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

Comparing libev/Changes (file contents):
Revision 1.198 by root, Sun Oct 24 21:51:03 2010 UTC vs.
Revision 1.313 by root, Wed Dec 28 04:22:06 2016 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: include ev_xyz_start in each example? 34.24 Wed Dec 28 05:19:55 CET 2016
4TODO: which supports signals and child events 4 - bump version to 4.24, as the release tarball inexplicably
5TODO: pointer races 5 didn't have the right version in ev.h, even though the cvs-tagged
6TODO: ev_default_destroy, ev_default_fork 6 version did have the right one (reported by Ales Teska).
7
84.23 Wed Nov 16 18:23:41 CET 2016
9 - move some declarations at the beginning to help certain retarded
10 microsoft compilers, even though their documentation claims
11 otherwise (reported by Ruslan Osmanov).
12
134.22 Sun Dec 20 22:11:50 CET 2015
14 - when epoll detects unremovable fds in the fd set, rebuild
15 only the epoll descriptor, not the signal pipe, to avoid
16 SIGPIPE in ev_async_send. This doesn't solve it on fork,
17 so document what needs to be done in ev_loop_fork
18 (analyzed by Benjamin Mahler).
19 - remove superfluous sys/timeb.h include on win32
20 (analyzed by Jason Madden).
21 - updated libecb.
22
234.20 Sat Jun 20 13:01:43 CEST 2015
24 - prefer noexcept over throw () with C++ 11.
25 - update ecb.h due to incompatibilities with c11.
26 - fix a potential aliasing issue when reading and writing
27 watcher callbacks.
28
294.19 Thu Sep 25 08:18:25 CEST 2014
30 - ev.h wasn't valid C++ anymore, which tripped compilers other than
31 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
32 C++ doesn't support typedefs for function pointers fully, so the affected
33 declarations have to spell out the types each time.
34 - when not using autoconf, tighten the check for clock_gettime and related
35 functionality.
36
374.18 Fri Sep 5 17:55:26 CEST 2014
38 - events on files were not always generated properly with the
39 epoll backend (testcase by Assaf Inbal).
40 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
41 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
42 - use a reasonable fallback for EV_NSIG instead of erroring out
43 when we can't detect the signal set size.
44 - in the absence of autoconf, do not use the clock syscall
45 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
46 doing clock_gettime in userspace).
47 - ensure extern "C" function pointers are used for externally-visible
48 loop callbacks (not watcher callbacks yet).
49 - (ecb) work around memory barriers and volatile apparently both being
50 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
51
524.15 Fri Mar 1 12:04:50 CET 2013
53 - destroying a non-default loop would stop the global waitpid
54 watcher (Denis Bilenko).
55 - queueing pending watchers of higher priority from a watcher now invokes
56 them in a timely fashion (reported by Denis Bilenko).
57 - add throw() to all libev functions that cannot throw exceptions, for
58 further code size decrease when compiling for C++.
59 - add throw () to callbacks that must not throw exceptions (allocator,
60 syserr, loop acquire/release, periodic reschedule cbs).
61 - fix event_base_loop return code, add event_get_callback, event_base_new,
62 event_base_get_method calls to improve libevent 1.x emulation and add
63 some libevent 2.x functionality (based on a patch by Jeff Davey).
64 - add more memory fences to fix a bug reported by Jeff Davey. Better
65 be overfenced than underprotected.
66 - ev_run now returns a boolean status (true meaning watchers are
67 still active).
68 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
69 libev's EV_ERROR (reported by 191919).
70 - (ecb) add memory fence support for xlC (Darin McBride).
71 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
72 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
73 - work around some kernels losing file descriptors by leaking
74 the kqueue descriptor in the child.
75 - work around linux inotify not reporting IN_ATTRIB changes for directories
76 in many cases.
77 - include sys/syscall.h instead of plain syscall.h.
78 - check for io watcher loops in ev_verify, check for the most
79 common reported usage bug in ev_io_start.
80 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
81 - always use WSASend/WSARecv directly on windows, hoping that this
82 works in all cases (unlike read/write/send/recv...).
83 - try to detect signals around a fork faster (test program by
84 Denis Bilenko).
85 - work around recent glibc versions that leak memory in realloc.
86 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
87 the watcher base set (loop) method.
88 - rewrite the async/signal pipe logic to always keep a valid fd, which
89 simplifies (and hopefully correctifies :) the race checking
90 on fork, at the cost of one extra fd.
91 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
92 inotify-supporting filesystems.
93 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
94 versions ignore it before
95 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
96 - add some untested android support.
97 - enum expressions must be of type int (reported by Juan Pablo L).
98
994.11 Sat Feb 4 19:52:39 CET 2012
100 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
101 was documented already, but not implemented in the repeating case.
102 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
103 - fix a race where the workaround against the epoll fork bugs
104 caused signals to not be handled anymore.
105 - correct backend_fudge for most backends, and implement a windows
106 specific workaround to avoid looping because we call both
107 select and Sleep, both with different time resolutions.
108 - document range and guarantees of ev_sleep.
109 - document reasonable ranges for periodics interval and offset.
110 - rename backend_fudge to backend_mintime to avoid future confusion :)
111 - change the default periodic reschedule function to hopefully be more
112 exact and correct even in corner cases or in the far future.
113 - do not rely on -lm anymore: use it when available but use our
114 own floor () if it is missing. This should make it easier to embed,
115 as no external libraries are required.
116 - strategically import macros from libecb and mark rarely-used functions
117 as cache-cold (saving almost 2k code size on typical amd64 setups).
118 - add Symbols.ev and Symbols.event files, that were missing.
119 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
120 - fix #3 "be smart about timeouts" to not "deadlock" when
121 timeout == now, also improve the section overall.
122 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
123 - support new EV_API_STATIC mode to make all libev symbols
124 static.
125 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
126 were empty.
127
1284.04 Wed Feb 16 09:01:51 CET 2011
129 - fix two problems in the native win32 backend, where reuse of fd's
130 with different underlying handles caused handles not to be removed
131 or added to the select set (analyzed and tested by Bert Belder).
132 - do no rely on ceil() in ev_e?poll.c.
133 - backport libev to HP-UX versions before 11 v3.
134 - configure did not detect nanosleep and clock_gettime properly when
135 they are available in the libc (as opposed to -lrt).
136
1374.03 Tue Jan 11 14:37:25 CET 2011
138 - officially support polling files with all backends.
139 - support files, /dev/zero etc. the same way as select in the epoll
140 backend, by generating events on our own.
141 - ports backend: work around solaris bug 6874410 and many related ones
142 (EINTR, maybe more), with no performance loss (note that the solaris
143 bug report is actually wrong, reality is far more bizarre and broken
144 than that).
145 - define EV_READ/EV_WRITE as macros in event.h, as some programs use
146 #ifdef to test for them.
147 - new (experimental) function: ev_feed_signal.
148 - new (to become default) EVFLAG_NOSIGMASK flag.
149 - new EVBACKEND_MASK symbol.
150 - updated COMMON IDIOMS SECTION.
151
1524.01 Fri Nov 5 21:51:29 CET 2010
153 - automake fucked it up, apparently, --add-missing -f is not quite enough
154 to make it update its files, so 4.00 didn't install ev++.h and
155 event.h on make install. grrr.
156 - ev_loop(count|depth) didn't return anything (Robin Haberkorn).
157 - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
158 - use "(libev) " prefix for all libev error messages now.
159
1604.00 Mon Oct 25 12:32:12 CEST 2010
7 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading. 161 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
8 - ev_embed_stop did not correctly stop the watcher (very good 162 - ev_embed_stop did not correctly stop the watcher (very good
9 testcase by Vladimir Timofeev). 163 testcase by Vladimir Timofeev).
10 - ev_run will now always update the current loop time - it erroneously 164 - ev_run will now always update the current loop time - it erroneously
11 didn't when idle watchers were active, causing timers not to fire. 165 didn't when idle watchers were active, causing timers not to fire.
… …
63 catch race conditions, but the users have spoken. 217 catch race conditions, but the users have spoken.
64 - point out the unspecified signal mask in the documentation, and 218 - point out the unspecified signal mask in the documentation, and
65 that this is a race condition regardless of EV_SIGNALFD. 219 that this is a race condition regardless of EV_SIGNALFD.
66 - backport inotify code to C89. 220 - backport inotify code to C89.
67 - inotify file descriptors could leak into child processes. 221 - inotify file descriptors could leak into child processes.
68 - ev_stat watchers could keep an errornous extra ref on the loop, 222 - ev_stat watchers could keep an erroneous extra ref on the loop,
69 preventing exit when unregistering all watchers (testcases 223 preventing exit when unregistering all watchers (testcases
70 provided by ry@tinyclouds.org). 224 provided by ry@tinyclouds.org).
71 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration 225 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
72 symbols to make it easier for apps to do their own fd management. 226 symbols to make it easier for apps to do their own fd management.
73 - support EV_IDLE_ENABLE being disabled in ev++.h 227 - support EV_IDLE_ENABLE being disabled in ev++.h
… …
131 - correctly zero the idx values for arrays, so destroying and 285 - correctly zero the idx values for arrays, so destroying and
132 reinitialising the default loop actually works (patch by 286 reinitialising the default loop actually works (patch by
133 Malek Hadj-Ali). 287 Malek Hadj-Ali).
134 - implement ev_suspend and ev_resume. 288 - implement ev_suspend and ev_resume.
135 - new EV_CUSTOM revents flag for use by applications. 289 - new EV_CUSTOM revents flag for use by applications.
136 - add documentation section about priorites. 290 - add documentation section about priorities.
137 - add a glossary to the dcoumentation. 291 - add a glossary to the documentation.
138 - extend the ev_fork description slightly. 292 - extend the ev_fork description slightly.
139 - optimize a jump out of call_pending. 293 - optimize a jump out of call_pending.
140 294
1413.53 Sun Feb 15 02:38:20 CET 2009 2953.53 Sun Feb 15 02:38:20 CET 2009
142 - fix a bug in event pipe creation on win32 that would cause a 296 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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