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

Comparing Coro/Changes (file contents):
Revision 1.45 by root, Wed Sep 12 16:45:59 2001 UTC vs.
Revision 1.266 by root, Sun Oct 7 13:53:37 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
30.49 3TODO: localise SIG{__DIE__} and/or __WARN__
4TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO.
5TODO: maybe implement a default message channel, very much like Erlang
6 (which is cool in a lot of important aspects (failures!),
7 but very lacking in others (higher level ipc)).
8
94.1
10 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
11 be local to each coro (see Coro::State).
12 - possibly bring back 5.10 compatibility (untested).
13 - work around stupid (and wrong) warning on 5.10 :(.
14 - overlay the saved state over the context stack. This saves
15 a few hundred bytes per coroutine on average and also
16 speeds up context switching a bit.
17 - further tune default stack sizes.
18 - (more) correctly calculate stack usage in coro_rss.
19
204.03 Sat Oct 6 21:24:00 CEST 2007
21 - added Coro::throw method.
22 - minor code cleanups.
23
244.02 Sat Oct 6 02:36:47 CEST 2007
25 - fix a very minor per-coroutine memleak (a single codereference).
26 - fixed a bug where the currently in-use c context would be freed
27 prematurely (can happen only when programs change the stacksize
28 or use tracing).
29 - tracing can no longer keep a coro alive after it terminated.
30 - do static branch prediction in the common path for gcc. gives
31 about 2-5% speed improvement here.
32
334.01 Fri Oct 5 22:10:49 CEST 2007
34 - instead of recreating *a* standard output handle we simply
35 use STDOUT, which is faster and hopefully more robust.
36
374.0 Fri Oct 5 12:56:00 CEST 2007
38 - incompatibly changed Coro::Storable::freeze.
39 - major new feature: added Coro::Debug, for interactive coroutine
40 debugging, tracing and much more.
41 - major bug fix: unbelievable, but true: $_, $/ and many other
42 "saved" variables actually weren't being saved. This has been fixed,
43 of course, while increasing performance while losing all the save
44 flags.
45 - save flags are gone, and all the api functions dealing with them.
46 - added Coro::Semaphore::adjust.
47 - added Coro::Util::fork_eval.
48 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
49 - added Coro::killall.
50 - reduce initial stack sizes to allow for "micro-coroutines".
51 - better async_pool resource management, moved parts of async_pool
52 handling to XS (major speed improvement).
53 - actually croak before modifying important data structures.
54 - refuse to transfer while compiling.
55 - possibly support eval EXPR better now.
56 - enable assembly per default on linux+bsd x86+amd64.
57 - all internal members were renamed _something for easier subclassing.
58 - many minor tweaks.
59
603.63 Wed May 16 14:10:06 CEST 2007
61 - implement handcoded assembly for x86/amd64 SVR ABI.
62
633.62 Fri Apr 27 21:36:06 CEST 2007
64 - upgrade libcoro (which might set unwind info correctly).
65 - change default on linux to setjmp/longjmp.
66
673.61 Thu Apr 19 12:36:18 CEST 2007
68 - Coro::Storable caused an endless loop when thawing invalid
69 pst's sometimes.
70 - use a Semaphore in Coro::Storable, as Storable doesn't
71 seem to be reentrant (although it is documented to
72 be threadsafe...).
73 - fix Coro::Signal to bring back the original unreliable
74 but stateful semantics.
75 - fixed a lot of typos in Coro.pm (patch submitted by David
76 Steinbrunner, which applied flawlessly).
77
783.6 Sat Apr 14 17:13:31 CEST 2007
79 - added some bugfixes to get eg/myhttpd working again.
80 - added Coro::Storable for often-cede'ing freeze/thaw.
81 - try to do a clean exit when a coroutine calls exit
82 (EXPERIMENTAL).
83 - got rid of indirect call through _coro_init.
84 - updated the partly antique examples in eg/ to
85 work again and be a bit less magic, too.
86 - fixed Coro::Signal semantics to work as documented again.
87
883.55 Sun Mar 25 01:20:47 CET 2007
89 - add SAVE_DEFFH to save the default filehandle and enable
90 it by default.
91 - finally move socket-operations from Coro::Socket to Coro::Handle
92 to be able to unblock foreign sockets.
93 - add Coro::State::save_also and guarded_save.
94 - add count accessor to Coro::Semaphore.
95 - add Coro::State::cctx_stacksize.
96 - just for the fun of it, do not rely on implicit context,
97 which can dramatically improve performance, but people
98 using windows-process-emulation perls do not care much about
99 performance.
100
1013.51 Sun Mar 4 14:18:04 CET 2007
102 - fixed a problem when you weakened references to Coro::State's
103 (patch by Yuval Kogman).
104
1053.501 Wed Feb 28 12:44:07 CET 2007
106 - rename some global symbols as macosx from hell redefines
107 them without asking.
108
1093.5 Tue Feb 13 20:22:53 CET 2007
110 - do AnyEvent model detection earlier, avoiding problems
111 caused by first using AnyEvent and later Coro::Event.
112 - implement and document Coro::Event event objects.
113 - fix a potential problem in Coro::Event causing crashes.
114 - initialise PL_comppad when creating a new coroutine,
115 avoids crashes on early coro destruction.
116
1173.41 Mon Jan 22 19:19:49 CET 2007
118 - readline on Coro::Handle did not support undefined $/,
119 nor did it deliver partial lines on EOF or error.
120 - implement malloc fallback for stack allocation because
121 stupid broken idiotic OSX has a stupid broken
122 idiotic fits-the-whole-os mmap "implementation" and
123 my dick feels longer if Coro is portable even to
124 obsolete platforms.
125
1263.4 Fri Jan 19 21:52:54 CET 2007
127 - remove t/09_timer.t, as it isn't really testing much
128 but was rather flaky in practise.
129 - async_pool coro would keep arguments and callback alive until
130 it was reused.
131 - cancellation of a coroutine could cause spurious idle calls
132 in cede_notself.
133
1343.3 Sat Jan 6 03:45:00 CET 2007
135 - implement $coro->on_destroy.
136 - Coro::Event blocking semantics have been changed,
137 documented and - hopefully - improved.
138 - fix nice adding, not subtracting, from priority.
139 - fix ->prio and api_is_ready (patch by Mark Hinds).
140 - fixed an assert ("... == prev__cctx->idle_te")
141 that could errronously trigger.
142 - fix various large and small memleaks.
143 - use a (hopefully) more stable cancel implementation
144 that immediately frees the coroutine data.
145 - cede/cede_notself return a status now.
146 - added Coro::guard function.
147 - added a global coroutine pool for jobs (on my machine,
148 I can create and execute 48k simple coros/s with async,
149 and 128k coros with async_pool).
150 - Coro::AIO now uses the coroutine priority as io priority.
151
1523.2 Fri Dec 22 05:07:09 CET 2006
153 - improve portability to slightly older perls.
154 - use cleaner coroutine destruction.
155 - simplify configuration for users.
156 - optionally (unrecommended) prefer perl functions over
157 their coro replacements.
158
1593.11 Tue Dec 5 13:11:24 CET 2006
160 - fixed some bogus assert's, but as perl.h disables assert even
161 without NDEBUG (thank you very much), not too many people should
162 notice (that did include myself). Andreas König noticed, though :)
163 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
164 it. Also noticed by Andreas König :)
165 - save/restore tainted status.
166 - verified to pass the testsuite on my 5.9.5.
167
1683.1 Mon Dec 4 23:03:40 CET 2006
169 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
170 - incompatible change: transfer flags are now per-state.
171 - give a better error message on deadlock.
172 - document Coro::nready.
173 - enhanced testsuite.
174
1753.01 Sun Dec 3 23:47:42 CET 2006
176 - forgot to include Coro::Timer.
177
1783.0 Sun Dec 3 22:57:25 CET 2006
179 - the "FINALLY COMPLETELY STABLE" release (coming soon:
180 the "FAMOUS LAST WORDS" release).
181 - implement a new stack sharing algorithm, which uses a stack
182 pool (size currently hardcoded).
183 - make stack sharing mandatory (it no longer uses a heuristic).
184 - eval/die no longer cause weird problems under heavy use.
185 - Coro::Event could cause livelocks if it was used but
186 no Coro::Event watchers were used.
187 - Coro::Event now uses asynccheck as prepare does not
188 check for changed watchers.
189 - Coro::Event allows multiple waiting coros and will wake up one
190 per event.
191 - Coro::Event should be cleaner and more efficient now.
192 - new utility function Coro::unblock_sub.
193 - document the sad fact that Event is no longer reentrant.
194 - putting a coroutine into the ready queue twice could under
195 some circumstances lead to stack corruption.
196 - minor incompatibility: subclassing Coro::State is supported
197 directly now without going through a _coro_state member.
198 - state/coro switching is much faster now.
199 - very minor optimisations and code/documentation cleanup.
200 - avoid problems due to compiler inlining.
201 - removed timers from Coro::Timer -> use AnyEvent instead.
202 - replaced Coro::idle coroutine by (cleaner) idle handler.
203 - updated to newest libcoro.
204 - implement enhanced support for valgrind.
205 - implement is_ready and return value for ready.
206 - removed Coro::Cont, it was a misnomer (it's generators),
207 and it was somewhat annoying to get it right. Will come back
208 if somebody asks for it :)
209 - many ->wait methods and Coro::Event could return
210 spuriously without the event having happened.
211
2122.5 Tue Nov 7 12:22:33 CET 2006
213 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
214 use AnyEvent, moved them to Coro/.
215 - added Coro::LWP which contains all the uglyness required to
216 make LWP non-blocking.
217 - should work with perl 5.9.x now (Andreas König made me do it).
218 - fixed another bug in Coro::Select when the timeout was undef.
219 - reuse PL_start_env for all coros, saves some memory per coroutine.
220 - manage PL_top_env differently, hopefully to avoid panic: top_env.
221 - timeout argument was not properly used in Coro::Socket.
222 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
223 - emulate undocumented(!) functionality of IO::Socket required
224 by LWP(!!).
225 - updated eg/lwp to work with newer lwp's.
226 - remove "FATAL: uncaught exception" prefix. Coroutines that die
227 kill the whole process, just as exceptions in the main "coroutine"
228 did already.
229
2302.1 Wed Nov 1 23:01:13 CET 2006
231 - fix a long-standing bug in Coro::Select where select with
232 zero timeout would instead change the current default filehandle.
233 - use a simpler and hopefully more robust way to clone padlists
234 (uses less memory and a perl function instead of our own).
235 - coro can now create a stack guard area on many architectures.
236 - Coro::AIO properly reexports additional functions from IO::AIO.
237 - updated libcoro with a workaround for OS X,
238 pach and testing by Michael Schwern.
239
2402.0 Tue Oct 24 05:47:17 CEST 2006
241 - support additional aio requests in Coro::AIO.
242
2431.9 (never properly released due to a glitch)
244
2451.8 Thu Feb 2 00:59:06 CET 2006
246 - applied suggested patch by SAMV to avoid problems during stupid
247 mark & sweep gc run.
248 - applied patch by Scott Walters for 5.9.3 compatibility.
249
2501.7 Tue Dec 27 01:41:58 CET 2005
251 - added Coro::AIO, a thin wrapper around IO::AIO.
252 - improved Makefile.PL explanations.
253
2541.6 Mon Dec 12 21:30:05 CET 2005
255 - additionally save PM_curpm might fix as-of-yet
256 unseen problems with regex matches being attributed to the
257 wrong package.
258 - add t/10_bugs.t, which currently checks against imho broken
259 perls that use 0.26MB of stack space per Perl_magic_get
260 invocation (newer linuxes) as opposed to the 0.0002MB perl
261 normally uses.
262 - make stacksize configurable for the ultimate debian experience.
263
2641.51 Mon Dec 12 18:48:36 CET 2005
265 - remove debugging warn() accidentally left in Coro::Select.
266
2671.5 Tue Nov 29 13:32:44 CET 2005
268 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
269 - fix Coro::Select 'select'.
270 - strict'ify some modules.
271 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
272 - use XSLoader in selected modules.
273 - remove some 5.6 compatibility cruft.
274
2751.4 Tue Sep 6 00:11:05 CEST 2005
276 - libcoro did not take into account the trampoline on amd64,
277 when 'l'inux method was used.
278
2791.31 Tue Aug 30 23:31:33 CEST 2005
280 - some portability fixes/workarounds.
281
2821.3 Sat Aug 20 03:08:56 CEST 2005
283 - no code changes. module seems to work fine.
284 - improve Coro::State docs, remove reference to nonexisting
285 Coro::State::flush method.
286 - no longer autodetect windows, present it as an option instead.
287
2881.2 Mon May 16 02:00:55 CEST 2005
289 - included libcoro.c earlier in State.xs, might improve portability.
290 - use a faster and possibly more stable Coro::Cont implementation.
291 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
292 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
293 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
294
2951.11 Thu Mar 3 18:00:52 CET 2005
296 - change of contact address.
297
2981.1 Tue Feb 22 20:51:16 CET 2005
299 - support [l]inux method on amd64.
300 - allow some context switches while compiling/eval'ing.
301
3021.0 Tue Aug 10 03:47:27 CEST 2004
303 - Event 0.89 fixes the bug again.
304 - fix the remaining(??) memleak. I didn't free the cache-AV for
305 padlists, so one AV per created anon-sub leaked.
306
3070.97 Fri May 14 15:23:32 CEST 2004
308 - remove stupid gcc extension that was an experiment only anyway.
309
3100.96 Thu May 13 18:09:29 CEST 2004
311 - tell people if they have yet another broken Event variant.
312 - made Coro to work _better_ with threads (it seems that, as long
313 as you use Coros in one thread only everything will work fine.
314 Looking at the perl source the problem might be missing locking
315 between Coro and some perl internal routines. That's not easily
316 fixable).
317 - (hopefully) got rid of the static xnext inside transfer.
318
3190.95 Sun Feb 15 03:20:28 CET 2004
320 - removed Coro::State::flush, implemented a new and much
321 faster caching mechanism by attaching magic to the CV.
322 - a memleak with (real) closures remains.
323 - dramatically reduced size of stacks to 65536 (linux/x86), after
324 observing that even myhttpd never had more than a kilobyte
325 of stack in use (stack is still only physically allocated in
326 pagesize increments for systems suporting mmap).
327 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
328 - fix for threaded perl.
329
3300.9 Sun Nov 23 04:32:39 CET 2003
331 - fix a few very hard to track down but (of course) stupid
332 memory leaks.
333 - stringify version number to avoid locale problems :(.
334 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
335 hopefully ignoring it "just works".
336
3370.8 Wed Nov 5 19:38:40 CET 2003
338 - port to perl5.005_03, but only 5.8.x+ is supported!
339 - honor LocalAddr even without LocalPort in Coro::Socket.
340
3410.7 Tue May 27 03:12:38 CEST 2003
342 - the version jump indicates some level of testing, not gobs
343 of new features.
344 - uh... I found the prompt function in ExtUtils::MakeMaker.
345 Highly correct stuff, that is...
346 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
347 the fix is not well understood by the author, so beware :(.
348
3490.652 Thu May 8 02:54:46 CEST 2003
350 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
351 cause version 4 doesn't have ucontext.
352 - Benjamin Reed reported that setjmp works fine on darwin,
353 so preselect it.
354
3550.651 Sat Mar 29 15:00:23 CET 2003
356 - fix a bug in Coro::Handle where some bytes could get lost
357 on reading, reported by jason@nichego.net.
358
3590.65 Sun Mar 23 00:08:26 CET 2003
360 - added a README.linux-glibc.
361 - new module Coro::Select.
362 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
363 (I never claimed that I know what I am doing ;).
364 - set default to setjmp/longjmp on non-x86-linux.
365
3660.6 Thu Nov 21 11:09:06 CET 2002
367 - made lazy context switching the default.
368 - fixed the bug where SemaphoreSet::guard
369 was the same as timed_guard and timed_guard was missing.
370 - fix a memleak in Coro::Timer.
371 - reclassified context sharing as not experimental.
372
3730.534 Sun Apr 14 03:05:12 CEST 2002
374 - fixed backspace => tab in header parsing (myhttpd).
375 - added eg/readline, for elmex.
376 - Coro::Event now calls ->start on first call instead of
377 ->again, which created wrong timeouts for the first ->next.
378 - fixed a bug where Coro::Socket returned a closed socket instead
379 of nothing on ECONNREFUSED and similar error conditions.
380
3810.533 Mon Feb 18 18:49:40 CET 2002
382 - adapt to finally working Event-0.85.
383 - fixed buggy event initialization (trapped under windows, but how
384 could it possibly have worked anywhere?)
385
3860.532 Wed Jan 16 02:45:32 CET 2002
387 - removed the reference to pp_entersub, might work on
388 (some) win32 perls now (testsuite works under cygwin,
389 without Event, which freezes).
390
3910.531 Mon Dec 10 22:18:44 CET 2001
392 - Fixed a bug in SemaphoreSet::timed_down.
393
3940.53 Tue Nov 27 21:11:13 CET 2001
395 - some tests for Event.
396 - slightly different internal architecture to get refcnt into the object
397 and not the reference. might fix a few bugs, certainly did introduce
398 new ones.
399 - Coro::Timer (independent of Event).
400 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
401
4020.52 Tue Nov 6 21:36:18 CET 2001
403 - ported to cygwin (trivial).
404 patch by Gerrit P. Haase <gp@familiehaase.de>.
405 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
406
4070.51 Thu Nov 1 20:39:01 CET 2001
408 - terminate/cancel now work properly (otherwise termination
409 could cause a "next coroutine is not and contains not..."
410 error.
411 - added Coro::Socket::shutdown.
412 - Coro::Event::loop no is the same as Event::loop.
413 - implemented terminate with args + join.
414
4150.5 Fri Sep 28 16:15:35 CEST 2001
416 - fixed "print" on a Coro::Handle. The print method worked.
417 - small tweaks (seem to reduce memory consumption a lot)
418 in various modules.
419 - splendid use of "no warnings" scattered throughout.
420 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
421
4220.49 Sun Sep 16 02:42:45 CEST 2001
4 - changed some method calls to function calls for speed 423 - changed some method calls to function calls for speed
5 inside Coro::Handle. 424 inside Coro::Handle.
6 - make Coro::Handle use an array instead of a hash for 425 - make Coro::Handle use an array instead of a hash for
7 speed reasons. 426 speed reasons.
8 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?". 427 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
9 sjlj and ucontext should now work. 428 sjlj and ucontext should now work.
10 - IRIX-specific port for libcoro. 429 - IRIX-specific port for libcoro.
11 - swapped order of accept results to match IO::Socket. 430 - swapped order of accept results to match IO::Socket.
431 - changed getsock/peername to sock/peername to match IO::Socket.
432 - fixed a bug that caused segfault when returning to main task
433 under some circumstances.
434 - other bugfixes.
12 435
130.45 Sun Sep 2 02:54:01 CEST 2001 4360.45 Sun Sep 2 02:54:01 CEST 2001
14 - new method Coro::Handle::timeout. 437 - new method Coro::Handle::timeout.
15 - corrected speling of set...name to get...name in Coro::Socket. 438 - corrected speling of set...name to get...name in Coro::Socket.
16 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle. 439 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines