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

Comparing Coro/Changes (file contents):
Revision 1.479 by root, Sun Dec 14 22:03:40 2008 UTC vs.
Revision 1.614 by root, Wed Mar 6 06:00:07 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
8new -> ready 76.28 Wed Mar 6 06:58:02 CET 2013
9new -> canceled 8 - clean remnants of existing __DIE__ and __WARN__ handlers so
10ready -> running 9 they lose their magic and will not cause segfaults later
11ready -> canceled 10 (testcase by Andrey Sagulin).
12running -> ready 11 - improved Coro::State documentation a bit.
13running -> blocked 12 - Coro::Debug::command now flushes the output.
14running -> canceled 13 - add hack detection code for x32 abi, because the braindead slugs
15blocked -> ready 14 who designed that made it look exactly like x86_64 without
16blocked -> canceled 15 providing proper compile time symbols to test for it. as a result,
16 this detection cannot work reliably.
17 - valgrind stack registering was broken.
18 - do not rely on Time::HiRes anymore in Coro::Debug.
17 19
185.13 206.23 Fri Dec 7 23:36:37 CET 2012
21 - use experimental fiber implementation on native windows
22 perls.
23 - use sizeof (void *) as multiplication factor for stack sizes,
24 to accomodate the totally braindamaged microsoft 64 bit "os".
25 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
26 activeperl 5.16 crashes when PerlIO_define_layer is called due
27 to some bug in the perl dll, strawberry perl at least passes
28 the testsuite.
29 - implement Coro::Handle->peeraddr/host/port, for slightly
30 improved compatibility with LWP.
31 - implement 5.17 compatibility by almost blindly applying a
32 good-looking patch by Father Chrysostomos.
33 - move stack management functions into libcoro 3.
34 - libcoro version 3 "released".
35 - support magic values as timed_io_once args.
36 - recommend AnyEvent 7+ or EV 4+, also require EV
37 version 4 or newer for Coro::EV.
38
396.10 Tue Oct 9 01:14:27 CEST 2012
40 - updated ecb.h, it had a typo that caused it to not compile on many
41 big endian systems (reported by many people).
42 - disable memory fences in ecb.h to improve portability.
43
446.09 Sat Oct 6 23:25:02 CEST 2012
45 - Coro::EV I/O watchers were not interruptible by exceptions
46 (Coro::State::throw) (testcase by sten).
47 - ->throw now puts threads into the ready queue, as this seems to
48 be expected by existing code, and code that doesn't cope with spurious
49 wakeups needs fixing anyway.
50 - use fd -1 in mmap.
51 - cast I32 to int in error message printf.
52 - warn about broken so-called "hardened" kernels.
53
546.08 Fri Apr 13 12:05:47 CEST 2012
55 - be more aggressive about exiting like perl does - formerly,
56 exiting from the non-main thread would not execute END blocks.
57
586.07 Fri Nov 11 21:21:48 CET 2011
59 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
60 - update ecb.h.
61
626.06 Mon Aug 8 23:59:48 CEST 2011
63 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
64 backend, which is an order of magnitude slower. unfortunately, cygwins
65 pthread implementation isn't very complete either, so allocate the stack
66 twice just to be sure.
67 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
68 in its ithreaded perl - best recompile cygwin and use the 'w'indows
69 backend for much better performance. also disable ithreads for
70 even better performance...).
71
726.05 Thu Aug 4 21:36:36 CEST 2011
73 - blush, condvar values would not be propagated from send to recv anymore
74 (reported by Chip Salzenberg).
75 - use exponential increase for the readline buffer length in
76 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
77
786.04 Wed Aug 3 17:15:45 CEST 2011
79 - use even more efficient and more compatible condvars for
80 compatibility to AnyEvent 6.x :)
81 - more inconsequential ecb.h updates.
82
836.03 Wed Aug 3 11:41:30 CEST 2011
84 - change how Coro patches AnyEvent condvars for compatibility to
85 AnyEvent 6.x.
86 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
87 hasn't been included.
88
896.02 Wed Jul 13 04:35:19 CEST 2011
90 - "improve portability to Gentoo" - gentoo manages to put perl variables
91 in memory areas that are farther than 2gb apart, which the jit couldn't
92 handle and barfed. now it's just a bit slower on gentoo and similar
93 systems.
94
956.01 Sun Jul 3 12:31:14 CEST 2011
96 - workarounds are good, but the test for whether pthreads are used
97 was not good. thisone should be better.
98 - check differently whether gcc generates cfi instructions itself.
99
1006.0 Wed Jun 29 19:43:35 CEST 2011
101 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
102 destroyed and cleaned up automatically (e.g. async { schedule }).
103 - implement a JIT compiler for part of the thread switch code,
104 which gives a 50% speed improvement on threaded perls, and
105 about 4% on non-thraeded perls (so threaded perls now finally
106 reach about half the speed of non-threaded perls).
107 - slightly modernise Coro::Intro, add section about rouse functions.
108 - avoid DEFSV and ERRSV, giving another 10% improvement
109 in thread switching.
110 - Coro::State->is_destroyed is now called is_zombie.
111 - implement a Coro->safe_cancel method that might fail, but
112 cancels in a "safer" way if it succeeds.
113 - add preliminary support for DEBUGGING perls.
114 - get rid of two hash-accesses when initialising a new Coro - this
115 speeds up coro creation by almost a factor of two.
116 - croak when a coro that is being cancelled tries to block
117 (e.g. while executing a guard block), instead of crashing or
118 deadlocking.
119 - use a more robust and also faster method to identify Coro::State
120 objects - speeds up everything a bit.
121 - implement Coro->cancel in XS for a 20% speed improvement, and to
122 be able to implement mutual cancellation.
123 - speed up context switches by a percent or two by more efficiently
124 allocating context stack entries.
125 - implement Coro->join and Coro->on_destroy in XS for a speedup and
126 a reduction in memory use.
127 - cancelling a coro while it itself is cancelling another coro is
128 now supported and working, instead of triggering an assertion.
129 - be a bit more crash-resistant when calling (buggy) on_destroy
130 callbacks (best effort).
131 - move on_destroy into the slf_frame, to allow extension slf
132 functions to have destructors.
133 - get rid if coro refcounting - simply crash in other interpreter
134 threads by nulling the pointers on clone.
135 - simplify warn/die hook handling when loading Coro - the convoluted
136 logic seems to be no longer neccessary.
137 - use libecb instead of our own home-grown gcc hacks.
138 - document alternatives to Coro::LWP. Please use them :)
139 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
140 sigaltstack. Really. wine suffers from it, erlang suffers from it,
141 and it's known since at least 2006.
142
1435.372 Wed Feb 23 06:14:30 CET 2011
144 - apparently mingw doesn't provide a working gettimeofday, try to
145 work around that by relying on Time::HiRes (indirectly brought to
146 my attention by Max Maischein).
147 - fix some portability issues when Time::HiRes was used.
148
1495.371 Mon Feb 21 14:36:08 CET 2011
150 - backport to windows process emulation code again.
151
1525.37 Sat Feb 19 07:49:44 CET 2011
153 - add a big "Coro thread life cycle" section to "man Coro".
154 - try a tentative workaround against the breakage that 5.13 has
155 introduced without depreciation period. sigh.
156 - no longer use Time::HiRes if gettimeofday is available, which
157 saves quite a lot of memory.
158
1595.36 Sun Feb 13 05:33:41 CET 2011
160 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
161 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
162 method call.
163 - undocument Coro::Timer::sleep and obsolete whole module.
164 - optimise Coro::Timer::timeout memory and cpu usage.
165 - slightly updated Coro::Intro for recent changes.
166 - do not initialise PL_dirty anymore.
167
1685.25 Thu Nov 11 01:08:39 CET 2010
169 - try a different approach on netbsd - netbsd 5 finally has marginally
170 working pthreads, but still broken ucontext/sigaltstack.
171 - openbsd 4.8 finally got their act together, Coro works out of the box
172 with asm, setjmp and pthreads (no change, just informational).
173
1745.24 Sat Oct 23 11:27:12 CEST 2010
175 - port to the EV 4.0 API.
