| 1 | Revision history for Perl extension EV and the embedded libev library |
1 | Revision history for Perl extension EV |
|
|
2 | |
|
|
3 | Changes marked with (libev) are changes in libev, and have more |
|
|
4 | documentation in the libev Changes file. |
|
|
5 | |
|
|
6 | 4.00 |
|
|
7 | - fix a bug where inotify usage would parse the same event |
|
|
8 | multiple times, causing various forms of breakage. |
|
|
9 | - greatly reduce stack usage for inotify (8kb to <0.5kb). |
|
|
10 | - renamed loop_verify to verify, loop_count to iteration and |
|
|
11 | loop_depth to depth. |
|
|
12 | - expose ev_depth and ev_verify via the XS API. |
|
|
13 | - (libev) ev_embed_stop did not correctly stop the watcher. |
|
|
14 | - (libev) disable poll backend on AIX. |
|
|
15 | - (libev) rename EV_TIMEOUT to EV_TIMER. |
|
|
16 | - (libev) add section on accept() problems to the manpage. |
|
|
17 | - (libev) no child watchers on win32. |
|
|
18 | - make code more aliasing compliant, in case perl is ever |
|
|
19 | translated to C. |
|
|
20 | - ported to minix 3.1.7. |
|
|
21 | |
|
|
22 | 3.9 Thu Dec 31 07:59:59 CET 2009 |
|
|
23 | - disable t/07* under automatic testing. |
|
|
24 | - increase t/09* timeouts as netbsd has *horribly* broken |
|
|
25 | select/kevent that *usually* sleep >>0.6s instead of 0.3. |
|
|
26 | - add constants EV::FLAG_NOSIGFD, EV::FLAG_NOINOTIFY and |
|
|
27 | EV::BACKEND_ALL. |
|
|
28 | - (libev) signalfd is no longer used by default. |
|
|
29 | - (libev) backport inotify code to C89. |
|
|
30 | - (libev) inotify file descriptors could leak into child processes. |
|
|
31 | - (libev) ev_stat watchers could keep an errornous extra ref on the loop. |
|
|
32 | - (libev) take advantage of inotify_init1, if available. |
|
|
33 | - (libev) the signal handling pipe wasn't always initialised under windows. |
|
|
34 | - changed minimum glibc requirement from glibc 2.9 to 2.7, for signalfd. |
|
|
35 | - (libev) only replace ev_stat.prev when we detect an actual difference. |
|
|
36 | |
|
|
37 | 3.8 Sun Aug 9 15:30:10 CEST 2009 |
|
|
38 | - implement $loop->signal/signal_ns. |
|
|
39 | - (libev) incompatible change: do not necessarily reset signal |
|
|
40 | handler to SIG_DFL when a sighandler is stopped. |
|
|
41 | - (libev) ev_default_destroy did not properly free or zero some |
|
|
42 | members, potentially causing crashes and memory corruption on |
|
|
43 | repated ev_default_destroy/ev_default_loop calls. |
|
|
44 | - (libev) take advantage of signalfd on GNU/Linux systems. |
|
|
45 | - (libev) document that the signal mask might be in an unspecified |
|
|
46 | state when using libev's signal handling. |
|
|
47 | - (libev) take advantage of some GNU/Linux calls to set |
|
|
48 | cloexec/nonblock on fd creation, to avoid race conditions. |
|
|
49 | - implement internal glue code to interface more efficiently |
|
|
50 | with AnyEvent. |
|
|
51 | |
|
|
52 | 3.7 Fri Jul 17 16:49:16 CEST 2009 |
|
|
53 | - add EV::loop_verify and EV::loop_depth. |
|
|
54 | - use output filehandle for i/o watchers waiting |
|
|
55 | for EV_WRITE, otherwise input filehandle. |
|
|
56 | - use common schmorp.h header. |
|
|
57 | - add EV::Timer->remaining. |
|
|
58 | - allow for subclassing of EV::Loop (see the exciting |
|
|
59 | EV::Loop::Async module for an example). |
|
|
60 | - added EV::invoke_pending and EV::pending_count. |
|
|
61 | |
|
|
62 | - (libev) ev_unloop and ev_loop wrongly used a global variable to |
|
|
63 | exit loops, instead of using a per-loop variable. |
|
|
