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

Comparing Coro/Changes (file contents):
Revision 1.58 by root, Tue Nov 6 20:34:09 2001 UTC vs.
Revision 1.291 by root, Fri Nov 23 16:35:20 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
30.52 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
8 - fix a spurious memory read.
9 - Coro::EV no longer keeps the eventloop "alive".
10
114.2 Fri Nov 9 20:47:05 CET 2007
12 - enable/disable tracing from a new coroutine, not a pooled one.
13 - fix a memleak in Coro::Event.
14 - removed killall call from fork_eval.
15 - made sure store_fd is already loaded so that fork_eval
16 does not have to parse autoload in each subprocess.
17 - only use assembly method if -O switch is in $Config{optimize}.
18 - add (optional) Coro::EV module, so far the best event loop module
19 directly supported by Coro.
20 - if the event model is EV, use EV::DNS to resolve
21 stuff in Coro::Util.
22 - don't get confused by multiple event notifications in Coro::Handle.
23 - initial support for EV (libevent interface).
24 - require Event and EV using configure_requires, to force their existance.
25
264.13 Wed Oct 24 07:26:45 CEST 2007
27 - add Coro::Storable::blocking_thaw.
28 - use a vastly more complicated technique to localise
29 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
30 - use a coroutine for the idle callback Coro::Event,
31 instead of running Event in the current coroutine context.
32 This also catches recursive invocations.
33 - actually report fork errors in gethostbyname and inet_aton.
34
354.11 Thu Oct 11 02:40:24 CEST 2007
36 - port to threaded perls.
37
384.1 Thu Oct 11 02:38:16 CEST 2007
39 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
40 be local to each coro (see Coro::State).
41 - incompatible change: for very deep reasons, cede and cede_notself
42 cannot return anything, so nothing will be returned.
43 - possibly bring back 5.10 compatibility (untested).
44 - work around stupid (and wrong) warning on 5.10 :(.
45 - overlay the saved state over the context stack. This saves
46 a few hundred bytes per coroutine on average and also
47 speeds up context switching a bit.
48 - further tune default stack sizes.
49 - (more) correctly calculate stack usage in coro_rss.
50 - Coro::Storable::blocking_* did not properly lock
51 resulting in races between coroutines.
52 - added Coro::Storable::guard.
53 - stopping to trace a coroutine could destroy the cctx of
54 an unrelated coroutine.
55 - explain the relationship between Perl and C coroutines in
56 more detail in Coro::State.
57 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
58 causing a fork per resolve. This also affected Coro::Socket.
59 - switch to a separate stack in $coro->call/eval to avoid
60 invalidating pointers.
61
624.03 Sat Oct 6 21:24:00 CEST 2007
63 - added Coro::throw method.
64 - minor code cleanups.
65
664.02 Sat Oct 6 02:36:47 CEST 2007
67 - fix a very minor per-coroutine memleak (a single codereference).
68 - fixed a bug where the currently in-use c context would be freed
69 prematurely (can happen only when programs change the stacksize
70 or use tracing).
71 - tracing can no longer keep a coro alive after it terminated.
72 - do static branch prediction in the common path for gcc. gives
73 about 2-5% speed improvement here.
74
754.01 Fri Oct 5 22:10:49 CEST 2007
76 - instead of recreating *a* standard output handle we simply
77 use STDOUT, which is faster and hopefully more robust.
78
794.0 Fri Oct 5 12:56:00 CEST 2007
80 - incompatibly changed Coro::Storable::freeze.
81 - major new feature: added Coro::Debug, for interactive coroutine
82 debugging, tracing and much more.
83 - major bug fix: unbelievable, but true: $_, $/ and many other
84 "saved" variables actually weren't being saved. This has been fixed,
85 of course, while increasing performance while losing all the save
86 flags.
87 - save flags are gone, and all the api functions dealing with them.
88 - added Coro::Semaphore::adjust.
89 - added Coro::Util::fork_eval.
90 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
91 - added Coro::killall.
92 - reduce initial stack sizes to allow for "micro-coroutines".
93 - better async_pool resource management, moved parts of async_pool
94 handling to XS (major speed improvement).
95 - actually croak before modifying important data structures.
96 - refuse to transfer while compiling.
97 - possibly support eval EXPR better now.
98 - enable assembly per default on linux+bsd x86+amd64.
99 - all internal members were renamed _something for easier subclassing.
100 - many minor tweaks.
101
1023.63 Wed May 16 14:10:06 CEST 2007
103 - implement handcoded assembly for x86/amd64 SVR ABI.
104
1053.62 Fri Apr 27 21:36:06 CEST 2007
106 - upgrade libcoro (which might set unwind info correctly).
107 - change default on linux to setjmp/longjmp.
108
1093.61 Thu Apr 19 12:36:18 CEST 2007
110 - Coro::Storable caused an endless loop when thawing invalid
111 pst's sometimes.
112 - use a Semaphore in Coro::Storable, as Storable doesn't
113 seem to be reentrant (although it is documented to
114 be threadsafe...).
115 - fix Coro::Signal to bring back the original unreliable
116 but stateful semantics.
117 - fixed a lot of typos in Coro.pm (patch submitted by David
118 Steinbrunner, which applied flawlessly).
119
1203.6 Sat Apr 14 17:13:31 CEST 2007
121 - added some bugfixes to get eg/myhttpd working again.
122 - added Coro::Storable for often-cede'ing freeze/thaw.
123 - try to do a clean exit when a coroutine calls exit
124 (EXPERIMENTAL).
125 - got rid of indirect call through _coro_init.
126 - updated the partly antique examples in eg/ to
127 work again and be a bit less magic, too.
128 - fixed Coro::Signal semantics to work as documented again.
129
1303.55 Sun Mar 25 01:20:47 CET 2007
131 - add SAVE_DEFFH to save the default filehandle and enable
132 it by default.
133 - finally move socket-operations from Coro::Socket to Coro::Handle
134 to be able to unblock foreign sockets.
135 - add Coro::State::save_also and guarded_save.
136 - add count accessor to Coro::Semaphore.
137 - add Coro::State::cctx_stacksize.
138 - just for the fun of it, do not rely on implicit context,
139 which can dramatically improve performance, but people
140 using windows-process-emulation perls do not care much about
141 performance.
142
1433.51 Sun Mar 4 14:18:04 CET 2007
144 - fixed a problem when you weakened references to Coro::State's
145 (patch by Yuval Kogman).
146
1473.501 Wed Feb 28 12:44:07 CET 2007
148 - rename some global symbols as macosx from hell redefines
149 them without asking.
150
1513.5 Tue Feb 13 20:22:53 CET 2007
152 - do AnyEvent model detection earlier, avoiding problems
153 caused by first using AnyEvent and later Coro::Event.
154 - implement and document Coro::Event event objects.
155 - fix a potential problem in Coro::Event causing crashes.
156 - initialise PL_comppad when creating a new coroutine,
157 avoids crashes on early coro destruction.
158
1593.41 Mon Jan 22 19:19:49 CET 2007
160 - readline on Coro::Handle did not support undefined $/,
161 nor did it deliver partial lines on EOF or error.
162 - implement malloc fallback for stack allocation because
163 stupid broken idiotic OSX has a stupid broken
164 idiotic fits-the-whole-os mmap "implementation" and
165 my dick feels longer if Coro is portable even to
166 obsolete platforms.
167
1683.4 Fri Jan 19 21:52:54 CET 2007
169 - remove t/09_timer.t, as it isn't really testing much
170 but was rather flaky in practise.
171 - async_pool coro would keep arguments and callback alive until
172 it was reused.
173 - cancellation of a coroutine could cause spurious idle calls
174 in cede_notself.
175
1763.3 Sat Jan 6 03:45:00 CET 2007
177 - implement $coro->on_destroy.
178 - Coro::Event blocking semantics have been changed,
179 documented and - hopefully - improved.
180 - fix nice adding, not subtracting, from priority.
181 - fix ->prio and api_is_ready (patch by Mark Hinds).
182 - fixed an assert ("... == prev__cctx->idle_te")
183 that could errronously trigger.
184 - fix various large and small memleaks.
185 - use a (hopefully) more stable cancel implementation
186 that immediately frees the coroutine data.
187 - cede/cede_notself return a status now.
188 - added Coro::guard function.
189 - added a global coroutine pool for jobs (on my machine,
190 I can create and execute 48k simple coros/s with async,
191 and 128k coros with async_pool).
192 - Coro::AIO now uses the coroutine priority as io priority.
193
1943.2 Fri Dec 22 05:07:09 CET 2006
195 - improve portability to slightly older perls.
196 - use cleaner coroutine destruction.
197 - simplify configuration for users.
198 - optionally (unrecommended) prefer perl functions over
199 their coro replacements.
200
2013.11 Tue Dec 5 13:11:24 CET 2006
202 - fixed some bogus assert's, but as perl.h disables assert even
203 without NDEBUG (thank you very much), not too many people should
204 notice (that did include myself). Andreas König noticed, though :)
205 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
206 it. Also noticed by Andreas König :)
207 - save/restore tainted status.
208 - verified to pass the testsuite on my 5.9.5.
209
2103.1 Mon Dec 4 23:03:40 CET 2006
211 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
212 - incompatible change: transfer flags are now per-state.
213 - give a better error message on deadlock.
214 - document Coro::nready.
215 - enhanced testsuite.
216
2173.01 Sun Dec 3 23:47:42 CET 2006
218 - forgot to include Coro::Timer.
219
2203.0 Sun Dec 3 22:57:25 CET 2006
221 - the "FINALLY COMPLETELY STABLE" release (coming soon:
222 the "FAMOUS LAST WORDS" release).
223 - implement a new stack sharing algorithm, which uses a stack
224 pool (size currently hardcoded).
225 - make stack sharing mandatory (it no longer uses a heuristic).
226 - eval/die no longer cause weird problems under heavy use.
227 - Coro::Event could cause livelocks if it was used but
228 no Coro::Event watchers were used.
229 - Coro::Event now uses asynccheck as prepare does not
230 check for changed watchers.
231 - Coro::Event allows multiple waiting coros and will wake up one
232 per event.
233 - Coro::Event should be cleaner and more efficient now.
234 - new utility function Coro::unblock_sub.
235 - document the sad fact that Event is no longer reentrant.
236 - putting a coroutine into the ready queue twice could under
237 some circumstances lead to stack corruption.
238 - minor incompatibility: subclassing Coro::State is supported
239 directly now without going through a _coro_state member.
240 - state/coro switching is much faster now.
241 - very minor optimisations and code/documentation cleanup.
242 - avoid problems due to compiler inlining.
243 - removed timers from Coro::Timer -> use AnyEvent instead.
244 - replaced Coro::idle coroutine by (cleaner) idle handler.
245 - updated to newest libcoro.
246 - implement enhanced support for valgrind.
247 - implement is_ready and return value for ready.
248 - removed Coro::Cont, it was a misnomer (it's generators),
249 and it was somewhat annoying to get it right. Will come back
250 if somebody asks for it :)
251 - many ->wait methods and Coro::Event could return
252 spuriously without the event having happened.
253
2542.5 Tue Nov 7 12:22:33 CET 2006
255 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
256 use AnyEvent, moved them to Coro/.
257 - added Coro::LWP which contains all the uglyness required to
258 make LWP non-blocking.
259 - should work with perl 5.9.x now (Andreas König made me do it).
260 - fixed another bug in Coro::Select when the timeout was undef.
261 - reuse PL_start_env for all coros, saves some memory per coroutine.
262 - manage PL_top_env differently, hopefully to avoid panic: top_env.
263 - timeout argument was not properly used in Coro::Socket.
264 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
265 - emulate undocumented(!) functionality of IO::Socket required
266 by LWP(!!).
267 - updated eg/lwp to work with newer lwp's.
268 - remove "FATAL: uncaught exception" prefix. Coroutines that die
269 kill the whole process, just as exceptions in the main "coroutine"
270 did already.
271
2722.1 Wed Nov 1 23:01:13 CET 2006
273 - fix a long-standing bug in Coro::Select where select with
274 zero timeout would instead change the current default filehandle.
275 - use a simpler and hopefully more robust way to clone padlists
276 (uses less memory and a perl function instead of our own).
277 - coro can now create a stack guard area on many architectures.
278 - Coro::AIO properly reexports additional functions from IO::AIO.
279 - updated libcoro with a workaround for OS X,
280 pach and testing by Michael Schwern.
281
2822.0 Tue Oct 24 05:47:17 CEST 2006
283 - support additional aio requests in Coro::AIO.
284
2851.9 (never properly released due to a glitch)
286
2871.8 Thu Feb 2 00:59:06 CET 2006
288 - applied suggested patch by SAMV to avoid problems during stupid
289 mark & sweep gc run.
290 - applied patch by Scott Walters for 5.9.3 compatibility.
291
2921.7 Tue Dec 27 01:41:58 CET 2005
293 - added Coro::AIO, a thin wrapper around IO::AIO.
294 - improved Makefile.PL explanations.
295
2961.6 Mon Dec 12 21:30:05 CET 2005
297 - additionally save PM_curpm might fix as-of-yet
298 unseen problems with regex matches being attributed to the
299 wrong package.
300 - add t/10_bugs.t, which currently checks against imho broken
301 perls that use 0.26MB of stack space per Perl_magic_get
302 invocation (newer linuxes) as opposed to the 0.0002MB perl
303 normally uses.
304 - make stacksize configurable for the ultimate debian experience.
305
3061.51 Mon Dec 12 18:48:36 CET 2005
307 - remove debugging warn() accidentally left in Coro::Select.
308
3091.5 Tue Nov 29 13:32:44 CET 2005
310 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
311 - fix Coro::Select 'select'.
312 - strict'ify some modules.
313 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
314 - use XSLoader in selected modules.
315 - remove some 5.6 compatibility cruft.
316
3171.4 Tue Sep 6 00:11:05 CEST 2005
318 - libcoro did not take into account the trampoline on amd64,
319 when 'l'inux method was used.
320
3211.31 Tue Aug 30 23:31:33 CEST 2005
322 - some portability fixes/workarounds.
323
3241.3 Sat Aug 20 03:08:56 CEST 2005
325 - no code changes. module seems to work fine.
326 - improve Coro::State docs, remove reference to nonexisting
327 Coro::State::flush method.
328 - no longer autodetect windows, present it as an option instead.
329
3301.2 Mon May 16 02:00:55 CEST 2005
331 - included libcoro.c earlier in State.xs, might improve portability.
332 - use a faster and possibly more stable Coro::Cont implementation.
333 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
334 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
335 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
336
3371.11 Thu Mar 3 18:00:52 CET 2005
338 - change of contact address.
339
3401.1 Tue Feb 22 20:51:16 CET 2005
341 - support [l]inux method on amd64.
342 - allow some context switches while compiling/eval'ing.
343
3441.0 Tue Aug 10 03:47:27 CEST 2004
345 - Event 0.89 fixes the bug again.
346 - fix the remaining(??) memleak. I didn't free the cache-AV for
347 padlists, so one AV per created anon-sub leaked.
348
3490.97 Fri May 14 15:23:32 CEST 2004
350 - remove stupid gcc extension that was an experiment only anyway.
351
3520.96 Thu May 13 18:09:29 CEST 2004
353 - tell people if they have yet another broken Event variant.
354 - made Coro to work _better_ with threads (it seems that, as long
355 as you use Coros in one thread only everything will work fine.
356 Looking at the perl source the problem might be missing locking
357 between Coro and some perl internal routines. That's not easily
358 fixable).
359 - (hopefully) got rid of the static xnext inside transfer.
360
3610.95 Sun Feb 15 03:20:28 CET 2004
362 - removed Coro::State::flush, implemented a new and much
363 faster caching mechanism by attaching magic to the CV.
364 - a memleak with (real) closures remains.
365 - dramatically reduced size of stacks to 65536 (linux/x86), after
366 observing that even myhttpd never had more than a kilobyte
367 of stack in use (stack is still only physically allocated in
368 pagesize increments for systems suporting mmap).
369 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
370 - fix for threaded perl.
371
3720.9 Sun Nov 23 04:32:39 CET 2003
373 - fix a few very hard to track down but (of course) stupid
374 memory leaks.
375 - stringify version number to avoid locale problems :(.
376 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
377 hopefully ignoring it "just works".
378
3790.8 Wed Nov 5 19:38:40 CET 2003
380 - port to perl5.005_03, but only 5.8.x+ is supported!
381 - honor LocalAddr even without LocalPort in Coro::Socket.
382
3830.7 Tue May 27 03:12:38 CEST 2003
384 - the version jump indicates some level of testing, not gobs
385 of new features.
386 - uh... I found the prompt function in ExtUtils::MakeMaker.
387 Highly correct stuff, that is...
388 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
389 the fix is not well understood by the author, so beware :(.
390
3910.652 Thu May 8 02:54:46 CEST 2003
392 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
393 cause version 4 doesn't have ucontext.
394 - Benjamin Reed reported that setjmp works fine on darwin,
395 so preselect it.
396
3970.651 Sat Mar 29 15:00:23 CET 2003
398 - fix a bug in Coro::Handle where some bytes could get lost
399 on reading, reported by jason@nichego.net.
400
4010.65 Sun Mar 23 00:08:26 CET 2003
402 - added a README.linux-glibc.
403 - new module Coro::Select.
404 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
405 (I never claimed that I know what I am doing ;).
406 - set default to setjmp/longjmp on non-x86-linux.
407
4080.6 Thu Nov 21 11:09:06 CET 2002
409 - made lazy context switching the default.
410 - fixed the bug where SemaphoreSet::guard
411 was the same as timed_guard and timed_guard was missing.
412 - fix a memleak in Coro::Timer.
413 - reclassified context sharing as not experimental.
414
4150.534 Sun Apr 14 03:05:12 CEST 2002
416 - fixed backspace => tab in header parsing (myhttpd).
417 - added eg/readline, for elmex.
418 - Coro::Event now calls ->start on first call instead of
419 ->again, which created wrong timeouts for the first ->next.
420 - fixed a bug where Coro::Socket returned a closed socket instead
421 of nothing on ECONNREFUSED and similar error conditions.
422
4230.533 Mon Feb 18 18:49:40 CET 2002
424 - adapt to finally working Event-0.85.
425 - fixed buggy event initialization (trapped under windows, but how
426 could it possibly have worked anywhere?)
427
4280.532 Wed Jan 16 02:45:32 CET 2002
429 - removed the reference to pp_entersub, might work on
430 (some) win32 perls now (testsuite works under cygwin,
431 without Event, which freezes).
432
4330.531 Mon Dec 10 22:18:44 CET 2001
434 - Fixed a bug in SemaphoreSet::timed_down.
435
4360.53 Tue Nov 27 21:11:13 CET 2001
437 - some tests for Event.
438 - slightly different internal architecture to get refcnt into the object
439 and not the reference. might fix a few bugs, certainly did introduce
440 new ones.
441 - Coro::Timer (independent of Event).
442 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
443
4440.52 Tue Nov 6 21:36:18 CET 2001
4 - ported to cygwin (trivial). 445 - ported to cygwin (trivial).
5 patch by Gerrit P. Haase <gp@familiehaase.de>. 446 patch by Gerrit P. Haase <gp@familiehaase.de>.
6 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov. 447 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
7 448
80.51 Thu Nov 1 20:39:01 CET 2001 4490.51 Thu Nov 1 20:39:01 CET 2001
9 - terminate/cancel now work properly (otherwise termination 450 - terminate/cancel now work properly (otherwise termination
10 could cause a "next coroutine is not and contains not..." 451 could cause a "next coroutine is not and contains not..."
11 error. 452 error.
12 - added Coro::Socket::shutdown. 453 - added Coro::Socket::shutdown.
13 - Coro::Event::loop no is the same as Event::loop. 454 - Coro::Event::loop no is the same as Event::loop.
14 - implemented terminate with args + join. 455 - implemented terminate with args + join.
15 456
160.5 Fri Sep 28 16:15:35 CEST 2001 4570.5 Fri Sep 28 16:15:35 CEST 2001
17 - fixed "print" on a Coro::Handle. The print method worked. 458 - fixed "print" on a Coro::Handle. The print method worked.
18 - small tweaks (seem to reduce memory consumption a lot) 459 - small tweaks (seem to reduce memory consumption a lot)
19 in various modules. 460 in various modules.
20 - splendid use of "no warnings" scattered throughout. 461 - splendid use of "no warnings" scattered throughout.
21 - added Coro::Handle::rbuf, fixed Coro::Handle::fh. 462 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines