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

Comparing Coro/Changes (file contents):
Revision 1.444 by root, Wed Nov 19 16:01:32 2008 UTC vs.
Revision 1.600 by root, Wed Oct 31 16:35:30 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: async_pool to SLF 8TODO: add peerhost/peerport from IO::Socket::INET etc.
9TODO: _cancel to xs 9
105.0 10 - bad things could happen in Coro::EV::timed_once if the arguments
11 were wrong. Also support magic values as timed_io_once args.
12 - recommend AnyEvent 7+ or EV 4+, also require EV
13 version 4 or newer for Coro::EV.
14
156.10 Tue Oct 9 01:14:27 CEST 2012
16 - updated ecb.h, it had a typo that caused it to not compile on many
17 big endian systems (reported by many people).
18 - disable memory fences in ecb.h to improve portability.
19
206.09 Sat Oct 6 23:25:02 CEST 2012
21 - Coro::EV I/O watchers were not interruptible by exceptions
22 (Coro::State::throw) (testcase by sten).
23 - ->throw now puts threads into the ready queue, as this seems to
24 be expected by existing code, and code that doesn't cope with spurious
25 wakeups needs fixing anyway.
26 - use fd -1 in mmap.
27 - cast I32 to int in error message printf.
28 - warn about broken so-called "hardened" kernels.
29
306.08 Fri Apr 13 12:05:47 CEST 2012
31 - be more aggressive about exiting like perl does - formerly,
32 exiting from the non-main thread would not execute END blocks.
33
346.07 Fri Nov 11 21:21:48 CET 2011
35 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
36 - update ecb.h.
37
386.06 Mon Aug 8 23:59:48 CEST 2011
39 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
40 backend, which is an order of magnitude slower. unfortunately, cygwins
41 pthread implementation isn't very complete either, so allocate the stack
42 twice just to be sure.
43 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
44 in its ithreaded perl - best recompile cygwin and use the 'w'indows
45 backend for much better performance. also disable ithreads for
46 even better performance...).
47
486.05 Thu Aug 4 21:36:36 CEST 2011
49 - blush, condvar values would not be propagated from send to recv anymore
50 (reported by Chip Salzenberg).
51 - use exponential increase for the readline buffer length in
52 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
53
546.04 Wed Aug 3 17:15:45 CEST 2011
55 - use even more efficient and more compatible condvars for
56 compatibility to AnyEvent 6.x :)
57 - more inconsequential ecb.h updates.
58
596.03 Wed Aug 3 11:41:30 CEST 2011
60 - change how Coro patches AnyEvent condvars for compatibility to
61 AnyEvent 6.x.
62 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
63 hasn't been included.
64
656.02 Wed Jul 13 04:35:19 CEST 2011
66 - "improve portability to Gentoo" - gentoo manages to put perl variables
67 in memory areas that are farther than 2gb apart, which the jit couldn't
68 handle and barfed. now it's just a bit slower on gentoo and similar
69 systems.
70
716.01 Sun Jul 3 12:31:14 CEST 2011
72 - workarounds are good, but the test for whether pthreads are used
73 was not good. thisone should be better.
74 - check differently whether gcc generates cfi instructions itself.
75
766.0 Wed Jun 29 19:43:35 CEST 2011
77 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
78 destroyed and cleaned up automatically (e.g. async { schedule }).
79 - implement a JIT compiler for part of the thread switch code,
80 which gives a 50% speed improvement on threaded perls, and
81 about 4% on non-thraeded perls (so threaded perls now finally
82 reach about half the speed of non-threaded perls).
83 - slightly modernise Coro::Intro, add section about rouse functions.
84 - avoid DEFSV and ERRSV, giving another 10% improvement
85 in thread switching.
86 - Coro::State->is_destroyed is now called is_zombie.
87 - implement a Coro->safe_cancel method that might fail, but
88 cancels in a "safer" way if it succeeds.
89 - add preliminary support for DEBUGGING perls.
90 - get rid of two hash-accesses when initialising a new Coro - this
91 speeds up coro creation by almost a factor of two.
92 - croak when a coro that is being cancelled tries to block
93 (e.g. while executing a guard block), instead of crashing or
94 deadlocking.
95 - use a more robust and also faster method to identify Coro::State
96 objects - speeds up everything a bit.
97 - implement Coro->cancel in XS for a 20% speed improvement, and to
98 be able to implement mutual cancellation.
99 - speed up context switches by a percent or two by more efficiently
100 allocating context stack entries.
101 - implement Coro->join and Coro->on_destroy in XS for a speedup and
102 a reduction in memory use.
103 - cancelling a coro while it itself is cancelling another coro is
104 now supported and working, instead of triggering an assertion.
105 - be a bit more crash-resistant when calling (buggy) on_destroy
106 callbacks (best effort).
107 - move on_destroy into the slf_frame, to allow extension slf
108 functions to have destructors.
109 - get rid if coro refcounting - simply crash in other interpreter
110 threads by nulling the pointers on clone.
111 - simplify warn/die hook handling when loading Coro - the convoluted
112 logic seems to be no longer neccessary.
113 - use libecb instead of our own home-grown gcc hacks.
114 - document alternatives to Coro::LWP. Please use them :)
115 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
116 sigaltstack. Really. wine suffers from it, erlang suffers from it,
117 and it's known since at least 2006.
118
1195.372 Wed Feb 23 06:14:30 CET 2011
120 - apparently mingw doesn't provide a working gettimeofday, try to
121 work around that by relying on Time::HiRes (indirectly brought to
122 my attention by Max Maischein).
123 - fix some portability issues when Time::HiRes was used.
124
1255.371 Mon Feb 21 14:36:08 CET 2011
126 - backport to windows process emulation code again.
127
1285.37 Sat Feb 19 07:49:44 CET 2011
129 - add a big "Coro thread life cycle" section to "man Coro".
130 - try a tentative workaround against the breakage that 5.13 has
131 introduced without depreciation period. sigh.
132 - no longer use Time::HiRes if gettimeofday is available, which
133 saves quite a lot of memory.
134
1355.36 Sun Feb 13 05:33:41 CET 2011
136 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
137 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
138 method call.
139 - undocument Coro::Timer::sleep and obsolete whole module.
140 - optimise Coro::Timer::timeout memory and cpu usage.
141 - slightly updated Coro::Intro for recent changes.
142 - do not initialise PL_dirty anymore.
143
1445.25 Thu Nov 11 01:08:39 CET 2010
145 - try a different approach on netbsd - netbsd 5 finally has marginally
146 working pthreads, but still broken ucontext/sigaltstack.
147 - openbsd 4.8 finally got their act together, Coro works out of the box
148 with asm, setjmp and pthreads (no change, just informational).
149
1505.24 Sat Oct 23 11:27:12 CEST 2010
151 - port to the EV 4.0 API.
152 - work around bugs in mingw32, making strawberry perl work
153 out of the box.
154 - correctly modify Coro::AIO function prototypes
155 so that they reflect the "no optional parameters" rule.
156 - "ported" libcoro to C++.
157
1585.23 Mon May 17 18:50:42 CEST 2010
159 - be more resistant to ordering changes when initialising
160 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
161 Waldorf).
162 - document that perl 5.12 deliberately removed support for cloning.
163
1645.22 Wed Apr 14 03:55:35 CEST 2010
165 - correctly return udnef on errors in Coro::Handle::read/write
166 (testcase by Marc Mims).
167 - convert Coro::Util into a "perl compatibility wrapper" - the functions
168 are less useful now, but are drop-in replacements for existing
169 functions, listing better alternatives in the documentation. This also
170 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
171 with Coro::Util::inet_aton.
172 - do not override $Coro::idle unconditionally in Coro.pm, as other
173 modules could have provided their own idle coro already
174 (for exmaple, Coro::AnyEvent).
175 - fix Coro::Util::gethost* functions.
176 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
177
1785.21 Wed Dec 16 07:19:51 CET 2009
179 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
180 together.
181 - add some examples on how to combine other event loops with Coro in
182 Coro::AnyEvent, and how to run it (and not to block). Seems to be
183 the most common source of confusion.
184 - try to catch people naively blocking in an event callback.
185 - work around the perl filehandle bug issue in conjunction with
186 older common::sense (as indirectly pointed out by ZSystem).
187 - clarify the "not from signal handlers" section.
188
1895.2 Sun Oct 4 14:54:24 CEST 2009
190 - Coro::Storable destroyed the prototypes of the functions it wrapped.
191 - export rouse_cb and rouse_wait by default now.
192 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
193 - new method $state->swap_sv.
194 - added section on "windows process emulation" to the manpage,
195 after a not-so-fruitful (nor-friendly) "discussion" with chip
196 salzenberg (discussion implies arguments, but his only arguments
197 were ad-hominems, one wonders why he started it in the first
198 place). I hope this explains it well enough for him to understand,
199 and maybe well enough for others to understand.
200 - use common::sense everywhere now.
201 - idle callbacks are no longer supported, use idle coros instead.
202 - print a thread listing when a deadlock is detected.
203
2045.17 Sat Aug 22 23:09:31 CEST 2009
205 - work around a bug in the perl debugger causing crashes
206 when running under the debugger by marking _pool_handler as nodebug.
207 - speed up Coro::async considerably.
208 - try some hacks to get netbsd to work "more often" - their broken
209 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
210 - convert Coro to AE 5.0 API.
211
2125.162 Tue Jul 28 04:04:03 CEST 2009
213 - perl 5.8.2 is now minimum requirement.
214 - skip t/19_handle.t on broken windows perls.
215
2165.161 Wed Jul 22 04:47:38 CEST 2009
217 - Coro::AnyEvent::poll could have a different prototype when EV was
218 used as backend (analyzed by Tatsuhiko Miyagawa).
219 - Coro::AnyEvent errornously initialised the event loop when loaded,
220 not on demand.
221 - try to workaround rare */t/01_unblock.t failures.
222
2235.16 Tue Jul 21 01:44:37 CEST 2009
224 - Coro::AnyEvent failed to hook into the event loop
225 when no threads had been readied between detecting
226 the event loop and actually running it.
227 - considerably speed up Coro::Select by taking avdantage
228 of AnyEvent > 4.8 and some other optimisations.
229 - implement paragraph readline mode in Coro::Handle
230 (based on patches by Zsbán Ambrus).
231 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
232 and analyzed by Yasuhiro MATSUMOTO).
233 - clarified libcoro license and copyright.
234 - someone stole my EXTRA_META!!!
235 - implement Coro::Select::patch_pp_sselect and it's brother,
236 for hardcode select overriding.
237
2385.151 Mon Jul 6 05:41:57 CEST 2009
239 - backport to windows process emulation code again (patch by
240 Yasuhiro MATSUMOTO).
241 - slightly update Coro::MakeMaker.
242
2435.15 Tue Jun 30 10:28:06 CEST 2009
244 - deprecate Coro::Socket, document how to get ipv6 support via
245 AnyEvent::Socket instead.
246 - implement signal->wait ($cb) interface, similar to semaphores.
247 - work around SvOK not supporting getmagic, so we have to getmagic
248 to test for undef :/ (reported by Matthias Waldorf).
249 - load Coro::AnyEvent in all modules using AnyEvent.
250 - work around perl corrupting our internal data structures,
251 reported by Tokuhiro Matsuno.
252 - enable per-coroutine real and cpu time gathering
253 (Coro::State::enable_times).
254
2555.14 Wed Jun 24 01:37:48 CEST 2009
256 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
257 as this seems to have been a difficult concept (poll, sleep, idle,
258 idle_upto).
259 - add Coro::AnyEvent::readable/writable functions.
260 - clarify Coro::EV/Event/AnyEvent manpages.
261 - free per-thread global scalars in the thread calling ->cancel, to
262 avoid crashes when $_, $@ etc., are magical but some of those
263 had already been freed.
264 - "unexperimentalise" the callback interface for Coro::Semaphore.
265 - speed up ready queue management/context switching by using a linked
266 list instead of an array (~5-10%).
267 - implement "watch" command in Coro::Debug shells.
268 - for fun, implement time-slicing as an example in the manpage.
269 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
270 Coro::Event, use the normal AnyEvent handling instead of dieing
271 (the same is true for Coro::Handle).
272 - properly document Coro::EV::timed_io_once.
273 - avoid unneccessary ->cancel calls in Coro::Handle.
274 - maybe make it work on mingw32 with win32 backend
275 (based on patch by Yasuhiro Matsumoto).
276
2775.132 Fri May 29 09:00:39 CEST 2009
278 - do not keep a reference to the argument itself in
279 Coro::Semaphore::guard, as it could change later.
280 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
281 should support a prepare callback.
282
2835.131 Mon Mar 16 23:20:37 CET 2009
284 - implement and document Coro->suspend, Coro->resume.
285 - fix Coro::Select implementation to not (often) close
286 the passed file descriptors (testcase provided by pippijn).
287
2885.13 Mon Dec 15 21:51:42 CET 2008
289 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
290 - don't set diehook to C<undef> but instead to NULL, to avoid
291 spurious warnings.
292 - fix a lot of bugs in Coro::SemaphoreSet.
293 - Coro::SemaphoreSet will less often create a semaphore needlessly.
294 - add Coro::SemaphoreSet::count and wait methods.
295 - take advantage of the new Guard module.
296 - deprecate Coro::guard.
297 - try to fix the dreaded 01_unblock tests once more. I hate it when
298 testsuites need more fixing than the code they are supposed to test.
299 - croak in more cases when a required callback isn't resolvable.
300 - fix some minor issues in Coro::State->call/eval.
301 - use current coroutine context instead of a temporary one
302 when temporarily switching to another coroutine.
303 - do not call C-level on_destroy handlers during global destruction,
304 to avoid needless segfaults.
305
3065.12 Sun Dec 7 13:30:38 CET 2008
307 - add default config for MirOS, which seems to be a bug-to-bug
308 compatible fork of openbsd ("world domination by releasing
309 openbsd cvs before the openbsd folks do it" or so :).
310 - free_padlist did destroy the names pad, not good, but didn't
311 seem to bother perl - this could fix issues such as eval ""
312 inside a function called from multiple coroutines.
313 - use a different method to detect destruction time.
314 - be more careful when freeing padlists just before global
315 destruction.
316 - fixed and expanded the call/cc example.
317 - renamed _terminate to _coro_run.
318 - new method Coro::Channel->shutdown.
319 - try pthreads on openbsd <4.4 (broken sigaltstack, will
320 pthreads fare better?).
321 - be less picky about destroying "a" running coroutine.
322
3235.11 Tue Nov 25 21:49:05 CET 2008
324 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
325 working code with failed assertions, introducing more bugs than
326 it fixes, requiring elaborate workarounds :(
327
3285.1 Mon Nov 24 08:54:59 CET 2008
329 - wrote a small introductory tutorial - Coro::Intro.
330 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
331 - Coro::Select did errornously dup the file descriptors
332 and didn't work with all AnyEvent backends.
333 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
334 LWP data transfers.
335 - disassociate c contexts from coro objects - this is agruably more
336 correct, but mostly allows sharing of cctxs between coro and state
337 objects, for added memory savings and speed increases.
338 - bumped $Coro::POOL_RSS up to 32kb by default.
339 - no longer set the optype to OP_CUSTOM, as B::* understandably
340 doesn't like this very much (and we *are* a type of entersub).
341 - implement state cloning, just to prove that call/cc can be done.
342 - automatically load Coro::AnyEvent in Coro::Handle.
343 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
344 do not have this method (and don't need it either).
345 - speed up generic anyevent methods in Coro::Handle by using rouse
346 callbacks.
347 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
348 others. It also makes the debugger happier, as you can trace
349 through the idle threads now.
350 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
351 - eg/event was pretty broken.
352 - better 5.8.6 compatibility.
353
3545.0 Thu Nov 20 10:35:05 CET 2008
11 - NEW ARCHITECTURE: use the latest 4.x version if you experience 355 - NEW ARCHITECTURE: use the latest 4.x version if you experience
12 stability issues. 356 stability issues.
13 - bump API version to 7 - all dependents must be recompiled. 357 - bump API version to 7 - all dependents must be recompiled.
14 - removed timed_* functions - they were not being used anyways 358 - removed timed_* functions - they were not being used anyways
15 and should be replaced by a more generic mechanism - 359 and should be replaced by a more generic mechanism -
16 and were annoying to support anyways :) 360 and were annoying to support anyways :)
17 - removed SemaphoreSet's waiter method - use sem instead. 361 - removed SemaphoreSet's waiter method - use sem method instead.
18 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 362 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
19 - async_pool did free a scalar value twice 363 - async_pool did free a scalar value twice
20 ("Attempt to unreference..."). 364 ("Attempt to unreference...").
21 - fix a disastrous bug in the readline optimisation
22 introduced in 4.801.
23 - fix a longstanding bug where calling terminate on a coro that 365 - fix a longstanding bug where calling terminate on a coro that
24 was waiting for a semaphore that was just becoming available 366 was waiting for a semaphore that was just becoming available
25 would cause a deadlock (semaphore would get into a state where 367 would cause a deadlock (semaphore would get into a state where
26 it was available but waiters were still blocked). 368 it was available but waiters were still blocked).
27 - calling throw on a coroutine that is waiting for a semaphore will 369 - calling throw on a coroutine that is waiting for a semaphore will
28 no longer make it acquire the semaphore (and thus leak a count). 370 no longer make it acquire the semaphore (and thus leak a count).
29 - perl's process emulation is now not even theoretically supported 371 - perl's process emulation is now not even theoretically supported
30 anymore. 372 anymore.
31 - new functions Coro::rouse_cb and Coro::rouse_wait. 373 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
374 conversion of callback-style to blocking-style.
375 - new methods $coro->schedule_to and ->cede_to, to specifically
376 schedule or cede to a specific coroutine.
32 - new function Coro::Semaphore::wait. 377 - new function Coro::Semaphore::wait.
33 - use named constants in Coro::Channel (Richard Hundt). 378 - use named constants in Coro::Channel (Richard Hundt).
34 - directly patch the entersub opcode calling SLF functions (cede, 379 - directly patch the entersub opcode calling SLF functions (cede,
35 transfer and so on). this does speed up context switching, but 380 transfer and so on). this does speed up context switching, but
36 more importanly, it frees us from the hardcoded behaviour of 381 more importanly, it frees us from the hardcoded behaviour of
… …
53 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate 398 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
54 (in comparison) 20-40% speedup. 399 (in comparison) 20-40% speedup.
55 - used new SLF interface to reimplement Coro::Signal gaining 400 - used new SLF interface to reimplement Coro::Signal gaining
56 some unknown (because I was too lazy), but certain, speedup, and also 401 some unknown (because I was too lazy), but certain, speedup, and also
57 making signals reliable for the first time. 402 making signals reliable for the first time.
403 - used new SLF interface and other optimisations to speed up async_pool
404 by a factor of two. It also doesn't rely on perl's exception mechanism
405 to exit anymore. The overhead for terminating an async_pool, coro over
406 a normal async is now very small.
58 - sped up coroutine creation/destruction by 40%. 407 - sped up coroutine creation/destruction by 40%.
59 - forgot to include Coro/libcoro/README in the dist for all these years. 408 - forgot to include Coro/libcoro/README in the dist for all these years.
60 - work around a freebsd pthreads bug (manual testcancel is required as 409 - work around a freebsd pthreads bug (manual testcancel is required as
61 pthread_cond_wait isn't a cancellation point on freebsd). 410 pthread_cond_wait isn't a cancellation point on freebsd).
62 - use new rouse functions to speed up and simplify Coro::BDB. 411 - use new rouse functions to speed up and simplify Coro::BDB.
63 - make "prefer perl native functions" work with threaded perls. 412 - make "prefer perl native functions" work with threaded perls.
64 - condense Coro::Debug ps output, hint at v and w flags. 413 - condense Coro::Debug ps output, hint at v and w flags.
65 - (libcoro) lots of minor cleanups and portability improvements. 414 - (libcoro) lots of minor cleanups and portability improvements.
415
4164.914 Wed Nov 19 12:54:18 CET 2008
417 - fix a disastrous bug in the readline optimisation
418 introduced in 4.801.
66 419
674.913 Sat Nov 15 07:58:28 CET 2008 4204.913 Sat Nov 15 07:58:28 CET 2008
68 - async_pool did free a scalar value twice 421 - async_pool did free a scalar value twice
69 ("Attempt to unreference..."). 422 ("Attempt to unreference...").
70 423

Diff Legend

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