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

Comparing Coro/Changes (file contents):
Revision 1.57 by root, Thu Nov 1 19:40:22 2001 UTC vs.
Revision 1.303 by root, Mon Dec 17 06:42:05 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO.
4TODO: maybe implement a default message channel, very much like Erlang
5 (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7
84.32
9 - majorly improved Coro::Handle's performance with Coro::EV.
10 - implemented the until now mythical Coro::BDB module.
11 - specialcase EV in Coro::AIO and Coro::BDB.
12
134.31 Wed Dec 5 12:32:39 CET 2007
14 - remove warn statement form Coro::Util.
15
164.3 Tue Dec 4 20:33:14 CET 2007
17 - calls to the idle function could cause stack corruption
18 when the stack changed.
19 - do no longer rely on the presence of EV::DNS if EV is used
20 (because it is gone), but instead take avdantage of EV::ADNS
21 if available.
22 - add ($) prototypes to all functions in Coro::Storable.
23 - use a conventional (safer) idle callback in Coro::EV.
24 - do accept quickly in Coro::Debug to avoid endless loops.
25
264.22 Fri Nov 30 16:04:04 CET 2007
27 - really use optimised versions for Event and EV in Coro::Util
28 and Coro::Handle.
29
304.21 Sun Nov 25 10:48:59 CET 2007
31 - fix a spurious memory read.
32 - Coro::EV no longer keeps the eventloop "alive".
33
344.2 Fri Nov 9 20:47:05 CET 2007
35 - enable/disable tracing from a new coroutine, not a pooled one.
36 - fix a memleak in Coro::Event.
37 - removed killall call from fork_eval.
38 - made sure store_fd is already loaded so that fork_eval
39 does not have to parse autoload in each subprocess.
40 - only use assembly method if -O switch is in $Config{optimize}.
41 - add (optional) Coro::EV module, so far the best event loop module
42 directly supported by Coro.
43 - if the event model is EV, use EV::DNS to resolve
44 stuff in Coro::Util.
45 - don't get confused by multiple event notifications in Coro::Handle.
46 - initial support for EV (libevent interface).
47 - require Event and EV using configure_requires, to force their existance.
48
494.13 Wed Oct 24 07:26:45 CEST 2007
50 - add Coro::Storable::blocking_thaw.
51 - use a vastly more complicated technique to localise
52 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
53 - use a coroutine for the idle callback Coro::Event,
54 instead of running Event in the current coroutine context.
55 This also catches recursive invocations.
56 - actually report fork errors in gethostbyname and inet_aton.
57
584.11 Thu Oct 11 02:40:24 CEST 2007
59 - port to threaded perls.
60
614.1 Thu Oct 11 02:38:16 CEST 2007
62 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
63 be local to each coro (see Coro::State).
64 - incompatible change: for very deep reasons, cede and cede_notself
65 cannot return anything, so nothing will be returned.
66 - possibly bring back 5.10 compatibility (untested).
67 - work around stupid (and wrong) warning on 5.10 :(.
68 - overlay the saved state over the context stack. This saves
69 a few hundred bytes per coroutine on average and also
70 speeds up context switching a bit.
71 - further tune default stack sizes.
72 - (more) correctly calculate stack usage in coro_rss.
73 - Coro::Storable::blocking_* did not properly lock
74 resulting in races between coroutines.
75 - added Coro::Storable::guard.
76 - stopping to trace a coroutine could destroy the cctx of
77 an unrelated coroutine.
78 - explain the relationship between Perl and C coroutines in
79 more detail in Coro::State.
80 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
81 causing a fork per resolve. This also affected Coro::Socket.
82 - switch to a separate stack in $coro->call/eval to avoid
83 invalidating pointers.
84
854.03 Sat Oct 6 21:24:00 CEST 2007
86 - added Coro::throw method.
87 - minor code cleanups.
88
894.02 Sat Oct 6 02:36:47 CEST 2007
90 - fix a very minor per-coroutine memleak (a single codereference).
91 - fixed a bug where the currently in-use c context would be freed
92 prematurely (can happen only when programs change the stacksize
93 or use tracing).
94 - tracing can no longer keep a coro alive after it terminated.
95 - do static branch prediction in the common path for gcc. gives
96 about 2-5% speed improvement here.
97
984.01 Fri Oct 5 22:10:49 CEST 2007
99 - instead of recreating *a* standard output handle we simply
100 use STDOUT, which is faster and hopefully more robust.
101
1024.0 Fri Oct 5 12:56:00 CEST 2007
103 - incompatibly changed Coro::Storable::freeze.
104 - major new feature: added Coro::Debug, for interactive coroutine
105 debugging, tracing and much more.
106 - major bug fix: unbelievable, but true: $_, $/ and many other
107 "saved" variables actually weren't being saved. This has been fixed,
108 of course, while increasing performance while losing all the save
109 flags.
110 - save flags are gone, and all the api functions dealing with them.
111 - added Coro::Semaphore::adjust.
112 - added Coro::Util::fork_eval.
113 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
114 - added Coro::killall.
115 - reduce initial stack sizes to allow for "micro-coroutines".
116 - better async_pool resource management, moved parts of async_pool
117 handling to XS (major speed improvement).
118 - actually croak before modifying important data structures.
119 - refuse to transfer while compiling.
120 - possibly support eval EXPR better now.
121 - enable assembly per default on linux+bsd x86+amd64.
122 - all internal members were renamed _something for easier subclassing.
123 - many minor tweaks.
124
1253.63 Wed May 16 14:10:06 CEST 2007
126 - implement handcoded assembly for x86/amd64 SVR ABI.
127
1283.62 Fri Apr 27 21:36:06 CEST 2007
129 - upgrade libcoro (which might set unwind info correctly).
130 - change default on linux to setjmp/longjmp.
131
1323.61 Thu Apr 19 12:36:18 CEST 2007
133 - Coro::Storable caused an endless loop when thawing invalid
134 pst's sometimes.
135 - use a Semaphore in Coro::Storable, as Storable doesn't
136 seem to be reentrant (although it is documented to
137 be threadsafe...).
138 - fix Coro::Signal to bring back the original unreliable
139 but stateful semantics.
140 - fixed a lot of typos in Coro.pm (patch submitted by David
141 Steinbrunner, which applied flawlessly).
142
1433.6 Sat Apr 14 17:13:31 CEST 2007
144 - added some bugfixes to get eg/myhttpd working again.
145 - added Coro::Storable for often-cede'ing freeze/thaw.
146 - try to do a clean exit when a coroutine calls exit
147 (EXPERIMENTAL).
148 - got rid of indirect call through _coro_init.
149 - updated the partly antique examples in eg/ to
150 work again and be a bit less magic, too.
151 - fixed Coro::Signal semantics to work as documented again.
152
1533.55 Sun Mar 25 01:20:47 CET 2007
154 - add SAVE_DEFFH to save the default filehandle and enable
155 it by default.
156 - finally move socket-operations from Coro::Socket to Coro::Handle
157 to be able to unblock foreign sockets.
158 - add Coro::State::save_also and guarded_save.
159 - add count accessor to Coro::Semaphore.
160 - add Coro::State::cctx_stacksize.
161 - just for the fun of it, do not rely on implicit context,
162 which can dramatically improve performance, but people
163 using windows-process-emulation perls do not care much about
164 performance.
165
1663.51 Sun Mar 4 14:18:04 CET 2007
167 - fixed a problem when you weakened references to Coro::State's
168 (patch by Yuval Kogman).
169
1703.501 Wed Feb 28 12:44:07 CET 2007
171 - rename some global symbols as macosx from hell redefines
172 them without asking.
173
1743.5 Tue Feb 13 20:22:53 CET 2007
175 - do AnyEvent model detection earlier, avoiding problems
176 caused by first using AnyEvent and later Coro::Event.
177 - implement and document Coro::Event event objects.
178 - fix a potential problem in Coro::Event causing crashes.
179 - initialise PL_comppad when creating a new coroutine,
180 avoids crashes on early coro destruction.
181
1823.41 Mon Jan 22 19:19:49 CET 2007
183 - readline on Coro::Handle did not support undefined $/,
184 nor did it deliver partial lines on EOF or error.
185 - implement malloc fallback for stack allocation because
186 stupid broken idiotic OSX has a stupid broken
187 idiotic fits-the-whole-os mmap "implementation" and
188 my dick feels longer if Coro is portable even to
189 obsolete platforms.
190
1913.4 Fri Jan 19 21:52:54 CET 2007
192 - remove t/09_timer.t, as it isn't really testing much
193 but was rather flaky in practise.
194 - async_pool coro would keep arguments and callback alive until
195 it was reused.
196 - cancellation of a coroutine could cause spurious idle calls
197 in cede_notself.
