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 1.205 by root, Wed Nov 3 20:03:49 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? 3 - automake fucked it up, apparently, --add-missing -f is not quite enough
4 to make it update its files. grrr.
5 - change EV_UNDEF to 0xffffffff to silence some overzealous compilers.
6 - use "(libev) " prefix for all libev error messages now.
4 7
84.00 Mon Oct 25 12:32:12 CEST 2010
9 - "PORTING FROM LIBEV 3.X TO 4.X" (in ev.pod) is recommended reading.
10 - ev_embed_stop did not correctly stop the watcher (very good
11 testcase by Vladimir Timofeev).
12 - ev_run will now always update the current loop time - it erroneously
13 didn't when idle watchers were active, causing timers not to fire.
14 - fix a bug where a timeout of zero caused the timer not to fire
15 in the libevent emulation (testcase by Péter Szabó).
16 - applied win32 fixes by Michael Lenaghan (also James Mansion).
17 - replace EV_MINIMAL by EV_FEATURES.
18 - prefer EPOLL_CTL_ADD over EPOLL_CTL_MOD in some more cases, as it
19 seems the former is *much* faster than the latter.
20 - linux kernel version detection (for inotify bug workarounds)
21 did not work properly.
22 - reduce the number of spurious wake-ups with the ports backend.
23 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
24 - do async init within ev_async_start, not ev_async_set, which avoids
25 an API quirk where the set function must be called in the C++ API
26 even when there is nothing to set.
27 - add (undocumented) EV_ENABLE when adding events with kqueue,
28 this might help with OS X, which seems to need it despite documenting
29 not to need it (helpfully pointed out by Tilghman Lesher).
30 - do not use poll by default on freebsd, it's broken (what isn't
31 on freebsd...).
32 - allow to embed epoll on kernels >= 2.6.32.
33 - configure now prepends -O3, not appends it, so one can still
34 override it.
35 - ev.pod: greatly expanded the portability section, added a porting
36 section, a description of watcher states and made lots of minor fixes.
37 - disable poll backend on AIX, the poll header spams the namespace
38 and it's not worth working around dead platforms (reported
39 and analyzed by Aivars Kalvans).
40 - improve header file compatibility of the standalone eventfd code
41 in an obscure case.
42 - implement EV_AVOID_STDIO option.
43 - do not use sscanf to parse linux version number (smaller, faster,
44 no sscanf dependency).
45 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
46 - update libev.m4 HAVE_CLOCK_SYSCALL test for newer glibcs.
47 - add section on accept() problems to the manpage.
48 - rename EV_TIMEOUT to EV_TIMER.
49 - rename ev_loop_count/depth/verify/loop/unloop.
50 - remove ev_default_destroy and ev_default_fork.
51 - switch to two-digit minor version.
52 - work around an apparent gentoo compiler bug.
53 - define _DARWIN_UNLIMITED_SELECT. just so.
54 - use enum instead of #define for most constants.
55 - improve compatibility to older C++ compilers.
56 - (experimental) ev_run/ev_default_loop/ev_break/ev_loop_new have now
57 default arguments when compiled as C++.
58 - enable automake dependency tracking.
59 - ev_loop_new no longer leaks memory when loop creation failed.
60 - new ev_cleanup watcher type.
61
623.9 Thu Dec 31 07:59:59 CET 2009
63 - signalfd is no longer used by default and has to be requested
64 explicitly - this means that easy to catch bugs become hard to
65 catch race conditions, but the users have spoken.
66 - point out the unspecified signal mask in the documentation, and
67 that this is a race condition regardless of EV_SIGNALFD.
68 - backport inotify code to C89.
69 - inotify file descriptors could leak into child processes.
70 - ev_stat watchers could keep an errornous extra ref on the loop,
71 preventing exit when unregistering all watchers (testcases
72 provided by ry@tinyclouds.org).
73 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
74 symbols to make it easier for apps to do their own fd management.
75 - support EV_IDLE_ENABLE being disabled in ev++.h
76 (patch by Didier Spezia).
77 - take advantage of inotify_init1, if available, to set cloexec/nonblock
78 on fd creation, to avoid races.
79 - the signal handling pipe wasn't always initialised under windows
80 (analysed by lekma).
81 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
82 signalfd.
83 - add missing string.h include (Denis F. Latypoff).
84 - only replace ev_stat.prev when we detect an actual difference,
85 so prev is (almost) always different to attr. this might
86 have caused the problems with 04_stat.t.
87 - add ev::timer->remaining () method to C++ API.
88
893.8 Sun Aug 9 14:30:45 CEST 2009
90 - incompatible change: do not necessarily reset signal handler
91 to SIG_DFL when a sighandler is stopped.
92 - ev_default_destroy did not properly free or zero some members,
93 potentially causing crashes and memory corruption on repeated
94 ev_default_destroy/ev_default_loop calls.
95 - take advantage of signalfd on GNU/Linux systems.
96 - document that the signal mask might be in an unspecified
97 state when using libev's signal handling.
98 - take advantage of some GNU/Linux calls to set cloexec/nonblock
99 on fd creation, to avoid race conditions.
100
1013.7 Fri Jul 17 16:36:32 CEST 2009
102 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
103 instead of using a per-loop variable (bug caught by accident...).
104 - the ev_set_io_collect_interval interpretation has changed.
105 - add new functionality: ev_set_userdata, ev_userdata,
106 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
107 ev_invoke_pending, ev_pending_count, together with a long example
108 about thread locking.
109 - add ev_timer_remaining (as requested by Denis F. Latypoff).
110 - add ev_loop_depth.
111 - calling ev_unloop in fork/prepare watchers will no longer poll
112 for new events.
113 - Denis F. Latypoff corrected many typos in example code snippets.
114 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
115 check that the syscall number is available before trying to
116 use it (reported by ry@tinyclouds).
117 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
118 slightly higher accuracy.
119 - properly declare ev_loop_verify and ev_now_update even when
120 !EV_MULTIPLICITY.
121 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
122 - support EV_MINIMAL==2 for a reduced API.
123 - actually 0-initialise struct sigaction when installing signals.
124 - add section on hibernate and stopped processes to ev_timer docs.
125
1263.6 Tue Apr 28 02:49:30 CEST 2009
127 - multiple timers becoming ready within an event loop iteration
128 will be invoked in the "correct" order now.
129 - do not leave the event loop early just because we have no active
130 watchers, fixing a problem when embedding a kqueue loop
131 that has active kernel events but no registered watchers
132 (reported by blacksand blacksand).
133 - correctly zero the idx values for arrays, so destroying and
134 reinitialising the default loop actually works (patch by
135 Malek Hadj-Ali).
136 - implement ev_suspend and ev_resume.
137 - new EV_CUSTOM revents flag for use by applications.
138 - add documentation section about priorites.
139 - add a glossary to the dcoumentation.
140 - extend the ev_fork description slightly.
141 - optimize a jump out of call_pending.
142
1433.53 Sun Feb 15 02:38:20 CET 2009
144 - fix a bug in event pipe creation on win32 that would cause a
145 failed assertion on event loop creation (patch by Malek Hadj-Ali).
146 - probe for CLOCK_REALTIME support at runtime as well and fall
147 back to gettimeofday if there is an error, to support older
148 operating systems with newer header files/libraries.
149 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
150 (default most everywhere), otherwise not.
151
1523.52 Wed Jan 7 21:43:02 CET 2009
153 - fix compilation of select backend in fd_set mode when NFDBITS is
154 missing (to get it to compile on QNX, reported by Rodrigo Campos).
155 - better select-nfds handling when select backend is in fd_set mode.
156 - diagnose fd_set overruns when select backend is in fd_set mode.
157 - due to a thinko, instead of disabling everything but
158 select on the borked OS X platform, everything but select was
159 allowed (reported by Emanuele Giaquinta).
160 - actually verify that local and remote port are matching in
161 libev's socketpair emulation, which makes denial-of-service
162 attacks harder (but not impossible - it's windows). Make sure
163 it even works under vista, which thinks that getpeer/sockname
164 should return fantasy port numbers.
165 - include "libev" in all assertion messages for potentially
166 clearer diagnostics.
167 - event_get_version (libevent compatibility) returned
168 a useless string instead of the expected version string
169 (patch by W.C.A. Wijngaards).
170
1713.51 Wed Dec 24 23:00:11 CET 2008
172 - fix a bug where an inotify watcher was added twice, causing
173 freezes on hash collisions (reported and analysed by Graham Leggett).
174 - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
175 a direct syscall - slower, but no dependency on librt et al.
176 - assume negative return values != -1 signals success of port_getn
177 (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
178 (no known failure reports, but it doesn't hurt).
179 - fork detection in ev_embed now stops and restarts the watcher
180 automatically.
181 - EXPERIMENTAL: default the method to operator () in ev++.h,
182 to make it nicer to use functors (requested by Benedek László).
183 - fixed const object callbacks in ev++.h.
184 - replaced loop_ref argument of watcher.set (loop) by a direct
185 ev_loop * in ev++.h, to avoid clashes with functor patch.
186 - do not try to watch the empty string via inotify.
187 - inotify watchers could be leaked under certain circumstances.
188 - OS X 10.5 is actually even more broken than earlier versions,
189 so fall back to select on that piece of garbage.
190 - fixed some weirdness in the ev_embed documentation.
191
1923.49 Wed Nov 19 11:26:53 CET 2008
193 - ev_stat watchers will now use inotify as a mere hint on
194 kernels <2.6.25, or if the filesystem is not in the
195 "known to be good" list.
196 - better mingw32 compatibility (it's not as borked as native win32)
197 (analysed by Roger Pack).
198 - include stdio.h in the example program, as too many people are
199 confused by the weird C language otherwise. I guess the next thing
200 I get told is that the "..." ellipses in the examples don't compile
201 with their C compiler.
202
2033.48 Thu Oct 30 09:02:37 CET 2008
5 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll 204 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
6 backend by assuming the kernel event mask hasn't changed if 205 backend by assuming the kernel event mask hasn't changed if
7 ADD fails with EEXIST. 206 ADD fails with EEXIST.
8 - work around spurious event notification bugs in epoll by using 207 - work around spurious event notification bugs in epoll by using
9 an 8-bit generation counter. 208 a 32-bit generation counter. recreate kernel state if we receive
209 spurious notifications or unwanted events. this is very costly,
210 but I didn't come up with this horrible design.
10 - use memset to initialise most arrays now and do away with the 211 - use memset to initialise most arrays now and do away with the
11 init functions. 212 init functions.
12 - expand time-out strategies into a "Be smart about timeouts" section. 213 - expand time-out strategies into a "Be smart about timeouts" section.
13 - drop the "struct" from all ev_watcher declarations in the 214 - drop the "struct" from all ev_watcher declarations in the
14 documentation and did other clarifications (yeah, it was a mistake 215 documentation and did other clarifications (yeah, it was a mistake
15 to have a struct AND a function called ev_loop). 216 to have a struct AND a function called ev_loop).
16 - fix a bug where ev_default would not initialise the default 217 - fix a bug where ev_default would not initialise the default
17 loop again after it was destroyed with ev_default_destroy. 218 loop again after it was destroyed with ev_default_destroy.
219 - rename syserr to ev_syserr to avoid name clashes when embedding,
220 do similar changes for event.c.
18 221
193.45 Tue Oct 21 21:59:26 CEST 2008 2223.45 Tue Oct 21 21:59:26 CEST 2008
20 - disable inotify usage on linux <2.6.25, as it is broken 223 - disable inotify usage on linux <2.6.25, as it is broken
21 (reported by Yoann Vandoorselaere). 224 (reported by Yoann Vandoorselaere).
22 - ev_stat errornously would try to add inotify watchers 225 - ev_stat erroneously would try to add inotify watchers
23 even when inotify wasn't available (this should only 226 even when inotify wasn't available (this should only
24 have a performance impact). 227 have a performance impact).
25 - ev_once now passes both timeout and io to the callback if both 228 - ev_once now passes both timeout and io to the callback if both
26 occur concurrently, instead of giving timeouts precedence. 229 occur concurrently, instead of giving timeouts precedence.
27 - disable EV_USE_INOTIFY when sys/inotify.h is too old. 230 - disable EV_USE_INOTIFY when sys/inotify.h is too old.
… …
95 - enable inlining with gcc versions 2 and 3. 298 - enable inlining with gcc versions 2 and 3.
96 - work around broken poll implementations potentially 299 - work around broken poll implementations potentially
97 not clearing revents field in ev_poll (Brandon Black) 300 not clearing revents field in ev_poll (Brandon Black)
98 (no such systems are known at this time). 301 (no such systems are known at this time).
99 - work around a bug in realloc on openbsd and darwin, 302 - work around a bug in realloc on openbsd and darwin,
100 also makes the errornous valgrind complaints 303 also makes the erroneous valgrind complaints
101 go away (noted by various people). 304 go away (noted by various people).
102 - fix ev_async_pending, add c++ wrapper for ev_async 305 - fix ev_async_pending, add c++ wrapper for ev_async
103 (based on patch sent by Johannes Deisenhofer. 306 (based on patch sent by Johannes Deisenhofer).
104 - add sensible set method to ev::embed. 307 - add sensible set method to ev::embed.
105 - made integer constants type int in ev.h. 308 - made integer constants type int in ev.h.
106 309
1073.2 Wed Apr 2 17:11:19 CEST 2008 3103.2 Wed Apr 2 17:11:19 CEST 2008
108 - fix a 64 bit overflow issue in the select backend, 311 - fix a 64 bit overflow issue in the select backend,

Diff Legend

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