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

Comparing libev/Changes (file contents):
Revision 1.176 by root, Mon Mar 29 12:40:57 2010 UTC vs.
Revision 1.302 by root, Mon Jan 19 16:45:19 2015 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: ABI??? API????? Changes??? 3TODO: ev_loop_wakeup
4TODO: win32 write() to socket for signal handling 4TODO: EV_STANDALONE == NO_HASSEL (do not use clock_gettime in ev_standalone)
5TODO: poll asert 5TODO: faq, process a thing in each iteration
6TODO: dbeugging tips, ev_verify, ev_init twice
7TODO: ev_break for immediate exit (EVBREAK_NOW?)
8TODO: ev_feed_child_event
9TODO: document the special problem of signals around fork.
10TODO: store pid for each signal
11TODO: document file descriptor usage per loop
12TODO: store loop pid_t and compare isndie signal handler,store 1 for same, 2 for differign pid, clean up in loop_fork
13TODO: embed watchers need updating when fd changes
14TODO: document portability requirements for atomic pointer access
15TODO: possible cb aliasing?
16TODO: document requirements for function pointers and calling conventions.
17
18 - prefer noexcept over throw () with C++ 11.
19
204.19 Thu Sep 25 08:18:25 CEST 2014
21 - ev.h wasn't valid C++ anymore, which tripped compilers other than
22 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
23 C++ doesn't support typedefs for function pointers fully, so the affected
24 declarations have to spell out the types each time.
25 - when not using autoconf, tighten the check for clock_gettime and related
26 functionality.
27
284.18 Fri Sep 5 17:55:26 CEST 2014
29 - events on files were not always generated properly with the
30 epoll backend (testcase by Assaf Inbal).
31 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
32 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
33 - use a reasonable fallback for EV_NSIG instead of erroring out
34 when we can't detect the signal set size.
35 - in the absence of autoconf, do not use the clock syscall
36 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
37 doing clock_gettime in userspace).
38 - ensure extern "C" function pointers are used for externally-visible
39 loop callbacks (not watcher callbacks yet).
40 - (ecb) work around memory barriers and volatile apparently both being
41 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
42
434.15 Fri Mar 1 12:04:50 CET 2013
44 - destroying a non-default loop would stop the global waitpid
45 watcher (Denis Bilenko).
46 - queueing pending watchers of higher priority from a watcher now invokes
47 them in a timely fashion (reported by Denis Bilenko).
48 - add throw() to all libev functions that cannot throw exceptions, for
49 further code size decrease when compiling for C++.
50 - add throw () to callbacks that must not throw exceptions (allocator,
51 syserr, loop acquire/release, periodic reschedule cbs).
52 - fix event_base_loop return code, add event_get_callback, event_base_new,
53 event_base_get_method calls to improve libevent 1.x emulation and add
54 some libevent 2.x functionality (based on a patch by Jeff Davey).
55 - add more memory fences to fix a bug reported by Jeff Davey. Better
56 be overfenced than underprotected.
57 - ev_run now returns a boolean status (true meaning watchers are
58 still active).
59 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
60 libev's EV_ERROR (reported by 191919).
61 - (ecb) add memory fence support for xlC (Darin McBride).
62 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
63 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
64 - work around some kernels losing file descriptors by leaking
65 the kqueue descriptor in the child.
66 - work around linux inotify not reporting IN_ATTRIB changes for directories
67 in many cases.
68 - include sys/syscall.h instead of plain syscall.h.
69 - check for io watcher loops in ev_verify, check for the most
70 common reported usage bug in ev_io_start.
71 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
72 - always use WSASend/WSARecv directly on windows, hoping that this
73 works in all cases (unlike read/write/send/recv...).
74 - try to detect signals around a fork faster (test program by
75 Denis Bilenko).
76 - work around recent glibc versions that leak memory in realloc.
77 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
78 the watcher base set (loop) method.
79 - rewrite the async/signal pipe logic to always keep a valid fd, which
80 simplifies (and hopefully correctifies :) the race checking
81 on fork, at the cost of one extra fd.
82 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
83 inotify-supporting filesystems.
84 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
85 versions ignore it before
86 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
87 - add some untested android support.
88 - enum expressions must be of type int (reported by Juan Pablo L).
89
904.11 Sat Feb 4 19:52:39 CET 2012
91 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
92 was documented already, but not implemented in the repeating case.
93 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
94 - fix a race where the workaround against the epoll fork bugs
95 caused signals to not be handled anymore.
96 - correct backend_fudge for most backends, and implement a windows
97 specific workaround to avoid looping because we call both
98 select and Sleep, both with different time resolutions.
99 - document range and guarantees of ev_sleep.
100 - document reasonable ranges for periodics interval and offset.
101 - rename backend_fudge to backend_mintime to avoid future confusion :)
102 - change the default periodic reschedule function to hopefully be more
103 exact and correct even in corner cases or in the far future.
104 - do not rely on -lm anymore: use it when available but use our
105 own floor () if it is missing. This should make it easier to embed,
106 as no external libraries are required.
107 - strategically import macros from libecb and mark rarely-used functions
108 as cache-cold (saving almost 2k code size on typical amd64 setups).
109 - add Symbols.ev and Symbols.event files, that were missing.
110 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
111 - fix #3 "be smart about timeouts" to not "deadlock" when
112 timeout == now, also improve the section overall.
113 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
114 - support new EV_API_STATIC mode to make all libev symbols
115 static.
116 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
117 were empty.
118
1194.04 Wed Feb 16 09:01:51 CET 2011
120 - fix two problems in the native win32 backend, where reuse of fd's
121 with different underlying handles caused handles not to be removed
122 or added to the select set (analyzed and tested by Bert Belder).
123 - do no rely on ceil() in ev_e?poll.c.
124 - backport libev to HP-UX versions before 11 v3.
125 - configure did not detect nanosleep and clock_gettime properly when
126 they are available in the libc (as opposed to -lrt).
127
1284.03 Tue Jan 11 14:37:25 CET 2011
129 - officially support polling files with all backends.
130 - support files, /dev/zero etc. the same way as select in the epoll
131 backend, by generating events on our own.
132 - ports backend: work around solaris bug 6874410 and many related ones
133 (EINTR, maybe more), with no performance loss (note that the solaris
134 bug report is actually wrong, reality is far more bizarre and broken
135 than that).
136 - define EV_READ/EV_WRITE as macros in event.h, as some programs use
137 #ifdef to test for them.
138 - new (experimental) function: ev_feed_signal.
139 - new (to become default) EVFLAG_NOSIGMASK flag.
140 - new EVBACKEND_MASK symbol.
141 - updated COMMON IDIOMS SECTION.
142
1434.01 Fri Nov 5 21:51:29 CET 2010
144 - automake fucked it up, apparently, --add-missing -f is not quite enough
145 to make it update its files, so 4.00 didn't install ev++.h and
146 event.h on make install. grrr.
147 - ev_loop(count|depth) didn't return anything (Robin Haberkorn).
148 - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
149 - use "(libev) " prefix for all libev error messages now.
150
1514.00 Mon Oct 25 12:32:12 CEST 2010
6 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading. 152 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
7 - ev_embed_stop did not correctly stop the watcher (very good 153 - ev_embed_stop did not correctly stop the watcher (very good
8 testcase by Vladimir Timofeev). 154 testcase by Vladimir Timofeev).
155 - ev_run will now always update the current loop time - it erroneously
156 didn't when idle watchers were active, causing timers not to fire.
157 - fix a bug where a timeout of zero caused the timer not to fire
158 in the libevent emulation (testcase by Péter Szabó).
9 - applied win32 fixes by Michael Lenaghan (also James Mansion). 159 - applied win32 fixes by Michael Lenaghan (also James Mansion).
10 - replace EV_MINIMAL by EV_FEATURES. 160 - replace EV_MINIMAL by EV_FEATURES.
161 - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
162 seems the former is *much* faster than the latter.
163 - linux kernel version detection (for inotify bug workarounds)
164 did not work properly.
165 - reduce the number of spurious wake-ups with the ports backend.
11 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu). 166 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
167 - do async init within ev_async_start, not ev_async_set, which avoids
168 an API quirk where the set function must be called in the C++ API
169 even when there is nothing to set.
12 - add (undocumented) EV_ENABLE when adding events with kqueue, 170 - add (undocumented) EV_ENABLE when adding events with kqueue,
13 this might help with OS X, which seems to need it despite documenting 171 this might help with OS X, which seems to need it despite documenting
14 not to need it (helpfully pointed out by Tilghman Lesher). 172 not to need it (helpfully pointed out by Tilghman Lesher).
15 - do not use poll by default on freebsd, it's broken (what isn't 173 - do not use poll by default on freebsd, it's broken (what isn't
16 on freebsd...). 174 on freebsd...).
175 - allow to embed epoll on kernels >= 2.6.32.
17 - configure now prepends -O3, not appends it, so one can still 176 - configure now prepends -O3, not appends it, so one can still
18 override it. 177 override it.
178 - ev.pod: greatly expanded the portability section, added a porting
179 section, a description of watcher states and made lots of minor fixes.
19 - disable poll backend on AIX, the poll header spams the namespace 180 - disable poll backend on AIX, the poll header spams the namespace
20 and it's not worth working around dead platforms (reported 181 and it's not worth working around dead platforms (reported
21 and analyzed by Aivars Kalvans). 182 and analyzed by Aivars Kalvans).
22 - improve header file compatibility of the standalone eventfd code 183 - improve header file compatibility of the standalone eventfd code
23 in an obscure case. 184 in an obscure case.
… …
26 no sscanf dependency). 187 no sscanf dependency).
27 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings. 188 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
28 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs. 189 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
29 - add section on accept() problems to the manpage. 190 - add section on accept() problems to the manpage.
30 - rename EV_TIMEOUT to EV_TIMER. 191 - rename EV_TIMEOUT to EV_TIMER.
31 - rename ev_loop_count/depth/verify. 192 - rename ev_loop_count/depth/verify/loop/unloop.
193 - remove ev_default_destroy and ev_default_fork.
32 - switch to two-digit minor version. 194 - switch to two-digit minor version.
195 - work around an apparent gentoo compiler bug.
196 - define _DARWIN_UNLIMITED_SELECT. just so.
197 - use enum instead of #define for most constants.
198 - improve compatibility to older C++ compilers.
199 - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now
200 default arguments when compiled as C++.
201 - enable automake dependency tracking.
202 - ev_loop_new no longer leaks memory when loop creation failed.
203 - new ev_cleanup watcher type.
33 204
343.9 Thu Dec 31 07:59:59 CET 2009 2053.9 Thu Dec 31 07:59:59 CET 2009
35 - signalfd is no longer used by default and has to be requested 206 - signalfd is no longer used by default and has to be requested
36 explicitly - this means that easy to catch bugs become hard to 207 explicitly - this means that easy to catch bugs become hard to
37 catch race conditions, but the users have spoken. 208 catch race conditions, but the users have spoken.
38 - point out the unspecified signal mask in the documentation, and 209 - point out the unspecified signal mask in the documentation, and
39 that this is a race condition regardless of EV_SIGNALFD. 210 that this is a race condition regardless of EV_SIGNALFD.
40 - backport inotify code to C89. 211 - backport inotify code to C89.
41 - inotify file descriptors could leak into child processes. 212 - inotify file descriptors could leak into child processes.
42 - ev_stat watchers could keep an errornous extra ref on the loop, 213 - ev_stat watchers could keep an erroneous extra ref on the loop,
43 preventing exit when unregistering all watchers (testcases 214 preventing exit when unregistering all watchers (testcases
44 provided by ry@tinyclouds.org). 215 provided by ry@tinyclouds.org).
45 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration 216 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
46 symbols to make it easier for apps to do their own fd management. 217 symbols to make it easier for apps to do their own fd management.
47 - support EV_IDLE_ENABLE being disabled in ev++.h 218 - support EV_IDLE_ENABLE being disabled in ev++.h
… …
53 - changed minimum glibc requirement from glibc 2.9 to 2.7, for 224 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
54 signalfd. 225 signalfd.
55 - add missing string.h include (Denis F. Latypoff). 226 - add missing string.h include (Denis F. Latypoff).
56 - only replace ev_stat.prev when we detect an actual difference, 227 - only replace ev_stat.prev when we detect an actual difference,
57 so prev is (almost) always different to attr. this might 228 so prev is (almost) always different to attr. this might
58 have caused the probems with 04_stat.t. 229 have caused the problems with 04_stat.t.
59 - add ev::timer->remaining () method to C++ API. 230 - add ev::timer->remaining () method to C++ API.
60 231
613.8 Sun Aug 9 14:30:45 CEST 2009 2323.8 Sun Aug 9 14:30:45 CEST 2009
62 - incompatible change: do not necessarily reset signal handler 233 - incompatible change: do not necessarily reset signal handler
63 to SIG_DFL when a sighandler is stopped. 234 to SIG_DFL when a sighandler is stopped.
64 - ev_default_destroy did not properly free or zero some members, 235 - ev_default_destroy did not properly free or zero some members,
65 potentially causing crashes and memory corruption on repated 236 potentially causing crashes and memory corruption on repeated
66 ev_default_destroy/ev_default_loop calls. 237 ev_default_destroy/ev_default_loop calls.
67 - take advantage of signalfd on GNU/Linux systems. 238 - take advantage of signalfd on GNU/Linux systems.
68 - document that the signal mask might be in an unspecified 239 - document that the signal mask might be in an unspecified
69 state when using libev's signal handling. 240 state when using libev's signal handling.
70 - take advantage of some GNU/Linux calls to set cloexec/nonblock 241 - take advantage of some GNU/Linux calls to set cloexec/nonblock
… …
105 - correctly zero the idx values for arrays, so destroying and 276 - correctly zero the idx values for arrays, so destroying and
106 reinitialising the default loop actually works (patch by 277 reinitialising the default loop actually works (patch by
107 Malek Hadj-Ali). 278 Malek Hadj-Ali).
108 - implement ev_suspend and ev_resume. 279 - implement ev_suspend and ev_resume.
109 - new EV_CUSTOM revents flag for use by applications. 280 - new EV_CUSTOM revents flag for use by applications.
110 - add documentation section about priorites. 281 - add documentation section about priorities.
111 - add a glossary to the dcoumentation. 282 - add a glossary to the documentation.
112 - extend the ev_fork description slightly. 283 - extend the ev_fork description slightly.
113 - optimize a jump out of call_pending. 284 - optimize a jump out of call_pending.
114 285
1153.53 Sun Feb 15 02:38:20 CET 2009 2863.53 Sun Feb 15 02:38:20 CET 2009
116 - fix a bug in event pipe creation on win32 that would cause a 287 - fix a bug in event pipe creation on win32 that would cause a
… …
192 do similar changes for event.c. 363 do similar changes for event.c.
193 364
1943.45 Tue Oct 21 21:59:26 CEST 2008 3653.45 Tue Oct 21 21:59:26 CEST 2008
195 - disable inotify usage on linux <2.6.25, as it is broken 366 - disable inotify usage on linux <2.6.25, as it is broken
196 (reported by Yoann Vandoorselaere). 367 (reported by Yoann Vandoorselaere).
197 - ev_stat errornously would try to add inotify watchers 368 - ev_stat erroneously would try to add inotify watchers
198 even when inotify wasn't available (this should only 369 even when inotify wasn't available (this should only
199 have a performance impact). 370 have a performance impact).
200 - ev_once now passes both timeout and io to the callback if both 371 - ev_once now passes both timeout and io to the callback if both
201 occur concurrently, instead of giving timeouts precedence. 372 occur concurrently, instead of giving timeouts precedence.
202 - disable EV_USE_INOTIFY when sys/inotify.h is too old. 373 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
… …
270 - enable inlining with gcc versions 2 and 3. 441 - enable inlining with gcc versions 2 and 3.
271 - work around broken poll implementations potentially 442 - work around broken poll implementations potentially
272 not clearing revents field in ev_poll (Brandon Black) 443 not clearing revents field in ev_poll (Brandon Black)
273 (no such systems are known at this time). 444 (no such systems are known at this time).
274 - work around a bug in realloc on openbsd and darwin, 445 - work around a bug in realloc on openbsd and darwin,
275 also makes the errornous valgrind complaints 446 also makes the erroneous valgrind complaints
276 go away (noted by various people). 447 go away (noted by various people).
277 - fix ev_async_pending, add c++ wrapper for ev_async 448 - fix ev_async_pending, add c++ wrapper for ev_async
278 (based on patch sent by Johannes Deisenhofer. 449 (based on patch sent by Johannes Deisenhofer).
279 - add sensible set method to ev::embed. 450 - add sensible set method to ev::embed.
280 - made integer constants type int in ev.h. 451 - made integer constants type int in ev.h.
281 452
2823.2 Wed Apr 2 17:11:19 CEST 2008 4533.2 Wed Apr 2 17:11:19 CEST 2008
283 - fix a 64 bit overflow issue in the select backend, 454 - fix a 64 bit overflow issue in the select backend,

Diff Legend

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