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 1.167 by root, Tue Mar 16 00:43:22 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: ABI??? API????? Changes???
4 - ev_embed_stop did not correctly stop the watcher (very good
5 testcase by Vladimir Timofeev).
6 - applied win32 fixes by Michael Lenaghan (also James Mansion).
7 - replace EV_MINIMAL by EV_FEATURES.
8 - remove dependency on sys/queue.h on freebsd (patch by Vanilla Hsu).
9 - configure now prepends -O3, not appends it, so one can still
10 override it.
11 - disable poll backend on AIX, the poll header spams the namespace
12 and it's not worth working around dead platforms (reported
13 and analyzed by Aivars Kalvans).
14 - improve header file compatibility of the standalone eventfd code
15 in an obscure case.
16 - implement EV_AVOID_STDIO option.
17 - do not use sscanf to parse linux version number (smaller, faster,
18 no sscanf dependency).
19 - new EV_CHILD_ENABLE and EV_SIGNAL_ENABLE configurable settings.
20 - update libev.m4 HAVE_CLOCK_SYSCALL test for newwer glibcs.
21 - add section on accept() problems to the manpage.
4 22
233.9 Thu Dec 31 07:59:59 CET 2009
24 - signalfd is no longer used by default and has to be requested
25 explicitly - this means that easy to catch bugs become hard to
26 catch race conditions, but the users have spoken.
27 - point out the unspecified signal mask in the documentation, and
28 that this is a race condition regardless of EV_SIGNALFD.
29 - backport inotify code to C89.
30 - inotify file descriptors could leak into child processes.
31 - ev_stat watchers could keep an errornous extra ref on the loop,
32 preventing exit when unregistering all watchers (testcases
33 provided by ry@tinyclouds.org).
34 - implement EV_WIN32_HANDLE_TO_FD and EV_WIN32_CLOSE_FD configuration
35 symbols to make it easier for apps to do their own fd management.
36 - support EV_IDLE_ENABLE being disabled in ev++.h
37 (patch by Didier Spezia).
38 - take advantage of inotify_init1, if available, to set cloexec/nonblock
39 on fd creation, to avoid races.
40 - the signal handling pipe wasn't always initialised under windows
41 (analysed by lekma).
42 - changed minimum glibc requirement from glibc 2.9 to 2.7, for
43 signalfd.
44 - add missing string.h include (Denis F. Latypoff).
45 - only replace ev_stat.prev when we detect an actual difference,
46 so prev is (almost) always different to attr. this might
47 have caused the probems with 04_stat.t.
48 - add ev::timer->remaining () method to C++ API.
49
503.8 Sun Aug 9 14:30:45 CEST 2009
51 - incompatible change: do not necessarily reset signal handler
52 to SIG_DFL when a sighandler is stopped.
53 - ev_default_destroy did not properly free or zero some members,
54 potentially causing crashes and memory corruption on repated
55 ev_default_destroy/ev_default_loop calls.
56 - take advantage of signalfd on GNU/Linux systems.
57 - document that the signal mask might be in an unspecified
58 state when using libev's signal handling.
59 - take advantage of some GNU/Linux calls to set cloexec/nonblock
60 on fd creation, to avoid race conditions.
61
623.7 Fri Jul 17 16:36:32 CEST 2009
63 - ev_unloop and ev_loop wrongly used a global variable to exit loops,
64 instead of using a per-loop variable (bug caught by accident...).
65 - the ev_set_io_collect_interval interpretation has changed.
66 - add new functionality: ev_set_userdata, ev_userdata,
67 ev_set_invoke_pending_cb, ev_set_loop_release_cb,
68 ev_invoke_pending, ev_pending_count, together with a long example
69 about thread locking.
70 - add ev_timer_remaining (as requested by Denis F. Latypoff).
71 - add ev_loop_depth.
72 - calling ev_unloop in fork/prepare watchers will no longer poll
73 for new events.
74 - Denis F. Latypoff corrected many typos in example code snippets.
75 - honor autoconf detection of EV_USE_CLOCK_SYSCALL, also double-
76 check that the syscall number is available before trying to
77 use it (reported by ry@tinyclouds).
78 - use GetSystemTimeAsFileTime instead of _timeb on windows, for
79 slightly higher accuracy.
80 - properly declare ev_loop_verify and ev_now_update even when
81 !EV_MULTIPLICITY.
82 - do not compile in any priority code when EV_MAXPRI == EV_MINPRI.
83 - support EV_MINIMAL==2 for a reduced API.
84 - actually 0-initialise struct sigaction when installing signals.
85 - add section on hibernate and stopped processes to ev_timer docs.
86
873.6 Tue Apr 28 02:49:30 CEST 2009
88 - multiple timers becoming ready within an event loop iteration
89 will be invoked in the "correct" order now.
90 - do not leave the event loop early just because we have no active
91 watchers, fixing a problem when embedding a kqueue loop
92 that has active kernel events but no registered watchers
93 (reported by blacksand blacksand).
94 - correctly zero the idx values for arrays, so destroying and
95 reinitialising the default loop actually works (patch by
96 Malek Hadj-Ali).
97 - implement ev_suspend and ev_resume.
98 - new EV_CUSTOM revents flag for use by applications.
99 - add documentation section about priorites.
100 - add a glossary to the dcoumentation.
101 - extend the ev_fork description slightly.
102 - optimize a jump out of call_pending.
103
1043.53 Sun Feb 15 02:38:20 CET 2009
105 - fix a bug in event pipe creation on win32 that would cause a
106 failed assertion on event loop creation (patch by Malek Hadj-Ali).
107 - probe for CLOCK_REALTIME support at runtime as well and fall
108 back to gettimeofday if there is an error, to support older
109 operating systems with newer header files/libraries.
110 - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL
111 (default most everywhere), otherwise not.
112
1133.52 Wed Jan 7 21:43:02 CET 2009
114 - fix compilation of select backend in fd_set mode when NFDBITS is
115 missing (to get it to compile on QNX, reported by Rodrigo Campos).
116 - better select-nfds handling when select backend is in fd_set mode.
117 - diagnose fd_set overruns when select backend is in fd_set mode.
118 - due to a thinko, instead of disabling everything but
119 select on the borked OS X platform, everything but select was
120 allowed (reported by Emanuele Giaquinta).
121 - actually verify that local and remote port are matching in
122 libev's socketpair emulation, which makes denial-of-service
123 attacks harder (but not impossible - it's windows). Make sure
124 it even works under vista, which thinks that getpeer/sockname
125 should return fantasy port numbers.
126 - include "libev" in all assertion messages for potentially
127 clearer diagnostics.
128 - event_get_version (libevent compatibility) returned
129 a useless string instead of the expected version string
130 (patch by W.C.A. Wijngaards).
131
1323.51 Wed Dec 24 23:00:11 CET 2008
133 - fix a bug where an inotify watcher was added twice, causing
134 freezes on hash collisions (reported and analysed by Graham Leggett).
135 - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use
136 a direct syscall - slower, but no dependency on librt et al.
137 - assume negative return values != -1 signals success of port_getn
138 (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52)
139 (no known failure reports, but it doesn't hurt).
140 - fork detection in ev_embed now stops and restarts the watcher
141 automatically.
142 - EXPERIMENTAL: default the method to operator () in ev++.h,
143 to make it nicer to use functors (requested by Benedek László).
144 - fixed const object callbacks in ev++.h.
145 - replaced loop_ref argument of watcher.set (loop) by a direct
146 ev_loop * in ev++.h, to avoid clashes with functor patch.
147 - do not try to watch the empty string via inotify.
148 - inotify watchers could be leaked under certain circumstances.
149 - OS X 10.5 is actually even more broken than earlier versions,
150 so fall back to select on that piece of garbage.
151 - fixed some weirdness in the ev_embed documentation.
152
1533.49 Wed Nov 19 11:26:53 CET 2008
154 - ev_stat watchers will now use inotify as a mere hint on
155 kernels <2.6.25, or if the filesystem is not in the
156 "known to be good" list.
157 - better mingw32 compatibility (it's not as borked as native win32)
158 (analysed by Roger Pack).
159 - include stdio.h in the example program, as too many people are
160 confused by the weird C language otherwise. I guess the next thing
161 I get told is that the "..." ellipses in the examples don't compile
162 with their C compiler.
163
1643.48 Thu Oct 30 09:02:37 CET 2008
5 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll 165 - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll
6 backend by assuming the kernel event mask hasn't changed if 166 backend by assuming the kernel event mask hasn't changed if
7 ADD fails with EEXIST. 167 ADD fails with EEXIST.
8 - work around spurious event notification bugs in epoll by using 168 - work around spurious event notification bugs in epoll by using
9 an 8-bit generation counter. 169 a 32-bit generation counter. recreate kernel state if we receive
170 spurious notifications or unwanted events. this is very costly,
171 but I didn't come up with this horrible design.
10 - use memset to initialise most arrays now and do away with the 172 - use memset to initialise most arrays now and do away with the
11 init functions. 173 init functions.
12 - expand time-out strategies into a "Be smart about timeouts" section. 174 - expand time-out strategies into a "Be smart about timeouts" section.
13 - drop the "struct" from all ev_watcher declarations in the 175 - drop the "struct" from all ev_watcher declarations in the
14 documentation and did other clarifications (yeah, it was a mistake 176 documentation and did other clarifications (yeah, it was a mistake
15 to have a struct AND a function called ev_loop). 177 to have a struct AND a function called ev_loop).
16 - fix a bug where ev_default would not initialise the default 178 - fix a bug where ev_default would not initialise the default
17 loop again after it was destroyed with ev_default_destroy. 179 loop again after it was destroyed with ev_default_destroy.
180 - rename syserr to ev_syserr to avoid name clashes when embedding,
181 do similar changes for event.c.
18 182
193.45 Tue Oct 21 21:59:26 CEST 2008 1833.45 Tue Oct 21 21:59:26 CEST 2008
20 - disable inotify usage on linux <2.6.25, as it is broken 184 - disable inotify usage on linux <2.6.25, as it is broken
21 (reported by Yoann Vandoorselaere). 185 (reported by Yoann Vandoorselaere).
22 - ev_stat errornously would try to add inotify watchers 186 - ev_stat errornously would try to add inotify watchers

Diff Legend

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