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

Comparing Coro/Changes (file contents):
Revision 1.391 by root, Fri Nov 7 16:39:35 2008 UTC vs.
Revision 1.467 by root, Thu Nov 27 12:00:58 2008 UTC

… …
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)).
7 7
8 - add default config for MirOS, which seems to be a bug-to-bug
9 compatible fork of openbsd ("world domination by releasing
10 openbsd cvs before the openbsd folks do it" or so :).
11 - use a different method to detect destruction time.
12 - be more careful when freeing padlists just before global
13 destruction.
14
155.11 Tue Nov 25 21:49:05 CET 2008
16 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
17 working code with failed assertions, introducing more bugs than
18 it fixes, requiring elaborate workarounds :(
19
205.1 Mon Nov 24 08:54:59 CET 2008
21 - wrote a small introductory tutorial - Coro::Intro.
22 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
23 - Coro::Select did errornously dup the file descriptors
24 and didn't work with all AnyEvent backends.
25 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
26 LWP data transfers.
27 - disassociate c contexts from coro objects - this is agruably more
28 correct, but mostly allows sharing of cctxs between coro and state
29 objects, for added memory savings and speed increases.
30 - bumped $Coro::POOL_RSS up to 32kb by default.
31 - no longer set the optype to OP_CUSTOM, as B::* understandably
32 doesn't like this very much (and we *are* a type of entersub).
33 - implement state cloning, just to prove that call/cc can be done.
34 - automatically load Coro::AnyEvent in Coro::Handle.
35 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
36 do not have this method (and don't need it either).
37 - speed up generic anyevent methods in Coro::Handle by using rouse
38 callbacks.
39 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
40 others. It also makes the debugger happier, as you can trace
41 through the idle threads now.
42 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
43 - eg/event was pretty broken.
44 - better 5.8.6 compatibility.
45
465.0 Thu Nov 20 10:35:05 CET 2008
47 - NEW ARCHITECTURE: use the latest 4.x version if you experience
48 stability issues.
49 - bump API version to 7 - all dependents must be recompiled.
50 - removed timed_* functions - they were not being used anyways
51 and should be replaced by a more generic mechanism -
52 and were annoying to support anyways :)
53 - removed SemaphoreSet's waiter method - use sem method instead.
54 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
55 - async_pool did free a scalar value twice
56 ("Attempt to unreference...").
57 - fix a longstanding bug where calling terminate on a coro that
58 was waiting for a semaphore that was just becoming available
59 would cause a deadlock (semaphore would get into a state where
60 it was available but waiters were still blocked).
61 - calling throw on a coroutine that is waiting for a semaphore will
62 no longer make it acquire the semaphore (and thus leak a count).
63 - perl's process emulation is now not even theoretically supported
64 anymore.
65 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
66 conversion of callback-style to blocking-style.
67 - new methods $coro->schedule_to and ->cede_to, to specifically
68 schedule or cede to a specific coroutine.
69 - new function Coro::Semaphore::wait.
70 - use named constants in Coro::Channel (Richard Hundt).
71 - directly patch the entersub opcode calling SLF functions (cede,
72 transfer and so on). this does speed up context switching, but
73 more importanly, it frees us from the hardcoded behaviour of
74 entersub, so we might actually be able to return something from
75 those functions and atcually create new ones.
76 - take advantage of __builtin_frame_address on gcc.
77 - expose THX in coroapi (not sure whether this was a wise decision,
78 as "threaded" perls are running at half speed anyways).
79 - implement execute_slf (schedule-like-function) interface that makes
80 it possible to implement schedule-like-functions in XS.
81 - use new SLF interface to massively speed up Coro::EV by roughly a
82 factor of two.
83 - used new SLF interface to massively speed up Coro::Semaphore by a
84 factor of three.
85 - used new SLF interface to speed up Coro::AIO by roughly a factor of
86 four and reduce its memory usage considerably.
87 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
88 for a nice speedup and vastly more correct behaviour. Also implement
89 a new method "sem" to get at the underlying semaphore object.
90 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
91 (in comparison) 20-40% speedup.
92 - used new SLF interface to reimplement Coro::Signal gaining
93 some unknown (because I was too lazy), but certain, speedup, and also
94 making signals reliable for the first time.
95 - used new SLF interface and other optimisations to speed up async_pool
96 by a factor of two. It also doesn't rely on perl's exception mechanism
97 to exit anymore. The overhead for terminating an async_pool, coro over
98 a normal async is now very small.
99 - sped up coroutine creation/destruction by 40%.
100 - forgot to include Coro/libcoro/README in the dist for all these years.
101 - work around a freebsd pthreads bug (manual testcancel is required as
102 pthread_cond_wait isn't a cancellation point on freebsd).
103 - use new rouse functions to speed up and simplify Coro::BDB.
104 - make "prefer perl native functions" work with threaded perls.
105 - condense Coro::Debug ps output, hint at v and w flags.
106 - (libcoro) lots of minor cleanups and portability improvements.
107
1084.914 Wed Nov 19 12:54:18 CET 2008
109 - fix a disastrous bug in the readline optimisation
110 introduced in 4.801.
111
1124.913 Sat Nov 15 07:58:28 CET 2008
113 - async_pool did free a scalar value twice
114 ("Attempt to unreference...").
115
1164.912 Thu Nov 13 18:31:23 CET 2008
117 - minor cleanups.
118 - use much larger stacks on linux and perl < 5.8.8.
119 - Coro::Debug::new_unix_server did not unlink the socket
120 when destroyed.
121
1224.911 Tue Nov 11 04:26:17 CET 2008
123 - "port" to threaded perls.
124
1254.91 Mon Nov 10 05:36:38 CET 2008
126 - the ->throw exception object no longer leaks.
127 - creating a new cctx leaked a scopestack entry (memleak).
128 - new coroutines didn't get created with a zero flags field
129 (unknown impact).
130 - calling ->throw on a not-yet-started coroutine should now work
131 instead of being ignored.
132 - ->throw is now supported on Coro::State objects.
133 - clean up cctx creation code a bit.
134 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
135
1364.9 Sat Nov 8 17:45:27 CET 2008
8 - libcoro did not preserve rbp with CORO_ASM. 137 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
138 - (libcoro) no longer leak threads in the experimental pthread backend,
139 also speed it up considerably.
140 - (libcoro) do not rely on makecontext passing void *'s unscathed.
141 - fix compiletime dependencies on libcoro in the Makefile.
142 - cctx_count wasn't always updated properly.
143 - Coro::State::cctx_stacksize wasn't applied correctly.
144 - new function Coro::State::cctx_max_idle.
145 - the default max number of idle C contexts is now 4.
146 - (libcoro) try harder to get _setjmp/_longjmp.
147 - (libcoro) cleanup and extend the libcoro API to officially
148 allow the creation of empty source contexts.
149 - very experimental workaround for the totally broken netbsd platform.
150 - tried to hack around openbsd bugs.
9 151
104.804 Wed Nov 5 16:36:00 CET 2008 1524.804 Wed Nov 5 16:36:00 CET 2008
11 - Coro::Debug::new_unix_server would not create a non-blocking listening 153 - Coro::Debug::new_unix_server would not create a non-blocking listening
12 socket, sometimes causing freezes. 154 socket, sometimes causing freezes.
13 - fix misaligned stack points for setjmp and assembly methods, which 155 - (libcoro) fix misaligned stack points for setjmp and assembly
14 can cause crashes on x86/x86_64 with a sufficiently aggressive compiler. 156 methods, which can cause crashes on x86/x86_64 with a sufficiently
157 aggressive compiler.
15 - new function: Coro::Debug::new_tcp_server. 158 - new function: Coro::Debug::new_tcp_server.
16 - move ->throw into the Coro class because it only works on coro objects. 159 - move ->throw into the Coro class because it only works on coro objects.
17 160
184.803 Mon Nov 3 17:16:12 CET 2008 1614.803 Mon Nov 3 17:16:12 CET 2008
19 - use a global asm statement to become independent of gcc otpimisations 162 - (libcoro) use a global asm statement to become independent of gcc
20 for CORO_ASM (thanks to pippijn for the idea). 163 otpimisations for CORO_ASM (thanks to pippijn for the idea).
21 - try to workaround yet another broken bsd, this time dragonfly. 164 - try to workaround yet another broken bsd, this time dragonfly.
22 165
234.802 Thu Oct 30 10:56:12 CET 2008 1664.802 Thu Oct 30 10:56:12 CET 2008
24 - support -fno-omit-frame-pointer on x86 with the assembly method. 167 - support -fno-omit-frame-pointer on x86 with the assembly method.
25 - tune 01_unblock.t tests a bit. 168 - tune 01_unblock.t tests a bit.

Diff Legend

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