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

Comparing Coro/Changes (file contents):
Revision 1.157 by root, Sun Nov 26 17:35:42 2006 UTC vs.
Revision 1.556 by root, Fri Apr 29 17:38:56 2011 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: check wether stack-sharing is still effective 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: more effective stack sharing by having one "default stack" for "simple" coroutines? 4TODO: channel->maxsize(newsize)?
5TODO: verify exception frame changes and force stack copies? 5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx
6TODO: cancel calling another coro that cancels ourselves => crash (CF_DELAYCANCEL?) -> canceltest
6 7
8 - implement Coro->cancel in XS for a 20% speed improvement, and to
9 be able to implement mutual cancellation.
10 - cancelling a coro while it itself is cancelling another coro is
11 now supported and working, instead of triggering an assertion.
12 - be a bit more crash-resistant when calling (buggy) on_destroy
13 callbacks (best effort).
14
155.372 Wed Feb 23 06:14:30 CET 2011
16 - apparently mingw doesn't provide a working gettimeofday, try to
17 work around that by relying on Time::HiRes (indirectly brought to
18 my attention by Max Maischein).
19 - fix some portability issues when Time::HiRes was used.
20
215.371 Mon Feb 21 14:36:08 CET 2011
22 - backport to windows process emulation code again.
23
245.37 Sat Feb 19 07:49:44 CET 2011
25 - add a big "Coro thread life cycle" section to "man Coro".
26 - try a tentative workaround against the breakage that 5.13 has
27 introduced without depreciation period. sigh.
28 - no longer use Time::HiRes if gettimeofday is available, which
29 saves quite a lot of memory.
30
315.36 Sun Feb 13 05:33:41 CET 2011
32 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
33 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
34 method call.
35 - undocument Coro::Timer::sleep and obsolete whole module.
36 - optimise Coro::Timer::timeout memory and cpu usage.
37 - slightly updated Coro::Intro for recent changes.
38 - do not initialise PL_dirty anymore.
39
405.25 Thu Nov 11 01:08:39 CET 2010
41 - try a different approach on netbsd - netbsd 5 finally has marginally
42 working pthreads, but still broken ucontext/sigaltstack.
43 - openbsd 4.8 finally got their act together, Coro works out of the box
44 with asm, setjmp and pthreads (no change, just informational).
45
465.24 Sat Oct 23 11:27:12 CEST 2010
47 - port to the EV 4.0 API.
48 - work around bugs in mingw32, making strawberry perl work
49 out of the box.
50 - correctly modify Coro::AIO function prototypes
51 so that they reflect the "no optional parameters" rule.
52 - "ported" libcoro to C++.
53
545.23 Mon May 17 18:50:42 CEST 2010
55 - be more resistant to ordering changes when initialising
56 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
57 Waldorf).
58 - document that perl 5.12 deliberately removed support for cloning.
59
605.22 Wed Apr 14 03:55:35 CEST 2010
61 - correctly return udnef on errors in Coro::Handle::read/write
62 (testcase by Marc Mims).
63 - convert Coro::Util into a "perl compatibility wrapper" - the functions
64 are less useful now, but are drop-in replacements for existing
65 functions, listing better alternatives in the documentation. This also
66 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
67 with Coro::Util::inet_aton.
68 - do not override $Coro::idle unconditionally in Coro.pm, as other
69 modules could have provided their own idle coro already
70 (for exmaple, Coro::AnyEvent).
71 - fix Coro::Util::gethost* functions.
72 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
73
745.21 Wed Dec 16 07:19:51 CET 2009
75 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
76 together.
77 - add some examples on how to combine other event loops with Coro in
78 Coro::AnyEvent, and how to run it (and not to block). Seems to be
79 the most common source of confusion.
80 - try to catch people naively blocking in an event callback.
81 - work around the perl filehandle bug issue in conjunction with
82 older common::sense (as indirectly pointed out by ZSystem).
83 - clarify the "not from signal handlers" section.
84
855.2 Sun Oct 4 14:54:24 CEST 2009
86 - Coro::Storable destroyed the prototypes of the functions it wrapped.
87 - export rouse_cb and rouse_wait by default now.
88 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
89 - new method $state->swap_sv.
90 - added section on "windows process emulation" to the manpage,
91 after a not-so-fruitful (nor-friendly) "discussion" with chip
92 salzenberg (discussion implies arguments, but his only arguments
93 were ad-hominems, one wonders why he started it in the first
94 place). I hope this explains it well enough for him to understand,
95 and maybe well enough for others to understand.
96 - use common::sense everywhere now.
97 - idle callbacks are no longer supported, use idle coros instead.
98 - print a thread listing when a deadlock is detected.
99
1005.17 Sat Aug 22 23:09:31 CEST 2009
101 - work around a bug in the perl debugger causing crashes
102 when running under the debugger by marking _pool_handler as nodebug.