64 | - (libev) the ev_set_io_collect_interval interpretation has changed. |
|
|
65 | - add new functionality: EV::invoke_pending, EV::pending_count. |
|
|
66 | - add $timer->remaining. |
|
|
67 | - add EV::loop_depth. |
|
|
68 | - (libev) calling unloop in fork/prepare watchers will no longer poll |
|
|
69 | for new events. |
|
|
70 | - (libev) use GetSystemTimeAsFileTime instead of _timeb on windows, |
|
|
71 | for slightly higher accuracy. |
|
|
72 | - (libev) actually 0-initialise struct sigaction when installing |
|
|
73 | signals. |
|
|
74 | |
|
|
75 | 3.6 Tue Apr 28 02:50:37 CEST 2009 |
|
|
76 | - keepalive(0) could decrease the refcount of the loop |
|
|
77 | permanently. |
|
|
78 | - add ev_suspend/ev_resume, also make ev_now_update |
|
|
79 | accessible via the XS API. |
|
|
80 | - most EV::<eventtype> constants were missing and have been |
|
|
81 | added :/. |
|
|
82 | - add EV::VERSION_MAJOR/VERSION_MINOR constants. |
|
|
83 | - (libev) multiple timers becoming ready within an event loop |
|
|
84 | iteration will be invoked in the "correct" order now. |
|
|
85 | - (libev) do not leave the event loop early just because we have |
|
|
86 | no active watchers, fixing a problem when embedding a kqueue |
|
|
87 | loop that has active kernel events but no registered watchers |
|
|
88 | (reported by blacksand blacksand). |
|
|
89 | - (libev) correctly zero the idx values for arrays, so destroying |
|
|
90 | and reinitialising the default loop actually works (patch by |
|
|
91 | Malek Hadj-Ali). |
|
|
92 | - (libev) new EV::CUSTOM revents flag for use by applications. |
|
|
93 | - (libev) add documentation section about priorites. |
|
|
94 | - (libev) add a glossary to the dcoumentation. |
|
|
95 | - (libev) extend the ev_fork description slightly. |
|
|
96 | - (libev) optimize a jump out of call_pending. |
|
|
97 | - t/03_keepalive could fail when there was no actual error. |
|
|
98 | |
|
|
99 | 3.53 Sun Feb 15 02:38:20 CET 2009 |
|
|
100 | - (libev) on win32, the event loop creation could randomly fail |
|
|
101 | due to an initialised variable having the wrong value. |
|
|
102 | - (libev) probe for CLOCK_REALTIME support at runtime as well and |
|
|
103 | fall back to gettimeofday if there is an error, to support older |
|
|
104 | operating systems with newer header files/libraries. |
|
|
105 | - prefer gettimeofday over clock_gettime by default. |
|
|
106 | |
|
|
107 | 3.52 Wed Jan 7 21:46:14 CET 2009 |
|
|
108 | - (libev) fix some issues in the select backend when in fd_set mode. |
|
|
109 | - (libev) due to a thinko, instead of disabling everything but |
|
|
110 | select on the borked OS X platform, everything but select was |
|
|
111 | allowed (reported by Emanuele Giaquinta). |
|
|
112 | - (libev) actually verify that local and remote port are matching in |
|
|
113 | libev's socketpair emulation, which makes denial-of-service |
|
|
114 | attacks harder (but not impossible - it's windows). Make sure |
|
|
115 | it even works under vista, which thinks that getpeer/sockname |
|
|
116 | should return fantasy port numbers. |
|
|
117 | |
|
|
118 | 3.51 Wed Dec 24 23:01:59 CET 2008 |
|
|
119 | - do not cache the arguments passed to callbacks if the refcount |
|
|
120 | indicates that the callback has stolen them. |
|
|
121 | - (libev) try to avoid librt on GNU/Linux. |
|
|
122 | - (libev) check port_getn return value dfferently, might potentially |
|
|
123 | avoid problems. |
|
|
124 | - (libev) fix a bug with stat watchers possibly causing freezes. |
|
|
125 | - (libev) work around OS X 10.5 breaking poll, now select is the |
|
|
126 | only viable choice left on that pile of garbage. |
|
|
127 | - play tester whore: disable some tests that typically fail only |
|
|
128 | on cpan tester machines. |
|
|
129 | |
|
|
130 | 3.49 Wed Nov 19 11:26:53 CET 2008 |
|
|
131 | - fix typos in manpage (Alex Efros). |
|
|
132 | - increase timing even further, for the ever-overloaded cpan-tester |
|
|
133 | machines, and to the detriment of everybody else who wants a fast |
|
|
134 | make test. |
|
|
135 | - possible 5.6 compatibility. |
|
|
136 | - (libev) use inotify even on buggy kernels, but don't rely on it. |
|
|
137 | - (libev) use inotify only as an added hint on network filesystems. |
|
|
138 | |
|
|
139 | 3.48 Thu Oct 30 09:09:48 CET 2008 |
|
|
140 | - (libev) use a generation counter to detect spurious epoll events |
|
|
141 | and recreate the kernel in such a case (sorry, it's slow, but |
|
|
142 | I didn't design epoll...). |
|
|
143 | - (libev) optimise away an EPOLL_CTL_ADD/MOD combo in the epoll |
|
|
144 | backend in some cases. |
|
|
145 | - (libev) use memset to initialise most arrays now and do away |
|
|
146 | with the init functions. |
|
|
147 | - speed up event callback invocation further (~15%), by |
|
|
148 | not creating the object reference each time. |
|
|
149 | - EV::sleep was documented, but not implemented. fun. |
|
|
150 | |
|
|
151 | 3.45 Tue Oct 21 22:20:39 CEST 2008 |
|
|
152 | - (libev) disable inotify usage on linux <2.6.25 (kernel bug). |
|
|
153 | - (libev) ev_embed will now automatically follow fork. |
|
|
154 | - (libev) ev_once will now pass both io and timeout events to |
|
|
155 | the callback when both happen concurrently, instead of giving |
|
|
156 | one precedence. |
|
|
157 | - install C library documentation as EV::libev manpage (sorry, oesi). |
|
|
158 | - (libev) fix a minor performance bug in ev_stat handling. |
|
|
159 | |
|
|
160 | 3.44 Mon Sep 29 05:16:31 CEST 2008 |
|
|
161 | - (libev) add EV::now_update and $loop->now_update. |
|
|
162 | |
|
|
163 | 3.431 Sun Jul 13 00:19:02 CEST 2008 |
|
|
164 | - made the EV::embed callback optional again. reported by |
|
|
165 | Vladimir Timofeev. |
|
|
166 | |
|
|
167 | 3.43 Tue Jul 8 20:56:31 CEST 2008 |
|
|
168 | - disabled warnings in EV.pm, the CHECK issue is now understood |
|
|
169 | and harmless. |
|
|
170 | |
|
|
171 | 3.42 Mon May 26 07:36:46 CEST 2008 |
|
|
172 | - (libev) work around another bug in windows perls and |
|
|
173 | windows itself: failed connects do NOT set read or write |
|
|
174 | flags in select, but this version of libev will. |
| 2 | |
175 | |
| 3 | 3.41 Thu May 22 01:39:40 CEST 2008 |
176 | 3.41 Thu May 22 01:39:40 CEST 2008 |
| 4 | - (libev) fix many heap-related bugs (timers, periodics). |
177 | - (libev) fix many heap-related bugs (timers, periodics). |
| 5 | - (libev) improve timing stability of timers and periodics. |
178 | - (libev) improve timing stability of timers and periodics. |
|
|
179 | - expose ev_loop_verify to perl code. |
| 6 | - clarify documentation for periodic reschedule callbacks. |
180 | - clarify documentation for periodic reschedule callbacks. |
| 7 | - verify that the passed callback argument is indeed a code |
181 | - verify that the passed callback argument is indeed a code |
| 8 | reference, for earlier error reporting and a nice calling |
182 | reference, for earlier error reporting and a nice calling |
| 9 | speed increase (as well as saving memory). |
183 | speed increase (as well as saving memory). |
| 10 | |
184 | |
