ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
(Generate patch)

Comparing Coro/Changes (file contents):
Revision 1.427 by root, Tue Nov 18 05:51:38 2008 UTC vs.
Revision 1.518 by root, Tue Jul 28 02:04:21 2009 UTC

… …
2 2
3TODO: should explore PerlIO::coroaio (perl leaks like hell). 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!), 5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)). 6 but very lacking in others (higher level ipc)).
7TODO: unready_all
8TODO: myhttpd header parsing
9TODO: channel->maxsize(newsize)?
7 10
8TODO: testcancel 115.162 Tue Jul 28 04:04:03 CEST 2009
9TODO: aio 12 - perl 5.8.2 is now minimum requirement.
10TODO: Coro::Semaphore 13 - skip t/19_handle.t on broken windows perls.
11TODO: trow disturbs cleanup path inside pp_slf? 14
12TODO: throw special-case in transfer? 155.161 Wed Jul 22 04:47:38 CEST 2009
13TODO: document throw limitations 16 - Coro::AnyEvent::poll could have a different prototype when EV was
145.0 17 used as backend (analyzed by Tatsuhiko Miyagawa).
18 - Coro::AnyEvent errornously initialised the event loop when loaded,
19 not on demand.
20 - try to workaround rare */t/01_unblock.t failures.
21
225.16 Tue Jul 21 01:44:37 CEST 2009
23 - Coro::AnyEvent failed to hook into the event loop
24 when no threads had been readied between detecting
25 the event loop and actually running it.
26 - considerably speed up Coro::Select by taking avdantage
27 of AnyEvent > 4.8 and some other optimisations.
28 - implement paragraph readline mode in Coro::Handle
29 (based on patches by Zsbán Ambrus).
30 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
31 and analyzed by Yasuhiro MATSUMOTO).
32 - clarified libcoro license and copyright.
33 - someone stole my EXTRA_META!!!
34 - implement Coro::Select::patch_pp_sselect and it's brother,
35 for hardcode select overriding.
36
375.151 Mon Jul 6 05:41:57 CEST 2009
38 - backport to windows process emulation code again (patch by
39 Yasuhiro MATSUMOTO).
40 - slightly update Coro::MakeMaker.
41
425.15 Tue Jun 30 10:28:06 CEST 2009
43 - deprecate Coro::Socket, document how to get ipv6 support via
44 AnyEvent::Socket instead.
45 - implement signal->wait ($cb) interface, similar to semaphores.
46 - work around SvOK not supporting getmagic, so we have to getmagic
47 to test for undef :/ (reported by Matthias Waldorf).
48 - load Coro::AnyEvent in all modules using AnyEvent.
49 - work around perl corrupting our internal data structures,
50 reported by Tokuhiro Matsuno.
51 - enable per-coroutine real and cpu time gathering
52 (Coro::State::enable_times).
53
545.14 Wed Jun 24 01:37:48 CEST 2009
55 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
56 as this seems to have been a difficult concept (poll, sleep, idle,
57 idle_upto).
58 - add Coro::AnyEvent::readable/writable functions.
59 - clarify Coro::EV/Event/AnyEvent manpages.
60 - free per-thread global scalars in the thread calling ->cancel, to
61 avoid crashes when $_, $@ etc., are magical but some of those
62 had already been freed.
63 - "unexperimentalise" the callback interface for Coro::Semaphore.
64 - speed up ready queue management/context switching by using a linked
65 list instead of an array (~5-10%).
66 - implement "watch" command in Coro::Debug shells.
67 - for fun, implement time-slicing as an example in the manpage.
68 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
69 Coro::Event, use the normal AnyEvent handling instead of dieing
70 (the same is true for Coro::Handle).
71 - properly document Coro::EV::timed_io_once.
72 - avoid unneccessary ->cancel calls in Coro::Handle.
73 - maybe make it work on mingw32 with win32 backend
74 (based on patch by Yasuhiro Matsumoto).
75
765.132 Fri May 29 09:00:39 CEST 2009
77 - do not keep a reference to the argument itself in
78 Coro::Semaphore::guard, as it could change later.
79 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
80 should support a prepare callback.
81
825.131 Mon Mar 16 23:20:37 CET 2009
83 - implement and document Coro->suspend, Coro->resume.
84 - fix Coro::Select implementation to not (often) close
85 the passed file descriptors (testcase provided by pippijn).
86
875.13 Mon Dec 15 21:51:42 CET 2008
88 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
89 - don't set diehook to C<undef> but instead to NULL, to avoid
90 spurious warnings.
91 - fix a lot of bugs in Coro::SemaphoreSet.
92 - Coro::SemaphoreSet will less often create a semaphore needlessly.
93 - add Coro::SemaphoreSet::count and wait methods.
94 - take advantage of the new Guard module.
95 - deprecate Coro::guard.
96 - try to fix the dreaded 01_unblock tests once more. I hate it when
97 testsuites need more fixing than the code they are supposed to test.
98 - croak in more cases when a required callback isn't resolvable.
99 - fix some minor issues in Coro::State->call/eval.
100 - use current coroutine context instead of a temporary one
101 when temporarily switching to another coroutine.
102 - do not call C-level on_destroy handlers during global destruction,
103 to avoid needless segfaults.
104
1055.12 Sun Dec 7 13:30:38 CET 2008
106 - add default config for MirOS, which seems to be a bug-to-bug
107 compatible fork of openbsd ("world domination by releasing
108 openbsd cvs before the openbsd folks do it" or so :).
109 - free_padlist did destroy the names pad, not good, but didn't
110 seem to bother perl - this could fix issues such as eval ""
111 inside a function called from multiple coroutines.
112 - use a different method to detect destruction time.
113 - be more careful when freeing padlists just before global
114 destruction.
115 - fixed and expanded the call/cc example.
116 - renamed _terminate to _coro_run.
117 - new method Coro::Channel->shutdown.
118 - try pthreads on openbsd <4.4 (broken sigaltstack, will
119 pthreads fare better?).
120 - be less picky about destroying "a" running coroutine.
121
1225.11 Tue Nov 25 21:49:05 CET 2008
123 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
124 working code with failed assertions, introducing more bugs than
125 it fixes, requiring elaborate workarounds :(
126
1275.1 Mon Nov 24 08:54:59 CET 2008
128 - wrote a small introductory tutorial - Coro::Intro.
129 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
130 - Coro::Select did errornously dup the file descriptors
131 and didn't work with all AnyEvent backends.
132 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
133 LWP data transfers.
134 - disassociate c contexts from coro objects - this is agruably more
135 correct, but mostly allows sharing of cctxs between coro and state
136 objects, for added memory savings and speed increases.
137 - bumped $Coro::POOL_RSS up to 32kb by default.
138 - no longer set the optype to OP_CUSTOM, as B::* understandably
139 doesn't like this very much (and we *are* a type of entersub).
140 - implement state cloning, just to prove that call/cc can be done.
141 - automatically load Coro::AnyEvent in Coro::Handle.
142 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
143 do not have this method (and don't need it either).
144 - speed up generic anyevent methods in Coro::Handle by using rouse
145 callbacks.
146 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
147 others. It also makes the debugger happier, as you can trace
148 through the idle threads now.
149 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
150 - eg/event was pretty broken.
151 - better 5.8.6 compatibility.
152
1535.0 Thu Nov 20 10:35:05 CET 2008
15 - NEW ARCHITECTURE: use the latest 4.x version if you experience 154 - NEW ARCHITECTURE: use the latest 4.x version if you experience
16 stability issues. 155 stability issues.
17 - bump API version to 7 - all dependents must be recompiled. 156 - bump API version to 7 - all dependents must be recompiled.
157 - removed timed_* functions - they were not being used anyways
158 and should be replaced by a more generic mechanism -
159 and were annoying to support anyways :)
160 - removed SemaphoreSet's waiter method - use sem method instead.
18 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 161 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
19 - async_pool did free a scalar value twice 162 - async_pool did free a scalar value twice
20 ("Attempt to unreference..."). 163 ("Attempt to unreference...").
21 - fix a longstanding bug where calling terminate on a coro that 164 - fix a longstanding bug where calling terminate on a coro that
22 was waiting for a semaphore that was just becoming available 165 was waiting for a semaphore that was just becoming available
23 would cause a deadlock (semaphore would get into a state where 166 would cause a deadlock (semaphore would get into a state where
24 it was available but waiters were waiting(. 167 it was available but waiters were still blocked).
25 - avoid deadlock in Coro::Channel if maxsize == 1 (Richard Hundt). 168 - calling throw on a coroutine that is waiting for a semaphore will
169 no longer make it acquire the semaphore (and thus leak a count).
26 - perl's process emulation is now not even theoretically supported 170 - perl's process emulation is now not even theoretically supported
27 anymore. 171 anymore.
172 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
173 conversion of callback-style to blocking-style.
174 - new methods $coro->schedule_to and ->cede_to, to specifically
175 schedule or cede to a specific coroutine.
176 - new function Coro::Semaphore::wait.
177 - use named constants in Coro::Channel (Richard Hundt).
28 - directly patch the entersub opcode calling SLF functions (cede, 178 - directly patch the entersub opcode calling SLF functions (cede,
29 transfer and so on). this does speed up context switching, but 179 transfer and so on). this does speed up context switching, but
30 more importanly, it frees us from the hardcoded behaviour of 180 more importanly, it frees us from the hardcoded behaviour of
31 entersub, so we might actually be able to return something from 181 entersub, so we might actually be able to return something from
32 those functions and atcually create new ones. 182 those functions and atcually create new ones.
… …
35 as "threaded" perls are running at half speed anyways). 185 as "threaded" perls are running at half speed anyways).
36 - implement execute_slf (schedule-like-function) interface that makes 186 - implement execute_slf (schedule-like-function) interface that makes
37 it possible to implement schedule-like-functions in XS. 187 it possible to implement schedule-like-functions in XS.
38 - use new SLF interface to massively speed up Coro::EV by roughly a 188 - use new SLF interface to massively speed up Coro::EV by roughly a
39 factor of two. 189 factor of two.
40 - removed timed_down and timed_guard functions - they were not being
41 used anyways and should be replaced by a more generic mechanism -
42 and were annoying to support anyways :)
43 - used new SLF interface to massively speed up Coro::Semaphore by a 190 - used new SLF interface to massively speed up Coro::Semaphore by a
44 factor of three. 191 factor of three.
45 - used new SLF interface to speed up Coro::AIO by roughly a factor of 192 - used new SLF interface to speed up Coro::AIO by roughly a factor of
46 four and reduce its memory usage considerably. 193 four and reduce its memory usage considerably.
194 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
195 for a nice speedup and vastly more correct behaviour. Also implement
196 a new method "sem" to get at the underlying semaphore object.
197 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
198 (in comparison) 20-40% speedup.
199 - used new SLF interface to reimplement Coro::Signal gaining
200 some unknown (because I was too lazy), but certain, speedup, and also
201 making signals reliable for the first time.
202 - used new SLF interface and other optimisations to speed up async_pool
203 by a factor of two. It also doesn't rely on perl's exception mechanism
204 to exit anymore. The overhead for terminating an async_pool, coro over
205 a normal async is now very small.
206 - sped up coroutine creation/destruction by 40%.
47 - forgot to include Coro/libcoro/README in the dist for all these years. 207 - forgot to include Coro/libcoro/README in the dist for all these years.
48 - work around a freebsd pthreads bug (manual testcancel is required as 208 - work around a freebsd pthreads bug (manual testcancel is required as
49 pthread_cond_wait isn't a cancellation point on freebsd). 209 pthread_cond_wait isn't a cancellation point on freebsd).
210 - use new rouse functions to speed up and simplify Coro::BDB.
211 - make "prefer perl native functions" work with threaded perls.
212 - condense Coro::Debug ps output, hint at v and w flags.
213 - (libcoro) lots of minor cleanups and portability improvements.
214
2154.914 Wed Nov 19 12:54:18 CET 2008
216 - fix a disastrous bug in the readline optimisation
217 introduced in 4.801.
50 218
514.913 Sat Nov 15 07:58:28 CET 2008 2194.913 Sat Nov 15 07:58:28 CET 2008
52 - async_pool did free a scalar value twice 220 - async_pool did free a scalar value twice
53 ("Attempt to unreference..."). 221 ("Attempt to unreference...").
54 222

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines