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

Comparing Coro/Changes (file contents):
Revision 1.54 by root, Wed Oct 3 01:09:56 2001 UTC vs.
Revision 1.317 by root, Sun Jan 20 10:23:44 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
30.51 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
84.37
9 - soften the check for an active parser for perl 5.10 (avoiding
10 "Coro::State::transfer called while parsing" in many cases).
11
124.36 Sun Jan 13 10:53:56 CET 2008
13 - reset diehook when terminating from an async_pool as to not
14 trigger any __DIE__ handlers.
15
164.35 Sun Jan 13 04:14:13 CET 2008
17 - "bt" debug command now displays any exceptions
18 from longmess and also skips the innermost
19 stackframes, giving more useufl output.
20 - allow backtraces in coroutines blocked in Coro::EV,
21 at a <1% speed hit.
22 - handle localising of $SIG{__WARN__} and __DIE__
23 properly (with a proper amount of dirty hacking).
24
254.34 Sat Dec 22 17:49:53 CET 2007
26 - upgrade to EV version 2.0 API.
27
284.33 Mon Dec 17 08:36:12 CET 2007
29 - make Coro::AIO etc. loadable in the absence of EV.
30
314.32 Mon Dec 17 07:46:02 CET 2007
32 - majorly improved Coro::Handle's performance with Coro::EV.
33 - implemented the until now mythical Coro::BDB module.
34 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
35
364.31 Wed Dec 5 12:32:39 CET 2007
37 - remove warn statement form Coro::Util.
38
394.3 Tue Dec 4 20:33:14 CET 2007
40 - calls to the idle function could cause stack corruption
41 when the stack changed.
42 - do no longer rely on the presence of EV::DNS if EV is used
43 (because it is gone), but instead take avdantage of EV::ADNS
44 if available.
45 - add ($) prototypes to all functions in Coro::Storable.
46 - use a conventional (safer) idle callback in Coro::EV.
47 - do accept quickly in Coro::Debug to avoid endless loops.
48
494.22 Fri Nov 30 16:04:04 CET 2007
50 - really use optimised versions for Event and EV in Coro::Util
51 and Coro::Handle.
52
534.21 Sun Nov 25 10:48:59 CET 2007
54 - fix a spurious memory read.
55 - Coro::EV no longer keeps the eventloop "alive".
56
574.2 Fri Nov 9 20:47:05 CET 2007
58 - enable/disable tracing from a new coroutine, not a pooled one.
59 - fix a memleak in Coro::Event.
60 - removed killall call from fork_eval.
61 - made sure store_fd is already loaded so that fork_eval
62 does not have to parse autoload in each subprocess.
63 - only use assembly method if -O switch is in $Config{optimize}.
64 - add (optional) Coro::EV module, so far the best event loop module
65 directly supported by Coro.
66 - if the event model is EV, use EV::DNS to resolve
67 stuff in Coro::Util.
68 - don't get confused by multiple event notifications in Coro::Handle.
69 - initial support for EV (libevent interface).
70 - require Event and EV using configure_requires, to force their existance.
71
724.13 Wed Oct 24 07:26:45 CEST 2007
73 - add Coro::Storable::blocking_thaw.
74 - use a vastly more complicated technique to localise
75 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
76 - use a coroutine for the idle callback Coro::Event,
77 instead of running Event in the current coroutine context.
78 This also catches recursive invocations.
79 - actually report fork errors in gethostbyname and inet_aton.
80
814.11 Thu Oct 11 02:40:24 CEST 2007
82 - port to threaded perls.
83
844.1 Thu Oct 11 02:38:16 CEST 2007
85 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
86 be local to each coro (see Coro::State).
87 - incompatible change: for very deep reasons, cede and cede_notself
88 cannot return anything, so nothing will be returned.
89 - possibly bring back 5.10 compatibility (untested).
90 - work around stupid (and wrong) warning on 5.10 :(.
91 - overlay the saved state over the context stack. This saves
92 a few hundred bytes per coroutine on average and also
93 speeds up context switching a bit.
94 - further tune default stack sizes.
95 - (more) correctly calculate stack usage in coro_rss.
96 - Coro::Storable::blocking_* did not properly lock
97 resulting in races between coroutines.
98 - added Coro::Storable::guard.
99 - stopping to trace a coroutine could destroy the cctx of
100 an unrelated coroutine.
101 - explain the relationship between Perl and C coroutines in
102 more detail in Coro::State.
103 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
104 causing a fork per resolve. This also affected Coro::Socket.
105 - switch to a separate stack in $coro->call/eval to avoid
106 invalidating pointers.
