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

Comparing Coro/Changes (file contents):
Revision 1.419 by root, Sun Nov 16 08:59:16 2008 UTC vs.
Revision 1.504 by root, Mon Jun 29 06:14:23 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
8TODO: testcancel 11 - deprecate Coro::Socket, document how to get ipv6 support via
9TODO: aio 12 AnyEvent::Socket instead.
10TODO: Coro::Semaphore 13 - implement signal->wait ($cb) interface, similar to semaphores.
115.0 14 - work around SvOK not supporting getmagic, so we have to getmagic
12 - NEW ARCHITECTURE: use 4.912 for the stable version. 15 to test for undef :/ (reported by Matthias Waldorf).
16 - load Coro::AnyEvent in all modules using AnyEvent.
17 - work around perl corrupting our internal data structures,
18 reported by Tokuhiro Matsuno.
19 - enable per-coroutine real and cpu time gathering
20 (Coro::State::enable_times).
21
225.14 Wed Jun 24 01:37:48 CEST 2009
23 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
24 as this seems to have been a difficult concept (poll, sleep, idle,
25 idle_upto).
26 - add Coro::AnyEvent::readable/writable functions.
27 - clarify Coro::EV/Event/AnyEvent manpages.
28 - free per-thread global scalars in the thread calling ->cancel, to
29 avoid crashes when $_, $@ etc., are magical but some of those
30 had already been freed.
31 - "unexperimentalise" the callback interface for Coro::Semaphore.
32 - speed up ready queue management/context switching by using a linked
33 list instead of an array (~5-10%).
34 - implement "watch" command in Coro::Debug shells.
35 - for fun, implement time-slicing as an example in the manpage.
36 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
37 Coro::Event, use the normal AnyEvent handling instead of dieing
38 (the same is true for Coro::Handle).
39 - properly document Coro::EV::timed_io_once.
40 - avoid unneccessary ->cancel calls in Coro::Handle.
41 - maybe make it work on mingw32 with win32 backend
42 (based on patch by Yasuhiro Matsumoto).
43
445.132 Fri May 29 09:00:39 CEST 2009
45 - do not keep a reference to the argument itself in
46 Coro::Semaphore::guard, as it could change later.
47 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
48 should support a prepare callback.
49
505.131 Mon Mar 16 23:20:37 CET 2009
51 - implement and document Coro->suspend, Coro->resume.
52 - fix Coro::Select implementation to not (often) close
53 the passed file descriptors (testcase provided by pippijn).
54
555.13 Mon Dec 15 21:51:42 CET 2008
56 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
57 - don't set diehook to C<undef> but instead to NULL, to avoid
58 spurious warnings.
59 - fix a lot of bugs in Coro::SemaphoreSet.
60 - Coro::SemaphoreSet will less often create a semaphore needlessly.
61 - add Coro::SemaphoreSet::count and wait methods.
62 - take advantage of the new Guard module.
63 - deprecate Coro::guard.
64 - try to fix the dreaded 01_unblock tests once more. I hate it when
65 testsuites need more fixing than the code they are supposed to test.
66 - croak in more cases when a required callback isn't resolvable.
67 - fix some minor issues in Coro::State->call/eval.
68 - use current coroutine context instead of a temporary one
69 when temporarily switching to another coroutine.
70 - do not call C-level on_destroy handlers during global destruction,
71 to avoid needless segfaults.
72
735.12 Sun Dec 7 13:30:38 CET 2008
74 - add default config for MirOS, which seems to be a bug-to-bug
75 compatible fork of openbsd ("world domination by releasing
76 openbsd cvs before the openbsd folks do it" or so :).
77 - free_padlist did destroy the names pad, not good, but didn't
78 seem to bother perl - this could fix issues such as eval ""
79 inside a function called from multiple coroutines.
80 - use a different method to detect destruction time.
81 - be more careful when freeing padlists just before global
82 destruction.
83 - fixed and expanded the call/cc example.
84 - renamed _terminate to _coro_run.
85 - new method Coro::Channel->shutdown.
86 - try pthreads on openbsd <4.4 (broken sigaltstack, will
87 pthreads fare better?).
88 - be less picky about destroying "a" running coroutine.
89
905.11 Tue Nov 25 21:49:05 CET 2008
91 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
92 working code with failed assertions, introducing more bugs than
93 it fixes, requiring elaborate workarounds :(
94
955.1 Mon Nov 24 08:54:59 CET 2008
96 - wrote a small introductory tutorial - Coro::Intro.
97 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
98 - Coro::Select did errornously dup the file descriptors
99 and didn't work with all AnyEvent backends.
100 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
101 LWP data transfers.
102 - disassociate c contexts from coro objects - this is agruably more
103 correct, but mostly allows sharing of cctxs between coro and state
104 objects, for added memory savings and speed increases.
105 - bumped $Coro::POOL_RSS up to 32kb by default.
106 - no longer set the optype to OP_CUSTOM, as B::* understandably
107 doesn't like this very much (and we *are* a type of entersub).
108 - implement state cloning, just to prove that call/cc can be done.
109 - automatically load Coro::AnyEvent in Coro::Handle.
110 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
111 do not have this method (and don't need it either).
112 - speed up generic anyevent methods in Coro::Handle by using rouse
113 callbacks.
114 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
115 others. It also makes the debugger happier, as you can trace
116 through the idle threads now.
117 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
