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

Comparing Coro/Changes (file contents):
Revision 1.398 by root, Sun Nov 9 23:08:49 2008 UTC vs.
Revision 1.509 by root, Wed Jul 8 02:16:49 2009 UTC

… …
2 2
3TODO: should explore PerlIO::coroaio (perl leaks like hell). 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!), 5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)). 6 but very lacking in others (higher level ipc)).
7TODO: unready_all
8TODO: myhttpd header parsing
9TODO: channel->maxsize(newsize)?
7 10
84.91 11 - considerably speed up Coro::Select by taking avdantage
12 of AnyEvent > 4.8 and some other optimisations.
13
145.151 Mon Jul 6 05:41:57 CEST 2009
15 - backport to windows process emulation code again (patch by
16 Yasuhiro MATSUMOTO).
17 - slightly update Coro::MakeMaker.
18
195.15 Tue Jun 30 10:28:06 CEST 2009
20 - deprecate Coro::Socket, document how to get ipv6 support via
21 AnyEvent::Socket instead.
22 - implement signal->wait ($cb) interface, similar to semaphores.
23 - work around SvOK not supporting getmagic, so we have to getmagic
24 to test for undef :/ (reported by Matthias Waldorf).
25 - load Coro::AnyEvent in all modules using AnyEvent.
26 - work around perl corrupting our internal data structures,
27 reported by Tokuhiro Matsuno.
28 - enable per-coroutine real and cpu time gathering
29 (Coro::State::enable_times).
30
315.14 Wed Jun 24 01:37:48 CEST 2009
32 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
33 as this seems to have been a difficult concept (poll, sleep, idle,
34 idle_upto).
35 - add Coro::AnyEvent::readable/writable functions.
36 - clarify Coro::EV/Event/AnyEvent manpages.
37 - free per-thread global scalars in the thread calling ->cancel, to
38 avoid crashes when $_, $@ etc., are magical but some of those
39 had already been freed.
40 - "unexperimentalise" the callback interface for Coro::Semaphore.
41 - speed up ready queue management/context switching by using a linked
42 list instead of an array (~5-10%).
43 - implement "watch" command in Coro::Debug shells.
44 - for fun, implement time-slicing as an example in the manpage.
45 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
46 Coro::Event, use the normal AnyEvent handling instead of dieing
47 (the same is true for Coro::Handle).
48 - properly document Coro::EV::timed_io_once.
49 - avoid unneccessary ->cancel calls in Coro::Handle.
50 - maybe make it work on mingw32 with win32 backend
51 (based on patch by Yasuhiro Matsumoto).
52
535.132 Fri May 29 09:00:39 CEST 2009
54 - do not keep a reference to the argument itself in
55 Coro::Semaphore::guard, as it could change later.
56 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
57 should support a prepare callback.
58
595.131 Mon Mar 16 23:20:37 CET 2009
60 - implement and document Coro->suspend, Coro->resume.
61 - fix Coro::Select implementation to not (often) close
62 the passed file descriptors (testcase provided by pippijn).
63
645.13 Mon Dec 15 21:51:42 CET 2008
65 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
66 - don't set diehook to C<undef> but instead to NULL, to avoid
67 spurious warnings.
68 - fix a lot of bugs in Coro::SemaphoreSet.
69 - Coro::SemaphoreSet will less often create a semaphore needlessly.
70 - add Coro::SemaphoreSet::count and wait methods.
71 - take advantage of the new Guard module.
72 - deprecate Coro::guard.
73 - try to fix the dreaded 01_unblock tests once more. I hate it when
74 testsuites need more fixing than the code they are supposed to test.
75 - croak in more cases when a required callback isn't resolvable.
76 - fix some minor issues in Coro::State->call/eval.
77 - use current coroutine context instead of a temporary one
78 when temporarily switching to another coroutine.
79 - do not call C-level on_destroy handlers during global destruction,
80 to avoid needless segfaults.
81
825.12 Sun Dec 7 13:30:38 CET 2008
83 - add default config for MirOS, which seems to be a bug-to-bug
84 compatible fork of openbsd ("world domination by releasing
85 openbsd cvs before the openbsd folks do it" or so :).
86 - free_padlist did destroy the names pad, not good, but didn't
87 seem to bother perl - this could fix issues such as eval ""
88 inside a function called from multiple coroutines.
89 - use a different method to detect destruction time.
90 - be more careful when freeing padlists just before global
91 destruction.
92 - fixed and expanded the call/cc example.
93 - renamed _terminate to _coro_run.
94 - new method Coro::Channel->shutdown.
95 - try pthreads on openbsd <4.4 (broken sigaltstack, will
96 pthreads fare better?).
97 - be less picky about destroying "a" running coroutine.
98
995.11 Tue Nov 25 21:49:05 CET 2008
100 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
101 working code with failed assertions, introducing more bugs than
102 it fixes, requiring elaborate workarounds :(
103
1045.1 Mon Nov 24 08:54:59 CET 2008
105 - wrote a small introductory tutorial - Coro::Intro.
106 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
107 - Coro::Select did errornously dup the file descriptors
108 and didn't work with all AnyEvent backends.
109 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
110 LWP data transfers.
111 - disassociate c contexts from coro objects - this is agruably more
112 correct, but mostly allows sharing of cctxs between coro and state
113 objects, for added memory savings and speed increases.
114 - bumped $Coro::POOL_RSS up to 32kb by default.
115 - no longer set the optype to OP_CUSTOM, as B::* understandably
116 doesn't like this very much (and we *are* a type of entersub).
117 - implement state cloning, just to prove that call/cc can be done.
118 - automatically load Coro::AnyEvent in Coro::Handle.
119 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
120 do not have this method (and don't need it either).
121 - speed up generic anyevent methods in Coro::Handle by using rouse
122 callbacks.
123 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
124 others. It also makes the debugger happier, as you can trace
125 through the idle threads now.
126 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
127 - eg/event was pretty broken.
128 - better 5.8.6 compatibility.
129
1305.0 Thu Nov 20 10:35:05 CET 2008
131 - NEW ARCHITECTURE: use the latest 4.x version if you experience
132 stability issues.
133 - bump API version to 7 - all dependents must be recompiled.
134 - removed timed_* functions - they were not being used anyways
135 and should be replaced by a more generic mechanism -
136 and were annoying to support anyways :)
137 - removed SemaphoreSet's waiter method - use sem method instead.
138 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
139 - async_pool did free a scalar value twice
140 ("Attempt to unreference...").
141 - fix a longstanding bug where calling terminate on a coro that
142 was waiting for a semaphore that was just becoming available
143 would cause a deadlock (semaphore would get into a state where
144 it was available but waiters were still blocked).
145 - calling throw on a coroutine that is waiting for a semaphore will
146 no longer make it acquire the semaphore (and thus leak a count).
147 - perl's process emulation is now not even theoretically supported
148 anymore.
149 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
150 conversion of callback-style to blocking-style.
151 - new methods $coro->schedule_to and ->cede_to, to specifically
152 schedule or cede to a specific coroutine.
153 - new function Coro::Semaphore::wait.
154 - use named constants in Coro::Channel (Richard Hundt).
155 - directly patch the entersub opcode calling SLF functions (cede,
156 transfer and so on). this does speed up context switching, but
157 more importanly, it frees us from the hardcoded behaviour of
158 entersub, so we might actually be able to return something from
159 those functions and atcually create new ones.
160 - take advantage of __builtin_frame_address on gcc.
161 - expose THX in coroapi (not sure whether this was a wise decision,
162 as "threaded" perls are running at half speed anyways).
163 - implement execute_slf (schedule-like-function) interface that makes
164 it possible to implement schedule-like-functions in XS.
165 - use new SLF interface to massively speed up Coro::EV by roughly a
166 factor of two.
167 - used new SLF interface to massively speed up Coro::Semaphore by a
168 factor of three.
169 - used new SLF interface to speed up Coro::AIO by roughly a factor of
170 four and reduce its memory usage considerably.
171 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
172 for a nice speedup and vastly more correct behaviour. Also implement
173 a new method "sem" to get at the underlying semaphore object.
174 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
175 (in comparison) 20-40% speedup.
176 - used new SLF interface to reimplement Coro::Signal gaining
177 some unknown (because I was too lazy), but certain, speedup, and also
178 making signals reliable for the first time.
179 - used new SLF interface and other optimisations to speed up async_pool
180 by a factor of two. It also doesn't rely on perl's exception mechanism
181 to exit anymore. The overhead for terminating an async_pool, coro over
182 a normal async is now very small.
183 - sped up coroutine creation/destruction by 40%.
184 - forgot to include Coro/libcoro/README in the dist for all these years.
185 - work around a freebsd pthreads bug (manual testcancel is required as
186 pthread_cond_wait isn't a cancellation point on freebsd).
187 - use new rouse functions to speed up and simplify Coro::BDB.
188 - make "prefer perl native functions" work with threaded perls.
189 - condense Coro::Debug ps output, hint at v and w flags.
190 - (libcoro) lots of minor cleanups and portability improvements.
191
1924.914 Wed Nov 19 12:54:18 CET 2008
193 - fix a disastrous bug in the readline optimisation
194 introduced in 4.801.
195
1964.913 Sat Nov 15 07:58:28 CET 2008
197 - async_pool did free a scalar value twice
198 ("Attempt to unreference...").
199
2004.912 Thu Nov 13 18:31:23 CET 2008
201 - minor cleanups.
202 - use much larger stacks on linux and perl < 5.8.8.
203 - Coro::Debug::new_unix_server did not unlink the socket
204 when destroyed.
205
2064.911 Tue Nov 11 04:26:17 CET 2008
207 - "port" to threaded perls.
208
2094.91 Mon Nov 10 05:36:38 CET 2008
210 - the ->throw exception object no longer leaks.
211 - creating a new cctx leaked a scopestack entry (memleak).
9 - typo: new coroutines didn't get created with a zero flags field. 212 - new coroutines didn't get created with a zero flags field
213 (unknown impact).
214 - calling ->throw on a not-yet-started coroutine should now work
215 instead of being ignored.
216 - ->throw is now supported on Coro::State objects.
10 - clean up cctx creation code a bit. 217 - clean up cctx creation code a bit.
218 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
11 219
124.9 Sat Nov 8 17:45:27 CET 2008 2204.9 Sat Nov 8 17:45:27 CET 2008
13 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 221 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
14 - (libcoro) no longer leak threads in the experimental pthread backend, 222 - (libcoro) no longer leak threads in the experimental pthread backend,
15 also speed it up considerably. 223 also speed it up considerably.

Diff Legend

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