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

Comparing Coro/Changes (file contents):
Revision 1.386 by root, Mon Nov 3 16:05:37 2008 UTC vs.
Revision 1.566 by root, Mon May 16 22:08:24 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
84.803 7TODO: tutorial
8 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
9 destroyed and cleaned up automatically (e.g. async { schedule }).
10 - Coro::State->is_destroyed is now called is_zombie.
11 - implement a Coro->safe_cancel method that might fail, but
12 cancels in a "safer" way if it succeeds.
13 - get rid of two hash-accesses when initialising a new Coro - this
14 speeds up coro creation by almost a factor of two.
15 - croak when a coro that is being cancelled tries to block
16 (e.g. while executing a guard block), instead of crashing or
17 deadlocking.
18 - use a more robust and also faster method to identify Coro::State
19 objects - speeds up everything a bit.
20 - implement Coro->cancel in XS for a 20% speed improvement, and to
21 be able to implement mutual cancellation.
22 - speed up context switches by a percent or two by more efficiently
23 allocating context stack entries.
24 - implement Coro->join and Coro->on_destroy in XS for a speedup and
25 a reduction in memory use.
26 - cancelling a coro while it itself is cancelling another coro is
27 now supported and working, instead of triggering an assertion.
28 - be a bit more crash-resistant when calling (buggy) on_destroy
29 callbacks (best effort).
30 - move on_destroy into the slf_frame, to allow extension slf
31 functions to have destructors.
32 - get rid if coro refcounting - simply crash in other interpreter
33 threads by nulling the pointers on clone.
34 - simplify warn/die hook handling when loading Coro - the convoluted
35 logic seems to be no longer neccessary.
36
375.372 Wed Feb 23 06:14:30 CET 2011
38 - apparently mingw doesn't provide a working gettimeofday, try to
39 work around that by relying on Time::HiRes (indirectly brought to
40 my attention by Max Maischein).
41 - fix some portability issues when Time::HiRes was used.
42
435.371 Mon Feb 21 14:36:08 CET 2011
44 - backport to windows process emulation code again.
45
465.37 Sat Feb 19 07:49:44 CET 2011
47 - add a big "Coro thread life cycle" section to "man Coro".
48 - try a tentative workaround against the breakage that 5.13 has
49 introduced without depreciation period. sigh.
50 - no longer use Time::HiRes if gettimeofday is available, which
51 saves quite a lot of memory.
52
535.36 Sun Feb 13 05:33:41 CET 2011
54 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
55 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
56 method call.
57 - undocument Coro::Timer::sleep and obsolete whole module.
58 - optimise Coro::Timer::timeout memory and cpu usage.
59 - slightly updated Coro::Intro for recent changes.
60 - do not initialise PL_dirty anymore.
61
625.25 Thu Nov 11 01:08:39 CET 2010
63 - try a different approach on netbsd - netbsd 5 finally has marginally
64 working pthreads, but still broken ucontext/sigaltstack.
65 - openbsd 4.8 finally got their act together, Coro works out of the box
66 with asm, setjmp and pthreads (no change, just informational).
67
685.24 Sat Oct 23 11:27:12 CEST 2010
69 - port to the EV 4.0 API.
70 - work around bugs in mingw32, making strawberry perl work
71 out of the box.
72 - correctly modify Coro::AIO function prototypes
73 so that they reflect the "no optional parameters" rule.
74 - "ported" libcoro to C++.
75
765.23 Mon May 17 18:50:42 CEST 2010
77 - be more resistant to ordering changes when initialising
78 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
79 Waldorf).
80 - document that perl 5.12 deliberately removed support for cloning.
81
825.22 Wed Apr 14 03:55:35 CEST 2010
83 - correctly return udnef on errors in Coro::Handle::read/write
84 (testcase by Marc Mims).
85 - convert Coro::Util into a "perl compatibility wrapper" - the functions
86 are less useful now, but are drop-in replacements for existing
87 functions, listing better alternatives in the documentation. This also
88 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
89 with Coro::Util::inet_aton.
90 - do not override $Coro::idle unconditionally in Coro.pm, as other
91 modules could have provided their own idle coro already
92 (for exmaple, Coro::AnyEvent).
93 - fix Coro::Util::gethost* functions.
94 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
95
965.21 Wed Dec 16 07:19:51 CET 2009
97 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
98 together.
99 - add some examples on how to combine other event loops with Coro in
100 Coro::AnyEvent, and how to run it (and not to block). Seems to be
101 the most common source of confusion.
102 - try to catch people naively blocking in an event callback.
103 - work around the perl filehandle bug issue in conjunction with
104 older common::sense (as indirectly pointed out by ZSystem).
105 - clarify the "not from signal handlers" section.
106
1075.2 Sun Oct 4 14:54:24 CEST 2009
108 - Coro::Storable destroyed the prototypes of the functions it wrapped.
109 - export rouse_cb and rouse_wait by default now.
110 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
111 - new method $state->swap_sv.
112 - added section on "windows process emulation" to the manpage,
113 after a not-so-fruitful (nor-friendly) "discussion" with chip
114 salzenberg (discussion implies arguments, but his only arguments
115 were ad-hominems, one wonders why he started it in the first
116 place). I hope this explains it well enough for him to understand,
117 and maybe well enough for others to understand.
118 - use common::sense everywhere now.
119 - idle callbacks are no longer supported, use idle coros instead.
120 - print a thread listing when a deadlock is detected.
121
1225.17 Sat Aug 22 23:09:31 CEST 2009
123 - work around a bug in the perl debugger causing crashes
124 when running under the debugger by marking _pool_handler as nodebug.
125 - speed up Coro::async considerably.
126 - try some hacks to get netbsd to work "more often" - their broken
127 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
128 - convert Coro to AE 5.0 API.
129
1305.162 Tue Jul 28 04:04:03 CEST 2009
131 - perl 5.8.2 is now minimum requirement.
132 - skip t/19_handle.t on broken windows perls.
133
1345.161 Wed Jul 22 04:47:38 CEST 2009
135 - Coro::AnyEvent::poll could have a different prototype when EV was
136 used as backend (analyzed by Tatsuhiko Miyagawa).
137 - Coro::AnyEvent errornously initialised the event loop when loaded,
138 not on demand.
139 - try to workaround rare */t/01_unblock.t failures.
140
1415.16 Tue Jul 21 01:44:37 CEST 2009
142 - Coro::AnyEvent failed to hook into the event loop
143 when no threads had been readied between detecting
144 the event loop and actually running it.
145 - considerably speed up Coro::Select by taking avdantage
146 of AnyEvent > 4.8 and some other optimisations.
147 - implement paragraph readline mode in Coro::Handle
148 (based on patches by Zsbán Ambrus).
149 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
150 and analyzed by Yasuhiro MATSUMOTO).
151 - clarified libcoro license and copyright.
152 - someone stole my EXTRA_META!!!
153 - implement Coro::Select::patch_pp_sselect and it's brother,
154 for hardcode select overriding.
155
1565.151 Mon Jul 6 05:41:57 CEST 2009
157 - backport to windows process emulation code again (patch by
158 Yasuhiro MATSUMOTO).
159 - slightly update Coro::MakeMaker.
160
1615.15 Tue Jun 30 10:28:06 CEST 2009
162 - deprecate Coro::Socket, document how to get ipv6 support via
163 AnyEvent::Socket instead.
164 - implement signal->wait ($cb) interface, similar to semaphores.
165 - work around SvOK not supporting getmagic, so we have to getmagic
166 to test for undef :/ (reported by Matthias Waldorf).
167 - load Coro::AnyEvent in all modules using AnyEvent.
168 - work around perl corrupting our internal data structures,
169 reported by Tokuhiro Matsuno.
170 - enable per-coroutine real and cpu time gathering
171 (Coro::State::enable_times).
172
1735.14 Wed Jun 24 01:37:48 CEST 2009
174 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
175 as this seems to have been a difficult concept (poll, sleep, idle,
176 idle_upto).
177 - add Coro::AnyEvent::readable/writable functions.
178 - clarify Coro::EV/Event/AnyEvent manpages.
179 - free per-thread global scalars in the thread calling ->cancel, to
180 avoid crashes when $_, $@ etc., are magical but some of those
181 had already been freed.
182 - "unexperimentalise" the callback interface for Coro::Semaphore.
183 - speed up ready queue management/context switching by using a linked
184 list instead of an array (~5-10%).
185 - implement "watch" command in Coro::Debug shells.
186 - for fun, implement time-slicing as an example in the manpage.
187 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
188 Coro::Event, use the normal AnyEvent handling instead of dieing
189 (the same is true for Coro::Handle).
190 - properly document Coro::EV::timed_io_once.
191 - avoid unneccessary ->cancel calls in Coro::Handle.
192 - maybe make it work on mingw32 with win32 backend
193 (based on patch by Yasuhiro Matsumoto).
194
1955.132 Fri May 29 09:00:39 CEST 2009
196 - do not keep a reference to the argument itself in
197 Coro::Semaphore::guard, as it could change later.
198 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
199 should support a prepare callback.
200
2015.131 Mon Mar 16 23:20:37 CET 2009
202 - implement and document Coro->suspend, Coro->resume.
203 - fix Coro::Select implementation to not (often) close
204 the passed file descriptors (testcase provided by pippijn).
205
2065.13 Mon Dec 15 21:51:42 CET 2008
207 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
208 - don't set diehook to C<undef> but instead to NULL, to avoid
209 spurious warnings.
210 - fix a lot of bugs in Coro::SemaphoreSet.
211 - Coro::SemaphoreSet will less often create a semaphore needlessly.
212 - add Coro::SemaphoreSet::count and wait methods.
213 - take advantage of the new Guard module.
214 - deprecate Coro::guard.
215 - try to fix the dreaded 01_unblock tests once more. I hate it when
216 testsuites need more fixing than the code they are supposed to test.
217 - croak in more cases when a required callback isn't resolvable.
218 - fix some minor issues in Coro::State->call/eval.
219 - use current coroutine context instead of a temporary one
220 when temporarily switching to another coroutine.
221 - do not call C-level on_destroy handlers during global destruction,
222 to avoid needless segfaults.
223
2245.12 Sun Dec 7 13:30:38 CET 2008
225 - add default config for MirOS, which seems to be a bug-to-bug
226 compatible fork of openbsd ("world domination by releasing
227 openbsd cvs before the openbsd folks do it" or so :).
228 - free_padlist did destroy the names pad, not good, but didn't
229 seem to bother perl - this could fix issues such as eval ""
230 inside a function called from multiple coroutines.
231 - use a different method to detect destruction time.
232 - be more careful when freeing padlists just before global
233 destruction.
234 - fixed and expanded the call/cc example.
235 - renamed _terminate to _coro_run.
236 - new method Coro::Channel->shutdown.
237 - try pthreads on openbsd <4.4 (broken sigaltstack, will
238 pthreads fare better?).
239 - be less picky about destroying "a" running coroutine.
240
2415.11 Tue Nov 25 21:49:05 CET 2008
242 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
243 working code with failed assertions, introducing more bugs than
244 it fixes, requiring elaborate workarounds :(
245
2465.1 Mon Nov 24 08:54:59 CET 2008
247 - wrote a small introductory tutorial - Coro::Intro.
248 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
249 - Coro::Select did errornously dup the file descriptors
250 and didn't work with all AnyEvent backends.
251 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
252 LWP data transfers.
253 - disassociate c contexts from coro objects - this is agruably more
254 correct, but mostly allows sharing of cctxs between coro and state
255 objects, for added memory savings and speed increases.
256 - bumped $Coro::POOL_RSS up to 32kb by default.
257 - no longer set the optype to OP_CUSTOM, as B::* understandably
258 doesn't like this very much (and we *are* a type of entersub).
259 - implement state cloning, just to prove that call/cc can be done.
260 - automatically load Coro::AnyEvent in Coro::Handle.
261 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
262 do not have this method (and don't need it either).
263 - speed up generic anyevent methods in Coro::Handle by using rouse
264 callbacks.
265 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
266 others. It also makes the debugger happier, as you can trace
267 through the idle threads now.
268 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
269 - eg/event was pretty broken.
270 - better 5.8.6 compatibility.
271
2725.0 Thu Nov 20 10:35:05 CET 2008
273 - NEW ARCHITECTURE: use the latest 4.x version if you experience
274 stability issues.
275 - bump API version to 7 - all dependents must be recompiled.
276 - removed timed_* functions - they were not being used anyways
277 and should be replaced by a more generic mechanism -
278 and were annoying to support anyways :)
279 - removed SemaphoreSet's waiter method - use sem method instead.
280 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
281 - async_pool did free a scalar value twice
282 ("Attempt to unreference...").
283 - fix a longstanding bug where calling terminate on a coro that
284 was waiting for a semaphore that was just becoming available
285 would cause a deadlock (semaphore would get into a state where
286 it was available but waiters were still blocked).
287 - calling throw on a coroutine that is waiting for a semaphore will
288 no longer make it acquire the semaphore (and thus leak a count).
289 - perl's process emulation is now not even theoretically supported
290 anymore.
291 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
292 conversion of callback-style to blocking-style.
293 - new methods $coro->schedule_to and ->cede_to, to specifically
294 schedule or cede to a specific coroutine.
295 - new function Coro::Semaphore::wait.
296 - use named constants in Coro::Channel (Richard Hundt).
297 - directly patch the entersub opcode calling SLF functions (cede,
298 transfer and so on). this does speed up context switching, but
299 more importanly, it frees us from the hardcoded behaviour of
300 entersub, so we might actually be able to return something from
301 those functions and atcually create new ones.
302 - take advantage of __builtin_frame_address on gcc.
303 - expose THX in coroapi (not sure whether this was a wise decision,
304 as "threaded" perls are running at half speed anyways).
305 - implement execute_slf (schedule-like-function) interface that makes
306 it possible to implement schedule-like-functions in XS.
307 - use new SLF interface to massively speed up Coro::EV by roughly a
308 factor of two.
309 - used new SLF interface to massively speed up Coro::Semaphore by a
310 factor of three.
311 - used new SLF interface to speed up Coro::AIO by roughly a factor of
312 four and reduce its memory usage considerably.
313 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
314 for a nice speedup and vastly more correct behaviour. Also implement
315 a new method "sem" to get at the underlying semaphore object.
316 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
317 (in comparison) 20-40% speedup.
318 - used new SLF interface to reimplement Coro::Signal gaining
319 some unknown (because I was too lazy), but certain, speedup, and also
320 making signals reliable for the first time.
321 - used new SLF interface and other optimisations to speed up async_pool
322 by a factor of two. It also doesn't rely on perl's exception mechanism
323 to exit anymore. The overhead for terminating an async_pool, coro over
324 a normal async is now very small.
325 - sped up coroutine creation/destruction by 40%.
326 - forgot to include Coro/libcoro/README in the dist for all these years.
327 - work around a freebsd pthreads bug (manual testcancel is required as
328 pthread_cond_wait isn't a cancellation point on freebsd).
329 - use new rouse functions to speed up and simplify Coro::BDB.
330 - make "prefer perl native functions" work with threaded perls.
331 - condense Coro::Debug ps output, hint at v and w flags.
332 - (libcoro) lots of minor cleanups and portability improvements.
333
3344.914 Wed Nov 19 12:54:18 CET 2008
335 - fix a disastrous bug in the readline optimisation
336 introduced in 4.801.
337
3384.913 Sat Nov 15 07:58:28 CET 2008
339 - async_pool did free a scalar value twice
340 ("Attempt to unreference...").
341
3424.912 Thu Nov 13 18:31:23 CET 2008
343 - minor cleanups.
344 - use much larger stacks on linux and perl < 5.8.8.
345 - Coro::Debug::new_unix_server did not unlink the socket
346 when destroyed.
347
3484.911 Tue Nov 11 04:26:17 CET 2008
349 - "port" to threaded perls.
350
3514.91 Mon Nov 10 05:36:38 CET 2008
352 - the ->throw exception object no longer leaks.
353 - creating a new cctx leaked a scopestack entry (memleak).
354 - new coroutines didn't get created with a zero flags field
355 (unknown impact).
356 - calling ->throw on a not-yet-started coroutine should now work
357 instead of being ignored.
358 - ->throw is now supported on Coro::State objects.
359 - clean up cctx creation code a bit.
360 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
361
3624.9 Sat Nov 8 17:45:27 CET 2008
363 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
364 - (libcoro) no longer leak threads in the experimental pthread backend,
365 also speed it up considerably.
366 - (libcoro) do not rely on makecontext passing void *'s unscathed.
367 - fix compiletime dependencies on libcoro in the Makefile.
368 - cctx_count wasn't always updated properly.
369 - Coro::State::cctx_stacksize wasn't applied correctly.
370 - new function Coro::State::cctx_max_idle.
371 - the default max number of idle C contexts is now 4.
372 - (libcoro) try harder to get _setjmp/_longjmp.
373 - (libcoro) cleanup and extend the libcoro API to officially
374 allow the creation of empty source contexts.
375 - very experimental workaround for the totally broken netbsd platform.
376 - tried to hack around openbsd bugs.
377
3784.804 Wed Nov 5 16:36:00 CET 2008
379 - Coro::Debug::new_unix_server would not create a non-blocking listening
380 socket, sometimes causing freezes.
381 - (libcoro) fix misaligned stack points for setjmp and assembly
382 methods, which can cause crashes on x86/x86_64 with a sufficiently
383 aggressive compiler.
384 - new function: Coro::Debug::new_tcp_server.
385 - move ->throw into the Coro class because it only works on coro objects.
386
3874.803 Mon Nov 3 17:16:12 CET 2008
9 - use a global asm statement to become independent of gcc otpimisations 388 - (libcoro) use a global asm statement to become independent of gcc
10 for CORO_ASM (thanks to pippijn for the idea). 389 otpimisations for CORO_ASM (thanks to pippijn for the idea).
390 - try to workaround yet another broken bsd, this time dragonfly.
11 391
124.802 Thu Oct 30 10:56:12 CET 2008 3924.802 Thu Oct 30 10:56:12 CET 2008
13 - support -fno-omit-frame-pointer on x86 with the assembly method. 393 - support -fno-omit-frame-pointer on x86 with the assembly method.
14 - tune 01_unblock.t tests a bit. 394 - tune 01_unblock.t tests a bit.
15 395

Diff Legend

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