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

Comparing Coro/Changes (file contents):
Revision 1.422 by root, Sun Nov 16 11:12:57 2008 UTC vs.
Revision 1.579 by root, Tue Jul 26 11:47:08 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
8TODO: testcancel 7 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
9TODO: aio 8 hasn't been included.
10TODO: Coro::Semaphore 9
11TODO: trow disturbs cleanup path inside pp_slf? 106.02 Wed Jul 13 04:35:19 CEST 2011
12TODO: throw special-case in transfer? 11 - "improve portability to Gentoo" - gentoo manages to put perl variables
135.0 12 in memory areas that are farther than 2gb apart, which the jit couldn't
14 - NEW ARCHITECTURE: use 4.912 for the stable version. 13 handle and barfed. now it's just a bit slower on gentoo and similar
14 systems.
15
166.01 Sun Jul 3 12:31:14 CEST 2011
17 - workarounds are good, but the test for whether pthreads are used
18 was not good. thisone should be better.
19 - check differently whether gcc generates cfi instructions itself.
20
216.0 Wed Jun 29 19:43:35 CEST 2011
22 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
23 destroyed and cleaned up automatically (e.g. async { schedule }).
24 - implement a JIT compiler for part of the thread switch code,
25 which gives a 50% speed improvement on threaded perls, and
26 about 4% on non-thraeded perls (so threaded perls now finally
27 reach about half the speed of non-threaded perls).
28 - slightly modernise Coro::Intro, add section about rouse functions.
29 - avoid DEFSV and ERRSV, giving another 10% improvement
30 in thread switching.
31 - Coro::State->is_destroyed is now called is_zombie.
32 - implement a Coro->safe_cancel method that might fail, but
33 cancels in a "safer" way if it succeeds.
34 - add preliminary support for DEBUGGING perls.
35 - get rid of two hash-accesses when initialising a new Coro - this
36 speeds up coro creation by almost a factor of two.
37 - croak when a coro that is being cancelled tries to block
38 (e.g. while executing a guard block), instead of crashing or
39 deadlocking.
40 - use a more robust and also faster method to identify Coro::State
41 objects - speeds up everything a bit.
42 - implement Coro->cancel in XS for a 20% speed improvement, and to
43 be able to implement mutual cancellation.
44 - speed up context switches by a percent or two by more efficiently
45 allocating context stack entries.
46 - implement Coro->join and Coro->on_destroy in XS for a speedup and
47 a reduction in memory use.
48 - cancelling a coro while it itself is cancelling another coro is
49 now supported and working, instead of triggering an assertion.
50 - be a bit more crash-resistant when calling (buggy) on_destroy
51 callbacks (best effort).
52 - move on_destroy into the slf_frame, to allow extension slf
53 functions to have destructors.
54 - get rid if coro refcounting - simply crash in other interpreter
55 threads by nulling the pointers on clone.
56 - simplify warn/die hook handling when loading Coro - the convoluted
57 logic seems to be no longer neccessary.
58 - use libecb instead of our own home-grown gcc hacks.
59 - document alternatives to Coro::LWP. Please use them :)
60 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
61 sigaltstack. Really. wine suffers from it, erlang suffers from it,
62 and it's known since at least 2006.
63
645.372 Wed Feb 23 06:14:30 CET 2011
65 - apparently mingw doesn't provide a working gettimeofday, try to
66 work around that by relying on Time::HiRes (indirectly brought to
67 my attention by Max Maischein).
68 - fix some portability issues when Time::HiRes was used.
69
705.371 Mon Feb 21 14:36:08 CET 2011
71 - backport to windows process emulation code again.
72
735.37 Sat Feb 19 07:49:44 CET 2011
74 - add a big "Coro thread life cycle" section to "man Coro".
75 - try a tentative workaround against the breakage that 5.13 has
76 introduced without depreciation period. sigh.
77 - no longer use Time::HiRes if gettimeofday is available, which
78 saves quite a lot of memory.
79
805.36 Sun Feb 13 05:33:41 CET 2011
81 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
82 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
83 method call.
84 - undocument Coro::Timer::sleep and obsolete whole module.
85 - optimise Coro::Timer::timeout memory and cpu usage.
86 - slightly updated Coro::Intro for recent changes.
87 - do not initialise PL_dirty anymore.
88
895.25 Thu Nov 11 01:08:39 CET 2010
90 - try a different approach on netbsd - netbsd 5 finally has marginally
91 working pthreads, but still broken ucontext/sigaltstack.
92 - openbsd 4.8 finally got their act together, Coro works out of the box
93 with asm, setjmp and pthreads (no change, just informational).
94
955.24 Sat Oct 23 11:27:12 CEST 2010
96 - port to the EV 4.0 API.
97 - work around bugs in mingw32, making strawberry perl work
98 out of the box.
99 - correctly modify Coro::AIO function prototypes
100 so that they reflect the "no optional parameters" rule.
101 - "ported" libcoro to C++.
102
1035.23 Mon May 17 18:50:42 CEST 2010
104 - be more resistant to ordering changes when initialising
105 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
106 Waldorf).
107 - document that perl 5.12 deliberately removed support for cloning.
108
1095.22 Wed Apr 14 03:55:35 CEST 2010
110 - correctly return udnef on errors in Coro::Handle::read/write
111 (testcase by Marc Mims).
112 - convert Coro::Util into a "perl compatibility wrapper" - the functions
113 are less useful now, but are drop-in replacements for existing
114 functions, listing better alternatives in the documentation. This also
115 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
116 with Coro::Util::inet_aton.
117 - do not override $Coro::idle unconditionally in Coro.pm, as other
118 modules could have provided their own idle coro already
119 (for exmaple, Coro::AnyEvent).
120 - fix Coro::Util::gethost* functions.
121 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
122
1235.21 Wed Dec 16 07:19:51 CET 2009
124 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
125 together.
126 - add some examples on how to combine other event loops with Coro in
127 Coro::AnyEvent, and how to run it (and not to block). Seems to be
128 the most common source of confusion.
129 - try to catch people naively blocking in an event callback.
130 - work around the perl filehandle bug issue in conjunction with
131 older common::sense (as indirectly pointed out by ZSystem).
132 - clarify the "not from signal handlers" section.
133
1345.2 Sun Oct 4 14:54:24 CEST 2009
135 - Coro::Storable destroyed the prototypes of the functions it wrapped.
136 - export rouse_cb and rouse_wait by default now.
137 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
138 - new method $state->swap_sv.
139 - added section on "windows process emulation" to the manpage,
140 after a not-so-fruitful (nor-friendly) "discussion" with chip
141 salzenberg (discussion implies arguments, but his only arguments
142 were ad-hominems, one wonders why he started it in the first
143 place). I hope this explains it well enough for him to understand,
144 and maybe well enough for others to understand.
145 - use common::sense everywhere now.
146 - idle callbacks are no longer supported, use idle coros instead.
147 - print a thread listing when a deadlock is detected.
148
1495.17 Sat Aug 22 23:09:31 CEST 2009
150 - work around a bug in the perl debugger causing crashes
151 when running under the debugger by marking _pool_handler as nodebug.
152 - speed up Coro::async considerably.
153 - try some hacks to get netbsd to work "more often" - their broken
154 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
155 - convert Coro to AE 5.0 API.
156
1575.162 Tue Jul 28 04:04:03 CEST 2009
158 - perl 5.8.2 is now minimum requirement.
159 - skip t/19_handle.t on broken windows perls.
160
1615.161 Wed Jul 22 04:47:38 CEST 2009
162 - Coro::AnyEvent::poll could have a different prototype when EV was
163 used as backend (analyzed by Tatsuhiko Miyagawa).
164 - Coro::AnyEvent errornously initialised the event loop when loaded,
165 not on demand.
166 - try to workaround rare */t/01_unblock.t failures.
167
1685.16 Tue Jul 21 01:44:37 CEST 2009
169 - Coro::AnyEvent failed to hook into the event loop
170 when no threads had been readied between detecting
171 the event loop and actually running it.
172 - considerably speed up Coro::Select by taking avdantage
173 of AnyEvent > 4.8 and some other optimisations.
174 - implement paragraph readline mode in Coro::Handle
175 (based on patches by Zsbán Ambrus).
176 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
177 and analyzed by Yasuhiro MATSUMOTO).
178 - clarified libcoro license and copyright.
179 - someone stole my EXTRA_META!!!
180 - implement Coro::Select::patch_pp_sselect and it's brother,
181 for hardcode select overriding.
182
1835.151 Mon Jul 6 05:41:57 CEST 2009
184 - backport to windows process emulation code again (patch by
185 Yasuhiro MATSUMOTO).
186 - slightly update Coro::MakeMaker.
187
1885.15 Tue Jun 30 10:28:06 CEST 2009
189 - deprecate Coro::Socket, document how to get ipv6 support via
190 AnyEvent::Socket instead.
191 - implement signal->wait ($cb) interface, similar to semaphores.
192 - work around SvOK not supporting getmagic, so we have to getmagic
193 to test for undef :/ (reported by Matthias Waldorf).
194 - load Coro::AnyEvent in all modules using AnyEvent.
195 - work around perl corrupting our internal data structures,
196 reported by Tokuhiro Matsuno.
197 - enable per-coroutine real and cpu time gathering
198 (Coro::State::enable_times).
199
2005.14 Wed Jun 24 01:37:48 CEST 2009
201 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
202 as this seems to have been a difficult concept (poll, sleep, idle,
203 idle_upto).
204 - add Coro::AnyEvent::readable/writable functions.
205 - clarify Coro::EV/Event/AnyEvent manpages.
206 - free per-thread global scalars in the thread calling ->cancel, to
207 avoid crashes when $_, $@ etc., are magical but some of those
208 had already been freed.
209 - "unexperimentalise" the callback interface for Coro::Semaphore.
210 - speed up ready queue management/context switching by using a linked
211 list instead of an array (~5-10%).
212 - implement "watch" command in Coro::Debug shells.
213 - for fun, implement time-slicing as an example in the manpage.
214 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
215 Coro::Event, use the normal AnyEvent handling instead of dieing
216 (the same is true for Coro::Handle).
217 - properly document Coro::EV::timed_io_once.
218 - avoid unneccessary ->cancel calls in Coro::Handle.
219 - maybe make it work on mingw32 with win32 backend
220 (based on patch by Yasuhiro Matsumoto).
221
2225.132 Fri May 29 09:00:39 CEST 2009
223 - do not keep a reference to the argument itself in
224 Coro::Semaphore::guard, as it could change later.
225 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
226 should support a prepare callback.
227
2285.131 Mon Mar 16 23:20:37 CET 2009
229 - implement and document Coro->suspend, Coro->resume.
230 - fix Coro::Select implementation to not (often) close
231 the passed file descriptors (testcase provided by pippijn).
232
2335.13 Mon Dec 15 21:51:42 CET 2008
234 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
235 - don't set diehook to C<undef> but instead to NULL, to avoid
236 spurious warnings.
237 - fix a lot of bugs in Coro::SemaphoreSet.
238 - Coro::SemaphoreSet will less often create a semaphore needlessly.
239 - add Coro::SemaphoreSet::count and wait methods.
240 - take advantage of the new Guard module.
241 - deprecate Coro::guard.
242 - try to fix the dreaded 01_unblock tests once more. I hate it when
243 testsuites need more fixing than the code they are supposed to test.
244 - croak in more cases when a required callback isn't resolvable.
245 - fix some minor issues in Coro::State->call/eval.
246 - use current coroutine context instead of a temporary one
247 when temporarily switching to another coroutine.
248 - do not call C-level on_destroy handlers during global destruction,
249 to avoid needless segfaults.
250
2515.12 Sun Dec 7 13:30:38 CET 2008
252 - add default config for MirOS, which seems to be a bug-to-bug
253 compatible fork of openbsd ("world domination by releasing
254 openbsd cvs before the openbsd folks do it" or so :).
255 - free_padlist did destroy the names pad, not good, but didn't
256 seem to bother perl - this could fix issues such as eval ""
257 inside a function called from multiple coroutines.
258 - use a different method to detect destruction time.
259 - be more careful when freeing padlists just before global
260 destruction.
261 - fixed and expanded the call/cc example.
262 - renamed _terminate to _coro_run.
263 - new method Coro::Channel->shutdown.
264 - try pthreads on openbsd <4.4 (broken sigaltstack, will
265 pthreads fare better?).
266 - be less picky about destroying "a" running coroutine.
267
2685.11 Tue Nov 25 21:49:05 CET 2008
269 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
270 working code with failed assertions, introducing more bugs than
271 it fixes, requiring elaborate workarounds :(
272
2735.1 Mon Nov 24 08:54:59 CET 2008
274 - wrote a small introductory tutorial - Coro::Intro.
275 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
276 - Coro::Select did errornously dup the file descriptors
277 and didn't work with all AnyEvent backends.
278 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
279 LWP data transfers.
280 - disassociate c contexts from coro objects - this is agruably more
281 correct, but mostly allows sharing of cctxs between coro and state
282 objects, for added memory savings and speed increases.
283 - bumped $Coro::POOL_RSS up to 32kb by default.
284 - no longer set the optype to OP_CUSTOM, as B::* understandably
285 doesn't like this very much (and we *are* a type of entersub).
286 - implement state cloning, just to prove that call/cc can be done.
287 - automatically load Coro::AnyEvent in Coro::Handle.
288 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
289 do not have this method (and don't need it either).
290 - speed up generic anyevent methods in Coro::Handle by using rouse
291 callbacks.
292 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
293 others. It also makes the debugger happier, as you can trace
294 through the idle threads now.
295 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
296 - eg/event was pretty broken.
297 - better 5.8.6 compatibility.
298
2995.0 Thu Nov 20 10:35:05 CET 2008
300 - NEW ARCHITECTURE: use the latest 4.x version if you experience
301 stability issues.
15 - bump API version to 7 - all dependents must be recompiled. 302 - bump API version to 7 - all dependents must be recompiled.
303 - removed timed_* functions - they were not being used anyways
304 and should be replaced by a more generic mechanism -
305 and were annoying to support anyways :)
306 - removed SemaphoreSet's waiter method - use sem method instead.
16 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 307 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
17 - async_pool did free a scalar value twice 308 - async_pool did free a scalar value twice
18 ("Attempt to unreference..."). 309 ("Attempt to unreference...").
19 - fix a longstanding bug where calling terminate on a coro that 310 - fix a longstanding bug where calling terminate on a coro that
20 was waiting for a semaphore that was just becoming available 311 was waiting for a semaphore that was just becoming available
21 would cause a deadlock (semaphore would get into a state where 312 would cause a deadlock (semaphore would get into a state where
22 it was available but waiters were waiting(. 313 it was available but waiters were still blocked).
314 - calling throw on a coroutine that is waiting for a semaphore will
315 no longer make it acquire the semaphore (and thus leak a count).
23 - perl's process emulation is now not even theoretically supported 316 - perl's process emulation is now not even theoretically supported
24 anymore. 317 anymore.
318 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
319 conversion of callback-style to blocking-style.
320 - new methods $coro->schedule_to and ->cede_to, to specifically
321 schedule or cede to a specific coroutine.
322 - new function Coro::Semaphore::wait.
323 - use named constants in Coro::Channel (Richard Hundt).
25 - directly patch the entersub opcode calling SLF functions (cede, 324 - directly patch the entersub opcode calling SLF functions (cede,
26 transfer and so on). this does speed up context switching, but 325 transfer and so on). this does speed up context switching, but
27 more importanly, it frees us from the hardcoded behaviour of 326 more importanly, it frees us from the hardcoded behaviour of
28 entersub, so we might actually be able to return something from 327 entersub, so we might actually be able to return something from
29 those functions and atcually create new ones. 328 those functions and atcually create new ones.
30 - take advantage of __builtin_frame_address on gcc. 329 - take advantage of __builtin_frame_address on gcc.
31 - expose THX in coroapi (not sure whether this was a wise decision, 330 - expose THX in coroapi (not sure whether this was a wise decision,
32 as "threaded" perls are running at half speed anyways). 331 as "threaded" perls are running at half speed anyways).
33 - implement execute_slf (schedule-like-function) interface that makes 332 - implement execute_slf (schedule-like-function) interface that makes
34 it possible to implement schedule-like-functions in XS. 333 it possible to implement schedule-like-functions in XS.
35 - use new SLF interface to massively speed up Coro::EV by roughly a factor 334 - use new SLF interface to massively speed up Coro::EV by roughly a
36 of two. 335 factor of two.
37 - removed timed_down and timed_guard functions - they were not being used
38 anyways and should be replaced by a more generic mechanism - and were
39 annoying to support anyways :
40 - used new SLF interface to massively speed up Coro::Semaphore by a factor 336 - used new SLF interface to massively speed up Coro::Semaphore by a
41 of three. 337 factor of three.
338 - used new SLF interface to speed up Coro::AIO by roughly a factor of
339 four and reduce its memory usage considerably.
340 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
341 for a nice speedup and vastly more correct behaviour. Also implement
342 a new method "sem" to get at the underlying semaphore object.
343 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
344 (in comparison) 20-40% speedup.
345 - used new SLF interface to reimplement Coro::Signal gaining
346 some unknown (because I was too lazy), but certain, speedup, and also
347 making signals reliable for the first time.
348 - used new SLF interface and other optimisations to speed up async_pool
349 by a factor of two. It also doesn't rely on perl's exception mechanism
350 to exit anymore. The overhead for terminating an async_pool, coro over
351 a normal async is now very small.
352 - sped up coroutine creation/destruction by 40%.
42 - forgot to include Coro/libcoro/README in the dist for all these years. 353 - forgot to include Coro/libcoro/README in the dist for all these years.
43 - work around a freebsd pthreads bug (manual testcancel is required as 354 - work around a freebsd pthreads bug (manual testcancel is required as
44 pthread_cond_wait isn't a cancellation point on freebsd). 355 pthread_cond_wait isn't a cancellation point on freebsd).
356 - use new rouse functions to speed up and simplify Coro::BDB.
357 - make "prefer perl native functions" work with threaded perls.
358 - condense Coro::Debug ps output, hint at v and w flags.
359 - (libcoro) lots of minor cleanups and portability improvements.
360
3614.914 Wed Nov 19 12:54:18 CET 2008
362 - fix a disastrous bug in the readline optimisation
363 introduced in 4.801.
45 364
464.913 Sat Nov 15 07:58:28 CET 2008 3654.913 Sat Nov 15 07:58:28 CET 2008
47 - async_pool did free a scalar value twice 366 - async_pool did free a scalar value twice
48 ("Attempt to unreference..."). 367 ("Attempt to unreference...").
49 368

Diff Legend

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