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

Comparing Coro/Changes (file contents):
Revision 1.141 by root, Sun Nov 5 19:16:09 2006 UTC vs.
Revision 1.308 by root, Thu Dec 27 21:32:15 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: possibly convert some of the Event modules into AnyEvent users. 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)).
4 7
5 - converted Coro::Select to AnyEvent. 8 - "bt" debug command now displays any exceptions
9 from longmess.
10
114.34 Sat Dec 22 17:49:53 CET 2007
12 - upgrade to EV version 2.0 API.
13
144.33 Mon Dec 17 08:36:12 CET 2007
15 - make Coro::AIO etc. loadable in the absence of EV.
16
174.32 Mon Dec 17 07:46:02 CET 2007
18 - majorly improved Coro::Handle's performance with Coro::EV.
19 - implemented the until now mythical Coro::BDB module.
20 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
21
224.31 Wed Dec 5 12:32:39 CET 2007
23 - remove warn statement form Coro::Util.
24
254.3 Tue Dec 4 20:33:14 CET 2007
26 - calls to the idle function could cause stack corruption
27 when the stack changed.
28 - do no longer rely on the presence of EV::DNS if EV is used
29 (because it is gone), but instead take avdantage of EV::ADNS
30 if available.
31 - add ($) prototypes to all functions in Coro::Storable.
32 - use a conventional (safer) idle callback in Coro::EV.
33 - do accept quickly in Coro::Debug to avoid endless loops.
34
354.22 Fri Nov 30 16:04:04 CET 2007
36 - really use optimised versions for Event and EV in Coro::Util
37 and Coro::Handle.
38
394.21 Sun Nov 25 10:48:59 CET 2007
40 - fix a spurious memory read.
41 - Coro::EV no longer keeps the eventloop "alive".
42
434.2 Fri Nov 9 20:47:05 CET 2007
44 - enable/disable tracing from a new coroutine, not a pooled one.
45 - fix a memleak in Coro::Event.
46 - removed killall call from fork_eval.
47 - made sure store_fd is already loaded so that fork_eval
48 does not have to parse autoload in each subprocess.
49 - only use assembly method if -O switch is in $Config{optimize}.
50 - add (optional) Coro::EV module, so far the best event loop module
51 directly supported by Coro.
52 - if the event model is EV, use EV::DNS to resolve
53 stuff in Coro::Util.
54 - don't get confused by multiple event notifications in Coro::Handle.
55 - initial support for EV (libevent interface).
56 - require Event and EV using configure_requires, to force their existance.
57
584.13 Wed Oct 24 07:26:45 CEST 2007
59 - add Coro::Storable::blocking_thaw.
60 - use a vastly more complicated technique to localise
61 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
62 - use a coroutine for the idle callback Coro::Event,
63 instead of running Event in the current coroutine context.
64 This also catches recursive invocations.
65 - actually report fork errors in gethostbyname and inet_aton.
66
674.11 Thu Oct 11 02:40:24 CEST 2007
68 - port to threaded perls.
69
704.1 Thu Oct 11 02:38:16 CEST 2007
71 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
72 be local to each coro (see Coro::State).
73 - incompatible change: for very deep reasons, cede and cede_notself
74 cannot return anything, so nothing will be returned.
75 - possibly bring back 5.10 compatibility (untested).
76 - work around stupid (and wrong) warning on 5.10 :(.
77 - overlay the saved state over the context stack. This saves
78 a few hundred bytes per coroutine on average and also
79 speeds up context switching a bit.
80 - further tune default stack sizes.
81 - (more) correctly calculate stack usage in coro_rss.
82 - Coro::Storable::blocking_* did not properly lock
83 resulting in races between coroutines.
84 - added Coro::Storable::guard.
85 - stopping to trace a coroutine could destroy the cctx of
86 an unrelated coroutine.
87 - explain the relationship between Perl and C coroutines in
88 more detail in Coro::State.
89 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
90 causing a fork per resolve. This also affected Coro::Socket.
91 - switch to a separate stack in $coro->call/eval to avoid
92 invalidating pointers.
93
944.03 Sat Oct 6 21:24:00 CEST 2007
95 - added Coro::throw method.
96 - minor code cleanups.
97
984.02 Sat Oct 6 02:36:47 CEST 2007
99 - fix a very minor per-coroutine memleak (a single codereference).
100 - fixed a bug where the currently in-use c context would be freed
101 prematurely (can happen only when programs change the stacksize
102 or use tracing).
103 - tracing can no longer keep a coro alive after it terminated.
104 - do static branch prediction in the common path for gcc. gives
105 about 2-5% speed improvement here.
106
1074.01 Fri Oct 5 22:10:49 CEST 2007
108 - instead of recreating *a* standard output handle we simply
109 use STDOUT, which is faster and hopefully more robust.
110
1114.0 Fri Oct 5 12:56:00 CEST 2007
112 - incompatibly changed Coro::Storable::freeze.
113 - major new feature: added Coro::Debug, for interactive coroutine
114 debugging, tracing and much more.
115 - major bug fix: unbelievable, but true: $_, $/ and many other
116 "saved" variables actually weren't being saved. This has been fixed,
117 of course, while increasing performance while losing all the save
118 flags.
119 - save flags are gone, and all the api functions dealing with them.
120 - added Coro::Semaphore::adjust.
121 - added Coro::Util::fork_eval.
122 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
123 - added Coro::killall.
124 - reduce initial stack sizes to allow for "micro-coroutines".
125 - better async_pool resource management, moved parts of async_pool
126 handling to XS (major speed improvement).
127 - actually croak before modifying important data structures.
128 - refuse to transfer while compiling.
129 - possibly support eval EXPR better now.
130 - enable assembly per default on linux+bsd x86+amd64.
131 - all internal members were renamed _something for easier subclassing.
132 - many minor tweaks.
133
1343.63 Wed May 16 14:10:06 CEST 2007
135 - implement handcoded assembly for x86/amd64 SVR ABI.
136
1373.62 Fri Apr 27 21:36:06 CEST 2007
138 - upgrade libcoro (which might set unwind info correctly).
139 - change default on linux to setjmp/longjmp.
140
1413.61 Thu Apr 19 12:36:18 CEST 2007
142 - Coro::Storable caused an endless loop when thawing invalid
143 pst's sometimes.
144 - use a Semaphore in Coro::Storable, as Storable doesn't
145 seem to be reentrant (although it is documented to
146 be threadsafe...).
147 - fix Coro::Signal to bring back the original unreliable
148 but stateful semantics.
149 - fixed a lot of typos in Coro.pm (patch submitted by David
150 Steinbrunner, which applied flawlessly).
151
1523.6 Sat Apr 14 17:13:31 CEST 2007
153 - added some bugfixes to get eg/myhttpd working again.
154 - added Coro::Storable for often-cede'ing freeze/thaw.
155 - try to do a clean exit when a coroutine calls exit
156 (EXPERIMENTAL).
157 - got rid of indirect call through _coro_init.
158 - updated the partly antique examples in eg/ to
159 work again and be a bit less magic, too.
160 - fixed Coro::Signal semantics to work as documented again.
161
1623.55 Sun Mar 25 01:20:47 CET 2007
163 - add SAVE_DEFFH to save the default filehandle and enable
164 it by default.
165 - finally move socket-operations from Coro::Socket to Coro::Handle
166 to be able to unblock foreign sockets.
167 - add Coro::State::save_also and guarded_save.
168 - add count accessor to Coro::Semaphore.
169 - add Coro::State::cctx_stacksize.
170 - just for the fun of it, do not rely on implicit context,
171 which can dramatically improve performance, but people
172 using windows-process-emulation perls do not care much about
173 performance.
174
1753.51 Sun Mar 4 14:18:04 CET 2007
176 - fixed a problem when you weakened references to Coro::State's
177 (patch by Yuval Kogman).
178
1793.501 Wed Feb 28 12:44:07 CET 2007
180 - rename some global symbols as macosx from hell redefines
181 them without asking.
182
1833.5 Tue Feb 13 20:22:53 CET 2007
184 - do AnyEvent model detection earlier, avoiding problems
185 caused by first using AnyEvent and later Coro::Event.
186 - implement and document Coro::Event event objects.
187 - fix a potential problem in Coro::Event causing crashes.
188 - initialise PL_comppad when creating a new coroutine,
189 avoids crashes on early coro destruction.
190
1913.41 Mon Jan 22 19:19:49 CET 2007
192 - readline on Coro::Handle did not support undefined $/,
193 nor did it deliver partial lines on EOF or error.
194 - implement malloc fallback for stack allocation because
195 stupid broken idiotic OSX has a stupid broken
196 idiotic fits-the-whole-os mmap "implementation" and
197 my dick feels longer if Coro is portable even to
198 obsolete platforms.
199
2003.4 Fri Jan 19 21:52:54 CET 2007
201 - remove t/09_timer.t, as it isn't really testing much
202 but was rather flaky in practise.
203 - async_pool coro would keep arguments and callback alive until
204 it was reused.
205 - cancellation of a coroutine could cause spurious idle calls
206 in cede_notself.
207
2083.3 Sat Jan 6 03:45:00 CET 2007
209 - implement $coro->on_destroy.
210 - Coro::Event blocking semantics have been changed,
211 documented and - hopefully - improved.
212 - fix nice adding, not subtracting, from priority.
213 - fix ->prio and api_is_ready (patch by Mark Hinds).
214 - fixed an assert ("... == prev__cctx->idle_te")
215 that could errronously trigger.
216 - fix various large and small memleaks.
217 - use a (hopefully) more stable cancel implementation
218 that immediately frees the coroutine data.
219 - cede/cede_notself return a status now.
220 - added Coro::guard function.
221 - added a global coroutine pool for jobs (on my machine,
222 I can create and execute 48k simple coros/s with async,
223 and 128k coros with async_pool).
224 - Coro::AIO now uses the coroutine priority as io priority.
225
2263.2 Fri Dec 22 05:07:09 CET 2006
227 - improve portability to slightly older perls.
228 - use cleaner coroutine destruction.
229 - simplify configuration for users.
230 - optionally (unrecommended) prefer perl functions over
231 their coro replacements.
232
2333.11 Tue Dec 5 13:11:24 CET 2006
234 - fixed some bogus assert's, but as perl.h disables assert even
235 without NDEBUG (thank you very much), not too many people should
236 notice (that did include myself). Andreas König noticed, though :)
237 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
238 it. Also noticed by Andreas König :)
239 - save/restore tainted status.
240 - verified to pass the testsuite on my 5.9.5.
241
2423.1 Mon Dec 4 23:03:40 CET 2006
243 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
244 - incompatible change: transfer flags are now per-state.
245 - give a better error message on deadlock.
246 - document Coro::nready.
247 - enhanced testsuite.
248
2493.01 Sun Dec 3 23:47:42 CET 2006
250 - forgot to include Coro::Timer.
251
2523.0 Sun Dec 3 22:57:25 CET 2006
253 - the "FINALLY COMPLETELY STABLE" release (coming soon:
254 the "FAMOUS LAST WORDS" release).
255 - implement a new stack sharing algorithm, which uses a stack
256 pool (size currently hardcoded).
257 - make stack sharing mandatory (it no longer uses a heuristic).
258 - eval/die no longer cause weird problems under heavy use.
259 - Coro::Event could cause livelocks if it was used but
260 no Coro::Event watchers were used.
261 - Coro::Event now uses asynccheck as prepare does not
262 check for changed watchers.
263 - Coro::Event allows multiple waiting coros and will wake up one
264 per event.
265 - Coro::Event should be cleaner and more efficient now.
266 - new utility function Coro::unblock_sub.
267 - document the sad fact that Event is no longer reentrant.
268 - putting a coroutine into the ready queue twice could under
269 some circumstances lead to stack corruption.
270 - minor incompatibility: subclassing Coro::State is supported
271 directly now without going through a _coro_state member.
272 - state/coro switching is much faster now.
273 - very minor optimisations and code/documentation cleanup.
274 - avoid problems due to compiler inlining.
275 - removed timers from Coro::Timer -> use AnyEvent instead.
276 - replaced Coro::idle coroutine by (cleaner) idle handler.
277 - updated to newest libcoro.
278 - implement enhanced support for valgrind.
279 - implement is_ready and return value for ready.
280 - removed Coro::Cont, it was a misnomer (it's generators),
281 and it was somewhat annoying to get it right. Will come back
282 if somebody asks for it :)
283 - many ->wait methods and Coro::Event could return
284 spuriously without the event having happened.
285
2862.5 Tue Nov 7 12:22:33 CET 2006
287 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
288 use AnyEvent, moved them to Coro/.
6 - added Coro::LWP which contains all the uglyness required to 289 - added Coro::LWP which contains all the uglyness required to
7 make LWP non-blocking. 290 make LWP non-blocking.
8 - should work with perl 5.9.x now (Andreas König made me do it). 291 - should work with perl 5.9.x now (Andreas König made me do it).
9 - fixed another bug in Coro::Select when the timeout was undef. 292 - fixed another bug in Coro::Select when the timeout was undef.
10 - reuse PL_start_env for all coros, saves some memory per coroutine. 293 - reuse PL_start_env for all coros, saves some memory per coroutine.
… …
12 - timeout argument was not properly used in Coro::Socket. 295 - timeout argument was not properly used in Coro::Socket.
13 - allow limited forms of subclassing in Coro::Handle/Coro::Socket. 296 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
14 - emulate undocumented(!) functionality of IO::Socket required 297 - emulate undocumented(!) functionality of IO::Socket required
15 by LWP(!!). 298 by LWP(!!).
16 - updated eg/lwp to work with newer lwp's. 299 - updated eg/lwp to work with newer lwp's.
17 - made Coro::Util an AnyEvent user, moved it to Coro/.
18 - remove "FATAL: uncaught exception" prefix. Coroutines that die 300 - remove "FATAL: uncaught exception" prefix. Coroutines that die
19 kill the whole process, just as exceptions in the main "coroutine" 301 kill the whole process, just as exceptions in the main "coroutine"
20 did already. 302 did already.
21 303
222.1 Wed Nov 1 23:01:13 CET 2006 3042.1 Wed Nov 1 23:01:13 CET 2006

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines