| 1 | Revision history for libev, a high-performance and full-featured event loop. |
1 | Revision history for libev, a high-performance and full-featured event loop. |
| 2 | |
2 | |
|
|
3 | - correctly zero the idx values for arrays, so destroying and |
|
|
4 | reinitialising the default loop actually works (patch by |
|
|
5 | Malek Hadj-Ali). |
|
|
6 | |
|
|
7 | 3.53 Sun Feb 15 02:38:20 CET 2009 |
|
|
8 | - fix a bug in event pipe creation on win32 that would cause a |
|
|
9 | failed assertion on event loop creation (patch by Malek Hadj-Ali). |
|
|
10 | - probe for CLOCK_REALTIME support at runtime as well and fall |
|
|
11 | back to gettimeofday if there is an error, to support older |
|
|
12 | operating systems with newer header files/libraries. |
|
|
13 | - prefer gettimeofday over clock_gettime with USE_CLOCK_SYSCALL |
|
|
14 | (default most everywhere), otherwise not. |
|
|
15 | |
|
|
16 | 3.52 Wed Jan 7 21:43:02 CET 2009 |
|
|
17 | - fix compilation of select backend in fd_set mode when NFDBITS is |
|
|
18 | missing (to get it to compile on QNX, reported by Rodrigo Campos). |
|
|
19 | - better select-nfds handling when select backend is in fd_set mode. |
|
|
20 | - diagnose fd_set overruns when select backend is in fd_set mode. |
|
|
21 | - due to a thinko, instead of disabling everything but |
|
|
22 | select on the borked OS X platform, everything but select was |
|
|
23 | allowed (reported by Emanuele Giaquinta). |
|
|
24 | - actually verify that local and remote port are matching in |
|
|
25 | libev's socketpair emulation, which makes denial-of-service |
|
|
26 | attacks harder (but not impossible - it's windows). Make sure |
|
|
27 | it even works under vista, which thinks that getpeer/sockname |
|
|
28 | should return fantasy port numbers. |
|
|
29 | - include "libev" in all assertion messages for potentially |
|
|
30 | clearer diagnostics. |
|
|
31 | - event_get_version (libevent compatibility) returned |
|
|
32 | a useless string instead of the expected version string |
|
|
33 | (patch by W.C.A. Wijngaards). |
|
|
34 | |
|
|
35 | 3.51 Wed Dec 24 23:00:11 CET 2008 |
|
|
36 | - fix a bug where an inotify watcher was added twice, causing |
|
|
37 | freezes on hash collisions (reported and analysed by Graham Leggett). |
|
|
38 | - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use |
|
|
39 | a direct syscall - slower, but no dependency on librt et al. |
|
|
40 | - assume negative return values != -1 signals success of port_getn |
|
|
41 | (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52) |
|
|
42 | (no known failure reports, but it doesn't hurt). |
|
|
43 | - fork detection in ev_embed now stops and restarts the watcher |
|
|
44 | automatically. |
|
|
45 | - EXPERIMENTAL: default the method to operator () in ev++.h, |
|
|
46 | to make it nicer to use functors (requested by Benedek László). |
|
|
47 | - fixed const object callbacks in ev++.h. |
|
|
48 | - replaced loop_ref argument of watcher.set (loop) by a direct |
|
|
49 | ev_loop * in ev++.h, to avoid clashes with functor patch. |
|
|
50 | - do not try to watch the empty string via inotify. |
|
|
51 | - inotify watchers could be leaked under certain circumstances. |
|
|
52 | - OS X 10.5 is actually even more broken than earlier versions, |
|
|
53 | so fall back to select on that piece of garbage. |
|
|
54 | - fixed some weirdness in the ev_embed documentation. |
|
|
55 | |
|
|
56 | 3.49 Wed Nov 19 11:26:53 CET 2008 |
|
|
57 | - ev_stat watchers will now use inotify as a mere hint on |
|
|
58 | kernels <2.6.25, or if the filesystem is not in the |
|
|
59 | "known to be good" list. |
|
|
60 | - better mingw32 compatibility (it's not as borked as native win32) |
|
|
61 | (analysed by Roger Pack). |
|
|
62 | - include stdio.h in the example program, as too many people are |
|
|
63 | confused by the weird C language otherwise. I guess the next thing |
|
|
64 | I get told is that the "..." ellipses in the examples don't compile |
|
|
65 | with their C compiler. |
|
|
66 | |
|
|
67 | 3.48 Thu Oct 30 09:02:37 CET 2008 |
|
|
68 | - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll |
|
|
69 | backend by assuming the kernel event mask hasn't changed if |
|
|
70 | ADD fails with EEXIST. |
|
|
71 | - work around spurious event notification bugs in epoll by using |
|
|
72 | a 32-bit generation counter. recreate kernel state if we receive |
|
|
73 | spurious notifications or unwanted events. this is very costly, |
|
|
74 | but I didn't come up with this horrible design. |
|
|
75 | - use memset to initialise most arrays now and do away with the |
|
|
76 | init functions. |
|
|
77 | - expand time-out strategies into a "Be smart about timeouts" section. |
|
|
78 | - drop the "struct" from all ev_watcher declarations in the |
|
|
79 | documentation and did other clarifications (yeah, it was a mistake |
|
|
80 | to have a struct AND a function called ev_loop). |
|
|
81 | - fix a bug where ev_default would not initialise the default |
|
|
82 | loop again after it was destroyed with ev_default_destroy. |
|
|
83 | - rename syserr to ev_syserr to avoid name clashes when embedding, |
|
|
84 | do similar changes for event.c. |
|
|
85 | |
|
|
86 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
|
|
87 | - disable inotify usage on linux <2.6.25, as it is broken |
|
|
88 | (reported by Yoann Vandoorselaere). |
|
|
89 | - ev_stat errornously would try to add inotify watchers |
|
|
90 | even when inotify wasn't available (this should only |
|
|
91 | have a performance impact). |
|
|
92 | - ev_once now passes both timeout and io to the callback if both |
|
|
93 | occur concurrently, instead of giving timeouts precedence. |
|
|
94 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
|
|
95 | |
|
|
96 | 3.44 Mon Sep 29 05:18:39 CEST 2008 |
|
|
97 | - embed watchers now automatically invoke ev_loop_fork on the |
|
|
98 | embedded loop when the parent loop forks. |
|
|
99 | - new function: ev_now_update (loop). |
|
|
100 | - verify_watcher was not marked static. |
|
|
101 | - improve the "associating..." manpage section. |
|
|
102 | - documentation tweaks here and there. |
|
|
103 | |
|
|
104 | 3.43 Sun Jul 6 05:34:41 CEST 2008 |
|
|
105 | - include more include files on windows to get struct _stati64 |
|
|
106 | (reported by Chris Hulbert, but doesn't quite fix his issue). |
|
|
107 | - add missing #include <io.h> in ev.c on windows (reported by |
|
|
108 | Matt Tolton). |
|
|
109 | |
|
|
110 | 3.42 Tue Jun 17 12:12:07 CEST 2008 |
|
|
111 | - work around yet another windows bug: FD_SET actually adds fd's |
|
|
112 | multiple times to the fd_*SET*, despite official MSN docs claiming |
|
|
113 | otherwise. Reported and well-analysed by Matt Tolton. |
| 3 | - define NFSBITS to 0 when EV_SELECT_IS_WINSOCKET to make it compile |
114 | - define NFDBITS to 0 when EV_SELECT_IS_WINSOCKET to make it compile |
| 4 | (reported any analysed by Chris Hulbert). |
115 | (reported any analysed by Chris Hulbert). |
|
|
116 | - fix a bug in ev_ebadf (this function is only used to catch |
|
|
117 | programming errors in the libev user). reported by Matt Tolton. |
|
|
118 | - fix a bug in fd_intern on win32 (could lead to compile errors |
|
|
119 | under some circumstances, but would work correctly if it compiles). |
|
|
120 | reported by Matt Tolton. |
|
|
121 | - (try to) work around missing lstat on windows. |
| 5 | - pass in the write fd set as except fd set under windows. windows |
122 | - pass in the write fd set as except fd set under windows. windows |
| 6 | is so uncontrollably lame that it requires this. this means that |
123 | is so uncontrollably lame that it requires this. this means that |
| 7 | switching off oobinline is not supported (but tcp/ip doesn't |
124 | switching off oobinline is not supported (but tcp/ip doesn't |
| 8 | have oob, so that would be stupid anyways. |
125 | have oob, so that would be stupid anyways. |
| 9 | - use posix module symbol to auto-detect monotonic clock presence |
126 | - use posix module symbol to auto-detect monotonic clock presence |
