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

Comparing Coro/Changes (file contents):
Revision 1.372 by root, Wed Sep 24 07:30:12 2008 UTC vs.
Revision 1.563 by root, Thu May 12 23:24:27 2011 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: channel->maxsize(newsize)?
5 actor model (which is cool in a lot of important aspects (failures!), 5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx
6 but very lacking in others (higher level ipc)).
7 6
84.748 7TODO:mg_virtual check
8TODO:detah on refcnt
9TODO:slf_xx_yy=>slf_yy_xx
10 - croak when a coro that is being cancelled tries to block
11 (e.g. while executing a guard block), instead of crashing or
12 deadlocking.
13 - use a more robust and also faster method to identify Coro::State
14 objects - speeds up everything a bit.
15 - implement a Coro->safe_cancel method that might fail, but
16 cancels in a "safer" way if it succeeds.
17 - implement Coro->cancel in XS for a 20% speed improvement, and to
18 be able to implement mutual cancellation.
19 - speed up context switches by a percent or two by more efficiently
20 allocating context stack entries.
21 - implement Coro->join and Coro->on_destroy in XS for a speedup and
22 a reduction in memory use.
23 - cancelling a coro while it itself is cancelling another coro is
24 now supported and working, instead of triggering an assertion.
25 - be a bit more crash-resistant when calling (buggy) on_destroy
26 callbacks (best effort).
27 - move on_destroy into the slf_frame, to allow extension slf
28 functions to have destructors.
29 - get rid if coro refcounting - simply crash in other interpreter
30 threads by nulling the pointers on clone.
31
325.372 Wed Feb 23 06:14:30 CET 2011
33 - apparently mingw doesn't provide a working gettimeofday, try to
34 work around that by relying on Time::HiRes (indirectly brought to
35 my attention by Max Maischein).
36 - fix some portability issues when Time::HiRes was used.
37
385.371 Mon Feb 21 14:36:08 CET 2011
39 - backport to windows process emulation code again.
40
415.37 Sat Feb 19 07:49:44 CET 2011
42 - add a big "Coro thread life cycle" section to "man Coro".
43 - try a tentative workaround against the breakage that 5.13 has
44 introduced without depreciation period. sigh.
45 - no longer use Time::HiRes if gettimeofday is available, which
46 saves quite a lot of memory.
47
485.36 Sun Feb 13 05:33:41 CET 2011
49 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
50 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
51 method call.
52 - undocument Coro::Timer::sleep and obsolete whole module.
53 - optimise Coro::Timer::timeout memory and cpu usage.
54 - slightly updated Coro::Intro for recent changes.
55 - do not initialise PL_dirty anymore.
56
575.25 Thu Nov 11 01:08:39 CET 2010
58 - try a different approach on netbsd - netbsd 5 finally has marginally
59 working pthreads, but still broken ucontext/sigaltstack.
60 - openbsd 4.8 finally got their act together, Coro works out of the box
61 with asm, setjmp and pthreads (no change, just informational).
62
635.24 Sat Oct 23 11:27:12 CEST 2010
64 - port to the EV 4.0 API.
65 - work around bugs in mingw32, making strawberry perl work
66 out of the box.
67 - correctly modify Coro::AIO function prototypes
68 so that they reflect the "no optional parameters" rule.
69 - "ported" libcoro to C++.
70
715.23 Mon May 17 18:50:42 CEST 2010
72 - be more resistant to ordering changes when initialising
73 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
74 Waldorf).
75 - document that perl 5.12 deliberately removed support for cloning.
76
775.22 Wed Apr 14 03:55:35 CEST 2010
78 - correctly return udnef on errors in Coro::Handle::read/write
79 (testcase by Marc Mims).
80 - convert Coro::Util into a "perl compatibility wrapper" - the functions
81 are less useful now, but are drop-in replacements for existing
82 functions, listing better alternatives in the documentation. This also
83 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
84 with Coro::Util::inet_aton.
85 - do not override $Coro::idle unconditionally in Coro.pm, as other
86 modules could have provided their own idle coro already
87 (for exmaple, Coro::AnyEvent).
88 - fix Coro::Util::gethost* functions.
89 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
90
915.21 Wed Dec 16 07:19:51 CET 2009
92 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
93 together.
94 - add some examples on how to combine other event loops with Coro in
95 Coro::AnyEvent, and how to run it (and not to block). Seems to be
96 the most common source of confusion.
97 - try to catch people naively blocking in an event callback.
98 - work around the perl filehandle bug issue in conjunction with
99 older common::sense (as indirectly pointed out by ZSystem).
100 - clarify the "not from signal handlers" section.
101
1025.2 Sun Oct 4 14:54:24 CEST 2009
103 - Coro::Storable destroyed the prototypes of the functions it wrapped.
104 - export rouse_cb and rouse_wait by default now.
105 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
106 - new method $state->swap_sv.
107 - added section on "windows process emulation" to the manpage,
108 after a not-so-fruitful (nor-friendly) "discussion" with chip
109 salzenberg (discussion implies arguments, but his only arguments
110 were ad-hominems, one wonders why he started it in the first
111 place). I hope this explains it well enough for him to understand,
112 and maybe well enough for others to understand.
113 - use common::sense everywhere now.
114 - idle callbacks are no longer supported, use idle coros instead.
115 - print a thread listing when a deadlock is detected.
116
1175.17 Sat Aug 22 23:09:31 CEST 2009
118 - work around a bug in the perl debugger causing crashes
119 when running under the debugger by marking _pool_handler as nodebug.
120 - speed up Coro::async considerably.
121 - try some hacks to get netbsd to work "more often" - their broken
122 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
123 - convert Coro to AE 5.0 API.
124
1255.162 Tue Jul 28 04:04:03 CEST 2009
126 - perl 5.8.2 is now minimum requirement.
127 - skip t/19_handle.t on broken windows perls.
128
1295.161 Wed Jul 22 04:47:38 CEST 2009
130 - Coro::AnyEvent::poll could have a different prototype when EV was
131 used as backend (analyzed by Tatsuhiko Miyagawa).
132 - Coro::AnyEvent errornously initialised the event loop when loaded,
133 not on demand.
134 - try to workaround rare */t/01_unblock.t failures.
135
1365.16 Tue Jul 21 01:44:37 CEST 2009
137 - Coro::AnyEvent failed to hook into the event loop
138 when no threads had been readied between detecting
139 the event loop and actually running it.
140 - considerably speed up Coro::Select by taking avdantage
141 of AnyEvent > 4.8 and some other optimisations.
142 - implement paragraph readline mode in Coro::Handle
143 (based on patches by Zsbán Ambrus).
144 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
145 and analyzed by Yasuhiro MATSUMOTO).
146 - clarified libcoro license and copyright.
147 - someone stole my EXTRA_META!!!
148 - implement Coro::Select::patch_pp_sselect and it's brother,
149 for hardcode select overriding.
150
1515.151 Mon Jul 6 05:41:57 CEST 2009
152 - backport to windows process emulation code again (patch by
153 Yasuhiro MATSUMOTO).
154 - slightly update Coro::MakeMaker.
155
1565.15 Tue Jun 30 10:28:06 CEST 2009
157 - deprecate Coro::Socket, document how to get ipv6 support via
158 AnyEvent::Socket instead.
159 - implement signal->wait ($cb) interface, similar to semaphores.
160 - work around SvOK not supporting getmagic, so we have to getmagic
161 to test for undef :/ (reported by Matthias Waldorf).
162 - load Coro::AnyEvent in all modules using AnyEvent.
163 - work around perl corrupting our internal data structures,
164 reported by Tokuhiro Matsuno.
165 - enable per-coroutine real and cpu time gathering
166 (Coro::State::enable_times).
167
1685.14 Wed Jun 24 01:37:48 CEST 2009
169 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
170 as this seems to have been a difficult concept (poll, sleep, idle,
171 idle_upto).
172 - add Coro::AnyEvent::readable/writable functions.
173 - clarify Coro::EV/Event/AnyEvent manpages.
174 - free per-thread global scalars in the thread calling ->cancel, to
175 avoid crashes when $_, $@ etc., are magical but some of those
176 had already been freed.
177 - "unexperimentalise" the callback interface for Coro::Semaphore.
178 - speed up ready queue management/context switching by using a linked
179 list instead of an array (~5-10%).
180 - implement "watch" command in Coro::Debug shells.
181 - for fun, implement time-slicing as an example in the manpage.
182 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
183 Coro::Event, use the normal AnyEvent handling instead of dieing
184 (the same is true for Coro::Handle).
185 - properly document Coro::EV::timed_io_once.
186 - avoid unneccessary ->cancel calls in Coro::Handle.
187 - maybe make it work on mingw32 with win32 backend
188 (based on patch by Yasuhiro Matsumoto).
189
1905.132 Fri May 29 09:00:39 CEST 2009
191 - do not keep a reference to the argument itself in
192 Coro::Semaphore::guard, as it could change later.
193 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
194 should support a prepare callback.
195
1965.131 Mon Mar 16 23:20:37 CET 2009
197 - implement and document Coro->suspend, Coro->resume.
198 - fix Coro::Select implementation to not (often) close
199 the passed file descriptors (testcase provided by pippijn).
200
2015.13 Mon Dec 15 21:51:42 CET 2008
202 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
203 - don't set diehook to C<undef> but instead to NULL, to avoid
204 spurious warnings.
205 - fix a lot of bugs in Coro::SemaphoreSet.
206 - Coro::SemaphoreSet will less often create a semaphore needlessly.
207 - add Coro::SemaphoreSet::count and wait methods.
208 - take advantage of the new Guard module.
209 - deprecate Coro::guard.
210 - try to fix the dreaded 01_unblock tests once more. I hate it when
211 testsuites need more fixing than the code they are supposed to test.
212 - croak in more cases when a required callback isn't resolvable.
213 - fix some minor issues in Coro::State->call/eval.
214 - use current coroutine context instead of a temporary one
215 when temporarily switching to another coroutine.
216 - do not call C-level on_destroy handlers during global destruction,
217 to avoid needless segfaults.
218
2195.12 Sun Dec 7 13:30:38 CET 2008
220 - add default config for MirOS, which seems to be a bug-to-bug
221 compatible fork of openbsd ("world domination by releasing
222 openbsd cvs before the openbsd folks do it" or so :).
223 - free_padlist did destroy the names pad, not good, but didn't
224 seem to bother perl - this could fix issues such as eval ""
225 inside a function called from multiple coroutines.
226 - use a different method to detect destruction time.
227 - be more careful when freeing padlists just before global
228 destruction.
229 - fixed and expanded the call/cc example.
230 - renamed _terminate to _coro_run.
231 - new method Coro::Channel->shutdown.
232 - try pthreads on openbsd <4.4 (broken sigaltstack, will
233 pthreads fare better?).
234 - be less picky about destroying "a" running coroutine.
235
2365.11 Tue Nov 25 21:49:05 CET 2008
237 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
238 working code with failed assertions, introducing more bugs than
239 it fixes, requiring elaborate workarounds :(
240
2415.1 Mon Nov 24 08:54:59 CET 2008
242 - wrote a small introductory tutorial - Coro::Intro.
243 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
244 - Coro::Select did errornously dup the file descriptors
245 and didn't work with all AnyEvent backends.
246 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
247 LWP data transfers.
248 - disassociate c contexts from coro objects - this is agruably more
249 correct, but mostly allows sharing of cctxs between coro and state
250 objects, for added memory savings and speed increases.
251 - bumped $Coro::POOL_RSS up to 32kb by default.
252 - no longer set the optype to OP_CUSTOM, as B::* understandably
253 doesn't like this very much (and we *are* a type of entersub).
254 - implement state cloning, just to prove that call/cc can be done.
255 - automatically load Coro::AnyEvent in Coro::Handle.
256 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
257 do not have this method (and don't need it either).
258 - speed up generic anyevent methods in Coro::Handle by using rouse
259 callbacks.
260 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
261 others. It also makes the debugger happier, as you can trace
262 through the idle threads now.
263 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
264 - eg/event was pretty broken.
265 - better 5.8.6 compatibility.
266
2675.0 Thu Nov 20 10:35:05 CET 2008
268 - NEW ARCHITECTURE: use the latest 4.x version if you experience
269 stability issues.
270 - bump API version to 7 - all dependents must be recompiled.
271 - removed timed_* functions - they were not being used anyways
272 and should be replaced by a more generic mechanism -
273 and were annoying to support anyways :)
274 - removed SemaphoreSet's waiter method - use sem method instead.
275 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
276 - async_pool did free a scalar value twice
277 ("Attempt to unreference...").
278 - fix a longstanding bug where calling terminate on a coro that
279 was waiting for a semaphore that was just becoming available
280 would cause a deadlock (semaphore would get into a state where
281 it was available but waiters were still blocked).
282 - calling throw on a coroutine that is waiting for a semaphore will
283 no longer make it acquire the semaphore (and thus leak a count).
284 - perl's process emulation is now not even theoretically supported
285 anymore.
286 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
287 conversion of callback-style to blocking-style.
288 - new methods $coro->schedule_to and ->cede_to, to specifically
289 schedule or cede to a specific coroutine.
290 - new function Coro::Semaphore::wait.
291 - use named constants in Coro::Channel (Richard Hundt).
292 - directly patch the entersub opcode calling SLF functions (cede,
293 transfer and so on). this does speed up context switching, but
294 more importanly, it frees us from the hardcoded behaviour of
295 entersub, so we might actually be able to return something from
296 those functions and atcually create new ones.
297 - take advantage of __builtin_frame_address on gcc.
298 - expose THX in coroapi (not sure whether this was a wise decision,
299 as "threaded" perls are running at half speed anyways).
300 - implement execute_slf (schedule-like-function) interface that makes
301 it possible to implement schedule-like-functions in XS.
302 - use new SLF interface to massively speed up Coro::EV by roughly a
303 factor of two.
304 - used new SLF interface to massively speed up Coro::Semaphore by a
305 factor of three.
306 - used new SLF interface to speed up Coro::AIO by roughly a factor of
307 four and reduce its memory usage considerably.
308 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
309 for a nice speedup and vastly more correct behaviour. Also implement
310 a new method "sem" to get at the underlying semaphore object.
311 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
312 (in comparison) 20-40% speedup.
313 - used new SLF interface to reimplement Coro::Signal gaining
314 some unknown (because I was too lazy), but certain, speedup, and also
315 making signals reliable for the first time.
316 - used new SLF interface and other optimisations to speed up async_pool
317 by a factor of two. It also doesn't rely on perl's exception mechanism
318 to exit anymore. The overhead for terminating an async_pool, coro over
319 a normal async is now very small.
320 - sped up coroutine creation/destruction by 40%.
321 - forgot to include Coro/libcoro/README in the dist for all these years.
322 - work around a freebsd pthreads bug (manual testcancel is required as
323 pthread_cond_wait isn't a cancellation point on freebsd).
324 - use new rouse functions to speed up and simplify Coro::BDB.
325 - make "prefer perl native functions" work with threaded perls.
326 - condense Coro::Debug ps output, hint at v and w flags.
327 - (libcoro) lots of minor cleanups and portability improvements.
328
3294.914 Wed Nov 19 12:54:18 CET 2008
330 - fix a disastrous bug in the readline optimisation
331 introduced in 4.801.
332
3334.913 Sat Nov 15 07:58:28 CET 2008
334 - async_pool did free a scalar value twice
335 ("Attempt to unreference...").
