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

Comparing Coro/Changes (file contents):
Revision 1.362 by root, Tue Jul 8 20:08:40 2008 UTC vs.
Revision 1.474 by root, Sat Dec 6 00:55:59 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::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)).
7
8new -> ready
9new -> canceled
10ready -> running
11ready -> canceled
12running -> ready
13running -> blocked
14running -> canceled
15blocked -> ready
16blocked -> canceled
17
18
195.12
20 - add default config for MirOS, which seems to be a bug-to-bug
21 compatible fork of openbsd ("world domination by releasing
22 openbsd cvs before the openbsd folks do it" or so :).
23 - free_padlist did destroy the names pad, not good, but didn't
24 seem to bother perl - this could fix issues such as eval ""
25 inside a function called form multiple coroutines.
26 - use a different method to detect destruction time.
27 - be more careful when freeing padlists just before global
28 destruction.
29 - fixed and expanded the call/cc example.
30 - renamed _terminate to _coro_run.
31 - new method Coro::Channel->shutdown.
32 - try pthreads on openbsd <4.4 (broken sigaltstack, will
33 pthreads fare better?).
34 - be less picky about destroying "a" running coroutine.
35
365.11 Tue Nov 25 21:49:05 CET 2008
37 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
38 working code with failed assertions, introducing more bugs than
39 it fixes, requiring elaborate workarounds :(
40
415.1 Mon Nov 24 08:54:59 CET 2008
42 - wrote a small introductory tutorial - Coro::Intro.
43 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
44 - Coro::Select did errornously dup the file descriptors
45 and didn't work with all AnyEvent backends.
46 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
47 LWP data transfers.
48 - disassociate c contexts from coro objects - this is agruably more
49 correct, but mostly allows sharing of cctxs between coro and state
50 objects, for added memory savings and speed increases.
51 - bumped $Coro::POOL_RSS up to 32kb by default.
52 - no longer set the optype to OP_CUSTOM, as B::* understandably
53 doesn't like this very much (and we *are* a type of entersub).
54 - implement state cloning, just to prove that call/cc can be done.
55 - automatically load Coro::AnyEvent in Coro::Handle.
56 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
57 do not have this method (and don't need it either).
58 - speed up generic anyevent methods in Coro::Handle by using rouse
59 callbacks.
60 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
61 others. It also makes the debugger happier, as you can trace
62 through the idle threads now.
63 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
64 - eg/event was pretty broken.
65 - better 5.8.6 compatibility.
66
675.0 Thu Nov 20 10:35:05 CET 2008
68 - NEW ARCHITECTURE: use the latest 4.x version if you experience
69 stability issues.
70 - bump API version to 7 - all dependents must be recompiled.
71 - removed timed_* functions - they were not being used anyways
72 and should be replaced by a more generic mechanism -
73 and were annoying to support anyways :)
74 - removed SemaphoreSet's waiter method - use sem method instead.
75 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
76 - async_pool did free a scalar value twice
77 ("Attempt to unreference...").
78 - fix a longstanding bug where calling terminate on a coro that
79 was waiting for a semaphore that was just becoming available
80 would cause a deadlock (semaphore would get into a state where
81 it was available but waiters were still blocked).
82 - calling throw on a coroutine that is waiting for a semaphore will
83 no longer make it acquire the semaphore (and thus leak a count).
84 - perl's process emulation is now not even theoretically supported
85 anymore.
86 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
87 conversion of callback-style to blocking-style.
88 - new methods $coro->schedule_to and ->cede_to, to specifically
89 schedule or cede to a specific coroutine.
90 - new function Coro::Semaphore::wait.
91 - use named constants in Coro::Channel (Richard Hundt).
92 - directly patch the entersub opcode calling SLF functions (cede,
93 transfer and so on). this does speed up context switching, but
94 more importanly, it frees us from the hardcoded behaviour of
95 entersub, so we might actually be able to return something from
96 those functions and atcually create new ones.
97 - take advantage of __builtin_frame_address on gcc.
98 - expose THX in coroapi (not sure whether this was a wise decision,
99 as "threaded" perls are running at half speed anyways).
100 - implement execute_slf (schedule-like-function) interface that makes
101 it possible to implement schedule-like-functions in XS.
102 - use new SLF interface to massively speed up Coro::EV by roughly a
103 factor of two.
104 - used new SLF interface to massively speed up Coro::Semaphore by a
105 factor of three.
106 - used new SLF interface to speed up Coro::AIO by roughly a factor of
107 four and reduce its memory usage considerably.
108 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
109 for a nice speedup and vastly more correct behaviour. Also implement
110 a new method "sem" to get at the underlying semaphore object.
111 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
112 (in comparison) 20-40% speedup.
113 - used new SLF interface to reimplement Coro::Signal gaining
114 some unknown (because I was too lazy), but certain, speedup, and also
115 making signals reliable for the first time.
116 - used new SLF interface and other optimisations to speed up async_pool
117 by a factor of two. It also doesn't rely on perl's exception mechanism
118 to exit anymore. The overhead for terminating an async_pool, coro over
119 a normal async is now very small.
120 - sped up coroutine creation/destruction by 40%.
121 - forgot to include Coro/libcoro/README in the dist for all these years.
122 - work around a freebsd pthreads bug (manual testcancel is required as
123 pthread_cond_wait isn't a cancellation point on freebsd).
124 - use new rouse functions to speed up and simplify Coro::BDB.
125 - make "prefer perl native functions" work with threaded perls.
126 - condense Coro::Debug ps output, hint at v and w flags.
127 - (libcoro) lots of minor cleanups and portability improvements.
128
1294.914 Wed Nov 19 12:54:18 CET 2008
130 - fix a disastrous bug in the readline optimisation
131 introduced in 4.801.
132
1334.913 Sat Nov 15 07:58:28 CET 2008
134 - async_pool did free a scalar value twice
135 ("Attempt to unreference...").
136
1374.912 Thu Nov 13 18:31:23 CET 2008
138 - minor cleanups.
139 - use much larger stacks on linux and perl < 5.8.8.
140 - Coro::Debug::new_unix_server did not unlink the socket
141 when destroyed.
142
1434.911 Tue Nov 11 04:26:17 CET 2008
144 - "port" to threaded perls.
145
1464.91 Mon Nov 10 05:36:38 CET 2008
147 - the ->throw exception object no longer leaks.
148 - creating a new cctx leaked a scopestack entry (memleak).
149 - new coroutines didn't get created with a zero flags field
150 (unknown impact).
151 - calling ->throw on a not-yet-started coroutine should now work
152 instead of being ignored.
153 - ->throw is now supported on Coro::State objects.
154 - clean up cctx creation code a bit.
155 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
156
1574.9 Sat Nov 8 17:45:27 CET 2008
158 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
159 - (libcoro) no longer leak threads in the experimental pthread backend,
160 also speed it up considerably.
161 - (libcoro) do not rely on makecontext passing void *'s unscathed.
162 - fix compiletime dependencies on libcoro in the Makefile.
163 - cctx_count wasn't always updated properly.
164 - Coro::State::cctx_stacksize wasn't applied correctly.
165 - new function Coro::State::cctx_max_idle.
166 - the default max number of idle C contexts is now 4.
167 - (libcoro) try harder to get _setjmp/_longjmp.
168 - (libcoro) cleanup and extend the libcoro API to officially
169 allow the creation of empty source contexts.
170 - very experimental workaround for the totally broken netbsd platform.
171 - tried to hack around openbsd bugs.
172
1734.804 Wed Nov 5 16:36:00 CET 2008
174 - Coro::Debug::new_unix_server would not create a non-blocking listening
175 socket, sometimes causing freezes.
176 - (libcoro) fix misaligned stack points for setjmp and assembly
177 methods, which can cause crashes on x86/x86_64 with a sufficiently
178 aggressive compiler.
179 - new function: Coro::Debug::new_tcp_server.
180 - move ->throw into the Coro class because it only works on coro objects.
181
1824.803 Mon Nov 3 17:16:12 CET 2008
183 - (libcoro) use a global asm statement to become independent of gcc
184 otpimisations for CORO_ASM (thanks to pippijn for the idea).
185 - try to workaround yet another broken bsd, this time dragonfly.
186
1874.802 Thu Oct 30 10:56:12 CET 2008
188 - support -fno-omit-frame-pointer on x86 with the assembly method.
189 - tune 01_unblock.t tests a bit.
190
1914.801 Wed Oct 22 18:33:37 CEST 2008
192 - improve readline speed for very long "lines".
193 - backport to 5.8.8.
194
1954.8 Thu Oct 2 13:34:40 CEST 2008
196 - new function Coro::AIO::aio_wait.
197 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
198 - greatly speed up and reduce memory usage of Coro::AIO requests.
199 - implement some other ยต-optimisations.
200
2014.749 Mon Sep 29 14:40:12 CEST 2008
202 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
203
2044.748 Sat Sep 27 14:03:19 CEST 2008
205 - implement, but do not document, PerlIO::cede(granularity).
206 - Coro::Storable forgot to wrap Storable::pstore.
207 - work around the multitude of leaks and memory corruption
208 bugs in PerlIO::via by using our own C-level perliol. As a side
209 effect, Coro::Storable is now much, much, much faster.
210
2114.747 Tue Sep 23 01:59:41 CEST 2008
212 - fix a per-cv memleak (one empty array was leaked per
213 code reference).
214 - avoid a crash in coro->call|rss when the coroutine was already
215 destroyed (most noticably when using Coro::Debug::ps :)
216 - also protect *Storable::FILE.
217 - push up default storable granularity to 20ms.
218
2194.746 Sun Sep 21 03:22:20 CEST 2008
220 - be more insistent on locking Storable against reentrancy
221 in Coro::Storable.
222 - move swap_def?v and throw to Coro::State, as documented.
223
2244.745 Thu Jul 24 00:14:38 CEST 2008
225 - remove debugging code related to MgPV_nolen_const, also try to
226 make it compile with perl 5.8.6 (yes, apple apparently loves
227 outdated software). Reported by John S.
7 228
84.744 Tue Jul 8 22:06:35 CEST 2008 2294.744 Tue Jul 8 22:06:35 CEST 2008
9 - correctly provide default DIE/WARN handlers as documented. 230 - correctly provide default DIE/WARN handlers as documented.
10 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 231 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
11 implementations inside perl work fine for us. 232 implementations inside perl work fine for us.

Diff Legend

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