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

Comparing Coro/Changes (file contents):
Revision 1.458 by root, Sat Nov 22 05:13:34 2008 UTC vs.
Revision 1.611 by root, Fri Dec 21 04:48:16 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: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)).
7 6
7 - improved Coro::State documentation a bit.
8 - Coro::Debug::command now flushes the output.
9 - add hack detection code for x32 abi, because the braindead slugs
10 who designed that made it look exactly like x86_64 without
11 providing proper compile time symbols to test for it. as a result,
12 this detection cannot work reliably.
13 - valgrind stack registering was broken.
14
156.23 Fri Dec 7 23:36:37 CET 2012
16 - use experimental fiber implementation on native windows
17 perls.
18 - use sizeof (void *) as multiplication factor for stack sizes,
19 to accomodate the totally braindamaged microsoft 64 bit "os".
20 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
21 activeperl 5.16 crashes when PerlIO_define_layer is called due
22 to some bug in the perl dll, strawberry perl at least passes
23 the testsuite.
24 - implement Coro::Handle->peeraddr/host/port, for slightly
25 improved compatibility with LWP.
26 - implement 5.17 compatibility by almost blindly applying a
27 good-looking patch by Father Chrysostomos.
28 - move stack management functions into libcoro 3.
29 - libcoro version 3 "released".
30 - support magic values as timed_io_once args.
31 - recommend AnyEvent 7+ or EV 4+, also require EV
32 version 4 or newer for Coro::EV.
33
346.10 Tue Oct 9 01:14:27 CEST 2012
35 - updated ecb.h, it had a typo that caused it to not compile on many
36 big endian systems (reported by many people).
37 - disable memory fences in ecb.h to improve portability.
38
396.09 Sat Oct 6 23:25:02 CEST 2012
40 - Coro::EV I/O watchers were not interruptible by exceptions
41 (Coro::State::throw) (testcase by sten).
42 - ->throw now puts threads into the ready queue, as this seems to
43 be expected by existing code, and code that doesn't cope with spurious
44 wakeups needs fixing anyway.
45 - use fd -1 in mmap.
46 - cast I32 to int in error message printf.
47 - warn about broken so-called "hardened" kernels.
48
496.08 Fri Apr 13 12:05:47 CEST 2012
50 - be more aggressive about exiting like perl does - formerly,
51 exiting from the non-main thread would not execute END blocks.
52
536.07 Fri Nov 11 21:21:48 CET 2011
54 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
55 - update ecb.h.
56
576.06 Mon Aug 8 23:59:48 CEST 2011
58 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
59 backend, which is an order of magnitude slower. unfortunately, cygwins
60 pthread implementation isn't very complete either, so allocate the stack
61 twice just to be sure.
62 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
63 in its ithreaded perl - best recompile cygwin and use the 'w'indows
64 backend for much better performance. also disable ithreads for
65 even better performance...).
66
676.05 Thu Aug 4 21:36:36 CEST 2011
68 - blush, condvar values would not be propagated from send to recv anymore
69 (reported by Chip Salzenberg).
70 - use exponential increase for the readline buffer length in
71 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
72
736.04 Wed Aug 3 17:15:45 CEST 2011
74 - use even more efficient and more compatible condvars for
75 compatibility to AnyEvent 6.x :)
76 - more inconsequential ecb.h updates.
77
786.03 Wed Aug 3 11:41:30 CEST 2011
79 - change how Coro patches AnyEvent condvars for compatibility to
80 AnyEvent 6.x.
81 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
82 hasn't been included.
83
846.02 Wed Jul 13 04:35:19 CEST 2011
85 - "improve portability to Gentoo" - gentoo manages to put perl variables
86 in memory areas that are farther than 2gb apart, which the jit couldn't
87 handle and barfed. now it's just a bit slower on gentoo and similar
88 systems.
89
906.01 Sun Jul 3 12:31:14 CEST 2011
91 - workarounds are good, but the test for whether pthreads are used
92 was not good. thisone should be better.
93 - check differently whether gcc generates cfi instructions itself.
94
956.0 Wed Jun 29 19:43:35 CEST 2011
96 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
97 destroyed and cleaned up automatically (e.g. async { schedule }).
98 - implement a JIT compiler for part of the thread switch code,
99 which gives a 50% speed improvement on threaded perls, and
100 about 4% on non-thraeded perls (so threaded perls now finally
101 reach about half the speed of non-threaded perls).
102 - slightly modernise Coro::Intro, add section about rouse functions.
103 - avoid DEFSV and ERRSV, giving another 10% improvement
104 in thread switching.
105 - Coro::State->is_destroyed is now called is_zombie.
106 - implement a Coro->safe_cancel method that might fail, but
107 cancels in a "safer" way if it succeeds.
108 - add preliminary support for DEBUGGING perls.
109 - get rid of two hash-accesses when initialising a new Coro - this
110 speeds up coro creation by almost a factor of two.
111 - croak when a coro that is being cancelled tries to block
112 (e.g. while executing a guard block), instead of crashing or
113 deadlocking.
114 - use a more robust and also faster method to identify Coro::State
115 objects - speeds up everything a bit.
116 - implement Coro->cancel in XS for a 20% speed improvement, and to
117 be able to implement mutual cancellation.
118 - speed up context switches by a percent or two by more efficiently
119 allocating context stack entries.
120 - implement Coro->join and Coro->on_destroy in XS for a speedup and
121 a reduction in memory use.
122 - cancelling a coro while it itself is cancelling another coro is
123 now supported and working, instead of triggering an assertion.
124 - be a bit more crash-resistant when calling (buggy) on_destroy
125 callbacks (best effort).
126 - move on_destroy into the slf_frame, to allow extension slf
127 functions to have destructors.
128 - get rid if coro refcounting - simply crash in other interpreter
129 threads by nulling the pointers on clone.
130 - simplify warn/die hook handling when loading Coro - the convoluted
131 logic seems to be no longer neccessary.
132 - use libecb instead of our own home-grown gcc hacks.
133 - document alternatives to Coro::LWP. Please use them :)
134 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
135 sigaltstack. Really. wine suffers from it, erlang suffers from it,
136 and it's known since at least 2006.
137
1385.372 Wed Feb 23 06:14:30 CET 2011
139 - apparently mingw doesn't provide a working gettimeofday, try to
140 work around that by relying on Time::HiRes (indirectly brought to
141 my attention by Max Maischein).
142 - fix some portability issues when Time::HiRes was used.
143
1445.371 Mon Feb 21 14:36:08 CET 2011
145 - backport to windows process emulation code again.
146
1475.37 Sat Feb 19 07:49:44 CET 2011
148 - add a big "Coro thread life cycle" section to "man Coro".
149 - try a tentative workaround against the breakage that 5.13 has
150 introduced without depreciation period. sigh.
151 - no longer use Time::HiRes if gettimeofday is available, which
152 saves quite a lot of memory.
153
1545.36 Sun Feb 13 05:33:41 CET 2011
155 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
156 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
157 method call.
158 - undocument Coro::Timer::sleep and obsolete whole module.
159 - optimise Coro::Timer::timeout memory and cpu usage.
160 - slightly updated Coro::Intro for recent changes.
161 - do not initialise PL_dirty anymore.
162
1635.25 Thu Nov 11 01:08:39 CET 2010
164 - try a different approach on netbsd - netbsd 5 finally has marginally
165 working pthreads, but still broken ucontext/sigaltstack.
166 - openbsd 4.8 finally got their act together, Coro works out of the box
167 with asm, setjmp and pthreads (no change, just informational).
168
1695.24 Sat Oct 23 11:27:12 CEST 2010
170 - port to the EV 4.0 API.
171 - work around bugs in mingw32, making strawberry perl work
172 out of the box.
173 - correctly modify Coro::AIO function prototypes
174 so that they reflect the "no optional parameters" rule.
175 - "ported" libcoro to C++.
176
1775.23 Mon May 17 18:50:42 CEST 2010
178 - be more resistant to ordering changes when initialising
179 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
180 Waldorf).
181 - document that perl 5.12 deliberately removed support for cloning.
182
1835.22 Wed Apr 14 03:55:35 CEST 2010
184 - correctly return udnef on errors in Coro::Handle::read/write
185 (testcase by Marc Mims).
186 - convert Coro::Util into a "perl compatibility wrapper" - the functions
187 are less useful now, but are drop-in replacements for existing
188 functions, listing better alternatives in the documentation. This also
189 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
190 with Coro::Util::inet_aton.
191 - do not override $Coro::idle unconditionally in Coro.pm, as other
192 modules could have provided their own idle coro already
193 (for exmaple, Coro::AnyEvent).
194 - fix Coro::Util::gethost* functions.
195 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
196
1975.21 Wed Dec 16 07:19:51 CET 2009
198 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
199 together.
200 - add some examples on how to combine other event loops with Coro in
201 Coro::AnyEvent, and how to run it (and not to block). Seems to be
202 the most common source of confusion.
203 - try to catch people naively blocking in an event callback.
204 - work around the perl filehandle bug issue in conjunction with
205 older common::sense (as indirectly pointed out by ZSystem).
206 - clarify the "not from signal handlers" section.
207
2085.2 Sun Oct 4 14:54:24 CEST 2009
209 - Coro::Storable destroyed the prototypes of the functions it wrapped.
210 - export rouse_cb and rouse_wait by default now.
211 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
212 - new method $state->swap_sv.
213 - added section on "windows process emulation" to the manpage,
214 after a not-so-fruitful (nor-friendly) "discussion" with chip
215 salzenberg (discussion implies arguments, but his only arguments
216 were ad-hominems, one wonders why he started it in the first
217 place). I hope this explains it well enough for him to understand,
218 and maybe well enough for others to understand.
219 - use common::sense everywhere now.
220 - idle callbacks are no longer supported, use idle coros instead.
221 - print a thread listing when a deadlock is detected.
222
2235.17 Sat Aug 22 23:09:31 CEST 2009
224 - work around a bug in the perl debugger causing crashes
225 when running under the debugger by marking _pool_handler as nodebug.
226 - speed up Coro::async considerably.
227 - try some hacks to get netbsd to work "more often" - their broken
228 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
229 - convert Coro to AE 5.0 API.
230
2315.162 Tue Jul 28 04:04:03 CEST 2009
232 - perl 5.8.2 is now minimum requirement.
233 - skip t/19_handle.t on broken windows perls.
234
2355.161 Wed Jul 22 04:47:38 CEST 2009
236 - Coro::AnyEvent::poll could have a different prototype when EV was
237 used as backend (analyzed by Tatsuhiko Miyagawa).
238 - Coro::AnyEvent errornously initialised the event loop when loaded,
239 not on demand.
240 - try to workaround rare */t/01_unblock.t failures.
241
2425.16 Tue Jul 21 01:44:37 CEST 2009
243 - Coro::AnyEvent failed to hook into the event loop
244 when no threads had been readied between detecting
245 the event loop and actually running it.
246 - considerably speed up Coro::Select by taking avdantage
247 of AnyEvent > 4.8 and some other optimisations.
248 - implement paragraph readline mode in Coro::Handle
249 (based on patches by Zsbán Ambrus).
250 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
251 and analyzed by Yasuhiro MATSUMOTO).
252 - clarified libcoro license and copyright.
253 - someone stole my EXTRA_META!!!
254 - implement Coro::Select::patch_pp_sselect and it's brother,
255 for hardcode select overriding.
256
2575.151 Mon Jul 6 05:41:57 CEST 2009
258 - backport to windows process emulation code again (patch by
259 Yasuhiro MATSUMOTO).
260 - slightly update Coro::MakeMaker.
261
2625.15 Tue Jun 30 10:28:06 CEST 2009
263 - deprecate Coro::Socket, document how to get ipv6 support via
264 AnyEvent::Socket instead.
265 - implement signal->wait ($cb) interface, similar to semaphores.
266 - work around SvOK not supporting getmagic, so we have to getmagic
267 to test for undef :/ (reported by Matthias Waldorf).
268 - load Coro::AnyEvent in all modules using AnyEvent.
269 - work around perl corrupting our internal data structures,
270 reported by Tokuhiro Matsuno.
271 - enable per-coroutine real and cpu time gathering
272 (Coro::State::enable_times).
273
2745.14 Wed Jun 24 01:37:48 CEST 2009
275 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
276 as this seems to have been a difficult concept (poll, sleep, idle,
277 idle_upto).
278 - add Coro::AnyEvent::readable/writable functions.
279 - clarify Coro::EV/Event/AnyEvent manpages.
280 - free per-thread global scalars in the thread calling ->cancel, to
281 avoid crashes when $_, $@ etc., are magical but some of those
282 had already been freed.
283 - "unexperimentalise" the callback interface for Coro::Semaphore.
284 - speed up ready queue management/context switching by using a linked
285 list instead of an array (~5-10%).
286 - implement "watch" command in Coro::Debug shells.
287 - for fun, implement time-slicing as an example in the manpage.
288 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
289 Coro::Event, use the normal AnyEvent handling instead of dieing
290 (the same is true for Coro::Handle).
291 - properly document Coro::EV::timed_io_once.
292 - avoid unneccessary ->cancel calls in Coro::Handle.
