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

Comparing libev/Changes (file contents):
Revision 1.54 by root, Thu Jun 19 06:58:34 2008 UTC vs.
Revision 1.147 by root, Wed Aug 26 17:46:22 2009 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 - backport inotify code to C89.
4 - inotify file descriptors were not properly recreated after a fork,
5 and could leak into child processes.
6 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
7 symbols to make it easier for apps to do their own fd management.
8 - support EV_IDLE_ENABLE being disabled in ev++.h
9 (patch by Didier Spezia).
10 - point out the unspecified signal mask in the documentation.
11 - take advantage of inotify_init1, if available, to set cloexec/nonblock
12 on fd creation, to avoid races.
13 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
14 signalfd.
15
163.8 Sun Aug 9 14:30:45 CEST 2009
17 - incompatible change: do not necessarily reset signal handler
18 to SIG_DFL when a sighandler is stopped.
19 - ev_default_destroy did not properly free or zero some members,
20 potentially causing crashes and memory corruption on repated
21 ev_default_destroy/ev_default_loop calls.
22 - take advantage of signalfd on GNU/Linux systems.
23 - document that the signal mask might be in an unspecified
24 state when using libev's signal handling.
25 - take advantage of some GNU/Linux calls to set cloexec/nonblock
26 on fd creation, to avoid race conditions.
27
283.7 Fri Jul 17 16:36:32 CEST 2009
29 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
30 instead of using a per-loop variable (bug caught by accident...).
31 - the ev_set_io_collect_interval interpretation has changed.
32 - add new functionality: ev_set_userdata, ev_userdata,
33 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
34 ev_invoke_pending, ev_pending_count, together with a long example
35 about thread locking.
36 - add ev_timer_remaining (as requested by Denis F. Latypoff).
37 - add ev_loop_depth.
38 - calling ev_unloop in fork/prepare watchers will no longer poll
39 for new events.
40 - Denis F. Latypoff corrected many typos in example code snippets.
41 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
42 check that the syscall number is available before trying to
43 use it (reported by ry@tinyclouds).
44 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
45 slightly higher accuracy.
46 - properly declare ev_loop_verify and ev_now_update even when
47 !EV_MULTIPLICITY.
48 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
49 - support EV_MINIMAL==2 for a reduced API.
50 - actually 0-initialise struct sigaction when installing signals.
51 - add section on hibernate and stopped processes to ev_timer docs.
52
533.6 Tue Apr 28 02:49:30 CEST 2009
54 - multiple timers becoming ready within an event loop iteration
55 will be invoked in the "correct" order now.
56 - do not leave the event loop early just because we have no active
57 watchers, fixing a problem when embedding a kqueue loop
58 that has active kernel events but no registered watchers
59 (reported by blacksand blacksand).
60 - correctly zero the idx values for arrays, so destroying and
61 reinitialising the default loop actually works (patch by
62 Malek Hadj-Ali).
63 - implement ev_suspend and ev_resume.
64 - new EV_CUSTOM revents flag for use by applications.
65 - add documentation section about priorites.
66 - add a glossary to the dcoumentation.
67 - extend the ev_fork description slightly.
68 - optimize a jump out of call_pending.
69
703.53 Sun Feb 15 02:38:20 CET 2009
71 - fix a bug in event pipe creation on win32 that would cause a
72 failed assertion on event loop creation (patch by Malek Hadj-Ali).
73 - probe for CLOCK_REALTIME support at runtime as well and fall
74 back to gettimeofday if there is an error, to support older
75 operating systems with newer header files/libraries.
76 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
77 (default most everywhere), otherwise not.
78
793.52 Wed Jan 7 21:43:02 CET 2009
80 - fix compilation of select backend in fd_set mode when NFDBITS is
81 missing (to get it to compile on QNX, reported by Rodrigo Campos).
82 - better select-nfds handling when select backend is in fd_set mode.
83 - diagnose fd_set overruns when select backend is in fd_set mode.
84 - due to a thinko, instead of disabling everything but
85 select on the borked OS X platform, everything but select was
86 allowed (reported by Emanuele Giaquinta).
87 - actually verify that local and remote port are matching in
88 libev's socketpair emulation, which makes denial-of-service
89 attacks harder (but not impossible - it's windows). Make sure
90 it even works under vista, which thinks that getpeer/sockname
91 should return fantasy port numbers.
92 - include "libev" in all assertion messages for potentially
93 clearer diagnostics.
94 - event_get_version (libevent compatibility) returned
95 a useless string instead of the expected version string
96 (patch by W.C.A. Wijngaards).
97
983.51 Wed Dec 24 23:00:11 CET 2008
99 - fix a bug where an inotify watcher was added twice, causing
100 freezes on hash collisions (reported and analysed by Graham Leggett).
101 - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
102 a direct syscall - slower, but no dependency on librt et al.
103 - assume negative return values != -1 signals success of port_getn
104 (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
105 (no known failure reports, but it doesn't hurt).
106 - fork detection in ev_embed now stops and restarts the watcher
107 automatically.
108 - EXPERIMENTAL: default the method to operator () in ev++.h,
109 to make it nicer to use functors (requested by Benedek László).
110 - fixed const object callbacks in ev++.h.
111 - replaced loop_ref argument of watcher.set (loop) by a direct
112 ev_loop * in ev++.h, to avoid clashes with functor patch.
113 - do not try to watch the empty string via inotify.
114 - inotify watchers could be leaked under certain circumstances.
115 - OS X 10.5 is actually even more broken than earlier versions,
116 so fall back to select on that piece of garbage.
117 - fixed some weirdness in the ev_embed documentation.
118
1193.49 Wed Nov 19 11:26:53 CET 2008
120 - ev_stat watchers will now use inotify as a mere hint on
121 kernels <2.6.25, or if the filesystem is not in the
122 "known to be good" list.
123 - better mingw32 compatibility (it's not as borked as native win32)
124 (analysed by Roger Pack).
125 - include stdio.h in the example program, as too many people are
126 confused by the weird C language otherwise. I guess the next thing
127 I get told is that the "..." ellipses in the examples don't compile
128 with their C compiler.
129
1303.48 Thu Oct 30 09:02:37 CET 2008
131 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
132 backend by assuming the kernel event mask hasn't changed if
133 ADD fails with EEXIST.
134 - work around spurious event notification bugs in epoll by using
135 a 32-bit generation counter. recreate kernel state if we receive
136 spurious notifications or unwanted events. this is very costly,
137 but I didn't come up with this horrible design.
138 - use memset to initialise most arrays now and do away with the
139 init functions.
140 - expand time-out strategies into a "Be smart about timeouts" section.
141 - drop the "struct" from all ev_watcher declarations in the
142 documentation and did other clarifications (yeah, it was a mistake
143 to have a struct AND a function called ev_loop).
144 - fix a bug where ev_default would not initialise the default
145 loop again after it was destroyed with ev_default_destroy.
146 - rename syserr to ev_syserr to avoid name clashes when embedding,
147 do similar changes for event.c.
148
1493.45 Tue Oct 21 21:59:26 CEST 2008
150 - disable inotify usage on linux <2.6.25, as it is broken
151 (reported by Yoann Vandoorselaere).
152 - ev_stat errornously would try to add inotify watchers
153 even when inotify wasn't available (this should only
154 have a performance impact).
155 - ev_once now passes both timeout and io to the callback if both
156 occur concurrently, instead of giving timeouts precedence.
157 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
158
1593.44 Mon Sep 29 05:18:39 CEST 2008
160 - embed watchers now automatically invoke ev_loop_fork on the
161 embedded loop when the parent loop forks.
162 - new function: ev_now_update (loop).
163 - verify_watcher was not marked static.
164 - improve the "associating..." manpage section.
165 - documentation tweaks here and there.
166
1673.43 Sun Jul 6 05:34:41 CEST 2008
3 - include more include files on windows to get struct _stati64 168 - include more include files on windows to get struct _stati64
4 (reported by Chris Hulbert). 169 (reported by Chris Hulbert, but doesn't quite fix his issue).
5 - add missing #include <io.h> in ev.c on windows (reported by 170 - add missing #include <io.h> in ev.c on windows (reported by
6 Matt Tolton). 171 Matt Tolton).
7 172
83.42 Tue Jun 17 12:12:07 CEST 2008 1733.42 Tue Jun 17 12:12:07 CEST 2008
9 - work around yet another windows bug: FD_SET actually adds fd's 174 - work around yet another windows bug: FD_SET actually adds fd's

Diff Legend

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