107
1084.03 Sat Oct 6 21:24:00 CEST 2007
109 - added Coro::throw method.
110 - minor code cleanups.
111
1124.02 Sat Oct 6 02:36:47 CEST 2007
113 - fix a very minor per-coroutine memleak (a single codereference).
114 - fixed a bug where the currently in-use c context would be freed
115 prematurely (can happen only when programs change the stacksize
116 or use tracing).
117 - tracing can no longer keep a coro alive after it terminated.
118 - do static branch prediction in the common path for gcc. gives
119 about 2-5% speed improvement here.
120
1214.01 Fri Oct 5 22:10:49 CEST 2007
122 - instead of recreating *a* standard output handle we simply
123 use STDOUT, which is faster and hopefully more robust.
124
1254.0 Fri Oct 5 12:56:00 CEST 2007
126 - incompatibly changed Coro::Storable::freeze.
127 - major new feature: added Coro::Debug, for interactive coroutine
128 debugging, tracing and much more.
129 - major bug fix: unbelievable, but true: $_, $/ and many other
130 "saved" variables actually weren't being saved. This has been fixed,
131 of course, while increasing performance while losing all the save
132 flags.
133 - save flags are gone, and all the api functions dealing with them.
134 - added Coro::Semaphore::adjust.
135 - added Coro::Util::fork_eval.
136 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
137 - added Coro::killall.
138 - reduce initial stack sizes to allow for "micro-coroutines".
139 - better async_pool resource management, moved parts of async_pool
140 handling to XS (major speed improvement).
141 - actually croak before modifying important data structures.
142 - refuse to transfer while compiling.
143 - possibly support eval EXPR better now.
144 - enable assembly per default on linux+bsd x86+amd64.
145 - all internal members were renamed _something for easier subclassing.
146 - many minor tweaks.
147
1483.63 Wed May 16 14:10:06 CEST 2007
149 - implement handcoded assembly for x86/amd64 SVR ABI.
150
1513.62 Fri Apr 27 21:36:06 CEST 2007
152 - upgrade libcoro (which might set unwind info correctly).
153 - change default on linux to setjmp/longjmp.
154
1553.61 Thu Apr 19 12:36:18 CEST 2007
156 - Coro::Storable caused an endless loop when thawing invalid
157 pst's sometimes.
158 - use a Semaphore in Coro::Storable, as Storable doesn't
159 seem to be reentrant (although it is documented to
160 be threadsafe...).
161 - fix Coro::Signal to bring back the original unreliable
162 but stateful semantics.
163 - fixed a lot of typos in Coro.pm (patch submitted by David
164 Steinbrunner, which applied flawlessly).
165
1663.6 Sat Apr 14 17:13:31 CEST 2007
167 - added some bugfixes to get eg/myhttpd working again.
168 - added Coro::Storable for often-cede'ing freeze/thaw.
169 - try to do a clean exit when a coroutine calls exit
170 (EXPERIMENTAL).
171 - got rid of indirect call through _coro_init.
172 - updated the partly antique examples in eg/ to
173 work again and be a bit less magic, too.
174 - fixed Coro::Signal semantics to work as documented again.
175
1763.55 Sun Mar 25 01:20:47 CET 2007
177 - add SAVE_DEFFH to save the default filehandle and enable
178 it by default.
179 - finally move socket-operations from Coro::Socket to Coro::Handle
180 to be able to unblock foreign sockets.
181 - add Coro::State::save_also and guarded_save.
182 - add count accessor to Coro::Semaphore.
183 - add Coro::State::cctx_stacksize.
184 - just for the fun of it, do not rely on implicit context,
185 which can dramatically improve performance, but people
186 using windows-process-emulation perls do not care much about
187 performance.
188
1893.51 Sun Mar 4 14:18:04 CET 2007
190 - fixed a problem when you weakened references to Coro::State's
191 (patch by Yuval Kogman).
192
1933.501 Wed Feb 28 12:44:07 CET 2007
194 - rename some global symbols as macosx from hell redefines
195 them without asking.
196
1973.5 Tue Feb 13 20:22:53 CET 2007
198 - do AnyEvent model detection earlier, avoiding problems
199 caused by first using AnyEvent and later Coro::Event.
200 - implement and document Coro::Event event objects.
201 - fix a potential problem in Coro::Event causing crashes.
202 - initialise PL_comppad when creating a new coroutine,
203 avoids crashes on early coro destruction.
204
2053.41 Mon Jan 22 19:19:49 CET 2007
206 - readline on Coro::Handle did not support undefined $/,
207 nor did it deliver partial lines on EOF or error.
208 - implement malloc fallback for stack allocation because
209 stupid broken idiotic OSX has a stupid broken
210 idiotic fits-the-whole-os mmap "implementation" and
211 my dick feels longer if Coro is portable even to
212 obsolete platforms.
213
2143.4 Fri Jan 19 21:52:54 CET 2007
215 - remove t/09_timer.t, as it isn't really testing much
216 but was rather flaky in practise.
217 - async_pool coro would keep arguments and callback alive until
218 it was reused.
219 - cancellation of a coroutine could cause spurious idle calls
220 in cede_notself.
221
2223.3 Sat Jan 6 03:45:00 CET 2007
223 - implement $coro->on_destroy.
224 - Coro::Event blocking semantics have been changed,
225 documented and - hopefully - improved.
226 - fix nice adding, not subtracting, from priority.
227 - fix ->prio and api_is_ready (patch by Mark Hinds).
228 - fixed an assert ("... == prev__cctx->idle_te")
229 that could errronously trigger.
230 - fix various large and small memleaks.
231 - use a (hopefully) more stable cancel implementation
232 that immediately frees the coroutine data.
233 - cede/cede_notself return a status now.
234 - added Coro::guard function.
235 - added a global coroutine pool for jobs (on my machine,
236 I can create and execute 48k simple coros/s with async,
237 and 128k coros with async_pool).
238 - Coro::AIO now uses the coroutine priority as io priority.
239
2403.2 Fri Dec 22 05:07:09 CET 2006
241 - improve portability to slightly older perls.
242 - use cleaner coroutine destruction.
243 - simplify configuration for users.
244 - optionally (unrecommended) prefer perl functions over
245 their coro replacements.
246
2473.11 Tue Dec 5 13:11:24 CET 2006
248 - fixed some bogus assert's, but as perl.h disables assert even
249 without NDEBUG (thank you very much), not too many people should
250 notice (that did include myself). Andreas König noticed, though :)
251 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
252 it. Also noticed by Andreas König :)
253 - save/restore tainted status.
254 - verified to pass the testsuite on my 5.9.5.
255
2563.1 Mon Dec 4 23:03:40 CET 2006
257 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
258 - incompatible change: transfer flags are now per-state.
259 - give a better error message on deadlock.
260 - document Coro::nready.
261 - enhanced testsuite.
262
2633.01 Sun Dec 3 23:47:42 CET 2006
264 - forgot to include Coro::Timer.
265
2663.0 Sun Dec 3 22:57:25 CET 2006
267 - the "FINALLY COMPLETELY STABLE" release (coming soon:
268 the "FAMOUS LAST WORDS" release).
269 - implement a new stack sharing algorithm, which uses a stack
270 pool (size currently hardcoded).
271 - make stack sharing mandatory (it no longer uses a heuristic).
272 - eval/die no longer cause weird problems under heavy use.
273 - Coro::Event could cause livelocks if it was used but
274 no Coro::Event watchers were used.
275 - Coro::Event now uses asynccheck as prepare does not
276 check for changed watchers.
277 - Coro::Event allows multiple waiting coros and will wake up one
278 per event.
279 - Coro::Event should be cleaner and more efficient now.
280 - new utility function Coro::unblock_sub.
281 - document the sad fact that Event is no longer reentrant.
282 - putting a coroutine into the ready queue twice could under
283 some circumstances lead to stack corruption.
284 - minor incompatibility: subclassing Coro::State is supported
285 directly now without going through a _coro_state member.
286 - state/coro switching is much faster now.
287 - very minor optimisations and code/documentation cleanup.
288 - avoid problems due to compiler inlining.
289 - removed timers from Coro::Timer -> use AnyEvent instead.
290 - replaced Coro::idle coroutine by (cleaner) idle handler.
291 - updated to newest libcoro.
292 - implement enhanced support for valgrind.
293 - implement is_ready and return value for ready.
294 - removed Coro::Cont, it was a misnomer (it's generators),
295 and it was somewhat annoying to get it right. Will come back
296 if somebody asks for it :)
297 - many ->wait methods and Coro::Event could return
298 spuriously without the event having happened.
299
3002.5 Tue Nov 7 12:22:33 CET 2006
301 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
302 use AnyEvent, moved them to Coro/.
303 - added Coro::LWP which contains all the uglyness required to
304 make LWP non-blocking.
305 - should work with perl 5.9.x now (Andreas König made me do it).
306 - fixed another bug in Coro::Select when the timeout was undef.
307 - reuse PL_start_env for all coros, saves some memory per coroutine.
308 - manage PL_top_env differently, hopefully to avoid panic: top_env.
309 - timeout argument was not properly used in Coro::Socket.
310 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
311 - emulate undocumented(!) functionality of IO::Socket required
312 by LWP(!!).
313 - updated eg/lwp to work with newer lwp's.
314 - remove "FATAL: uncaught exception" prefix. Coroutines that die
315 kill the whole process, just as exceptions in the main "coroutine"
316 did already.
317
3182.1 Wed Nov 1 23:01:13 CET 2006
319 - fix a long-standing bug in Coro::Select where select with
320 zero timeout would instead change the current default filehandle.
321 - use a simpler and hopefully more robust way to clone padlists
322 (uses less memory and a perl function instead of our own).
323 - coro can now create a stack guard area on many architectures.
324 - Coro::AIO properly reexports additional functions from IO::AIO.
325 - updated libcoro with a workaround for OS X,
326 pach and testing by Michael Schwern.
327
3282.0 Tue Oct 24 05:47:17 CEST 2006
329 - support additional aio requests in Coro::AIO.
330
3311.9 (never properly released due to a glitch)
332
3331.8 Thu Feb 2 00:59:06 CET 2006
334 - applied suggested patch by SAMV to avoid problems during stupid
335 mark & sweep gc run.
336 - applied patch by Scott Walters for 5.9.3 compatibility.
337
3381.7 Tue Dec 27 01:41:58 CET 2005
339 - added Coro::AIO, a thin wrapper around IO::AIO.
340 - improved Makefile.PL explanations.
341
3421.6 Mon Dec 12 21:30:05 CET 2005
343 - additionally save PM_curpm might fix as-of-yet
344 unseen problems with regex matches being attributed to the
345 wrong package.
346 - add t/10_bugs.t, which currently checks against imho broken
347 perls that use 0.26MB of stack space per Perl_magic_get
348 invocation (newer linuxes) as opposed to the 0.0002MB perl
349 normally uses.
350 - make stacksize configurable for the ultimate debian experience.
351
3521.51 Mon Dec 12 18:48:36 CET 2005
353 - remove debugging warn() accidentally left in Coro::Select.
354
3551.5 Tue Nov 29 13:32:44 CET 2005
356 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
357 - fix Coro::Select 'select'.
358 - strict'ify some modules.
359 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
360 - use XSLoader in selected modules.
361 - remove some 5.6 compatibility cruft.
362
3631.4 Tue Sep 6 00:11:05 CEST 2005
364 - libcoro did not take into account the trampoline on amd64,
365 when 'l'inux method was used.
366
3671.31 Tue Aug 30 23:31:33 CEST 2005
368 - some portability fixes/workarounds.
369
3701.3 Sat Aug 20 03:08:56 CEST 2005
371 - no code changes. module seems to work fine.
372 - improve Coro::State docs, remove reference to nonexisting
373 Coro::State::flush method.
374 - no longer autodetect windows, present it as an option instead.
375
3761.2 Mon May 16 02:00:55 CEST 2005
377 - included libcoro.c earlier in State.xs, might improve portability.
378 - use a faster and possibly more stable Coro::Cont implementation.
379 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
380 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
381 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
382
3831.11 Thu Mar 3 18:00:52 CET 2005
384 - change of contact address.
385
3861.1 Tue Feb 22 20:51:16 CET 2005
387 - support [l]inux method on amd64.
388 - allow some context switches while compiling/eval'ing.
389
3901.0 Tue Aug 10 03:47:27 CEST 2004
391 - Event 0.89 fixes the bug again.
392 - fix the remaining(??) memleak. I didn't free the cache-AV for
393 padlists, so one AV per created anon-sub leaked.
394
3950.97 Fri May 14 15:23:32 CEST 2004
396 - remove stupid gcc extension that was an experiment only anyway.
397
3980.96 Thu May 13 18:09:29 CEST 2004
399 - tell people if they have yet another broken Event variant.
400 - made Coro to work _better_ with threads (it seems that, as long
401 as you use Coros in one thread only everything will work fine.
402 Looking at the perl source the problem might be missing locking
403 between Coro and some perl internal routines. That's not easily
404 fixable).
405 - (hopefully) got rid of the static xnext inside transfer.
406
4070.95 Sun Feb 15 03:20:28 CET 2004
408 - removed Coro::State::flush, implemented a new and much
409 faster caching mechanism by attaching magic to the CV.
410 - a memleak with (real) closures remains.
411 - dramatically reduced size of stacks to 65536 (linux/x86), after
412 observing that even myhttpd never had more than a kilobyte
413 of stack in use (stack is still only physically allocated in
414 pagesize increments for systems suporting mmap).
415 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
416 - fix for threaded perl.
417
4180.9 Sun Nov 23 04:32:39 CET 2003
419 - fix a few very hard to track down but (of course) stupid
420 memory leaks.
421 - stringify version number to avoid locale problems :(.
422 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
423 hopefully ignoring it "just works".
424
4250.8 Wed Nov 5 19:38:40 CET 2003
426 - port to perl5.005_03, but only 5.8.x+ is supported!
427 - honor LocalAddr even without LocalPort in Coro::Socket.
428
4290.7 Tue May 27 03:12:38 CEST 2003
430 - the version jump indicates some level of testing, not gobs
431 of new features.
432 - uh... I found the prompt function in ExtUtils::MakeMaker.
433 Highly correct stuff, that is...
434 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
435 the fix is not well understood by the author, so beware :(.
436
4370.652 Thu May 8 02:54:46 CEST 2003
438 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
439 cause version 4 doesn't have ucontext.
440 - Benjamin Reed reported that setjmp works fine on darwin,
441 so preselect it.
442
4430.651 Sat Mar 29 15:00:23 CET 2003
444 - fix a bug in Coro::Handle where some bytes could get lost
445 on reading, reported by jason@nichego.net.
446
4470.65 Sun Mar 23 00:08:26 CET 2003
448 - added a README.linux-glibc.
449 - new module Coro::Select.
450 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
451 (I never claimed that I know what I am doing ;).
452 - set default to setjmp/longjmp on non-x86-linux.
453
4540.6 Thu Nov 21 11:09:06 CET 2002
455 - made lazy context switching the default.
456 - fixed the bug where SemaphoreSet::guard
457 was the same as timed_guard and timed_guard was missing.
458 - fix a memleak in Coro::Timer.
459 - reclassified context sharing as not experimental.
460
4610.534 Sun Apr 14 03:05:12 CEST 2002
462 - fixed backspace => tab in header parsing (myhttpd).
463 - added eg/readline, for elmex.
464 - Coro::Event now calls ->start on first call instead of
465 ->again, which created wrong timeouts for the first ->next.
466 - fixed a bug where Coro::Socket returned a closed socket instead
467 of nothing on ECONNREFUSED and similar error conditions.
468
4690.533 Mon Feb 18 18:49:40 CET 2002
470 - adapt to finally working Event-0.85.
471 - fixed buggy event initialization (trapped under windows, but how
472 could it possibly have worked anywhere?)
473
4740.532 Wed Jan 16 02:45:32 CET 2002
475 - removed the reference to pp_entersub, might work on
476 (some) win32 perls now (testsuite works under cygwin,
477 without Event, which freezes).
478
4790.531 Mon Dec 10 22:18:44 CET 2001
480 - Fixed a bug in SemaphoreSet::timed_down.
481
4820.53 Tue Nov 27 21:11:13 CET 2001
483 - some tests for Event.
484 - slightly different internal architecture to get refcnt into the object
485 and not the reference. might fix a few bugs, certainly did introduce
486 new ones.
487 - Coro::Timer (independent of Event).
488 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
489
4900.52 Tue Nov 6 21:36:18 CET 2001
491 - ported to cygwin (trivial).
492 patch by Gerrit P. Haase <gp@familiehaase.de>.
493 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
494
4950.51 Thu Nov 1 20:39:01 CET 2001
4 - terminate/cancel now work properly (otherwise termination 496 - terminate/cancel now work properly (otherwise termination
5 could cause a "next coroutine is not and contains not..." 497 could cause a "next coroutine is not and contains not..."
6 error. 498 error.
499 - added Coro::Socket::shutdown.
500 - Coro::Event::loop no is the same as Event::loop.
501 - implemented terminate with args + join.
7 502
80.5 Fri Sep 28 16:15:35 CEST 2001 5030.5 Fri Sep 28 16:15:35 CEST 2001
9 - fixed "print" on a Coro::Handle. The print method worked. 504 - fixed "print" on a Coro::Handle. The print method worked.
10 - small tweaks (seem to reduce memory consumption a lot) 505 - small tweaks (seem to reduce memory consumption a lot)
11 in various modules. 506 in various modules.
12 - splendid use of "no warnings" scattered throughout. 507 - splendid use of "no warnings" scattered throughout.
13 - added Coro::Handle::rbuf, fixed Coro::Handle::fh. 508 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines