| 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 | WISH? monotonic clocks times/GetTickCount for coarse corrections? |
3 | - fix compilation of select backend in fd_set mode when NFDBITS is |
| 4 | TODO: make ev_once timeout=0 useful, and make I/O always win. |
4 | missing (to get it to compile on QNX, reported by Rodrigo Campos). |
|
|
5 | - better select-nfds handling when select backend is in fd_set mode. |
|
|
6 | - diagnose fd_set overruns when select backend is in fd_set mode. |
|
|
7 | - due to a thinko, instead of disabling everything but |
|
|
8 | select on the borked OS X platform, everything but select was |
|
|
9 | allowed (reported by Emanuele Giaquinta). |
|
|
10 | - actually verify that local and remote port are matching in |
|
|
11 | libev's socketpair emulation, which makes denial-of-service |
|
|
12 | attacks harder (but not impossible - it's windows). Make sure |
|
|
13 | it even works under vista, which thinks that getpeer/sockname |
|
|
14 | should return fantasy port numbers. |
|
|
15 | - include "libev" all assertion messages for potentially |
|
|
16 | clearer diagnostics. |
|
|
17 | |
|
|
18 | 3.51 Wed Dec 24 23:00:11 CET 2008 |
|
|
19 | - fix a bug where an inotify watcher was added twice, causing |
|
|
20 | freezes on hash collisions (reported and analysed by Graham Leggett). |
|
|
21 | - new config symbol, EV_USE_CLOCK_SYSCALL, to make libev use |
|
|
22 | a direct syscall - slower, but no dependency on librt et al. |
|
|
23 | - assume negative return values != -1 signals success of port_getn |
|
|
24 | (http://cvs.epicsol.org/cgi/viewcvs.cgi/epic5/source/newio.c?rev=1.52) |
|
|
25 | (no known failure reports, but it doesn't hurt). |
|
|
26 | - fork detection in ev_embed now stops and restarts the watcher |
|
|
27 | automatically. |
|
|
28 | - EXPERIMENTAL: default the method to operator () in ev++.h, |
|
|
29 | to make it nicer to use functors (requested by Benedek László). |
|
|
30 | - fixed const object callbacks in ev++.h. |
|
|
31 | - replaced loop_ref argument of watcher.set (loop) by a direct |
|
|
32 | ev_loop * in ev++.h, to avoid clashes with functor patch. |
|
|
33 | - do not try to watch the empty string via inotify. |
|
|
34 | - inotify watchers could be leaked under certain circumstances. |
|
|
35 | - OS X 10.5 is actually even more broken than earlier versions, |
|
|
36 | so fall back to select on that piece of garbage. |
|
|
37 | - fixed some weirdness in the ev_embed documentation. |
|
|
38 | |
|
|
39 | 3.49 Wed Nov 19 11:26:53 CET 2008 |
|
|
40 | - ev_stat watchers will now use inotify as a mere hint on |
|
|
41 | kernels <2.6.25, or if the filesystem is not in the |
|
|
42 | "known to be good" list. |
|
|
43 | - better mingw32 compatibility (it's not as borked as native win32) |
|
|
44 | (analysed by Roger Pack). |
|
|
45 | - include stdio.h in the example program, as too many people are |
|
|
46 | confused by the weird C language otherwise. I guess the next thing |
|
|
47 | I get told is that the "..." ellipses in the examples don't compile |
|
|
48 | with their C compiler. |
|
|
49 | |
|
|
50 | 3.48 Thu Oct 30 09:02:37 CET 2008 |
|
|
51 | - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll |
|
|
52 | backend by assuming the kernel event mask hasn't changed if |
|
|
53 | ADD fails with EEXIST. |
|
|
54 | - work around spurious event notification bugs in epoll by using |
|
|
55 | a 32-bit generation counter. recreate kernel state if we receive |
|
|
56 | spurious notifications or unwanted events. this is very costly, |
|
|
57 | but I didn't come up with this horrible design. |
|
|
58 | - use memset to initialise most arrays now and do away with the |
|
|
59 | init functions. |
|
|
60 | - expand time-out strategies into a "Be smart about timeouts" section. |
|
|
61 | - drop the "struct" from all ev_watcher declarations in the |
|
|
62 | documentation and did other clarifications (yeah, it was a mistake |
|
|
63 | to have a struct AND a function called ev_loop). |
|
|
64 | - fix a bug where ev_default would not initialise the default |
|
|
65 | loop again after it was destroyed with ev_default_destroy. |
|
|
66 | - rename syserr to ev_syserr to avoid name clashes when embedding, |
|
|
67 | do similar changes for event.c. |
|
|
68 | |
|
|
69 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
|
|
70 | - disable inotify usage on linux <2.6.25, as it is broken |
|
|
71 | (reported by Yoann Vandoorselaere). |
|
|
72 | - ev_stat errornously would try to add inotify watchers |
|
|
73 | even when inotify wasn't available (this should only |
|
|
74 | have a performance impact). |
|
|
75 | - ev_once now passes both timeout and io to the callback if both |
|
|
76 | occur concurrently, instead of giving timeouts precedence. |
|
|
77 | - disable EV_USE_INOTIFY when sys/inotify.h is too old. |
|
|
78 | |
| 5 | 3.44 Mon Sep 29 05:18:39 CEST 2008 |
79 | 3.44 Mon Sep 29 05:18:39 CEST 2008 |
| 6 | - embed watchers now automatically invoke ev_loop_fork on the |
80 | - embed watchers now automatically invoke ev_loop_fork on the |
| 7 | embedded loop when the parent loop forks. |
81 | embedded loop when the parent loop forks. |
| 8 | - new function: ev_now_update (loop). |
82 | - new function: ev_now_update (loop). |
| 9 | - verify_watcher was not marked static. |
83 | - verify_watcher was not marked static. |