103 - speed up Coro::async considerably.
104 - try some hacks to get netbsd to work "more often" - their broken
105 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
106 - convert Coro to AE 5.0 API.
107
1085.162 Tue Jul 28 04:04:03 CEST 2009
109 - perl 5.8.2 is now minimum requirement.
110 - skip t/19_handle.t on broken windows perls.
111
1125.161 Wed Jul 22 04:47:38 CEST 2009
113 - Coro::AnyEvent::poll could have a different prototype when EV was
114 used as backend (analyzed by Tatsuhiko Miyagawa).
115 - Coro::AnyEvent errornously initialised the event loop when loaded,
116 not on demand.
117 - try to workaround rare */t/01_unblock.t failures.
118
1195.16 Tue Jul 21 01:44:37 CEST 2009
120 - Coro::AnyEvent failed to hook into the event loop
121 when no threads had been readied between detecting
122 the event loop and actually running it.
123 - considerably speed up Coro::Select by taking avdantage
124 of AnyEvent > 4.8 and some other optimisations.
125 - implement paragraph readline mode in Coro::Handle
126 (based on patches by Zsbán Ambrus).
127 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
128 and analyzed by Yasuhiro MATSUMOTO).
129 - clarified libcoro license and copyright.
130 - someone stole my EXTRA_META!!!
131 - implement Coro::Select::patch_pp_sselect and it's brother,
132 for hardcode select overriding.
133
1345.151 Mon Jul 6 05:41:57 CEST 2009
135 - backport to windows process emulation code again (patch by
136 Yasuhiro MATSUMOTO).
137 - slightly update Coro::MakeMaker.
138
1395.15 Tue Jun 30 10:28:06 CEST 2009
140 - deprecate Coro::Socket, document how to get ipv6 support via
141 AnyEvent::Socket instead.
142 - implement signal->wait ($cb) interface, similar to semaphores.
143 - work around SvOK not supporting getmagic, so we have to getmagic
144 to test for undef :/ (reported by Matthias Waldorf).
145 - load Coro::AnyEvent in all modules using AnyEvent.
146 - work around perl corrupting our internal data structures,
147 reported by Tokuhiro Matsuno.
148 - enable per-coroutine real and cpu time gathering
149 (Coro::State::enable_times).
150
1515.14 Wed Jun 24 01:37:48 CEST 2009
152 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
153 as this seems to have been a difficult concept (poll, sleep, idle,
154 idle_upto).
155 - add Coro::AnyEvent::readable/writable functions.
156 - clarify Coro::EV/Event/AnyEvent manpages.
157 - free per-thread global scalars in the thread calling ->cancel, to
158 avoid crashes when $_, $@ etc., are magical but some of those
159 had already been freed.
160 - "unexperimentalise" the callback interface for Coro::Semaphore.
161 - speed up ready queue management/context switching by using a linked
162 list instead of an array (~5-10%).
163 - implement "watch" command in Coro::Debug shells.
164 - for fun, implement time-slicing as an example in the manpage.
165 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
166 Coro::Event, use the normal AnyEvent handling instead of dieing
167 (the same is true for Coro::Handle).
168 - properly document Coro::EV::timed_io_once.
169 - avoid unneccessary ->cancel calls in Coro::Handle.
170 - maybe make it work on mingw32 with win32 backend
171 (based on patch by Yasuhiro Matsumoto).
172
1735.132 Fri May 29 09:00:39 CEST 2009
174 - do not keep a reference to the argument itself in
175 Coro::Semaphore::guard, as it could change later.
176 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
177 should support a prepare callback.
178
1795.131 Mon Mar 16 23:20:37 CET 2009
180 - implement and document Coro->suspend, Coro->resume.
181 - fix Coro::Select implementation to not (often) close
182 the passed file descriptors (testcase provided by pippijn).
183
1845.13 Mon Dec 15 21:51:42 CET 2008
185 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
186 - don't set diehook to C<undef> but instead to NULL, to avoid
187 spurious warnings.
188 - fix a lot of bugs in Coro::SemaphoreSet.
189 - Coro::SemaphoreSet will less often create a semaphore needlessly.
190 - add Coro::SemaphoreSet::count and wait methods.
191 - take advantage of the new Guard module.
192 - deprecate Coro::guard.
193 - try to fix the dreaded 01_unblock tests once more. I hate it when
194 testsuites need more fixing than the code they are supposed to test.
195 - croak in more cases when a required callback isn't resolvable.
196 - fix some minor issues in Coro::State->call/eval.
197 - use current coroutine context instead of a temporary one
198 when temporarily switching to another coroutine.
199 - do not call C-level on_destroy handlers during global destruction,
200 to avoid needless segfaults.
201
2025.12 Sun Dec 7 13:30:38 CET 2008
203 - add default config for MirOS, which seems to be a bug-to-bug
204 compatible fork of openbsd ("world domination by releasing
205 openbsd cvs before the openbsd folks do it" or so :).
206 - free_padlist did destroy the names pad, not good, but didn't
207 seem to bother perl - this could fix issues such as eval ""
208 inside a function called from multiple coroutines.
209 - use a different method to detect destruction time.
210 - be more careful when freeing padlists just before global
211 destruction.
212 - fixed and expanded the call/cc example.
213 - renamed _terminate to _coro_run.
214 - new method Coro::Channel->shutdown.
215 - try pthreads on openbsd <4.4 (broken sigaltstack, will
216 pthreads fare better?).
217 - be less picky about destroying "a" running coroutine.
218
2195.11 Tue Nov 25 21:49:05 CET 2008
220 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
221 working code with failed assertions, introducing more bugs than
222 it fixes, requiring elaborate workarounds :(
223
2245.1 Mon Nov 24 08:54:59 CET 2008
225 - wrote a small introductory tutorial - Coro::Intro.
226 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
227 - Coro::Select did errornously dup the file descriptors
228 and didn't work with all AnyEvent backends.
229 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
230 LWP data transfers.
231 - disassociate c contexts from coro objects - this is agruably more
232 correct, but mostly allows sharing of cctxs between coro and state
233 objects, for added memory savings and speed increases.
234 - bumped $Coro::POOL_RSS up to 32kb by default.
235 - no longer set the optype to OP_CUSTOM, as B::* understandably
236 doesn't like this very much (and we *are* a type of entersub).
237 - implement state cloning, just to prove that call/cc can be done.
238 - automatically load Coro::AnyEvent in Coro::Handle.
239 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
240 do not have this method (and don't need it either).
241 - speed up generic anyevent methods in Coro::Handle by using rouse
242 callbacks.
243 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
244 others. It also makes the debugger happier, as you can trace
245 through the idle threads now.
246 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
247 - eg/event was pretty broken.
248 - better 5.8.6 compatibility.
249
2505.0 Thu Nov 20 10:35:05 CET 2008
251 - NEW ARCHITECTURE: use the latest 4.x version if you experience
252 stability issues.
253 - bump API version to 7 - all dependents must be recompiled.
254 - removed timed_* functions - they were not being used anyways
255 and should be replaced by a more generic mechanism -
256 and were annoying to support anyways :)
257 - removed SemaphoreSet's waiter method - use sem method instead.
258 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
259 - async_pool did free a scalar value twice
260 ("Attempt to unreference...").
261 - fix a longstanding bug where calling terminate on a coro that
262 was waiting for a semaphore that was just becoming available
263 would cause a deadlock (semaphore would get into a state where
264 it was available but waiters were still blocked).
265 - calling throw on a coroutine that is waiting for a semaphore will
266 no longer make it acquire the semaphore (and thus leak a count).
267 - perl's process emulation is now not even theoretically supported
268 anymore.
269 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
270 conversion of callback-style to blocking-style.
271 - new methods $coro->schedule_to and ->cede_to, to specifically
272 schedule or cede to a specific coroutine.
273 - new function Coro::Semaphore::wait.
274 - use named constants in Coro::Channel (Richard Hundt).
275 - directly patch the entersub opcode calling SLF functions (cede,
276 transfer and so on). this does speed up context switching, but
277 more importanly, it frees us from the hardcoded behaviour of
278 entersub, so we might actually be able to return something from
279 those functions and atcually create new ones.
280 - take advantage of __builtin_frame_address on gcc.
281 - expose THX in coroapi (not sure whether this was a wise decision,
282 as "threaded" perls are running at half speed anyways).
283 - implement execute_slf (schedule-like-function) interface that makes
284 it possible to implement schedule-like-functions in XS.
285 - use new SLF interface to massively speed up Coro::EV by roughly a
286 factor of two.
287 - used new SLF interface to massively speed up Coro::Semaphore by a
288 factor of three.
289 - used new SLF interface to speed up Coro::AIO by roughly a factor of
290 four and reduce its memory usage considerably.
291 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
292 for a nice speedup and vastly more correct behaviour. Also implement
293 a new method "sem" to get at the underlying semaphore object.
294 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
295 (in comparison) 20-40% speedup.
296 - used new SLF interface to reimplement Coro::Signal gaining
297 some unknown (because I was too lazy), but certain, speedup, and also
298 making signals reliable for the first time.
299 - used new SLF interface and other optimisations to speed up async_pool
300 by a factor of two. It also doesn't rely on perl's exception mechanism
301 to exit anymore. The overhead for terminating an async_pool, coro over
302 a normal async is now very small.
303 - sped up coroutine creation/destruction by 40%.
304 - forgot to include Coro/libcoro/README in the dist for all these years.
305 - work around a freebsd pthreads bug (manual testcancel is required as
306 pthread_cond_wait isn't a cancellation point on freebsd).
307 - use new rouse functions to speed up and simplify Coro::BDB.
308 - make "prefer perl native functions" work with threaded perls.
309 - condense Coro::Debug ps output, hint at v and w flags.
310 - (libcoro) lots of minor cleanups and portability improvements.
311
3124.914 Wed Nov 19 12:54:18 CET 2008
313 - fix a disastrous bug in the readline optimisation
314 introduced in 4.801.
315
3164.913 Sat Nov 15 07:58:28 CET 2008
317 - async_pool did free a scalar value twice
318 ("Attempt to unreference...").
319
3204.912 Thu Nov 13 18:31:23 CET 2008
321 - minor cleanups.
322 - use much larger stacks on linux and perl < 5.8.8.
323 - Coro::Debug::new_unix_server did not unlink the socket
324 when destroyed.
325
3264.911 Tue Nov 11 04:26:17 CET 2008
327 - "port" to threaded perls.
328
3294.91 Mon Nov 10 05:36:38 CET 2008
330 - the ->throw exception object no longer leaks.
331 - creating a new cctx leaked a scopestack entry (memleak).
332 - new coroutines didn't get created with a zero flags field
333 (unknown impact).
334 - calling ->throw on a not-yet-started coroutine should now work
335 instead of being ignored.
336 - ->throw is now supported on Coro::State objects.
337 - clean up cctx creation code a bit.
338 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
339
3404.9 Sat Nov 8 17:45:27 CET 2008
341 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
342 - (libcoro) no longer leak threads in the experimental pthread backend,
343 also speed it up considerably.
344 - (libcoro) do not rely on makecontext passing void *'s unscathed.
345 - fix compiletime dependencies on libcoro in the Makefile.
346 - cctx_count wasn't always updated properly.
347 - Coro::State::cctx_stacksize wasn't applied correctly.
348 - new function Coro::State::cctx_max_idle.
349 - the default max number of idle C contexts is now 4.
350 - (libcoro) try harder to get _setjmp/_longjmp.
351 - (libcoro) cleanup and extend the libcoro API to officially
352 allow the creation of empty source contexts.
353 - very experimental workaround for the totally broken netbsd platform.
354 - tried to hack around openbsd bugs.
355
3564.804 Wed Nov 5 16:36:00 CET 2008
357 - Coro::Debug::new_unix_server would not create a non-blocking listening
358 socket, sometimes causing freezes.
359 - (libcoro) fix misaligned stack points for setjmp and assembly
360 methods, which can cause crashes on x86/x86_64 with a sufficiently
361 aggressive compiler.
362 - new function: Coro::Debug::new_tcp_server.
363 - move ->throw into the Coro class because it only works on coro objects.
364
3654.803 Mon Nov 3 17:16:12 CET 2008
366 - (libcoro) use a global asm statement to become independent of gcc
367 otpimisations for CORO_ASM (thanks to pippijn for the idea).
368 - try to workaround yet another broken bsd, this time dragonfly.
369
3704.802 Thu Oct 30 10:56:12 CET 2008
371 - support -fno-omit-frame-pointer on x86 with the assembly method.
372 - tune 01_unblock.t tests a bit.
373
3744.801 Wed Oct 22 18:33:37 CEST 2008
375 - improve readline speed for very long "lines".
376 - backport to 5.8.8.
377
3784.8 Thu Oct 2 13:34:40 CEST 2008
379 - new function Coro::AIO::aio_wait.
380 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
381 - greatly speed up and reduce memory usage of Coro::AIO requests.
382 - implement some other µ-optimisations.
383
3844.749 Mon Sep 29 14:40:12 CEST 2008
385 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
386
3874.748 Sat Sep 27 14:03:19 CEST 2008
388 - implement, but do not document, PerlIO::cede(granularity).
389 - Coro::Storable forgot to wrap Storable::pstore.
390 - work around the multitude of leaks and memory corruption
391 bugs in PerlIO::via by using our own C-level perliol. As a side
392 effect, Coro::Storable is now much, much, much faster.
393
3944.747 Tue Sep 23 01:59:41 CEST 2008
395 - fix a per-cv memleak (one empty array was leaked per
396 code reference).
397 - avoid a crash in coro->call|rss when the coroutine was already
398 destroyed (most noticably when using Coro::Debug::ps :)
399 - also protect *Storable::FILE.
400 - push up default storable granularity to 20ms.
401
4024.746 Sun Sep 21 03:22:20 CEST 2008
403 - be more insistent on locking Storable against reentrancy
404 in Coro::Storable.
405 - move swap_def?v and throw to Coro::State, as documented.
406
4074.745 Thu Jul 24 00:14:38 CEST 2008
408 - remove debugging code related to MgPV_nolen_const, also try to
409 make it compile with perl 5.8.6 (yes, apple apparently loves
410 outdated software). Reported by John S.
411
4124.744 Tue Jul 8 22:06:35 CEST 2008
413 - correctly provide default DIE/WARN handlers as documented.
414 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
415 implementations inside perl work fine for us.
416
4174.743 Mon Jun 16 00:21:57 CEST 2008
418 - when using Coro::EV without running EV::loop it could
419 result in busy-waiting for events, this has been fixed.
420 - reduce codesize and improve performance by using EV_DEFAULT_UC.
421
4224.742 Sat May 31 14:10:21 CEST 2008
423 - implement a workaround for (some) perl <5.8.8 versions.
424 - require EV 3.3+.
425
4264.741 Fri May 30 23:33:09 CEST 2008
427 - tell netbsd how utterly broken their imitation of an OS is
428 and refuse to build by default if pthreads are in use.
429 - switch to "s" method on all bsd's by default, as their ucontext
430 stuff seems just too broken.
431 - fix a bug in Coro::Select.
432
4334.74 Thu May 29 20:05:31 CEST 2008
434 - do not test Coro::LWP for lack of dependencies.
435 - do not test Coro::Select for lack of working perls.
436
4374.73 Thu May 29 2008
438 - fix a bug in Coro::EV which would cause it to block despite
439 there being runnable coroutines.
440 - sprinkle "no warnings" freely over everything, also suppress
441 warnings for some other modules.
442 - fix typo in WSAEWOULDBLOCK.
443
4444.72 Sun May 25 05:14:36 CEST 2008
445 - tweak META.yaml a bit, unfortunately, there is no documented way
446 to have optional dependencies with CPAN. doh :(
447 - avoid running some tests on windows because they would fail due to
448 perl bug (broken fork, broken pipes...).
449 - work around perl on windows bugs where perl returns undocumented
450 error codes for sysread, syswrite etc. by taking advantage
451 of AnyEvent's workaround for that problem.
452 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
453 a common perl implementation bug on windows.
454 - use unix domain sockets in testsuite to work around
455 common perl implementation bugs on widows (they are emulated by
456 tcp sockets on windows. ugh.)
457
4584.71 Sat May 24 20:01:27 CEST 2008
459 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
460 - take advantage of async name resolution of AnyEvent::Util.
461 - work around brutal inet_aton override in Coro::LWP.
462 - take advantage of the readyhook in Coro::EV, for smoother
463 scheduling.
464
4654.7 Sun May 11 00:32:19 CEST 2008
466 - completely reworked the Coro manpage.
467 - added Coro::AnyEvent, generic event loop integration.
468 - implement cancel, ready and kill commands in Coro::Debug.
469 - document find_coro in Coro::Debug.
470 - incompatible change: rename has_stack to has_cctx.
471 - Coro::AIO and Coro::BDB no longer force event model detection,
472 use AnyEvent::AIO and AnyEvent::BDB.
473
4744.6 Sat Apr 26 10:05:14 CEST 2008
475 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
476 by the Coro module. It was a mistake to have it in the first place.
477 - (experimental) support for activestate perl 5.10 (method "w").
478 - (experimental) support for strawberry perl (method "a").
479 - coro_sigelem_set did not return a value although it had to,
480 actual impact unknown.
481
4824.51 Mon Apr 14 13:28:27 CEST 2008
483 - make it compile again on 5.8.
484
4854.50 Thu Apr 10 09:43:17 CEST 2008
486 - I did it twice! (see 4.49).
487
4884.49 Sun Apr 6 21:23:31 CEST 2008
489 - grr, instead of compiling the recent changes on 5.10 only they
490 were compiled on 5.8 only.
491
4924.48 Sun Apr 6 20:36:46 CEST 2008
493 - allow coroutine switches during eval's under 5.10.x, as it seems
494 the parser is a per-interpreter option now, so this is safe
495 (this might fix the odd crashing bug).
496 - drop support for 5.9.x versions: they are dead, jim.
497
4984.47 Sun Apr 6 00:37:52 CEST 2008
499 - force cctx allocation on API calls: we know we need to force one
500 and gcc actually manages to confuse our heuristic nowadays,
501 leading to crashes and worse.
502 - document force_cctx.
503
5044.46 Fri Apr 4 22:05:43 CEST 2008
505 - upgrade libcoro, resulting in pthread-backend (which was only created
506 to fulfill the rules of the programming languages shootout).
507
5084.45 Thu Mar 13 11:55:36 CET 2008
509 - fix a file leaking bug in eg/mhyttpd that would allow
510 downloading of any file (reported by oesi).
511 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
512 (also add testcase).
513 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
514 apparently having loads of fun with that).
515 - implement $state->swap_defsv, ->swap_defav and document ->throw.
516
5174.4 Wed Feb 13 16:44:29 CET 2008
518 - only assume .cfi pseudo isns exist on GNU/Linux.
519 - add get_timed method to Coro::Channels.
520 - fixed Coro::Debug manpage.
521 - perl 5.11 compatbility improvement based on patch by
522 Andreas König.
523
5244.37 Sun Jan 20 11:25:23 CET 2008
525 - soften the check for an active parser for perl 5.10 (avoiding
526 "Coro::State::transfer called while parsing" in many cases).
527
5284.36 Sun Jan 13 10:53:56 CET 2008
529 - reset diehook when terminating from an async_pool as to not
530 trigger any __DIE__ handlers.
531
5324.35 Sun Jan 13 04:14:13 CET 2008
533 - "bt" debug command now displays any exceptions
534 from longmess and also skips the innermost
535 stackframes, giving more useufl output.
536 - allow backtraces in coroutines blocked in Coro::EV,
537 at a <1% speed hit.
538 - handle localising of $SIG{__WARN__} and __DIE__
539 properly (with a proper amount of dirty hacking).
540
5414.34 Sat Dec 22 17:49:53 CET 2007
542 - upgrade to EV version 2.0 API.
543
5444.33 Mon Dec 17 08:36:12 CET 2007
545 - make Coro::AIO etc. loadable in the absence of EV.
546
5474.32 Mon Dec 17 07:46:02 CET 2007
548 - majorly improved Coro::Handle's performance with Coro::EV.
549 - implemented the until now mythical Coro::BDB module.
550 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
551
5524.31 Wed Dec 5 12:32:39 CET 2007
553 - remove warn statement form Coro::Util.
554
5554.3 Tue Dec 4 20:33:14 CET 2007
556 - calls to the idle function could cause stack corruption
557 when the stack changed.
558 - do no longer rely on the presence of EV::DNS if EV is used
559 (because it is gone), but instead take avdantage of EV::ADNS
560 if available.
561 - add ($) prototypes to all functions in Coro::Storable.
562 - use a conventional (safer) idle callback in Coro::EV.
563 - do accept quickly in Coro::Debug to avoid endless loops.
564
5654.22 Fri Nov 30 16:04:04 CET 2007
566 - really use optimised versions for Event and EV in Coro::Util
567 and Coro::Handle.
568
5694.21 Sun Nov 25 10:48:59 CET 2007
570 - fix a spurious memory read.
571 - Coro::EV no longer keeps the eventloop "alive".
572
5734.2 Fri Nov 9 20:47:05 CET 2007
574 - enable/disable tracing from a new coroutine, not a pooled one.
575 - fix a memleak in Coro::Event.
576 - removed killall call from fork_eval.
577 - made sure store_fd is already loaded so that fork_eval
578 does not have to parse autoload in each subprocess.
579 - only use assembly method if -O switch is in $Config{optimize}.
580 - add (optional) Coro::EV module, so far the best event loop module
581 directly supported by Coro.
582 - if the event model is EV, use EV::DNS to resolve
583 stuff in Coro::Util.
584 - don't get confused by multiple event notifications in Coro::Handle.
585 - initial support for EV (libevent interface).
586 - require Event and EV using configure_requires, to force their existance.
587
5884.13 Wed Oct 24 07:26:45 CEST 2007
589 - add Coro::Storable::blocking_thaw.
590 - use a vastly more complicated technique to localise
591 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
592 - use a coroutine for the idle callback Coro::Event,
593 instead of running Event in the current coroutine context.
594 This also catches recursive invocations.
595 - actually report fork errors in gethostbyname and inet_aton.
596
5974.11 Thu Oct 11 02:40:24 CEST 2007
598 - port to threaded perls.
599
6004.1 Thu Oct 11 02:38:16 CEST 2007
601 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
602 be local to each coro (see Coro::State).
603 - incompatible change: for very deep reasons, cede and cede_notself
604 cannot return anything, so nothing will be returned.
605 - possibly bring back 5.10 compatibility (untested).
606 - work around stupid (and wrong) warning on 5.10 :(.
607 - overlay the saved state over the context stack. This saves
608 a few hundred bytes per coroutine on average and also
609 speeds up context switching a bit.
610 - further tune default stack sizes.
611 - (more) correctly calculate stack usage in coro_rss.
612 - Coro::Storable::blocking_* did not properly lock
613 resulting in races between coroutines.
614 - added Coro::Storable::guard.
615 - stopping to trace a coroutine could destroy the cctx of
616 an unrelated coroutine.
617 - explain the relationship between Perl and C coroutines in
618 more detail in Coro::State.
619 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
620 causing a fork per resolve. This also affected Coro::Socket.
621 - switch to a separate stack in $coro->call/eval to avoid
622 invalidating pointers.
623
6244.03 Sat Oct 6 21:24:00 CEST 2007
625 - added Coro::throw method.
626 - minor code cleanups.
627
6284.02 Sat Oct 6 02:36:47 CEST 2007
629 - fix a very minor per-coroutine memleak (a single codereference).
630 - fixed a bug where the currently in-use c context would be freed
631 prematurely (can happen only when programs change the stacksize
632 or use tracing).
633 - tracing can no longer keep a coro alive after it terminated.
634 - do static branch prediction in the common path for gcc. gives
635 about 2-5% speed improvement here.
636
6374.01 Fri Oct 5 22:10:49 CEST 2007
638 - instead of recreating *a* standard output handle we simply
639 use STDOUT, which is faster and hopefully more robust.
640
6414.0 Fri Oct 5 12:56:00 CEST 2007
642 - incompatibly changed Coro::Storable::freeze.
643 - major new feature: added Coro::Debug, for interactive coroutine
644 debugging, tracing and much more.
645 - major bug fix: unbelievable, but true: $_, $/ and many other
646 "saved" variables actually weren't being saved. This has been fixed,
647 of course, while increasing performance while losing all the save
648 flags.
649 - save flags are gone, and all the api functions dealing with them.
650 - added Coro::Semaphore::adjust.
651 - added Coro::Util::fork_eval.
652 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
653 - added Coro::killall.
654 - reduce initial stack sizes to allow for "micro-coroutines".
655 - better async_pool resource management, moved parts of async_pool
656 handling to XS (major speed improvement).
657 - actually croak before modifying important data structures.
658 - refuse to transfer while compiling.
659 - possibly support eval EXPR better now.
660 - enable assembly per default on linux+bsd x86+amd64.
661 - all internal members were renamed _something for easier subclassing.
662 - many minor tweaks.
663
6643.63 Wed May 16 14:10:06 CEST 2007
665 - implement handcoded assembly for x86/amd64 SVR ABI.
666
6673.62 Fri Apr 27 21:36:06 CEST 2007
668 - upgrade libcoro (which might set unwind info correctly).
669 - change default on linux to setjmp/longjmp.
670
6713.61 Thu Apr 19 12:36:18 CEST 2007
672 - Coro::Storable caused an endless loop when thawing invalid
673 pst's sometimes.
674 - use a Semaphore in Coro::Storable, as Storable doesn't
675 seem to be reentrant (although it is documented to
676 be threadsafe...).
677 - fix Coro::Signal to bring back the original unreliable
678 but stateful semantics.
679 - fixed a lot of typos in Coro.pm (patch submitted by David
680 Steinbrunner, which applied flawlessly).
681
6823.6 Sat Apr 14 17:13:31 CEST 2007
683 - added some bugfixes to get eg/myhttpd working again.
684 - added Coro::Storable for often-cede'ing freeze/thaw.
685 - try to do a clean exit when a coroutine calls exit
686 (EXPERIMENTAL).
687 - got rid of indirect call through _coro_init.
688 - updated the partly antique examples in eg/ to
689 work again and be a bit less magic, too.
690 - fixed Coro::Signal semantics to work as documented again.
691
6923.55 Sun Mar 25 01:20:47 CET 2007
693 - add SAVE_DEFFH to save the default filehandle and enable
694 it by default.
695 - finally move socket-operations from Coro::Socket to Coro::Handle
696 to be able to unblock foreign sockets.
697 - add Coro::State::save_also and guarded_save.
698 - add count accessor to Coro::Semaphore.
699 - add Coro::State::cctx_stacksize.
700 - just for the fun of it, do not rely on implicit context,
701 which can dramatically improve performance, but people
702 using windows-process-emulation perls do not care much about
703 performance.
704
7053.51 Sun Mar 4 14:18:04 CET 2007
706 - fixed a problem when you weakened references to Coro::State's
707 (patch by Yuval Kogman).
708
7093.501 Wed Feb 28 12:44:07 CET 2007
710 - rename some global symbols as macosx from hell redefines
711 them without asking.
712
7133.5 Tue Feb 13 20:22:53 CET 2007
714 - do AnyEvent model detection earlier, avoiding problems
715 caused by first using AnyEvent and later Coro::Event.
716 - implement and document Coro::Event event objects.
717 - fix a potential problem in Coro::Event causing crashes.
718 - initialise PL_comppad when creating a new coroutine,
719 avoids crashes on early coro destruction.
720
7213.41 Mon Jan 22 19:19:49 CET 2007
722 - readline on Coro::Handle did not support undefined $/,
723 nor did it deliver partial lines on EOF or error.
724 - implement malloc fallback for stack allocation because
725 stupid broken idiotic OSX has a stupid broken
726 idiotic fits-the-whole-os mmap "implementation" and
727 my dick feels longer if Coro is portable even to
728 obsolete platforms.
729
7303.4 Fri Jan 19 21:52:54 CET 2007
731 - remove t/09_timer.t, as it isn't really testing much
732 but was rather flaky in practise.
733 - async_pool coro would keep arguments and callback alive until
734 it was reused.
735 - cancellation of a coroutine could cause spurious idle calls
736 in cede_notself.
737
7383.3 Sat Jan 6 03:45:00 CET 2007
739 - implement $coro->on_destroy.
740 - Coro::Event blocking semantics have been changed,
741 documented and - hopefully - improved.
742 - fix nice adding, not subtracting, from priority.
743 - fix ->prio and api_is_ready (patch by Mark Hinds).
744 - fixed an assert ("... == prev__cctx->idle_te")
745 that could errronously trigger.
746 - fix various large and small memleaks.
747 - use a (hopefully) more stable cancel implementation
748 that immediately frees the coroutine data.
749 - cede/cede_notself return a status now.
750 - added Coro::guard function.
751 - added a global coroutine pool for jobs (on my machine,
752 I can create and execute 48k simple coros/s with async,
753 and 128k coros with async_pool).
754 - Coro::AIO now uses the coroutine priority as io priority.
755
7563.2 Fri Dec 22 05:07:09 CET 2006
757 - improve portability to slightly older perls.
758 - use cleaner coroutine destruction.
759 - simplify configuration for users.
760 - optionally (unrecommended) prefer perl functions over
761 their coro replacements.
762
7633.11 Tue Dec 5 13:11:24 CET 2006
764 - fixed some bogus assert's, but as perl.h disables assert even
765 without NDEBUG (thank you very much), not too many people should
766 notice (that did include myself). Andreas König noticed, though :)
767 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
768 it. Also noticed by Andreas König :)
769 - save/restore tainted status.
770 - verified to pass the testsuite on my 5.9.5.
771
7723.1 Mon Dec 4 23:03:40 CET 2006
773 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
774 - incompatible change: transfer flags are now per-state.
775 - give a better error message on deadlock.
776 - document Coro::nready.
777 - enhanced testsuite.
778
7793.01 Sun Dec 3 23:47:42 CET 2006
780 - forgot to include Coro::Timer.
781
7823.0 Sun Dec 3 22:57:25 CET 2006
783 - the "FINALLY COMPLETELY STABLE" release (coming soon:
784 the "FAMOUS LAST WORDS" release).
7 - reimplement a new stack sharing algorithm. 785 - implement a new stack sharing algorithm, which uses a stack
786 pool (size currently hardcoded).
8 - make stack sharing mandatory (it no longer uses a heuristic). 787 - make stack sharing mandatory (it no longer uses a heuristic).
9 - eval/die no longer cause weird problems under heavy use. 788 - eval/die no longer cause weird problems under heavy use.
10 - Coro::Event could cause livelocks if it was used but 789 - Coro::Event could cause livelocks if it was used but
11 Event watchers were used directly. 790 no Coro::Event watchers were used.
791 - Coro::Event now uses asynccheck as prepare does not
792 check for changed watchers.
793 - Coro::Event allows multiple waiting coros and will wake up one
794 per event.
12 - Coro::Event should be cleaner and more efficient now. 795 - Coro::Event should be cleaner and more efficient now.
796 - new utility function Coro::unblock_sub.
797 - document the sad fact that Event is no longer reentrant.
798 - putting a coroutine into the ready queue twice could under
799 some circumstances lead to stack corruption.
13 - minor incompatibility: subclassing is supported directly now 800 - minor incompatibility: subclassing Coro::State is supported
14 without going through a _coro_state member. 801 directly now without going through a _coro_state member.
15 - state/coro switching is much faster now. 802 - state/coro switching is much faster now.
16 - very minor optimisations and code/documentation cleanup. 803 - very minor optimisations and code/documentation cleanup.
17 - avoid problems due to compiler inlining. 804 - avoid problems due to compiler inlining.
18 - removed timers from Coro::Timer -> use AnyEvent instead. 805 - removed timers from Coro::Timer -> use AnyEvent instead.
19 - replaced Coro::idle coroutine by (cleaner) idle handler. 806 - replaced Coro::idle coroutine by (cleaner) idle handler.
20 - updated to newest libcoro. 807 - updated to newest libcoro.
808 - implement enhanced support for valgrind.
809 - implement is_ready and return value for ready.
810 - removed Coro::Cont, it was a misnomer (it's generators),
811 and it was somewhat annoying to get it right. Will come back
812 if somebody asks for it :)
813 - many ->wait methods and Coro::Event could return
814 spuriously without the event having happened.
21 815
222.5 Tue Nov 7 12:22:33 CET 2006 8162.5 Tue Nov 7 12:22:33 CET 2006
23 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket 817 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
24 use AnyEvent, moved them to Coro/. 818 use AnyEvent, moved them to Coro/.
25 - added Coro::LWP which contains all the uglyness required to 819 - added Coro::LWP which contains all the uglyness required to

Diff Legend

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