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

Comparing Coro/Changes (file contents):
Revision 1.374 by root, Wed Sep 24 21:20:05 2008 UTC vs.
Revision 1.587 by root, Tue Aug 30 07:48:53 2011 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::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 7 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
8 - update ecb.h.
9
106.06 Mon Aug 8 23:59:48 CEST 2011
11 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
12 backend, which is an order of magnitude slower. unfortunately, cygwins
13 pthread implementation isn't very complete either, so allocate the stack
14 twice just to be sure.
15 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
16 in its ithreaded perl - best recompile cygwin and use the 'w'indows
17 backend for much better performance. also disable ithreads for
18 even better performance...).
19
206.05 Thu Aug 4 21:36:36 CEST 2011
21 - blush, condvar values would not be propagated from send to recv anymore
22 (reported by Chip Salzenberg).
23 - use exponential increase for the readline buffer length in
24 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
25
266.04 Wed Aug 3 17:15:45 CEST 2011
27 - use even more efficient and more compatible condvars for
28 compatibility to AnyEvent 6.x :)
29 - more inconsequential ecb.h updates.
30
316.03 Wed Aug 3 11:41:30 CEST 2011
32 - change how Coro patches AnyEvent condvars for compatibility to
33 AnyEvent 6.x.
34 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
35 hasn't been included.
36
376.02 Wed Jul 13 04:35:19 CEST 2011
38 - "improve portability to Gentoo" - gentoo manages to put perl variables
39 in memory areas that are farther than 2gb apart, which the jit couldn't
40 handle and barfed. now it's just a bit slower on gentoo and similar
41 systems.
42
436.01 Sun Jul 3 12:31:14 CEST 2011
44 - workarounds are good, but the test for whether pthreads are used
45 was not good. thisone should be better.
46 - check differently whether gcc generates cfi instructions itself.
47
486.0 Wed Jun 29 19:43:35 CEST 2011
49 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
50 destroyed and cleaned up automatically (e.g. async { schedule }).
51 - implement a JIT compiler for part of the thread switch code,
52 which gives a 50% speed improvement on threaded perls, and
53 about 4% on non-thraeded perls (so threaded perls now finally
54 reach about half the speed of non-threaded perls).
55 - slightly modernise Coro::Intro, add section about rouse functions.
56 - avoid DEFSV and ERRSV, giving another 10% improvement
57 in thread switching.
58 - Coro::State->is_destroyed is now called is_zombie.
59 - implement a Coro->safe_cancel method that might fail, but
60 cancels in a "safer" way if it succeeds.
61 - add preliminary support for DEBUGGING perls.
62 - get rid of two hash-accesses when initialising a new Coro - this
63 speeds up coro creation by almost a factor of two.
64 - croak when a coro that is being cancelled tries to block
65 (e.g. while executing a guard block), instead of crashing or
66 deadlocking.
67 - use a more robust and also faster method to identify Coro::State
68 objects - speeds up everything a bit.
69 - implement Coro->cancel in XS for a 20% speed improvement, and to
70 be able to implement mutual cancellation.
71 - speed up context switches by a percent or two by more efficiently
72 allocating context stack entries.
73 - implement Coro->join and Coro->on_destroy in XS for a speedup and
74 a reduction in memory use.
75 - cancelling a coro while it itself is cancelling another coro is
76 now supported and working, instead of triggering an assertion.
77 - be a bit more crash-resistant when calling (buggy) on_destroy
78 callbacks (best effort).
79 - move on_destroy into the slf_frame, to allow extension slf
80 functions to have destructors.
81 - get rid if coro refcounting - simply crash in other interpreter
82 threads by nulling the pointers on clone.
83 - simplify warn/die hook handling when loading Coro - the convoluted
84 logic seems to be no longer neccessary.
85 - use libecb instead of our own home-grown gcc hacks.
86 - document alternatives to Coro::LWP. Please use them :)
87 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
88 sigaltstack. Really. wine suffers from it, erlang suffers from it,
89 and it's known since at least 2006.
90
915.372 Wed Feb 23 06:14:30 CET 2011
92 - apparently mingw doesn't provide a working gettimeofday, try to
93 work around that by relying on Time::HiRes (indirectly brought to
94 my attention by Max Maischein).
95 - fix some portability issues when Time::HiRes was used.
96
975.371 Mon Feb 21 14:36:08 CET 2011
98 - backport to windows process emulation code again.
99
1005.37 Sat Feb 19 07:49:44 CET 2011
101 - add a big "Coro thread life cycle" section to "man Coro".
102 - try a tentative workaround against the breakage that 5.13 has
103 introduced without depreciation period. sigh.
104 - no longer use Time::HiRes if gettimeofday is available, which
105 saves quite a lot of memory.
106
1075.36 Sun Feb 13 05:33:41 CET 2011
108 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
109 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
110 method call.
111 - undocument Coro::Timer::sleep and obsolete whole module.
112 - optimise Coro::Timer::timeout memory and cpu usage.
113 - slightly updated Coro::Intro for recent changes.
114 - do not initialise PL_dirty anymore.
115
1165.25 Thu Nov 11 01:08:39 CET 2010
117 - try a different approach on netbsd - netbsd 5 finally has marginally
118 working pthreads, but still broken ucontext/sigaltstack.
119 - openbsd 4.8 finally got their act together, Coro works out of the box
120 with asm, setjmp and pthreads (no change, just informational).
121
1225.24 Sat Oct 23 11:27:12 CEST 2010
123 - port to the EV 4.0 API.
124 - work around bugs in mingw32, making strawberry perl work
125 out of the box.
126 - correctly modify Coro::AIO function prototypes
127 so that they reflect the "no optional parameters" rule.
128 - "ported" libcoro to C++.
129
1305.23 Mon May 17 18:50:42 CEST 2010
131 - be more resistant to ordering changes when initialising
132 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
133 Waldorf).
134 - document that perl 5.12 deliberately removed support for cloning.
135
1365.22 Wed Apr 14 03:55:35 CEST 2010
137 - correctly return udnef on errors in Coro::Handle::read/write
138 (testcase by Marc Mims).
139 - convert Coro::Util into a "perl compatibility wrapper" - the functions
140 are less useful now, but are drop-in replacements for existing
141 functions, listing better alternatives in the documentation. This also
142 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
143 with Coro::Util::inet_aton.
144 - do not override $Coro::idle unconditionally in Coro.pm, as other
145 modules could have provided their own idle coro already
146 (for exmaple, Coro::AnyEvent).
147 - fix Coro::Util::gethost* functions.
148 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
149
1505.21 Wed Dec 16 07:19:51 CET 2009
151 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
152 together.
153 - add some examples on how to combine other event loops with Coro in
154 Coro::AnyEvent, and how to run it (and not to block). Seems to be
155 the most common source of confusion.
156 - try to catch people naively blocking in an event callback.
157 - work around the perl filehandle bug issue in conjunction with
158 older common::sense (as indirectly pointed out by ZSystem).
159 - clarify the "not from signal handlers" section.
160
1615.2 Sun Oct 4 14:54:24 CEST 2009
162 - Coro::Storable destroyed the prototypes of the functions it wrapped.
163 - export rouse_cb and rouse_wait by default now.
164 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
165 - new method $state->swap_sv.
166 - added section on "windows process emulation" to the manpage,
167 after a not-so-fruitful (nor-friendly) "discussion" with chip
168 salzenberg (discussion implies arguments, but his only arguments
169 were ad-hominems, one wonders why he started it in the first
170 place). I hope this explains it well enough for him to understand,
171 and maybe well enough for others to understand.
172 - use common::sense everywhere now.
173 - idle callbacks are no longer supported, use idle coros instead.
174 - print a thread listing when a deadlock is detected.
175
1765.17 Sat Aug 22 23:09:31 CEST 2009
177 - work around a bug in the perl debugger causing crashes
178 when running under the debugger by marking _pool_handler as nodebug.
179 - speed up Coro::async considerably.
180 - try some hacks to get netbsd to work "more often" - their broken
181 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
182 - convert Coro to AE 5.0 API.
183
1845.162 Tue Jul 28 04:04:03 CEST 2009
185 - perl 5.8.2 is now minimum requirement.
186 - skip t/19_handle.t on broken windows perls.
187
1885.161 Wed Jul 22 04:47:38 CEST 2009
189 - Coro::AnyEvent::poll could have a different prototype when EV was
190 used as backend (analyzed by Tatsuhiko Miyagawa).
191 - Coro::AnyEvent errornously initialised the event loop when loaded,
192 not on demand.
193 - try to workaround rare */t/01_unblock.t failures.
194
1955.16 Tue Jul 21 01:44:37 CEST 2009
196 - Coro::AnyEvent failed to hook into the event loop
197 when no threads had been readied between detecting
198 the event loop and actually running it.
199 - considerably speed up Coro::Select by taking avdantage
200 of AnyEvent > 4.8 and some other optimisations.
201 - implement paragraph readline mode in Coro::Handle
202 (based on patches by Zsbán Ambrus).
203 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
204 and analyzed by Yasuhiro MATSUMOTO).
205 - clarified libcoro license and copyright.
206 - someone stole my EXTRA_META!!!
207 - implement Coro::Select::patch_pp_sselect and it's brother,
208 for hardcode select overriding.
209
2105.151 Mon Jul 6 05:41:57 CEST 2009
211 - backport to windows process emulation code again (patch by
212 Yasuhiro MATSUMOTO).
213 - slightly update Coro::MakeMaker.
214
2155.15 Tue Jun 30 10:28:06 CEST 2009
216 - deprecate Coro::Socket, document how to get ipv6 support via
217 AnyEvent::Socket instead.
218 - implement signal->wait ($cb) interface, similar to semaphores.
219 - work around SvOK not supporting getmagic, so we have to getmagic
220 to test for undef :/ (reported by Matthias Waldorf).
221 - load Coro::AnyEvent in all modules using AnyEvent.
222 - work around perl corrupting our internal data structures,
223 reported by Tokuhiro Matsuno.
224 - enable per-coroutine real and cpu time gathering
225 (Coro::State::enable_times).
226
2275.14 Wed Jun 24 01:37:48 CEST 2009
228 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
229 as this seems to have been a difficult concept (poll, sleep, idle,
230 idle_upto).
231 - add Coro::AnyEvent::readable/writable functions.
232 - clarify Coro::EV/Event/AnyEvent manpages.
233 - free per-thread global scalars in the thread calling ->cancel, to
234 avoid crashes when $_, $@ etc., are magical but some of those
235 had already been freed.
236 - "unexperimentalise" the callback interface for Coro::Semaphore.
237 - speed up ready queue management/context switching by using a linked
238 list instead of an array (~5-10%).
239 - implement "watch" command in Coro::Debug shells.
240 - for fun, implement time-slicing as an example in the manpage.
241 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
242 Coro::Event, use the normal AnyEvent handling instead of dieing
243 (the same is true for Coro::Handle).
244 - properly document Coro::EV::timed_io_once.
245 - avoid unneccessary ->cancel calls in Coro::Handle.
246 - maybe make it work on mingw32 with win32 backend
247 (based on patch by Yasuhiro Matsumoto).
248
2495.132 Fri May 29 09:00:39 CEST 2009
250 - do not keep a reference to the argument itself in
251 Coro::Semaphore::guard, as it could change later.
252 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
253 should support a prepare callback.
254
2555.131 Mon Mar 16 23:20:37 CET 2009
256 - implement and document Coro->suspend, Coro->resume.
257 - fix Coro::Select implementation to not (often) close
258 the passed file descriptors (testcase provided by pippijn).
259
2605.13 Mon Dec 15 21:51:42 CET 2008
261 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
262 - don't set diehook to C<undef> but instead to NULL, to avoid
263 spurious warnings.
264 - fix a lot of bugs in Coro::SemaphoreSet.
265 - Coro::SemaphoreSet will less often create a semaphore needlessly.
266 - add Coro::SemaphoreSet::count and wait methods.
267 - take advantage of the new Guard module.
268 - deprecate Coro::guard.
269 - try to fix the dreaded 01_unblock tests once more. I hate it when
270 testsuites need more fixing than the code they are supposed to test.
271 - croak in more cases when a required callback isn't resolvable.
272 - fix some minor issues in Coro::State->call/eval.
273 - use current coroutine context instead of a temporary one
274 when temporarily switching to another coroutine.
275 - do not call C-level on_destroy handlers during global destruction,
276 to avoid needless segfaults.
277
2785.12 Sun Dec 7 13:30:38 CET 2008
279 - add default config for MirOS, which seems to be a bug-to-bug
280 compatible fork of openbsd ("world domination by releasing
281 openbsd cvs before the openbsd folks do it" or so :).
282 - free_padlist did destroy the names pad, not good, but didn't
283 seem to bother perl - this could fix issues such as eval ""
284 inside a function called from multiple coroutines.
285 - use a different method to detect destruction time.
286 - be more careful when freeing padlists just before global
287 destruction.
288 - fixed and expanded the call/cc example.
289 - renamed _terminate to _coro_run.
290 - new method Coro::Channel->shutdown.
291 - try pthreads on openbsd <4.4 (broken sigaltstack, will
292 pthreads fare better?).
293 - be less picky about destroying "a" running coroutine.
294
2955.11 Tue Nov 25 21:49:05 CET 2008
296 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
297 working code with failed assertions, introducing more bugs than
298 it fixes, requiring elaborate workarounds :(
299
3005.1 Mon Nov 24 08:54:59 CET 2008
301 - wrote a small introductory tutorial - Coro::Intro.
302 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
303 - Coro::Select did errornously dup the file descriptors
304 and didn't work with all AnyEvent backends.
305 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
306 LWP data transfers.
307 - disassociate c contexts from coro objects - this is agruably more
308 correct, but mostly allows sharing of cctxs between coro and state
309 objects, for added memory savings and speed increases.
310 - bumped $Coro::POOL_RSS up to 32kb by default.
311 - no longer set the optype to OP_CUSTOM, as B::* understandably
312 doesn't like this very much (and we *are* a type of entersub).
313 - implement state cloning, just to prove that call/cc can be done.
314 - automatically load Coro::AnyEvent in Coro::Handle.
315 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
316 do not have this method (and don't need it either).
317 - speed up generic anyevent methods in Coro::Handle by using rouse
318 callbacks.
319 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
320 others. It also makes the debugger happier, as you can trace
321 through the idle threads now.
322 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
323 - eg/event was pretty broken.
324 - better 5.8.6 compatibility.
325
3265.0 Thu Nov 20 10:35:05 CET 2008
327 - NEW ARCHITECTURE: use the latest 4.x version if you experience
328 stability issues.
329 - bump API version to 7 - all dependents must be recompiled.
330 - removed timed_* functions - they were not being used anyways
331 and should be replaced by a more generic mechanism -
332 and were annoying to support anyways :)
333 - removed SemaphoreSet's waiter method - use sem method instead.
334 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
335 - async_pool did free a scalar value twice
336 ("Attempt to unreference...").
337 - fix a longstanding bug where calling terminate on a coro that
338 was waiting for a semaphore that was just becoming available
339 would cause a deadlock (semaphore would get into a state where
340 it was available but waiters were still blocked).
341 - calling throw on a coroutine that is waiting for a semaphore will
342 no longer make it acquire the semaphore (and thus leak a count).
343 - perl's process emulation is now not even theoretically supported
344 anymore.
345 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
346 conversion of callback-style to blocking-style.
347 - new methods $coro->schedule_to and ->cede_to, to specifically
348 schedule or cede to a specific coroutine.
349 - new function Coro::Semaphore::wait.
350 - use named constants in Coro::Channel (Richard Hundt).
351 - directly patch the entersub opcode calling SLF functions (cede,
352 transfer and so on). this does speed up context switching, but
353 more importanly, it frees us from the hardcoded behaviour of
354 entersub, so we might actually be able to return something from
355 those functions and atcually create new ones.
356 - take advantage of __builtin_frame_address on gcc.
357 - expose THX in coroapi (not sure whether this was a wise decision,
358 as "threaded" perls are running at half speed anyways).
359 - implement execute_slf (schedule-like-function) interface that makes
360 it possible to implement schedule-like-functions in XS.
361 - use new SLF interface to massively speed up Coro::EV by roughly a
362 factor of two.
363 - used new SLF interface to massively speed up Coro::Semaphore by a
364 factor of three.
365 - used new SLF interface to speed up Coro::AIO by roughly a factor of
366 four and reduce its memory usage considerably.
367 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
368 for a nice speedup and vastly more correct behaviour. Also implement
369 a new method "sem" to get at the underlying semaphore object.
370 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
371 (in comparison) 20-40% speedup.
372 - used new SLF interface to reimplement Coro::Signal gaining
373 some unknown (because I was too lazy), but certain, speedup, and also
374 making signals reliable for the first time.
375 - used new SLF interface and other optimisations to speed up async_pool
376 by a factor of two. It also doesn't rely on perl's exception mechanism
377 to exit anymore. The overhead for terminating an async_pool, coro over
378 a normal async is now very small.
379 - sped up coroutine creation/destruction by 40%.
380 - forgot to include Coro/libcoro/README in the dist for all these years.
381 - work around a freebsd pthreads bug (manual testcancel is required as
382 pthread_cond_wait isn't a cancellation point on freebsd).
383 - use new rouse functions to speed up and simplify Coro::BDB.
384 - make "prefer perl native functions" work with threaded perls.
385 - condense Coro::Debug ps output, hint at v and w flags.
386 - (libcoro) lots of minor cleanups and portability improvements.
387
3884.914 Wed Nov 19 12:54:18 CET 2008
389 - fix a disastrous bug in the readline optimisation
390 introduced in 4.801.
391
3924.913 Sat Nov 15 07:58:28 CET 2008
393 - async_pool did free a scalar value twice
394 ("Attempt to unreference...").
395
3964.912 Thu Nov 13 18:31:23 CET 2008
397 - minor cleanups.
398 - use much larger stacks on linux and perl < 5.8.8.
399 - Coro::Debug::new_unix_server did not unlink the socket
400 when destroyed.
401
4024.911 Tue Nov 11 04:26:17 CET 2008
403 - "port" to threaded perls.
404
4054.91 Mon Nov 10 05:36:38 CET 2008
406 - the ->throw exception object no longer leaks.
407 - creating a new cctx leaked a scopestack entry (memleak).
408 - new coroutines didn't get created with a zero flags field
409 (unknown impact).
410 - calling ->throw on a not-yet-started coroutine should now work
411 instead of being ignored.
412 - ->throw is now supported on Coro::State objects.
413 - clean up cctx creation code a bit.
414 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
415
4164.9 Sat Nov 8 17:45:27 CET 2008
417 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
418 - (libcoro) no longer leak threads in the experimental pthread backend,
419 also speed it up considerably.
420 - (libcoro) do not rely on makecontext passing void *'s unscathed.
421 - fix compiletime dependencies on libcoro in the Makefile.
422 - cctx_count wasn't always updated properly.
423 - Coro::State::cctx_stacksize wasn't applied correctly.
424 - new function Coro::State::cctx_max_idle.
425 - the default max number of idle C contexts is now 4.
426 - (libcoro) try harder to get _setjmp/_longjmp.
427 - (libcoro) cleanup and extend the libcoro API to officially
428 allow the creation of empty source contexts.
429 - very experimental workaround for the totally broken netbsd platform.
430 - tried to hack around openbsd bugs.
431
4324.804 Wed Nov 5 16:36:00 CET 2008
433 - Coro::Debug::new_unix_server would not create a non-blocking listening
434 socket, sometimes causing freezes.
435 - (libcoro) fix misaligned stack points for setjmp and assembly
436 methods, which can cause crashes on x86/x86_64 with a sufficiently
437 aggressive compiler.
438 - new function: Coro::Debug::new_tcp_server.
439 - move ->throw into the Coro class because it only works on coro objects.
440
4414.803 Mon Nov 3 17:16:12 CET 2008
442 - (libcoro) use a global asm statement to become independent of gcc
443 otpimisations for CORO_ASM (thanks to pippijn for the idea).
444 - try to workaround yet another broken bsd, this time dragonfly.
445
4464.802 Thu Oct 30 10:56:12 CET 2008
447 - support -fno-omit-frame-pointer on x86 with the assembly method.
448 - tune 01_unblock.t tests a bit.
449
4504.801 Wed Oct 22 18:33:37 CEST 2008
451 - improve readline speed for very long "lines".
452 - backport to 5.8.8.
453
4544.8 Thu Oct 2 13:34:40 CEST 2008
455 - new function Coro::AIO::aio_wait.
456 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
457 - greatly speed up and reduce memory usage of Coro::AIO requests.
458 - implement some other µ-optimisations.
459
4604.749 Mon Sep 29 14:40:12 CEST 2008
461 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
462
4634.748 Sat Sep 27 14:03:19 CEST 2008
9 - implement, but do not document, PerlIO::cede(granularity). 464 - implement, but do not document, PerlIO::cede(granularity).
10 - Coro::Storable forgot to wrap Storable::pstore. 465 - Coro::Storable forgot to wrap Storable::pstore.
11 - work around the multitude of leaks and memory corruption 466 - work around the multitude of leaks and memory corruption
12 bugs in PerlIO::via by using our own C-level perliol. 467 bugs in PerlIO::via by using our own C-level perliol. As a side
468 effect, Coro::Storable is now much, much, much faster.
13 469
144.747 Tue Sep 23 01:59:41 CEST 2008 4704.747 Tue Sep 23 01:59:41 CEST 2008
15 - fix a per-cv memleak (one empty array was leaked per 471 - fix a per-cv memleak (one empty array was leaked per
16 code reference). 472 code reference).
17 - avoid a crash in coro->call|rss when the coroutine was already 473 - avoid a crash in coro->call|rss when the coroutine was already

Diff Legend

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