293 - maybe make it work on mingw32 with win32 backend
294 (based on patch by Yasuhiro Matsumoto).
295
2965.132 Fri May 29 09:00:39 CEST 2009
297 - do not keep a reference to the argument itself in
298 Coro::Semaphore::guard, as it could change later.
299 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
300 should support a prepare callback.
301
3025.131 Mon Mar 16 23:20:37 CET 2009
303 - implement and document Coro->suspend, Coro->resume.
304 - fix Coro::Select implementation to not (often) close
305 the passed file descriptors (testcase provided by pippijn).
306
3075.13 Mon Dec 15 21:51:42 CET 2008
308 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
309 - don't set diehook to C<undef> but instead to NULL, to avoid
310 spurious warnings.
311 - fix a lot of bugs in Coro::SemaphoreSet.
312 - Coro::SemaphoreSet will less often create a semaphore needlessly.
313 - add Coro::SemaphoreSet::count and wait methods.
314 - take advantage of the new Guard module.
315 - deprecate Coro::guard.
316 - try to fix the dreaded 01_unblock tests once more. I hate it when
317 testsuites need more fixing than the code they are supposed to test.
318 - croak in more cases when a required callback isn't resolvable.
319 - fix some minor issues in Coro::State->call/eval.
320 - use current coroutine context instead of a temporary one
321 when temporarily switching to another coroutine.
322 - do not call C-level on_destroy handlers during global destruction,
323 to avoid needless segfaults.
324
3255.12 Sun Dec 7 13:30:38 CET 2008
326 - add default config for MirOS, which seems to be a bug-to-bug
327 compatible fork of openbsd ("world domination by releasing
328 openbsd cvs before the openbsd folks do it" or so :).
329 - free_padlist did destroy the names pad, not good, but didn't
330 seem to bother perl - this could fix issues such as eval ""
331 inside a function called from multiple coroutines.
332 - use a different method to detect destruction time.
333 - be more careful when freeing padlists just before global
334 destruction.
335 - fixed and expanded the call/cc example.
336 - renamed _terminate to _coro_run.
337 - new method Coro::Channel->shutdown.
338 - try pthreads on openbsd <4.4 (broken sigaltstack, will
339 pthreads fare better?).
340 - be less picky about destroying "a" running coroutine.
341
3425.11 Tue Nov 25 21:49:05 CET 2008
343 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
344 working code with failed assertions, introducing more bugs than
345 it fixes, requiring elaborate workarounds :(
346
3475.1 Mon Nov 24 08:54:59 CET 2008
348 - wrote a small introductory tutorial - Coro::Intro.
8 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. 349 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
9 - Coro::Select did errornously dup the file descriptors 350 - Coro::Select did errornously dup the file descriptors
10 and didn't work with all AnyEvent backends. 351 and didn't work with all AnyEvent backends.
352 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
353 LWP data transfers.
11 - disassociate c contexts from coro objects - this is agruably more 354 - disassociate c contexts from coro objects - this is agruably more
12 correct, but mostly allows sharing of cctxs between coro and state 355 correct, but mostly allows sharing of cctxs between coro and state
13 objects, for added memory savings and speed increases. 356 objects, for added memory savings and speed increases.
14 - bumped $Coro::POOL_RSS up to 32kb by default. 357 - bumped $Coro::POOL_RSS up to 32kb by default.
15 - no longer set the optype to OP_CUSTOM, as B::* understandably 358 - no longer set the optype to OP_CUSTOM, as B::* understandably
16 doesn't like this very much (and we *are* a type of entersub). 359 doesn't like this very much (and we *are* a type of entersub).
17 - implement state cloning, just to prove that call/cc can be done. 360 - implement state cloning, just to prove that call/cc can be done.
361 - automatically load Coro::AnyEvent in Coro::Handle.
362 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
363 do not have this method (and don't need it either).
364 - speed up generic anyevent methods in Coro::Handle by using rouse
365 callbacks.
366 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
367 others. It also makes the debugger happier, as you can trace
368 through the idle threads now.
369 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
370 - eg/event was pretty broken.
18 - better 5.8.6 compatibility. 371 - better 5.8.6 compatibility.
19 372
205.0 Thu Nov 20 10:35:05 CET 2008 3735.0 Thu Nov 20 10:35:05 CET 2008
21 - NEW ARCHITECTURE: use the latest 4.x version if you experience 374 - NEW ARCHITECTURE: use the latest 4.x version if you experience
22 stability issues. 375 stability issues.

Diff Legend

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