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

Comparing libev/Changes (file contents):
Revision 1.75 by root, Mon Oct 27 11:08:29 2008 UTC vs.
Revision 1.192 by root, Fri Oct 22 05:57:55 2010 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
3WISH? monotonic clocks times/GetTickCount for coarse corrections? 3TODO: include ev_xyz_start in each example?
4 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
5 - ev_embed_stop did not correctly stop the watcher (very good
6 testcase by Vladimir Timofeev).
7 - ev_run will now always update the current loop time - it erroneously
8 didn't when idle watchers were active, causing timers not to fire.
9 - fix a bug where a timeout of zero caused the timer not to fire
10 in the libevent emulation (testcase by Péter Szabó).
11 - applied win32 fixes by Michael Lenaghan (also James Mansion).
12 - replace EV_MINIMAL by EV_FEATURES.
13 - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
14 seems the former is *much* faster than the latter.
15 - reduce the number of spurious wake-ups with the ports backend.
16 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
17 - do async init within ev_async_start, not ev_async_set, which avoids
18 an API quirk where the set function must be called in the C++ API
19 even when there is nothing to set.
20 - add (undocumented) EV_ENABLE when adding events with kqueue,
21 this might help with OS X, which seems to need it despite documenting
22 not to need it (helpfully pointed out by Tilghman Lesher).
23 - do not use poll by default on freebsd, it's broken (what isn't
24 on freebsd...).
25 - configure now prepends -O3, not appends it, so one can still
26 override it.
27 - ev.pod: greatly expanded the portability section, added a porting
28 section, a description of watcher states and made lots of minor fixes.
29 - disable poll backend on AIX, the poll header spams the namespace
30 and it's not worth working around dead platforms (reported
31 and analyzed by Aivars Kalvans).
32 - improve header file compatibility of the standalone eventfd code
33 in an obscure case.
34 - implement EV_AVOID_STDIO option.
35 - do not use sscanf to parse linux version number (smaller, faster,
36 no sscanf dependency).
37 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
38 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
39 - add section on accept() problems to the manpage.
40 - rename EV_TIMEOUT to EV_TIMER.
41 - rename ev_loop_count/depth/verify/loop/unloop.
42 - switch to two-digit minor version.
43 - work around an apparent gentoo compiler bug.
44 - define _DARWIN_UNLIMITED_SELECT. just so.
45 - use enum instead of #define for most constants.
46 - improve compatibility to older C++ compilers.
47 - ev_run/ev_default_loop/ev_break/ev_loop_new have now default
48 arguments when compiled as C++.
4 49
503.9 Thu Dec 31 07:59:59 CET 2009
51 - signalfd is no longer used by default and has to be requested
52 explicitly - this means that easy to catch bugs become hard to
53 catch race conditions, but the users have spoken.
54 - point out the unspecified signal mask in the documentation, and
55 that this is a race condition regardless of EV_SIGNALFD.
56 - backport inotify code to C89.
57 - inotify file descriptors could leak into child processes.
58 - ev_stat watchers could keep an errornous extra ref on the loop,
59 preventing exit when unregistering all watchers (testcases
60 provided by ry@tinyclouds.org).
61 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
62 symbols to make it easier for apps to do their own fd management.
63 - support EV_IDLE_ENABLE being disabled in ev++.h
64 (patch by Didier Spezia).
65 - take advantage of inotify_init1, if available, to set cloexec/nonblock
66 on fd creation, to avoid races.
67 - the signal handling pipe wasn't always initialised under windows
68 (analysed by lekma).
69 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
70 signalfd.
71 - add missing string.h include (Denis F. Latypoff).
72 - only replace ev_stat.prev when we detect an actual difference,
73 so prev is (almost) always different to attr. this might
74 have caused the problems with 04_stat.t.
75 - add ev::timer->remaining () method to C++ API.
76
773.8 Sun Aug 9 14:30:45 CEST 2009
78 - incompatible change: do not necessarily reset signal handler
79 to SIG_DFL when a sighandler is stopped.
80 - ev_default_destroy did not properly free or zero some members,
81 potentially causing crashes and memory corruption on repeated
82 ev_default_destroy/ev_default_loop calls.
83 - take advantage of signalfd on GNU/Linux systems.
84 - document that the signal mask might be in an unspecified
85 state when using libev's signal handling.
86 - take advantage of some GNU/Linux calls to set cloexec/nonblock
87 on fd creation, to avoid race conditions.
88
893.7 Fri Jul 17 16:36:32 CEST 2009
90 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
91 instead of using a per-loop variable (bug caught by accident...).
92 - the ev_set_io_collect_interval interpretation has changed.
93 - add new functionality: ev_set_userdata, ev_userdata,
94 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
95 ev_invoke_pending, ev_pending_count, together with a long example
96 about thread locking.
97 - add ev_timer_remaining (as requested by Denis F. Latypoff).
98 - add ev_loop_depth.
99 - calling ev_unloop in fork/prepare watchers will no longer poll
100 for new events.
101 - Denis F. Latypoff corrected many typos in example code snippets.
102 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
103 check that the syscall number is available before trying to
104 use it (reported by ry@tinyclouds).
105 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
106 slightly higher accuracy.
107 - properly declare ev_loop_verify and ev_now_update even when
108 !EV_MULTIPLICITY.
109 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
110 - support EV_MINIMAL==2 for a reduced API.
111 - actually 0-initialise struct sigaction when installing signals.
112 - add section on hibernate and stopped processes to ev_timer docs.
113
1143.6 Tue Apr 28 02:49:30 CEST 2009
115 - multiple timers becoming ready within an event loop iteration
116 will be invoked in the "correct" order now.
117 - do not leave the event loop early just because we have no active
118 watchers, fixing a problem when embedding a kqueue loop
119 that has active kernel events but no registered watchers
120 (reported by blacksand blacksand).
121 - correctly zero the idx values for arrays, so destroying and
122 reinitialising the default loop actually works (patch by
123 Malek Hadj-Ali).
124 - implement ev_suspend and ev_resume.
125 - new EV_CUSTOM revents flag for use by applications.
126 - add documentation section about priorites.
127 - add a glossary to the dcoumentation.
128 - extend the ev_fork description slightly.
129 - optimize a jump out of call_pending.
130
1313.53 Sun Feb 15 02:38:20 CET 2009
132 - fix a bug in event pipe creation on win32 that would cause a
133 failed assertion on event loop creation (patch by Malek Hadj-Ali).
134 - probe for CLOCK_REALTIME support at runtime as well and fall
135 back to gettimeofday if there is an error, to support older
136 operating systems with newer header files/libraries.
137 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
138 (default most everywhere), otherwise not.
139
1403.52 Wed Jan 7 21:43:02 CET 2009
141 - fix compilation of select backend in fd_set mode when NFDBITS is
142 missing (to get it to compile on QNX, reported by Rodrigo Campos).
143 - better select-nfds handling when select backend is in fd_set mode.
144 - diagnose fd_set overruns when select backend is in fd_set mode.
145 - due to a thinko, instead of disabling everything but
146 select on the borked OS X platform, everything but select was
147 allowed (reported by Emanuele Giaquinta).
148 - actually verify that local and remote port are matching in
149 libev's socketpair emulation, which makes denial-of-service
150 attacks harder (but not impossible - it's windows). Make sure
151 it even works under vista, which thinks that getpeer/sockname
152 should return fantasy port numbers.
153 - include "libev" in all assertion messages for potentially
154 clearer diagnostics.
155 - event_get_version (libevent compatibility) returned
156 a useless string instead of the expected version string
157 (patch by W.C.A. Wijngaards).
158
1593.51 Wed Dec 24 23:00:11 CET 2008
160 - fix a bug where an inotify watcher was added twice, causing
161 freezes on hash collisions (reported and analysed by Graham Leggett).
162 - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
163 a direct syscall - slower, but no dependency on librt et al.
164 - assume negative return values != -1 signals success of port_getn
165 (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
166 (no known failure reports, but it doesn't hurt).
167 - fork detection in ev_embed now stops and restarts the watcher
168 automatically.
169 - EXPERIMENTAL: default the method to operator () in ev++.h,
170 to make it nicer to use functors (requested by Benedek László).
171 - fixed const object callbacks in ev++.h.
172 - replaced loop_ref argument of watcher.set (loop) by a direct
173 ev_loop * in ev++.h, to avoid clashes with functor patch.
174 - do not try to watch the empty string via inotify.
175 - inotify watchers could be leaked under certain circumstances.
176 - OS X 10.5 is actually even more broken than earlier versions,
177 so fall back to select on that piece of garbage.
178 - fixed some weirdness in the ev_embed documentation.
179
1803.49 Wed Nov 19 11:26:53 CET 2008
181 - ev_stat watchers will now use inotify as a mere hint on
182 kernels <2.6.25, or if the filesystem is not in the
183 "known to be good" list.
184 - better mingw32 compatibility (it's not as borked as native win32)
185 (analysed by Roger Pack).
186 - include stdio.h in the example program, as too many people are
187 confused by the weird C language otherwise. I guess the next thing
188 I get told is that the "..." ellipses in the examples don't compile
189 with their C compiler.
190
1913.48 Thu Oct 30 09:02:37 CET 2008
5 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll 192 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
6 backend by assuming the kernel event mask hasn't changed if 193 backend by assuming the kernel event mask hasn't changed if
7 ADD fails with EEXIST. 194 ADD fails with EEXIST.
8 - work around spurious event notification bugs in epoll by using 195 - work around spurious event notification bugs in epoll by using
9 an 8-bit generation counter. 196 a 32-bit generation counter. recreate kernel state if we receive
197 spurious notifications or unwanted events. this is very costly,
198 but I didn't come up with this horrible design.
10 - use memset to initialise most arrays now and do away with the 199 - use memset to initialise most arrays now and do away with the
11 init functions. 200 init functions.
12 - expand time-out strategies into a "Be smart about timeouts" section. 201 - expand time-out strategies into a "Be smart about timeouts" section.
13 - drop the "struct" from all ev_watcher declarations in the 202 - drop the "struct" from all ev_watcher declarations in the
14 documentation and did other clarifications (yeah, it was a mistake 203 documentation and did other clarifications (yeah, it was a mistake
15 to have a struct AND a function called ev_loop). 204 to have a struct AND a function called ev_loop).
16 - fix a bug where ev_default would not initialise the default 205 - fix a bug where ev_default would not initialise the default
17 loop again after it was destroyed with ev_default_destroy. 206 loop again after it was destroyed with ev_default_destroy.
207 - rename syserr to ev_syserr to avoid name clashes when embedding,
208 do similar changes for event.c.
18 209
193.45 Tue Oct 21 21:59:26 CEST 2008 2103.45 Tue Oct 21 21:59:26 CEST 2008
20 - disable inotify usage on linux <2.6.25, as it is broken 211 - disable inotify usage on linux <2.6.25, as it is broken
21 (reported by Yoann Vandoorselaere). 212 (reported by Yoann Vandoorselaere).
22 - ev_stat errornously would try to add inotify watchers 213 - ev_stat erroneously would try to add inotify watchers
23 even when inotify wasn't available (this should only 214 even when inotify wasn't available (this should only
24 have a performance impact). 215 have a performance impact).
25 - ev_once now passes both timeout and io to the callback if both 216 - ev_once now passes both timeout and io to the callback if both
26 occur concurrently, instead of giving timeouts precedence. 217 occur concurrently, instead of giving timeouts precedence.
27 - disable EV_USE_INOTIFY when sys/inotify.h is too old. 218 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
… …
95 - enable inlining with gcc versions 2 and 3. 286 - enable inlining with gcc versions 2 and 3.
96 - work around broken poll implementations potentially 287 - work around broken poll implementations potentially
97 not clearing revents field in ev_poll (Brandon Black) 288 not clearing revents field in ev_poll (Brandon Black)
98 (no such systems are known at this time). 289 (no such systems are known at this time).
99 - work around a bug in realloc on openbsd and darwin, 290 - work around a bug in realloc on openbsd and darwin,
100 also makes the errornous valgrind complaints 291 also makes the erroneous valgrind complaints
101 go away (noted by various people). 292 go away (noted by various people).
102 - fix ev_async_pending, add c++ wrapper for ev_async 293 - fix ev_async_pending, add c++ wrapper for ev_async
103 (based on patch sent by Johannes Deisenhofer. 294 (based on patch sent by Johannes Deisenhofer).
104 - add sensible set method to ev::embed. 295 - add sensible set method to ev::embed.
105 - made integer constants type int in ev.h. 296 - made integer constants type int in ev.h.
106 297
1073.2 Wed Apr 2 17:11:19 CEST 2008 2983.2 Wed Apr 2 17:11:19 CEST 2008
108 - fix a 64 bit overflow issue in the select backend, 299 - fix a 64 bit overflow issue in the select backend,

Diff Legend

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