198
1993.3 Sat Jan 6 03:45:00 CET 2007
200 - implement $coro->on_destroy.
201 - Coro::Event blocking semantics have been changed,
202 documented and - hopefully - improved.
203 - fix nice adding, not subtracting, from priority.
204 - fix ->prio and api_is_ready (patch by Mark Hinds).
205 - fixed an assert ("... == prev__cctx->idle_te")
206 that could errronously trigger.
207 - fix various large and small memleaks.
208 - use a (hopefully) more stable cancel implementation
209 that immediately frees the coroutine data.
210 - cede/cede_notself return a status now.
211 - added Coro::guard function.
212 - added a global coroutine pool for jobs (on my machine,
213 I can create and execute 48k simple coros/s with async,
214 and 128k coros with async_pool).
215 - Coro::AIO now uses the coroutine priority as io priority.
216
2173.2 Fri Dec 22 05:07:09 CET 2006
218 - improve portability to slightly older perls.
219 - use cleaner coroutine destruction.
220 - simplify configuration for users.
221 - optionally (unrecommended) prefer perl functions over
222 their coro replacements.
223
2243.11 Tue Dec 5 13:11:24 CET 2006
225 - fixed some bogus assert's, but as perl.h disables assert even
226 without NDEBUG (thank you very much), not too many people should
227 notice (that did include myself). Andreas König noticed, though :)
228 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
229 it. Also noticed by Andreas König :)
230 - save/restore tainted status.
231 - verified to pass the testsuite on my 5.9.5.
232
2333.1 Mon Dec 4 23:03:40 CET 2006
234 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
235 - incompatible change: transfer flags are now per-state.
236 - give a better error message on deadlock.
237 - document Coro::nready.
238 - enhanced testsuite.
239
2403.01 Sun Dec 3 23:47:42 CET 2006
241 - forgot to include Coro::Timer.
242
2433.0 Sun Dec 3 22:57:25 CET 2006
244 - the "FINALLY COMPLETELY STABLE" release (coming soon:
245 the "FAMOUS LAST WORDS" release).
246 - implement a new stack sharing algorithm, which uses a stack
247 pool (size currently hardcoded).
248 - make stack sharing mandatory (it no longer uses a heuristic).
249 - eval/die no longer cause weird problems under heavy use.
250 - Coro::Event could cause livelocks if it was used but
251 no Coro::Event watchers were used.
252 - Coro::Event now uses asynccheck as prepare does not
253 check for changed watchers.
254 - Coro::Event allows multiple waiting coros and will wake up one
255 per event.
256 - Coro::Event should be cleaner and more efficient now.
257 - new utility function Coro::unblock_sub.
258 - document the sad fact that Event is no longer reentrant.
259 - putting a coroutine into the ready queue twice could under
260 some circumstances lead to stack corruption.
261 - minor incompatibility: subclassing Coro::State is supported
262 directly now without going through a _coro_state member.
263 - state/coro switching is much faster now.
264 - very minor optimisations and code/documentation cleanup.
265 - avoid problems due to compiler inlining.
266 - removed timers from Coro::Timer -> use AnyEvent instead.
267 - replaced Coro::idle coroutine by (cleaner) idle handler.
268 - updated to newest libcoro.
269 - implement enhanced support for valgrind.
270 - implement is_ready and return value for ready.
271 - removed Coro::Cont, it was a misnomer (it's generators),
272 and it was somewhat annoying to get it right. Will come back
273 if somebody asks for it :)
274 - many ->wait methods and Coro::Event could return
275 spuriously without the event having happened.
276
2772.5 Tue Nov 7 12:22:33 CET 2006
278 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
279 use AnyEvent, moved them to Coro/.
280 - added Coro::LWP which contains all the uglyness required to
281 make LWP non-blocking.
282 - should work with perl 5.9.x now (Andreas König made me do it).
283 - fixed another bug in Coro::Select when the timeout was undef.
284 - reuse PL_start_env for all coros, saves some memory per coroutine.
285 - manage PL_top_env differently, hopefully to avoid panic: top_env.
286 - timeout argument was not properly used in Coro::Socket.
287 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
288 - emulate undocumented(!) functionality of IO::Socket required
289 by LWP(!!).
290 - updated eg/lwp to work with newer lwp's.
291 - remove "FATAL: uncaught exception" prefix. Coroutines that die
292 kill the whole process, just as exceptions in the main "coroutine"
293 did already.
294
2952.1 Wed Nov 1 23:01:13 CET 2006
296 - fix a long-standing bug in Coro::Select where select with
297 zero timeout would instead change the current default filehandle.
298 - use a simpler and hopefully more robust way to clone padlists
299 (uses less memory and a perl function instead of our own).
300 - coro can now create a stack guard area on many architectures.
301 - Coro::AIO properly reexports additional functions from IO::AIO.
302 - updated libcoro with a workaround for OS X,
303 pach and testing by Michael Schwern.
304
3052.0 Tue Oct 24 05:47:17 CEST 2006
306 - support additional aio requests in Coro::AIO.
307
3081.9 (never properly released due to a glitch)
309
3101.8 Thu Feb 2 00:59:06 CET 2006
311 - applied suggested patch by SAMV to avoid problems during stupid
312 mark & sweep gc run.
313 - applied patch by Scott Walters for 5.9.3 compatibility.
314
3151.7 Tue Dec 27 01:41:58 CET 2005
316 - added Coro::AIO, a thin wrapper around IO::AIO.
317 - improved Makefile.PL explanations.
318
3191.6 Mon Dec 12 21:30:05 CET 2005
320 - additionally save PM_curpm might fix as-of-yet
321 unseen problems with regex matches being attributed to the
322 wrong package.
323 - add t/10_bugs.t, which currently checks against imho broken
324 perls that use 0.26MB of stack space per Perl_magic_get
325 invocation (newer linuxes) as opposed to the 0.0002MB perl
326 normally uses.
327 - make stacksize configurable for the ultimate debian experience.
328
3291.51 Mon Dec 12 18:48:36 CET 2005
330 - remove debugging warn() accidentally left in Coro::Select.
331
3321.5 Tue Nov 29 13:32:44 CET 2005
333 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
334 - fix Coro::Select 'select'.
335 - strict'ify some modules.
336 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
337 - use XSLoader in selected modules.
338 - remove some 5.6 compatibility cruft.
339
3401.4 Tue Sep 6 00:11:05 CEST 2005
341 - libcoro did not take into account the trampoline on amd64,
342 when 'l'inux method was used.
343
3441.31 Tue Aug 30 23:31:33 CEST 2005
345 - some portability fixes/workarounds.
346
3471.3 Sat Aug 20 03:08:56 CEST 2005
348 - no code changes. module seems to work fine.
349 - improve Coro::State docs, remove reference to nonexisting
350 Coro::State::flush method.
351 - no longer autodetect windows, present it as an option instead.
352
3531.2 Mon May 16 02:00:55 CEST 2005
354 - included libcoro.c earlier in State.xs, might improve portability.
355 - use a faster and possibly more stable Coro::Cont implementation.
356 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
357 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
358 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
359
3601.11 Thu Mar 3 18:00:52 CET 2005
361 - change of contact address.
362
3631.1 Tue Feb 22 20:51:16 CET 2005
364 - support [l]inux method on amd64.
365 - allow some context switches while compiling/eval'ing.
366
3671.0 Tue Aug 10 03:47:27 CEST 2004
368 - Event 0.89 fixes the bug again.
369 - fix the remaining(??) memleak. I didn't free the cache-AV for
370 padlists, so one AV per created anon-sub leaked.
371
3720.97 Fri May 14 15:23:32 CEST 2004
373 - remove stupid gcc extension that was an experiment only anyway.
374
3750.96 Thu May 13 18:09:29 CEST 2004
376 - tell people if they have yet another broken Event variant.
377 - made Coro to work _better_ with threads (it seems that, as long
378 as you use Coros in one thread only everything will work fine.
379 Looking at the perl source the problem might be missing locking
380 between Coro and some perl internal routines. That's not easily
381 fixable).
382 - (hopefully) got rid of the static xnext inside transfer.
383
3840.95 Sun Feb 15 03:20:28 CET 2004
385 - removed Coro::State::flush, implemented a new and much
386 faster caching mechanism by attaching magic to the CV.
387 - a memleak with (real) closures remains.
388 - dramatically reduced size of stacks to 65536 (linux/x86), after
389 observing that even myhttpd never had more than a kilobyte
390 of stack in use (stack is still only physically allocated in
391 pagesize increments for systems suporting mmap).
392 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
393 - fix for threaded perl.
394
3950.9 Sun Nov 23 04:32:39 CET 2003
396 - fix a few very hard to track down but (of course) stupid
397 memory leaks.
398 - stringify version number to avoid locale problems :(.
399 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
400 hopefully ignoring it "just works".
401
4020.8 Wed Nov 5 19:38:40 CET 2003
403 - port to perl5.005_03, but only 5.8.x+ is supported!
404 - honor LocalAddr even without LocalPort in Coro::Socket.
405
4060.7 Tue May 27 03:12:38 CEST 2003
407 - the version jump indicates some level of testing, not gobs
408 of new features.
409 - uh... I found the prompt function in ExtUtils::MakeMaker.
410 Highly correct stuff, that is...
411 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
412 the fix is not well understood by the author, so beware :(.
413
4140.652 Thu May 8 02:54:46 CEST 2003
415 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
416 cause version 4 doesn't have ucontext.
417 - Benjamin Reed reported that setjmp works fine on darwin,
418 so preselect it.
419
4200.651 Sat Mar 29 15:00:23 CET 2003
421 - fix a bug in Coro::Handle where some bytes could get lost
422 on reading, reported by jason@nichego.net.
423
4240.65 Sun Mar 23 00:08:26 CET 2003
425 - added a README.linux-glibc.
426 - new module Coro::Select.
427 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
428 (I never claimed that I know what I am doing ;).
429 - set default to setjmp/longjmp on non-x86-linux.
430
4310.6 Thu Nov 21 11:09:06 CET 2002
432 - made lazy context switching the default.
433 - fixed the bug where SemaphoreSet::guard
434 was the same as timed_guard and timed_guard was missing.
435 - fix a memleak in Coro::Timer.
436 - reclassified context sharing as not experimental.
437
4380.534 Sun Apr 14 03:05:12 CEST 2002
439 - fixed backspace => tab in header parsing (myhttpd).
440 - added eg/readline, for elmex.
441 - Coro::Event now calls ->start on first call instead of
442 ->again, which created wrong timeouts for the first ->next.
443 - fixed a bug where Coro::Socket returned a closed socket instead
444 of nothing on ECONNREFUSED and similar error conditions.
445
4460.533 Mon Feb 18 18:49:40 CET 2002
447 - adapt to finally working Event-0.85.
448 - fixed buggy event initialization (trapped under windows, but how
449 could it possibly have worked anywhere?)
450
4510.532 Wed Jan 16 02:45:32 CET 2002
452 - removed the reference to pp_entersub, might work on
453 (some) win32 perls now (testsuite works under cygwin,
454 without Event, which freezes).
455
4560.531 Mon Dec 10 22:18:44 CET 2001
457 - Fixed a bug in SemaphoreSet::timed_down.
458
4590.53 Tue Nov 27 21:11:13 CET 2001
460 - some tests for Event.
461 - slightly different internal architecture to get refcnt into the object
462 and not the reference. might fix a few bugs, certainly did introduce
463 new ones.
464 - Coro::Timer (independent of Event).
465 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
466
4670.52 Tue Nov 6 21:36:18 CET 2001
468 - ported to cygwin (trivial).
469 patch by Gerrit P. Haase <gp@familiehaase.de>.
470 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
471
30.51 Thu Nov 1 20:39:01 CET 2001 4720.51 Thu Nov 1 20:39:01 CET 2001
4 - terminate/cancel now work properly (otherwise termination 473 - terminate/cancel now work properly (otherwise termination
5 could cause a "next coroutine is not and contains not..." 474 could cause a "next coroutine is not and contains not..."
6 error. 475 error.
7 - added Coro::Socket::shutdown. 476 - added Coro::Socket::shutdown.
8 - Coro::Event::loop no is the same as Event::loop. 477 - Coro::Event::loop no is the same as Event::loop.
9 - implemented terminate with args + join. 478 - implemented terminate with args + join.
10 479
110.5 Fri Sep 28 16:15:35 CEST 2001 4800.5 Fri Sep 28 16:15:35 CEST 2001
12 - fixed "print" on a Coro::Handle. The print method worked. 481 - fixed "print" on a Coro::Handle. The print method worked.
13 - small tweaks (seem to reduce memory consumption a lot) 482 - small tweaks (seem to reduce memory consumption a lot)
14 in various modules. 483 in various modules.
15 - splendid use of "no warnings" scattered throughout. 484 - splendid use of "no warnings" scattered throughout.
16 - added Coro::Handle::rbuf, fixed Coro::Handle::fh. 485 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines