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

Comparing Coro/Changes (file contents):
Revision 1.88 by pcg, Sat Feb 14 12:39:41 2004 UTC vs.
Revision 1.289 by root, Fri Nov 9 19:50:15 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
30.95 Fri Dec 12 20:50:57 CET 2003 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
84.2 Fri Nov 9 20:47:05 CET 2007
9 - enable/disable tracing from a new coroutine, not a pooled one.
10 - fix a memleak in Coro::Event.
11 - removed killall call from fork_eval.
12 - made sure store_fd is already loaded so that fork_eval
13 does not have to parse autoload in each subprocess.
14 - only use assembly method if -O switch is in $Config{optimize}.
15 - add (optional) Coro::EV module, so far the best event loop module
16 directly supported by Coro.
17 - if the event model is EV, use EV::DNS to resolve
18 stuff in Coro::Util.
19 - don't get confused by multiple event notifications in Coro::Handle.
20 - initial support for EV (libevent interface).
21 - require Event and EV using configure_requires, to force their existance.
22
234.13 Wed Oct 24 07:26:45 CEST 2007
24 - add Coro::Storable::blocking_thaw.
25 - use a vastly more complicated technique to localise
26 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
27 - use a coroutine for the idle callback Coro::Event,
28 instead of running Event in the current coroutine context.
29 This also catches recursive invocations.
30 - actually report fork errors in gethostbyname and inet_aton.
31
324.11 Thu Oct 11 02:40:24 CEST 2007
33 - port to threaded perls.
34
354.1 Thu Oct 11 02:38:16 CEST 2007
36 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
37 be local to each coro (see Coro::State).
38 - incompatible change: for very deep reasons, cede and cede_notself
39 cannot return anything, so nothing will be returned.
40 - possibly bring back 5.10 compatibility (untested).
41 - work around stupid (and wrong) warning on 5.10 :(.
42 - overlay the saved state over the context stack. This saves
43 a few hundred bytes per coroutine on average and also
44 speeds up context switching a bit.
45 - further tune default stack sizes.
46 - (more) correctly calculate stack usage in coro_rss.
47 - Coro::Storable::blocking_* did not properly lock
48 resulting in races between coroutines.
49 - added Coro::Storable::guard.
50 - stopping to trace a coroutine could destroy the cctx of
51 an unrelated coroutine.
52 - explain the relationship between Perl and C coroutines in
53 more detail in Coro::State.
54 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
55 causing a fork per resolve. This also affected Coro::Socket.
56 - switch to a separate stack in $coro->call/eval to avoid
57 invalidating pointers.
58
594.03 Sat Oct 6 21:24:00 CEST 2007
60 - added Coro::throw method.
61 - minor code cleanups.
62
634.02 Sat Oct 6 02:36:47 CEST 2007
64 - fix a very minor per-coroutine memleak (a single codereference).
65 - fixed a bug where the currently in-use c context would be freed
66 prematurely (can happen only when programs change the stacksize
67 or use tracing).
68 - tracing can no longer keep a coro alive after it terminated.
69 - do static branch prediction in the common path for gcc. gives
70 about 2-5% speed improvement here.
71
724.01 Fri Oct 5 22:10:49 CEST 2007
73 - instead of recreating *a* standard output handle we simply
74 use STDOUT, which is faster and hopefully more robust.
75
764.0 Fri Oct 5 12:56:00 CEST 2007
77 - incompatibly changed Coro::Storable::freeze.
78 - major new feature: added Coro::Debug, for interactive coroutine
79 debugging, tracing and much more.
80 - major bug fix: unbelievable, but true: $_, $/ and many other
81 "saved" variables actually weren't being saved. This has been fixed,
82 of course, while increasing performance while losing all the save
83 flags.
84 - save flags are gone, and all the api functions dealing with them.
85 - added Coro::Semaphore::adjust.
86 - added Coro::Util::fork_eval.
87 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
88 - added Coro::killall.
89 - reduce initial stack sizes to allow for "micro-coroutines".
90 - better async_pool resource management, moved parts of async_pool
91 handling to XS (major speed improvement).
92 - actually croak before modifying important data structures.
93 - refuse to transfer while compiling.
94 - possibly support eval EXPR better now.
95 - enable assembly per default on linux+bsd x86+amd64.
96 - all internal members were renamed _something for easier subclassing.
97 - many minor tweaks.
98
993.63 Wed May 16 14:10:06 CEST 2007
100 - implement handcoded assembly for x86/amd64 SVR ABI.
101
1023.62 Fri Apr 27 21:36:06 CEST 2007
103 - upgrade libcoro (which might set unwind info correctly).
104 - change default on linux to setjmp/longjmp.
105
1063.61 Thu Apr 19 12:36:18 CEST 2007
107 - Coro::Storable caused an endless loop when thawing invalid
108 pst's sometimes.
109 - use a Semaphore in Coro::Storable, as Storable doesn't
110 seem to be reentrant (although it is documented to
111 be threadsafe...).
112 - fix Coro::Signal to bring back the original unreliable
113 but stateful semantics.
114 - fixed a lot of typos in Coro.pm (patch submitted by David
115 Steinbrunner, which applied flawlessly).
116
1173.6 Sat Apr 14 17:13:31 CEST 2007
118 - added some bugfixes to get eg/myhttpd working again.
119 - added Coro::Storable for often-cede'ing freeze/thaw.
120 - try to do a clean exit when a coroutine calls exit
121 (EXPERIMENTAL).
122 - got rid of indirect call through _coro_init.
123 - updated the partly antique examples in eg/ to
124 work again and be a bit less magic, too.
125 - fixed Coro::Signal semantics to work as documented again.
126
1273.55 Sun Mar 25 01:20:47 CET 2007
128 - add SAVE_DEFFH to save the default filehandle and enable
129 it by default.
130 - finally move socket-operations from Coro::Socket to Coro::Handle
131 to be able to unblock foreign sockets.
132 - add Coro::State::save_also and guarded_save.
133 - add count accessor to Coro::Semaphore.
134 - add Coro::State::cctx_stacksize.
135 - just for the fun of it, do not rely on implicit context,
136 which can dramatically improve performance, but people
137 using windows-process-emulation perls do not care much about
138 performance.
139
1403.51 Sun Mar 4 14:18:04 CET 2007
141 - fixed a problem when you weakened references to Coro::State's
142 (patch by Yuval Kogman).
143
1443.501 Wed Feb 28 12:44:07 CET 2007
145 - rename some global symbols as macosx from hell redefines
146 them without asking.
147
1483.5 Tue Feb 13 20:22:53 CET 2007
149 - do AnyEvent model detection earlier, avoiding problems
150 caused by first using AnyEvent and later Coro::Event.
151 - implement and document Coro::Event event objects.
152 - fix a potential problem in Coro::Event causing crashes.
153 - initialise PL_comppad when creating a new coroutine,
154 avoids crashes on early coro destruction.
155
1563.41 Mon Jan 22 19:19:49 CET 2007
157 - readline on Coro::Handle did not support undefined $/,
158 nor did it deliver partial lines on EOF or error.
159 - implement malloc fallback for stack allocation because
160 stupid broken idiotic OSX has a stupid broken
161 idiotic fits-the-whole-os mmap "implementation" and
162 my dick feels longer if Coro is portable even to
163 obsolete platforms.
164
1653.4 Fri Jan 19 21:52:54 CET 2007
166 - remove t/09_timer.t, as it isn't really testing much
167 but was rather flaky in practise.
168 - async_pool coro would keep arguments and callback alive until
169 it was reused.
170 - cancellation of a coroutine could cause spurious idle calls
171 in cede_notself.
172
1733.3 Sat Jan 6 03:45:00 CET 2007
174 - implement $coro->on_destroy.
175 - Coro::Event blocking semantics have been changed,
176 documented and - hopefully - improved.
177 - fix nice adding, not subtracting, from priority.
178 - fix ->prio and api_is_ready (patch by Mark Hinds).
179 - fixed an assert ("... == prev__cctx->idle_te")
180 that could errronously trigger.
181 - fix various large and small memleaks.
182 - use a (hopefully) more stable cancel implementation
183 that immediately frees the coroutine data.
184 - cede/cede_notself return a status now.
185 - added Coro::guard function.
186 - added a global coroutine pool for jobs (on my machine,
187 I can create and execute 48k simple coros/s with async,
188 and 128k coros with async_pool).
189 - Coro::AIO now uses the coroutine priority as io priority.
190
1913.2 Fri Dec 22 05:07:09 CET 2006
192 - improve portability to slightly older perls.
193 - use cleaner coroutine destruction.
194 - simplify configuration for users.
195 - optionally (unrecommended) prefer perl functions over
196 their coro replacements.
197
1983.11 Tue Dec 5 13:11:24 CET 2006
199 - fixed some bogus assert's, but as perl.h disables assert even
200 without NDEBUG (thank you very much), not too many people should
201 notice (that did include myself). Andreas König noticed, though :)
202 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
203 it. Also noticed by Andreas König :)
204 - save/restore tainted status.
205 - verified to pass the testsuite on my 5.9.5.
206
2073.1 Mon Dec 4 23:03:40 CET 2006
208 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
209 - incompatible change: transfer flags are now per-state.
210 - give a better error message on deadlock.
211 - document Coro::nready.
212 - enhanced testsuite.
213
2143.01 Sun Dec 3 23:47:42 CET 2006
215 - forgot to include Coro::Timer.
216
2173.0 Sun Dec 3 22:57:25 CET 2006
218 - the "FINALLY COMPLETELY STABLE" release (coming soon:
219 the "FAMOUS LAST WORDS" release).
220 - implement a new stack sharing algorithm, which uses a stack
221 pool (size currently hardcoded).
222 - make stack sharing mandatory (it no longer uses a heuristic).
223 - eval/die no longer cause weird problems under heavy use.
224 - Coro::Event could cause livelocks if it was used but
225 no Coro::Event watchers were used.
226 - Coro::Event now uses asynccheck as prepare does not
227 check for changed watchers.
228 - Coro::Event allows multiple waiting coros and will wake up one
229 per event.
230 - Coro::Event should be cleaner and more efficient now.
231 - new utility function Coro::unblock_sub.
232 - document the sad fact that Event is no longer reentrant.
233 - putting a coroutine into the ready queue twice could under
234 some circumstances lead to stack corruption.
235 - minor incompatibility: subclassing Coro::State is supported
236 directly now without going through a _coro_state member.
237 - state/coro switching is much faster now.
238 - very minor optimisations and code/documentation cleanup.
239 - avoid problems due to compiler inlining.
240 - removed timers from Coro::Timer -> use AnyEvent instead.
241 - replaced Coro::idle coroutine by (cleaner) idle handler.
242 - updated to newest libcoro.
243 - implement enhanced support for valgrind.
244 - implement is_ready and return value for ready.
245 - removed Coro::Cont, it was a misnomer (it's generators),
246 and it was somewhat annoying to get it right. Will come back
247 if somebody asks for it :)
248 - many ->wait methods and Coro::Event could return
249 spuriously without the event having happened.
250
2512.5 Tue Nov 7 12:22:33 CET 2006
252 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
253 use AnyEvent, moved them to Coro/.
254 - added Coro::LWP which contains all the uglyness required to
255 make LWP non-blocking.
256 - should work with perl 5.9.x now (Andreas König made me do it).
257 - fixed another bug in Coro::Select when the timeout was undef.
258 - reuse PL_start_env for all coros, saves some memory per coroutine.
259 - manage PL_top_env differently, hopefully to avoid panic: top_env.
260 - timeout argument was not properly used in Coro::Socket.
261 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
262 - emulate undocumented(!) functionality of IO::Socket required
263 by LWP(!!).
264 - updated eg/lwp to work with newer lwp's.
265 - remove "FATAL: uncaught exception" prefix. Coroutines that die
266 kill the whole process, just as exceptions in the main "coroutine"
267 did already.
268
2692.1 Wed Nov 1 23:01:13 CET 2006
270 - fix a long-standing bug in Coro::Select where select with
271 zero timeout would instead change the current default filehandle.
272 - use a simpler and hopefully more robust way to clone padlists
273 (uses less memory and a perl function instead of our own).
274 - coro can now create a stack guard area on many architectures.
275 - Coro::AIO properly reexports additional functions from IO::AIO.
276 - updated libcoro with a workaround for OS X,
277 pach and testing by Michael Schwern.
278
2792.0 Tue Oct 24 05:47:17 CEST 2006
280 - support additional aio requests in Coro::AIO.
281
2821.9 (never properly released due to a glitch)
283
2841.8 Thu Feb 2 00:59:06 CET 2006
285 - applied suggested patch by SAMV to avoid problems during stupid
286 mark & sweep gc run.
287 - applied patch by Scott Walters for 5.9.3 compatibility.
288
2891.7 Tue Dec 27 01:41:58 CET 2005
290 - added Coro::AIO, a thin wrapper around IO::AIO.
291 - improved Makefile.PL explanations.
292
2931.6 Mon Dec 12 21:30:05 CET 2005
294 - additionally save PM_curpm might fix as-of-yet
295 unseen problems with regex matches being attributed to the
296 wrong package.
297 - add t/10_bugs.t, which currently checks against imho broken
298 perls that use 0.26MB of stack space per Perl_magic_get
299 invocation (newer linuxes) as opposed to the 0.0002MB perl
300 normally uses.
301 - make stacksize configurable for the ultimate debian experience.
302
3031.51 Mon Dec 12 18:48:36 CET 2005
304 - remove debugging warn() accidentally left in Coro::Select.
305
3061.5 Tue Nov 29 13:32:44 CET 2005
307 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
308 - fix Coro::Select 'select'.
309 - strict'ify some modules.
310 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
311 - use XSLoader in selected modules.
312 - remove some 5.6 compatibility cruft.
313
3141.4 Tue Sep 6 00:11:05 CEST 2005
315 - libcoro did not take into account the trampoline on amd64,
316 when 'l'inux method was used.
317
3181.31 Tue Aug 30 23:31:33 CEST 2005
319 - some portability fixes/workarounds.
320
3211.3 Sat Aug 20 03:08:56 CEST 2005
322 - no code changes. module seems to work fine.
323 - improve Coro::State docs, remove reference to nonexisting
324 Coro::State::flush method.
325 - no longer autodetect windows, present it as an option instead.
326
3271.2 Mon May 16 02:00:55 CEST 2005
328 - included libcoro.c earlier in State.xs, might improve portability.
329 - use a faster and possibly more stable Coro::Cont implementation.
330 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
331 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
332 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
333
3341.11 Thu Mar 3 18:00:52 CET 2005
335 - change of contact address.
336
3371.1 Tue Feb 22 20:51:16 CET 2005
338 - support [l]inux method on amd64.
339 - allow some context switches while compiling/eval'ing.
340
3411.0 Tue Aug 10 03:47:27 CEST 2004
342 - Event 0.89 fixes the bug again.
343 - fix the remaining(??) memleak. I didn't free the cache-AV for
344 padlists, so one AV per created anon-sub leaked.
345
3460.97 Fri May 14 15:23:32 CEST 2004
347 - remove stupid gcc extension that was an experiment only anyway.
348
3490.96 Thu May 13 18:09:29 CEST 2004
350 - tell people if they have yet another broken Event variant.
351 - made Coro to work _better_ with threads (it seems that, as long
352 as you use Coros in one thread only everything will work fine.
353 Looking at the perl source the problem might be missing locking
354 between Coro and some perl internal routines. That's not easily
355 fixable).
356 - (hopefully) got rid of the static xnext inside transfer.
357
3580.95 Sun Feb 15 03:20:28 CET 2004
4 - removed Coro::State::flush, implemented a new and much 359 - removed Coro::State::flush, implemented a new and much
5 faster caching mechanism by attaching magic to the CV. 360 faster caching mechanism by attaching magic to the CV.
6 - a memleak with (real) closures remains. 361 - a memleak with (real) closures remains.
7 - dramatically reduced size of stacks to 65536 (linux/x86), after 362 - dramatically reduced size of stacks to 65536 (linux/x86), after
8 observing that even myhttpd never had more than a kilobyte 363 observing that even myhttpd never had more than a kilobyte
9 of stack in use (stack is still only physically allocated in 364 of stack in use (stack is still only physically allocated in
10 pagesize increments for systems suporting mmap). 365 pagesize increments for systems suporting mmap).
366 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
367 - fix for threaded perl.
11 368
120.9 Sun Nov 23 04:32:39 CET 2003 3690.9 Sun Nov 23 04:32:39 CET 2003
13 - fix a few very hard to track down but (of course) stupid 370 - fix a few very hard to track down but (of course) stupid
14 memory leaks. 371 memory leaks.
15 - stringify version number to avoid locale problems :(. 372 - stringify version number to avoid locale problems :(.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines