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

Comparing Coro/Changes (file contents):
Revision 1.411 by root, Fri Nov 14 07:22:11 2008 UTC vs.
Revision 1.602 by root, Wed Dec 5 00:14:22 2012 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: http://www.microsoft.com/msj/archive/s2ce.aspx
6 but very lacking in others (higher level ipc)). 6TODO: __GCC_HAVE_DWARF2_CFI_ASM
7 7
85.0 8TODO: add peerhost/peerport from IO::Socket::INET etc.
9 - NEW ARCHITECTURE: use 4.912 for the stable version. 9
10 - implement 5.17 compatibility by almost blindly applying a
11 good-looking patch by Father Chrysostomos.
12 - support magic values as timed_io_once args.
13 - recommend AnyEvent 7+ or EV 4+, also require EV
14 version 4 or newer for Coro::EV.
15
166.10 Tue Oct 9 01:14:27 CEST 2012
17 - updated ecb.h, it had a typo that caused it to not compile on many
18 big endian systems (reported by many people).
19 - disable memory fences in ecb.h to improve portability.
20
216.09 Sat Oct 6 23:25:02 CEST 2012
22 - Coro::EV I/O watchers were not interruptible by exceptions
23 (Coro::State::throw) (testcase by sten).
24 - ->throw now puts threads into the ready queue, as this seems to
25 be expected by existing code, and code that doesn't cope with spurious
26 wakeups needs fixing anyway.
27 - use fd -1 in mmap.
28 - cast I32 to int in error message printf.
29 - warn about broken so-called "hardened" kernels.
30
316.08 Fri Apr 13 12:05:47 CEST 2012
32 - be more aggressive about exiting like perl does - formerly,
33 exiting from the non-main thread would not execute END blocks.
34
356.07 Fri Nov 11 21:21:48 CET 2011
36 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
37 - update ecb.h.
38
396.06 Mon Aug 8 23:59:48 CEST 2011
40 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
41 backend, which is an order of magnitude slower. unfortunately, cygwins
42 pthread implementation isn't very complete either, so allocate the stack
43 twice just to be sure.
44 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
45 in its ithreaded perl - best recompile cygwin and use the 'w'indows
46 backend for much better performance. also disable ithreads for
47 even better performance...).
48
496.05 Thu Aug 4 21:36:36 CEST 2011
50 - blush, condvar values would not be propagated from send to recv anymore
51 (reported by Chip Salzenberg).
52 - use exponential increase for the readline buffer length in
53 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
54
556.04 Wed Aug 3 17:15:45 CEST 2011
56 - use even more efficient and more compatible condvars for
57 compatibility to AnyEvent 6.x :)
58 - more inconsequential ecb.h updates.
59
606.03 Wed Aug 3 11:41:30 CEST 2011
61 - change how Coro patches AnyEvent condvars for compatibility to
62 AnyEvent 6.x.
63 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
64 hasn't been included.
65
666.02 Wed Jul 13 04:35:19 CEST 2011
67 - "improve portability to Gentoo" - gentoo manages to put perl variables
68 in memory areas that are farther than 2gb apart, which the jit couldn't
69 handle and barfed. now it's just a bit slower on gentoo and similar
70 systems.
71
726.01 Sun Jul 3 12:31:14 CEST 2011
73 - workarounds are good, but the test for whether pthreads are used
74 was not good. thisone should be better.
75 - check differently whether gcc generates cfi instructions itself.
76
776.0 Wed Jun 29 19:43:35 CEST 2011
78 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
79 destroyed and cleaned up automatically (e.g. async { schedule }).
80 - implement a JIT compiler for part of the thread switch code,
81 which gives a 50% speed improvement on threaded perls, and
82 about 4% on non-thraeded perls (so threaded perls now finally
83 reach about half the speed of non-threaded perls).
84 - slightly modernise Coro::Intro, add section about rouse functions.
85 - avoid DEFSV and ERRSV, giving another 10% improvement
86 in thread switching.
87 - Coro::State->is_destroyed is now called is_zombie.
88 - implement a Coro->safe_cancel method that might fail, but
89 cancels in a "safer" way if it succeeds.
90 - add preliminary support for DEBUGGING perls.
91 - get rid of two hash-accesses when initialising a new Coro - this
92 speeds up coro creation by almost a factor of two.
93 - croak when a coro that is being cancelled tries to block
94 (e.g. while executing a guard block), instead of crashing or
95 deadlocking.
96 - use a more robust and also faster method to identify Coro::State
97 objects - speeds up everything a bit.
98 - implement Coro->cancel in XS for a 20% speed improvement, and to
99 be able to implement mutual cancellation.
100 - speed up context switches by a percent or two by more efficiently
101 allocating context stack entries.
102 - implement Coro->join and Coro->on_destroy in XS for a speedup and
103 a reduction in memory use.
104 - cancelling a coro while it itself is cancelling another coro is
105 now supported and working, instead of triggering an assertion.
106 - be a bit more crash-resistant when calling (buggy) on_destroy
107 callbacks (best effort).
108 - move on_destroy into the slf_frame, to allow extension slf
109 functions to have destructors.
110 - get rid if coro refcounting - simply crash in other interpreter
111 threads by nulling the pointers on clone.
112 - simplify warn/die hook handling when loading Coro - the convoluted
113 logic seems to be no longer neccessary.
114 - use libecb instead of our own home-grown gcc hacks.
115 - document alternatives to Coro::LWP. Please use them :)
116 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
117 sigaltstack. Really. wine suffers from it, erlang suffers from it,
118 and it's known since at least 2006.
119
1205.372 Wed Feb 23 06:14:30 CET 2011
121 - apparently mingw doesn't provide a working gettimeofday, try to
122 work around that by relying on Time::HiRes (indirectly brought to
123 my attention by Max Maischein).
124 - fix some portability issues when Time::HiRes was used.
125
1265.371 Mon Feb 21 14:36:08 CET 2011
127 - backport to windows process emulation code again.
128
1295.37 Sat Feb 19 07:49:44 CET 2011
130 - add a big "Coro thread life cycle" section to "man Coro".
131 - try a tentative workaround against the breakage that 5.13 has
132 introduced without depreciation period. sigh.
133 - no longer use Time::HiRes if gettimeofday is available, which
134 saves quite a lot of memory.
135
1365.36 Sun Feb 13 05:33:41 CET 2011
137 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
138 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
139 method call.
140 - undocument Coro::Timer::sleep and obsolete whole module.
141 - optimise Coro::Timer::timeout memory and cpu usage.
142 - slightly updated Coro::Intro for recent changes.
143 - do not initialise PL_dirty anymore.
144
1455.25 Thu Nov 11 01:08:39 CET 2010
146 - try a different approach on netbsd - netbsd 5 finally has marginally
147 working pthreads, but still broken ucontext/sigaltstack.
148 - openbsd 4.8 finally got their act together, Coro works out of the box
149 with asm, setjmp and pthreads (no change, just informational).
150
1515.24 Sat Oct 23 11:27:12 CEST 2010
152 - port to the EV 4.0 API.
153 - work around bugs in mingw32, making strawberry perl work
154 out of the box.
155 - correctly modify Coro::AIO function prototypes
156 so that they reflect the "no optional parameters" rule.
157 - "ported" libcoro to C++.
158
1595.23 Mon May 17 18:50:42 CEST 2010
160 - be more resistant to ordering changes when initialising
161 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
162 Waldorf).
163 - document that perl 5.12 deliberately removed support for cloning.
164
1655.22 Wed Apr 14 03:55:35 CEST 2010
166 - correctly return udnef on errors in Coro::Handle::read/write
167 (testcase by Marc Mims).
168 - convert Coro::Util into a "perl compatibility wrapper" - the functions
169 are less useful now, but are drop-in replacements for existing
170 functions, listing better alternatives in the documentation. This also
171 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
172 with Coro::Util::inet_aton.
173 - do not override $Coro::idle unconditionally in Coro.pm, as other
174 modules could have provided their own idle coro already
175 (for exmaple, Coro::AnyEvent).
176 - fix Coro::Util::gethost* functions.
177 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
178
1795.21 Wed Dec 16 07:19:51 CET 2009
180 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
181 together.
182 - add some examples on how to combine other event loops with Coro in
183 Coro::AnyEvent, and how to run it (and not to block). Seems to be
184 the most common source of confusion.
185 - try to catch people naively blocking in an event callback.
186 - work around the perl filehandle bug issue in conjunction with
187 older common::sense (as indirectly pointed out by ZSystem).
188 - clarify the "not from signal handlers" section.
189
1905.2 Sun Oct 4 14:54:24 CEST 2009
191 - Coro::Storable destroyed the prototypes of the functions it wrapped.
192 - export rouse_cb and rouse_wait by default now.
193 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
194 - new method $state->swap_sv.
195 - added section on "windows process emulation" to the manpage,
196 after a not-so-fruitful (nor-friendly) "discussion" with chip
197 salzenberg (discussion implies arguments, but his only arguments
198 were ad-hominems, one wonders why he started it in the first
199 place). I hope this explains it well enough for him to understand,
200 and maybe well enough for others to understand.
201 - use common::sense everywhere now.
202 - idle callbacks are no longer supported, use idle coros instead.
203 - print a thread listing when a deadlock is detected.
204
2055.17 Sat Aug 22 23:09:31 CEST 2009
206 - work around a bug in the perl debugger causing crashes
207 when running under the debugger by marking _pool_handler as nodebug.
208 - speed up Coro::async considerably.
209 - try some hacks to get netbsd to work "more often" - their broken
210 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
211 - convert Coro to AE 5.0 API.
212
2135.162 Tue Jul 28 04:04:03 CEST 2009
214 - perl 5.8.2 is now minimum requirement.
215 - skip t/19_handle.t on broken windows perls.
216
2175.161 Wed Jul 22 04:47:38 CEST 2009
218 - Coro::AnyEvent::poll could have a different prototype when EV was
219 used as backend (analyzed by Tatsuhiko Miyagawa).
220 - Coro::AnyEvent errornously initialised the event loop when loaded,
221 not on demand.
222 - try to workaround rare */t/01_unblock.t failures.
223
2245.16 Tue Jul 21 01:44:37 CEST 2009
225 - Coro::AnyEvent failed to hook into the event loop
226 when no threads had been readied between detecting
227 the event loop and actually running it.
228 - considerably speed up Coro::Select by taking avdantage
229 of AnyEvent > 4.8 and some other optimisations.
230 - implement paragraph readline mode in Coro::Handle
231 (based on patches by Zsbán Ambrus).
232 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
233 and analyzed by Yasuhiro MATSUMOTO).
234 - clarified libcoro license and copyright.
235 - someone stole my EXTRA_META!!!
236 - implement Coro::Select::patch_pp_sselect and it's brother,
237 for hardcode select overriding.
238
2395.151 Mon Jul 6 05:41:57 CEST 2009
240 - backport to windows process emulation code again (patch by
241 Yasuhiro MATSUMOTO).
242 - slightly update Coro::MakeMaker.
243
2445.15 Tue Jun 30 10:28:06 CEST 2009
245 - deprecate Coro::Socket, document how to get ipv6 support via
246 AnyEvent::Socket instead.
247 - implement signal->wait ($cb) interface, similar to semaphores.
248 - work around SvOK not supporting getmagic, so we have to getmagic
249 to test for undef :/ (reported by Matthias Waldorf).
250 - load Coro::AnyEvent in all modules using AnyEvent.
251 - work around perl corrupting our internal data structures,
252 reported by Tokuhiro Matsuno.
253 - enable per-coroutine real and cpu time gathering
254 (Coro::State::enable_times).
255
2565.14 Wed Jun 24 01:37:48 CEST 2009
257 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
258 as this seems to have been a difficult concept (poll, sleep, idle,
259 idle_upto).
260 - add Coro::AnyEvent::readable/writable functions.
261 - clarify Coro::EV/Event/AnyEvent manpages.
262 - free per-thread global scalars in the thread calling ->cancel, to
263 avoid crashes when $_, $@ etc., are magical but some of those
264 had already been freed.
265 - "unexperimentalise" the callback interface for Coro::Semaphore.
266 - speed up ready queue management/context switching by using a linked
267 list instead of an array (~5-10%).
268 - implement "watch" command in Coro::Debug shells.
269 - for fun, implement time-slicing as an example in the manpage.
270 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
271 Coro::Event, use the normal AnyEvent handling instead of dieing
272 (the same is true for Coro::Handle).
273 - properly document Coro::EV::timed_io_once.
274 - avoid unneccessary ->cancel calls in Coro::Handle.
275 - maybe make it work on mingw32 with win32 backend
276 (based on patch by Yasuhiro Matsumoto).
277
2785.132 Fri May 29 09:00:39 CEST 2009
279 - do not keep a reference to the argument itself in
280 Coro::Semaphore::guard, as it could change later.
281 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
282 should support a prepare callback.
283
2845.131 Mon Mar 16 23:20:37 CET 2009
285 - implement and document Coro->suspend, Coro->resume.
286 - fix Coro::Select implementation to not (often) close
287 the passed file descriptors (testcase provided by pippijn).
288
2895.13 Mon Dec 15 21:51:42 CET 2008
290 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
291 - don't set diehook to C<undef> but instead to NULL, to avoid
292 spurious warnings.
293 - fix a lot of bugs in Coro::SemaphoreSet.
294 - Coro::SemaphoreSet will less often create a semaphore needlessly.
295 - add Coro::SemaphoreSet::count and wait methods.
296 - take advantage of the new Guard module.
297 - deprecate Coro::guard.
298 - try to fix the dreaded 01_unblock tests once more. I hate it when
299 testsuites need more fixing than the code they are supposed to test.
300 - croak in more cases when a required callback isn't resolvable.
301 - fix some minor issues in Coro::State->call/eval.
302 - use current coroutine context instead of a temporary one
303 when temporarily switching to another coroutine.
304 - do not call C-level on_destroy handlers during global destruction,
305 to avoid needless segfaults.
306
3075.12 Sun Dec 7 13:30:38 CET 2008
308 - add default config for MirOS, which seems to be a bug-to-bug
309 compatible fork of openbsd ("world domination by releasing
310 openbsd cvs before the openbsd folks do it" or so :).
311 - free_padlist did destroy the names pad, not good, but didn't
312 seem to bother perl - this could fix issues such as eval ""
313 inside a function called from multiple coroutines.
314 - use a different method to detect destruction time.
315 - be more careful when freeing padlists just before global
316 destruction.
317 - fixed and expanded the call/cc example.
318 - renamed _terminate to _coro_run.
319 - new method Coro::Channel->shutdown.
320 - try pthreads on openbsd <4.4 (broken sigaltstack, will
321 pthreads fare better?).
322 - be less picky about destroying "a" running coroutine.
323
3245.11 Tue Nov 25 21:49:05 CET 2008
325 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
326 working code with failed assertions, introducing more bugs than
327 it fixes, requiring elaborate workarounds :(
328
3295.1 Mon Nov 24 08:54:59 CET 2008
330 - wrote a small introductory tutorial - Coro::Intro.
331 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
332 - Coro::Select did errornously dup the file descriptors
333 and didn't work with all AnyEvent backends.
334 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
335 LWP data transfers.
336 - disassociate c contexts from coro objects - this is agruably more
337 correct, but mostly allows sharing of cctxs between coro and state
338 objects, for added memory savings and speed increases.
339 - bumped $Coro::POOL_RSS up to 32kb by default.
340 - no longer set the optype to OP_CUSTOM, as B::* understandably
341 doesn't like this very much (and we *are* a type of entersub).
342 - implement state cloning, just to prove that call/cc can be done.
343 - automatically load Coro::AnyEvent in Coro::Handle.
344 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
345 do not have this method (and don't need it either).
346 - speed up generic anyevent methods in Coro::Handle by using rouse
347 callbacks.
348 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
349 others. It also makes the debugger happier, as you can trace
350 through the idle threads now.
351 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
352 - eg/event was pretty broken.
353 - better 5.8.6 compatibility.
354
3555.0 Thu Nov 20 10:35:05 CET 2008
356 - NEW ARCHITECTURE: use the latest 4.x version if you experience
357 stability issues.
10 - bump API version to 7 - all dependents must be recompiled. 358 - bump API version to 7 - all dependents must be recompiled.
359 - removed timed_* functions - they were not being used anyways
360 and should be replaced by a more generic mechanism -
361 and were annoying to support anyways :)
362 - removed SemaphoreSet's waiter method - use sem method instead.
363 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
364 - async_pool did free a scalar value twice
365 ("Attempt to unreference...").
366 - fix a longstanding bug where calling terminate on a coro that
367 was waiting for a semaphore that was just becoming available
368 would cause a deadlock (semaphore would get into a state where
369 it was available but waiters were still blocked).
370 - calling throw on a coroutine that is waiting for a semaphore will
371 no longer make it acquire the semaphore (and thus leak a count).
372 - perl's process emulation is now not even theoretically supported
373 anymore.
374 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
375 conversion of callback-style to blocking-style.
376 - new methods $coro->schedule_to and ->cede_to, to specifically
377 schedule or cede to a specific coroutine.
378 - new function Coro::Semaphore::wait.
379 - use named constants in Coro::Channel (Richard Hundt).
11 - directly patch the entersub opcode calling ssl functions (cede, 380 - directly patch the entersub opcode calling SLF functions (cede,
12 transfer and so on). this does speed up context switching by 381 transfer and so on). this does speed up context switching, but
13 up to 10%. More importanly, it frees us from the hardcoded 382 more importanly, it frees us from the hardcoded behaviour of
14 behaviour of entersub, so we might actually be able to return 383 entersub, so we might actually be able to return something from
15 something from those functions. 384 those functions and atcually create new ones.
16 - take advanatge of __builtin_frame_address on gcc. 385 - take advantage of __builtin_frame_address on gcc.
17 - expose THX in coroapi (not sure whether this was a wise decision). 386 - expose THX in coroapi (not sure whether this was a wise decision,
387 as "threaded" perls are running at half speed anyways).
388 - implement execute_slf (schedule-like-function) interface that makes
389 it possible to implement schedule-like-functions in XS.
390 - use new SLF interface to massively speed up Coro::EV by roughly a
391 factor of two.
392 - used new SLF interface to massively speed up Coro::Semaphore by a
393 factor of three.
394 - used new SLF interface to speed up Coro::AIO by roughly a factor of
395 four and reduce its memory usage considerably.
396 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
397 for a nice speedup and vastly more correct behaviour. Also implement
398 a new method "sem" to get at the underlying semaphore object.
399 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
400 (in comparison) 20-40% speedup.
401 - used new SLF interface to reimplement Coro::Signal gaining
402 some unknown (because I was too lazy), but certain, speedup, and also
403 making signals reliable for the first time.
404 - used new SLF interface and other optimisations to speed up async_pool
405 by a factor of two. It also doesn't rely on perl's exception mechanism
406 to exit anymore. The overhead for terminating an async_pool, coro over
407 a normal async is now very small.
408 - sped up coroutine creation/destruction by 40%.
409 - forgot to include Coro/libcoro/README in the dist for all these years.
410 - work around a freebsd pthreads bug (manual testcancel is required as
411 pthread_cond_wait isn't a cancellation point on freebsd).
412 - use new rouse functions to speed up and simplify Coro::BDB.
413 - make "prefer perl native functions" work with threaded perls.
414 - condense Coro::Debug ps output, hint at v and w flags.
415 - (libcoro) lots of minor cleanups and portability improvements.
416
4174.914 Wed Nov 19 12:54:18 CET 2008
418 - fix a disastrous bug in the readline optimisation
419 introduced in 4.801.
420
4214.913 Sat Nov 15 07:58:28 CET 2008
422 - async_pool did free a scalar value twice
423 ("Attempt to unreference...").
18 424
194.912 Thu Nov 13 18:31:23 CET 2008 4254.912 Thu Nov 13 18:31:23 CET 2008
20 - minor cleanups. 426 - minor cleanups.
21 - use much larger stacks on linux and perl < 5.8.8. 427 - use much larger stacks on linux and perl < 5.8.8.
22 - Coro::Debug::new_unix_server did not unlink the socket 428 - Coro::Debug::new_unix_server did not unlink the socket

Diff Legend

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