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

Comparing Coro/Changes (file contents):
Revision 1.455 by root, Fri Nov 21 02:39:40 2008 UTC vs.
Revision 1.593 by root, Mon Jul 2 23:16:48 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
8 - use fd -1 in mmap.
9 - cast I32 to int in error message printf.
10 - warn about broken so-called "hardened" kernels.
11
126.08 Fri Apr 13 12:05:47 CEST 2012
13 - be more aggressive about exiting like perl does - formerly,
14 exiting from the non-main thread would not execute END blocks.
15
166.07 Fri Nov 11 21:21:48 CET 2011
17 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
18 - update ecb.h.
19
206.06 Mon Aug 8 23:59:48 CEST 2011
21 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
22 backend, which is an order of magnitude slower. unfortunately, cygwins
23 pthread implementation isn't very complete either, so allocate the stack
24 twice just to be sure.
25 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
26 in its ithreaded perl - best recompile cygwin and use the 'w'indows
27 backend for much better performance. also disable ithreads for
28 even better performance...).
29
306.05 Thu Aug 4 21:36:36 CEST 2011
31 - blush, condvar values would not be propagated from send to recv anymore
32 (reported by Chip Salzenberg).
33 - use exponential increase for the readline buffer length in
34 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
35
366.04 Wed Aug 3 17:15:45 CEST 2011
37 - use even more efficient and more compatible condvars for
38 compatibility to AnyEvent 6.x :)
39 - more inconsequential ecb.h updates.
40
416.03 Wed Aug 3 11:41:30 CEST 2011
42 - change how Coro patches AnyEvent condvars for compatibility to
43 AnyEvent 6.x.
44 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
45 hasn't been included.
46
476.02 Wed Jul 13 04:35:19 CEST 2011
48 - "improve portability to Gentoo" - gentoo manages to put perl variables
49 in memory areas that are farther than 2gb apart, which the jit couldn't
50 handle and barfed. now it's just a bit slower on gentoo and similar
51 systems.
52
536.01 Sun Jul 3 12:31:14 CEST 2011
54 - workarounds are good, but the test for whether pthreads are used
55 was not good. thisone should be better.
56 - check differently whether gcc generates cfi instructions itself.
57
586.0 Wed Jun 29 19:43:35 CEST 2011
59 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
60 destroyed and cleaned up automatically (e.g. async { schedule }).
61 - implement a JIT compiler for part of the thread switch code,
62 which gives a 50% speed improvement on threaded perls, and
63 about 4% on non-thraeded perls (so threaded perls now finally
64 reach about half the speed of non-threaded perls).
65 - slightly modernise Coro::Intro, add section about rouse functions.
66 - avoid DEFSV and ERRSV, giving another 10% improvement
67 in thread switching.
68 - Coro::State->is_destroyed is now called is_zombie.
69 - implement a Coro->safe_cancel method that might fail, but
70 cancels in a "safer" way if it succeeds.
71 - add preliminary support for DEBUGGING perls.
72 - get rid of two hash-accesses when initialising a new Coro - this
73 speeds up coro creation by almost a factor of two.
74 - croak when a coro that is being cancelled tries to block
75 (e.g. while executing a guard block), instead of crashing or
76 deadlocking.
77 - use a more robust and also faster method to identify Coro::State
78 objects - speeds up everything a bit.
79 - implement Coro->cancel in XS for a 20% speed improvement, and to
80 be able to implement mutual cancellation.
81 - speed up context switches by a percent or two by more efficiently
82 allocating context stack entries.
83 - implement Coro->join and Coro->on_destroy in XS for a speedup and
84 a reduction in memory use.
85 - cancelling a coro while it itself is cancelling another coro is
86 now supported and working, instead of triggering an assertion.
87 - be a bit more crash-resistant when calling (buggy) on_destroy
88 callbacks (best effort).
89 - move on_destroy into the slf_frame, to allow extension slf
90 functions to have destructors.
91 - get rid if coro refcounting - simply crash in other interpreter
92 threads by nulling the pointers on clone.
93 - simplify warn/die hook handling when loading Coro - the convoluted
94 logic seems to be no longer neccessary.
95 - use libecb instead of our own home-grown gcc hacks.
96 - document alternatives to Coro::LWP. Please use them :)
97 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
98 sigaltstack. Really. wine suffers from it, erlang suffers from it,
99 and it's known since at least 2006.
100
1015.372 Wed Feb 23 06:14:30 CET 2011
102 - apparently mingw doesn't provide a working gettimeofday, try to
103 work around that by relying on Time::HiRes (indirectly brought to
104 my attention by Max Maischein).
105 - fix some portability issues when Time::HiRes was used.
106
1075.371 Mon Feb 21 14:36:08 CET 2011
108 - backport to windows process emulation code again.
109
1105.37 Sat Feb 19 07:49:44 CET 2011
111 - add a big "Coro thread life cycle" section to "man Coro".
112 - try a tentative workaround against the breakage that 5.13 has
113 introduced without depreciation period. sigh.
114 - no longer use Time::HiRes if gettimeofday is available, which
115 saves quite a lot of memory.
116
1175.36 Sun Feb 13 05:33:41 CET 2011
118 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
119 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
120 method call.
121 - undocument Coro::Timer::sleep and obsolete whole module.
122 - optimise Coro::Timer::timeout memory and cpu usage.
123 - slightly updated Coro::Intro for recent changes.
124 - do not initialise PL_dirty anymore.
125
1265.25 Thu Nov 11 01:08:39 CET 2010
127 - try a different approach on netbsd - netbsd 5 finally has marginally
128 working pthreads, but still broken ucontext/sigaltstack.
129 - openbsd 4.8 finally got their act together, Coro works out of the box
130 with asm, setjmp and pthreads (no change, just informational).
131
1325.24 Sat Oct 23 11:27:12 CEST 2010
133 - port to the EV 4.0 API.
134 - work around bugs in mingw32, making strawberry perl work
135 out of the box.
136 - correctly modify Coro::AIO function prototypes
137 so that they reflect the "no optional parameters" rule.
138 - "ported" libcoro to C++.
139
1405.23 Mon May 17 18:50:42 CEST 2010
141 - be more resistant to ordering changes when initialising
142 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
143 Waldorf).
144 - document that perl 5.12 deliberately removed support for cloning.
145
1465.22 Wed Apr 14 03:55:35 CEST 2010
147 - correctly return udnef on errors in Coro::Handle::read/write
148 (testcase by Marc Mims).
149 - convert Coro::Util into a "perl compatibility wrapper" - the functions
150 are less useful now, but are drop-in replacements for existing
151 functions, listing better alternatives in the documentation. This also
152 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
153 with Coro::Util::inet_aton.
154 - do not override $Coro::idle unconditionally in Coro.pm, as other
155 modules could have provided their own idle coro already
156 (for exmaple, Coro::AnyEvent).
157 - fix Coro::Util::gethost* functions.
158 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
159
1605.21 Wed Dec 16 07:19:51 CET 2009
161 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
162 together.
163 - add some examples on how to combine other event loops with Coro in
164 Coro::AnyEvent, and how to run it (and not to block). Seems to be
165 the most common source of confusion.
166 - try to catch people naively blocking in an event callback.
167 - work around the perl filehandle bug issue in conjunction with
168 older common::sense (as indirectly pointed out by ZSystem).
169 - clarify the "not from signal handlers" section.
170
1715.2 Sun Oct 4 14:54:24 CEST 2009
172 - Coro::Storable destroyed the prototypes of the functions it wrapped.
173 - export rouse_cb and rouse_wait by default now.
174 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
175 - new method $state->swap_sv.
176 - added section on "windows process emulation" to the manpage,
177 after a not-so-fruitful (nor-friendly) "discussion" with chip
178 salzenberg (discussion implies arguments, but his only arguments
179 were ad-hominems, one wonders why he started it in the first
180 place). I hope this explains it well enough for him to understand,
181 and maybe well enough for others to understand.
182 - use common::sense everywhere now.
183 - idle callbacks are no longer supported, use idle coros instead.
184 - print a thread listing when a deadlock is detected.
185
1865.17 Sat Aug 22 23:09:31 CEST 2009
187 - work around a bug in the perl debugger causing crashes
188 when running under the debugger by marking _pool_handler as nodebug.
189 - speed up Coro::async considerably.
190 - try some hacks to get netbsd to work "more often" - their broken
191 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
192 - convert Coro to AE 5.0 API.
193
1945.162 Tue Jul 28 04:04:03 CEST 2009
195 - perl 5.8.2 is now minimum requirement.
196 - skip t/19_handle.t on broken windows perls.
197
1985.161 Wed Jul 22 04:47:38 CEST 2009
199 - Coro::AnyEvent::poll could have a different prototype when EV was
200 used as backend (analyzed by Tatsuhiko Miyagawa).
201 - Coro::AnyEvent errornously initialised the event loop when loaded,
202 not on demand.
203 - try to workaround rare */t/01_unblock.t failures.
204
2055.16 Tue Jul 21 01:44:37 CEST 2009
206 - Coro::AnyEvent failed to hook into the event loop
207 when no threads had been readied between detecting
208 the event loop and actually running it.
209 - considerably speed up Coro::Select by taking avdantage
210 of AnyEvent > 4.8 and some other optimisations.
211 - implement paragraph readline mode in Coro::Handle
212 (based on patches by Zsbán Ambrus).
213 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
214 and analyzed by Yasuhiro MATSUMOTO).
215 - clarified libcoro license and copyright.
216 - someone stole my EXTRA_META!!!
217 - implement Coro::Select::patch_pp_sselect and it's brother,
218 for hardcode select overriding.
219
2205.151 Mon Jul 6 05:41:57 CEST 2009
221 - backport to windows process emulation code again (patch by
222 Yasuhiro MATSUMOTO).
223 - slightly update Coro::MakeMaker.
224
2255.15 Tue Jun 30 10:28:06 CEST 2009
226 - deprecate Coro::Socket, document how to get ipv6 support via
227 AnyEvent::Socket instead.
228 - implement signal->wait ($cb) interface, similar to semaphores.
229 - work around SvOK not supporting getmagic, so we have to getmagic
230 to test for undef :/ (reported by Matthias Waldorf).
231 - load Coro::AnyEvent in all modules using AnyEvent.
232 - work around perl corrupting our internal data structures,
233 reported by Tokuhiro Matsuno.
234 - enable per-coroutine real and cpu time gathering
235 (Coro::State::enable_times).
236
2375.14 Wed Jun 24 01:37:48 CEST 2009
238 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
239 as this seems to have been a difficult concept (poll, sleep, idle,
240 idle_upto).
241 - add Coro::AnyEvent::readable/writable functions.
242 - clarify Coro::EV/Event/AnyEvent manpages.
243 - free per-thread global scalars in the thread calling ->cancel, to
244 avoid crashes when $_, $@ etc., are magical but some of those
245 had already been freed.
246 - "unexperimentalise" the callback interface for Coro::Semaphore.
247 - speed up ready queue management/context switching by using a linked
248 list instead of an array (~5-10%).
249 - implement "watch" command in Coro::Debug shells.
250 - for fun, implement time-slicing as an example in the manpage.
251 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
252 Coro::Event, use the normal AnyEvent handling instead of dieing
253 (the same is true for Coro::Handle).
254 - properly document Coro::EV::timed_io_once.
255 - avoid unneccessary ->cancel calls in Coro::Handle.
256 - maybe make it work on mingw32 with win32 backend
257 (based on patch by Yasuhiro Matsumoto).
258
2595.132 Fri May 29 09:00:39 CEST 2009
260 - do not keep a reference to the argument itself in
261 Coro::Semaphore::guard, as it could change later.
262 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
263 should support a prepare callback.
264
2655.131 Mon Mar 16 23:20:37 CET 2009
266 - implement and document Coro->suspend, Coro->resume.
267 - fix Coro::Select implementation to not (often) close
268 the passed file descriptors (testcase provided by pippijn).
269
2705.13 Mon Dec 15 21:51:42 CET 2008
271 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
272 - don't set diehook to C<undef> but instead to NULL, to avoid
273 spurious warnings.
274 - fix a lot of bugs in Coro::SemaphoreSet.
275 - Coro::SemaphoreSet will less often create a semaphore needlessly.
276 - add Coro::SemaphoreSet::count and wait methods.
277 - take advantage of the new Guard module.
278 - deprecate Coro::guard.
279 - try to fix the dreaded 01_unblock tests once more. I hate it when
280 testsuites need more fixing than the code they are supposed to test.
281 - croak in more cases when a required callback isn't resolvable.
282 - fix some minor issues in Coro::State->call/eval.
283 - use current coroutine context instead of a temporary one
284 when temporarily switching to another coroutine.
285 - do not call C-level on_destroy handlers during global destruction,
286 to avoid needless segfaults.
287
2885.12 Sun Dec 7 13:30:38 CET 2008
289 - add default config for MirOS, which seems to be a bug-to-bug
290 compatible fork of openbsd ("world domination by releasing
291 openbsd cvs before the openbsd folks do it" or so :).
292 - free_padlist did destroy the names pad, not good, but didn't
293 seem to bother perl - this could fix issues such as eval ""
294 inside a function called from multiple coroutines.
295 - use a different method to detect destruction time.
296 - be more careful when freeing padlists just before global
297 destruction.
298 - fixed and expanded the call/cc example.
299 - renamed _terminate to _coro_run.
300 - new method Coro::Channel->shutdown.
301 - try pthreads on openbsd <4.4 (broken sigaltstack, will
302 pthreads fare better?).
303 - be less picky about destroying "a" running coroutine.
304
3055.11 Tue Nov 25 21:49:05 CET 2008
306 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
307 working code with failed assertions, introducing more bugs than
308 it fixes, requiring elaborate workarounds :(
309
3105.1 Mon Nov 24 08:54:59 CET 2008
311 - wrote a small introductory tutorial - Coro::Intro.
8 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. 312 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
9 - Coro::Select did errornously dup the file descriptors 313 - Coro::Select did errornously dup the file descriptors
10 and didn't work with all AnyEvent backends. 314 and didn't work with all AnyEvent backends.
315 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
316 LWP data transfers.
317 - disassociate c contexts from coro objects - this is agruably more
318 correct, but mostly allows sharing of cctxs between coro and state
319 objects, for added memory savings and speed increases.
11 - bumped $Coro::POOL_RSS up to 32kb by default. 320 - bumped $Coro::POOL_RSS up to 32kb by default.
321 - no longer set the optype to OP_CUSTOM, as B::* understandably
322 doesn't like this very much (and we *are* a type of entersub).
323 - implement state cloning, just to prove that call/cc can be done.
324 - automatically load Coro::AnyEvent in Coro::Handle.
325 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
326 do not have this method (and don't need it either).
327 - speed up generic anyevent methods in Coro::Handle by using rouse
328 callbacks.
329 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
330 others. It also makes the debugger happier, as you can trace
331 through the idle threads now.
332 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
333 - eg/event was pretty broken.
334 - better 5.8.6 compatibility.
12 335
135.0 Thu Nov 20 10:35:05 CET 2008 3365.0 Thu Nov 20 10:35:05 CET 2008
14 - NEW ARCHITECTURE: use the latest 4.x version if you experience 337 - NEW ARCHITECTURE: use the latest 4.x version if you experience
15 stability issues. 338 stability issues.
16 - bump API version to 7 - all dependents must be recompiled. 339 - bump API version to 7 - all dependents must be recompiled.

Diff Legend

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