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

Comparing Coro/Changes (file contents):
Revision 1.123 by root, Mon Dec 26 18:48:11 2005 UTC vs.
Revision 1.336 by root, Sat Apr 19 19:06:02 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
31.7 3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO.
4TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7
8 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
9 by the Coro module. It was a mistake.
10
114.51 Mon Apr 14 13:28:27 CEST 2008
12 - make it compile again on 5.8.
13
144.50 Thu Apr 10 09:43:17 CEST 2008
15 - I did it twice! (see 4.49).
16
174.49 Sun Apr 6 21:23:31 CEST 2008
18 - grr, instead of compiling the recent changes on 5.10 only they
19 were compiled on 5.8 only.
20
214.48 Sun Apr 6 20:36:46 CEST 2008
22 - allow coroutine switches during eval's under 5.10.x, as it seems
23 the parser is a per-interpreter option now, so this is safe
24 (this might fix the odd crashing bug).
25 - drop support for 5.9.x versions: they are dead, jim.
26
274.47 Sun Apr 6 00:37:52 CEST 2008
28 - force cctx allocation on API calls: we know we need to force one
29 and gcc actually manages to confuse our heuristic nowadays,
30 leading to crashes and worse.
31 - document force_cctx.
32
334.46 Fri Apr 4 22:05:43 CEST 2008
34 - upgrade libcoro, resulting in pthread-backend (which was only created
35 to fulfill the rules of the programming languages shootout).
36
374.45 Thu Mar 13 11:55:36 CET 2008
38 - fix a file leaking bug in eg/mhyttpd that would allow
39 downloading of any file (reported by oesi).
40 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
41 (also add testcase).
42 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
43 apparently having loads of fun with that).
44 - implement $state->swap_defsv, ->swap_defav and document ->throw.
45
464.4 Wed Feb 13 16:44:29 CET 2008
47 - only assume .cfi pseudo isns exist on GNU/Linux.
48 - add get_timed method to Coro::Channels.
49 - fixed Coro::Debug manpage.
50 - perl 5.11 compatbility improvement based on patch by
51 Andreas König.
52
534.37 Sun Jan 20 11:25:23 CET 2008
54 - soften the check for an active parser for perl 5.10 (avoiding
55 "Coro::State::transfer called while parsing" in many cases).
56
574.36 Sun Jan 13 10:53:56 CET 2008
58 - reset diehook when terminating from an async_pool as to not
59 trigger any __DIE__ handlers.
60
614.35 Sun Jan 13 04:14:13 CET 2008
62 - "bt" debug command now displays any exceptions
63 from longmess and also skips the innermost
64 stackframes, giving more useufl output.
65 - allow backtraces in coroutines blocked in Coro::EV,
66 at a <1% speed hit.
67 - handle localising of $SIG{__WARN__} and __DIE__
68 properly (with a proper amount of dirty hacking).
69
704.34 Sat Dec 22 17:49:53 CET 2007
71 - upgrade to EV version 2.0 API.
72
734.33 Mon Dec 17 08:36:12 CET 2007
74 - make Coro::AIO etc. loadable in the absence of EV.
75
764.32 Mon Dec 17 07:46:02 CET 2007
77 - majorly improved Coro::Handle's performance with Coro::EV.
78 - implemented the until now mythical Coro::BDB module.
79 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
80
814.31 Wed Dec 5 12:32:39 CET 2007
82 - remove warn statement form Coro::Util.
83
844.3 Tue Dec 4 20:33:14 CET 2007
85 - calls to the idle function could cause stack corruption
86 when the stack changed.
87 - do no longer rely on the presence of EV::DNS if EV is used
88 (because it is gone), but instead take avdantage of EV::ADNS
89 if available.
90 - add ($) prototypes to all functions in Coro::Storable.
91 - use a conventional (safer) idle callback in Coro::EV.
92 - do accept quickly in Coro::Debug to avoid endless loops.
93
944.22 Fri Nov 30 16:04:04 CET 2007
95 - really use optimised versions for Event and EV in Coro::Util
96 and Coro::Handle.
97
984.21 Sun Nov 25 10:48:59 CET 2007
99 - fix a spurious memory read.
100 - Coro::EV no longer keeps the eventloop "alive".
101
1024.2 Fri Nov 9 20:47:05 CET 2007
103 - enable/disable tracing from a new coroutine, not a pooled one.
104 - fix a memleak in Coro::Event.
105 - removed killall call from fork_eval.
106 - made sure store_fd is already loaded so that fork_eval
107 does not have to parse autoload in each subprocess.
108 - only use assembly method if -O switch is in $Config{optimize}.
109 - add (optional) Coro::EV module, so far the best event loop module
110 directly supported by Coro.
111 - if the event model is EV, use EV::DNS to resolve
112 stuff in Coro::Util.
113 - don't get confused by multiple event notifications in Coro::Handle.
114 - initial support for EV (libevent interface).
115 - require Event and EV using configure_requires, to force their existance.
116
1174.13 Wed Oct 24 07:26:45 CEST 2007
118 - add Coro::Storable::blocking_thaw.
119 - use a vastly more complicated technique to localise
120 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
121 - use a coroutine for the idle callback Coro::Event,
122 instead of running Event in the current coroutine context.
123 This also catches recursive invocations.
124 - actually report fork errors in gethostbyname and inet_aton.
125
1264.11 Thu Oct 11 02:40:24 CEST 2007
127 - port to threaded perls.
128
1294.1 Thu Oct 11 02:38:16 CEST 2007
130 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
131 be local to each coro (see Coro::State).
132 - incompatible change: for very deep reasons, cede and cede_notself
133 cannot return anything, so nothing will be returned.
134 - possibly bring back 5.10 compatibility (untested).
135 - work around stupid (and wrong) warning on 5.10 :(.
136 - overlay the saved state over the context stack. This saves
137 a few hundred bytes per coroutine on average and also
138 speeds up context switching a bit.
139 - further tune default stack sizes.
140 - (more) correctly calculate stack usage in coro_rss.
141 - Coro::Storable::blocking_* did not properly lock
142 resulting in races between coroutines.
143 - added Coro::Storable::guard.
144 - stopping to trace a coroutine could destroy the cctx of
145 an unrelated coroutine.
146 - explain the relationship between Perl and C coroutines in
147 more detail in Coro::State.
148 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
149 causing a fork per resolve. This also affected Coro::Socket.
150 - switch to a separate stack in $coro->call/eval to avoid
151 invalidating pointers.
152
1534.03 Sat Oct 6 21:24:00 CEST 2007
154 - added Coro::throw method.
155 - minor code cleanups.
156
1574.02 Sat Oct 6 02:36:47 CEST 2007
158 - fix a very minor per-coroutine memleak (a single codereference).
159 - fixed a bug where the currently in-use c context would be freed
160 prematurely (can happen only when programs change the stacksize
161 or use tracing).
162 - tracing can no longer keep a coro alive after it terminated.
163 - do static branch prediction in the common path for gcc. gives
164 about 2-5% speed improvement here.
165
1664.01 Fri Oct 5 22:10:49 CEST 2007
167 - instead of recreating *a* standard output handle we simply
168 use STDOUT, which is faster and hopefully more robust.
169
1704.0 Fri Oct 5 12:56:00 CEST 2007
171 - incompatibly changed Coro::Storable::freeze.
172 - major new feature: added Coro::Debug, for interactive coroutine
173 debugging, tracing and much more.
174 - major bug fix: unbelievable, but true: $_, $/ and many other
175 "saved" variables actually weren't being saved. This has been fixed,
176 of course, while increasing performance while losing all the save
177 flags.
178 - save flags are gone, and all the api functions dealing with them.
179 - added Coro::Semaphore::adjust.
180 - added Coro::Util::fork_eval.
181 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
182 - added Coro::killall.
183 - reduce initial stack sizes to allow for "micro-coroutines".
184 - better async_pool resource management, moved parts of async_pool
185 handling to XS (major speed improvement).
186 - actually croak before modifying important data structures.
187 - refuse to transfer while compiling.
188 - possibly support eval EXPR better now.
189 - enable assembly per default on linux+bsd x86+amd64.
190 - all internal members were renamed _something for easier subclassing.
191 - many minor tweaks.
192
1933.63 Wed May 16 14:10:06 CEST 2007
194 - implement handcoded assembly for x86/amd64 SVR ABI.
195
1963.62 Fri Apr 27 21:36:06 CEST 2007
197 - upgrade libcoro (which might set unwind info correctly).
198 - change default on linux to setjmp/longjmp.
199
2003.61 Thu Apr 19 12:36:18 CEST 2007
201 - Coro::Storable caused an endless loop when thawing invalid
202 pst's sometimes.
203 - use a Semaphore in Coro::Storable, as Storable doesn't
204 seem to be reentrant (although it is documented to
205 be threadsafe...).
206 - fix Coro::Signal to bring back the original unreliable
207 but stateful semantics.
208 - fixed a lot of typos in Coro.pm (patch submitted by David
209 Steinbrunner, which applied flawlessly).
210
2113.6 Sat Apr 14 17:13:31 CEST 2007
212 - added some bugfixes to get eg/myhttpd working again.
213 - added Coro::Storable for often-cede'ing freeze/thaw.
214 - try to do a clean exit when a coroutine calls exit
215 (EXPERIMENTAL).
216 - got rid of indirect call through _coro_init.
217 - updated the partly antique examples in eg/ to
218 work again and be a bit less magic, too.
219 - fixed Coro::Signal semantics to work as documented again.
220
2213.55 Sun Mar 25 01:20:47 CET 2007
222 - add SAVE_DEFFH to save the default filehandle and enable
223 it by default.
224 - finally move socket-operations from Coro::Socket to Coro::Handle
225 to be able to unblock foreign sockets.
226 - add Coro::State::save_also and guarded_save.
227 - add count accessor to Coro::Semaphore.
228 - add Coro::State::cctx_stacksize.
229 - just for the fun of it, do not rely on implicit context,
230 which can dramatically improve performance, but people
231 using windows-process-emulation perls do not care much about
232 performance.
233
2343.51 Sun Mar 4 14:18:04 CET 2007
235 - fixed a problem when you weakened references to Coro::State's
236 (patch by Yuval Kogman).
237
2383.501 Wed Feb 28 12:44:07 CET 2007
239 - rename some global symbols as macosx from hell redefines
240 them without asking.
241
2423.5 Tue Feb 13 20:22:53 CET 2007
243 - do AnyEvent model detection earlier, avoiding problems
244 caused by first using AnyEvent and later Coro::Event.
245 - implement and document Coro::Event event objects.
246 - fix a potential problem in Coro::Event causing crashes.
247 - initialise PL_comppad when creating a new coroutine,
248 avoids crashes on early coro destruction.
249
2503.41 Mon Jan 22 19:19:49 CET 2007
251 - readline on Coro::Handle did not support undefined $/,
252 nor did it deliver partial lines on EOF or error.
253 - implement malloc fallback for stack allocation because
254 stupid broken idiotic OSX has a stupid broken
255 idiotic fits-the-whole-os mmap "implementation" and
256 my dick feels longer if Coro is portable even to
257 obsolete platforms.
258
2593.4 Fri Jan 19 21:52:54 CET 2007
260 - remove t/09_timer.t, as it isn't really testing much
261 but was rather flaky in practise.
262 - async_pool coro would keep arguments and callback alive until
263 it was reused.
264 - cancellation of a coroutine could cause spurious idle calls
265 in cede_notself.
266
2673.3 Sat Jan 6 03:45:00 CET 2007
268 - implement $coro->on_destroy.
269 - Coro::Event blocking semantics have been changed,
270 documented and - hopefully - improved.
271 - fix nice adding, not subtracting, from priority.
272 - fix ->prio and api_is_ready (patch by Mark Hinds).
273 - fixed an assert ("... == prev__cctx->idle_te")
274 that could errronously trigger.
275 - fix various large and small memleaks.
276 - use a (hopefully) more stable cancel implementation
277 that immediately frees the coroutine data.
278 - cede/cede_notself return a status now.
279 - added Coro::guard function.
280 - added a global coroutine pool for jobs (on my machine,
281 I can create and execute 48k simple coros/s with async,
282 and 128k coros with async_pool).
283 - Coro::AIO now uses the coroutine priority as io priority.
284
2853.2 Fri Dec 22 05:07:09 CET 2006
286 - improve portability to slightly older perls.
287 - use cleaner coroutine destruction.
288 - simplify configuration for users.
289 - optionally (unrecommended) prefer perl functions over
290 their coro replacements.
291
2923.11 Tue Dec 5 13:11:24 CET 2006
293 - fixed some bogus assert's, but as perl.h disables assert even
294 without NDEBUG (thank you very much), not too many people should
295 notice (that did include myself). Andreas König noticed, though :)
296 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
297 it. Also noticed by Andreas König :)
298 - save/restore tainted status.
299 - verified to pass the testsuite on my 5.9.5.
300
3013.1 Mon Dec 4 23:03:40 CET 2006
302 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
303 - incompatible change: transfer flags are now per-state.
304 - give a better error message on deadlock.
305 - document Coro::nready.
306 - enhanced testsuite.
307
3083.01 Sun Dec 3 23:47:42 CET 2006
309 - forgot to include Coro::Timer.
310
3113.0 Sun Dec 3 22:57:25 CET 2006
312 - the "FINALLY COMPLETELY STABLE" release (coming soon:
313 the "FAMOUS LAST WORDS" release).
314 - implement a new stack sharing algorithm, which uses a stack
315 pool (size currently hardcoded).
316 - make stack sharing mandatory (it no longer uses a heuristic).
317 - eval/die no longer cause weird problems under heavy use.
318 - Coro::Event could cause livelocks if it was used but
319 no Coro::Event watchers were used.
320 - Coro::Event now uses asynccheck as prepare does not
321 check for changed watchers.
322 - Coro::Event allows multiple waiting coros and will wake up one
323 per event.
324 - Coro::Event should be cleaner and more efficient now.
325 - new utility function Coro::unblock_sub.
326 - document the sad fact that Event is no longer reentrant.
327 - putting a coroutine into the ready queue twice could under
328 some circumstances lead to stack corruption.
329 - minor incompatibility: subclassing Coro::State is supported
330 directly now without going through a _coro_state member.
331 - state/coro switching is much faster now.
332 - very minor optimisations and code/documentation cleanup.
333 - avoid problems due to compiler inlining.
334 - removed timers from Coro::Timer -> use AnyEvent instead.
335 - replaced Coro::idle coroutine by (cleaner) idle handler.
336 - updated to newest libcoro.
337 - implement enhanced support for valgrind.
338 - implement is_ready and return value for ready.
339 - removed Coro::Cont, it was a misnomer (it's generators),
340 and it was somewhat annoying to get it right. Will come back
341 if somebody asks for it :)
342 - many ->wait methods and Coro::Event could return
343 spuriously without the event having happened.
344
3452.5 Tue Nov 7 12:22:33 CET 2006
346 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
347 use AnyEvent, moved them to Coro/.
348 - added Coro::LWP which contains all the uglyness required to
349 make LWP non-blocking.
350 - should work with perl 5.9.x now (Andreas König made me do it).
351 - fixed another bug in Coro::Select when the timeout was undef.
352 - reuse PL_start_env for all coros, saves some memory per coroutine.
353 - manage PL_top_env differently, hopefully to avoid panic: top_env.
354 - timeout argument was not properly used in Coro::Socket.
355 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
356 - emulate undocumented(!) functionality of IO::Socket required
357 by LWP(!!).
358 - updated eg/lwp to work with newer lwp's.
359 - remove "FATAL: uncaught exception" prefix. Coroutines that die
360 kill the whole process, just as exceptions in the main "coroutine"
361 did already.
362
3632.1 Wed Nov 1 23:01:13 CET 2006
364 - fix a long-standing bug in Coro::Select where select with
365 zero timeout would instead change the current default filehandle.
366 - use a simpler and hopefully more robust way to clone padlists
367 (uses less memory and a perl function instead of our own).
368 - coro can now create a stack guard area on many architectures.
369 - Coro::AIO properly reexports additional functions from IO::AIO.
370 - updated libcoro with a workaround for OS X,
371 pach and testing by Michael Schwern.
372
3732.0 Tue Oct 24 05:47:17 CEST 2006
374 - support additional aio requests in Coro::AIO.
375
3761.9 (never properly released due to a glitch)
377
3781.8 Thu Feb 2 00:59:06 CET 2006
379 - applied suggested patch by SAMV to avoid problems during stupid
380 mark & sweep gc run.
381 - applied patch by Scott Walters for 5.9.3 compatibility.
382
3831.7 Tue Dec 27 01:41:58 CET 2005
4 - added Coro::AIO, a thin wrapper around IO::AIO. 384 - added Coro::AIO, a thin wrapper around IO::AIO.
5 - improved Makefile.PL explanations. 385 - improved Makefile.PL explanations.
6 386
71.6 Mon Dec 12 21:30:05 CET 2005 3871.6 Mon Dec 12 21:30:05 CET 2005
8 - additionally save PM_curpm might fix as-of-yet 388 - additionally save PM_curpm might fix as-of-yet

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines