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

Comparing Coro/Changes (file contents):
Revision 1.378 by root, Tue Sep 30 17:12:34 2008 UTC vs.
Revision 1.498 by root, Tue Jun 23 00:56:14 2009 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::coroaio. 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
84.8 11 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
12 as this seems to have been a difficult concept (poll, sleep, idle,
13 idle_upto).
14 - add Coro::AnyEvent::readable/writable functions.
15 - clarify Coro::EV/Event/AnyEvent manpages.
16 - free per-thread global scalars in the thread calling ->cancel, to
17 avoid crashes when $_, $@ etc., are magical but some of those
18 had already been freed.
19 - "unexperimentalize" the callback interface for Coro::Semaphore.
20 - speed up ready queue management/context switching by using a linked
21 list instead of an array (~5-10%).
22 - implement "watch" command in Coro::Debug shells.
23 - for fun, implement time-slicing as an example in the manpage.
24 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
25 Coro::Event, use the normal AnyEvent handling instead of dieing
26 (the same is true for Coro::Handle).
27 - properly document Coro::EV::timed_io_once.
28 - avoid unneccessary ->cancel calls in Coro::Handle.
29 - maybe make it work on mingw32 with win32 backend
30 (based on patch by Yasuhiro Matsumoto).
31
325.132 Fri May 29 09:00:39 CEST 2009
33 - do not keep a reference to the argument itself in
34 Coro::Semaphore::guard, as it could change later.
35 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
36 should support a prepare callback.
37
385.131 Mon Mar 16 23:20:37 CET 2009
39 - implement and document Coro->suspend, Coro->resume.
40 - fix Coro::Select implementation to not (often) close
41 the passed file descriptors (testcase provided by pippijn).
42
435.13 Mon Dec 15 21:51:42 CET 2008
44 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
45 - don't set diehook to C<undef> but instead to NULL, to avoid
46 spurious warnings.
47 - fix a lot of bugs in Coro::SemaphoreSet.
48 - Coro::SemaphoreSet will less often create a semaphore needlessly.
49 - add Coro::SemaphoreSet::count and wait methods.
50 - take advantage of the new Guard module.
51 - deprecate Coro::guard.
52 - try to fix the dreaded 01_unblock tests once more. I hate it when
53 testsuites need more fixing than the code they are supposed to test.
54 - croak in more cases when a required callback isn't resolvable.
55 - fix some minor issues in Coro::State->call/eval.
56 - use current coroutine context instead of a temporary one
57 when temporarily switching to another coroutine.
58 - do not call C-level on_destroy handlers during global destruction,
59 to avoid needless segfaults.
60
615.12 Sun Dec 7 13:30:38 CET 2008
62 - add default config for MirOS, which seems to be a bug-to-bug
63 compatible fork of openbsd ("world domination by releasing
64 openbsd cvs before the openbsd folks do it" or so :).
65 - free_padlist did destroy the names pad, not good, but didn't
66 seem to bother perl - this could fix issues such as eval ""
67 inside a function called from multiple coroutines.
68 - use a different method to detect destruction time.
69 - be more careful when freeing padlists just before global
70 destruction.
71 - fixed and expanded the call/cc example.
72 - renamed _terminate to _coro_run.
73 - new method Coro::Channel->shutdown.
74 - try pthreads on openbsd <4.4 (broken sigaltstack, will
75 pthreads fare better?).
76 - be less picky about destroying "a" running coroutine.
77
785.11 Tue Nov 25 21:49:05 CET 2008
79 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
80 working code with failed assertions, introducing more bugs than
81 it fixes, requiring elaborate workarounds :(
82
835.1 Mon Nov 24 08:54:59 CET 2008
84 - wrote a small introductory tutorial - Coro::Intro.
85 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
86 - Coro::Select did errornously dup the file descriptors
87 and didn't work with all AnyEvent backends.
88 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
89 LWP data transfers.
90 - disassociate c contexts from coro objects - this is agruably more
91 correct, but mostly allows sharing of cctxs between coro and state
92 objects, for added memory savings and speed increases.
93 - bumped $Coro::POOL_RSS up to 32kb by default.
94 - no longer set the optype to OP_CUSTOM, as B::* understandably
95 doesn't like this very much (and we *are* a type of entersub).
96 - implement state cloning, just to prove that call/cc can be done.
97 - automatically load Coro::AnyEvent in Coro::Handle.
98 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
99 do not have this method (and don't need it either).
100 - speed up generic anyevent methods in Coro::Handle by using rouse
101 callbacks.
102 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
103 others. It also makes the debugger happier, as you can trace
104 through the idle threads now.
105 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
106 - eg/event was pretty broken.
107 - better 5.8.6 compatibility.
108
1095.0 Thu Nov 20 10:35:05 CET 2008
110 - NEW ARCHITECTURE: use the latest 4.x version if you experience
111 stability issues.
112 - bump API version to 7 - all dependents must be recompiled.
113 - removed timed_* functions - they were not being used anyways
114 and should be replaced by a more generic mechanism -
115 and were annoying to support anyways :)
116 - removed SemaphoreSet's waiter method - use sem method instead.
117 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
118 - async_pool did free a scalar value twice
119 ("Attempt to unreference...").
120 - fix a longstanding bug where calling terminate on a coro that
121 was waiting for a semaphore that was just becoming available
122 would cause a deadlock (semaphore would get into a state where
123 it was available but waiters were still blocked).
124 - calling throw on a coroutine that is waiting for a semaphore will
125 no longer make it acquire the semaphore (and thus leak a count).
126 - perl's process emulation is now not even theoretically supported
127 anymore.
128 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
129 conversion of callback-style to blocking-style.
130 - new methods $coro->schedule_to and ->cede_to, to specifically
131 schedule or cede to a specific coroutine.
132 - new function Coro::Semaphore::wait.
133 - use named constants in Coro::Channel (Richard Hundt).
134 - directly patch the entersub opcode calling SLF functions (cede,
135 transfer and so on). this does speed up context switching, but
136 more importanly, it frees us from the hardcoded behaviour of
137 entersub, so we might actually be able to return something from
138 those functions and atcually create new ones.
139 - take advantage of __builtin_frame_address on gcc.
140 - expose THX in coroapi (not sure whether this was a wise decision,
141 as "threaded" perls are running at half speed anyways).
142 - implement execute_slf (schedule-like-function) interface that makes
143 it possible to implement schedule-like-functions in XS.
144 - use new SLF interface to massively speed up Coro::EV by roughly a
145 factor of two.
146 - used new SLF interface to massively speed up Coro::Semaphore by a
147 factor of three.
148 - used new SLF interface to speed up Coro::AIO by roughly a factor of
149 four and reduce its memory usage considerably.
150 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
151 for a nice speedup and vastly more correct behaviour. Also implement
152 a new method "sem" to get at the underlying semaphore object.
153 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
154 (in comparison) 20-40% speedup.
155 - used new SLF interface to reimplement Coro::Signal gaining
156 some unknown (because I was too lazy), but certain, speedup, and also
157 making signals reliable for the first time.
158 - used new SLF interface and other optimisations to speed up async_pool
159 by a factor of two. It also doesn't rely on perl's exception mechanism
160 to exit anymore. The overhead for terminating an async_pool, coro over
161 a normal async is now very small.
162 - sped up coroutine creation/destruction by 40%.
163 - forgot to include Coro/libcoro/README in the dist for all these years.
164 - work around a freebsd pthreads bug (manual testcancel is required as
165 pthread_cond_wait isn't a cancellation point on freebsd).
166 - use new rouse functions to speed up and simplify Coro::BDB.
167 - make "prefer perl native functions" work with threaded perls.
168 - condense Coro::Debug ps output, hint at v and w flags.
169 - (libcoro) lots of minor cleanups and portability improvements.
170
1714.914 Wed Nov 19 12:54:18 CET 2008
172 - fix a disastrous bug in the readline optimisation
173 introduced in 4.801.
174
1754.913 Sat Nov 15 07:58:28 CET 2008
176 - async_pool did free a scalar value twice
177 ("Attempt to unreference...").
178
1794.912 Thu Nov 13 18:31:23 CET 2008
180 - minor cleanups.
181 - use much larger stacks on linux and perl < 5.8.8.
182 - Coro::Debug::new_unix_server did not unlink the socket
183 when destroyed.
184
1854.911 Tue Nov 11 04:26:17 CET 2008
186 - "port" to threaded perls.
187
1884.91 Mon Nov 10 05:36:38 CET 2008
189 - the ->throw exception object no longer leaks.
190 - creating a new cctx leaked a scopestack entry (memleak).
191 - new coroutines didn't get created with a zero flags field
192 (unknown impact).
193 - calling ->throw on a not-yet-started coroutine should now work
194 instead of being ignored.
195 - ->throw is now supported on Coro::State objects.
196 - clean up cctx creation code a bit.
197 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
198
1994.9 Sat Nov 8 17:45:27 CET 2008
200 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
201 - (libcoro) no longer leak threads in the experimental pthread backend,
202 also speed it up considerably.
203 - (libcoro) do not rely on makecontext passing void *'s unscathed.
204 - fix compiletime dependencies on libcoro in the Makefile.
205 - cctx_count wasn't always updated properly.
206 - Coro::State::cctx_stacksize wasn't applied correctly.
207 - new function Coro::State::cctx_max_idle.
208 - the default max number of idle C contexts is now 4.
209 - (libcoro) try harder to get _setjmp/_longjmp.
210 - (libcoro) cleanup and extend the libcoro API to officially
211 allow the creation of empty source contexts.
212 - very experimental workaround for the totally broken netbsd platform.
213 - tried to hack around openbsd bugs.
214
2154.804 Wed Nov 5 16:36:00 CET 2008
216 - Coro::Debug::new_unix_server would not create a non-blocking listening
217 socket, sometimes causing freezes.
218 - (libcoro) fix misaligned stack points for setjmp and assembly
219 methods, which can cause crashes on x86/x86_64 with a sufficiently
220 aggressive compiler.
221 - new function: Coro::Debug::new_tcp_server.
222 - move ->throw into the Coro class because it only works on coro objects.
223
2244.803 Mon Nov 3 17:16:12 CET 2008
225 - (libcoro) use a global asm statement to become independent of gcc
226 otpimisations for CORO_ASM (thanks to pippijn for the idea).
227 - try to workaround yet another broken bsd, this time dragonfly.
228
2294.802 Thu Oct 30 10:56:12 CET 2008
230 - support -fno-omit-frame-pointer on x86 with the assembly method.
231 - tune 01_unblock.t tests a bit.
232
2334.801 Wed Oct 22 18:33:37 CEST 2008
234 - improve readline speed for very long "lines".
235 - backport to 5.8.8.
236
2374.8 Thu Oct 2 13:34:40 CEST 2008
9 - new function Coro::AIO::aio_wait. 238 - new function Coro::AIO::aio_wait.
10 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". 239 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
11 - greatly speed up and reduce memory usage of Coro::AIO requests. 240 - greatly speed up and reduce memory usage of Coro::AIO requests.
241 - implement some other ยต-optimisations.
12 242
134.749 Mon Sep 29 14:40:12 CEST 2008 2434.749 Mon Sep 29 14:40:12 CEST 2008
14 - port to slow and broken pseudo-threaded perls. (courtesy pippijn). 244 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
15 245
164.748 Sat Sep 27 14:03:19 CEST 2008 2464.748 Sat Sep 27 14:03:19 CEST 2008

Diff Legend

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