118 - eg/event was pretty broken.
119 - better 5.8.6 compatibility.
120
1215.0 Thu Nov 20 10:35:05 CET 2008
122 - NEW ARCHITECTURE: use the latest 4.x version if you experience
123 stability issues.
13 - bump API version to 7 - all dependents must be recompiled. 124 - bump API version to 7 - all dependents must be recompiled.
125 - removed timed_* functions - they were not being used anyways
126 and should be replaced by a more generic mechanism -
127 and were annoying to support anyways :)
128 - removed SemaphoreSet's waiter method - use sem method instead.
14 - Coro::Semaphore->adjust didn't correctly wake up enough waiters. 129 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
15 - async_pool did free a scalar value twice 130 - async_pool did free a scalar value twice
16 ("Attempt to unreference..."). 131 ("Attempt to unreference...").
132 - fix a longstanding bug where calling terminate on a coro that
133 was waiting for a semaphore that was just becoming available
134 would cause a deadlock (semaphore would get into a state where
135 it was available but waiters were still blocked).
136 - calling throw on a coroutine that is waiting for a semaphore will
137 no longer make it acquire the semaphore (and thus leak a count).
17 - perl's process emulation is now not even theoretically supported 138 - perl's process emulation is now not even theoretically supported
18 anymore. 139 anymore.
140 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
141 conversion of callback-style to blocking-style.
142 - new methods $coro->schedule_to and ->cede_to, to specifically
143 schedule or cede to a specific coroutine.
144 - new function Coro::Semaphore::wait.
145 - use named constants in Coro::Channel (Richard Hundt).
19 - directly patch the entersub opcode calling SLF functions (cede, 146 - directly patch the entersub opcode calling SLF functions (cede,
20 transfer and so on). this does speed up context switching, but 147 transfer and so on). this does speed up context switching, but
21 more importanly, it frees us from the hardcoded behaviour of 148 more importanly, it frees us from the hardcoded behaviour of
22 entersub, so we might actually be able to return something from 149 entersub, so we might actually be able to return something from
23 those functions and atcually create new ones. 150 those functions and atcually create new ones.
24 - take advantage of __builtin_frame_address on gcc. 151 - take advantage of __builtin_frame_address on gcc.
25 - expose THX in coroapi (not sure whether this was a wise decision, 152 - expose THX in coroapi (not sure whether this was a wise decision,
26 as "threaded" perls are running at half speed anyways). 153 as "threaded" perls are running at half speed anyways).
27 - implement execute_slf (schedule-like-function) interface that makes 154 - implement execute_slf (schedule-like-function) interface that makes
28 it possible to implement schedule-like-functions in XS. 155 it possible to implement schedule-like-functions in XS.
29 - use new SLF interface to massively speed up Coro::EV by roughly a factor 156 - use new SLF interface to massively speed up Coro::EV by roughly a
30 of two. 157 factor of two.
31 - removed timed_down and timed_guard functions - they were not being used
32 anyways and should be replaced by a more generic mechanism - and were
33 annoying to support anyways :
34 - used new SLF interface to massively speed up Coro::Semaphore by a factor 158 - used new SLF interface to massively speed up Coro::Semaphore by a
35 of three. 159 factor of three.
160 - used new SLF interface to speed up Coro::AIO by roughly a factor of
161 four and reduce its memory usage considerably.
162 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
163 for a nice speedup and vastly more correct behaviour. Also implement
164 a new method "sem" to get at the underlying semaphore object.
165 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
166 (in comparison) 20-40% speedup.
167 - used new SLF interface to reimplement Coro::Signal gaining
168 some unknown (because I was too lazy), but certain, speedup, and also
169 making signals reliable for the first time.
170 - used new SLF interface and other optimisations to speed up async_pool
171 by a factor of two. It also doesn't rely on perl's exception mechanism
172 to exit anymore. The overhead for terminating an async_pool, coro over
173 a normal async is now very small.
174 - sped up coroutine creation/destruction by 40%.
36 - forgot to include Coro/libcoro/README in the dist for all these years. 175 - forgot to include Coro/libcoro/README in the dist for all these years.
37 - work around a freebsd pthreads bug (manual testcancel is required as 176 - work around a freebsd pthreads bug (manual testcancel is required as
38 pthread_cond_wait isn't a cancellation point on freebsd). 177 pthread_cond_wait isn't a cancellation point on freebsd).
178 - use new rouse functions to speed up and simplify Coro::BDB.
179 - make "prefer perl native functions" work with threaded perls.
180 - condense Coro::Debug ps output, hint at v and w flags.
181 - (libcoro) lots of minor cleanups and portability improvements.
182
1834.914 Wed Nov 19 12:54:18 CET 2008
184 - fix a disastrous bug in the readline optimisation
185 introduced in 4.801.
39 186
404.913 Sat Nov 15 07:58:28 CET 2008 1874.913 Sat Nov 15 07:58:28 CET 2008
41 - async_pool did free a scalar value twice 188 - async_pool did free a scalar value twice
42 ("Attempt to unreference..."). 189 ("Attempt to unreference...").
43 190

Diff Legend

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