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

Comparing Coro/Changes (file contents):
Revision 1.414 by root, Sat Nov 15 00:37:38 2008 UTC vs.
Revision 1.582 by root, Wed Aug 3 15:13:19 2011 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)).
7 6
85.0 76.04
9 - NEW ARCHITECTURE: use 4.912 for the stable version. 8 - use even more effciient and more compatible condvars for
9 compatibility to AnyEvent 6.x :)
10 - more inconsequential ecb.h updates.
11
126.03 Wed Aug 3 11:41:30 CEST 2011
13 - change how Coro patches AnyEvent condvars for compatibility to
14 AnyEvent 6.x.
15 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
16 hasn't been included.
17
186.02 Wed Jul 13 04:35:19 CEST 2011
19 - "improve portability to Gentoo" - gentoo manages to put perl variables
20 in memory areas that are farther than 2gb apart, which the jit couldn't
21 handle and barfed. now it's just a bit slower on gentoo and similar
22 systems.
23
246.01 Sun Jul 3 12:31:14 CEST 2011
25 - workarounds are good, but the test for whether pthreads are used
26 was not good. thisone should be better.
27 - check differently whether gcc generates cfi instructions itself.
28
296.0 Wed Jun 29 19:43:35 CEST 2011
30 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
31 destroyed and cleaned up automatically (e.g. async { schedule }).
32 - implement a JIT compiler for part of the thread switch code,
33 which gives a 50% speed improvement on threaded perls, and
34 about 4% on non-thraeded perls (so threaded perls now finally
35 reach about half the speed of non-threaded perls).
36 - slightly modernise Coro::Intro, add section about rouse functions.
37 - avoid DEFSV and ERRSV, giving another 10% improvement
38 in thread switching.
39 - Coro::State->is_destroyed is now called is_zombie.
40 - implement a Coro->safe_cancel method that might fail, but
41 cancels in a "safer" way if it succeeds.
42 - add preliminary support for DEBUGGING perls.
43 - get rid of two hash-accesses when initialising a new Coro - this
44 speeds up coro creation by almost a factor of two.
45 - croak when a coro that is being cancelled tries to block
46 (e.g. while executing a guard block), instead of crashing or
47 deadlocking.
48 - use a more robust and also faster method to identify Coro::State
49 objects - speeds up everything a bit.
50 - implement Coro->cancel in XS for a 20% speed improvement, and to
51 be able to implement mutual cancellation.
52 - speed up context switches by a percent or two by more efficiently
53 allocating context stack entries.
54 - implement Coro->join and Coro->on_destroy in XS for a speedup and
55 a reduction in memory use.
56 - cancelling a coro while it itself is cancelling another coro is
57 now supported and working, instead of triggering an assertion.
58 - be a bit more crash-resistant when calling (buggy) on_destroy
59 callbacks (best effort).
60 - move on_destroy into the slf_frame, to allow extension slf
61 functions to have destructors.
62 - get rid if coro refcounting - simply crash in other interpreter
63 threads by nulling the pointers on clone.
64 - simplify warn/die hook handling when loading Coro - the convoluted
65 logic seems to be no longer neccessary.
66 - use libecb instead of our own home-grown gcc hacks.
67 - document alternatives to Coro::LWP. Please use them :)
68 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
69 sigaltstack. Really. wine suffers from it, erlang suffers from it,
70 and it's known since at least 2006.
71
725.372 Wed Feb 23 06:14:30 CET 2011
73 - apparently mingw doesn't provide a working gettimeofday, try to
74 work around that by relying on Time::HiRes (indirectly brought to
75 my attention by Max Maischein).
76 - fix some portability issues when Time::HiRes was used.
77
785.371 Mon Feb 21 14:36:08 CET 2011
79 - backport to windows process emulation code again.
80
815.37 Sat Feb 19 07:49:44 CET 2011
82 - add a big "Coro thread life cycle" section to "man Coro".
83 - try a tentative workaround against the breakage that 5.13 has
84 introduced without depreciation period. sigh.
85 - no longer use Time::HiRes if gettimeofday is available, which
86 saves quite a lot of memory.
87
885.36 Sun Feb 13 05:33:41 CET 2011
89 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
90 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
91 method call.
92 - undocument Coro::Timer::sleep and obsolete whole module.
93 - optimise Coro::Timer::timeout memory and cpu usage.
94 - slightly updated Coro::Intro for recent changes.
95 - do not initialise PL_dirty anymore.
96
975.25 Thu Nov 11 01:08:39 CET 2010
98 - try a different approach on netbsd - netbsd 5 finally has marginally
99 working pthreads, but still broken ucontext/sigaltstack.
100 - openbsd 4.8 finally got their act together, Coro works out of the box
101 with asm, setjmp and pthreads (no change, just informational).
102
1035.24 Sat Oct 23 11:27:12 CEST 2010
104 - port to the EV 4.0 API.
105 - work around bugs in mingw32, making strawberry perl work
106 out of the box.
107 - correctly modify Coro::AIO function prototypes
108 so that they reflect the "no optional parameters" rule.
109 - "ported" libcoro to C++.
110
1115.23 Mon May 17 18:50:42 CEST 2010
112 - be more resistant to ordering changes when initialising
113 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
114 Waldorf).
115 - document that perl 5.12 deliberately removed support for cloning.
116
1175.22 Wed Apr 14 03:55:35 CEST 2010
118 - correctly return udnef on errors in Coro::Handle::read/write
119 (testcase by Marc Mims).
120 - convert Coro::Util into a "perl compatibility wrapper" - the functions
121 are less useful now, but are drop-in replacements for existing
122 functions, listing better alternatives in the documentation. This also
123 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
124 with Coro::Util::inet_aton.
125 - do not override $Coro::idle unconditionally in Coro.pm, as other
126 modules could have provided their own idle coro already
127 (for exmaple, Coro::AnyEvent).
128 - fix Coro::Util::gethost* functions.
129 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
130
1315.21 Wed Dec 16 07:19:51 CET 2009
132 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
133 together.
134 - add some examples on how to combine other event loops with Coro in
135 Coro::AnyEvent, and how to run it (and not to block). Seems to be
136 the most common source of confusion.
137 - try to catch people naively blocking in an event callback.
138 - work around the perl filehandle bug issue in conjunction with
139 older common::sense (as indirectly pointed out by ZSystem).
140 - clarify the "not from signal handlers" section.
141
1425.2 Sun Oct 4 14:54:24 CEST 2009
143 - Coro::Storable destroyed the prototypes of the functions it wrapped.
144 - export rouse_cb and rouse_wait by default now.
145 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
146 - new method $state->swap_sv.
147 - added section on "windows process emulation" to the manpage,
148 after a not-so-fruitful (nor-friendly) "discussion" with chip
149 salzenberg (discussion implies arguments, but his only arguments
150 were ad-hominems, one wonders why he started it in the first
151 place). I hope this explains it well enough for him to understand,
152 and maybe well enough for others to understand.
153 - use common::sense everywhere now.
154 - idle callbacks are no longer supported, use idle coros instead.
155 - print a thread listing when a deadlock is detected.
156
1575.17 Sat Aug 22 23:09:31 CEST 2009
158 - work around a bug in the perl debugger causing crashes
159 when running under the debugger by marking _pool_handler as nodebug.
160 - speed up Coro::async considerably.
161 - try some hacks to get netbsd to work "more often" - their broken
162 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
163 - convert Coro to AE 5.0 API.
164
1655.162 Tue Jul 28 04:04:03 CEST 2009
166 - perl 5.8.2 is now minimum requirement.
167 - skip t/19_handle.t on broken windows perls.
168
1695.161 Wed Jul 22 04:47:38 CEST 2009
170 - Coro::AnyEvent::poll could have a different prototype when EV was
171 used as backend (analyzed by Tatsuhiko Miyagawa).
172 - Coro::AnyEvent errornously initialised the event loop when loaded,
173 not on demand.
174 - try to workaround rare */t/01_unblock.t failures.
175
1765.16 Tue Jul 21 01:44:37 CEST 2009
177 - Coro::AnyEvent failed to hook into the event loop
178 when no threads had been readied between detecting
179 the event loop and actually running it.
180 - considerably speed up Coro::Select by taking avdantage
181 of AnyEvent > 4.8 and some other optimisations.
182 - implement paragraph readline mode in Coro::Handle
183 (based on patches by Zsbán Ambrus).
184 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
185 and analyzed by Yasuhiro MATSUMOTO).
186 - clarified libcoro license and copyright.
187 - someone stole my EXTRA_META!!!
188 - implement Coro::Select::patch_pp_sselect and it's brother,
189 for hardcode select overriding.
190
1915.151 Mon Jul 6 05:41:57 CEST 2009
192 - backport to windows process emulation code again (patch by
193 Yasuhiro MATSUMOTO).
194 - slightly update Coro::MakeMaker.
195
1965.15 Tue Jun 30 10:28:06 CEST 2009
197 - deprecate Coro::Socket, document how to get ipv6 support via
198 AnyEvent::Socket instead.
199 - implement signal->wait ($cb) interface, similar to semaphores.
200 - work around SvOK not supporting getmagic, so we have to getmagic
201 to test for undef :/ (reported by Matthias Waldorf).
202 - load Coro::AnyEvent in all modules using AnyEvent.
203 - work around perl corrupting our internal data structures,
204 reported by Tokuhiro Matsuno.
205 - enable per-coroutine real and cpu time gathering
206 (Coro::State::enable_times).
207
2085.14 Wed Jun 24 01:37:48 CEST 2009
209 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
210 as this seems to have been a difficult concept (poll, sleep, idle,
211 idle_upto).
212 - add Coro::AnyEvent::readable/writable functions.
213 - clarify Coro::EV/Event/AnyEvent manpages.
214 - free per-thread global scalars in the thread calling ->cancel, to
215 avoid crashes when $_, $@ etc., are magical but some of those
216 had already been freed.
217 - "unexperimentalise" the callback interface for Coro::Semaphore.
218 - speed up ready queue management/context switching by using a linked
219 list instead of an array (~5-10%).
220 - implement "watch" command in Coro::Debug shells.
221 - for fun, implement time-slicing as an example in the manpage.
222 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
223 Coro::Event, use the normal AnyEvent handling instead of dieing
224 (the same is true for Coro::Handle).
225 - properly document Coro::EV::timed_io_once.
226 - avoid unneccessary ->cancel calls in Coro::Handle.
227 - maybe make it work on mingw32 with win32 backend
228 (based on patch by Yasuhiro Matsumoto).
229
2305.132 Fri May 29 09:00:39 CEST 2009
231 - do not keep a reference to the argument itself in
232 Coro::Semaphore::guard, as it could change later.
233 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
234 should support a prepare callback.
235
2365.131 Mon Mar 16 23:20:37 CET 2009
237 - implement and document Coro->suspend, Coro->resume.
238 - fix Coro::Select implementation to not (often) close
239 the passed file descriptors (testcase provided by pippijn).
240
2415.13 Mon Dec 15 21:51:42 CET 2008
242 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
243 - don't set diehook to C<undef> but instead to NULL, to avoid
244 spurious warnings.
245 - fix a lot of bugs in Coro::SemaphoreSet.
246 - Coro::SemaphoreSet will less often create a semaphore needlessly.
247 - add Coro::SemaphoreSet::count and wait methods.
248 - take advantage of the new Guard module.
249 - deprecate Coro::guard.
250 - try to fix the dreaded 01_unblock tests once more. I hate it when
251 testsuites need more fixing than the code they are supposed to test.
252 - croak in more cases when a required callback isn't resolvable.
253 - fix some minor issues in Coro::State->call/eval.
254 - use current coroutine context instead of a temporary one
255 when temporarily switching to another coroutine.
256 - do not call C-level on_destroy handlers during global destruction,
257 to avoid needless segfaults.
258
2595.12 Sun Dec 7 13:30:38 CET 2008
260 - add default config for MirOS, which seems to be a bug-to-bug
261 compatible fork of openbsd ("world domination by releasing
262 openbsd cvs before the openbsd folks do it" or so :).
263 - free_padlist did destroy the names pad, not good, but didn't
264 seem to bother perl - this could fix issues such as eval ""
265 inside a function called from multiple coroutines.
266 - use a different method to detect destruction time.
267 - be more careful when freeing padlists just before global
268 destruction.
269 - fixed and expanded the call/cc example.
270 - renamed _terminate to _coro_run.
271 - new method Coro::Channel->shutdown.
272 - try pthreads on openbsd <4.4 (broken sigaltstack, will
273 pthreads fare better?).
274 - be less picky about destroying "a" running coroutine.
275
2765.11 Tue Nov 25 21:49:05 CET 2008
277 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
278 working code with failed assertions, introducing more bugs than
279 it fixes, requiring elaborate workarounds :(
280
2815.1 Mon Nov 24 08:54:59 CET 2008
282 - wrote a small introductory tutorial - Coro::Intro.
283 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
284 - Coro::Select did errornously dup the file descriptors
285 and didn't work with all AnyEvent backends.
286 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
287 LWP data transfers.
288 - disassociate c contexts from coro objects - this is agruably more
289 correct, but mostly allows sharing of cctxs between coro and state
290 objects, for added memory savings and speed increases.
291 - bumped $Coro::POOL_RSS up to 32kb by default.
292 - no longer set the optype to OP_CUSTOM, as B::* understandably
293 doesn't like this very much (and we *are* a type of entersub).
294 - implement state cloning, just to prove that call/cc can be done.
295 - automatically load Coro::AnyEvent in Coro::Handle.
296 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
297 do not have this method (and don't need it either).
298 - speed up generic anyevent methods in Coro::Handle by using rouse
299 callbacks.
300 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
301 others. It also makes the debugger happier, as you can trace
302 through the idle threads now.
303 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
304 - eg/event was pretty broken.
305 - better 5.8.6 compatibility.
306
3075.0 Thu Nov 20 10:35:05 CET 2008
308 - NEW ARCHITECTURE: use the latest 4.x version if you experience
309 stability issues.
10 - bump API version to 7 - all dependents must be recompiled. 310 - bump API version to 7 - all dependents must be recompiled.
311 - removed timed_* functions - they were not being used anyways
312 and should be replaced by a more generic mechanism -
313 and were annoying to support anyways :)
314 - removed SemaphoreSet's waiter method - use sem method instead.
315 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
316 - async_pool did free a scalar value twice
317 ("Attempt to unreference...").
318 - fix a longstanding bug where calling terminate on a coro that
319 was waiting for a semaphore that was just becoming available
320 would cause a deadlock (semaphore would get into a state where
321 it was available but waiters were still blocked).
322 - calling throw on a coroutine that is waiting for a semaphore will
323 no longer make it acquire the semaphore (and thus leak a count).
11 - perl's process emulation is now not even theoretically supported 324 - perl's process emulation is now not even theoretically supported
12 anymore. 325 anymore.
326 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
327 conversion of callback-style to blocking-style.
328 - new methods $coro->schedule_to and ->cede_to, to specifically
329 schedule or cede to a specific coroutine.
330 - new function Coro::Semaphore::wait.
331 - use named constants in Coro::Channel (Richard Hundt).
13 - directly patch the entersub opcode calling SLF functions (cede, 332 - directly patch the entersub opcode calling SLF functions (cede,
14 transfer and so on). this does speed up context switching, but 333 transfer and so on). this does speed up context switching, but
15 more importanly, it frees us from the hardcoded behaviour of 334 more importanly, it frees us from the hardcoded behaviour of
16 entersub, so we might actually be able to return something from 335 entersub, so we might actually be able to return something from
17 those functions and atcually create new ones. 336 those functions and atcually create new ones.
18 - take advantage of __builtin_frame_address on gcc. 337 - take advantage of __builtin_frame_address on gcc.
19 - expose THX in coroapi (not sure whether this was a wise decision, 338 - expose THX in coroapi (not sure whether this was a wise decision,
20 as "threaded" perls are running at half speed anyways). 339 as "threaded" perls are running at half speed anyways).
21 - implement execute_slf (schedule-like-function) interface that makes 340 - implement execute_slf (schedule-like-function) interface that makes
22 it possible to implement schedule-like-functions in XS. 341 it possible to implement schedule-like-functions in XS.
23 - use new SLF interface to massively speed up Coro::EV (it's roughly 342 - use new SLF interface to massively speed up Coro::EV by roughly a
24 twice as fast now). 343 factor of two.
344 - used new SLF interface to massively speed up Coro::Semaphore by a
345 factor of three.
346 - used new SLF interface to speed up Coro::AIO by roughly a factor of
347 four and reduce its memory usage considerably.
348 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
349 for a nice speedup and vastly more correct behaviour. Also implement
350 a new method "sem" to get at the underlying semaphore object.
351 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
352 (in comparison) 20-40% speedup.
353 - used new SLF interface to reimplement Coro::Signal gaining
354 some unknown (because I was too lazy), but certain, speedup, and also
355 making signals reliable for the first time.
356 - used new SLF interface and other optimisations to speed up async_pool
357 by a factor of two. It also doesn't rely on perl's exception mechanism
358 to exit anymore. The overhead for terminating an async_pool, coro over
359 a normal async is now very small.
360 - sped up coroutine creation/destruction by 40%.
361 - forgot to include Coro/libcoro/README in the dist for all these years.
362 - work around a freebsd pthreads bug (manual testcancel is required as
363 pthread_cond_wait isn't a cancellation point on freebsd).
364 - use new rouse functions to speed up and simplify Coro::BDB.
365 - make "prefer perl native functions" work with threaded perls.
366 - condense Coro::Debug ps output, hint at v and w flags.
367 - (libcoro) lots of minor cleanups and portability improvements.
368
3694.914 Wed Nov 19 12:54:18 CET 2008
370 - fix a disastrous bug in the readline optimisation
371 introduced in 4.801.
372
3734.913 Sat Nov 15 07:58:28 CET 2008
374 - async_pool did free a scalar value twice
375 ("Attempt to unreference...").
25 376
264.912 Thu Nov 13 18:31:23 CET 2008 3774.912 Thu Nov 13 18:31:23 CET 2008
27 - minor cleanups. 378 - minor cleanups.
28 - use much larger stacks on linux and perl < 5.8.8. 379 - use much larger stacks on linux and perl < 5.8.8.
29 - Coro::Debug::new_unix_server did not unlink the socket 380 - Coro::Debug::new_unix_server did not unlink the socket

Diff Legend

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