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

Comparing Coro/Changes (file contents):
Revision 1.370 by root, Mon Sep 22 05:40:21 2008 UTC vs.
Revision 1.584 by root, Thu Aug 4 10:37:31 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
84.747 7 - use exponential increase for the readline buffer length in
8 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
9
106.04 Wed Aug 3 17:15:45 CEST 2011
11 - use even more efficient and more compatible condvars for
12 compatibility to AnyEvent 6.x :)
13 - more inconsequential ecb.h updates.
14
156.03 Wed Aug 3 11:41:30 CEST 2011
16 - change how Coro patches AnyEvent condvars for compatibility to
17 AnyEvent 6.x.
18 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
19 hasn't been included.
20
216.02 Wed Jul 13 04:35:19 CEST 2011
22 - "improve portability to Gentoo" - gentoo manages to put perl variables
23 in memory areas that are farther than 2gb apart, which the jit couldn't
24 handle and barfed. now it's just a bit slower on gentoo and similar
25 systems.
26
276.01 Sun Jul 3 12:31:14 CEST 2011
28 - workarounds are good, but the test for whether pthreads are used
29 was not good. thisone should be better.
30 - check differently whether gcc generates cfi instructions itself.
31
326.0 Wed Jun 29 19:43:35 CEST 2011
33 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
34 destroyed and cleaned up automatically (e.g. async { schedule }).
35 - implement a JIT compiler for part of the thread switch code,
36 which gives a 50% speed improvement on threaded perls, and
37 about 4% on non-thraeded perls (so threaded perls now finally
38 reach about half the speed of non-threaded perls).
39 - slightly modernise Coro::Intro, add section about rouse functions.
40 - avoid DEFSV and ERRSV, giving another 10% improvement
41 in thread switching.
42 - Coro::State->is_destroyed is now called is_zombie.
43 - implement a Coro->safe_cancel method that might fail, but
44 cancels in a "safer" way if it succeeds.
45 - add preliminary support for DEBUGGING perls.
46 - get rid of two hash-accesses when initialising a new Coro - this
47 speeds up coro creation by almost a factor of two.
48 - croak when a coro that is being cancelled tries to block
49 (e.g. while executing a guard block), instead of crashing or
50 deadlocking.
51 - use a more robust and also faster method to identify Coro::State
52 objects - speeds up everything a bit.
53 - implement Coro->cancel in XS for a 20% speed improvement, and to
54 be able to implement mutual cancellation.
55 - speed up context switches by a percent or two by more efficiently
56 allocating context stack entries.
57 - implement Coro->join and Coro->on_destroy in XS for a speedup and
58 a reduction in memory use.
59 - cancelling a coro while it itself is cancelling another coro is
60 now supported and working, instead of triggering an assertion.
61 - be a bit more crash-resistant when calling (buggy) on_destroy
62 callbacks (best effort).
63 - move on_destroy into the slf_frame, to allow extension slf
64 functions to have destructors.
65 - get rid if coro refcounting - simply crash in other interpreter
66 threads by nulling the pointers on clone.
67 - simplify warn/die hook handling when loading Coro - the convoluted
68 logic seems to be no longer neccessary.
69 - use libecb instead of our own home-grown gcc hacks.
70 - document alternatives to Coro::LWP. Please use them :)
71 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
72 sigaltstack. Really. wine suffers from it, erlang suffers from it,
73 and it's known since at least 2006.
74
755.372 Wed Feb 23 06:14:30 CET 2011
76 - apparently mingw doesn't provide a working gettimeofday, try to
77 work around that by relying on Time::HiRes (indirectly brought to
78 my attention by Max Maischein).
79 - fix some portability issues when Time::HiRes was used.
80
815.371 Mon Feb 21 14:36:08 CET 2011
82 - backport to windows process emulation code again.
83
845.37 Sat Feb 19 07:49:44 CET 2011
85 - add a big "Coro thread life cycle" section to "man Coro".
86 - try a tentative workaround against the breakage that 5.13 has
87 introduced without depreciation period. sigh.
88 - no longer use Time::HiRes if gettimeofday is available, which
89 saves quite a lot of memory.
90
915.36 Sun Feb 13 05:33:41 CET 2011
92 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
93 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
94 method call.
95 - undocument Coro::Timer::sleep and obsolete whole module.
96 - optimise Coro::Timer::timeout memory and cpu usage.
97 - slightly updated Coro::Intro for recent changes.
98 - do not initialise PL_dirty anymore.
99
1005.25 Thu Nov 11 01:08:39 CET 2010
101 - try a different approach on netbsd - netbsd 5 finally has marginally
102 working pthreads, but still broken ucontext/sigaltstack.
103 - openbsd 4.8 finally got their act together, Coro works out of the box
104 with asm, setjmp and pthreads (no change, just informational).
105
1065.24 Sat Oct 23 11:27:12 CEST 2010
107 - port to the EV 4.0 API.
108 - work around bugs in mingw32, making strawberry perl work
109 out of the box.
110 - correctly modify Coro::AIO function prototypes
111 so that they reflect the "no optional parameters" rule.
112 - "ported" libcoro to C++.
113
1145.23 Mon May 17 18:50:42 CEST 2010
115 - be more resistant to ordering changes when initialising
116 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
117 Waldorf).
118 - document that perl 5.12 deliberately removed support for cloning.
119
1205.22 Wed Apr 14 03:55:35 CEST 2010
121 - correctly return udnef on errors in Coro::Handle::read/write
122 (testcase by Marc Mims).
123 - convert Coro::Util into a "perl compatibility wrapper" - the functions
124 are less useful now, but are drop-in replacements for existing
125 functions, listing better alternatives in the documentation. This also
126 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
127 with Coro::Util::inet_aton.
128 - do not override $Coro::idle unconditionally in Coro.pm, as other
129 modules could have provided their own idle coro already
130 (for exmaple, Coro::AnyEvent).
131 - fix Coro::Util::gethost* functions.
132 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
133
1345.21 Wed Dec 16 07:19:51 CET 2009
135 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
136 together.
137 - add some examples on how to combine other event loops with Coro in
138 Coro::AnyEvent, and how to run it (and not to block). Seems to be
139 the most common source of confusion.
140 - try to catch people naively blocking in an event callback.
141 - work around the perl filehandle bug issue in conjunction with
142 older common::sense (as indirectly pointed out by ZSystem).
143 - clarify the "not from signal handlers" section.
144
1455.2 Sun Oct 4 14:54:24 CEST 2009
146 - Coro::Storable destroyed the prototypes of the functions it wrapped.
147 - export rouse_cb and rouse_wait by default now.
148 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
149 - new method $state->swap_sv.
150 - added section on "windows process emulation" to the manpage,
151 after a not-so-fruitful (nor-friendly) "discussion" with chip
152 salzenberg (discussion implies arguments, but his only arguments
153 were ad-hominems, one wonders why he started it in the first
154 place). I hope this explains it well enough for him to understand,
155 and maybe well enough for others to understand.
156 - use common::sense everywhere now.
157 - idle callbacks are no longer supported, use idle coros instead.
158 - print a thread listing when a deadlock is detected.
159
1605.17 Sat Aug 22 23:09:31 CEST 2009
161 - work around a bug in the perl debugger causing crashes
162 when running under the debugger by marking _pool_handler as nodebug.
163 - speed up Coro::async considerably.
164 - try some hacks to get netbsd to work "more often" - their broken
165 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
166 - convert Coro to AE 5.0 API.
167
1685.162 Tue Jul 28 04:04:03 CEST 2009
169 - perl 5.8.2 is now minimum requirement.
170 - skip t/19_handle.t on broken windows perls.
171
1725.161 Wed Jul 22 04:47:38 CEST 2009
173 - Coro::AnyEvent::poll could have a different prototype when EV was
174 used as backend (analyzed by Tatsuhiko Miyagawa).
175 - Coro::AnyEvent errornously initialised the event loop when loaded,
176 not on demand.
177 - try to workaround rare */t/01_unblock.t failures.
178
1795.16 Tue Jul 21 01:44:37 CEST 2009
180 - Coro::AnyEvent failed to hook into the event loop
181 when no threads had been readied between detecting
182 the event loop and actually running it.
183 - considerably speed up Coro::Select by taking avdantage
184 of AnyEvent > 4.8 and some other optimisations.
185 - implement paragraph readline mode in Coro::Handle
186 (based on patches by Zsbán Ambrus).
187 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
188 and analyzed by Yasuhiro MATSUMOTO).
189 - clarified libcoro license and copyright.
190 - someone stole my EXTRA_META!!!
191 - implement Coro::Select::patch_pp_sselect and it's brother,
192 for hardcode select overriding.
193
1945.151 Mon Jul 6 05:41:57 CEST 2009
195 - backport to windows process emulation code again (patch by
196 Yasuhiro MATSUMOTO).
197 - slightly update Coro::MakeMaker.
198
1995.15 Tue Jun 30 10:28:06 CEST 2009
200 - deprecate Coro::Socket, document how to get ipv6 support via
201 AnyEvent::Socket instead.
202 - implement signal->wait ($cb) interface, similar to semaphores.
203 - work around SvOK not supporting getmagic, so we have to getmagic
204 to test for undef :/ (reported by Matthias Waldorf).
205 - load Coro::AnyEvent in all modules using AnyEvent.
206 - work around perl corrupting our internal data structures,
207 reported by Tokuhiro Matsuno.
208 - enable per-coroutine real and cpu time gathering
209 (Coro::State::enable_times).
210
2115.14 Wed Jun 24 01:37:48 CEST 2009
212 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
213 as this seems to have been a difficult concept (poll, sleep, idle,
214 idle_upto).
215 - add Coro::AnyEvent::readable/writable functions.
216 - clarify Coro::EV/Event/AnyEvent manpages.
217 - free per-thread global scalars in the thread calling ->cancel, to
218 avoid crashes when $_, $@ etc., are magical but some of those
219 had already been freed.
220 - "unexperimentalise" the callback interface for Coro::Semaphore.
221 - speed up ready queue management/context switching by using a linked
222 list instead of an array (~5-10%).
223 - implement "watch" command in Coro::Debug shells.
224 - for fun, implement time-slicing as an example in the manpage.
225 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
226 Coro::Event, use the normal AnyEvent handling instead of dieing
227 (the same is true for Coro::Handle).
228 - properly document Coro::EV::timed_io_once.
229 - avoid unneccessary ->cancel calls in Coro::Handle.
230 - maybe make it work on mingw32 with win32 backend
231 (based on patch by Yasuhiro Matsumoto).
232
2335.132 Fri May 29 09:00:39 CEST 2009
234 - do not keep a reference to the argument itself in
235 Coro::Semaphore::guard, as it could change later.
236 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
237 should support a prepare callback.
238
2395.131 Mon Mar 16 23:20:37 CET 2009
240 - implement and document Coro->suspend, Coro->resume.
241 - fix Coro::Select implementation to not (often) close
242 the passed file descriptors (testcase provided by pippijn).
243
2445.13 Mon Dec 15 21:51:42 CET 2008
245 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
246 - don't set diehook to C<undef> but instead to NULL, to avoid
247 spurious warnings.
248 - fix a lot of bugs in Coro::SemaphoreSet.
249 - Coro::SemaphoreSet will less often create a semaphore needlessly.
250 - add Coro::SemaphoreSet::count and wait methods.
251 - take advantage of the new Guard module.
252 - deprecate Coro::guard.
253 - try to fix the dreaded 01_unblock tests once more. I hate it when
254 testsuites need more fixing than the code they are supposed to test.
255 - croak in more cases when a required callback isn't resolvable.
256 - fix some minor issues in Coro::State->call/eval.
257 - use current coroutine context instead of a temporary one
258 when temporarily switching to another coroutine.
259 - do not call C-level on_destroy handlers during global destruction,
260 to avoid needless segfaults.
261
2625.12 Sun Dec 7 13:30:38 CET 2008
263 - add default config for MirOS, which seems to be a bug-to-bug
264 compatible fork of openbsd ("world domination by releasing
265 openbsd cvs before the openbsd folks do it" or so :).
266 - free_padlist did destroy the names pad, not good, but didn't
267 seem to bother perl - this could fix issues such as eval ""
268 inside a function called from multiple coroutines.
269 - use a different method to detect destruction time.
270 - be more careful when freeing padlists just before global
271 destruction.
272 - fixed and expanded the call/cc example.
273 - renamed _terminate to _coro_run.
274 - new method Coro::Channel->shutdown.
275 - try pthreads on openbsd <4.4 (broken sigaltstack, will
276 pthreads fare better?).
277 - be less picky about destroying "a" running coroutine.
278
2795.11 Tue Nov 25 21:49:05 CET 2008
280 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
281 working code with failed assertions, introducing more bugs than
282 it fixes, requiring elaborate workarounds :(
283
2845.1 Mon Nov 24 08:54:59 CET 2008
285 - wrote a small introductory tutorial - Coro::Intro.
286 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
287 - Coro::Select did errornously dup the file descriptors
288 and didn't work with all AnyEvent backends.
289 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
290 LWP data transfers.
291 - disassociate c contexts from coro objects - this is agruably more
292 correct, but mostly allows sharing of cctxs between coro and state
293 objects, for added memory savings and speed increases.
294 - bumped $Coro::POOL_RSS up to 32kb by default.
295 - no longer set the optype to OP_CUSTOM, as B::* understandably
296 doesn't like this very much (and we *are* a type of entersub).
297 - implement state cloning, just to prove that call/cc can be done.
298 - automatically load Coro::AnyEvent in Coro::Handle.
299 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
300 do not have this method (and don't need it either).
301 - speed up generic anyevent methods in Coro::Handle by using rouse
302 callbacks.
303 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
304 others. It also makes the debugger happier, as you can trace
305 through the idle threads now.
306 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
307 - eg/event was pretty broken.
308 - better 5.8.6 compatibility.
309
3105.0 Thu Nov 20 10:35:05 CET 2008
311 - NEW ARCHITECTURE: use the latest 4.x version if you experience
312 stability issues.
313 - bump API version to 7 - all dependents must be recompiled.
314 - removed timed_* functions - they were not being used anyways
315 and should be replaced by a more generic mechanism -
316 and were annoying to support anyways :)
317 - removed SemaphoreSet's waiter method - use sem method instead.
318 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
319 - async_pool did free a scalar value twice
320 ("Attempt to unreference...").
321 - fix a longstanding bug where calling terminate on a coro that
322 was waiting for a semaphore that was just becoming available
323 would cause a deadlock (semaphore would get into a state where
324 it was available but waiters were still blocked).
325 - calling throw on a coroutine that is waiting for a semaphore will
326 no longer make it acquire the semaphore (and thus leak a count).
327 - perl's process emulation is now not even theoretically supported
328 anymore.
329 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
330 conversion of callback-style to blocking-style.
331 - new methods $coro->schedule_to and ->cede_to, to specifically
332 schedule or cede to a specific coroutine.
333 - new function Coro::Semaphore::wait.
334 - use named constants in Coro::Channel (Richard Hundt).
335 - directly patch the entersub opcode calling SLF functions (cede,
336 transfer and so on). this does speed up context switching, but
337 more importanly, it frees us from the hardcoded behaviour of
338 entersub, so we might actually be able to return something from
339 those functions and atcually create new ones.
340 - take advantage of __builtin_frame_address on gcc.
341 - expose THX in coroapi (not sure whether this was a wise decision,
342 as "threaded" perls are running at half speed anyways).
343 - implement execute_slf (schedule-like-function) interface that makes
344 it possible to implement schedule-like-functions in XS.
345 - use new SLF interface to massively speed up Coro::EV by roughly a
346 factor of two.
347 - used new SLF interface to massively speed up Coro::Semaphore by a
348 factor of three.
349 - used new SLF interface to speed up Coro::AIO by roughly a factor of
350 four and reduce its memory usage considerably.
351 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
352 for a nice speedup and vastly more correct behaviour. Also implement
353 a new method "sem" to get at the underlying semaphore object.
354 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
355 (in comparison) 20-40% speedup.
356 - used new SLF interface to reimplement Coro::Signal gaining
357 some unknown (because I was too lazy), but certain, speedup, and also
358 making signals reliable for the first time.
359 - used new SLF interface and other optimisations to speed up async_pool
360 by a factor of two. It also doesn't rely on perl's exception mechanism
361 to exit anymore. The overhead for terminating an async_pool, coro over
362 a normal async is now very small.
363 - sped up coroutine creation/destruction by 40%.
364 - forgot to include Coro/libcoro/README in the dist for all these years.
365 - work around a freebsd pthreads bug (manual testcancel is required as
366 pthread_cond_wait isn't a cancellation point on freebsd).
367 - use new rouse functions to speed up and simplify Coro::BDB.
368 - make "prefer perl native functions" work with threaded perls.
369 - condense Coro::Debug ps output, hint at v and w flags.
370 - (libcoro) lots of minor cleanups and portability improvements.
371
3724.914 Wed Nov 19 12:54:18 CET 2008
373 - fix a disastrous bug in the readline optimisation
374 introduced in 4.801.
375
3764.913 Sat Nov 15 07:58:28 CET 2008
377 - async_pool did free a scalar value twice
378 ("Attempt to unreference...").
379
3804.912 Thu Nov 13 18:31:23 CET 2008
381 - minor cleanups.
382 - use much larger stacks on linux and perl < 5.8.8.
383 - Coro::Debug::new_unix_server did not unlink the socket
384 when destroyed.
385
3864.911 Tue Nov 11 04:26:17 CET 2008
387 - "port" to threaded perls.
388
3894.91 Mon Nov 10 05:36:38 CET 2008
390 - the ->throw exception object no longer leaks.
391 - creating a new cctx leaked a scopestack entry (memleak).
392 - new coroutines didn't get created with a zero flags field
393 (unknown impact).
394 - calling ->throw on a not-yet-started coroutine should now work
395 instead of being ignored.
396 - ->throw is now supported on Coro::State objects.
397 - clean up cctx creation code a bit.
398 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
399
4004.9 Sat Nov 8 17:45:27 CET 2008
401 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
402 - (libcoro) no longer leak threads in the experimental pthread backend,
403 also speed it up considerably.
404 - (libcoro) do not rely on makecontext passing void *'s unscathed.
405 - fix compiletime dependencies on libcoro in the Makefile.
406 - cctx_count wasn't always updated properly.
407 - Coro::State::cctx_stacksize wasn't applied correctly.
408 - new function Coro::State::cctx_max_idle.
409 - the default max number of idle C contexts is now 4.
410 - (libcoro) try harder to get _setjmp/_longjmp.
411 - (libcoro) cleanup and extend the libcoro API to officially
412 allow the creation of empty source contexts.
413 - very experimental workaround for the totally broken netbsd platform.
414 - tried to hack around openbsd bugs.
415
4164.804 Wed Nov 5 16:36:00 CET 2008
417 - Coro::Debug::new_unix_server would not create a non-blocking listening
418 socket, sometimes causing freezes.
419 - (libcoro) fix misaligned stack points for setjmp and assembly
420 methods, which can cause crashes on x86/x86_64 with a sufficiently
421 aggressive compiler.
422 - new function: Coro::Debug::new_tcp_server.
423 - move ->throw into the Coro class because it only works on coro objects.
424
4254.803 Mon Nov 3 17:16:12 CET 2008
426 - (libcoro) use a global asm statement to become independent of gcc
427 otpimisations for CORO_ASM (thanks to pippijn for the idea).
428 - try to workaround yet another broken bsd, this time dragonfly.
429
4304.802 Thu Oct 30 10:56:12 CET 2008
431 - support -fno-omit-frame-pointer on x86 with the assembly method.
432 - tune 01_unblock.t tests a bit.
433
4344.801 Wed Oct 22 18:33:37 CEST 2008
435 - improve readline speed for very long "lines".
436 - backport to 5.8.8.
437
4384.8 Thu Oct 2 13:34:40 CEST 2008
439 - new function Coro::AIO::aio_wait.
440 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
441 - greatly speed up and reduce memory usage of Coro::AIO requests.
442 - implement some other µ-optimisations.
443
4444.749 Mon Sep 29 14:40:12 CEST 2008
445 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
446
4474.748 Sat Sep 27 14:03:19 CEST 2008
448 - implement, but do not document, PerlIO::cede(granularity).
449 - Coro::Storable forgot to wrap Storable::pstore.
450 - work around the multitude of leaks and memory corruption
451 bugs in PerlIO::via by using our own C-level perliol. As a side
452 effect, Coro::Storable is now much, much, much faster.
453
4544.747 Tue Sep 23 01:59:41 CEST 2008
9 - fix a per-cv memleak (one empty array was leaked per 455 - fix a per-cv memleak (one empty array was leaked per
10 code reference). 456 code reference).
11 - avoid a crash in coro->call|rss when the coroutine was already 457 - avoid a crash in coro->call|rss when the coroutine was already
12 destroyed (most noticably when using Coro::Debug::ps :) 458 destroyed (most noticably when using Coro::Debug::ps :)
13 - also protect *Storable::FILE. 459 - also protect *Storable::FILE.
460 - push up default storable granularity to 20ms.
14 461
154.746 Sun Sep 21 03:22:20 CEST 2008 4624.746 Sun Sep 21 03:22:20 CEST 2008
16 - be more insistent on locking Storable against reentrancy 463 - be more insistent on locking Storable against reentrancy
17 in Coro::Storable. 464 in Coro::Storable.
18 - move swap_def?v and throw to Coro::State, as documented. 465 - move swap_def?v and throw to Coro::State, as documented.

Diff Legend

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