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

Comparing Coro/Changes (file contents):
Revision 1.459 by root, Sat Nov 22 15:48:43 2008 UTC vs.
Revision 1.617 by root, Thu May 9 05:40:13 2013 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: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)).
7 6
8 TODO: Coro/Intro 76.31 Thu May 9 07:39:48 CEST 2013
8 - Coro::AIO requests would crash if the thread was ready'd
9 while the request was ongoing.
10
116.29 Wed May 8 02:55:18 CEST 2013
12 - when an on_destroy handler destructs the coro currently being
13 destructed a perl scalar could be accessed after being freed,
14 likely causing a crash.
15
166.28 Wed Mar 6 06:58:02 CET 2013
17 - clean remnants of existing __DIE__ and __WARN__ handlers so
18 they lose their magic and will not cause segfaults later
19 (testcase by Andrey Sagulin).
20 - improved Coro::State documentation a bit.
21 - Coro::Debug::command now flushes the output.
22 - add hack detection code for x32 abi, because the braindead slugs
23 who designed that made it look exactly like x86_64 without
24 providing proper compile time symbols to test for it. as a result,
25 this detection cannot work reliably.
26 - valgrind stack registering was broken.
27 - do not rely on Time::HiRes anymore in Coro::Debug.
28
296.23 Fri Dec 7 23:36:37 CET 2012
30 - use experimental fiber implementation on native windows
31 perls.
32 - use sizeof (void *) as multiplication factor for stack sizes,
33 to accomodate the totally braindamaged microsoft 64 bit "os".
34 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
35 activeperl 5.16 crashes when PerlIO_define_layer is called due
36 to some bug in the perl dll, strawberry perl at least passes
37 the testsuite.
38 - implement Coro::Handle->peeraddr/host/port, for slightly
39 improved compatibility with LWP.
40 - implement 5.17 compatibility by almost blindly applying a
41 good-looking patch by Father Chrysostomos.
42 - move stack management functions into libcoro 3.
43 - libcoro version 3 "released".
44 - support magic values as timed_io_once args.
45 - recommend AnyEvent 7+ or EV 4+, also require EV
46 version 4 or newer for Coro::EV.
47
486.10 Tue Oct 9 01:14:27 CEST 2012
49 - updated ecb.h, it had a typo that caused it to not compile on many
50 big endian systems (reported by many people).
51 - disable memory fences in ecb.h to improve portability.
52
536.09 Sat Oct 6 23:25:02 CEST 2012
54 - Coro::EV I/O watchers were not interruptible by exceptions
55 (Coro::State::throw) (testcase by sten).
56 - ->throw now puts threads into the ready queue, as this seems to
57 be expected by existing code, and code that doesn't cope with spurious
58 wakeups needs fixing anyway.
59 - use fd -1 in mmap.
60 - cast I32 to int in error message printf.
61 - warn about broken so-called "hardened" kernels.
62
636.08 Fri Apr 13 12:05:47 CEST 2012
64 - be more aggressive about exiting like perl does - formerly,
65 exiting from the non-main thread would not execute END blocks.
66
676.07 Fri Nov 11 21:21:48 CET 2011
68 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
69 - update ecb.h.
70
716.06 Mon Aug 8 23:59:48 CEST 2011
72 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
73 backend, which is an order of magnitude slower. unfortunately, cygwins
74 pthread implementation isn't very complete either, so allocate the stack
75 twice just to be sure.
76 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
77 in its ithreaded perl - best recompile cygwin and use the 'w'indows
78 backend for much better performance. also disable ithreads for
79 even better performance...).
80
816.05 Thu Aug 4 21:36:36 CEST 2011
82 - blush, condvar values would not be propagated from send to recv anymore
83 (reported by Chip Salzenberg).
84 - use exponential increase for the readline buffer length in
85 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
86
876.04 Wed Aug 3 17:15:45 CEST 2011
88 - use even more efficient and more compatible condvars for
89 compatibility to AnyEvent 6.x :)
90 - more inconsequential ecb.h updates.
91
926.03 Wed Aug 3 11:41:30 CEST 2011
93 - change how Coro patches AnyEvent condvars for compatibility to
94 AnyEvent 6.x.
95 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
96 hasn't been included.
97
986.02 Wed Jul 13 04:35:19 CEST 2011
99 - "improve portability to Gentoo" - gentoo manages to put perl variables
100 in memory areas that are farther than 2gb apart, which the jit couldn't
101 handle and barfed. now it's just a bit slower on gentoo and similar
102 systems.
103
1046.01 Sun Jul 3 12:31:14 CEST 2011
105 - workarounds are good, but the test for whether pthreads are used
106 was not good. thisone should be better.
107 - check differently whether gcc generates cfi instructions itself.
108
1096.0 Wed Jun 29 19:43:35 CEST 2011
110 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
111 destroyed and cleaned up automatically (e.g. async { schedule }).
112 - implement a JIT compiler for part of the thread switch code,
113 which gives a 50% speed improvement on threaded perls, and
114 about 4% on non-thraeded perls (so threaded perls now finally
115 reach about half the speed of non-threaded perls).
116 - slightly modernise Coro::Intro, add section about rouse functions.
117 - avoid DEFSV and ERRSV, giving another 10% improvement
118 in thread switching.
119 - Coro::State->is_destroyed is now called is_zombie.
120 - implement a Coro->safe_cancel method that might fail, but
121 cancels in a "safer" way if it succeeds.
122 - add preliminary support for DEBUGGING perls.
123 - get rid of two hash-accesses when initialising a new Coro - this
124 speeds up coro creation by almost a factor of two.
125 - croak when a coro that is being cancelled tries to block
126 (e.g. while executing a guard block), instead of crashing or
127 deadlocking.
128 - use a more robust and also faster method to identify Coro::State
129 objects - speeds up everything a bit.
130 - implement Coro->cancel in XS for a 20% speed improvement, and to
131 be able to implement mutual cancellation.
132 - speed up context switches by a percent or two by more efficiently
133 allocating context stack entries.
134 - implement Coro->join and Coro->on_destroy in XS for a speedup and
135 a reduction in memory use.
136 - cancelling a coro while it itself is cancelling another coro is
137 now supported and working, instead of triggering an assertion.
138 - be a bit more crash-resistant when calling (buggy) on_destroy
139 callbacks (best effort).
140 - move on_destroy into the slf_frame, to allow extension slf
141 functions to have destructors.
142 - get rid if coro refcounting - simply crash in other interpreter
143 threads by nulling the pointers on clone.
144 - simplify warn/die hook handling when loading Coro - the convoluted
145 logic seems to be no longer neccessary.
146 - use libecb instead of our own home-grown gcc hacks.
147 - document alternatives to Coro::LWP. Please use them :)
148 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
149 sigaltstack. Really. wine suffers from it, erlang suffers from it,
150 and it's known since at least 2006.
151
1525.372 Wed Feb 23 06:14:30 CET 2011
153 - apparently mingw doesn't provide a working gettimeofday, try to
154 work around that by relying on Time::HiRes (indirectly brought to
155 my attention by Max Maischein).
156 - fix some portability issues when Time::HiRes was used.
157
1585.371 Mon Feb 21 14:36:08 CET 2011
159 - backport to windows process emulation code again.
160
1615.37 Sat Feb 19 07:49:44 CET 2011
162 - add a big "Coro thread life cycle" section to "man Coro".
163 - try a tentative workaround against the breakage that 5.13 has
164 introduced without depreciation period. sigh.
165 - no longer use Time::HiRes if gettimeofday is available, which
166 saves quite a lot of memory.
167
1685.36 Sun Feb 13 05:33:41 CET 2011
169 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
170 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
171 method call.
172 - undocument Coro::Timer::sleep and obsolete whole module.
173 - optimise Coro::Timer::timeout memory and cpu usage.
174 - slightly updated Coro::Intro for recent changes.
175 - do not initialise PL_dirty anymore.
176
1775.25 Thu Nov 11 01:08:39 CET 2010
178 - try a different approach on netbsd - netbsd 5 finally has marginally
179 working pthreads, but still broken ucontext/sigaltstack.
180 - openbsd 4.8 finally got their act together, Coro works out of the box
181 with asm, setjmp and pthreads (no change, just informational).
182
1835.24 Sat Oct 23 11:27:12 CEST 2010
184 - port to the EV 4.0 API.
185 - work around bugs in mingw32, making strawberry perl work
186 out of the box.
187 - correctly modify Coro::AIO function prototypes
188 so that they reflect the "no optional parameters" rule.
189 - "ported" libcoro to C++.
190
1915.23 Mon May 17 18:50:42 CEST 2010
192 - be more resistant to ordering changes when initialising
193 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
194 Waldorf).
195 - document that perl 5.12 deliberately removed support for cloning.
196
1975.22 Wed Apr 14 03:55:35 CEST 2010
198 - correctly return udnef on errors in Coro::Handle::read/write
199 (testcase by Marc Mims).
200 - convert Coro::Util into a "perl compatibility wrapper" - the functions
201 are less useful now, but are drop-in replacements for existing
202 functions, listing better alternatives in the documentation. This also
203 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
204 with Coro::Util::inet_aton.
205 - do not override $Coro::idle unconditionally in Coro.pm, as other
206 modules could have provided their own idle coro already
207 (for exmaple, Coro::AnyEvent).
208 - fix Coro::Util::gethost* functions.
209 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
210
2115.21 Wed Dec 16 07:19:51 CET 2009
212 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
213 together.
214 - add some examples on how to combine other event loops with Coro in
215 Coro::AnyEvent, and how to run it (and not to block). Seems to be
216 the most common source of confusion.
217 - try to catch people naively blocking in an event callback.
218 - work around the perl filehandle bug issue in conjunction with
219 older common::sense (as indirectly pointed out by ZSystem).
220 - clarify the "not from signal handlers" section.
221
2225.2 Sun Oct 4 14:54:24 CEST 2009
223 - Coro::Storable destroyed the prototypes of the functions it wrapped.
224 - export rouse_cb and rouse_wait by default now.
225 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
226 - new method $state->swap_sv.
227 - added section on "windows process emulation" to the manpage,
228 after a not-so-fruitful (nor-friendly) "discussion" with chip
229 salzenberg (discussion implies arguments, but his only arguments
230 were ad-hominems, one wonders why he started it in the first
231 place). I hope this explains it well enough for him to understand,
232 and maybe well enough for others to understand.
233 - use common::sense everywhere now.
234 - idle callbacks are no longer supported, use idle coros instead.
235 - print a thread listing when a deadlock is detected.
236
2375.17 Sat Aug 22 23:09:31 CEST 2009
238 - work around a bug in the perl debugger causing crashes
239 when running under the debugger by marking _pool_handler as nodebug.
240 - speed up Coro::async considerably.
241 - try some hacks to get netbsd to work "more often" - their broken
242 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
243 - convert Coro to AE 5.0 API.
244
2455.162 Tue Jul 28 04:04:03 CEST 2009
246 - perl 5.8.2 is now minimum requirement.
247 - skip t/19_handle.t on broken windows perls.
248
2495.161 Wed Jul 22 04:47:38 CEST 2009
250 - Coro::AnyEvent::poll could have a different prototype when EV was
251 used as backend (analyzed by Tatsuhiko Miyagawa).
252 - Coro::AnyEvent errornously initialised the event loop when loaded,
253 not on demand.
254 - try to workaround rare */t/01_unblock.t failures.
255
2565.16 Tue Jul 21 01:44:37 CEST 2009
257 - Coro::AnyEvent failed to hook into the event loop
258 when no threads had been readied between detecting
259 the event loop and actually running it.
260 - considerably speed up Coro::Select by taking avdantage
261 of AnyEvent > 4.8 and some other optimisations.
262 - implement paragraph readline mode in Coro::Handle
263 (based on patches by Zsbán Ambrus).
264 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
265 and analyzed by Yasuhiro MATSUMOTO).
266 - clarified libcoro license and copyright.
267 - someone stole my EXTRA_META!!!
268 - implement Coro::Select::patch_pp_sselect and it's brother,
269 for hardcode select overriding.
270
2715.151 Mon Jul 6 05:41:57 CEST 2009
272 - backport to windows process emulation code again (patch by
273 Yasuhiro MATSUMOTO).
274 - slightly update Coro::MakeMaker.
275
2765.15 Tue Jun 30 10:28:06 CEST 2009
277 - deprecate Coro::Socket, document how to get ipv6 support via
278 AnyEvent::Socket instead.
279 - implement signal->wait ($cb) interface, similar to semaphores.
280 - work around SvOK not supporting getmagic, so we have to getmagic
281 to test for undef :/ (reported by Matthias Waldorf).
282 - load Coro::AnyEvent in all modules using AnyEvent.
283 - work around perl corrupting our internal data structures,
284 reported by Tokuhiro Matsuno.
285 - enable per-coroutine real and cpu time gathering
286 (Coro::State::enable_times).
287
2885.14 Wed Jun 24 01:37:48 CEST 2009
289 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
290 as this seems to have been a difficult concept (poll, sleep, idle,
291 idle_upto).
292 - add Coro::AnyEvent::readable/writable functions.
293 - clarify Coro::EV/Event/AnyEvent manpages.
294 - free per-thread global scalars in the thread calling ->cancel, to
295 avoid crashes when $_, $@ etc., are magical but some of those
296 had already been freed.
297 - "unexperimentalise" the callback interface for Coro::Semaphore.
298 - speed up ready queue management/context switching by using a linked
299 list instead of an array (~5-10%).
300 - implement "watch" command in Coro::Debug shells.
301 - for fun, implement time-slicing as an example in the manpage.
302 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
303 Coro::Event, use the normal AnyEvent handling instead of dieing
304 (the same is true for Coro::Handle).
305 - properly document Coro::EV::timed_io_once.
306 - avoid unneccessary ->cancel calls in Coro::Handle.
307 - maybe make it work on mingw32 with win32 backend
308 (based on patch by Yasuhiro Matsumoto).
309
3105.132 Fri May 29 09:00:39 CEST 2009
311 - do not keep a reference to the argument itself in
312 Coro::Semaphore::guard, as it could change later.
313 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
314 should support a prepare callback.
315
3165.131 Mon Mar 16 23:20:37 CET 2009
317 - implement and document Coro->suspend, Coro->resume.
318 - fix Coro::Select implementation to not (often) close
319 the passed file descriptors (testcase provided by pippijn).
320
3215.13 Mon Dec 15 21:51:42 CET 2008
322 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
323 - don't set diehook to C<undef> but instead to NULL, to avoid
324 spurious warnings.
325 - fix a lot of bugs in Coro::SemaphoreSet.
326 - Coro::SemaphoreSet will less often create a semaphore needlessly.
327 - add Coro::SemaphoreSet::count and wait methods.
328 - take advantage of the new Guard module.
329 - deprecate Coro::guard.
330 - try to fix the dreaded 01_unblock tests once more. I hate it when
331 testsuites need more fixing than the code they are supposed to test.
332 - croak in more cases when a required callback isn't resolvable.
333 - fix some minor issues in Coro::State->call/eval.
334 - use current coroutine context instead of a temporary one
335 when temporarily switching to another coroutine.
336 - do not call C-level on_destroy handlers during global destruction,
337 to avoid needless segfaults.
338
3395.12 Sun Dec 7 13:30:38 CET 2008
340 - add default config for MirOS, which seems to be a bug-to-bug
341 compatible fork of openbsd ("world domination by releasing
342 openbsd cvs before the openbsd folks do it" or so :).
343 - free_padlist did destroy the names pad, not good, but didn't
344 seem to bother perl - this could fix issues such as eval ""
345 inside a function called from multiple coroutines.
346 - use a different method to detect destruction time.
347 - be more careful when freeing padlists just before global
348 destruction.
349 - fixed and expanded the call/cc example.
350 - renamed _terminate to _coro_run.
351 - new method Coro::Channel->shutdown.
352 - try pthreads on openbsd <4.4 (broken sigaltstack, will
353 pthreads fare better?).
354 - be less picky about destroying "a" running coroutine.
355
3565.11 Tue Nov 25 21:49:05 CET 2008
357 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
358 working code with failed assertions, introducing more bugs than
359 it fixes, requiring elaborate workarounds :(
360
3615.1 Mon Nov 24 08:54:59 CET 2008
362 - wrote a small introductory tutorial - Coro::Intro.
9 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. 363 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
10 - Coro::Select did errornously dup the file descriptors 364 - Coro::Select did errornously dup the file descriptors
11 and didn't work with all AnyEvent backends. 365 and didn't work with all AnyEvent backends.
366 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
367 LWP data transfers.
12 - disassociate c contexts from coro objects - this is agruably more 368 - disassociate c contexts from coro objects - this is agruably more
13 correct, but mostly allows sharing of cctxs between coro and state 369 correct, but mostly allows sharing of cctxs between coro and state
14 objects, for added memory savings and speed increases. 370 objects, for added memory savings and speed increases.
15 - bumped $Coro::POOL_RSS up to 32kb by default. 371 - bumped $Coro::POOL_RSS up to 32kb by default.
16 - no longer set the optype to OP_CUSTOM, as B::* understandably 372 - no longer set the optype to OP_CUSTOM, as B::* understandably
17 doesn't like this very much (and we *are* a type of entersub). 373 doesn't like this very much (and we *are* a type of entersub).
18 - implement state cloning, just to prove that call/cc can be done. 374 - implement state cloning, just to prove that call/cc can be done.
375 - automatically load Coro::AnyEvent in Coro::Handle.
376 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
377 do not have this method (and don't need it either).
378 - speed up generic anyevent methods in Coro::Handle by using rouse
379 callbacks.
380 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
381 others. It also makes the debugger happier, as you can trace
382 through the idle threads now.
383 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
384 - eg/event was pretty broken.
19 - better 5.8.6 compatibility. 385 - better 5.8.6 compatibility.
20 386
215.0 Thu Nov 20 10:35:05 CET 2008 3875.0 Thu Nov 20 10:35:05 CET 2008
22 - NEW ARCHITECTURE: use the latest 4.x version if you experience 388 - NEW ARCHITECTURE: use the latest 4.x version if you experience
23 stability issues. 389 stability issues.

Diff Legend

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