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

Comparing libev/Changes (file contents):
Revision 1.208 by root, Wed Nov 10 14:36:42 2010 UTC vs.
Revision 1.324 by root, Sat Dec 29 14:23:20 2018 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: move some other examples to common idioms? combining watchers, 3 - removed redundant 0-ptr check in ev_once.
4thread usage, coroutine switch? 4
54.25 Fri Dec 21 07:49:20 CET 2018
6 - INCOMPATIBLE CHANGE: EV_THROW was renamed to EV_NOEXCEPT
7 (EV_THROW still provided) and now uses noexcept on C++11 or newer.
8 - move the darwin select workaround highe rin ev.c, as newer versions of
9 darwin managed to break their broken select even more.
10 - ANDROID => __ANDROID__ (reported by enh@google.com).
11 - disable epoll_create1 on android because it has broken header files
12 and google is unwilling to fix them (reported by enh@google.com).
13 - avoid a minor compilation warning on win32.
14 - c++: remove deprecated dynamic throw() specifications.
15 - c++: improve the (unsupported) bad_loop exception class.
16 - backport perl ev_periodic example to C, untested.
17 - update libecb, biggets change is to include a memory fence
18 in ECB_MEMORY_FENCE_RELEASE on x86/amd64.
19 - minor autoconf/automake modernisation.
20
214.24 Wed Dec 28 05:19:55 CET 2016
22 - bump version to 4.24, as the release tarball inexplicably
23 didn't have the right version in ev.h, even though the cvs-tagged
24 version did have the right one (reported by Ales Teska).
25
264.23 Wed Nov 16 18:23:41 CET 2016
27 - move some declarations at the beginning to help certain retarded
28 microsoft compilers, even though their documentation claims
29 otherwise (reported by Ruslan Osmanov).
30
314.22 Sun Dec 20 22:11:50 CET 2015
32 - when epoll detects unremovable fds in the fd set, rebuild
33 only the epoll descriptor, not the signal pipe, to avoid
34 SIGPIPE in ev_async_send. This doesn't solve it on fork,
35 so document what needs to be done in ev_loop_fork
36 (analyzed by Benjamin Mahler).
37 - remove superfluous sys/timeb.h include on win32
38 (analyzed by Jason Madden).
39 - updated libecb.
40
414.20 Sat Jun 20 13:01:43 CEST 2015
42 - prefer noexcept over throw () with C++ 11.
43 - update ecb.h due to incompatibilities with c11.
44 - fix a potential aliasing issue when reading and writing
45 watcher callbacks.
46
474.19 Thu Sep 25 08:18:25 CEST 2014
48 - ev.h wasn't valid C++ anymore, which tripped compilers other than
49 clang, msvc or gcc (analyzed by Raphael 'kena' Poss). Unfortunately,
50 C++ doesn't support typedefs for function pointers fully, so the affected
51 declarations have to spell out the types each time.
52 - when not using autoconf, tighten the check for clock_gettime and related
53 functionality.
54
554.18 Fri Sep 5 17:55:26 CEST 2014
56 - events on files were not always generated properly with the
57 epoll backend (testcase by Assaf Inbal).
58 - mark event pipe fd as cloexec after a fork (analyzed by Sami Farin).
59 - (ecb) support m68k, m88k and sh (patch by Miod Vallat).
60 - use a reasonable fallback for EV_NSIG instead of erroring out
61 when we can't detect the signal set size.
62 - in the absence of autoconf, do not use the clock syscall
63 on glibc >= 2.17 (avoids the syscall AND -lrt on systems
64 doing clock_gettime in userspace).
65 - ensure extern "C" function pointers are used for externally-visible
66 loop callbacks (not watcher callbacks yet).
67 - (ecb) work around memory barriers and volatile apparently both being
68 broken in visual studio 2008 and later (analysed and patch by Nicolas Noble).
69
704.15 Fri Mar 1 12:04:50 CET 2013
71 - destroying a non-default loop would stop the global waitpid
72 watcher (Denis Bilenko).
73 - queueing pending watchers of higher priority from a watcher now invokes
74 them in a timely fashion (reported by Denis Bilenko).
75 - add throw() to all libev functions that cannot throw exceptions, for
76 further code size decrease when compiling for C++.
77 - add throw () to callbacks that must not throw exceptions (allocator,
78 syserr, loop acquire/release, periodic reschedule cbs).
79 - fix event_base_loop return code, add event_get_callback, event_base_new,
80 event_base_get_method calls to improve libevent 1.x emulation and add
81 some libevent 2.x functionality (based on a patch by Jeff Davey).
82 - add more memory fences to fix a bug reported by Jeff Davey. Better
83 be overfenced than underprotected.
84 - ev_run now returns a boolean status (true meaning watchers are
85 still active).
86 - ev_once: undef EV_ERROR in ev_kqueue.c, to avoid clashing with
87 libev's EV_ERROR (reported by 191919).
88 - (ecb) add memory fence support for xlC (Darin McBride).
89 - (ecb) add memory fence support for gcc-mips (Anton Kirilov).
90 - (ecb) add memory fence support for gcc-alpha (Christian Weisgerber).
91 - work around some kernels losing file descriptors by leaking
92 the kqueue descriptor in the child.
93 - work around linux inotify not reporting IN_ATTRIB changes for directories
94 in many cases.
95 - include sys/syscall.h instead of plain syscall.h.
96 - check for io watcher loops in ev_verify, check for the most
97 common reported usage bug in ev_io_start.
98 - choose socket vs. WSASocket at compiletime using EV_USE_WSASOCKET.
99 - always use WSASend/WSARecv directly on windows, hoping that this
100 works in all cases (unlike read/write/send/recv...).
101 - try to detect signals around a fork faster (test program by
102 Denis Bilenko).
103 - work around recent glibc versions that leak memory in realloc.
104 - rename ev::embed::set to ev::embed::set_embed to avoid clashing
105 the watcher base set (loop) method.
106 - rewrite the async/signal pipe logic to always keep a valid fd, which
107 simplifies (and hopefully correctifies :) the race checking
108 on fork, at the cost of one extra fd.
109 - add fat, msdos, jffs2, ramfs, ntfs and btrfs to the list of
110 inotify-supporting filesystems.
111 - move orig_CFLAGS assignment to after AC_INIT, as newer autoconf
112 versions ignore it before
113 (https://bugzilla.redhat.com/show_bug.cgi?id=908096).
114 - add some untested android support.
115 - enum expressions must be of type int (reported by Juan Pablo L).
116
1174.11 Sat Feb 4 19:52:39 CET 2012
118 - INCOMPATIBLE CHANGE: ev_timer_again now clears the pending status, as
119 was documented already, but not implemented in the repeating case.
120 - new compiletime symbols: EV_NO_SMP and EV_NO_THREADS.
121 - fix a race where the workaround against the epoll fork bugs
122 caused signals to not be handled anymore.
123 - correct backend_fudge for most backends, and implement a windows
124 specific workaround to avoid looping because we call both
125 select and Sleep, both with different time resolutions.
126 - document range and guarantees of ev_sleep.
127 - document reasonable ranges for periodics interval and offset.
128 - rename backend_fudge to backend_mintime to avoid future confusion :)
129 - change the default periodic reschedule function to hopefully be more
130 exact and correct even in corner cases or in the far future.
131 - do not rely on -lm anymore: use it when available but use our
132 own floor () if it is missing. This should make it easier to embed,
133 as no external libraries are required.
134 - strategically import macros from libecb and mark rarely-used functions
135 as cache-cold (saving almost 2k code size on typical amd64 setups).
136 - add Symbols.ev and Symbols.event files, that were missing.
137 - fix backend_mintime value for epoll (was 1/1024, is 1/1000 now).
138 - fix #3 "be smart about timeouts" to not "deadlock" when
139 timeout == now, also improve the section overall.
140 - avoid "AVOIDING FINISHING BEFORE RETURNING" idiom.
141 - support new EV_API_STATIC mode to make all libev symbols
142 static.
143 - supply default CFLAGS of -g -O3 with gcc when original CFLAGS
144 were empty.
145
1464.04 Wed Feb 16 09:01:51 CET 2011
147 - fix two problems in the native win32 backend, where reuse of fd's
148 with different underlying handles caused handles not to be removed
149 or added to the select set (analyzed and tested by Bert Belder).
150 - do no rely on ceil() in ev_e?poll.c.
151 - backport libev to HP-UX versions before 11 v3.
152 - configure did not detect nanosleep and clock_gettime properly when
153 they are available in the libc (as opposed to -lrt).
154
1554.03 Tue Jan 11 14:37:25 CET 2011
156 - officially support polling files with all backends.
157 - support files, /dev/zero etc. the same way as select in the epoll
158 backend, by generating events on our own.
159 - ports backend: work around solaris bug 6874410 and many related ones
160 (EINTR, maybe more), with no performance loss (note that the solaris
161 bug report is actually wrong, reality is far more bizarre and broken
162 than that).
163 - define EV_READ/EV_WRITE as macros in event.h, as some programs use
164 #ifdef to test for them.
165 - new (experimental) function: ev_feed_signal.
166 - new (to become default) EVFLAG_NOSIGMASK flag.
167 - new EVBACKEND_MASK symbol.
168 - updated COMMON IDIOMS SECTION.
5 169
64.01 Fri Nov 5 21:51:29 CET 2010 1704.01 Fri Nov 5 21:51:29 CET 2010
7 - automake fucked it up, apparently, --add-missing -f is not quite enough 171 - automake fucked it up, apparently, --add-missing -f is not quite enough
8 to make it update its files, so 4.00 didn't install ev++.h and 172 to make it update its files, so 4.00 didn't install ev++.h and
9 event.h on make install. grrr. 173 event.h on make install. grrr.
… …
71 catch race conditions, but the users have spoken. 235 catch race conditions, but the users have spoken.
72 - point out the unspecified signal mask in the documentation, and 236 - point out the unspecified signal mask in the documentation, and
73 that this is a race condition regardless of EV_SIGNALFD. 237 that this is a race condition regardless of EV_SIGNALFD.
74 - backport inotify code to C89. 238 - backport inotify code to C89.
75 - inotify file descriptors could leak into child processes. 239 - inotify file descriptors could leak into child processes.
76 - ev_stat watchers could keep an errornous extra ref on the loop, 240 - ev_stat watchers could keep an erroneous extra ref on the loop,
77 preventing exit when unregistering all watchers (testcases 241 preventing exit when unregistering all watchers (testcases
78 provided by ry@tinyclouds.org). 242 provided by ry@tinyclouds.org).
79 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration 243 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
80 symbols to make it easier for apps to do their own fd management. 244 symbols to make it easier for apps to do their own fd management.
81 - support EV_IDLE_ENABLE being disabled in ev++.h 245 - support EV_IDLE_ENABLE being disabled in ev++.h
… …
139 - correctly zero the idx values for arrays, so destroying and 303 - correctly zero the idx values for arrays, so destroying and
140 reinitialising the default loop actually works (patch by 304 reinitialising the default loop actually works (patch by
141 Malek Hadj-Ali). 305 Malek Hadj-Ali).
142 - implement ev_suspend and ev_resume. 306 - implement ev_suspend and ev_resume.
143 - new EV_CUSTOM revents flag for use by applications. 307 - new EV_CUSTOM revents flag for use by applications.
144 - add documentation section about priorites. 308 - add documentation section about priorities.
145 - add a glossary to the dcoumentation. 309 - add a glossary to the documentation.
146 - extend the ev_fork description slightly. 310 - extend the ev_fork description slightly.
147 - optimize a jump out of call_pending. 311 - optimize a jump out of call_pending.
148 312
1493.53 Sun Feb 15 02:38:20 CET 2009 3133.53 Sun Feb 15 02:38:20 CET 2009
150 - fix a bug in event pipe creation on win32 that would cause a 314 - fix a bug in event pipe creation on win32 that would cause a

Diff Legend

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