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

Comparing Coro/Changes (file contents):
Revision 1.148 by root, Mon Nov 20 20:29:33 2006 UTC vs.
Revision 1.298 by root, Mon Dec 3 22:48:04 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: check wether stack-sharing is still effective 3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO.
4TODO: more effective stack sharing by having one "default stack" for "simple" coroutines? 4TODO: maybe implement a default message channel, very much like Erlang
5TODO: verify exception frame changes and force stack copies? 5 (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
6 7
7 - avoid deadlock in Coro::Event's idle process when it is called 8 - do no longer rely on the presence of EV::DNS if EV is used
8 early with no active watchers. 9 (because it is gone), but instead take avdantage of EV::ADNS
9 - minor documentation cleanup. 10 if available.
11 - add ($) prototypes to all functions in Coro::Storable.
12 - use a conventional (safer) idle callback in Coro::EV.
13 - do accept quickly in Coro::Debug to avoid endless loops.
14
154.22 Fri Nov 30 16:04:04 CET 2007
16 - really use optimised versions for Event and EV in Coro::Util
17 and Coro::Handle.
18
194.21 Sun Nov 25 10:48:59 CET 2007
20 - fix a spurious memory read.
21 - Coro::EV no longer keeps the eventloop "alive".
22
234.2 Fri Nov 9 20:47:05 CET 2007
24 - enable/disable tracing from a new coroutine, not a pooled one.
25 - fix a memleak in Coro::Event.
26 - removed killall call from fork_eval.
27 - made sure store_fd is already loaded so that fork_eval
28 does not have to parse autoload in each subprocess.
29 - only use assembly method if -O switch is in $Config{optimize}.
30 - add (optional) Coro::EV module, so far the best event loop module
31 directly supported by Coro.
32 - if the event model is EV, use EV::DNS to resolve
33 stuff in Coro::Util.
34 - don't get confused by multiple event notifications in Coro::Handle.
35 - initial support for EV (libevent interface).
36 - require Event and EV using configure_requires, to force their existance.
37
384.13 Wed Oct 24 07:26:45 CEST 2007
39 - add Coro::Storable::blocking_thaw.
40 - use a vastly more complicated technique to localise
41 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
42 - use a coroutine for the idle callback Coro::Event,
43 instead of running Event in the current coroutine context.
44 This also catches recursive invocations.
45 - actually report fork errors in gethostbyname and inet_aton.
46
474.11 Thu Oct 11 02:40:24 CEST 2007
48 - port to threaded perls.
49
504.1 Thu Oct 11 02:38:16 CEST 2007
51 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
52 be local to each coro (see Coro::State).
53 - incompatible change: for very deep reasons, cede and cede_notself
54 cannot return anything, so nothing will be returned.
55 - possibly bring back 5.10 compatibility (untested).
56 - work around stupid (and wrong) warning on 5.10 :(.
57 - overlay the saved state over the context stack. This saves
58 a few hundred bytes per coroutine on average and also
59 speeds up context switching a bit.
60 - further tune default stack sizes.
61 - (more) correctly calculate stack usage in coro_rss.
62 - Coro::Storable::blocking_* did not properly lock
63 resulting in races between coroutines.
64 - added Coro::Storable::guard.
65 - stopping to trace a coroutine could destroy the cctx of
66 an unrelated coroutine.
67 - explain the relationship between Perl and C coroutines in
68 more detail in Coro::State.
69 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
70 causing a fork per resolve. This also affected Coro::Socket.
71 - switch to a separate stack in $coro->call/eval to avoid
72 invalidating pointers.
73
744.03 Sat Oct 6 21:24:00 CEST 2007
75 - added Coro::throw method.
76 - minor code cleanups.
77
784.02 Sat Oct 6 02:36:47 CEST 2007
79 - fix a very minor per-coroutine memleak (a single codereference).
80 - fixed a bug where the currently in-use c context would be freed
81 prematurely (can happen only when programs change the stacksize
82 or use tracing).
83 - tracing can no longer keep a coro alive after it terminated.
84 - do static branch prediction in the common path for gcc. gives
85 about 2-5% speed improvement here.
86
874.01 Fri Oct 5 22:10:49 CEST 2007
88 - instead of recreating *a* standard output handle we simply
89 use STDOUT, which is faster and hopefully more robust.
90
914.0 Fri Oct 5 12:56:00 CEST 2007
92 - incompatibly changed Coro::Storable::freeze.
93 - major new feature: added Coro::Debug, for interactive coroutine
94 debugging, tracing and much more.
95 - major bug fix: unbelievable, but true: $_, $/ and many other
96 "saved" variables actually weren't being saved. This has been fixed,
97 of course, while increasing performance while losing all the save
98 flags.
99 - save flags are gone, and all the api functions dealing with them.
100 - added Coro::Semaphore::adjust.
101 - added Coro::Util::fork_eval.
102 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
103 - added Coro::killall.
104 - reduce initial stack sizes to allow for "micro-coroutines".
105 - better async_pool resource management, moved parts of async_pool
106 handling to XS (major speed improvement).
107 - actually croak before modifying important data structures.
108 - refuse to transfer while compiling.
109 - possibly support eval EXPR better now.
110 - enable assembly per default on linux+bsd x86+amd64.
111 - all internal members were renamed _something for easier subclassing.
112 - many minor tweaks.
113
1143.63 Wed May 16 14:10:06 CEST 2007
115 - implement handcoded assembly for x86/amd64 SVR ABI.
116
1173.62 Fri Apr 27 21:36:06 CEST 2007
118 - upgrade libcoro (which might set unwind info correctly).
119 - change default on linux to setjmp/longjmp.
120
1213.61 Thu Apr 19 12:36:18 CEST 2007
122 - Coro::Storable caused an endless loop when thawing invalid
123 pst's sometimes.
124 - use a Semaphore in Coro::Storable, as Storable doesn't
125 seem to be reentrant (although it is documented to
126 be threadsafe...).
127 - fix Coro::Signal to bring back the original unreliable
128 but stateful semantics.
129 - fixed a lot of typos in Coro.pm (patch submitted by David
130 Steinbrunner, which applied flawlessly).
131
1323.6 Sat Apr 14 17:13:31 CEST 2007
133 - added some bugfixes to get eg/myhttpd working again.
134 - added Coro::Storable for often-cede'ing freeze/thaw.
135 - try to do a clean exit when a coroutine calls exit
136 (EXPERIMENTAL).
137 - got rid of indirect call through _coro_init.
138 - updated the partly antique examples in eg/ to
139 work again and be a bit less magic, too.
140 - fixed Coro::Signal semantics to work as documented again.
141
1423.55 Sun Mar 25 01:20:47 CET 2007
143 - add SAVE_DEFFH to save the default filehandle and enable
144 it by default.
145 - finally move socket-operations from Coro::Socket to Coro::Handle
146 to be able to unblock foreign sockets.
147 - add Coro::State::save_also and guarded_save.
148 - add count accessor to Coro::Semaphore.
149 - add Coro::State::cctx_stacksize.
150 - just for the fun of it, do not rely on implicit context,
151 which can dramatically improve performance, but people
152 using windows-process-emulation perls do not care much about
153 performance.
154
1553.51 Sun Mar 4 14:18:04 CET 2007
156 - fixed a problem when you weakened references to Coro::State's
157 (patch by Yuval Kogman).
158
1593.501 Wed Feb 28 12:44:07 CET 2007
160 - rename some global symbols as macosx from hell redefines
161 them without asking.
162
1633.5 Tue Feb 13 20:22:53 CET 2007
164 - do AnyEvent model detection earlier, avoiding problems
165 caused by first using AnyEvent and later Coro::Event.
166 - implement and document Coro::Event event objects.
167 - fix a potential problem in Coro::Event causing crashes.
168 - initialise PL_comppad when creating a new coroutine,
169 avoids crashes on early coro destruction.
170
1713.41 Mon Jan 22 19:19:49 CET 2007
172 - readline on Coro::Handle did not support undefined $/,
173 nor did it deliver partial lines on EOF or error.
174 - implement malloc fallback for stack allocation because
175 stupid broken idiotic OSX has a stupid broken
176 idiotic fits-the-whole-os mmap "implementation" and
177 my dick feels longer if Coro is portable even to
178 obsolete platforms.
179
1803.4 Fri Jan 19 21:52:54 CET 2007
181 - remove t/09_timer.t, as it isn't really testing much
182 but was rather flaky in practise.
183 - async_pool coro would keep arguments and callback alive until
184 it was reused.
185 - cancellation of a coroutine could cause spurious idle calls
186 in cede_notself.
187
1883.3 Sat Jan 6 03:45:00 CET 2007
189 - implement $coro->on_destroy.
190 - Coro::Event blocking semantics have been changed,
191 documented and - hopefully - improved.
192 - fix nice adding, not subtracting, from priority.
193 - fix ->prio and api_is_ready (patch by Mark Hinds).
194 - fixed an assert ("... == prev__cctx->idle_te")
195 that could errronously trigger.
196 - fix various large and small memleaks.
197 - use a (hopefully) more stable cancel implementation
198 that immediately frees the coroutine data.
199 - cede/cede_notself return a status now.
200 - added Coro::guard function.
201 - added a global coroutine pool for jobs (on my machine,
202 I can create and execute 48k simple coros/s with async,
203 and 128k coros with async_pool).
204 - Coro::AIO now uses the coroutine priority as io priority.
205
2063.2 Fri Dec 22 05:07:09 CET 2006
207 - improve portability to slightly older perls.
208 - use cleaner coroutine destruction.
209 - simplify configuration for users.
210 - optionally (unrecommended) prefer perl functions over
211 their coro replacements.
212
2133.11 Tue Dec 5 13:11:24 CET 2006
214 - fixed some bogus assert's, but as perl.h disables assert even
215 without NDEBUG (thank you very much), not too many people should
216 notice (that did include myself). Andreas König noticed, though :)
217 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
218 it. Also noticed by Andreas König :)
219 - save/restore tainted status.
220 - verified to pass the testsuite on my 5.9.5.
221
2223.1 Mon Dec 4 23:03:40 CET 2006
223 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
224 - incompatible change: transfer flags are now per-state.
225 - give a better error message on deadlock.
226 - document Coro::nready.
227 - enhanced testsuite.
228
2293.01 Sun Dec 3 23:47:42 CET 2006
230 - forgot to include Coro::Timer.
231
2323.0 Sun Dec 3 22:57:25 CET 2006
233 - the "FINALLY COMPLETELY STABLE" release (coming soon:
234 the "FAMOUS LAST WORDS" release).
235 - implement a new stack sharing algorithm, which uses a stack
236 pool (size currently hardcoded).
237 - make stack sharing mandatory (it no longer uses a heuristic).
238 - eval/die no longer cause weird problems under heavy use.
239 - Coro::Event could cause livelocks if it was used but
240 no Coro::Event watchers were used.
241 - Coro::Event now uses asynccheck as prepare does not
242 check for changed watchers.
243 - Coro::Event allows multiple waiting coros and will wake up one
244 per event.
245 - Coro::Event should be cleaner and more efficient now.
246 - new utility function Coro::unblock_sub.
247 - document the sad fact that Event is no longer reentrant.
248 - putting a coroutine into the ready queue twice could under
249 some circumstances lead to stack corruption.
250 - minor incompatibility: subclassing Coro::State is supported
251 directly now without going through a _coro_state member.
252 - state/coro switching is much faster now.
253 - very minor optimisations and code/documentation cleanup.
254 - avoid problems due to compiler inlining.
255 - removed timers from Coro::Timer -> use AnyEvent instead.
256 - replaced Coro::idle coroutine by (cleaner) idle handler.
257 - updated to newest libcoro.
258 - implement enhanced support for valgrind.
259 - implement is_ready and return value for ready.
260 - removed Coro::Cont, it was a misnomer (it's generators),
261 and it was somewhat annoying to get it right. Will come back
262 if somebody asks for it :)
263 - many ->wait methods and Coro::Event could return
264 spuriously without the event having happened.
10 265
112.5 Tue Nov 7 12:22:33 CET 2006 2662.5 Tue Nov 7 12:22:33 CET 2006
12 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket 267 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
13 use AnyEvent, moved them to Coro/. 268 use AnyEvent, moved them to Coro/.
14 - added Coro::LWP which contains all the uglyness required to 269 - added Coro::LWP which contains all the uglyness required to

Diff Legend

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