336
3374.912 Thu Nov 13 18:31:23 CET 2008
338 - minor cleanups.
339 - use much larger stacks on linux and perl < 5.8.8.
340 - Coro::Debug::new_unix_server did not unlink the socket
341 when destroyed.
342
3434.911 Tue Nov 11 04:26:17 CET 2008
344 - "port" to threaded perls.
345
3464.91 Mon Nov 10 05:36:38 CET 2008
347 - the ->throw exception object no longer leaks.
348 - creating a new cctx leaked a scopestack entry (memleak).
349 - new coroutines didn't get created with a zero flags field
350 (unknown impact).
351 - calling ->throw on a not-yet-started coroutine should now work
352 instead of being ignored.
353 - ->throw is now supported on Coro::State objects.
354 - clean up cctx creation code a bit.
355 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
356
3574.9 Sat Nov 8 17:45:27 CET 2008
358 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
359 - (libcoro) no longer leak threads in the experimental pthread backend,
360 also speed it up considerably.
361 - (libcoro) do not rely on makecontext passing void *'s unscathed.
362 - fix compiletime dependencies on libcoro in the Makefile.
363 - cctx_count wasn't always updated properly.
364 - Coro::State::cctx_stacksize wasn't applied correctly.
365 - new function Coro::State::cctx_max_idle.
366 - the default max number of idle C contexts is now 4.
367 - (libcoro) try harder to get _setjmp/_longjmp.
368 - (libcoro) cleanup and extend the libcoro API to officially
369 allow the creation of empty source contexts.
370 - very experimental workaround for the totally broken netbsd platform.
371 - tried to hack around openbsd bugs.
372
3734.804 Wed Nov 5 16:36:00 CET 2008
374 - Coro::Debug::new_unix_server would not create a non-blocking listening
375 socket, sometimes causing freezes.
376 - (libcoro) fix misaligned stack points for setjmp and assembly
377 methods, which can cause crashes on x86/x86_64 with a sufficiently
378 aggressive compiler.
379 - new function: Coro::Debug::new_tcp_server.
380 - move ->throw into the Coro class because it only works on coro objects.
381
3824.803 Mon Nov 3 17:16:12 CET 2008
383 - (libcoro) use a global asm statement to become independent of gcc
384 otpimisations for CORO_ASM (thanks to pippijn for the idea).
385 - try to workaround yet another broken bsd, this time dragonfly.
386
3874.802 Thu Oct 30 10:56:12 CET 2008
388 - support -fno-omit-frame-pointer on x86 with the assembly method.
389 - tune 01_unblock.t tests a bit.
390
3914.801 Wed Oct 22 18:33:37 CEST 2008
392 - improve readline speed for very long "lines".
393 - backport to 5.8.8.
394
3954.8 Thu Oct 2 13:34:40 CEST 2008
396 - new function Coro::AIO::aio_wait.
397 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
398 - greatly speed up and reduce memory usage of Coro::AIO requests.
399 - implement some other µ-optimisations.
400
4014.749 Mon Sep 29 14:40:12 CEST 2008
402 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
403
4044.748 Sat Sep 27 14:03:19 CEST 2008
405 - implement, but do not document, PerlIO::cede(granularity).
9 - Coro::Storable forgot to wrap Storable::pstore. 406 - Coro::Storable forgot to wrap Storable::pstore.
407 - work around the multitude of leaks and memory corruption
408 bugs in PerlIO::via by using our own C-level perliol. As a side
409 effect, Coro::Storable is now much, much, much faster.
10 410
114.747 Tue Sep 23 01:59:41 CEST 2008 4114.747 Tue Sep 23 01:59:41 CEST 2008
12 - fix a per-cv memleak (one empty array was leaked per 412 - fix a per-cv memleak (one empty array was leaked per
13 code reference). 413 code reference).
14 - avoid a crash in coro->call|rss when the coroutine was already 414 - avoid a crash in coro->call|rss when the coroutine was already

Diff Legend

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