1 |
Revision history for libev, a high-performance and full-featured event loop. |
2 |
|
3 |
- fix a bug where an inotify watcher was added twice, causing |
4 |
freezes on hash collisions (reported and analysed by Graham Leggett). |
5 |
- new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use |
6 |
a direct syscall - slower, but no dependency on librt et al. |
7 |
- assume negative return values != -1 signals success of port_getn |
8 |
(http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52) |
9 |
(no known failure reports, but it doesn't hurt). |
10 |
- fork detection in ev_embed now stops and restarts the watcher |
11 |
automatically. |
12 |
- EXPERIMENTAL: default the method to operator () in ev++.h, |
13 |
to make it nicer to use functors (requested by Benedek László). |
14 |
- fixed const object callbacks in ev++.h. |
15 |
- replaced loop_ref argument of watcher.set (loop) by a direct |
16 |
ev_loop * in ev++.h, to avoid clashes with functor patch. |
17 |
- do not try to watch the empty string via inotify. |
18 |
- inotify watchers could be leaked under certain circumstances. |
19 |
- OS X 10.5 is actually even more broken than earlier versions, |
20 |
so fall back to select on that piece of garbage. |
21 |
- fixed some weirdness in the ev_embed documentation. |
22 |
|
23 |
3.49 Wed Nov 19 11:26:53 CET 2008 |
24 |
- ev_stat watchers will now use inotify as a mere hint on |
25 |
kernels <2.6.25, or if the filesystem is not in the |
26 |
"known to be good" list. |
27 |
- better mingw32 compatibility (it's not as borked as native win32) |
28 |
(analysed by Roger Pack). |
29 |
- include stdio.h in the example program, as too many people are |
30 |
confused by the weird C language otherwise. I guess the next thing |
31 |
I get told is that the "..." ellipses in the examples don't compile |
32 |
with their C compiler. |
33 |
|
34 |
3.48 Thu Oct 30 09:02:37 CET 2008 |
35 |
- further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll |
36 |
backend by assuming the kernel event mask hasn't changed if |
37 |
ADD fails with EEXIST. |
38 |
- work around spurious event notification bugs in epoll by using |
39 |
a 32-bit generation counter. recreate kernel state if we receive |
40 |
spurious notifications or unwanted events. this is very costly, |
41 |
but I didn't come up with this horrible design. |
42 |
- use memset to initialise most arrays now and do away with the |
43 |
init functions. |
44 |
- expand time-out strategies into a "Be smart about timeouts" section. |
45 |
- drop the "struct" from all ev_watcher declarations in the |
46 |
documentation and did other clarifications (yeah, it was a mistake |
47 |
to have a struct AND a function called ev_loop). |
48 |
- fix a bug where ev_default would not initialise the default |
49 |
loop again after it was destroyed with ev_default_destroy. |
50 |
- rename syserr to ev_syserr to avoid name clashes when embedding, |
51 |
do similar changes for event.c. |
52 |
|
53 |
3.45 Tue Oct 21 21:59:26 CEST 2008 |
54 |
- disable inotify usage on linux <2.6.25, as it is broken |
55 |
(reported by Yoann Vandoorselaere). |
56 |
- ev_stat errornously would try to add inotify watchers |
57 |
even when inotify wasn't available (this should only |
58 |
have a performance impact). |
59 |
- ev_once now passes both timeout and io to the callback if both |
60 |
occur concurrently, instead of giving timeouts precedence. |
61 |
- disable EV_USE_INOTIFY when sys/inotify.h is too old. |
62 |
|
63 |
3.44 Mon Sep 29 05:18:39 CEST 2008 |
64 |
- embed watchers now automatically invoke ev_loop_fork on the |
65 |
embedded loop when the parent loop forks. |
66 |
- new function: ev_now_update (loop). |
67 |
- verify_watcher was not marked static. |
68 |
- improve the "associating..." manpage section. |
69 |
- documentation tweaks here and there. |
70 |
|
71 |
3.43 Sun Jul 6 05:34:41 CEST 2008 |
72 |
- include more include files on windows to get struct _stati64 |
73 |
(reported by Chris Hulbert, but doesn't quite fix his issue). |
74 |
- add missing #include <io.h> in ev.c on windows (reported by |
75 |
Matt Tolton). |
76 |
|
77 |
3.42 Tue Jun 17 12:12:07 CEST 2008 |
78 |
- work around yet another windows bug: FD_SET actually adds fd's |
79 |
multiple times to the fd_*SET*, despite official MSN docs claiming |
80 |
otherwise. Reported and well-analysed by Matt Tolton. |
81 |
- define NFDBITS to 0 when EV_SELECT_IS_WINSOCKET to make it compile |
82 |
(reported any analysed by Chris Hulbert). |
83 |
- fix a bug in ev_ebadf (this function is only used to catch |
84 |
programming errors in the libev user). reported by Matt Tolton. |
85 |
- fix a bug in fd_intern on win32 (could lead to compile errors |
86 |
under some circumstances, but would work correctly if it compiles). |
87 |
reported by Matt Tolton. |
88 |
- (try to) work around missing lstat on windows. |
89 |
- pass in the write fd set as except fd set under windows. windows |
90 |
is so uncontrollably lame that it requires this. this means that |
91 |
switching off oobinline is not supported (but tcp/ip doesn't |
92 |
have oob, so that would be stupid anyways. |
93 |
- use posix module symbol to auto-detect monotonic clock presence |
94 |
and some other default values. |
95 |
|
96 |
3.41 Fri May 23 18:42:54 CEST 2008 |
97 |
- work around an obscure bug in winsocket select: if you |
98 |
provide only empty fd sets then select returns WSAEINVAL. how sucky. |
99 |
- improve timer scheduling stability and reduce use of time_epsilon. |
100 |
- use 1-based 2-heap for EV_MINIMAL, simplifies code, reduces |
101 |
codesize and makes for better cache-efficiency. |
102 |
- use 3-based 4-heap for !EV_MINIMAL. this makes better use |
103 |
of cpu cache lines and gives better growth behaviour than |
104 |
2-based heaps. |
105 |
- cache timestamp within heap for !EV_MINIMAL, to avoid random |
106 |
memory accesses. |
107 |
- document/add EV_USE_4HEAP and EV_HEAP_CACHE_AT. |
108 |
- fix a potential aliasing issue in ev_timer_again. |
109 |
- add/document ev_periodic_at, retract direct access to ->at. |
110 |
- improve ev_stat docs. |
111 |
- add portability requirements section. |
112 |
- fix manpage headers etc. |
113 |
- normalise WSA error codes to lower range on windows. |
114 |
- add consistency check code that can be called automatically |
115 |
or on demand to check for internal structures (ev_loop_verify). |
116 |
|
117 |
3.31 Wed Apr 16 20:45:04 CEST 2008 |
118 |
- added last minute fix for ev_poll.c by Brandon Black. |
119 |
|
120 |
3.3 Wed Apr 16 19:04:10 CEST 2008 |
121 |
- event_base_loopexit should return 0 on success |
122 |
(W.C.A. Wijngaards). |
123 |
- added linux eventfd support. |
124 |
- try to autodetect epoll and inotify support |
125 |
by libc header version if not using autoconf. |
126 |
- new symbols: EV_DEFAULT_UC and EV_DEFAULT_UC_. |
127 |
- declare functions defined in ev.h as inline if |
128 |
C99 or gcc are available. |
129 |
- enable inlining with gcc versions 2 and 3. |
130 |
- work around broken poll implementations potentially |
131 |
not clearing revents field in ev_poll (Brandon Black) |
132 |
(no such systems are known at this time). |
133 |
- work around a bug in realloc on openbsd and darwin, |
134 |
also makes the errornous valgrind complaints |
135 |
go away (noted by various people). |
136 |
- fix ev_async_pending, add c++ wrapper for ev_async |
137 |
(based on patch sent by Johannes Deisenhofer. |
138 |
- add sensible set method to ev::embed. |
139 |
- made integer constants type int in ev.h. |
140 |
|
141 |
3.2 Wed Apr 2 17:11:19 CEST 2008 |
142 |
- fix a 64 bit overflow issue in the select backend, |
143 |
by using fd_mask instead of int for the mask. |
144 |
- rename internal sighandler to avoid clash with very old perls. |
145 |
- entering ev_loop will not clear the ONESHOT or NONBLOCKING |
146 |
flags of any outer loops anymore. |
147 |
- add ev_async_pending. |
148 |
|
149 |
3.1 Thu Mar 13 13:45:22 CET 2008 |
150 |
- implement ev_async watchers. |
151 |
- only initialise signal pipe on demand. |
152 |
- make use of sig_atomic_t configurable. |
153 |
- improved documentation. |
154 |
|
155 |
3.0 Mon Jan 28 13:14:47 CET 2008 |
156 |
- API/ABI bump to version 3.0. |
157 |
- ev++.h includes "ev.h" by default now, not <ev.h>. |
158 |
- slightly improved documentation. |
159 |
- speed up signal detection after a fork. |
160 |
- only optionally return trace status changed in ev_child |
161 |
watchers. |
162 |
- experimental (and undocumented) loop wrappers for ev++.h. |
163 |
|
164 |
2.01 Tue Dec 25 08:04:41 CET 2007 |
165 |
- separate Changes file. |
166 |
- fix ev_path_set => ev_stat_set typo. |
167 |
- remove event_compat.h from the libev tarball. |
168 |
- change how include files are found. |
169 |
- doc updates. |
170 |
- update licenses, explicitly allow for GPL relicensing. |
171 |
|
172 |
2.0 Sat Dec 22 17:47:03 CET 2007 |
173 |
- new ev_sleep, ev_set_(io|timeout)_collect_interval. |
174 |
- removed epoll from embeddable fd set. |
175 |
- fix embed watchers. |
176 |
- renamed ev_embed.loop to other. |
177 |
- added exported Symbol tables. |
178 |
- undefine member wrapper macros at the end of ev.c. |
179 |
- respect EV_H in ev++.h. |
180 |
|
181 |
1.86 Tue Dec 18 02:36:57 CET 2007 |
182 |
- fix memleak on loop destroy (not relevant for perl). |
183 |
|
184 |
1.85 Fri Dec 14 20:32:40 CET 2007 |
185 |
- fix some aliasing issues w.r.t. timers and periodics |
186 |
(not relevant for perl). |
187 |
|
188 |
(for historic versions refer to EV/Changes, found in the Perl interface) |
189 |
|
190 |
0.1 Wed Oct 31 21:31:48 CET 2007 |
191 |
- original version; hacked together in <24h. |
192 |
|