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

Comparing Coro/Changes (file contents):
Revision 1.399 by root, Mon Nov 10 00:02:29 2008 UTC vs.
Revision 1.581 by root, Wed Aug 3 14:52:18 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
84.91 76.04
8 - use even more effciient and more compatible condvars for
9 compatibility to AnyEvent 6.x :)
10
116.03 Wed Aug 3 11:41:30 CEST 2011
12 - change how Coro patches AnyEvent condvars for compatibility to
13 AnyEvent 6.x.
14 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
15 hasn't been included.
16
176.02 Wed Jul 13 04:35:19 CEST 2011
18 - "improve portability to Gentoo" - gentoo manages to put perl variables
19 in memory areas that are farther than 2gb apart, which the jit couldn't
20 handle and barfed. now it's just a bit slower on gentoo and similar
21 systems.
22
236.01 Sun Jul 3 12:31:14 CEST 2011
24 - workarounds are good, but the test for whether pthreads are used
25 was not good. thisone should be better.
26 - check differently whether gcc generates cfi instructions itself.
27
286.0 Wed Jun 29 19:43:35 CEST 2011
29 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
30 destroyed and cleaned up automatically (e.g. async { schedule }).
31 - implement a JIT compiler for part of the thread switch code,
32 which gives a 50% speed improvement on threaded perls, and
33 about 4% on non-thraeded perls (so threaded perls now finally
34 reach about half the speed of non-threaded perls).
35 - slightly modernise Coro::Intro, add section about rouse functions.
36 - avoid DEFSV and ERRSV, giving another 10% improvement
37 in thread switching.
38 - Coro::State->is_destroyed is now called is_zombie.
39 - implement a Coro->safe_cancel method that might fail, but
40 cancels in a "safer" way if it succeeds.
41 - add preliminary support for DEBUGGING perls.
42 - get rid of two hash-accesses when initialising a new Coro - this
43 speeds up coro creation by almost a factor of two.
44 - croak when a coro that is being cancelled tries to block
45 (e.g. while executing a guard block), instead of crashing or
46 deadlocking.
47 - use a more robust and also faster method to identify Coro::State
48 objects - speeds up everything a bit.
49 - implement Coro->cancel in XS for a 20% speed improvement, and to
50 be able to implement mutual cancellation.
51 - speed up context switches by a percent or two by more efficiently
52 allocating context stack entries.
53 - implement Coro->join and Coro->on_destroy in XS for a speedup and
54 a reduction in memory use.
55 - cancelling a coro while it itself is cancelling another coro is
56 now supported and working, instead of triggering an assertion.
57 - be a bit more crash-resistant when calling (buggy) on_destroy
58 callbacks (best effort).
59 - move on_destroy into the slf_frame, to allow extension slf
60 functions to have destructors.
61 - get rid if coro refcounting - simply crash in other interpreter
62 threads by nulling the pointers on clone.
63 - simplify warn/die hook handling when loading Coro - the convoluted
64 logic seems to be no longer neccessary.
65 - use libecb instead of our own home-grown gcc hacks.
66 - document alternatives to Coro::LWP. Please use them :)
67 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
68 sigaltstack. Really. wine suffers from it, erlang suffers from it,
69 and it's known since at least 2006.
70
715.372 Wed Feb 23 06:14:30 CET 2011
72 - apparently mingw doesn't provide a working gettimeofday, try to
73 work around that by relying on Time::HiRes (indirectly brought to
74 my attention by Max Maischein).
75 - fix some portability issues when Time::HiRes was used.
76
775.371 Mon Feb 21 14:36:08 CET 2011
78 - backport to windows process emulation code again.
79
805.37 Sat Feb 19 07:49:44 CET 2011
81 - add a big "Coro thread life cycle" section to "man Coro".
82 - try a tentative workaround against the breakage that 5.13 has
83 introduced without depreciation period. sigh.
84 - no longer use Time::HiRes if gettimeofday is available, which
85 saves quite a lot of memory.
86
875.36 Sun Feb 13 05:33:41 CET 2011
88 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
89 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
90 method call.
91 - undocument Coro::Timer::sleep and obsolete whole module.
92 - optimise Coro::Timer::timeout memory and cpu usage.
93 - slightly updated Coro::Intro for recent changes.
94 - do not initialise PL_dirty anymore.
95
965.25 Thu Nov 11 01:08:39 CET 2010
97 - try a different approach on netbsd - netbsd 5 finally has marginally
98 working pthreads, but still broken ucontext/sigaltstack.
99 - openbsd 4.8 finally got their act together, Coro works out of the box
100 with asm, setjmp and pthreads (no change, just informational).
101
1025.24 Sat Oct 23 11:27:12 CEST 2010
103 - port to the EV 4.0 API.
104 - work around bugs in mingw32, making strawberry perl work
105 out of the box.
106 - correctly modify Coro::AIO function prototypes
107 so that they reflect the "no optional parameters" rule.
108 - "ported" libcoro to C++.
109
1105.23 Mon May 17 18:50:42 CEST 2010
111 - be more resistant to ordering changes when initialising
112 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
113 Waldorf).
114 - document that perl 5.12 deliberately removed support for cloning.
115
1165.22 Wed Apr 14 03:55:35 CEST 2010
117 - correctly return udnef on errors in Coro::Handle::read/write
118 (testcase by Marc Mims).
119 - convert Coro::Util into a "perl compatibility wrapper" - the functions
120 are less useful now, but are drop-in replacements for existing
121 functions, listing better alternatives in the documentation. This also
122 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
123 with Coro::Util::inet_aton.
124 - do not override $Coro::idle unconditionally in Coro.pm, as other
125 modules could have provided their own idle coro already
126 (for exmaple, Coro::AnyEvent).
127 - fix Coro::Util::gethost* functions.
128 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
129
1305.21 Wed Dec 16 07:19:51 CET 2009
131 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
132 together.
133 - add some examples on how to combine other event loops with Coro in
134 Coro::AnyEvent, and how to run it (and not to block). Seems to be
135 the most common source of confusion.
136 - try to catch people naively blocking in an event callback.
137 - work around the perl filehandle bug issue in conjunction with
138 older common::sense (as indirectly pointed out by ZSystem).
139 - clarify the "not from signal handlers" section.
140
1415.2 Sun Oct 4 14:54:24 CEST 2009
142 - Coro::Storable destroyed the prototypes of the functions it wrapped.
143 - export rouse_cb and rouse_wait by default now.
144 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
145 - new method $state->swap_sv.
146 - added section on "windows process emulation" to the manpage,
147 after a not-so-fruitful (nor-friendly) "discussion" with chip
148 salzenberg (discussion implies arguments, but his only arguments
149 were ad-hominems, one wonders why he started it in the first
150 place). I hope this explains it well enough for him to understand,
151 and maybe well enough for others to understand.
152 - use common::sense everywhere now.
153 - idle callbacks are no longer supported, use idle coros instead.
154 - print a thread listing when a deadlock is detected.
155
1565.17 Sat Aug 22 23:09:31 CEST 2009
157 - work around a bug in the perl debugger causing crashes
158 when running under the debugger by marking _pool_handler as nodebug.
159 - speed up Coro::async considerably.
160 - try some hacks to get netbsd to work "more often" - their broken
161 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
162 - convert Coro to AE 5.0 API.
163
1645.162 Tue Jul 28 04:04:03 CEST 2009
165 - perl 5.8.2 is now minimum requirement.
166 - skip t/19_handle.t on broken windows perls.
167
1685.161 Wed Jul 22 04:47:38 CEST 2009
169 - Coro::AnyEvent::poll could have a different prototype when EV was
170 used as backend (analyzed by Tatsuhiko Miyagawa).
171 - Coro::AnyEvent errornously initialised the event loop when loaded,
172 not on demand.
173 - try to workaround rare */t/01_unblock.t failures.
174
1755.16 Tue Jul 21 01:44:37 CEST 2009
176 - Coro::AnyEvent failed to hook into the event loop
177 when no threads had been readied between detecting
178 the event loop and actually running it.
179 - considerably speed up Coro::Select by taking avdantage
180 of AnyEvent > 4.8 and some other optimisations.
181 - implement paragraph readline mode in Coro::Handle
182 (based on patches by Zsbán Ambrus).
183 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
184 and analyzed by Yasuhiro MATSUMOTO).
185 - clarified libcoro license and copyright.
186 - someone stole my EXTRA_META!!!
187 - implement Coro::Select::patch_pp_sselect and it's brother,
188 for hardcode select overriding.
189
1905.151 Mon Jul 6 05:41:57 CEST 2009
191 - backport to windows process emulation code again (patch by
192 Yasuhiro MATSUMOTO).
193 - slightly update Coro::MakeMaker.
194
1955.15 Tue Jun 30 10:28:06 CEST 2009
196 - deprecate Coro::Socket, document how to get ipv6 support via
197 AnyEvent::Socket instead.
198 - implement signal->wait ($cb) interface, similar to semaphores.
199 - work around SvOK not supporting getmagic, so we have to getmagic
200 to test for undef :/ (reported by Matthias Waldorf).
201 - load Coro::AnyEvent in all modules using AnyEvent.
202 - work around perl corrupting our internal data structures,
203 reported by Tokuhiro Matsuno.
204 - enable per-coroutine real and cpu time gathering
205 (Coro::State::enable_times).
206
2075.14 Wed Jun 24 01:37:48 CEST 2009
208 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
209 as this seems to have been a difficult concept (poll, sleep, idle,
210 idle_upto).
211 - add Coro::AnyEvent::readable/writable functions.
212 - clarify Coro::EV/Event/AnyEvent manpages.
213 - free per-thread global scalars in the thread calling ->cancel, to
214 avoid crashes when $_, $@ etc., are magical but some of those
215 had already been freed.
216 - "unexperimentalise" the callback interface for Coro::Semaphore.
217 - speed up ready queue management/context switching by using a linked
218 list instead of an array (~5-10%).
219 - implement "watch" command in Coro::Debug shells.
220 - for fun, implement time-slicing as an example in the manpage.
221 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
222 Coro::Event, use the normal AnyEvent handling instead of dieing
223 (the same is true for Coro::Handle).
224 - properly document Coro::EV::timed_io_once.
225 - avoid unneccessary ->cancel calls in Coro::Handle.
226 - maybe make it work on mingw32 with win32 backend
227 (based on patch by Yasuhiro Matsumoto).
228
2295.132 Fri May 29 09:00:39 CEST 2009
230 - do not keep a reference to the argument itself in
231 Coro::Semaphore::guard, as it could change later.
232 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
233 should support a prepare callback.
234
2355.131 Mon Mar 16 23:20:37 CET 2009
236 - implement and document Coro->suspend, Coro->resume.
237 - fix Coro::Select implementation to not (often) close
238 the passed file descriptors (testcase provided by pippijn).
239
2405.13 Mon Dec 15 21:51:42 CET 2008
241 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
242 - don't set diehook to C<undef> but instead to NULL, to avoid
243 spurious warnings.
244 - fix a lot of bugs in Coro::SemaphoreSet.
245 - Coro::SemaphoreSet will less often create a semaphore needlessly.
246 - add Coro::SemaphoreSet::count and wait methods.
247 - take advantage of the new Guard module.
248 - deprecate Coro::guard.
249 - try to fix the dreaded 01_unblock tests once more. I hate it when
250 testsuites need more fixing than the code they are supposed to test.
251 - croak in more cases when a required callback isn't resolvable.
252 - fix some minor issues in Coro::State->call/eval.
253 - use current coroutine context instead of a temporary one
254 when temporarily switching to another coroutine.
255 - do not call C-level on_destroy handlers during global destruction,
256 to avoid needless segfaults.
257
2585.12 Sun Dec 7 13:30:38 CET 2008
259 - add default config for MirOS, which seems to be a bug-to-bug
260 compatible fork of openbsd ("world domination by releasing
261 openbsd cvs before the openbsd folks do it" or so :).
262 - free_padlist did destroy the names pad, not good, but didn't
263 seem to bother perl - this could fix issues such as eval ""
264 inside a function called from multiple coroutines.
265 - use a different method to detect destruction time.
266 - be more careful when freeing padlists just before global
267 destruction.
268 - fixed and expanded the call/cc example.
269 - renamed _terminate to _coro_run.
270 - new method Coro::Channel->shutdown.
271 - try pthreads on openbsd <4.4 (broken sigaltstack, will
272 pthreads fare better?).
273 - be less picky about destroying "a" running coroutine.
274
2755.11 Tue Nov 25 21:49:05 CET 2008
276 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
277 working code with failed assertions, introducing more bugs than
278 it fixes, requiring elaborate workarounds :(
279
2805.1 Mon Nov 24 08:54:59 CET 2008
281 - wrote a small introductory tutorial - Coro::Intro.
282 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
283 - Coro::Select did errornously dup the file descriptors
284 and didn't work with all AnyEvent backends.
285 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
286 LWP data transfers.
287 - disassociate c contexts from coro objects - this is agruably more
288 correct, but mostly allows sharing of cctxs between coro and state
289 objects, for added memory savings and speed increases.
290 - bumped $Coro::POOL_RSS up to 32kb by default.
291 - no longer set the optype to OP_CUSTOM, as B::* understandably
292 doesn't like this very much (and we *are* a type of entersub).
293 - implement state cloning, just to prove that call/cc can be done.
294 - automatically load Coro::AnyEvent in Coro::Handle.
295 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
296 do not have this method (and don't need it either).
297 - speed up generic anyevent methods in Coro::Handle by using rouse
298 callbacks.
299 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
300 others. It also makes the debugger happier, as you can trace
301 through the idle threads now.
302 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
303 - eg/event was pretty broken.
304 - better 5.8.6 compatibility.
305
3065.0 Thu Nov 20 10:35:05 CET 2008
307 - NEW ARCHITECTURE: use the latest 4.x version if you experience
308 stability issues.
309 - bump API version to 7 - all dependents must be recompiled.
310 - removed timed_* functions - they were not being used anyways
311 and should be replaced by a more generic mechanism -
312 and were annoying to support anyways :)
313 - removed SemaphoreSet's waiter method - use sem method instead.
314 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
315 - async_pool did free a scalar value twice
316 ("Attempt to unreference...").
317 - fix a longstanding bug where calling terminate on a coro that
318 was waiting for a semaphore that was just becoming available
319 would cause a deadlock (semaphore would get into a state where
320 it was available but waiters were still blocked).
321 - calling throw on a coroutine that is waiting for a semaphore will
322 no longer make it acquire the semaphore (and thus leak a count).
323 - perl's process emulation is now not even theoretically supported
324 anymore.
325 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
326 conversion of callback-style to blocking-style.
327 - new methods $coro->schedule_to and ->cede_to, to specifically
328 schedule or cede to a specific coroutine.
329 - new function Coro::Semaphore::wait.
330 - use named constants in Coro::Channel (Richard Hundt).
331 - directly patch the entersub opcode calling SLF functions (cede,
332 transfer and so on). this does speed up context switching, but
333 more importanly, it frees us from the hardcoded behaviour of
334 entersub, so we might actually be able to return something from
335 those functions and atcually create new ones.
336 - take advantage of __builtin_frame_address on gcc.
337 - expose THX in coroapi (not sure whether this was a wise decision,
338 as "threaded" perls are running at half speed anyways).
339 - implement execute_slf (schedule-like-function) interface that makes
340 it possible to implement schedule-like-functions in XS.
341 - use new SLF interface to massively speed up Coro::EV by roughly a
342 factor of two.
343 - used new SLF interface to massively speed up Coro::Semaphore by a
344 factor of three.
345 - used new SLF interface to speed up Coro::AIO by roughly a factor of
346 four and reduce its memory usage considerably.
347 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
348 for a nice speedup and vastly more correct behaviour. Also implement
349 a new method "sem" to get at the underlying semaphore object.
350 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
351 (in comparison) 20-40% speedup.
352 - used new SLF interface to reimplement Coro::Signal gaining
353 some unknown (because I was too lazy), but certain, speedup, and also
354 making signals reliable for the first time.
355 - used new SLF interface and other optimisations to speed up async_pool
356 by a factor of two. It also doesn't rely on perl's exception mechanism
357 to exit anymore. The overhead for terminating an async_pool, coro over
358 a normal async is now very small.
359 - sped up coroutine creation/destruction by 40%.
360 - forgot to include Coro/libcoro/README in the dist for all these years.
361 - work around a freebsd pthreads bug (manual testcancel is required as
362 pthread_cond_wait isn't a cancellation point on freebsd).
363 - use new rouse functions to speed up and simplify Coro::BDB.
364 - make "prefer perl native functions" work with threaded perls.
365 - condense Coro::Debug ps output, hint at v and w flags.
366 - (libcoro) lots of minor cleanups and portability improvements.
367
3684.914 Wed Nov 19 12:54:18 CET 2008
369 - fix a disastrous bug in the readline optimisation
370 introduced in 4.801.
371
3724.913 Sat Nov 15 07:58:28 CET 2008
373 - async_pool did free a scalar value twice
374 ("Attempt to unreference...").
375
3764.912 Thu Nov 13 18:31:23 CET 2008
377 - minor cleanups.
378 - use much larger stacks on linux and perl < 5.8.8.
379 - Coro::Debug::new_unix_server did not unlink the socket
380 when destroyed.
381
3824.911 Tue Nov 11 04:26:17 CET 2008
383 - "port" to threaded perls.
384
3854.91 Mon Nov 10 05:36:38 CET 2008
386 - the ->throw exception object no longer leaks.
387 - creating a new cctx leaked a scopestack entry (memleak).
388 - new coroutines didn't get created with a zero flags field
389 (unknown impact).
9 - calling ->throw on a not-yet-started coroutine should now work 390 - calling ->throw on a not-yet-started coroutine should now work
10 instead of being ignored. 391 instead of being ignored.
11 - typo: new coroutines didn't get created with a zero flags field 392 - ->throw is now supported on Coro::State objects.
12 (unknown impact).
13 - clean up cctx creation code a bit. 393 - clean up cctx creation code a bit.
394 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
14 395
154.9 Sat Nov 8 17:45:27 CET 2008 3964.9 Sat Nov 8 17:45:27 CET 2008
16 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 397 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
17 - (libcoro) no longer leak threads in the experimental pthread backend, 398 - (libcoro) no longer leak threads in the experimental pthread backend,
18 also speed it up considerably. 399 also speed it up considerably.

Diff Legend

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