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

Comparing Coro/Changes (file contents):
Revision 1.163 by root, Fri Dec 1 02:17:37 2006 UTC vs.
Revision 1.280 by root, Tue Oct 23 01:48:33 2007 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
33.0 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
84.13
9 - add Coro::Storable::blocking_thaw.
10 - use a vastly more complicated technique to localise
11 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
12 - use a coroutine for the idle callback Coro::Event,
13 instead of running Event in the current coroutine context.
14 This also catches recursive invocations.
15 - actually report fork errors in gethostbyname and inet_aton.
16
174.11 Thu Oct 11 02:40:24 CEST 2007
18 - port to threaded perls.
19
204.1 Thu Oct 11 02:38:16 CEST 2007
21 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
22 be local to each coro (see Coro::State).
23 - incompatible change: for very deep reasons, cede and cede_notself
24 cannot return anything, so nothing will be returned.
25 - possibly bring back 5.10 compatibility (untested).
26 - work around stupid (and wrong) warning on 5.10 :(.
27 - overlay the saved state over the context stack. This saves
28 a few hundred bytes per coroutine on average and also
29 speeds up context switching a bit.
30 - further tune default stack sizes.
31 - (more) correctly calculate stack usage in coro_rss.
32 - Coro::Storable::blocking_* did not properly lock
33 resulting in races between coroutines.
34 - added Coro::Storable::guard.
35 - stopping to trace a coroutine could destroy the cctx of
36 an unrelated coroutine.
37 - explain the relationship between Perl and C coroutines in
38 more detail in Coro::State.
39 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
40 causing a fork per resolve. This also affected Coro::Socket.
41 - switch to a separate stack in $coro->call/eval to avoid
42 invalidating pointers.
43
444.03 Sat Oct 6 21:24:00 CEST 2007
45 - added Coro::throw method.
46 - minor code cleanups.
47
484.02 Sat Oct 6 02:36:47 CEST 2007
49 - fix a very minor per-coroutine memleak (a single codereference).
50 - fixed a bug where the currently in-use c context would be freed
51 prematurely (can happen only when programs change the stacksize
52 or use tracing).
53 - tracing can no longer keep a coro alive after it terminated.
54 - do static branch prediction in the common path for gcc. gives
55 about 2-5% speed improvement here.
56
574.01 Fri Oct 5 22:10:49 CEST 2007
58 - instead of recreating *a* standard output handle we simply
59 use STDOUT, which is faster and hopefully more robust.
60
614.0 Fri Oct 5 12:56:00 CEST 2007
62 - incompatibly changed Coro::Storable::freeze.
63 - major new feature: added Coro::Debug, for interactive coroutine
64 debugging, tracing and much more.
65 - major bug fix: unbelievable, but true: $_, $/ and many other
66 "saved" variables actually weren't being saved. This has been fixed,
67 of course, while increasing performance while losing all the save
68 flags.
69 - save flags are gone, and all the api functions dealing with them.
70 - added Coro::Semaphore::adjust.
71 - added Coro::Util::fork_eval.
72 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
73 - added Coro::killall.
74 - reduce initial stack sizes to allow for "micro-coroutines".
75 - better async_pool resource management, moved parts of async_pool
76 handling to XS (major speed improvement).
77 - actually croak before modifying important data structures.
78 - refuse to transfer while compiling.
79 - possibly support eval EXPR better now.
80 - enable assembly per default on linux+bsd x86+amd64.
81 - all internal members were renamed _something for easier subclassing.
82 - many minor tweaks.
83
843.63 Wed May 16 14:10:06 CEST 2007
85 - implement handcoded assembly for x86/amd64 SVR ABI.
86
873.62 Fri Apr 27 21:36:06 CEST 2007
88 - upgrade libcoro (which might set unwind info correctly).
89 - change default on linux to setjmp/longjmp.
90
913.61 Thu Apr 19 12:36:18 CEST 2007
92 - Coro::Storable caused an endless loop when thawing invalid
93 pst's sometimes.
94 - use a Semaphore in Coro::Storable, as Storable doesn't
95 seem to be reentrant (although it is documented to
96 be threadsafe...).
97 - fix Coro::Signal to bring back the original unreliable
98 but stateful semantics.
99 - fixed a lot of typos in Coro.pm (patch submitted by David
100 Steinbrunner, which applied flawlessly).
101
1023.6 Sat Apr 14 17:13:31 CEST 2007
103 - added some bugfixes to get eg/myhttpd working again.
104 - added Coro::Storable for often-cede'ing freeze/thaw.
105 - try to do a clean exit when a coroutine calls exit
106 (EXPERIMENTAL).
107 - got rid of indirect call through _coro_init.
108 - updated the partly antique examples in eg/ to
109 work again and be a bit less magic, too.
110 - fixed Coro::Signal semantics to work as documented again.
111
1123.55 Sun Mar 25 01:20:47 CET 2007
113 - add SAVE_DEFFH to save the default filehandle and enable
114 it by default.
115 - finally move socket-operations from Coro::Socket to Coro::Handle
116 to be able to unblock foreign sockets.
117 - add Coro::State::save_also and guarded_save.
118 - add count accessor to Coro::Semaphore.
119 - add Coro::State::cctx_stacksize.
120 - just for the fun of it, do not rely on implicit context,
121 which can dramatically improve performance, but people
122 using windows-process-emulation perls do not care much about
123 performance.
124
1253.51 Sun Mar 4 14:18:04 CET 2007
126 - fixed a problem when you weakened references to Coro::State's
127 (patch by Yuval Kogman).
128
1293.501 Wed Feb 28 12:44:07 CET 2007
130 - rename some global symbols as macosx from hell redefines
131 them without asking.
132
1333.5 Tue Feb 13 20:22:53 CET 2007
134 - do AnyEvent model detection earlier, avoiding problems
135 caused by first using AnyEvent and later Coro::Event.
136 - implement and document Coro::Event event objects.
137 - fix a potential problem in Coro::Event causing crashes.
138 - initialise PL_comppad when creating a new coroutine,
139 avoids crashes on early coro destruction.
140
1413.41 Mon Jan 22 19:19:49 CET 2007
142 - readline on Coro::Handle did not support undefined $/,
143 nor did it deliver partial lines on EOF or error.
144 - implement malloc fallback for stack allocation because
145 stupid broken idiotic OSX has a stupid broken
146 idiotic fits-the-whole-os mmap "implementation" and
147 my dick feels longer if Coro is portable even to
148 obsolete platforms.
149
1503.4 Fri Jan 19 21:52:54 CET 2007
151 - remove t/09_timer.t, as it isn't really testing much
152 but was rather flaky in practise.
153 - async_pool coro would keep arguments and callback alive until
154 it was reused.
155 - cancellation of a coroutine could cause spurious idle calls
156 in cede_notself.
157
1583.3 Sat Jan 6 03:45:00 CET 2007
159 - implement $coro->on_destroy.
160 - Coro::Event blocking semantics have been changed,
161 documented and - hopefully - improved.
162 - fix nice adding, not subtracting, from priority.
163 - fix ->prio and api_is_ready (patch by Mark Hinds).
164 - fixed an assert ("... == prev__cctx->idle_te")
165 that could errronously trigger.
166 - fix various large and small memleaks.
167 - use a (hopefully) more stable cancel implementation
168 that immediately frees the coroutine data.
169 - cede/cede_notself return a status now.
170 - added Coro::guard function.
171 - added a global coroutine pool for jobs (on my machine,
172 I can create and execute 48k simple coros/s with async,
173 and 128k coros with async_pool).
174 - Coro::AIO now uses the coroutine priority as io priority.
175
1763.2 Fri Dec 22 05:07:09 CET 2006
177 - improve portability to slightly older perls.
178 - use cleaner coroutine destruction.
179 - simplify configuration for users.
180 - optionally (unrecommended) prefer perl functions over
181 their coro replacements.
182
1833.11 Tue Dec 5 13:11:24 CET 2006
184 - fixed some bogus assert's, but as perl.h disables assert even
185 without NDEBUG (thank you very much), not too many people should
186 notice (that did include myself). Andreas König noticed, though :)
187 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
188 it. Also noticed by Andreas König :)
189 - save/restore tainted status.
190 - verified to pass the testsuite on my 5.9.5.
191
1923.1 Mon Dec 4 23:03:40 CET 2006
193 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
194 - incompatible change: transfer flags are now per-state.
195 - give a better error message on deadlock.
196 - document Coro::nready.
197 - enhanced testsuite.
198
1993.01 Sun Dec 3 23:47:42 CET 2006
200 - forgot to include Coro::Timer.
201
2023.0 Sun Dec 3 22:57:25 CET 2006
4 - the "FINALLY COMPLETELY STABLE" release (coming soon: 203 - the "FINALLY COMPLETELY STABLE" release (coming soon:
5 the "FAMOUS LAST WORDS" release). 204 the "FAMOUS LAST WORDS" release).
6 - implement a new stack sharing algorithm. 205 - implement a new stack sharing algorithm, which uses a stack
206 pool (size currently hardcoded).
7 - make stack sharing mandatory (it no longer uses a heuristic). 207 - make stack sharing mandatory (it no longer uses a heuristic).
8 - eval/die no longer cause weird problems under heavy use. 208 - eval/die no longer cause weird problems under heavy use.
9 - Coro::Event could cause livelocks if it was used but 209 - Coro::Event could cause livelocks if it was used but
10 no Coro::Event watchers were used. 210 no Coro::Event watchers were used.
211 - Coro::Event now uses asynccheck as prepare does not
212 check for changed watchers.
213 - Coro::Event allows multiple waiting coros and will wake up one
214 per event.
11 - Coro::Event should be cleaner and more efficient now. 215 - Coro::Event should be cleaner and more efficient now.
216 - new utility function Coro::unblock_sub.
217 - document the sad fact that Event is no longer reentrant.
12 - putting a coroutine into the ready queue twice could under 218 - putting a coroutine into the ready queue twice could under
13 some circumstances lead to stack corruption. 219 some circumstances lead to stack corruption.
14 - minor incompatibility: subclassing Coro::State is supported 220 - minor incompatibility: subclassing Coro::State is supported
15 directly now without going through a _coro_state member. 221 directly now without going through a _coro_state member.
16 - state/coro switching is much faster now. 222 - state/coro switching is much faster now.
… …
19 - removed timers from Coro::Timer -> use AnyEvent instead. 225 - removed timers from Coro::Timer -> use AnyEvent instead.
20 - replaced Coro::idle coroutine by (cleaner) idle handler. 226 - replaced Coro::idle coroutine by (cleaner) idle handler.
21 - updated to newest libcoro. 227 - updated to newest libcoro.
22 - implement enhanced support for valgrind. 228 - implement enhanced support for valgrind.
23 - implement is_ready and return value for ready. 229 - implement is_ready and return value for ready.
230 - removed Coro::Cont, it was a misnomer (it's generators),
231 and it was somewhat annoying to get it right. Will come back
232 if somebody asks for it :)
24 - many ->wait methods and Coro::Event could return 233 - many ->wait methods and Coro::Event could return
25 spuriously without the event having happened. 234 spuriously without the event having happened.
26 235
272.5 Tue Nov 7 12:22:33 CET 2006 2362.5 Tue Nov 7 12:22:33 CET 2006
28 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket 237 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket

Diff Legend

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