend now uses epoll_create1 also after fork.
9
104.25 Fri Dec 21 07:49:20 CET 2018
11 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
12 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
13 - move the darwin select workaround highe rin ev.c, as newer versions of
14 darwin managed to break their broken select even more.
15 - ANDROID => __ANDROID__ (reported by enh@google.com).
16 - disable epoll_create1 on android because it has broken header files
17 and google is unwilling to fix them (reported by enh@google.com).
18 - avoid a minor compilation warning on win32.
19 - c++: remove deprecated dynamic throw() specifications.
20 - c++: improve the (unsupported) bad_loop exception class.
21 - backport perl ev_periodic example to C, untested.
22 - update libecb, biggets change is to include a memory fence
23 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
24 - minor autoconf/automake modernisation.
25
264.24 Wed Dec 28 05:19:55 CET 2016
27 - bump version to 4.24, as the release tarball inexplicably
28 didn't have the right version in ev.h, even though the cvs-tagged
29 version did have the right one (reported by Ales Teska).
30
314.23 Wed Nov 16 18:23:41 CET 2016
32 - move some declarations at the beginning to help certain retarded
33 microsoft compilers, even though their documentation claims
34 otherwise (reported by Ruslan Osmanov).
35
364.22 Sun Dec 20 22:11:50 CET 2015
37 - when epoll detects unremovable fds in the fd set, rebuild
38 only the epoll descriptor, not the signal pipe, to avoid
39 SIGPIPE in ev_async_send. This doesn't solve it on fork,
40 so document what needs to be done in ev_loop_fork
41 (analyzed by Benjamin Mahler).
42 - remove superfluous sys/timeb.h include on win32
43 (analyzed by Jason Madden).
44 - updated libecb.
45
464.20 Sat Jun 20 13:01:43 CEST 2015
47 - prefer noexcept over throw () with C++ 11.
48 - update ecb.h due to incompatibilities with c11.
49 - fix a potential aliasing issue when reading and writing
50 watcher callbacks.
51
524.19 Thu Sep 25 08:18:25 CEST 2014
53 - ev.h wasn't valid C++ anymore, which tripped compilers other than
54 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
55 C++ doesn't support typedefs for function pointers fully, so the affected
56 declarations have to spell out the types each time.
57 - when not using autoconf, tighten the check for clock_gettime and related
58 functionality.
59
604.18 Fri Sep 5 17:55:26 CEST 2014
61 - events on files were not always generated properly with the
62 epoll backend (testcase by Assaf Inbal).
63 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
64 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
65 - use a reasonable fallback for EV_NSIG instead of erroring out
66 when we can't detect the signal set size.
67 - in the absence of autoconf, do not use the clock syscall
68 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
69 doing clock_gettime in userspace).
70 - ensure extern "C" function pointers are used for externally-visible
71 loop callbacks (not watcher callbacks yet).
72 - (ecb) work around memory barriers and volatile apparently both being
73 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
74
754.15 Fri Mar 1 12:04:50 CET 2013
76 - destroying a non-default loop would stop the global waitpid
77 watcher (Denis Bilenko).
78 - queueing pending watchers of higher priority from a watcher now invokes
79 them in a timely fashion (reported by Denis Bilenko).
80 - add throw() to all libev functions that cannot throw exceptions, for
81 further code size decrease when compiling for C++.
82 - add throw () to callbacks that must not throw exceptions (allocator,
83 syserr, loop acquire/release, periodic reschedule cbs).
84 - fix event_base_loop return code, add event_get_callback, event_base_new,
85 event_base_get_method calls to improve libevent 1.x emulation and add
86 some libevent 2.x functionality (based on a patch by Jeff Davey).
87 - add more memory fences to fix a bug reported by Jeff Davey. Better
88 be overfenced than underprotected.
89 - ev_run now returns a boolean status (true meaning watchers are
90 still active).
91 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
92 libev's EV_ERROR (reported by 191919).
93 - (ecb) add memory fence support for xlC (Darin McBride).
94 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
95 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
96 - work around some kernels losing file descriptors by leaking
97 the kqueue descriptor in the child.
98 - work around linux inotify not reporting IN_ATTRIB changes for directories
99 in many cases.
100 - include sys/syscall.h instead of plain syscall.h.
101 - check for io watcher loops in ev_verify, check for the most
102 common reported usage bug in ev_io_start.
103 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
104 - always use WSASend/WSARecv directly on windows, hoping that this
105 works in all cases (unlike read/write/send/recv...).
106 - try to detect signals around a fork faster (test program by
107 Denis Bilenko).
108 - work around recent glibc versions that leak memory in realloc.
109 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
110 the watcher base set (loop) method.
111 - rewrite the async/signal pipe logic to always keep a valid fd, which
112 simplifies (and hopefully correctifies :) the race checking
113 on fork, at the cost of one extra fd.
114 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
115 inotify-supporting filesystems.
116 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
117 versions ignore it before
118 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
119 - add some untested android support.
120 - enum expressions must be of type int (reported by Juan Pablo L).
121
1224.11 Sat Feb 4 19:52:39 CET 2012
123 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
124 was documented already, but not implemented in the repeating case.
125 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
126 - fix a race where the workaround against the epoll fork bugs
127 caused signals to not be handled anymore.
128 - correct backend_fudge for most backends, and implement a windows
129 specific workaround to avoid looping because we call both
130 select and Sleep, both with different time resolutions.
131 - document range and guarantees of ev_sleep.
132 - document reasonable ranges for periodics interval and offset.
133 - rename backend_fudge to backend_mintime to avoid future confusion :)
134 - change the default periodic reschedule function to hopefully be more
135 exact and correct even in corner cases or in the far future.
136 - do not rely on -lm anymore: use it when available but use our
137 own floor () if it is missing. This should make it easier to embed,
138 as no external libraries are required.
139 - strategically import macros from libecb and mark rarely-used functions
140 as cache-cold (saving almost 2k code size on typical amd64 setups).
141 - add Symbols.ev and Symbols.event files, that were missing.
142 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
143 - fix #3 "be smart about timeouts" to not "deadlock" when
144 timeout == now, also improve the section overall.
145 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
146 - support new EV_API_STATIC mode to make all libev symbols
147 static.
148 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
149 were empty.
150
1514.04 Wed Feb 16 09:01:51 CET 2011
152 - fix two problems in the native win32 backend, where reuse of fd's
153 with different underlying handles caused handles not to be removed
154 or added to the select set (analyzed and tested by Bert Belder).
155 - do no rely on ceil() in ev_e?poll.c.
156 - backport libev to HP-UX versions before 11 v3.
157 - configure did not detect nanosleep and clock_gettime properly when
158 they are available in the libc (as opposed to -lrt).
159
1604.03 Tue Jan 11 14:37:25 CET 2011
161 - officially support polling files with all backends.
162 - support files, /dev/zero etc. the same way as select in the epoll
163 backend, by generating events on our own.
164 - ports backend: work around solaris bug 6874410 and many related ones
165 (EINTR, maybe more), with no performance loss (note that the solaris
166 bug report is actually wrong, reality is far more bizarre and broken
167 than that).
168 - define EV_READ/EV_WRITE as macros in event.h, as some programs use
169 #ifdef to test for them.
170 - new (experimental) function: ev_feed_signal.
171 - new (to become default) EVFLAG_NOSIGMASK flag.
172 - new EVBACKEND_MASK symbol.
173 - updated COMMON IDIOMS SECTION.
174
1754.01 Fri Nov 5 21:51:29 CET 2010
176 - automake fucked it up, apparently, --add-missing -f is not quite enough
177 to make it update its files, so 4.00 didn't install ev++.h and
178 event.h on make install. grrr.
179 - ev_loop(count|depth) didn't return anything (Robin Haberkorn).
180 - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
181 - use "(libev) " prefix for all libev error messages now.
182
1834.00 Mon Oct 25 12:32:12 CEST 2010
184 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
185 - ev_embed_stop did not correctly stop the watcher (very good
186 testcase by Vladimir Timofeev).
187 - ev_run will now always update the current loop time - it erroneously
188 didn't when idle watchers were active, causing timers not to fire.
189 - fix a bug where a timeout of zero caused the timer not to fire
190 in the libevent emulation (testcase by Péter Szabó).
191 - applied win32 fixes by Michael Lenaghan (also James Mansion).
192 - replace EV_MINIMAL by EV_FEATURES.
193 - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
194 seems the former is *much* faster than the latter.
195 - linux kernel version detection (for inotify bug workarounds)
196 did not work properly.
197 - reduce the number of spurious wake-ups with the ports backend.
198 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
199 - do async init within ev_async_start, not ev_async_set, which avoids
200 an API quirk where the set function must be called in the C++ API
201 even when there is nothing to set.
202 - add (undocumented) EV_ENABLE when adding events with kqueue,
203 this might help with OS X, which seems to need it despite documenting
204 not to need it (helpfully pointed out by Tilghman Lesher).
205 - do not use poll by default on freebsd, it's broken (what isn't
206 on freebsd...).
207 - allow to embed epoll on kernels >= 2.6.32.
208 - configure now prepends -O3, not appends it, so one can still
209 override it.
210 - ev.pod: greatly expanded the portability section, added a porting
211 section, a description of watcher states and made lots of minor fixes.
212 - disable poll backend on AIX, the poll header spams the namespace
213 and it's not worth working around dead platforms (reported
214 and analyzed by Aivars Kalvans).
215 - improve header file compatibility of the standalone eventfd code
216 in an obscure case.
217 - implement EV_AVOID_STDIO option.
218 - do not use sscanf to parse linux version number (smaller, faster,
219 no sscanf dependency).
220 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
221 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
222 - add section on accept() problems to the manpage.
223 - rename EV_TIMEOUT to EV_TIMER.
224 - rename ev_loop_count/depth/verify/loop/unloop.
225 - remove ev_default_destroy and ev_default_fork.
226 - switch to two-digit minor version.
227 - work around an apparent gentoo compiler bug.
228 - define _DARWIN_UNLIMITED_SELECT. just so.
229 - use enum instead of #define for most constants.
230 - improve compatibility to older C++ compilers.
231 - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now
232 default arguments when compiled as C++.
233 - enable automake dependency tracking.
234 - ev_loop_new no longer leaks memory when loop creation failed.
235 - new ev_cleanup watcher type.
236
2373.9 Thu Dec 31 07:59:59 CET 2009
238 - signalfd is no longer used by default and has to be requested
239 explicitly - this means that easy to catch bugs become hard to
240 catch race conditions, but the users have spoken.
241 - point out the unspecified signal mask in the documentation, and
242 that this is a race condition regardless of EV_SIGNALFD.
243 - backport inotify code to C89.
244 - inotify file descriptors could leak into child processes.
245 - ev_stat watchers could keep an erroneous extra ref on the loop,
246 preventing exit when unregistering all watchers (testcases
247 provided by ry@tinyclouds.org).
248 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
249 symbols to make it easier for apps to do their own fd management.
250 - support EV_IDLE_ENABLE being disabled in ev++.h
251 (patch by Didier Spezia).
252 - take advantage of inotify_init1, if available, to set cloexec/nonblock
253 on fd creation, to avoid races.
254 - the signal handling pipe wasn't always initialised under windows
255 (analysed by lekma).
256 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
257 signalfd.
258 - add missing string.h include (Denis F. Latypoff).
259 - only replace ev_stat.prev when we detect an actual difference,
260 so prev is (almost) always different to attr. this might
261 have caused the problems with 04_stat.t.
262 - add ev::timer->remaining () method to C++ API.
263
2643.8 Sun Aug 9 14:30:45 CEST 2009
265 - incompatible change: do not necessarily reset signal handler
266 to SIG_DFL when a sighandler is stopped.
267 - ev_default_destroy did not properly free or zero some members,
268 potentially causing crashes and memory corruption on repeated
269 ev_default_destroy/ev_default_loop calls.
270 - take advantage of signalfd on GNU/Linux systems.
271 - document that the signal mask might be in an unspecified
272 state when using libev's signal handling.
273 - take advantage of some GNU/Linux calls to set cloexec/nonblock
274 on fd creation, to avoid race conditions.
275
2763.7 Fri Jul 17 16:36:32 CEST 2009
277 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
278 instead of using a per-loop variable (bug caught by accident...).
279 - the ev_set_io_collect_interval interpretation has changed.
280 - add new functionality: ev_set_userdata, ev_userdata,
281 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
282 ev_invoke_pending, ev_pending_count, together with a long example
283 about thread locking.
284 - add ev_timer_remaining (as requested by Denis F. Latypoff).
285 - add ev_loop_depth.
286 - calling ev_unloop in fork/prepare watchers will no longer poll
287 for new events.
288 - Denis F. Latypoff corrected many typos in example code snippets.
289 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
290 check that the syscall number is available before trying to
291 use it (reported by ry@tinyclouds).
292 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
293 slightly higher accuracy.
294 - properly declare ev_loop_verify and ev_now_update even when
295 !EV_MULTIPLICITY.
296 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
297 - support EV_MINIMAL==2 for a reduced API.
298 - actually 0-initialise struct sigaction when installing signals.
299 - add section on hibernate and stopped processes to ev_timer docs.
300
3013.6 Tue Apr 28 02:49:30 CEST 2009
302 - multiple timers becoming ready within an event loop iteration
303 will be invoked in the "correct" order now.
304 - do not leave the event loop early just because we have no active
305 watchers, fixing a problem when embedding a kqueue loop
306 that has active kernel events but no registered watchers
307 (reported by blacksand blacksand).
3 - correctly zero the idx values for arrays, so destroying and 308 - correctly zero the idx values for arrays, so destroying and
4 reinitialising the default loop actually works (patch by 309 reinitialising the default loop actually works (patch by
5 Malek Hadj-Ali). 310 Malek Hadj-Ali).
311 - implement ev_suspend and ev_resume.
312 - new EV_CUSTOM revents flag for use by applications.
313 - add documentation section about priorities.
314 - add a glossary to the documentation.
315 - extend the ev_fork description slightly.
316 - optimize a jump out of call_pending.
6 317
73.53 Sun Feb 15 02:38:20 CET 2009 3183.53 Sun Feb 15 02:38:20 CET 2009
8 - fix a bug in event pipe creation on win32 that would cause a 319 - fix a bug in event pipe creation on win32 that would cause a
9 failed assertion on event loop creation (patch by Malek Hadj-Ali). 320 failed assertion on event loop creation (patch by Malek Hadj-Ali).
10 - probe for CLOCK_REALTIME support at runtime as well and fall 321 - probe for CLOCK_REALTIME support at runtime as well and fall
… …
84 do similar changes for event.c. 395 do similar changes for event.c.
85 396
863.45 Tue Oct 21 21:59:26 CEST 2008 3973.45 Tue Oct 21 21:59:26 CEST 2008
87 - disable inotify usage on linux <2.6.25, as it is broken 398 - disable inotify usage on linux <2.6.25, as it is broken
88 (reported by Yoann Vandoorselaere). 399 (reported by Yoann Vandoorselaere).
89 - ev_stat errornously would try to add inotify watchers 400 - ev_stat erroneously would try to add inotify watchers
90 even when inotify wasn't available (this should only 401 even when inotify wasn't available (this should only
91 have a performance impact). 402 have a performance impact).
92 - ev_once now passes both timeout and io to the callback if both 403 - ev_once now passes both timeout and io to the callback if both
93 occur concurrently, instead of giving timeouts precedence. 404 occur concurrently, instead of giving timeouts precedence.
94 - disable EV_USE_INOTIFY when sys/inotify.h is too old. 405 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
… …
162 - enable inlining with gcc versions 2 and 3. 473 - enable inlining with gcc versions 2 and 3.
163 - work around broken poll implementations potentially 474 - work around broken poll implementations potentially
164 not clearing revents field in ev_poll (Brandon Black) 475 not clearing revents field in ev_poll (Brandon Black)
165 (no such systems are known at this time). 476 (no such systems are known at this time).
166 - work around a bug in realloc on openbsd and darwin, 477 - work around a bug in realloc on openbsd and darwin,
167 also makes the errornous valgrind complaints 478 also makes the erroneous valgrind complaints
168 go away (noted by various people). 479 go away (noted by various people).
169 - fix ev_async_pending, add c++ wrapper for ev_async 480 - fix ev_async_pending, add c++ wrapper for ev_async
170 (based on patch sent by Johannes Deisenhofer. 481 (based on patch sent by Johannes Deisenhofer).
171 - add sensible set method to ev::embed. 482 - add sensible set method to ev::embed.
172 - made integer constants type int in ev.h. 483 - made integer constants type int in ev.h.
173 484
1743.2 Wed Apr 2 17:11:19 CEST 2008 4853.2 Wed Apr 2 17:11:19 CEST 2008
175 - fix a 64 bit overflow issue in the select backend, 486 - fix a 64 bit overflow issue in the select backend,

Diff Legend

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