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

Comparing Coro/Changes (file contents):
Revision 1.426 by root, Mon Nov 17 07:19:31 2008 UTC vs.
Revision 1.591 by root, Sun May 27 16:36:05 2012 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)). 6TODO: __GCC_HAVE_DWARF2_CFI_ASM
7 7
8TODO: testcancel 8 - use fd = -1 in mmap.
9TODO: aio 9
10TODO: Coro::Semaphore 106.08 Fri Apr 13 12:05:47 CEST 2012
11TODO: trow disturbs cleanup path inside pp_slf? 11 - be more aggressive about exiting like perl does - formerly,
12TODO: throw special-case in transfer? 12 exiting from the non-main thread would not execute END blocks.
135.0 13
146.07 Fri Nov 11 21:21:48 CET 2011
15 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
16 - update ecb.h.
17
186.06 Mon Aug 8 23:59:48 CEST 2011
19 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
20 backend, which is an order of magnitude slower. unfortunately, cygwins
21 pthread implementation isn't very complete either, so allocate the stack
22 twice just to be sure.
23 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
24 in its ithreaded perl - best recompile cygwin and use the 'w'indows
25 backend for much better performance. also disable ithreads for
26 even better performance...).
27
286.05 Thu Aug 4 21:36:36 CEST 2011
29 - blush, condvar values would not be propagated from send to recv anymore
30 (reported by Chip Salzenberg).
31 - use exponential increase for the readline buffer length in
32 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
33
346.04 Wed Aug 3 17:15:45 CEST 2011
35 - use even more efficient and more compatible condvars for
36 compatibility to AnyEvent 6.x :)
37 - more inconsequential ecb.h updates.
38
396.03 Wed Aug 3 11:41:30 CEST 2011
40 - change how Coro patches AnyEvent condvars for compatibility to
41 AnyEvent 6.x.
42 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
43 hasn't been included.
44
456.02 Wed Jul 13 04:35:19 CEST 2011
46 - "improve portability to Gentoo" - gentoo manages to put perl variables
47 in memory areas that are farther than 2gb apart, which the jit couldn't
48 handle and barfed. now it's just a bit slower on gentoo and similar
49 systems.
50
516.01 Sun Jul 3 12:31:14 CEST 2011
52 - workarounds are good, but the test for whether pthreads are used
53 was not good. thisone should be better.
54 - check differently whether gcc generates cfi instructions itself.
55
566.0 Wed Jun 29 19:43:35 CEST 2011
57 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
58 destroyed and cleaned up automatically (e.g. async { schedule }).
59 - implement a JIT compiler for part of the thread switch code,
60 which gives a 50% speed improvement on threaded perls, and
61 about 4% on non-thraeded perls (so threaded perls now finally
62 reach about half the speed of non-threaded perls).
63 - slightly modernise Coro::Intro, add section about rouse functions.
64 - avoid DEFSV and ERRSV, giving another 10% improvement
65 in thread switching.
66 - Coro::State->is_destroyed is now called is_zombie.
67 - implement a Coro->safe_cancel method that might fail, but
68 cancels in a "safer" way if it succeeds.
69 - add preliminary support for DEBUGGING perls.
70 - get rid of two hash-accesses when initialising a new Coro - this
71 speeds up coro creation by almost a factor of two.
72 - croak when a coro that is being cancelled tries to block
73 (e.g. while executing a guard block), instead of crashing or
74 deadlocking.
75 - use a more robust and also faster method to identify Coro::State
76 objects - speeds up everything a bit.
77 - implement Coro->cancel in XS for a 20% speed improvement, and to
78 be able to implement mutual cancellation.
79 - speed up context switches by a percent or two by more efficiently
80 allocating context stack entries.
81 - implement Coro->join and Coro->on_destroy in XS for a speedup and
82 a reduction in memory use.
83 - cancelling a coro while it itself is cancelling another coro is
84 now supported and working, instead of triggering an assertion.
85 - be a bit more crash-resistant when calling (buggy) on_destroy
86 callbacks (best effort).
87 - move on_destroy into the slf_frame, to allow extension slf
88 functions to have destructors.
89 - get rid if coro refcounting - simply crash in other interpreter
90 threads by nulling the pointers on clone.
91 - simplify warn/die hook handling when loading Coro - the convoluted
92 logic seems to be no longer neccessary.
93 - use libecb instead of our own home-grown gcc hacks.
94 - document alternatives to Coro::LWP. Please use them :)
95 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
96 sigaltstack. Really. wine suffers from it, erlang suffers from it,
97 and it's known since at least 2006.
98
995.372 Wed Feb 23 06:14:30 CET 2011
100 - apparently mingw doesn't provide a working gettimeofday, try to
101 work around that by relying on Time::HiRes (indirectly brought to
102 my attention by Max Maischein).
103 - fix some portability issues when Time::HiRes was used.
104
1055.371 Mon Feb 21 14:36:08 CET 2011
106 - backport to windows process emulation code again.
107
1085.37 Sat Feb 19 07:49:44 CET 2011
109 - add a big "Coro thread life cycle" section to "man Coro".
110 - try a tentative workaround against the breakage that 5.13 has
111 introduced without depreciation period. sigh.
112 - no longer use Time::HiRes if gettimeofday is available, which
113 saves quite a lot of memory.
114
1155.36 Sun Feb 13 05:33:41 CET 2011
116 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
117 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
118 method call.
119 - undocument Coro::Timer::sleep and obsolete whole module.
120 - optimise Coro::Timer::timeout memory and cpu usage.
121 - slightly updated Coro::Intro for recent changes.
122 - do not initialise PL_dirty anymore.
123
1245.25 Thu Nov 11 01:08:39 CET 2010
125 - try a different approach on netbsd - netbsd 5 finally has marginally
126 working pthreads, but still broken ucontext/sigaltstack.
127 - openbsd 4.8 finally got their act together, Coro works out of the box
128 with asm, setjmp and pthreads (no change, just informational).
129
1305.24 Sat Oct 23 11:27:12 CEST 2010
131 - port to the EV 4.0 API.
132 - work around bugs in mingw32, making strawberry perl work
133 out of the box.
134 - correctly modify Coro::AIO function prototypes
135 so that they reflect the "no optional parameters" rule.
136 - "ported" libcoro to C++.
137
1385.23 Mon May 17 18:50:42 CEST 2010
139 - be more resistant to ordering changes when initialising
140 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
141 Waldorf).
142 - document that perl 5.12 deliberately removed support for cloning.
143
1445.22 Wed Apr 14 03:55:35 CEST 2010
145 - correctly return udnef on errors in Coro::Handle::read/write
146 (testcase by Marc Mims).
147 - convert Coro::Util into a "perl compatibility wrapper" - the functions
148 are less useful now, but are drop-in replacements for existing
149 functions, listing better alternatives in the documentation. This also
150 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
151 with Coro::Util::inet_aton.
152 - do not override $Coro::idle unconditionally in Coro.pm, as other
153 modules could have provided their own idle coro already
154 (for exmaple, Coro::AnyEvent).
155 - fix Coro::Util::gethost* functions.
156 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
157
1585.21 Wed Dec 16 07:19:51 CET 2009
159 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
160 together.
161 - add some examples on how to combine other event loops with Coro in
162 Coro::AnyEvent, and how to run it (and not to block). Seems to be
163 the most common source of confusion.
164 - try to catch people naively blocking in an event callback.
165 - work around the perl filehandle bug issue in conjunction with
166 older common::sense (as indirectly pointed out by ZSystem).
167 - clarify the "not from signal handlers" section.
168
1695.2 Sun Oct 4 14:54:24 CEST 2009
170 - Coro::Storable destroyed the prototypes of the functions it wrapped.
171 - export rouse_cb and rouse_wait by default now.
172 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
173 - new method $state->swap_sv.
174 - added section on "windows process emulation" to the manpage,
175 after a not-so-fruitful (nor-friendly) "discussion" with chip
176 salzenberg (discussion implies arguments, but his only arguments
177 were ad-hominems, one wonders why he started it in the first
178 place). I hope this explains it well enough for him to understand,
179 and maybe well enough for others to understand.
180 - use common::sense everywhere now.
181 - idle callbacks are no longer supported, use idle coros instead.
182 - print a thread listing when a deadlock is detected.
183
1845.17 Sat Aug 22 23:09:31 CEST 2009
185 - work around a bug in the perl debugger causing crashes
186 when running under the debugger by marking _pool_handler as nodebug.
187 - speed up Coro::async considerably.
188 - try some hacks to get netbsd to work "more often" - their broken
189 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
190 - convert Coro to AE 5.0 API.
191
1925.162 Tue Jul 28 04:04:03 CEST 2009
193 - perl 5.8.2 is now minimum requirement.
194 - skip t/19_handle.t on broken windows perls.
195
1965.161 Wed Jul 22 04:47:38 CEST 2009
197 - Coro::AnyEvent::poll could have a different prototype when EV was
198 used as backend (analyzed by Tatsuhiko Miyagawa).
199 - Coro::AnyEvent errornously initialised the event loop when loaded,
200 not on demand.
201 - try to workaround rare */t/01_unblock.t failures.
202
2035.16 Tue Jul 21 01:44:37 CEST 2009
204 - Coro::AnyEvent failed to hook into the event loop
205 when no threads had been readied between detecting
206 the event loop and actually running it.
207 - considerably speed up Coro::Select by taking avdantage
208 of AnyEvent > 4.8 and some other optimisations.
209 - implement paragraph readline mode in Coro::Handle
210 (based on patches by Zsbán Ambrus).
211 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
212 and analyzed by Yasuhiro MATSUMOTO).
213 - clarified libcoro license and copyright.
214 - someone stole my EXTRA_META!!!
215 - implement Coro::Select::patch_pp_sselect and it's brother,
216 for hardcode select overriding.
217
2185.151 Mon Jul 6 05:41:57 CEST 2009
219 - backport to windows process emulation code again (patch by
220 Yasuhiro MATSUMOTO).
221 - slightly update Coro::MakeMaker.
222
2235.15 Tue Jun 30 10:28:06 CEST 2009
224 - deprecate Coro::Socket, document how to get ipv6 support via
225 AnyEvent::Socket instead.
226 - implement signal->wait ($cb) interface, similar to semaphores.
227 - work around SvOK not supporting getmagic, so we have to getmagic
228 to test for undef :/ (reported by Matthias Waldorf).
229 - load Coro::AnyEvent in all modules using AnyEvent.
230 - work around perl corrupting our internal data structures,
231 reported by Tokuhiro Matsuno.
232 - enable per-coroutine real and cpu time gathering
233 (Coro::State::enable_times).
234
2355.14 Wed Jun 24 01:37:48 CEST 2009
236 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
237 as this seems to have been a difficult concept (poll, sleep, idle,
238 idle_upto).
239 - add Coro::AnyEvent::readable/writable functions.
240 - clarify Coro::EV/Event/AnyEvent manpages.
241 - free per-thread global scalars in the thread calling ->cancel, to
242 avoid crashes when $_, $@ etc., are magical but some of those
243 had already been freed.
244 - "unexperimentalise" the callback interface for Coro::Semaphore.
245 - speed up ready queue management/context switching by using a linked
246 list instead of an array (~5-10%).
247 - implement "watch" command in Coro::Debug shells.
248 - for fun, implement time-slicing as an example in the manpage.
249 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
250 Coro::Event, use the normal AnyEvent handling instead of dieing
251 (the same is true for Coro::Handle).
252 - properly document Coro::EV::timed_io_once.
253 - avoid unneccessary ->cancel calls in Coro::Handle.
254 - maybe make it work on mingw32 with win32 backend
255 (based on patch by Yasuhiro Matsumoto).
256
2575.132 Fri May 29 09:00:39 CEST 2009
258 - do not keep a reference to the argument itself in
259 Coro::Semaphore::guard, as it could change later.
260 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
261 should support a prepare callback.
262
2635.131 Mon Mar 16 23:20:37 CET 2009
264 - implement and document Coro->suspend, Coro->resume.
265 - fix Coro::Select implementation to not (often) close
266 the passed file descriptors (testcase provided by pippijn).
267
2685.13 Mon Dec 15 21:51:42 CET 2008
269 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
270 - don't set diehook to C<undef> but instead to NULL, to avoid
271 spurious warnings.
272 - fix a lot of bugs in Coro::SemaphoreSet.
273 - Coro::SemaphoreSet will less often create a semaphore needlessly.
274 - add Coro::SemaphoreSet::count and wait methods.
275 - take advantage of the new Guard module.
276 - deprecate Coro::guard.
277 - try to fix the dreaded 01_unblock tests once more. I hate it when
278 testsuites need more fixing than the code they are supposed to test.
279 - croak in more cases when a required callback isn't resolvable.
280 - fix some minor issues in Coro::State->call/eval.
281 - use current coroutine context instead of a temporary one
282 when temporarily switching to another coroutine.
283 - do not call C-level on_destroy handlers during global destruction,
284 to avoid needless segfaults.
285
2865.12 Sun Dec 7 13:30:38 CET 2008
287 - add default config for MirOS, which seems to be a bug-to-bug
288 compatible fork of openbsd ("world domination by releasing
289 openbsd cvs before the openbsd folks do it" or so :).
290 - free_padlist did destroy the names pad, not good, but didn't
291 seem to bother perl - this could fix issues such as eval ""
292 inside a function called from multiple coroutines.
293 - use a different method to detect destruction time.
294 - be more careful when freeing padlists just before global
295 destruction.
296 - fixed and expanded the call/cc example.
297 - renamed _terminate to _coro_run.
298 - new method Coro::Channel->shutdown.
299 - try pthreads on openbsd <4.4 (broken sigaltstack, will
300 pthreads fare better?).
301 - be less picky about destroying "a" running coroutine.
302
3035.11 Tue Nov 25 21:49:05 CET 2008
304 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
305 working code with failed assertions, introducing more bugs than
306 it fixes, requiring elaborate workarounds :(
307
3085.1 Mon Nov 24 08:54:59 CET 2008
309 - wrote a small introductory tutorial - Coro::Intro.
310 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
311 - Coro::Select did errornously dup the file descriptors
312 and didn't work with all AnyEvent backends.
313 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
314 LWP data transfers.
315 - disassociate c contexts from coro objects - this is agruably more
316 correct, but mostly allows sharing of cctxs between coro and state
317 objects, for added memory savings and speed increases.
318 - bumped $Coro::POOL_RSS up to 32kb by default.
319 - no longer set the optype to OP_CUSTOM, as B::* understandably
320 doesn't like this very much (and we *are* a type of entersub).
321 - implement state cloning, just to prove that call/cc can be done.
322 - automatically load Coro::AnyEvent in Coro::Handle.
323 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
324 do not have this method (and don't need it either).
325 - speed up generic anyevent methods in Coro::Handle by using rouse
326 callbacks.
327 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
328 others. It also makes the debugger happier, as you can trace
329 through the idle threads now.
330 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
331 - eg/event was pretty broken.
332 - better 5.8.6 compatibility.
333
3345.0 Thu Nov 20 10:35:05 CET 2008
14 - NEW ARCHITECTURE: use the latest 4.x version if you experience 335 - NEW ARCHITECTURE: use the latest 4.x version if you experience
15 stability issues. 336 stability issues.
16 - bump API version to 7 - all dependents must be recompiled. 337 - bump API version to 7 - all dependents must be recompiled.
338 - removed timed_* functions - they were not being used anyways
339 and should be replaced by a more generic mechanism -
340 and were annoying to support anyways :)
341 - removed SemaphoreSet's waiter method - use sem method instead.
17 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 342 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
18 - async_pool did free a scalar value twice 343 - async_pool did free a scalar value twice
19 ("Attempt to unreference..."). 344 ("Attempt to unreference...").
20 - fix a longstanding bug where calling terminate on a coro that 345 - fix a longstanding bug where calling terminate on a coro that
21 was waiting for a semaphore that was just becoming available 346 was waiting for a semaphore that was just becoming available
22 would cause a deadlock (semaphore would get into a state where 347 would cause a deadlock (semaphore would get into a state where
23 it was available but waiters were waiting(. 348 it was available but waiters were still blocked).
24 - avoid deadlock in Coro::Channel if maxsize == 1 (Richard Hundt). 349 - calling throw on a coroutine that is waiting for a semaphore will
350 no longer make it acquire the semaphore (and thus leak a count).
25 - perl's process emulation is now not even theoretically supported 351 - perl's process emulation is now not even theoretically supported
26 anymore. 352 anymore.
353 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
354 conversion of callback-style to blocking-style.
355 - new methods $coro->schedule_to and ->cede_to, to specifically
356 schedule or cede to a specific coroutine.
357 - new function Coro::Semaphore::wait.
358 - use named constants in Coro::Channel (Richard Hundt).
27 - directly patch the entersub opcode calling SLF functions (cede, 359 - directly patch the entersub opcode calling SLF functions (cede,
28 transfer and so on). this does speed up context switching, but 360 transfer and so on). this does speed up context switching, but
29 more importanly, it frees us from the hardcoded behaviour of 361 more importanly, it frees us from the hardcoded behaviour of
30 entersub, so we might actually be able to return something from 362 entersub, so we might actually be able to return something from
31 those functions and atcually create new ones. 363 those functions and atcually create new ones.
… …
34 as "threaded" perls are running at half speed anyways). 366 as "threaded" perls are running at half speed anyways).
35 - implement execute_slf (schedule-like-function) interface that makes 367 - implement execute_slf (schedule-like-function) interface that makes
36 it possible to implement schedule-like-functions in XS. 368 it possible to implement schedule-like-functions in XS.
37 - use new SLF interface to massively speed up Coro::EV by roughly a 369 - use new SLF interface to massively speed up Coro::EV by roughly a
38 factor of two. 370 factor of two.
39 - removed timed_down and timed_guard functions - they were not being
40 used anyways and should be replaced by a more generic mechanism -
41 and were annoying to support anyways :)
42 - used new SLF interface to massively speed up Coro::Semaphore by a 371 - used new SLF interface to massively speed up Coro::Semaphore by a
43 factor of three. 372 factor of three.
44 - used new SLF interface to speed up Coro::AIO by roughly a factor of 373 - used new SLF interface to speed up Coro::AIO by roughly a factor of
45 four and reduce its memory usage considerably. 374 four and reduce its memory usage considerably.
375 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
376 for a nice speedup and vastly more correct behaviour. Also implement
377 a new method "sem" to get at the underlying semaphore object.
378 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
379 (in comparison) 20-40% speedup.
380 - used new SLF interface to reimplement Coro::Signal gaining
381 some unknown (because I was too lazy), but certain, speedup, and also
382 making signals reliable for the first time.
383 - used new SLF interface and other optimisations to speed up async_pool
384 by a factor of two. It also doesn't rely on perl's exception mechanism
385 to exit anymore. The overhead for terminating an async_pool, coro over
386 a normal async is now very small.
387 - sped up coroutine creation/destruction by 40%.
46 - forgot to include Coro/libcoro/README in the dist for all these years. 388 - forgot to include Coro/libcoro/README in the dist for all these years.
47 - work around a freebsd pthreads bug (manual testcancel is required as 389 - work around a freebsd pthreads bug (manual testcancel is required as
48 pthread_cond_wait isn't a cancellation point on freebsd). 390 pthread_cond_wait isn't a cancellation point on freebsd).
391 - use new rouse functions to speed up and simplify Coro::BDB.
392 - make "prefer perl native functions" work with threaded perls.
393 - condense Coro::Debug ps output, hint at v and w flags.
394 - (libcoro) lots of minor cleanups and portability improvements.
395
3964.914 Wed Nov 19 12:54:18 CET 2008
397 - fix a disastrous bug in the readline optimisation
398 introduced in 4.801.
49 399
504.913 Sat Nov 15 07:58:28 CET 2008 4004.913 Sat Nov 15 07:58:28 CET 2008
51 - async_pool did free a scalar value twice 401 - async_pool did free a scalar value twice
52 ("Attempt to unreference..."). 402 ("Attempt to unreference...").
53 403

Diff Legend

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