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

Comparing Coro/Changes (file contents):
Revision 1.388 by root, Tue Nov 4 12:08:44 2008 UTC vs.
Revision 1.569 by root, Sat Jun 11 13:01:25 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
7TODO: tutorial
8TODO: in_eval, PL_tained
9 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
10 destroyed and cleaned up automatically (e.g. async { schedule }).
11 - implement a JIT compiler for part of the thread switch code,
12 which gives a 50% speed improvement on threaded perls, and
13 about 4% on non-thraeded perls (so threaded perls now finally
14 reach about half the speed of non-threaded perls).
15 - avoid DEFSV and ERRSV, giving another 10% improvement
16 in thread switching.
17 - Coro::State->is_destroyed is now called is_zombie.
18 - implement a Coro->safe_cancel method that might fail, but
19 cancels in a "safer" way if it succeeds.
20 - get rid of two hash-accesses when initialising a new Coro - this
21 speeds up coro creation by almost a factor of two.
22 - croak when a coro that is being cancelled tries to block
23 (e.g. while executing a guard block), instead of crashing or
24 deadlocking.
25 - use a more robust and also faster method to identify Coro::State
26 objects - speeds up everything a bit.
27 - implement Coro->cancel in XS for a 20% speed improvement, and to
28 be able to implement mutual cancellation.
29 - speed up context switches by a percent or two by more efficiently
30 allocating context stack entries.
31 - implement Coro->join and Coro->on_destroy in XS for a speedup and
32 a reduction in memory use.
33 - cancelling a coro while it itself is cancelling another coro is
34 now supported and working, instead of triggering an assertion.
35 - be a bit more crash-resistant when calling (buggy) on_destroy
36 callbacks (best effort).
37 - move on_destroy into the slf_frame, to allow extension slf
38 functions to have destructors.
39 - get rid if coro refcounting - simply crash in other interpreter
40 threads by nulling the pointers on clone.
41 - simplify warn/die hook handling when loading Coro - the convoluted
42 logic seems to be no longer neccessary.
43 - use libecb instead of our own home-grown gcc hacks.
44
455.372 Wed Feb 23 06:14:30 CET 2011
46 - apparently mingw doesn't provide a working gettimeofday, try to
47 work around that by relying on Time::HiRes (indirectly brought to
48 my attention by Max Maischein).
49 - fix some portability issues when Time::HiRes was used.
50
515.371 Mon Feb 21 14:36:08 CET 2011
52 - backport to windows process emulation code again.
53
545.37 Sat Feb 19 07:49:44 CET 2011
55 - add a big "Coro thread life cycle" section to "man Coro".
56 - try a tentative workaround against the breakage that 5.13 has
57 introduced without depreciation period. sigh.
58 - no longer use Time::HiRes if gettimeofday is available, which
59 saves quite a lot of memory.
60
615.36 Sun Feb 13 05:33:41 CET 2011
62 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
63 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
64 method call.
65 - undocument Coro::Timer::sleep and obsolete whole module.
66 - optimise Coro::Timer::timeout memory and cpu usage.
67 - slightly updated Coro::Intro for recent changes.
68 - do not initialise PL_dirty anymore.
69
705.25 Thu Nov 11 01:08:39 CET 2010
71 - try a different approach on netbsd - netbsd 5 finally has marginally
72 working pthreads, but still broken ucontext/sigaltstack.
73 - openbsd 4.8 finally got their act together, Coro works out of the box
74 with asm, setjmp and pthreads (no change, just informational).
75
765.24 Sat Oct 23 11:27:12 CEST 2010
77 - port to the EV 4.0 API.
78 - work around bugs in mingw32, making strawberry perl work
79 out of the box.
80 - correctly modify Coro::AIO function prototypes
81 so that they reflect the "no optional parameters" rule.
82 - "ported" libcoro to C++.
83
845.23 Mon May 17 18:50:42 CEST 2010
85 - be more resistant to ordering changes when initialising
86 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
87 Waldorf).
88 - document that perl 5.12 deliberately removed support for cloning.
89
905.22 Wed Apr 14 03:55:35 CEST 2010
91 - correctly return udnef on errors in Coro::Handle::read/write
92 (testcase by Marc Mims).
93 - convert Coro::Util into a "perl compatibility wrapper" - the functions
94 are less useful now, but are drop-in replacements for existing
95 functions, listing better alternatives in the documentation. This also
96 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
97 with Coro::Util::inet_aton.
98 - do not override $Coro::idle unconditionally in Coro.pm, as other
99 modules could have provided their own idle coro already
100 (for exmaple, Coro::AnyEvent).
101 - fix Coro::Util::gethost* functions.
102 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
103
1045.21 Wed Dec 16 07:19:51 CET 2009
105 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
106 together.
107 - add some examples on how to combine other event loops with Coro in
108 Coro::AnyEvent, and how to run it (and not to block). Seems to be
109 the most common source of confusion.
110 - try to catch people naively blocking in an event callback.
111 - work around the perl filehandle bug issue in conjunction with
112 older common::sense (as indirectly pointed out by ZSystem).
113 - clarify the "not from signal handlers" section.
114
1155.2 Sun Oct 4 14:54:24 CEST 2009
116 - Coro::Storable destroyed the prototypes of the functions it wrapped.
117 - export rouse_cb and rouse_wait by default now.
118 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
119 - new method $state->swap_sv.
120 - added section on "windows process emulation" to the manpage,
121 after a not-so-fruitful (nor-friendly) "discussion" with chip
122 salzenberg (discussion implies arguments, but his only arguments
123 were ad-hominems, one wonders why he started it in the first
124 place). I hope this explains it well enough for him to understand,
125 and maybe well enough for others to understand.
126 - use common::sense everywhere now.
127 - idle callbacks are no longer supported, use idle coros instead.
128 - print a thread listing when a deadlock is detected.
129
1305.17 Sat Aug 22 23:09:31 CEST 2009
131 - work around a bug in the perl debugger causing crashes
132 when running under the debugger by marking _pool_handler as nodebug.
133 - speed up Coro::async considerably.
134 - try some hacks to get netbsd to work "more often" - their broken
135 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
136 - convert Coro to AE 5.0 API.
137
1385.162 Tue Jul 28 04:04:03 CEST 2009
139 - perl 5.8.2 is now minimum requirement.
140 - skip t/19_handle.t on broken windows perls.
141
1425.161 Wed Jul 22 04:47:38 CEST 2009
143 - Coro::AnyEvent::poll could have a different prototype when EV was
144 used as backend (analyzed by Tatsuhiko Miyagawa).
145 - Coro::AnyEvent errornously initialised the event loop when loaded,
146 not on demand.
147 - try to workaround rare */t/01_unblock.t failures.
148
1495.16 Tue Jul 21 01:44:37 CEST 2009
150 - Coro::AnyEvent failed to hook into the event loop
151 when no threads had been readied between detecting
152 the event loop and actually running it.
153 - considerably speed up Coro::Select by taking avdantage
154 of AnyEvent > 4.8 and some other optimisations.
155 - implement paragraph readline mode in Coro::Handle
156 (based on patches by Zsbán Ambrus).
157 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
158 and analyzed by Yasuhiro MATSUMOTO).
159 - clarified libcoro license and copyright.
160 - someone stole my EXTRA_META!!!
161 - implement Coro::Select::patch_pp_sselect and it's brother,
162 for hardcode select overriding.
163
1645.151 Mon Jul 6 05:41:57 CEST 2009
165 - backport to windows process emulation code again (patch by
166 Yasuhiro MATSUMOTO).
167 - slightly update Coro::MakeMaker.
168
1695.15 Tue Jun 30 10:28:06 CEST 2009
170 - deprecate Coro::Socket, document how to get ipv6 support via
171 AnyEvent::Socket instead.
172 - implement signal->wait ($cb) interface, similar to semaphores.
173 - work around SvOK not supporting getmagic, so we have to getmagic
174 to test for undef :/ (reported by Matthias Waldorf).
175 - load Coro::AnyEvent in all modules using AnyEvent.
176 - work around perl corrupting our internal data structures,
177 reported by Tokuhiro Matsuno.
178 - enable per-coroutine real and cpu time gathering
179 (Coro::State::enable_times).
180
1815.14 Wed Jun 24 01:37:48 CEST 2009
182 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
183 as this seems to have been a difficult concept (poll, sleep, idle,
184 idle_upto).
185 - add Coro::AnyEvent::readable/writable functions.
186 - clarify Coro::EV/Event/AnyEvent manpages.
187 - free per-thread global scalars in the thread calling ->cancel, to
188 avoid crashes when $_, $@ etc., are magical but some of those
189 had already been freed.
190 - "unexperimentalise" the callback interface for Coro::Semaphore.
191 - speed up ready queue management/context switching by using a linked
192 list instead of an array (~5-10%).
193 - implement "watch" command in Coro::Debug shells.
194 - for fun, implement time-slicing as an example in the manpage.
195 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
196 Coro::Event, use the normal AnyEvent handling instead of dieing
197 (the same is true for Coro::Handle).
198 - properly document Coro::EV::timed_io_once.
199 - avoid unneccessary ->cancel calls in Coro::Handle.
200 - maybe make it work on mingw32 with win32 backend
201 (based on patch by Yasuhiro Matsumoto).
202
2035.132 Fri May 29 09:00:39 CEST 2009
204 - do not keep a reference to the argument itself in
205 Coro::Semaphore::guard, as it could change later.
206 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
207 should support a prepare callback.
208
2095.131 Mon Mar 16 23:20:37 CET 2009
210 - implement and document Coro->suspend, Coro->resume.
211 - fix Coro::Select implementation to not (often) close
212 the passed file descriptors (testcase provided by pippijn).
213
2145.13 Mon Dec 15 21:51:42 CET 2008
215 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
216 - don't set diehook to C<undef> but instead to NULL, to avoid
217 spurious warnings.
218 - fix a lot of bugs in Coro::SemaphoreSet.
219 - Coro::SemaphoreSet will less often create a semaphore needlessly.
220 - add Coro::SemaphoreSet::count and wait methods.
221 - take advantage of the new Guard module.
222 - deprecate Coro::guard.
223 - try to fix the dreaded 01_unblock tests once more. I hate it when
224 testsuites need more fixing than the code they are supposed to test.
225 - croak in more cases when a required callback isn't resolvable.
226 - fix some minor issues in Coro::State->call/eval.
227 - use current coroutine context instead of a temporary one
228 when temporarily switching to another coroutine.
229 - do not call C-level on_destroy handlers during global destruction,
230 to avoid needless segfaults.
231
2325.12 Sun Dec 7 13:30:38 CET 2008
233 - add default config for MirOS, which seems to be a bug-to-bug
234 compatible fork of openbsd ("world domination by releasing
235 openbsd cvs before the openbsd folks do it" or so :).
236 - free_padlist did destroy the names pad, not good, but didn't
237 seem to bother perl - this could fix issues such as eval ""
238 inside a function called from multiple coroutines.
239 - use a different method to detect destruction time.
240 - be more careful when freeing padlists just before global
241 destruction.
242 - fixed and expanded the call/cc example.
243 - renamed _terminate to _coro_run.
244 - new method Coro::Channel->shutdown.
245 - try pthreads on openbsd <4.4 (broken sigaltstack, will
246 pthreads fare better?).
247 - be less picky about destroying "a" running coroutine.
248
2495.11 Tue Nov 25 21:49:05 CET 2008
250 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
251 working code with failed assertions, introducing more bugs than
252 it fixes, requiring elaborate workarounds :(
253
2545.1 Mon Nov 24 08:54:59 CET 2008
255 - wrote a small introductory tutorial - Coro::Intro.
256 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
257 - Coro::Select did errornously dup the file descriptors
258 and didn't work with all AnyEvent backends.
259 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
260 LWP data transfers.
261 - disassociate c contexts from coro objects - this is agruably more
262 correct, but mostly allows sharing of cctxs between coro and state
263 objects, for added memory savings and speed increases.
264 - bumped $Coro::POOL_RSS up to 32kb by default.
265 - no longer set the optype to OP_CUSTOM, as B::* understandably
266 doesn't like this very much (and we *are* a type of entersub).
267 - implement state cloning, just to prove that call/cc can be done.
268 - automatically load Coro::AnyEvent in Coro::Handle.
269 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
270 do not have this method (and don't need it either).
271 - speed up generic anyevent methods in Coro::Handle by using rouse
272 callbacks.
273 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
274 others. It also makes the debugger happier, as you can trace
275 through the idle threads now.
276 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
277 - eg/event was pretty broken.
278 - better 5.8.6 compatibility.
279
2805.0 Thu Nov 20 10:35:05 CET 2008
281 - NEW ARCHITECTURE: use the latest 4.x version if you experience
282 stability issues.
283 - bump API version to 7 - all dependents must be recompiled.
284 - removed timed_* functions - they were not being used anyways
285 and should be replaced by a more generic mechanism -
286 and were annoying to support anyways :)
287 - removed SemaphoreSet's waiter method - use sem method instead.
288 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
289 - async_pool did free a scalar value twice
290 ("Attempt to unreference...").
291 - fix a longstanding bug where calling terminate on a coro that
292 was waiting for a semaphore that was just becoming available
293 would cause a deadlock (semaphore would get into a state where
294 it was available but waiters were still blocked).
295 - calling throw on a coroutine that is waiting for a semaphore will
296 no longer make it acquire the semaphore (and thus leak a count).
297 - perl's process emulation is now not even theoretically supported
298 anymore.
299 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
300 conversion of callback-style to blocking-style.
301 - new methods $coro->schedule_to and ->cede_to, to specifically
302 schedule or cede to a specific coroutine.
303 - new function Coro::Semaphore::wait.
304 - use named constants in Coro::Channel (Richard Hundt).
305 - directly patch the entersub opcode calling SLF functions (cede,
306 transfer and so on). this does speed up context switching, but
307 more importanly, it frees us from the hardcoded behaviour of
308 entersub, so we might actually be able to return something from
309 those functions and atcually create new ones.
310 - take advantage of __builtin_frame_address on gcc.
311 - expose THX in coroapi (not sure whether this was a wise decision,
312 as "threaded" perls are running at half speed anyways).
313 - implement execute_slf (schedule-like-function) interface that makes
314 it possible to implement schedule-like-functions in XS.
315 - use new SLF interface to massively speed up Coro::EV by roughly a
316 factor of two.
317 - used new SLF interface to massively speed up Coro::Semaphore by a
318 factor of three.
319 - used new SLF interface to speed up Coro::AIO by roughly a factor of
320 four and reduce its memory usage considerably.
321 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
322 for a nice speedup and vastly more correct behaviour. Also implement
323 a new method "sem" to get at the underlying semaphore object.
324 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
325 (in comparison) 20-40% speedup.
326 - used new SLF interface to reimplement Coro::Signal gaining
327 some unknown (because I was too lazy), but certain, speedup, and also
328 making signals reliable for the first time.
329 - used new SLF interface and other optimisations to speed up async_pool
330 by a factor of two. It also doesn't rely on perl's exception mechanism
331 to exit anymore. The overhead for terminating an async_pool, coro over
332 a normal async is now very small.
333 - sped up coroutine creation/destruction by 40%.
334 - forgot to include Coro/libcoro/README in the dist for all these years.
335 - work around a freebsd pthreads bug (manual testcancel is required as
336 pthread_cond_wait isn't a cancellation point on freebsd).
337 - use new rouse functions to speed up and simplify Coro::BDB.
338 - make "prefer perl native functions" work with threaded perls.
339 - condense Coro::Debug ps output, hint at v and w flags.
340 - (libcoro) lots of minor cleanups and portability improvements.
341
3424.914 Wed Nov 19 12:54:18 CET 2008
343 - fix a disastrous bug in the readline optimisation
344 introduced in 4.801.
345
3464.913 Sat Nov 15 07:58:28 CET 2008
347 - async_pool did free a scalar value twice
348 ("Attempt to unreference...").
349
3504.912 Thu Nov 13 18:31:23 CET 2008
351 - minor cleanups.
352 - use much larger stacks on linux and perl < 5.8.8.
353 - Coro::Debug::new_unix_server did not unlink the socket
354 when destroyed.
355
3564.911 Tue Nov 11 04:26:17 CET 2008
357 - "port" to threaded perls.
358
3594.91 Mon Nov 10 05:36:38 CET 2008
360 - the ->throw exception object no longer leaks.
361 - creating a new cctx leaked a scopestack entry (memleak).
362 - new coroutines didn't get created with a zero flags field
363 (unknown impact).
364 - calling ->throw on a not-yet-started coroutine should now work
365 instead of being ignored.
366 - ->throw is now supported on Coro::State objects.
367 - clean up cctx creation code a bit.
368 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
369
3704.9 Sat Nov 8 17:45:27 CET 2008
371 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
372 - (libcoro) no longer leak threads in the experimental pthread backend,
373 also speed it up considerably.
374 - (libcoro) do not rely on makecontext passing void *'s unscathed.
375 - fix compiletime dependencies on libcoro in the Makefile.
376 - cctx_count wasn't always updated properly.
377 - Coro::State::cctx_stacksize wasn't applied correctly.
378 - new function Coro::State::cctx_max_idle.
379 - the default max number of idle C contexts is now 4.
380 - (libcoro) try harder to get _setjmp/_longjmp.
381 - (libcoro) cleanup and extend the libcoro API to officially
382 allow the creation of empty source contexts.
383 - very experimental workaround for the totally broken netbsd platform.
384 - tried to hack around openbsd bugs.
385
3864.804 Wed Nov 5 16:36:00 CET 2008
387 - Coro::Debug::new_unix_server would not create a non-blocking listening
388 socket, sometimes causing freezes.
389 - (libcoro) fix misaligned stack points for setjmp and assembly
390 methods, which can cause crashes on x86/x86_64 with a sufficiently
391 aggressive compiler.
392 - new function: Coro::Debug::new_tcp_server.
8 - move ->throw into the Coro class because it only works on coro objects. 393 - move ->throw into the Coro class because it only works on coro objects.
9 394
104.803 Mon Nov 3 17:16:12 CET 2008 3954.803 Mon Nov 3 17:16:12 CET 2008
11 - use a global asm statement to become independent of gcc otpimisations 396 - (libcoro) use a global asm statement to become independent of gcc
12 for CORO_ASM (thanks to pippijn for the idea). 397 otpimisations for CORO_ASM (thanks to pippijn for the idea).
13 - try to workaround yet another broken bsd, this time dragonfly. 398 - try to workaround yet another broken bsd, this time dragonfly.
14 399
154.802 Thu Oct 30 10:56:12 CET 2008 4004.802 Thu Oct 30 10:56:12 CET 2008
16 - support -fno-omit-frame-pointer on x86 with the assembly method. 401 - support -fno-omit-frame-pointer on x86 with the assembly method.
17 - tune 01_unblock.t tests a bit. 402 - tune 01_unblock.t tests a bit.

Diff Legend

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