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

Comparing Coro/Changes (file contents):
Revision 1.401 by root, Mon Nov 10 20:38:33 2008 UTC vs.
Revision 1.496 by root, Sat Jun 20 08:58:00 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
11TODO: make Coro::EV/event optional in Coro::AnyEvent
12 - clarify Coro::EV/Event/AnyEvent manpages: you have to actually run
13 the event loop somewhere to be able to poll for events
14 in the presence of ready threads.
15 - free per-thread global scalars in the thread calling ->cancel, to
16 avoid crashes when $_, $@ etc., are magical but some of those
17 had already been freed.
18 - speed up ready queue management/context switching by using a linked
19 list instead of an array (~5-10%).
20 - implement "watch" command in Coro::Debug shells.
21 - for fun, implement time-slicing as an example in the manpage.
22
235.132 Fri May 29 09:00:39 CEST 2009
24 - do not keep a reference to the argument itself in
25 Coro::Semaphore::guard, as it could change later.
26 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
27 should support a prepare callback.
28
295.131 Mon Mar 16 23:20:37 CET 2009
30 - implement and document Coro->suspend, Coro->resume.
31 - fix Coro::Select implementation to not (often) close
32 the passed file descriptors (testcase provided by pippijn).
33
345.13 Mon Dec 15 21:51:42 CET 2008
35 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
36 - don't set diehook to C<undef> but instead to NULL, to avoid
37 spurious warnings.
38 - fix a lot of bugs in Coro::SemaphoreSet.
39 - Coro::SemaphoreSet will less often create a semaphore needlessly.
40 - add Coro::SemaphoreSet::count and wait methods.
41 - take advantage of the new Guard module.
42 - deprecate Coro::guard.
43 - try to fix the dreaded 01_unblock tests once more. I hate it when
44 testsuites need more fixing than the code they are supposed to test.
45 - croak in more cases when a required callback isn't resolvable.
46 - fix some minor issues in Coro::State->call/eval.
47 - use current coroutine context instead of a temporary one
48 when temporarily switching to another coroutine.
49 - do not call C-level on_destroy handlers during global destruction,
50 to avoid needless segfaults.
51
525.12 Sun Dec 7 13:30:38 CET 2008
53 - add default config for MirOS, which seems to be a bug-to-bug
54 compatible fork of openbsd ("world domination by releasing
55 openbsd cvs before the openbsd folks do it" or so :).
56 - free_padlist did destroy the names pad, not good, but didn't
57 seem to bother perl - this could fix issues such as eval ""
58 inside a function called from multiple coroutines.
59 - use a different method to detect destruction time.
60 - be more careful when freeing padlists just before global
61 destruction.
62 - fixed and expanded the call/cc example.
63 - renamed _terminate to _coro_run.
64 - new method Coro::Channel->shutdown.
65 - try pthreads on openbsd <4.4 (broken sigaltstack, will
66 pthreads fare better?).
67 - be less picky about destroying "a" running coroutine.
68
695.11 Tue Nov 25 21:49:05 CET 2008
70 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
71 working code with failed assertions, introducing more bugs than
72 it fixes, requiring elaborate workarounds :(
73
745.1 Mon Nov 24 08:54:59 CET 2008
75 - wrote a small introductory tutorial - Coro::Intro.
76 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
77 - Coro::Select did errornously dup the file descriptors
78 and didn't work with all AnyEvent backends.
79 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
80 LWP data transfers.
81 - disassociate c contexts from coro objects - this is agruably more
82 correct, but mostly allows sharing of cctxs between coro and state
83 objects, for added memory savings and speed increases.
84 - bumped $Coro::POOL_RSS up to 32kb by default.
85 - no longer set the optype to OP_CUSTOM, as B::* understandably
86 doesn't like this very much (and we *are* a type of entersub).
87 - implement state cloning, just to prove that call/cc can be done.
88 - automatically load Coro::AnyEvent in Coro::Handle.
89 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
90 do not have this method (and don't need it either).
91 - speed up generic anyevent methods in Coro::Handle by using rouse
92 callbacks.
93 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
94 others. It also makes the debugger happier, as you can trace
95 through the idle threads now.
96 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
97 - eg/event was pretty broken.
98 - better 5.8.6 compatibility.
99
1005.0 Thu Nov 20 10:35:05 CET 2008
101 - NEW ARCHITECTURE: use the latest 4.x version if you experience
102 stability issues.
103 - bump API version to 7 - all dependents must be recompiled.
104 - removed timed_* functions - they were not being used anyways
105 and should be replaced by a more generic mechanism -
106 and were annoying to support anyways :)
107 - removed SemaphoreSet's waiter method - use sem method instead.
108 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
109 - async_pool did free a scalar value twice
110 ("Attempt to unreference...").
111 - fix a longstanding bug where calling terminate on a coro that
112 was waiting for a semaphore that was just becoming available
113 would cause a deadlock (semaphore would get into a state where
114 it was available but waiters were still blocked).
115 - calling throw on a coroutine that is waiting for a semaphore will
116 no longer make it acquire the semaphore (and thus leak a count).
117 - perl's process emulation is now not even theoretically supported
118 anymore.
119 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
120 conversion of callback-style to blocking-style.
121 - new methods $coro->schedule_to and ->cede_to, to specifically
122 schedule or cede to a specific coroutine.
123 - new function Coro::Semaphore::wait.
124 - use named constants in Coro::Channel (Richard Hundt).
125 - directly patch the entersub opcode calling SLF functions (cede,
126 transfer and so on). this does speed up context switching, but
127 more importanly, it frees us from the hardcoded behaviour of
128 entersub, so we might actually be able to return something from
129 those functions and atcually create new ones.
130 - take advantage of __builtin_frame_address on gcc.
131 - expose THX in coroapi (not sure whether this was a wise decision,
132 as "threaded" perls are running at half speed anyways).
133 - implement execute_slf (schedule-like-function) interface that makes
134 it possible to implement schedule-like-functions in XS.
135 - use new SLF interface to massively speed up Coro::EV by roughly a
136 factor of two.
137 - used new SLF interface to massively speed up Coro::Semaphore by a
138 factor of three.
139 - used new SLF interface to speed up Coro::AIO by roughly a factor of
140 four and reduce its memory usage considerably.
141 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
142 for a nice speedup and vastly more correct behaviour. Also implement
143 a new method "sem" to get at the underlying semaphore object.
144 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
145 (in comparison) 20-40% speedup.
146 - used new SLF interface to reimplement Coro::Signal gaining
147 some unknown (because I was too lazy), but certain, speedup, and also
148 making signals reliable for the first time.
149 - used new SLF interface and other optimisations to speed up async_pool
150 by a factor of two. It also doesn't rely on perl's exception mechanism
151 to exit anymore. The overhead for terminating an async_pool, coro over
152 a normal async is now very small.
153 - sped up coroutine creation/destruction by 40%.
154 - forgot to include Coro/libcoro/README in the dist for all these years.
155 - work around a freebsd pthreads bug (manual testcancel is required as
156 pthread_cond_wait isn't a cancellation point on freebsd).
157 - use new rouse functions to speed up and simplify Coro::BDB.
158 - make "prefer perl native functions" work with threaded perls.
159 - condense Coro::Debug ps output, hint at v and w flags.
160 - (libcoro) lots of minor cleanups and portability improvements.
161
1624.914 Wed Nov 19 12:54:18 CET 2008
163 - fix a disastrous bug in the readline optimisation
164 introduced in 4.801.
165
1664.913 Sat Nov 15 07:58:28 CET 2008
167 - async_pool did free a scalar value twice
168 ("Attempt to unreference...").
169
1704.912 Thu Nov 13 18:31:23 CET 2008
171 - minor cleanups.
172 - use much larger stacks on linux and perl < 5.8.8.
173 - Coro::Debug::new_unix_server did not unlink the socket
174 when destroyed.
175
1764.911 Tue Nov 11 04:26:17 CET 2008
8 - "port" to threaded perls. 177 - "port" to threaded perls.
9 178
104.91 Mon Nov 10 05:36:38 CET 2008 1794.91 Mon Nov 10 05:36:38 CET 2008
11 - the ->throw exception object no longer leaks. 180 - the ->throw exception object no longer leaks.
12 - creating a new cctx leaked a scopestack entry (memleak). 181 - creating a new cctx leaked a scopestack entry (memleak).

Diff Legend

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