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

Comparing Coro/Changes (file contents):
Revision 1.435 by root, Wed Nov 19 02:07:48 2008 UTC vs.
Revision 1.612 by root, Fri Feb 8 22:29:18 2013 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::coroaio (perl leaks like hell). 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: channel->maxsize(newsize)?
5 actor model (which is cool in a lot of important aspects (failures!), 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)).
7 6
8TODO: testcancel 7 - improved Coro::State documentation a bit.
9TODO: guard as SLF - to avoid deadlocks. 8 - Coro::Debug::command now flushes the output.
10 - avoid deadlock in Coro::Channel if maxsize == 1 (Richard Hundt). 9 - add hack detection code for x32 abi, because the braindead slugs
10 who designed that made it look exactly like x86_64 without
11 providing proper compile time symbols to test for it. as a result,
12 this detection cannot work reliably.
13 - valgrind stack registering was broken.
14 - do not rely on Time::HiRes anymore in Coro::Debug.
11 15
125.0 166.23 Fri Dec 7 23:36:37 CET 2012
17 - use experimental fiber implementation on native windows
18 perls.
19 - use sizeof (void *) as multiplication factor for stack sizes,
20 to accomodate the totally braindamaged microsoft 64 bit "os".
21 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
22 activeperl 5.16 crashes when PerlIO_define_layer is called due
23 to some bug in the perl dll, strawberry perl at least passes
24 the testsuite.
25 - implement Coro::Handle->peeraddr/host/port, for slightly
26 improved compatibility with LWP.
27 - implement 5.17 compatibility by almost blindly applying a
28 good-looking patch by Father Chrysostomos.
29 - move stack management functions into libcoro 3.
30 - libcoro version 3 "released".
31 - support magic values as timed_io_once args.
32 - recommend AnyEvent 7+ or EV 4+, also require EV
33 version 4 or newer for Coro::EV.
34
356.10 Tue Oct 9 01:14:27 CEST 2012
36 - updated ecb.h, it had a typo that caused it to not compile on many
37 big endian systems (reported by many people).
38 - disable memory fences in ecb.h to improve portability.
39
406.09 Sat Oct 6 23:25:02 CEST 2012
41 - Coro::EV I/O watchers were not interruptible by exceptions
42 (Coro::State::throw) (testcase by sten).
43 - ->throw now puts threads into the ready queue, as this seems to
44 be expected by existing code, and code that doesn't cope with spurious
45 wakeups needs fixing anyway.
46 - use fd -1 in mmap.
47 - cast I32 to int in error message printf.
48 - warn about broken so-called "hardened" kernels.
49
506.08 Fri Apr 13 12:05:47 CEST 2012
51 - be more aggressive about exiting like perl does - formerly,
52 exiting from the non-main thread would not execute END blocks.
53
546.07 Fri Nov 11 21:21:48 CET 2011
55 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
56 - update ecb.h.
57
586.06 Mon Aug 8 23:59:48 CEST 2011
59 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
60 backend, which is an order of magnitude slower. unfortunately, cygwins
61 pthread implementation isn't very complete either, so allocate the stack
62 twice just to be sure.
63 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
64 in its ithreaded perl - best recompile cygwin and use the 'w'indows
65 backend for much better performance. also disable ithreads for
66 even better performance...).
67
686.05 Thu Aug 4 21:36:36 CEST 2011
69 - blush, condvar values would not be propagated from send to recv anymore
70 (reported by Chip Salzenberg).
71 - use exponential increase for the readline buffer length in
72 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
73
746.04 Wed Aug 3 17:15:45 CEST 2011
75 - use even more efficient and more compatible condvars for
76 compatibility to AnyEvent 6.x :)
77 - more inconsequential ecb.h updates.
78
796.03 Wed Aug 3 11:41:30 CEST 2011
80 - change how Coro patches AnyEvent condvars for compatibility to
81 AnyEvent 6.x.
82 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
83 hasn't been included.
84
856.02 Wed Jul 13 04:35:19 CEST 2011
86 - "improve portability to Gentoo" - gentoo manages to put perl variables
87 in memory areas that are farther than 2gb apart, which the jit couldn't
88 handle and barfed. now it's just a bit slower on gentoo and similar
89 systems.
90
916.01 Sun Jul 3 12:31:14 CEST 2011
92 - workarounds are good, but the test for whether pthreads are used
93 was not good. thisone should be better.
94 - check differently whether gcc generates cfi instructions itself.
95
966.0 Wed Jun 29 19:43:35 CEST 2011
97 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
98 destroyed and cleaned up automatically (e.g. async { schedule }).
99 - implement a JIT compiler for part of the thread switch code,
100 which gives a 50% speed improvement on threaded perls, and
101 about 4% on non-thraeded perls (so threaded perls now finally
102 reach about half the speed of non-threaded perls).
103 - slightly modernise Coro::Intro, add section about rouse functions.
104 - avoid DEFSV and ERRSV, giving another 10% improvement
105 in thread switching.
106 - Coro::State->is_destroyed is now called is_zombie.
107 - implement a Coro->safe_cancel method that might fail, but
108 cancels in a "safer" way if it succeeds.
109 - add preliminary support for DEBUGGING perls.
110 - get rid of two hash-accesses when initialising a new Coro - this
111 speeds up coro creation by almost a factor of two.
112 - croak when a coro that is being cancelled tries to block
113 (e.g. while executing a guard block), instead of crashing or
114 deadlocking.
115 - use a more robust and also faster method to identify Coro::State
116 objects - speeds up everything a bit.
117 - implement Coro->cancel in XS for a 20% speed improvement, and to
118 be able to implement mutual cancellation.
119 - speed up context switches by a percent or two by more efficiently
120 allocating context stack entries.
121 - implement Coro->join and Coro->on_destroy in XS for a speedup and
122 a reduction in memory use.
123 - cancelling a coro while it itself is cancelling another coro is
124 now supported and working, instead of triggering an assertion.
125 - be a bit more crash-resistant when calling (buggy) on_destroy
126 callbacks (best effort).
127 - move on_destroy into the slf_frame, to allow extension slf
128 functions to have destructors.
129 - get rid if coro refcounting - simply crash in other interpreter
130 threads by nulling the pointers on clone.
131 - simplify warn/die hook handling when loading Coro - the convoluted
132 logic seems to be no longer neccessary.
133 - use libecb instead of our own home-grown gcc hacks.
134 - document alternatives to Coro::LWP. Please use them :)
135 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
136 sigaltstack. Really. wine suffers from it, erlang suffers from it,
137 and it's known since at least 2006.
138
1395.372 Wed Feb 23 06:14:30 CET 2011
140 - apparently mingw doesn't provide a working gettimeofday, try to
141 work around that by relying on Time::HiRes (indirectly brought to
142 my attention by Max Maischein).
143 - fix some portability issues when Time::HiRes was used.
144
1455.371 Mon Feb 21 14:36:08 CET 2011
146 - backport to windows process emulation code again.
147
1485.37 Sat Feb 19 07:49:44 CET 2011
149 - add a big "Coro thread life cycle" section to "man Coro".
150 - try a tentative workaround against the breakage that 5.13 has
151 introduced without depreciation period. sigh.
152 - no longer use Time::HiRes if gettimeofday is available, which
153 saves quite a lot of memory.
154
1555.36 Sun Feb 13 05:33:41 CET 2011
156 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
157 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
158 method call.
159 - undocument Coro::Timer::sleep and obsolete whole module.
160 - optimise Coro::Timer::timeout memory and cpu usage.
161 - slightly updated Coro::Intro for recent changes.
162 - do not initialise PL_dirty anymore.
163
1645.25 Thu Nov 11 01:08:39 CET 2010
165 - try a different approach on netbsd - netbsd 5 finally has marginally
166 working pthreads, but still broken ucontext/sigaltstack.
167 - openbsd 4.8 finally got their act together, Coro works out of the box
168 with asm, setjmp and pthreads (no change, just informational).
169
1705.24 Sat Oct 23 11:27:12 CEST 2010
171 - port to the EV 4.0 API.
172 - work around bugs in mingw32, making strawberry perl work
173 out of the box.
174 - correctly modify Coro::AIO function prototypes
175 so that they reflect the "no optional parameters" rule.
176 - "ported" libcoro to C++.
177
1785.23 Mon May 17 18:50:42 CEST 2010
179 - be more resistant to ordering changes when initialising
180 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
181 Waldorf).
182 - document that perl 5.12 deliberately removed support for cloning.
183
1845.22 Wed Apr 14 03:55:35 CEST 2010
185 - correctly return udnef on errors in Coro::Handle::read/write
186 (testcase by Marc Mims).
187 - convert Coro::Util into a "perl compatibility wrapper" - the functions
188 are less useful now, but are drop-in replacements for existing
189 functions, listing better alternatives in the documentation. This also
190 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
191 with Coro::Util::inet_aton.
192 - do not override $Coro::idle unconditionally in Coro.pm, as other
193 modules could have provided their own idle coro already
194 (for exmaple, Coro::AnyEvent).
195 - fix Coro::Util::gethost* functions.
196 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
197
1985.21 Wed Dec 16 07:19:51 CET 2009
199 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
200 together.
201 - add some examples on how to combine other event loops with Coro in
202 Coro::AnyEvent, and how to run it (and not to block). Seems to be
203 the most common source of confusion.
204 - try to catch people naively blocking in an event callback.
205 - work around the perl filehandle bug issue in conjunction with
206 older common::sense (as indirectly pointed out by ZSystem).
207 - clarify the "not from signal handlers" section.
208
2095.2 Sun Oct 4 14:54:24 CEST 2009
210 - Coro::Storable destroyed the prototypes of the functions it wrapped.
211 - export rouse_cb and rouse_wait by default now.
212 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
213 - new method $state->swap_sv.
214 - added section on "windows process emulation" to the manpage,
215 after a not-so-fruitful (nor-friendly) "discussion" with chip
216 salzenberg (discussion implies arguments, but his only arguments
217 were ad-hominems, one wonders why he started it in the first
218 place). I hope this explains it well enough for him to understand,
219 and maybe well enough for others to understand.
220 - use common::sense everywhere now.
221 - idle callbacks are no longer supported, use idle coros instead.
222 - print a thread listing when a deadlock is detected.
223
2245.17 Sat Aug 22 23:09:31 CEST 2009
225 - work around a bug in the perl debugger causing crashes
226 when running under the debugger by marking _pool_handler as nodebug.
227 - speed up Coro::async considerably.
228 - try some hacks to get netbsd to work "more often" - their broken
229 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
230 - convert Coro to AE 5.0 API.
231
2325.162 Tue Jul 28 04:04:03 CEST 2009
233 - perl 5.8.2 is now minimum requirement.
234 - skip t/19_handle.t on broken windows perls.
235
2365.161 Wed Jul 22 04:47:38 CEST 2009
237 - Coro::AnyEvent::poll could have a different prototype when EV was
238 used as backend (analyzed by Tatsuhiko Miyagawa).
239 - Coro::AnyEvent errornously initialised the event loop when loaded,
240 not on demand.
241 - try to workaround rare */t/01_unblock.t failures.
242
2435.16 Tue Jul 21 01:44:37 CEST 2009
244 - Coro::AnyEvent failed to hook into the event loop
245 when no threads had been readied between detecting
246 the event loop and actually running it.
247 - considerably speed up Coro::Select by taking avdantage
248 of AnyEvent > 4.8 and some other optimisations.
249 - implement paragraph readline mode in Coro::Handle
250 (based on patches by Zsbán Ambrus).
251 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
252 and analyzed by Yasuhiro MATSUMOTO).
253 - clarified libcoro license and copyright.
254 - someone stole my EXTRA_META!!!
255 - implement Coro::Select::patch_pp_sselect and it's brother,
256 for hardcode select overriding.
257
2585.151 Mon Jul 6 05:41:57 CEST 2009
259 - backport to windows process emulation code again (patch by
260 Yasuhiro MATSUMOTO).
261 - slightly update Coro::MakeMaker.
262
2635.15 Tue Jun 30 10:28:06 CEST 2009
264 - deprecate Coro::Socket, document how to get ipv6 support via
265 AnyEvent::Socket instead.
266 - implement signal->wait ($cb) interface, similar to semaphores.
267 - work around SvOK not supporting getmagic, so we have to getmagic
268 to test for undef :/ (reported by Matthias Waldorf).
269 - load Coro::AnyEvent in all modules using AnyEvent.
270 - work around perl corrupting our internal data structures,
271 reported by Tokuhiro Matsuno.
272 - enable per-coroutine real and cpu time gathering
273 (Coro::State::enable_times).
274
2755.14 Wed Jun 24 01:37:48 CEST 2009
276 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
277 as this seems to have been a difficult concept (poll, sleep, idle,
278 idle_upto).
279 - add Coro::AnyEvent::readable/writable functions.
280 - clarify Coro::EV/Event/AnyEvent manpages.
281 - free per-thread global scalars in the thread calling ->cancel, to
282 avoid crashes when $_, $@ etc., are magical but some of those
283 had already been freed.
284 - "unexperimentalise" the callback interface for Coro::Semaphore.
285 - speed up ready queue management/context switching by using a linked
286 list instead of an array (~5-10%).
287 - implement "watch" command in Coro::Debug shells.
288 - for fun, implement time-slicing as an example in the manpage.
289 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
290 Coro::Event, use the normal AnyEvent handling instead of dieing
291 (the same is true for Coro::Handle).
292 - properly document Coro::EV::timed_io_once.
293 - avoid unneccessary ->cancel calls in Coro::Handle.
294 - maybe make it work on mingw32 with win32 backend
295 (based on patch by Yasuhiro Matsumoto).
296
2975.132 Fri May 29 09:00:39 CEST 2009
298 - do not keep a reference to the argument itself in
299 Coro::Semaphore::guard, as it could change later.
300 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
301 should support a prepare callback.
302
3035.131 Mon Mar 16 23:20:37 CET 2009
304 - implement and document Coro->suspend, Coro->resume.
305 - fix Coro::Select implementation to not (often) close
306 the passed file descriptors (testcase provided by pippijn).
307
3085.13 Mon Dec 15 21:51:42 CET 2008
309 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
310 - don't set diehook to C<undef> but instead to NULL, to avoid
311 spurious warnings.
312 - fix a lot of bugs in Coro::SemaphoreSet.
313 - Coro::SemaphoreSet will less often create a semaphore needlessly.
314 - add Coro::SemaphoreSet::count and wait methods.
315 - take advantage of the new Guard module.
316 - deprecate Coro::guard.
317 - try to fix the dreaded 01_unblock tests once more. I hate it when
318 testsuites need more fixing than the code they are supposed to test.
319 - croak in more cases when a required callback isn't resolvable.
320 - fix some minor issues in Coro::State->call/eval.
321 - use current coroutine context instead of a temporary one
322 when temporarily switching to another coroutine.
323 - do not call C-level on_destroy handlers during global destruction,
324 to avoid needless segfaults.
325
3265.12 Sun Dec 7 13:30:38 CET 2008
327 - add default config for MirOS, which seems to be a bug-to-bug
328 compatible fork of openbsd ("world domination by releasing
329 openbsd cvs before the openbsd folks do it" or so :).
330 - free_padlist did destroy the names pad, not good, but didn't
331 seem to bother perl - this could fix issues such as eval ""
332 inside a function called from multiple coroutines.
333 - use a different method to detect destruction time.
334 - be more careful when freeing padlists just before global
335 destruction.
336 - fixed and expanded the call/cc example.
337 - renamed _terminate to _coro_run.
338 - new method Coro::Channel->shutdown.
339 - try pthreads on openbsd <4.4 (broken sigaltstack, will
340 pthreads fare better?).
341 - be less picky about destroying "a" running coroutine.
342
3435.11 Tue Nov 25 21:49:05 CET 2008
344 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
345 working code with failed assertions, introducing more bugs than
346 it fixes, requiring elaborate workarounds :(
347
3485.1 Mon Nov 24 08:54:59 CET 2008
349 - wrote a small introductory tutorial - Coro::Intro.
350 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
351 - Coro::Select did errornously dup the file descriptors
352 and didn't work with all AnyEvent backends.
353 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
354 LWP data transfers.
355 - disassociate c contexts from coro objects - this is agruably more
356 correct, but mostly allows sharing of cctxs between coro and state
357 objects, for added memory savings and speed increases.
358 - bumped $Coro::POOL_RSS up to 32kb by default.
359 - no longer set the optype to OP_CUSTOM, as B::* understandably
360 doesn't like this very much (and we *are* a type of entersub).
361 - implement state cloning, just to prove that call/cc can be done.
362 - automatically load Coro::AnyEvent in Coro::Handle.
363 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
364 do not have this method (and don't need it either).
365 - speed up generic anyevent methods in Coro::Handle by using rouse
366 callbacks.
367 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
368 others. It also makes the debugger happier, as you can trace
369 through the idle threads now.
370 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
371 - eg/event was pretty broken.
372 - better 5.8.6 compatibility.
373
3745.0 Thu Nov 20 10:35:05 CET 2008
13 - NEW ARCHITECTURE: use the latest 4.x version if you experience 375 - NEW ARCHITECTURE: use the latest 4.x version if you experience
14 stability issues. 376 stability issues.
15 - bump API version to 7 - all dependents must be recompiled. 377 - bump API version to 7 - all dependents must be recompiled.
16 - removed timed_* functions - they were not being used anyways 378 - removed timed_* functions - they were not being used anyways
17 and should be replaced by a more generic mechanism - 379 and should be replaced by a more generic mechanism -
18 and were annoying to support anyways :) 380 and were annoying to support anyways :)
19 - removed SemaphoreSet's waiter method - use sem instead. 381 - removed SemaphoreSet's waiter method - use sem method instead.
20 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 382 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
21 - async_pool did free a scalar value twice 383 - async_pool did free a scalar value twice
22 ("Attempt to unreference..."). 384 ("Attempt to unreference...").
23 - fix a longstanding bug where calling terminate on a coro that 385 - fix a longstanding bug where calling terminate on a coro that
24 was waiting for a semaphore that was just becoming available 386 was waiting for a semaphore that was just becoming available
… …
26 it was available but waiters were still blocked). 388 it was available but waiters were still blocked).
27 - calling throw on a coroutine that is waiting for a semaphore will 389 - calling throw on a coroutine that is waiting for a semaphore will
28 no longer make it acquire the semaphore (and thus leak a count). 390 no longer make it acquire the semaphore (and thus leak a count).
29 - perl's process emulation is now not even theoretically supported 391 - perl's process emulation is now not even theoretically supported
30 anymore. 392 anymore.
393 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
394 conversion of callback-style to blocking-style.
395 - new methods $coro->schedule_to and ->cede_to, to specifically
396 schedule or cede to a specific coroutine.
31 - new function Coro::Semaphore::wait. 397 - new function Coro::Semaphore::wait.
32 - use named constants in Coro::Channel (Richard Hundt). 398 - use named constants in Coro::Channel (Richard Hundt).
33 - directly patch the entersub opcode calling SLF functions (cede, 399 - directly patch the entersub opcode calling SLF functions (cede,
34 transfer and so on). this does speed up context switching, but 400 transfer and so on). this does speed up context switching, but
35 more importanly, it frees us from the hardcoded behaviour of 401 more importanly, it frees us from the hardcoded behaviour of
… …
52 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate 418 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
53 (in comparison) 20-40% speedup. 419 (in comparison) 20-40% speedup.
54 - used new SLF interface to reimplement Coro::Signal gaining 420 - used new SLF interface to reimplement Coro::Signal gaining
55 some unknown (because I was too lazy), but certain, speedup, and also 421 some unknown (because I was too lazy), but certain, speedup, and also
56 making signals reliable for the first time. 422 making signals reliable for the first time.
423 - used new SLF interface and other optimisations to speed up async_pool
424 by a factor of two. It also doesn't rely on perl's exception mechanism
425 to exit anymore. The overhead for terminating an async_pool, coro over
426 a normal async is now very small.
427 - sped up coroutine creation/destruction by 40%.
57 - forgot to include Coro/libcoro/README in the dist for all these years. 428 - forgot to include Coro/libcoro/README in the dist for all these years.
58 - work around a freebsd pthreads bug (manual testcancel is required as 429 - work around a freebsd pthreads bug (manual testcancel is required as
59 pthread_cond_wait isn't a cancellation point on freebsd). 430 pthread_cond_wait isn't a cancellation point on freebsd).
431 - use new rouse functions to speed up and simplify Coro::BDB.
432 - make "prefer perl native functions" work with threaded perls.
433 - condense Coro::Debug ps output, hint at v and w flags.
434 - (libcoro) lots of minor cleanups and portability improvements.
435
4364.914 Wed Nov 19 12:54:18 CET 2008
437 - fix a disastrous bug in the readline optimisation
438 introduced in 4.801.
60 439
614.913 Sat Nov 15 07:58:28 CET 2008 4404.913 Sat Nov 15 07:58:28 CET 2008
62 - async_pool did free a scalar value twice 441 - async_pool did free a scalar value twice
63 ("Attempt to unreference..."). 442 ("Attempt to unreference...").
64 443

Diff Legend

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