| 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 | 3.48 Thu Oct 30 09:02:37 CET 2008 |
|
|
4 | - further optimise away the EPOLL_CTL_ADD/MOD combo in the epoll |
|
|
5 | backend by assuming the kernel event mask hasn't changed if |
|
|
6 | ADD fails with EEXIST. |
|
|
7 | - work around spurious event notification bugs in epoll by using |
|
|
8 | a 32-bit generation counter. recreate kernel state if we receive |
|
|
9 | spurious notifications or unwanted events. this is very costly, |
|
|
10 | but I didn't come up with this horrible design. |
|
|
11 | - use memset to initialise most arrays now and do away with the |
|
|
12 | init functions. |
|
|
13 | - expand time-out strategies into a "Be smart about timeouts" section. |
|
|
14 | - drop the "struct" from all ev_watcher declarations in the |
|
|
15 | documentation and did other clarifications (yeah, it was a mistake |
|
|
16 | to have a struct AND a function called ev_loop). |
|
|
17 | - fix a bug where ev_default would not initialise the default |
|
|
18 | loop again after it was destroyed with ev_default_destroy. |
|
|
19 | - rename syserr to ev_syserr to avoid name clashes when embedding, |
|
|
20 | do similar changes for event.c. |
| 3 | |
21 | |
| 4 | WISH? monotonic clocks times/GetTickCount for coarse corrections? |
22 | 3.45 Tue Oct 21 21:59:26 CEST 2008 |
| 5 | TODO: document 2.6.24-- breakage of inotify |
|
|
| 6 | - disable inotify usage on linux <2.6.25, as it is broken |
23 | - disable inotify usage on linux <2.6.25, as it is broken |
| 7 | (reported by Yoann Vandoorselaere). |
24 | (reported by Yoann Vandoorselaere). |
| 8 | - ev_stat errornously would try to add inotify watchers |
25 | - ev_stat errornously would try to add inotify watchers |
| 9 | even when inotify wasn't available (this should only |
26 | even when inotify wasn't available (this should only |
| 10 | have a performance impact). |
27 | have a performance impact). |
