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

Comparing Coro/Changes (file contents):
Revision 1.270 by root, Tue Oct 9 02:15:58 2007 UTC vs.
Revision 1.636 by root, Sat Sep 6 20:14:25 2014 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
8TODO: croak when async_pool triesd to run canceled thread?
7 9
84.1 106.41 Sat Sep 6 22:08:46 CEST 2014
11 - restore portability to perl 5.8.x.
12 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
13 up some memory, but fixes "use feature" and similar modules
14 when used inside a Coro.
15 - allow zero as argument to Coro::Channel to mean the same thing
16 as no argument. this works with older versions as well,
17 but wasn't legal until now.
18 - slightly better c header file detection.
19
206.39 Mon Jun 2 00:00:08 CEST 2014
21 - work around more incompatible changes in 5.20.
22
236.38 Sun Jun 1 21:54:23 CEST 2014
24 - check that perl slots actually have enough space to hold
25 interpreter variables.
26 - untested port to perl 5.19 (64 bit tmps indices) (reported
27 by Andreas König).
28 - croak when cancel is called without a thread context.
29
306.37 Tue Mar 4 13:27:33 CET 2014
31 - *sigh*, removed leftover debugging code from debugging a
32 perl bug, of all things.
33
346.36 Tue Mar 4 07:11:59 CET 2014
35 - semaphores would not clear the destroy hook when interrupted
36 by ->throw, causing segfaults or worse.
37 - ->throw on a thread waiting for a semaphore did not acquire
38 the semaphore, but also didn't wake up other waiters,
39 possibly causing a deadlock.
40 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
41 to report a full stacktrace, which should help find out
42 where the actual call is.
43 - "atomically" destroy data in slf_destroy, because it is
44 the right thing to do, just in case.
45 - disable ecb.h libm dependency, to compile on ancient systems
46 or under adverse conditions.
47
486.33 Mon Nov 18 11:26:27 CET 2013
49 - do not crash when freeing padlists with holes (in 5.18).
50 - tentative SVt_BIND 5.19 port/fix.
51
526.32 Tue Nov 5 15:35:35 CET 2013
53 - use a new algorithm to derive padlists for perl 5.18. The old
54 one could lead to 0-pointer accesses inside perl (reported
55 by Darin McBride).
56
576.31 Thu May 9 07:39:48 CEST 2013
58 - Coro::AIO requests would crash if the thread was ready'd
59 while the request was ongoing.
60
616.29 Wed May 8 02:55:18 CEST 2013
62 - when an on_destroy handler destructs the coro currently being
63 destructed a perl scalar could be accessed after being freed,
64 likely causing a crash.
65
666.28 Wed Mar 6 06:58:02 CET 2013
67 - clean remnants of existing __DIE__ and __WARN__ handlers so
68 they lose their magic and will not cause segfaults later
69 (testcase by Andrey Sagulin).
70 - improved Coro::State documentation a bit.
71 - Coro::Debug::command now flushes the output.
72 - add hack detection code for x32 abi, because the braindead slugs
73 who designed that made it look exactly like x86_64 without
74 providing proper compile time symbols to test for it. as a result,
75 this detection cannot work reliably.
76 - valgrind stack registering was broken.
77 - do not rely on Time::HiRes anymore in Coro::Debug.
78
796.23 Fri Dec 7 23:36:37 CET 2012
80 - use experimental fiber implementation on native windows
81 perls.
82 - use sizeof (void *) as multiplication factor for stack sizes,
83 to accomodate the totally braindamaged microsoft 64 bit "os".
84 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
85 activeperl 5.16 crashes when PerlIO_define_layer is called due
86 to some bug in the perl dll, strawberry perl at least passes
87 the testsuite.
88 - implement Coro::Handle->peeraddr/host/port, for slightly
89 improved compatibility with LWP.
90 - implement 5.17 compatibility by almost blindly applying a
91 good-looking patch by Father Chrysostomos.
92 - move stack management functions into libcoro 3.
93 - libcoro version 3 "released".
94 - support magic values as timed_io_once args.
95 - recommend AnyEvent 7+ or EV 4+, also require EV
96 version 4 or newer for Coro::EV.
97
986.10 Tue Oct 9 01:14:27 CEST 2012
99 - updated ecb.h, it had a typo that caused it to not compile on many
100 big endian systems (reported by many people).
101 - disable memory fences in ecb.h to improve portability.
102
1036.09 Sat Oct 6 23:25:02 CEST 2012
104 - Coro::EV I/O watchers were not interruptible by exceptions
105 (Coro::State::throw) (testcase by sten).
106 - ->throw now puts threads into the ready queue, as this seems to
107 be expected by existing code, and code that doesn't cope with spurious
108 wakeups needs fixing anyway.
109 - use fd -1 in mmap.
110 - cast I32 to int in error message printf.
111 - warn about broken so-called "hardened" kernels.
112
1136.08 Fri Apr 13 12:05:47 CEST 2012
114 - be more aggressive about exiting like perl does - formerly,
115 exiting from the non-main thread would not execute END blocks.
116
1176.07 Fri Nov 11 21:21:48 CET 2011
118 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
119 - update ecb.h.
120
1216.06 Mon Aug 8 23:59:48 CEST 2011
122 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
123 backend, which is an order of magnitude slower. unfortunately, cygwins
124 pthread implementation isn't very complete either, so allocate the stack
125 twice just to be sure.
126 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
127 in its ithreaded perl - best recompile cygwin and use the 'w'indows
128 backend for much better performance. also disable ithreads for
129 even better performance...).
130
1316.05 Thu Aug 4 21:36:36 CEST 2011
132 - blush, condvar values would not be propagated from send to recv anymore
133 (reported by Chip Salzenberg).
134 - use exponential increase for the readline buffer length in
135 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
136
1376.04 Wed Aug 3 17:15:45 CEST 2011
138 - use even more efficient and more compatible condvars for
139 compatibility to AnyEvent 6.x :)
140 - more inconsequential ecb.h updates.
141
1426.03 Wed Aug 3 11:41:30 CEST 2011
143 - change how Coro patches AnyEvent condvars for compatibility to
144 AnyEvent 6.x.
145 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
146 hasn't been included.
147
1486.02 Wed Jul 13 04:35:19 CEST 2011
149 - "improve portability to Gentoo" - gentoo manages to put perl variables
150 in memory areas that are farther than 2gb apart, which the jit couldn't
151 handle and barfed. now it's just a bit slower on gentoo and similar
152 systems.
153
1546.01 Sun Jul 3 12:31:14 CEST 2011
155 - workarounds are good, but the test for whether pthreads are used
156 was not good. this one should be better.
157 - check differently whether gcc generates cfi instructions itself.
158
1596.0 Wed Jun 29 19:43:35 CEST 2011
160 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
161 destroyed and cleaned up automatically (e.g. async { schedule }).
162 - implement a JIT compiler for part of the thread switch code,
163 which gives a 50% speed improvement on threaded perls, and
164 about 4% on non-threaded perls (so threaded perls now finally
165 reach about half the speed of non-threaded perls).
166 - slightly modernise Coro::Intro, add section about rouse functions.
167 - avoid DEFSV and ERRSV, giving another 10% improvement
168 in thread switching.
169 - Coro::State->is_destroyed is now called is_zombie.
170 - implement a Coro->safe_cancel method that might fail, but
171 cancels in a "safer" way if it succeeds.
172 - add preliminary support for DEBUGGING perls.
173 - get rid of two hash-accesses when initialising a new Coro - this
174 speeds up coro creation by almost a factor of two.
175 - croak when a coro that is being cancelled tries to block
176 (e.g. while executing a guard block), instead of crashing or
177 deadlocking.
178 - use a more robust and also faster method to identify Coro::State
179 objects - speeds up everything a bit.
180 - implement Coro->cancel in XS for a 20% speed improvement, and to
181 be able to implement mutual cancellation.
182 - speed up context switches by a percent or two by more efficiently
183 allocating context stack entries.
184 - implement Coro->join and Coro->on_destroy in XS for a speedup and
185 a reduction in memory use.
186 - cancelling a coro while it itself is cancelling another coro is
187 now supported and working, instead of triggering an assertion.
188 - be a bit more crash-resistant when calling (buggy) on_destroy
189 callbacks (best effort).
190 - move on_destroy into the slf_frame, to allow extension slf
191 functions to have destructors.
192 - get rid if coro refcounting - simply crash in other interpreter
193 threads by nulling the pointers on clone.
194 - simplify warn/die hook handling when loading Coro - the convoluted
195 logic seems to be no longer neccessary.
196 - use libecb instead of our own home-grown gcc hacks.
197 - document alternatives to Coro::LWP. Please use them :)
198 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
199 sigaltstack. Really. wine suffers from it, erlang suffers from it,
200 and it's known since at least 2006.
201
2025.372 Wed Feb 23 06:14:30 CET 2011
203 - apparently mingw doesn't provide a working gettimeofday, try to
204 work around that by relying on Time::HiRes (indirectly brought to
205 my attention by Max Maischein).
206 - fix some portability issues when Time::HiRes was used.
207
2085.371 Mon Feb 21 14:36:08 CET 2011
209 - backport to windows process emulation code again.
210
2115.37 Sat Feb 19 07:49:44 CET 2011
212 - add a big "Coro thread life cycle" section to "man Coro".
213 - try a tentative workaround against the breakage that 5.13 has
214 introduced without depreciation period. sigh.
215 - no longer use Time::HiRes if gettimeofday is available, which
216 saves quite a lot of memory.
217
2185.36 Sun Feb 13 05:33:41 CET 2011
219 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
220 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
221 method call.
222 - undocument Coro::Timer::sleep and obsolete whole module.
223 - optimise Coro::Timer::timeout memory and cpu usage.
224 - slightly updated Coro::Intro for recent changes.
225 - do not initialise PL_dirty anymore.
226
2275.25 Thu Nov 11 01:08:39 CET 2010
228 - try a different approach on netbsd - netbsd 5 finally has marginally
229 working pthreads, but still broken ucontext/sigaltstack.
230 - openbsd 4.8 finally got their act together, Coro works out of the box
231 with asm, setjmp and pthreads (no change, just informational).
232
2335.24 Sat Oct 23 11:27:12 CEST 2010
234 - port to the EV 4.0 API.
235 - work around bugs in mingw32, making strawberry perl work
236 out of the box.
237 - correctly modify Coro::AIO function prototypes
238 so that they reflect the "no optional parameters" rule.
239 - "ported" libcoro to C++.
240
2415.23 Mon May 17 18:50:42 CEST 2010
242 - be more resistant to ordering changes when initialising
243 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
244 Waldorf).
245 - document that perl 5.12 deliberately removed support for cloning.
246
2475.22 Wed Apr 14 03:55:35 CEST 2010
248 - correctly return udnef on errors in Coro::Handle::read/write
249 (testcase by Marc Mims).
250 - convert Coro::Util into a "perl compatibility wrapper" - the functions
251 are less useful now, but are drop-in replacements for existing
252 functions, listing better alternatives in the documentation. This also
253 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
254 with Coro::Util::inet_aton.
255 - do not override $Coro::idle unconditionally in Coro.pm, as other
256 modules could have provided their own idle coro already
257 (for exmaple, Coro::AnyEvent).
258 - fix Coro::Util::gethost* functions.
259 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
260
2615.21 Wed Dec 16 07:19:51 CET 2009
262 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
263 together.
264 - add some examples on how to combine other event loops with Coro in
265 Coro::AnyEvent, and how to run it (and not to block). Seems to be
266 the most common source of confusion.
267 - try to catch people naively blocking in an event callback.
268 - work around the perl filehandle bug issue in conjunction with
269 older common::sense (as indirectly pointed out by ZSystem).
270 - clarify the "not from signal handlers" section.
271
2725.2 Sun Oct 4 14:54:24 CEST 2009
273 - Coro::Storable destroyed the prototypes of the functions it wrapped.
274 - export rouse_cb and rouse_wait by default now.
275 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
276 - new method $state->swap_sv.
277 - added section on "windows process emulation" to the manpage,
278 after a not-so-fruitful (nor-friendly) "discussion" with chip
279 salzenberg (discussion implies arguments, but his only arguments
280 were ad-hominems, one wonders why he started it in the first
281 place). I hope this explains it well enough for him to understand,
282 and maybe well enough for others to understand.
283 - use common::sense everywhere now.
284 - idle callbacks are no longer supported, use idle coros instead.
285 - print a thread listing when a deadlock is detected.
286
2875.17 Sat Aug 22 23:09:31 CEST 2009
288 - work around a bug in the perl debugger causing crashes
289 when running under the debugger by marking _pool_handler as nodebug.
290 - speed up Coro::async considerably.
291 - try some hacks to get netbsd to work "more often" - their broken
292 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
293 - convert Coro to AE 5.0 API.
294
2955.162 Tue Jul 28 04:04:03 CEST 2009
296 - perl 5.8.2 is now minimum requirement.
297 - skip t/19_handle.t on broken windows perls.
298
2995.161 Wed Jul 22 04:47:38 CEST 2009
300 - Coro::AnyEvent::poll could have a different prototype when EV was
301 used as backend (analyzed by Tatsuhiko Miyagawa).
302 - Coro::AnyEvent errornously initialised the event loop when loaded,
303 not on demand.
304 - try to workaround rare */t/01_unblock.t failures.
305
3065.16 Tue Jul 21 01:44:37 CEST 2009
307 - Coro::AnyEvent failed to hook into the event loop
308 when no threads had been readied between detecting
309 the event loop and actually running it.
310 - considerably speed up Coro::Select by taking avdantage
311 of AnyEvent > 4.8 and some other optimisations.
312 - implement paragraph readline mode in Coro::Handle
313 (based on patches by Zsbán Ambrus).
314 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
315 and analyzed by Yasuhiro MATSUMOTO).
316 - clarified libcoro license and copyright.
317 - someone stole my EXTRA_META!!!
318 - implement Coro::Select::patch_pp_sselect and it's brother,
319 for hardcode select overriding.
320
3215.151 Mon Jul 6 05:41:57 CEST 2009
322 - backport to windows process emulation code again (patch by
323 Yasuhiro MATSUMOTO).
324 - slightly update Coro::MakeMaker.
325
3265.15 Tue Jun 30 10:28:06 CEST 2009
327 - deprecate Coro::Socket, document how to get ipv6 support via
328 AnyEvent::Socket instead.
329 - implement signal->wait ($cb) interface, similar to semaphores.
330 - work around SvOK not supporting getmagic, so we have to getmagic
331 to test for undef :/ (reported by Matthias Waldorf).
332 - load Coro::AnyEvent in all modules using AnyEvent.
333 - work around perl corrupting our internal data structures,
334 reported by Tokuhiro Matsuno.
335 - enable per-coroutine real and cpu time gathering
336 (Coro::State::enable_times).
337
3385.14 Wed Jun 24 01:37:48 CEST 2009
339 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
340 as this seems to have been a difficult concept (poll, sleep, idle,
341 idle_upto).
342 - add Coro::AnyEvent::readable/writable functions.
343 - clarify Coro::EV/Event/AnyEvent manpages.
344 - free per-thread global scalars in the thread calling ->cancel, to
345 avoid crashes when $_, $@ etc., are magical but some of those
346 had already been freed.
347 - "unexperimentalise" the callback interface for Coro::Semaphore.
348 - speed up ready queue management/context switching by using a linked
349 list instead of an array (~5-10%).
350 - implement "watch" command in Coro::Debug shells.
351 - for fun, implement time-slicing as an example in the manpage.
352 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
353 Coro::Event, use the normal AnyEvent handling instead of dieing
354 (the same is true for Coro::Handle).
355 - properly document Coro::EV::timed_io_once.
356 - avoid unneccessary ->cancel calls in Coro::Handle.
357 - maybe make it work on mingw32 with win32 backend
358 (based on patch by Yasuhiro Matsumoto).
359
3605.132 Fri May 29 09:00:39 CEST 2009
361 - do not keep a reference to the argument itself in
362 Coro::Semaphore::guard, as it could change later.
363 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
364 should support a prepare callback.
365
3665.131 Mon Mar 16 23:20:37 CET 2009
367 - implement and document Coro->suspend, Coro->resume.
368 - fix Coro::Select implementation to not (often) close
369 the passed file descriptors (testcase provided by pippijn).
370
3715.13 Mon Dec 15 21:51:42 CET 2008
372 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
373 - don't set diehook to C<undef> but instead to NULL, to avoid
374 spurious warnings.
375 - fix a lot of bugs in Coro::SemaphoreSet.
376 - Coro::SemaphoreSet will less often create a semaphore needlessly.
377 - add Coro::SemaphoreSet::count and wait methods.
378 - take advantage of the new Guard module.
379 - deprecate Coro::guard.
380 - try to fix the dreaded 01_unblock tests once more. I hate it when
381 testsuites need more fixing than the code they are supposed to test.
382 - croak in more cases when a required callback isn't resolvable.
383 - fix some minor issues in Coro::State->call/eval.
384 - use current coroutine context instead of a temporary one
385 when temporarily switching to another coroutine.
386 - do not call C-level on_destroy handlers during global destruction,
387 to avoid needless segfaults.
388
3895.12 Sun Dec 7 13:30:38 CET 2008
390 - add default config for MirOS, which seems to be a bug-to-bug
391 compatible fork of openbsd ("world domination by releasing
392 openbsd cvs before the openbsd folks do it" or so :).
393 - free_padlist did destroy the names pad, not good, but didn't
394 seem to bother perl - this could fix issues such as eval ""
395 inside a function called from multiple coroutines.
396 - use a different method to detect destruction time.
397 - be more careful when freeing padlists just before global
398 destruction.
399 - fixed and expanded the call/cc example.
400 - renamed _terminate to _coro_run.
401 - new method Coro::Channel->shutdown.
402 - try pthreads on openbsd <4.4 (broken sigaltstack, will
403 pthreads fare better?).
404 - be less picky about destroying "a" running coroutine.
405
4065.11 Tue Nov 25 21:49:05 CET 2008
407 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
408 working code with failed assertions, introducing more bugs than
409 it fixes, requiring elaborate workarounds :(
410
4115.1 Mon Nov 24 08:54:59 CET 2008
412 - wrote a small introductory tutorial - Coro::Intro.
413 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
414 - Coro::Select did errornously dup the file descriptors
415 and didn't work with all AnyEvent backends.
416 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
417 LWP data transfers.
418 - disassociate c contexts from coro objects - this is agruably more
419 correct, but mostly allows sharing of cctxs between coro and state
420 objects, for added memory savings and speed increases.
421 - bumped $Coro::POOL_RSS up to 32kb by default.
422 - no longer set the optype to OP_CUSTOM, as B::* understandably
423 doesn't like this very much (and we *are* a type of entersub).
424 - implement state cloning, just to prove that call/cc can be done.
425 - automatically load Coro::AnyEvent in Coro::Handle.
426 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
427 do not have this method (and don't need it either).
428 - speed up generic anyevent methods in Coro::Handle by using rouse
429 callbacks.
430 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
431 others. It also makes the debugger happier, as you can trace
432 through the idle threads now.
433 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
434 - eg/event was pretty broken.
435 - better 5.8.6 compatibility.
436
4375.0 Thu Nov 20 10:35:05 CET 2008
438 - NEW ARCHITECTURE: use the latest 4.x version if you experience
439 stability issues.
440 - bump API version to 7 - all dependents must be recompiled.
441 - removed timed_* functions - they were not being used anyways
442 and should be replaced by a more generic mechanism -
443 and were annoying to support anyways :)
444 - removed SemaphoreSet's waiter method - use sem method instead.
445 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
446 - async_pool did free a scalar value twice
447 ("Attempt to unreference...").
448 - fix a longstanding bug where calling terminate on a coro that
449 was waiting for a semaphore that was just becoming available
450 would cause a deadlock (semaphore would get into a state where
451 it was available but waiters were still blocked).
452 - calling throw on a coroutine that is waiting for a semaphore will
453 no longer make it acquire the semaphore (and thus leak a count).
454 - perl's process emulation is now not even theoretically supported
455 anymore.
456 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
457 conversion of callback-style to blocking-style.
458 - new methods $coro->schedule_to and ->cede_to, to specifically
459 schedule or cede to a specific coroutine.
460 - new function Coro::Semaphore::wait.
461 - use named constants in Coro::Channel (Richard Hundt).
462 - directly patch the entersub opcode calling SLF functions (cede,
463 transfer and so on). this does speed up context switching, but
464 more importanly, it frees us from the hardcoded behaviour of
465 entersub, so we might actually be able to return something from
466 those functions and atcually create new ones.
467 - take advantage of __builtin_frame_address on gcc.
468 - expose THX in coroapi (not sure whether this was a wise decision,
469 as "threaded" perls are running at half speed anyways).
470 - implement execute_slf (schedule-like-function) interface that makes
471 it possible to implement schedule-like-functions in XS.
472 - use new SLF interface to massively speed up Coro::EV by roughly a
473 factor of two.
474 - used new SLF interface to massively speed up Coro::Semaphore by a
475 factor of three.
476 - used new SLF interface to speed up Coro::AIO by roughly a factor of
477 four and reduce its memory usage considerably.
478 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
479 for a nice speedup and vastly more correct behaviour. Also implement
480 a new method "sem" to get at the underlying semaphore object.
481 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
482 (in comparison) 20-40% speedup.
483 - used new SLF interface to reimplement Coro::Signal gaining
484 some unknown (because I was too lazy), but certain, speedup, and also
485 making signals reliable for the first time.
486 - used new SLF interface and other optimisations to speed up async_pool
487 by a factor of two. It also doesn't rely on perl's exception mechanism
488 to exit anymore. The overhead for terminating an async_pool, coro over
489 a normal async is now very small.
490 - sped up coroutine creation/destruction by 40%.
491 - forgot to include Coro/libcoro/README in the dist for all these years.
492 - work around a freebsd pthreads bug (manual testcancel is required as
493 pthread_cond_wait isn't a cancellation point on freebsd).
494 - use new rouse functions to speed up and simplify Coro::BDB.
495 - make "prefer perl native functions" work with threaded perls.
496 - condense Coro::Debug ps output, hint at v and w flags.
497 - (libcoro) lots of minor cleanups and portability improvements.
498
4994.914 Wed Nov 19 12:54:18 CET 2008
500 - fix a disastrous bug in the readline optimisation
501 introduced in 4.801.
502
5034.913 Sat Nov 15 07:58:28 CET 2008
504 - async_pool did free a scalar value twice
505 ("Attempt to unreference...").
506
5074.912 Thu Nov 13 18:31:23 CET 2008
508 - minor cleanups.
509 - use much larger stacks on linux and perl < 5.8.8.
510 - Coro::Debug::new_unix_server did not unlink the socket
511 when destroyed.
512
5134.911 Tue Nov 11 04:26:17 CET 2008
514 - "port" to threaded perls.
515
5164.91 Mon Nov 10 05:36:38 CET 2008
517 - the ->throw exception object no longer leaks.
518 - creating a new cctx leaked a scopestack entry (memleak).
519 - new coroutines didn't get created with a zero flags field
520 (unknown impact).
521 - calling ->throw on a not-yet-started coroutine should now work
522 instead of being ignored.
523 - ->throw is now supported on Coro::State objects.
524 - clean up cctx creation code a bit.
525 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
526
5274.9 Sat Nov 8 17:45:27 CET 2008
528 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
529 - (libcoro) no longer leak threads in the experimental pthread backend,
530 also speed it up considerably.
531 - (libcoro) do not rely on makecontext passing void *'s unscathed.
532 - fix compiletime dependencies on libcoro in the Makefile.
533 - cctx_count wasn't always updated properly.
534 - Coro::State::cctx_stacksize wasn't applied correctly.
535 - new function Coro::State::cctx_max_idle.
536 - the default max number of idle C contexts is now 4.
537 - (libcoro) try harder to get _setjmp/_longjmp.
538 - (libcoro) cleanup and extend the libcoro API to officially
539 allow the creation of empty source contexts.
540 - very experimental workaround for the totally broken netbsd platform.
541 - tried to hack around openbsd bugs.
542
5434.804 Wed Nov 5 16:36:00 CET 2008
544 - Coro::Debug::new_unix_server would not create a non-blocking listening
545 socket, sometimes causing freezes.
546 - (libcoro) fix misaligned stack points for setjmp and assembly
547 methods, which can cause crashes on x86/x86_64 with a sufficiently
548 aggressive compiler.
549 - new function: Coro::Debug::new_tcp_server.
550 - move ->throw into the Coro class because it only works on coro objects.
551
5524.803 Mon Nov 3 17:16:12 CET 2008
553 - (libcoro) use a global asm statement to become independent of gcc
554 otpimisations for CORO_ASM (thanks to pippijn for the idea).
555 - try to workaround yet another broken bsd, this time dragonfly.
556
5574.802 Thu Oct 30 10:56:12 CET 2008
558 - support -fno-omit-frame-pointer on x86 with the assembly method.
559 - tune 01_unblock.t tests a bit.
560
5614.801 Wed Oct 22 18:33:37 CEST 2008
562 - improve readline speed for very long "lines".
563 - backport to 5.8.8.
564
5654.8 Thu Oct 2 13:34:40 CEST 2008
566 - new function Coro::AIO::aio_wait.
567 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
568 - greatly speed up and reduce memory usage of Coro::AIO requests.
569 - implement some other µ-optimisations.
570
5714.749 Mon Sep 29 14:40:12 CEST 2008
572 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
573
5744.748 Sat Sep 27 14:03:19 CEST 2008
575 - implement, but do not document, PerlIO::cede(granularity).
576 - Coro::Storable forgot to wrap Storable::pstore.
577 - work around the multitude of leaks and memory corruption
578 bugs in PerlIO::via by using our own C-level perliol. As a side
579 effect, Coro::Storable is now much, much, much faster.
580
5814.747 Tue Sep 23 01:59:41 CEST 2008
582 - fix a per-cv memleak (one empty array was leaked per
583 code reference).
584 - avoid a crash in coro->call|rss when the coroutine was already
585 destroyed (most noticably when using Coro::Debug::ps :)
586 - also protect *Storable::FILE.
587 - push up default storable granularity to 20ms.
588
5894.746 Sun Sep 21 03:22:20 CEST 2008
590 - be more insistent on locking Storable against reentrancy
591 in Coro::Storable.
592 - move swap_def?v and throw to Coro::State, as documented.
593
5944.745 Thu Jul 24 00:14:38 CEST 2008
595 - remove debugging code related to MgPV_nolen_const, also try to
596 make it compile with perl 5.8.6 (yes, apple apparently loves
597 outdated software). Reported by John S.
598
5994.744 Tue Jul 8 22:06:35 CEST 2008
600 - correctly provide default DIE/WARN handlers as documented.
601 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
602 implementations inside perl work fine for us.
603
6044.743 Mon Jun 16 00:21:57 CEST 2008
605 - when using Coro::EV without running EV::loop it could
606 result in busy-waiting for events, this has been fixed.
607 - reduce codesize and improve performance by using EV_DEFAULT_UC.
608
6094.742 Sat May 31 14:10:21 CEST 2008
610 - implement a workaround for (some) perl <5.8.8 versions.
611 - require EV 3.3+.
612
6134.741 Fri May 30 23:33:09 CEST 2008
614 - tell netbsd how utterly broken their imitation of an OS is
615 and refuse to build by default if pthreads are in use.
616 - switch to "s" method on all bsd's by default, as their ucontext
617 stuff seems just too broken.
618 - fix a bug in Coro::Select.
619
6204.74 Thu May 29 20:05:31 CEST 2008
621 - do not test Coro::LWP for lack of dependencies.
622 - do not test Coro::Select for lack of working perls.
623
6244.73 Thu May 29 2008
625 - fix a bug in Coro::EV which would cause it to block despite
626 there being runnable coroutines.
627 - sprinkle "no warnings" freely over everything, also suppress
628 warnings for some other modules.
629 - fix typo in WSAEWOULDBLOCK.
630
6314.72 Sun May 25 05:14:36 CEST 2008
632 - tweak META.yaml a bit, unfortunately, there is no documented way
633 to have optional dependencies with CPAN. doh :(
634 - avoid running some tests on windows because they would fail due to
635 perl bug (broken fork, broken pipes...).
636 - work around perl on windows bugs where perl returns undocumented
637 error codes for sysread, syswrite etc. by taking advantage
638 of AnyEvent's workaround for that problem.
639 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
640 a common perl implementation bug on windows.
641 - use unix domain sockets in testsuite to work around
642 common perl implementation bugs on widows (they are emulated by
643 tcp sockets on windows. ugh.)
644
6454.71 Sat May 24 20:01:27 CEST 2008
646 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
647 - take advantage of async name resolution of AnyEvent::Util.
648 - work around brutal inet_aton override in Coro::LWP.
649 - take advantage of the readyhook in Coro::EV, for smoother
650 scheduling.
651
6524.7 Sun May 11 00:32:19 CEST 2008
653 - completely reworked the Coro manpage.
654 - added Coro::AnyEvent, generic event loop integration.
655 - implement cancel, ready and kill commands in Coro::Debug.
656 - document find_coro in Coro::Debug.
657 - incompatible change: rename has_stack to has_cctx.
658 - Coro::AIO and Coro::BDB no longer force event model detection,
659 use AnyEvent::AIO and AnyEvent::BDB.
660
6614.6 Sat Apr 26 10:05:14 CEST 2008
662 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
663 by the Coro module. It was a mistake to have it in the first place.
664 - (experimental) support for activestate perl 5.10 (method "w").
665 - (experimental) support for strawberry perl (method "a").
666 - coro_sigelem_set did not return a value although it had to,
667 actual impact unknown.
668
6694.51 Mon Apr 14 13:28:27 CEST 2008
670 - make it compile again on 5.8.
671
6724.50 Thu Apr 10 09:43:17 CEST 2008
673 - I did it twice! (see 4.49).
674
6754.49 Sun Apr 6 21:23:31 CEST 2008
676 - grr, instead of compiling the recent changes on 5.10 only they
677 were compiled on 5.8 only.
678
6794.48 Sun Apr 6 20:36:46 CEST 2008
680 - allow coroutine switches during eval's under 5.10.x, as it seems
681 the parser is a per-interpreter option now, so this is safe
682 (this might fix the odd crashing bug).
683 - drop support for 5.9.x versions: they are dead, jim.
684
6854.47 Sun Apr 6 00:37:52 CEST 2008
686 - force cctx allocation on API calls: we know we need to force one
687 and gcc actually manages to confuse our heuristic nowadays,
688 leading to crashes and worse.
689 - document force_cctx.
690
6914.46 Fri Apr 4 22:05:43 CEST 2008
692 - upgrade libcoro, resulting in pthread-backend (which was only created
693 to fulfill the rules of the programming languages shootout).
694
6954.45 Thu Mar 13 11:55:36 CET 2008
696 - fix a file leaking bug in eg/mhyttpd that would allow
697 downloading of any file (reported by oesi).
698 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
699 (also add testcase).
700 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
701 apparently having loads of fun with that).
702 - implement $state->swap_defsv, ->swap_defav and document ->throw.
703
7044.4 Wed Feb 13 16:44:29 CET 2008
705 - only assume .cfi pseudo isns exist on GNU/Linux.
706 - add get_timed method to Coro::Channels.
707 - fixed Coro::Debug manpage.
708 - perl 5.11 compatbility improvement based on patch by
709 Andreas König.
710
7114.37 Sun Jan 20 11:25:23 CET 2008
712 - soften the check for an active parser for perl 5.10 (avoiding
713 "Coro::State::transfer called while parsing" in many cases).
714
7154.36 Sun Jan 13 10:53:56 CET 2008
716 - reset diehook when terminating from an async_pool as to not
717 trigger any __DIE__ handlers.
718
7194.35 Sun Jan 13 04:14:13 CET 2008
720 - "bt" debug command now displays any exceptions
721 from longmess and also skips the innermost
722 stackframes, giving more useufl output.
723 - allow backtraces in coroutines blocked in Coro::EV,
724 at a <1% speed hit.
725 - handle localising of $SIG{__WARN__} and __DIE__
726 properly (with a proper amount of dirty hacking).
727
7284.34 Sat Dec 22 17:49:53 CET 2007
729 - upgrade to EV version 2.0 API.
730
7314.33 Mon Dec 17 08:36:12 CET 2007
732 - make Coro::AIO etc. loadable in the absence of EV.
733
7344.32 Mon Dec 17 07:46:02 CET 2007
735 - majorly improved Coro::Handle's performance with Coro::EV.
736 - implemented the until now mythical Coro::BDB module.
737 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
738
7394.31 Wed Dec 5 12:32:39 CET 2007
740 - remove warn statement form Coro::Util.
741
7424.3 Tue Dec 4 20:33:14 CET 2007
743 - calls to the idle function could cause stack corruption
744 when the stack changed.
745 - do no longer rely on the presence of EV::DNS if EV is used
746 (because it is gone), but instead take avdantage of EV::ADNS
747 if available.
748 - add ($) prototypes to all functions in Coro::Storable.
749 - use a conventional (safer) idle callback in Coro::EV.
750 - do accept quickly in Coro::Debug to avoid endless loops.
751
7524.22 Fri Nov 30 16:04:04 CET 2007
753 - really use optimised versions for Event and EV in Coro::Util
754 and Coro::Handle.
755
7564.21 Sun Nov 25 10:48:59 CET 2007
757 - fix a spurious memory read.
758 - Coro::EV no longer keeps the eventloop "alive".
759
7604.2 Fri Nov 9 20:47:05 CET 2007
761 - enable/disable tracing from a new coroutine, not a pooled one.
762 - fix a memleak in Coro::Event.
763 - removed killall call from fork_eval.
764 - made sure store_fd is already loaded so that fork_eval
765 does not have to parse autoload in each subprocess.
766 - only use assembly method if -O switch is in $Config{optimize}.
767 - add (optional) Coro::EV module, so far the best event loop module
768 directly supported by Coro.
769 - if the event model is EV, use EV::DNS to resolve
770 stuff in Coro::Util.
771 - don't get confused by multiple event notifications in Coro::Handle.
772 - initial support for EV (libevent interface).
773 - require Event and EV using configure_requires, to force their existance.
774
7754.13 Wed Oct 24 07:26:45 CEST 2007
776 - add Coro::Storable::blocking_thaw.
777 - use a vastly more complicated technique to localise
778 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
779 - use a coroutine for the idle callback Coro::Event,
780 instead of running Event in the current coroutine context.
781 This also catches recursive invocations.
782 - actually report fork errors in gethostbyname and inet_aton.
783
7844.11 Thu Oct 11 02:40:24 CEST 2007
785 - port to threaded perls.
786
7874.1 Thu Oct 11 02:38:16 CEST 2007
9 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 788 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
10 be local to each coro (see Coro::State). 789 be local to each coro (see Coro::State).
790 - incompatible change: for very deep reasons, cede and cede_notself
791 cannot return anything, so nothing will be returned.
11 - possibly bring back 5.10 compatibility (untested). 792 - possibly bring back 5.10 compatibility (untested).
12 - work around stupid (and wrong) warning on 5.10 :(. 793 - work around stupid (and wrong) warning on 5.10 :(.
13 - overlay the saved state over the context stack. This saves 794 - overlay the saved state over the context stack. This saves
14 a few hundred bytes per coroutine on average and also 795 a few hundred bytes per coroutine on average and also
15 speeds up context switching a bit. 796 speeds up context switching a bit.
… …
22 an unrelated coroutine. 803 an unrelated coroutine.
23 - explain the relationship between Perl and C coroutines in 804 - explain the relationship between Perl and C coroutines in
24 more detail in Coro::State. 805 more detail in Coro::State.
25 - Coro::Util::inet_aton did not short-circuit dotted quad forms, 806 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
26 causing a fork per resolve. This also affected Coro::Socket. 807 causing a fork per resolve. This also affected Coro::Socket.
808 - switch to a separate stack in $coro->call/eval to avoid
809 invalidating pointers.
27 810
284.03 Sat Oct 6 21:24:00 CEST 2007 8114.03 Sat Oct 6 21:24:00 CEST 2007
29 - added Coro::throw method. 812 - added Coro::throw method.
30 - minor code cleanups. 813 - minor code cleanups.
31 814
… …
82 but stateful semantics. 865 but stateful semantics.
83 - fixed a lot of typos in Coro.pm (patch submitted by David 866 - fixed a lot of typos in Coro.pm (patch submitted by David
84 Steinbrunner, which applied flawlessly). 867 Steinbrunner, which applied flawlessly).
85 868
863.6 Sat Apr 14 17:13:31 CEST 2007 8693.6 Sat Apr 14 17:13:31 CEST 2007
87 - added some bugfixes to get eg/myhttpd working again. 870 - added some bugfixes to get eg/myhttpd working again.
88 - added Coro::Storable for often-cede'ing freeze/thaw. 871 - added Coro::Storable for often-cede'ing freeze/thaw.
89 - try to do a clean exit when a coroutine calls exit 872 - try to do a clean exit when a coroutine calls exit
90 (EXPERIMENTAL). 873 (EXPERIMENTAL).
91 - got rid of indirect call through _coro_init. 874 - got rid of indirect call through _coro_init.
92 - updated the partly antique examples in eg/ to 875 - updated the partly antique examples in eg/ to

Diff Legend

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