| 1 | Revision history for Perl extension EV |
1 | Revision history for Perl extension EV |
| 2 | |
2 | |
|
|
3 | 3.7 |
|
|
4 | TODO: libev changes etc. etc. |
|
|
5 | - add EV::loop_verify and EV::loop_depth. |
|
|
6 | - use output filehandle for i/o watchers waiting |
|
|
7 | for EV_WRITE, otherwise input filehandle. |
|
|
8 | - use common schmorp.h header. |
|
|
9 | - allow for subclassing of EV::Loop (see the exciting |
|
|
10 | EV::Loop::Async module for an example). |
|
|
11 | - added EV::invoke_pending and EV::pending_count. |
|
|
12 | |
|
|
13 | 3.6 Tue Apr 28 02:50:37 CEST 2009 |
|
|
14 | - keepalive(0) could decrease the refcount of the loop |
|
|
15 | permanently. |
|
|
16 | - add ev_suspend/ev_resume, also make ev_now_update |
|
|
17 | accessible via the XS API. |
|
|
18 | - most EV::<eventtype> constants were missing and have been |
|
|
19 | added :/. |
|
|
20 | - add EV::VERSION_MAJOR/VERSION_MINOR constants. |
|
|
21 | - (libev) multiple timers becoming ready within an event loop |
|
|
22 | iteration will be invoked in the "correct" order now. |
|
|
23 | - (libev) do not leave the event loop early just because we have |
|
|
24 | no active watchers, fixing a problem when embedding a kqueue |
|
|
25 | loop that has active kernel events but no registered watchers |
|
|
26 | (reported by blacksand blacksand). |
|
|
27 | - (libev) correctly zero the idx values for arrays, so destroying |
|
|
28 | and reinitialising the default loop actually works (patch by |
|
|
29 | Malek Hadj-Ali). |
|
|
30 | - (libev) new EV::CUSTOM revents flag for use by applications. |
|
|
31 | - (libev) add documentation section about priorites. |
|
|
32 | - (libev) add a glossary to the dcoumentation. |
|
|
33 | - (libev) extend the ev_fork description slightly. |
|
|
34 | - (libev) optimize a jump out of call_pending. |
|
|
35 | - t/03_keepalive could fail when there was no actual error. |
|
|
36 | |
|
|
37 | 3.53 Sun Feb 15 02:38:20 CET 2009 |
|
|
38 | - (libev) on win32, the event loop creation could randomly fail |
|
|
39 | due to an initialised variable having the wrong value. |
|
|
40 | - (libev) probe for CLOCK_REALTIME support at runtime as well and |
|
|
41 | fall back to gettimeofday if there is an error, to support older |
|
|
42 | operating systems with newer header files/libraries. |
|
|
43 | - prefer gettimeofday over clock_gettime by default. |
|
|
44 | |
|
|
45 | 3.52 Wed Jan 7 21:46:14 CET 2009 |
|
|
46 | - (libev) fix some issues in the select backend when in fd_set mode. |
|
|
47 | - (libev) due to a thinko, instead of disabling everything but |
|
|
48 | select on the borked OS X platform, everything but select was |
|
|
49 | allowed (reported by Emanuele Giaquinta). |
|
|
50 | - (libev) actually verify that local and remote port are matching in |
|
|
51 | libev's socketpair emulation, which makes denial-of-service |
|
|
52 | attacks harder (but not impossible - it's windows). Make sure |
|
|
53 | it even works under vista, which thinks that getpeer/sockname |
|
|
54 | should return fantasy port numbers. |
|
|
55 | |
|
|
56 | 3.51 Wed Dec 24 23:01:59 CET 2008 |
|
|
57 | - do not cache the arguments passed to callbacks if the refcount |
|
|
58 | indicates that the callback has stolen them. |
| 3 | - (libev) try to avoid librt on GNU/Linux. |
59 | - (libev) try to avoid librt on GNU/Linux. |
|
|
60 | - (libev) check port_getn return value dfferently, might potentially |
|
|
61 | avoid problems. |
|
|
62 | - (libev) fix a bug with stat watchers possibly causing freezes. |
|
|
63 | - (libev) work around OS X 10.5 breaking poll, now select is the |
|
|
64 | only viable choice left on that pile of garbage. |
|
|
65 | - play tester whore: disable some tests that typically fail only |
|
|
66 | on cpan tester machines. |
| 4 | |
67 | |
| 5 | 3.49 Wed Nov 19 11:26:53 CET 2008 |
68 | 3.49 Wed Nov 19 11:26:53 CET 2008 |
| 6 | - fix typos in manpage (Alex Efros). |
69 | - fix typos in manpage (Alex Efros). |
| 7 | - increase timing even further, for the ever-overloaded cpan-tester |
70 | - increase timing even further, for the ever-overloaded cpan-tester |
| 8 | machines, and to the detriment of everybody else who wants a fast |
71 | machines, and to the detriment of everybody else who wants a fast |
