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

Comparing Coro/Changes (file contents):
Revision 1.254 by root, Fri Oct 5 10:57:40 2007 UTC vs.
Revision 1.619 by root, Tue Nov 5 14:38:46 2013 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 4TODO: channel->maxsize(newsize)?
5 (which is cool in a lot of important aspects (failures!), 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)). 6TODO: swap_sv, maybe add scope_swap_sv, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
8
96.32 Tue Nov 5 15:35:35 CET 2013
10 - use a new algorithm to derive padlists for perl 5.18. The old
11 one could lead to 0-pointer accesses inside perl (reported
12 by Darin McBride).
13
146.31 Thu May 9 07:39:48 CEST 2013
15 - Coro::AIO requests would crash if the thread was ready'd
16 while the request was ongoing.
17
186.29 Wed May 8 02:55:18 CEST 2013
19 - when an on_destroy handler destructs the coro currently being
20 destructed a perl scalar could be accessed after being freed,
21 likely causing a crash.
22
236.28 Wed Mar 6 06:58:02 CET 2013
24 - clean remnants of existing __DIE__ and __WARN__ handlers so
25 they lose their magic and will not cause segfaults later
26 (testcase by Andrey Sagulin).
27 - improved Coro::State documentation a bit.
28 - Coro::Debug::command now flushes the output.
29 - add hack detection code for x32 abi, because the braindead slugs
30 who designed that made it look exactly like x86_64 without
31 providing proper compile time symbols to test for it. as a result,
32 this detection cannot work reliably.
33 - valgrind stack registering was broken.
34 - do not rely on Time::HiRes anymore in Coro::Debug.
35
366.23 Fri Dec 7 23:36:37 CET 2012
37 - use experimental fiber implementation on native windows
38 perls.
39 - use sizeof (void *) as multiplication factor for stack sizes,
40 to accomodate the totally braindamaged microsoft 64 bit "os".
41 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
42 activeperl 5.16 crashes when PerlIO_define_layer is called due
43 to some bug in the perl dll, strawberry perl at least passes
44 the testsuite.
45 - implement Coro::Handle->peeraddr/host/port, for slightly
46 improved compatibility with LWP.
47 - implement 5.17 compatibility by almost blindly applying a
48 good-looking patch by Father Chrysostomos.
49 - move stack management functions into libcoro 3.
50 - libcoro version 3 "released".
51 - support magic values as timed_io_once args.
52 - recommend AnyEvent 7+ or EV 4+, also require EV
53 version 4 or newer for Coro::EV.
54
556.10 Tue Oct 9 01:14:27 CEST 2012
56 - updated ecb.h, it had a typo that caused it to not compile on many
57 big endian systems (reported by many people).
58 - disable memory fences in ecb.h to improve portability.
59
606.09 Sat Oct 6 23:25:02 CEST 2012
61 - Coro::EV I/O watchers were not interruptible by exceptions
62 (Coro::State::throw) (testcase by sten).
63 - ->throw now puts threads into the ready queue, as this seems to
64 be expected by existing code, and code that doesn't cope with spurious
65 wakeups needs fixing anyway.
66 - use fd -1 in mmap.
67 - cast I32 to int in error message printf.
68 - warn about broken so-called "hardened" kernels.
69
706.08 Fri Apr 13 12:05:47 CEST 2012
71 - be more aggressive about exiting like perl does - formerly,
72 exiting from the non-main thread would not execute END blocks.
73
746.07 Fri Nov 11 21:21:48 CET 2011
75 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
76 - update ecb.h.
77
786.06 Mon Aug 8 23:59:48 CEST 2011
79 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
80 backend, which is an order of magnitude slower. unfortunately, cygwins
81 pthread implementation isn't very complete either, so allocate the stack
82 twice just to be sure.
83 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
84 in its ithreaded perl - best recompile cygwin and use the 'w'indows
85 backend for much better performance. also disable ithreads for
86 even better performance...).
87
886.05 Thu Aug 4 21:36:36 CEST 2011
89 - blush, condvar values would not be propagated from send to recv anymore
90 (reported by Chip Salzenberg).
91 - use exponential increase for the readline buffer length in
92 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
93
946.04 Wed Aug 3 17:15:45 CEST 2011
95 - use even more efficient and more compatible condvars for
96 compatibility to AnyEvent 6.x :)
97 - more inconsequential ecb.h updates.
98
996.03 Wed Aug 3 11:41:30 CEST 2011
100 - change how Coro patches AnyEvent condvars for compatibility to
101 AnyEvent 6.x.
102 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
103 hasn't been included.
104
1056.02 Wed Jul 13 04:35:19 CEST 2011
106 - "improve portability to Gentoo" - gentoo manages to put perl variables
107 in memory areas that are farther than 2gb apart, which the jit couldn't
108 handle and barfed. now it's just a bit slower on gentoo and similar
109 systems.
110
1116.01 Sun Jul 3 12:31:14 CEST 2011
112 - workarounds are good, but the test for whether pthreads are used
113 was not good. thisone should be better.
114 - check differently whether gcc generates cfi instructions itself.
115
1166.0 Wed Jun 29 19:43:35 CEST 2011
117 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
118 destroyed and cleaned up automatically (e.g. async { schedule }).
119 - implement a JIT compiler for part of the thread switch code,
120 which gives a 50% speed improvement on threaded perls, and
121 about 4% on non-thraeded perls (so threaded perls now finally
122 reach about half the speed of non-threaded perls).
123 - slightly modernise Coro::Intro, add section about rouse functions.
124 - avoid DEFSV and ERRSV, giving another 10% improvement
125 in thread switching.
126 - Coro::State->is_destroyed is now called is_zombie.
127 - implement a Coro->safe_cancel method that might fail, but
128 cancels in a "safer" way if it succeeds.
129 - add preliminary support for DEBUGGING perls.
130 - get rid of two hash-accesses when initialising a new Coro - this
131 speeds up coro creation by almost a factor of two.
132 - croak when a coro that is being cancelled tries to block
133 (e.g. while executing a guard block), instead of crashing or
134 deadlocking.
135 - use a more robust and also faster method to identify Coro::State
136 objects - speeds up everything a bit.
137 - implement Coro->cancel in XS for a 20% speed improvement, and to
138 be able to implement mutual cancellation.
139 - speed up context switches by a percent or two by more efficiently
140 allocating context stack entries.
141 - implement Coro->join and Coro->on_destroy in XS for a speedup and
142 a reduction in memory use.
143 - cancelling a coro while it itself is cancelling another coro is
144 now supported and working, instead of triggering an assertion.
145 - be a bit more crash-resistant when calling (buggy) on_destroy
146 callbacks (best effort).
147 - move on_destroy into the slf_frame, to allow extension slf
148 functions to have destructors.
149 - get rid if coro refcounting - simply crash in other interpreter
150 threads by nulling the pointers on clone.
151 - simplify warn/die hook handling when loading Coro - the convoluted
152 logic seems to be no longer neccessary.
153 - use libecb instead of our own home-grown gcc hacks.
154 - document alternatives to Coro::LWP. Please use them :)
155 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
156 sigaltstack. Really. wine suffers from it, erlang suffers from it,
157 and it's known since at least 2006.
158
1595.372 Wed Feb 23 06:14:30 CET 2011
160 - apparently mingw doesn't provide a working gettimeofday, try to
161 work around that by relying on Time::HiRes (indirectly brought to
162 my attention by Max Maischein).
163 - fix some portability issues when Time::HiRes was used.
164
1655.371 Mon Feb 21 14:36:08 CET 2011
166 - backport to windows process emulation code again.
167
1685.37 Sat Feb 19 07:49:44 CET 2011
169 - add a big "Coro thread life cycle" section to "man Coro".
170 - try a tentative workaround against the breakage that 5.13 has
171 introduced without depreciation period. sigh.
172 - no longer use Time::HiRes if gettimeofday is available, which
173 saves quite a lot of memory.
174
1755.36 Sun Feb 13 05:33:41 CET 2011
176 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
177 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
178 method call.
179 - undocument Coro::Timer::sleep and obsolete whole module.
180 - optimise Coro::Timer::timeout memory and cpu usage.
181 - slightly updated Coro::Intro for recent changes.
182 - do not initialise PL_dirty anymore.
183
1845.25 Thu Nov 11 01:08:39 CET 2010
185 - try a different approach on netbsd - netbsd 5 finally has marginally
186 working pthreads, but still broken ucontext/sigaltstack.
187 - openbsd 4.8 finally got their act together, Coro works out of the box
188 with asm, setjmp and pthreads (no change, just informational).
189
1905.24 Sat Oct 23 11:27:12 CEST 2010
191 - port to the EV 4.0 API.
192 - work around bugs in mingw32, making strawberry perl work
193 out of the box.
194 - correctly modify Coro::AIO function prototypes
195 so that they reflect the "no optional parameters" rule.
196 - "ported" libcoro to C++.
197
1985.23 Mon May 17 18:50:42 CEST 2010
199 - be more resistant to ordering changes when initialising
200 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
201 Waldorf).
202 - document that perl 5.12 deliberately removed support for cloning.
203
2045.22 Wed Apr 14 03:55:35 CEST 2010
205 - correctly return udnef on errors in Coro::Handle::read/write
206 (testcase by Marc Mims).
207 - convert Coro::Util into a "perl compatibility wrapper" - the functions
208 are less useful now, but are drop-in replacements for existing
209 functions, listing better alternatives in the documentation. This also
210 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
211 with Coro::Util::inet_aton.
212 - do not override $Coro::idle unconditionally in Coro.pm, as other
213 modules could have provided their own idle coro already
214 (for exmaple, Coro::AnyEvent).
215 - fix Coro::Util::gethost* functions.
216 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
217
2185.21 Wed Dec 16 07:19:51 CET 2009
219 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
220 together.
221 - add some examples on how to combine other event loops with Coro in
222 Coro::AnyEvent, and how to run it (and not to block). Seems to be
223 the most common source of confusion.
224 - try to catch people naively blocking in an event callback.
225 - work around the perl filehandle bug issue in conjunction with
226 older common::sense (as indirectly pointed out by ZSystem).
227 - clarify the "not from signal handlers" section.
228
2295.2 Sun Oct 4 14:54:24 CEST 2009
230 - Coro::Storable destroyed the prototypes of the functions it wrapped.
231 - export rouse_cb and rouse_wait by default now.
232 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
233 - new method $state->swap_sv.
234 - added section on "windows process emulation" to the manpage,
235 after a not-so-fruitful (nor-friendly) "discussion" with chip
236 salzenberg (discussion implies arguments, but his only arguments
237 were ad-hominems, one wonders why he started it in the first
238 place). I hope this explains it well enough for him to understand,
239 and maybe well enough for others to understand.
240 - use common::sense everywhere now.
241 - idle callbacks are no longer supported, use idle coros instead.
242 - print a thread listing when a deadlock is detected.
243
2445.17 Sat Aug 22 23:09:31 CEST 2009
245 - work around a bug in the perl debugger causing crashes
246 when running under the debugger by marking _pool_handler as nodebug.
247 - speed up Coro::async considerably.
248 - try some hacks to get netbsd to work "more often" - their broken
249 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
250 - convert Coro to AE 5.0 API.
251
2525.162 Tue Jul 28 04:04:03 CEST 2009
253 - perl 5.8.2 is now minimum requirement.
254 - skip t/19_handle.t on broken windows perls.
255
2565.161 Wed Jul 22 04:47:38 CEST 2009
257 - Coro::AnyEvent::poll could have a different prototype when EV was
258 used as backend (analyzed by Tatsuhiko Miyagawa).
259 - Coro::AnyEvent errornously initialised the event loop when loaded,
260 not on demand.
261 - try to workaround rare */t/01_unblock.t failures.
262
2635.16 Tue Jul 21 01:44:37 CEST 2009
264 - Coro::AnyEvent failed to hook into the event loop
265 when no threads had been readied between detecting
266 the event loop and actually running it.
267 - considerably speed up Coro::Select by taking avdantage
268 of AnyEvent > 4.8 and some other optimisations.
269 - implement paragraph readline mode in Coro::Handle
270 (based on patches by Zsbán Ambrus).
271 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
272 and analyzed by Yasuhiro MATSUMOTO).
273 - clarified libcoro license and copyright.
274 - someone stole my EXTRA_META!!!
275 - implement Coro::Select::patch_pp_sselect and it's brother,
276 for hardcode select overriding.
277
2785.151 Mon Jul 6 05:41:57 CEST 2009
279 - backport to windows process emulation code again (patch by
280 Yasuhiro MATSUMOTO).
281 - slightly update Coro::MakeMaker.
282
2835.15 Tue Jun 30 10:28:06 CEST 2009
284 - deprecate Coro::Socket, document how to get ipv6 support via
285 AnyEvent::Socket instead.
286 - implement signal->wait ($cb) interface, similar to semaphores.
287 - work around SvOK not supporting getmagic, so we have to getmagic
288 to test for undef :/ (reported by Matthias Waldorf).
289 - load Coro::AnyEvent in all modules using AnyEvent.
290 - work around perl corrupting our internal data structures,
291 reported by Tokuhiro Matsuno.
292 - enable per-coroutine real and cpu time gathering
293 (Coro::State::enable_times).
294
2955.14 Wed Jun 24 01:37:48 CEST 2009
296 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
297 as this seems to have been a difficult concept (poll, sleep, idle,
298 idle_upto).
299 - add Coro::AnyEvent::readable/writable functions.
300 - clarify Coro::EV/Event/AnyEvent manpages.
301 - free per-thread global scalars in the thread calling ->cancel, to
302 avoid crashes when $_, $@ etc., are magical but some of those
303 had already been freed.
304 - "unexperimentalise" the callback interface for Coro::Semaphore.
305 - speed up ready queue management/context switching by using a linked
306 list instead of an array (~5-10%).
307 - implement "watch" command in Coro::Debug shells.
308 - for fun, implement time-slicing as an example in the manpage.
309 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
310 Coro::Event, use the normal AnyEvent handling instead of dieing
311 (the same is true for Coro::Handle).
312 - properly document Coro::EV::timed_io_once.
313 - avoid unneccessary ->cancel calls in Coro::Handle.
314 - maybe make it work on mingw32 with win32 backend
315 (based on patch by Yasuhiro Matsumoto).
316
3175.132 Fri May 29 09:00:39 CEST 2009
318 - do not keep a reference to the argument itself in
319 Coro::Semaphore::guard, as it could change later.
320 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
321 should support a prepare callback.
322
3235.131 Mon Mar 16 23:20:37 CET 2009
324 - implement and document Coro->suspend, Coro->resume.
325 - fix Coro::Select implementation to not (often) close
326 the passed file descriptors (testcase provided by pippijn).
327
3285.13 Mon Dec 15 21:51:42 CET 2008
329 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
330 - don't set diehook to C<undef> but instead to NULL, to avoid
331 spurious warnings.
332 - fix a lot of bugs in Coro::SemaphoreSet.
333 - Coro::SemaphoreSet will less often create a semaphore needlessly.
334 - add Coro::SemaphoreSet::count and wait methods.
335 - take advantage of the new Guard module.
336 - deprecate Coro::guard.
337 - try to fix the dreaded 01_unblock tests once more. I hate it when
338 testsuites need more fixing than the code they are supposed to test.
339 - croak in more cases when a required callback isn't resolvable.
340 - fix some minor issues in Coro::State->call/eval.
341 - use current coroutine context instead of a temporary one
342 when temporarily switching to another coroutine.
343 - do not call C-level on_destroy handlers during global destruction,
344 to avoid needless segfaults.
345
3465.12 Sun Dec 7 13:30:38 CET 2008
347 - add default config for MirOS, which seems to be a bug-to-bug
348 compatible fork of openbsd ("world domination by releasing
349 openbsd cvs before the openbsd folks do it" or so :).
350 - free_padlist did destroy the names pad, not good, but didn't
351 seem to bother perl - this could fix issues such as eval ""
352 inside a function called from multiple coroutines.
353 - use a different method to detect destruction time.
354 - be more careful when freeing padlists just before global
355 destruction.
356 - fixed and expanded the call/cc example.
357 - renamed _terminate to _coro_run.
358 - new method Coro::Channel->shutdown.
359 - try pthreads on openbsd <4.4 (broken sigaltstack, will
360 pthreads fare better?).
361 - be less picky about destroying "a" running coroutine.
362
3635.11 Tue Nov 25 21:49:05 CET 2008
364 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
365 working code with failed assertions, introducing more bugs than
366 it fixes, requiring elaborate workarounds :(
367
3685.1 Mon Nov 24 08:54:59 CET 2008
369 - wrote a small introductory tutorial - Coro::Intro.
370 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
371 - Coro::Select did errornously dup the file descriptors
372 and didn't work with all AnyEvent backends.
373 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
374 LWP data transfers.
375 - disassociate c contexts from coro objects - this is agruably more
376 correct, but mostly allows sharing of cctxs between coro and state
377 objects, for added memory savings and speed increases.
378 - bumped $Coro::POOL_RSS up to 32kb by default.
379 - no longer set the optype to OP_CUSTOM, as B::* understandably
380 doesn't like this very much (and we *are* a type of entersub).
381 - implement state cloning, just to prove that call/cc can be done.
382 - automatically load Coro::AnyEvent in Coro::Handle.
383 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
384 do not have this method (and don't need it either).
385 - speed up generic anyevent methods in Coro::Handle by using rouse
386 callbacks.
387 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
388 others. It also makes the debugger happier, as you can trace
389 through the idle threads now.
390 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
391 - eg/event was pretty broken.
392 - better 5.8.6 compatibility.
393
3945.0 Thu Nov 20 10:35:05 CET 2008
395 - NEW ARCHITECTURE: use the latest 4.x version if you experience
396 stability issues.
397 - bump API version to 7 - all dependents must be recompiled.
398 - removed timed_* functions - they were not being used anyways
399 and should be replaced by a more generic mechanism -
400 and were annoying to support anyways :)
401 - removed SemaphoreSet's waiter method - use sem method instead.
402 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
403 - async_pool did free a scalar value twice
404 ("Attempt to unreference...").
405 - fix a longstanding bug where calling terminate on a coro that
406 was waiting for a semaphore that was just becoming available
407 would cause a deadlock (semaphore would get into a state where
408 it was available but waiters were still blocked).
409 - calling throw on a coroutine that is waiting for a semaphore will
410 no longer make it acquire the semaphore (and thus leak a count).
411 - perl's process emulation is now not even theoretically supported
412 anymore.
413 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
414 conversion of callback-style to blocking-style.
415 - new methods $coro->schedule_to and ->cede_to, to specifically
416 schedule or cede to a specific coroutine.
417 - new function Coro::Semaphore::wait.
418 - use named constants in Coro::Channel (Richard Hundt).
419 - directly patch the entersub opcode calling SLF functions (cede,
420 transfer and so on). this does speed up context switching, but
421 more importanly, it frees us from the hardcoded behaviour of
422 entersub, so we might actually be able to return something from
423 those functions and atcually create new ones.
424 - take advantage of __builtin_frame_address on gcc.
425 - expose THX in coroapi (not sure whether this was a wise decision,
426 as "threaded" perls are running at half speed anyways).
427 - implement execute_slf (schedule-like-function) interface that makes
428 it possible to implement schedule-like-functions in XS.
429 - use new SLF interface to massively speed up Coro::EV by roughly a
430 factor of two.
431 - used new SLF interface to massively speed up Coro::Semaphore by a
432 factor of three.
433 - used new SLF interface to speed up Coro::AIO by roughly a factor of
434 four and reduce its memory usage considerably.
435 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
436 for a nice speedup and vastly more correct behaviour. Also implement
437 a new method "sem" to get at the underlying semaphore object.
438 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
439 (in comparison) 20-40% speedup.
440 - used new SLF interface to reimplement Coro::Signal gaining
441 some unknown (because I was too lazy), but certain, speedup, and also
442 making signals reliable for the first time.
443 - used new SLF interface and other optimisations to speed up async_pool
444 by a factor of two. It also doesn't rely on perl's exception mechanism
445 to exit anymore. The overhead for terminating an async_pool, coro over
446 a normal async is now very small.
447 - sped up coroutine creation/destruction by 40%.
448 - forgot to include Coro/libcoro/README in the dist for all these years.
449 - work around a freebsd pthreads bug (manual testcancel is required as
450 pthread_cond_wait isn't a cancellation point on freebsd).
451 - use new rouse functions to speed up and simplify Coro::BDB.
452 - make "prefer perl native functions" work with threaded perls.
453 - condense Coro::Debug ps output, hint at v and w flags.
454 - (libcoro) lots of minor cleanups and portability improvements.
455
4564.914 Wed Nov 19 12:54:18 CET 2008
457 - fix a disastrous bug in the readline optimisation
458 introduced in 4.801.
459
4604.913 Sat Nov 15 07:58:28 CET 2008
461 - async_pool did free a scalar value twice
462 ("Attempt to unreference...").
463
4644.912 Thu Nov 13 18:31:23 CET 2008
465 - minor cleanups.
466 - use much larger stacks on linux and perl < 5.8.8.
467 - Coro::Debug::new_unix_server did not unlink the socket
468 when destroyed.
469
4704.911 Tue Nov 11 04:26:17 CET 2008
471 - "port" to threaded perls.
472
4734.91 Mon Nov 10 05:36:38 CET 2008
474 - the ->throw exception object no longer leaks.
475 - creating a new cctx leaked a scopestack entry (memleak).
476 - new coroutines didn't get created with a zero flags field
477 (unknown impact).
478 - calling ->throw on a not-yet-started coroutine should now work
479 instead of being ignored.
480 - ->throw is now supported on Coro::State objects.
481 - clean up cctx creation code a bit.
482 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
483
4844.9 Sat Nov 8 17:45:27 CET 2008
485 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
486 - (libcoro) no longer leak threads in the experimental pthread backend,
487 also speed it up considerably.
488 - (libcoro) do not rely on makecontext passing void *'s unscathed.
489 - fix compiletime dependencies on libcoro in the Makefile.
490 - cctx_count wasn't always updated properly.
491 - Coro::State::cctx_stacksize wasn't applied correctly.
492 - new function Coro::State::cctx_max_idle.
493 - the default max number of idle C contexts is now 4.
494 - (libcoro) try harder to get _setjmp/_longjmp.
495 - (libcoro) cleanup and extend the libcoro API to officially
496 allow the creation of empty source contexts.
497 - very experimental workaround for the totally broken netbsd platform.
498 - tried to hack around openbsd bugs.
499
5004.804 Wed Nov 5 16:36:00 CET 2008
501 - Coro::Debug::new_unix_server would not create a non-blocking listening
502 socket, sometimes causing freezes.
503 - (libcoro) fix misaligned stack points for setjmp and assembly
504 methods, which can cause crashes on x86/x86_64 with a sufficiently
505 aggressive compiler.
506 - new function: Coro::Debug::new_tcp_server.
507 - move ->throw into the Coro class because it only works on coro objects.
508
5094.803 Mon Nov 3 17:16:12 CET 2008
510 - (libcoro) use a global asm statement to become independent of gcc
511 otpimisations for CORO_ASM (thanks to pippijn for the idea).
512 - try to workaround yet another broken bsd, this time dragonfly.
513
5144.802 Thu Oct 30 10:56:12 CET 2008
515 - support -fno-omit-frame-pointer on x86 with the assembly method.
516 - tune 01_unblock.t tests a bit.
517
5184.801 Wed Oct 22 18:33:37 CEST 2008
519 - improve readline speed for very long "lines".
520 - backport to 5.8.8.
521
5224.8 Thu Oct 2 13:34:40 CEST 2008
523 - new function Coro::AIO::aio_wait.
524 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
525 - greatly speed up and reduce memory usage of Coro::AIO requests.
526 - implement some other µ-optimisations.
527
5284.749 Mon Sep 29 14:40:12 CEST 2008
529 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
530
5314.748 Sat Sep 27 14:03:19 CEST 2008
532 - implement, but do not document, PerlIO::cede(granularity).
533 - Coro::Storable forgot to wrap Storable::pstore.
534 - work around the multitude of leaks and memory corruption
535 bugs in PerlIO::via by using our own C-level perliol. As a side
536 effect, Coro::Storable is now much, much, much faster.
537
5384.747 Tue Sep 23 01:59:41 CEST 2008
539 - fix a per-cv memleak (one empty array was leaked per
540 code reference).
541 - avoid a crash in coro->call|rss when the coroutine was already
542 destroyed (most noticably when using Coro::Debug::ps :)
543 - also protect *Storable::FILE.
544 - push up default storable granularity to 20ms.
545
5464.746 Sun Sep 21 03:22:20 CEST 2008
547 - be more insistent on locking Storable against reentrancy
548 in Coro::Storable.
549 - move swap_def?v and throw to Coro::State, as documented.
550
5514.745 Thu Jul 24 00:14:38 CEST 2008
552 - remove debugging code related to MgPV_nolen_const, also try to
553 make it compile with perl 5.8.6 (yes, apple apparently loves
554 outdated software). Reported by John S.
555
5564.744 Tue Jul 8 22:06:35 CEST 2008
557 - correctly provide default DIE/WARN handlers as documented.
558 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
559 implementations inside perl work fine for us.
560
5614.743 Mon Jun 16 00:21:57 CEST 2008
562 - when using Coro::EV without running EV::loop it could
563 result in busy-waiting for events, this has been fixed.
564 - reduce codesize and improve performance by using EV_DEFAULT_UC.
565
5664.742 Sat May 31 14:10:21 CEST 2008
567 - implement a workaround for (some) perl <5.8.8 versions.
568 - require EV 3.3+.
569
5704.741 Fri May 30 23:33:09 CEST 2008
571 - tell netbsd how utterly broken their imitation of an OS is
572 and refuse to build by default if pthreads are in use.
573 - switch to "s" method on all bsd's by default, as their ucontext
574 stuff seems just too broken.
575 - fix a bug in Coro::Select.
576
5774.74 Thu May 29 20:05:31 CEST 2008
578 - do not test Coro::LWP for lack of dependencies.
579 - do not test Coro::Select for lack of working perls.
580
5814.73 Thu May 29 2008
582 - fix a bug in Coro::EV which would cause it to block despite
583 there being runnable coroutines.
584 - sprinkle "no warnings" freely over everything, also suppress
585 warnings for some other modules.
586 - fix typo in WSAEWOULDBLOCK.
587
5884.72 Sun May 25 05:14:36 CEST 2008
589 - tweak META.yaml a bit, unfortunately, there is no documented way
590 to have optional dependencies with CPAN. doh :(
591 - avoid running some tests on windows because they would fail due to
592 perl bug (broken fork, broken pipes...).
593 - work around perl on windows bugs where perl returns undocumented
594 error codes for sysread, syswrite etc. by taking advantage
595 of AnyEvent's workaround for that problem.
596 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
597 a common perl implementation bug on windows.
598 - use unix domain sockets in testsuite to work around
599 common perl implementation bugs on widows (they are emulated by
600 tcp sockets on windows. ugh.)
601
6024.71 Sat May 24 20:01:27 CEST 2008
603 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
604 - take advantage of async name resolution of AnyEvent::Util.
605 - work around brutal inet_aton override in Coro::LWP.
606 - take advantage of the readyhook in Coro::EV, for smoother
607 scheduling.
608
6094.7 Sun May 11 00:32:19 CEST 2008
610 - completely reworked the Coro manpage.
611 - added Coro::AnyEvent, generic event loop integration.
612 - implement cancel, ready and kill commands in Coro::Debug.
613 - document find_coro in Coro::Debug.
614 - incompatible change: rename has_stack to has_cctx.
615 - Coro::AIO and Coro::BDB no longer force event model detection,
616 use AnyEvent::AIO and AnyEvent::BDB.
617
6184.6 Sat Apr 26 10:05:14 CEST 2008
619 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
620 by the Coro module. It was a mistake to have it in the first place.
621 - (experimental) support for activestate perl 5.10 (method "w").
622 - (experimental) support for strawberry perl (method "a").
623 - coro_sigelem_set did not return a value although it had to,
624 actual impact unknown.
625
6264.51 Mon Apr 14 13:28:27 CEST 2008
627 - make it compile again on 5.8.
628
6294.50 Thu Apr 10 09:43:17 CEST 2008
630 - I did it twice! (see 4.49).
631
6324.49 Sun Apr 6 21:23:31 CEST 2008
633 - grr, instead of compiling the recent changes on 5.10 only they
634 were compiled on 5.8 only.
635
6364.48 Sun Apr 6 20:36:46 CEST 2008
637 - allow coroutine switches during eval's under 5.10.x, as it seems
638 the parser is a per-interpreter option now, so this is safe
639 (this might fix the odd crashing bug).
640 - drop support for 5.9.x versions: they are dead, jim.
641
6424.47 Sun Apr 6 00:37:52 CEST 2008
643 - force cctx allocation on API calls: we know we need to force one
644 and gcc actually manages to confuse our heuristic nowadays,
645 leading to crashes and worse.
646 - document force_cctx.
647
6484.46 Fri Apr 4 22:05:43 CEST 2008
649 - upgrade libcoro, resulting in pthread-backend (which was only created
650 to fulfill the rules of the programming languages shootout).
651
6524.45 Thu Mar 13 11:55:36 CET 2008
653 - fix a file leaking bug in eg/mhyttpd that would allow
654 downloading of any file (reported by oesi).
655 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
656 (also add testcase).
657 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
658 apparently having loads of fun with that).
659 - implement $state->swap_defsv, ->swap_defav and document ->throw.
660
6614.4 Wed Feb 13 16:44:29 CET 2008
662 - only assume .cfi pseudo isns exist on GNU/Linux.
663 - add get_timed method to Coro::Channels.
664 - fixed Coro::Debug manpage.
665 - perl 5.11 compatbility improvement based on patch by
666 Andreas König.
667
6684.37 Sun Jan 20 11:25:23 CET 2008
669 - soften the check for an active parser for perl 5.10 (avoiding
670 "Coro::State::transfer called while parsing" in many cases).
671
6724.36 Sun Jan 13 10:53:56 CET 2008
673 - reset diehook when terminating from an async_pool as to not
674 trigger any __DIE__ handlers.
675
6764.35 Sun Jan 13 04:14:13 CET 2008
677 - "bt" debug command now displays any exceptions
678 from longmess and also skips the innermost
679 stackframes, giving more useufl output.
680 - allow backtraces in coroutines blocked in Coro::EV,
681 at a <1% speed hit.
682 - handle localising of $SIG{__WARN__} and __DIE__
683 properly (with a proper amount of dirty hacking).
684
6854.34 Sat Dec 22 17:49:53 CET 2007
686 - upgrade to EV version 2.0 API.
687
6884.33 Mon Dec 17 08:36:12 CET 2007
689 - make Coro::AIO etc. loadable in the absence of EV.
690
6914.32 Mon Dec 17 07:46:02 CET 2007
692 - majorly improved Coro::Handle's performance with Coro::EV.
693 - implemented the until now mythical Coro::BDB module.
694 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
695
6964.31 Wed Dec 5 12:32:39 CET 2007
697 - remove warn statement form Coro::Util.
698
6994.3 Tue Dec 4 20:33:14 CET 2007
700 - calls to the idle function could cause stack corruption
701 when the stack changed.
702 - do no longer rely on the presence of EV::DNS if EV is used
703 (because it is gone), but instead take avdantage of EV::ADNS
704 if available.
705 - add ($) prototypes to all functions in Coro::Storable.
706 - use a conventional (safer) idle callback in Coro::EV.
707 - do accept quickly in Coro::Debug to avoid endless loops.
708
7094.22 Fri Nov 30 16:04:04 CET 2007
710 - really use optimised versions for Event and EV in Coro::Util
711 and Coro::Handle.
712
7134.21 Sun Nov 25 10:48:59 CET 2007
714 - fix a spurious memory read.
715 - Coro::EV no longer keeps the eventloop "alive".
716
7174.2 Fri Nov 9 20:47:05 CET 2007
718 - enable/disable tracing from a new coroutine, not a pooled one.
719 - fix a memleak in Coro::Event.
720 - removed killall call from fork_eval.
721 - made sure store_fd is already loaded so that fork_eval
722 does not have to parse autoload in each subprocess.
723 - only use assembly method if -O switch is in $Config{optimize}.
724 - add (optional) Coro::EV module, so far the best event loop module
725 directly supported by Coro.
726 - if the event model is EV, use EV::DNS to resolve
727 stuff in Coro::Util.
728 - don't get confused by multiple event notifications in Coro::Handle.
729 - initial support for EV (libevent interface).
730 - require Event and EV using configure_requires, to force their existance.
731
7324.13 Wed Oct 24 07:26:45 CEST 2007
733 - add Coro::Storable::blocking_thaw.
734 - use a vastly more complicated technique to localise
735 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
736 - use a coroutine for the idle callback Coro::Event,
737 instead of running Event in the current coroutine context.
738 This also catches recursive invocations.
739 - actually report fork errors in gethostbyname and inet_aton.
740
7414.11 Thu Oct 11 02:40:24 CEST 2007
742 - port to threaded perls.
743
7444.1 Thu Oct 11 02:38:16 CEST 2007
745 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
746 be local to each coro (see Coro::State).
747 - incompatible change: for very deep reasons, cede and cede_notself
748 cannot return anything, so nothing will be returned.
749 - possibly bring back 5.10 compatibility (untested).
750 - work around stupid (and wrong) warning on 5.10 :(.
751 - overlay the saved state over the context stack. This saves
752 a few hundred bytes per coroutine on average and also
753 speeds up context switching a bit.
754 - further tune default stack sizes.
755 - (more) correctly calculate stack usage in coro_rss.
756 - Coro::Storable::blocking_* did not properly lock
757 resulting in races between coroutines.
758 - added Coro::Storable::guard.
759 - stopping to trace a coroutine could destroy the cctx of
760 an unrelated coroutine.
761 - explain the relationship between Perl and C coroutines in
762 more detail in Coro::State.
763 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
764 causing a fork per resolve. This also affected Coro::Socket.
765 - switch to a separate stack in $coro->call/eval to avoid
766 invalidating pointers.
767
7684.03 Sat Oct 6 21:24:00 CEST 2007
769 - added Coro::throw method.
770 - minor code cleanups.
771
7724.02 Sat Oct 6 02:36:47 CEST 2007
773 - fix a very minor per-coroutine memleak (a single codereference).
774 - fixed a bug where the currently in-use c context would be freed
775 prematurely (can happen only when programs change the stacksize
776 or use tracing).
777 - tracing can no longer keep a coro alive after it terminated.
778 - do static branch prediction in the common path for gcc. gives
779 about 2-5% speed improvement here.
780
7814.01 Fri Oct 5 22:10:49 CEST 2007
782 - instead of recreating *a* standard output handle we simply
783 use STDOUT, which is faster and hopefully more robust.
7 784
84.0 Fri Oct 5 12:56:00 CEST 2007 7854.0 Fri Oct 5 12:56:00 CEST 2007
9 - incompatibly changed Coro::Storable::freeze. 786 - incompatibly changed Coro::Storable::freeze.
10 - major new feature: added Coro::Debug, for interactive coroutine 787 - major new feature: added Coro::Debug, for interactive coroutine
11 debugging, tracing and much more. 788 debugging, tracing and much more.
… …
45 but stateful semantics. 822 but stateful semantics.
46 - fixed a lot of typos in Coro.pm (patch submitted by David 823 - fixed a lot of typos in Coro.pm (patch submitted by David
47 Steinbrunner, which applied flawlessly). 824 Steinbrunner, which applied flawlessly).
48 825
493.6 Sat Apr 14 17:13:31 CEST 2007 8263.6 Sat Apr 14 17:13:31 CEST 2007
50 - added some bugfixes to get eg/myhttpd working again. 827 - added some bugfixes to get eg/myhttpd working again.
51 - added Coro::Storable for often-cede'ing freeze/thaw. 828 - added Coro::Storable for often-cede'ing freeze/thaw.
52 - try to do a clean exit when a coroutine calls exit 829 - try to do a clean exit when a coroutine calls exit
53 (EXPERIMENTAL). 830 (EXPERIMENTAL).
54 - got rid of indirect call through _coro_init. 831 - got rid of indirect call through _coro_init.
55 - updated the partly antique examples in eg/ to 832 - updated the partly antique examples in eg/ to

Diff Legend

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