176 - work around bugs in mingw32, making strawberry perl work
177 out of the box.
178 - correctly modify Coro::AIO function prototypes
179 so that they reflect the "no optional parameters" rule.
180 - "ported" libcoro to C++.
181
1825.23 Mon May 17 18:50:42 CEST 2010
183 - be more resistant to ordering changes when initialising
184 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
185 Waldorf).
186 - document that perl 5.12 deliberately removed support for cloning.
187
1885.22 Wed Apr 14 03:55:35 CEST 2010
189 - correctly return udnef on errors in Coro::Handle::read/write
190 (testcase by Marc Mims).
191 - convert Coro::Util into a "perl compatibility wrapper" - the functions
192 are less useful now, but are drop-in replacements for existing
193 functions, listing better alternatives in the documentation. This also
194 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
195 with Coro::Util::inet_aton.
196 - do not override $Coro::idle unconditionally in Coro.pm, as other
197 modules could have provided their own idle coro already
198 (for exmaple, Coro::AnyEvent).
199 - fix Coro::Util::gethost* functions.
200 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
201
2025.21 Wed Dec 16 07:19:51 CET 2009
203 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
204 together.
205 - add some examples on how to combine other event loops with Coro in
206 Coro::AnyEvent, and how to run it (and not to block). Seems to be
207 the most common source of confusion.
208 - try to catch people naively blocking in an event callback.
209 - work around the perl filehandle bug issue in conjunction with
210 older common::sense (as indirectly pointed out by ZSystem).
211 - clarify the "not from signal handlers" section.
212
2135.2 Sun Oct 4 14:54:24 CEST 2009
214 - Coro::Storable destroyed the prototypes of the functions it wrapped.
215 - export rouse_cb and rouse_wait by default now.
216 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
217 - new method $state->swap_sv.
218 - added section on "windows process emulation" to the manpage,
219 after a not-so-fruitful (nor-friendly) "discussion" with chip
220 salzenberg (discussion implies arguments, but his only arguments
221 were ad-hominems, one wonders why he started it in the first
222 place). I hope this explains it well enough for him to understand,
223 and maybe well enough for others to understand.
224 - use common::sense everywhere now.
225 - idle callbacks are no longer supported, use idle coros instead.
226 - print a thread listing when a deadlock is detected.
227
2285.17 Sat Aug 22 23:09:31 CEST 2009
229 - work around a bug in the perl debugger causing crashes
230 when running under the debugger by marking _pool_handler as nodebug.
231 - speed up Coro::async considerably.
232 - try some hacks to get netbsd to work "more often" - their broken
233 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
234 - convert Coro to AE 5.0 API.
235
2365.162 Tue Jul 28 04:04:03 CEST 2009
237 - perl 5.8.2 is now minimum requirement.
238 - skip t/19_handle.t on broken windows perls.
239
2405.161 Wed Jul 22 04:47:38 CEST 2009
241 - Coro::AnyEvent::poll could have a different prototype when EV was
242 used as backend (analyzed by Tatsuhiko Miyagawa).
243 - Coro::AnyEvent errornously initialised the event loop when loaded,
244 not on demand.
245 - try to workaround rare */t/01_unblock.t failures.
246
2475.16 Tue Jul 21 01:44:37 CEST 2009
248 - Coro::AnyEvent failed to hook into the event loop
249 when no threads had been readied between detecting
250 the event loop and actually running it.
251 - considerably speed up Coro::Select by taking avdantage
252 of AnyEvent > 4.8 and some other optimisations.
253 - implement paragraph readline mode in Coro::Handle
254 (based on patches by Zsbán Ambrus).
255 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
256 and analyzed by Yasuhiro MATSUMOTO).
257 - clarified libcoro license and copyright.
258 - someone stole my EXTRA_META!!!
259 - implement Coro::Select::patch_pp_sselect and it's brother,
260 for hardcode select overriding.
261
2625.151 Mon Jul 6 05:41:57 CEST 2009
263 - backport to windows process emulation code again (patch by
264 Yasuhiro MATSUMOTO).
265 - slightly update Coro::MakeMaker.
266
2675.15 Tue Jun 30 10:28:06 CEST 2009
268 - deprecate Coro::Socket, document how to get ipv6 support via
269 AnyEvent::Socket instead.
270 - implement signal->wait ($cb) interface, similar to semaphores.
271 - work around SvOK not supporting getmagic, so we have to getmagic
272 to test for undef :/ (reported by Matthias Waldorf).
273 - load Coro::AnyEvent in all modules using AnyEvent.
274 - work around perl corrupting our internal data structures,
275 reported by Tokuhiro Matsuno.
276 - enable per-coroutine real and cpu time gathering
277 (Coro::State::enable_times).
278
2795.14 Wed Jun 24 01:37:48 CEST 2009
280 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
281 as this seems to have been a difficult concept (poll, sleep, idle,
282 idle_upto).
283 - add Coro::AnyEvent::readable/writable functions.
284 - clarify Coro::EV/Event/AnyEvent manpages.
285 - free per-thread global scalars in the thread calling ->cancel, to
286 avoid crashes when $_, $@ etc., are magical but some of those
287 had already been freed.
288 - "unexperimentalise" the callback interface for Coro::Semaphore.
289 - speed up ready queue management/context switching by using a linked
290 list instead of an array (~5-10%).
291 - implement "watch" command in Coro::Debug shells.
292 - for fun, implement time-slicing as an example in the manpage.
293 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
294 Coro::Event, use the normal AnyEvent handling instead of dieing
295 (the same is true for Coro::Handle).
296 - properly document Coro::EV::timed_io_once.
297 - avoid unneccessary ->cancel calls in Coro::Handle.
298 - maybe make it work on mingw32 with win32 backend
299 (based on patch by Yasuhiro Matsumoto).
300
3015.132 Fri May 29 09:00:39 CEST 2009
302 - do not keep a reference to the argument itself in
303 Coro::Semaphore::guard, as it could change later.
304 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
305 should support a prepare callback.
306
3075.131 Mon Mar 16 23:20:37 CET 2009
308 - implement and document Coro->suspend, Coro->resume.
309 - fix Coro::Select implementation to not (often) close
310 the passed file descriptors (testcase provided by pippijn).
311
3125.13 Mon Dec 15 21:51:42 CET 2008
313 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
19 - don't set diehook to C<undef> but instead to NULL, to avoid 314 - don't set diehook to C<undef> but instead to NULL, to avoid
20 spurious warnings. 315 spurious warnings.
21 - fix a lot of bugs in Coro::SemaphoreSet. 316 - fix a lot of bugs in Coro::SemaphoreSet.
22 - Coro::SemaphoreSet will less often create a semaphore needlessly. 317 - Coro::SemaphoreSet will less often create a semaphore needlessly.
23 - add Coro::SemaphoreSet::count and wait methods. 318 - add Coro::SemaphoreSet::count and wait methods.
24 - take advantage of the new Guard module. 319 - take advantage of the new Guard module.
25 - deprecate Coro::guard. 320 - deprecate Coro::guard.
26 - try to fix the dreaded 01_unblock tests once more. I hate it when 321 - try to fix the dreaded 01_unblock tests once more. I hate it when
27 testsuites need more fixing than the code they are supposed to test. 322 testsuites need more fixing than the code they are supposed to test.
323 - croak in more cases when a required callback isn't resolvable.
324 - fix some minor issues in Coro::State->call/eval.
325 - use current coroutine context instead of a temporary one
326 when temporarily switching to another coroutine.
327 - do not call C-level on_destroy handlers during global destruction,
328 to avoid needless segfaults.
28 329
295.12 Sun Dec 7 13:30:38 CET 2008 3305.12 Sun Dec 7 13:30:38 CET 2008
30 - add default config for MirOS, which seems to be a bug-to-bug 331 - add default config for MirOS, which seems to be a bug-to-bug
31 compatible fork of openbsd ("world domination by releasing 332 compatible fork of openbsd ("world domination by releasing
32 openbsd cvs before the openbsd folks do it" or so :). 333 openbsd cvs before the openbsd folks do it" or so :).

Diff Legend

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