1 | Revision history for Perl extension Coro. |
1 | Revision history for Perl extension Coro. |
2 | |
2 | |
3 | TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. |
3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
4 | TODO: maybe implement a default message channel, very much like Erlang's |
4 | TODO: 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 | TODO: unready_all |
|
|
8 | |
|
|
9 | TODO: myhttpd header parsing |
|
|
10 | - do not keep a reference to the argument itself in |
|
|
11 | Coro::Senmaphore::guard, as it could change later. |
|
|
12 | |
|
|
13 | 5.131 Mon Mar 16 23:20:37 CET 2009 |
|
|
14 | - implement and document Coro->suspend, Coro->resume. |
|
|
15 | - fix Coro::Select implementation to not (often) close |
|
|
16 | the passed file descriptors (testcase provided by pippijn). |
|
|
17 | |
|
|
18 | 5.13 Mon Dec 15 21:51:42 CET 2008 |
|
|
19 | - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave). |
|
|
20 | - don't set diehook to C<undef> but instead to NULL, to avoid |
|
|
21 | spurious warnings. |
|
|
22 | - fix a lot of bugs in Coro::SemaphoreSet. |
|
|
23 | - Coro::SemaphoreSet will less often create a semaphore needlessly. |
|
|
24 | - add Coro::SemaphoreSet::count and wait methods. |
|
|
25 | - take advantage of the new Guard module. |
|
|
26 | - deprecate Coro::guard. |
|
|
27 | - try to fix the dreaded 01_unblock tests once more. I hate it when |
|
|
28 | testsuites need more fixing than the code they are supposed to test. |
|
|
29 | - croak in more cases when a required callback isn't resolvable. |
|
|
30 | - fix some minor issues in Coro::State->call/eval. |
|
|
31 | - use current coroutine context instead of a temporary one |
|
|
32 | when temporarily switching to another coroutine. |
|
|
33 | - do not call C-level on_destroy handlers during global destruction, |
|
|
34 | to avoid needless segfaults. |
|
|
35 | |
|
|
36 | 5.12 Sun Dec 7 13:30:38 CET 2008 |
|
|
37 | - add default config for MirOS, which seems to be a bug-to-bug |
|
|
38 | compatible fork of openbsd ("world domination by releasing |
|
|
39 | openbsd cvs before the openbsd folks do it" or so :). |
|
|
40 | - free_padlist did destroy the names pad, not good, but didn't |
|
|
41 | seem to bother perl - this could fix issues such as eval "" |
|
|
42 | inside a function called from multiple coroutines. |
|
|
43 | - use a different method to detect destruction time. |
|
|
44 | - be more careful when freeing padlists just before global |
|
|
45 | destruction. |
|
|
46 | - fixed and expanded the call/cc example. |
|
|
47 | - renamed _terminate to _coro_run. |
|
|
48 | - new method Coro::Channel->shutdown. |
|
|
49 | - try pthreads on openbsd <4.4 (broken sigaltstack, will |
|
|
50 | pthreads fare better?). |
|
|
51 | - be less picky about destroying "a" running coroutine. |
|
|
52 | |
|
|
53 | 5.11 Tue Nov 25 21:49:05 CET 2008 |
|
|
54 | - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly |
|
|
55 | working code with failed assertions, introducing more bugs than |
|
|
56 | it fixes, requiring elaborate workarounds :( |
|
|
57 | |
|
|
58 | 5.1 Mon Nov 24 08:54:59 CET 2008 |
|
|
59 | - wrote a small introductory tutorial - Coro::Intro. |
|
|
60 | - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. |
|
|
61 | - Coro::Select did errornously dup the file descriptors |
|
|
62 | and didn't work with all AnyEvent backends. |
|
|
63 | - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking |
|
|
64 | LWP data transfers. |
|
|
65 | - disassociate c contexts from coro objects - this is agruably more |
|
|
66 | correct, but mostly allows sharing of cctxs between coro and state |
|
|
67 | objects, for added memory savings and speed increases. |
|
|
68 | - bumped $Coro::POOL_RSS up to 32kb by default. |
|
|
69 | - no longer set the optype to OP_CUSTOM, as B::* understandably |
|
|
70 | doesn't like this very much (and we *are* a type of entersub). |
|
|
71 | - implement state cloning, just to prove that call/cc can be done. |
|
|
72 | - automatically load Coro::AnyEvent in Coro::Handle. |
|
|
73 | - wrap ->cancel calls in eval inside Coro::Handle as EV watchers |
|
|
74 | do not have this method (and don't need it either). |
|
|
75 | - speed up generic anyevent methods in Coro::Handle by using rouse |
|
|
76 | callbacks. |
|
|
77 | - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and |
|
|
78 | others. It also makes the debugger happier, as you can trace |
|
|
79 | through the idle threads now. |
|
|
80 | - add comppad_name* and hints ($^H, %^H) to per-thread variables. |
|
|
81 | - eg/event was pretty broken. |
|
|
82 | - better 5.8.6 compatibility. |
|
|
83 | |
|
|
84 | 5.0 Thu Nov 20 10:35:05 CET 2008 |
|
|
85 | - NEW ARCHITECTURE: use the latest 4.x version if you experience |
|
|
86 | stability issues. |
|
|
87 | - bump API version to 7 - all dependents must be recompiled. |
|
|
88 | - removed timed_* functions - they were not being used anyways |
|
|
89 | and should be replaced by a more generic mechanism - |
|
|
90 | and were annoying to support anyways :) |
|
|
91 | - removed SemaphoreSet's waiter method - use sem method instead. |
|
|
92 | - Coro::Semaphore->adjust didn't correctly wake up enough waiters. |
|
|
93 | - async_pool did free a scalar value twice |
|
|
94 | ("Attempt to unreference..."). |
|
|
95 | - fix a longstanding bug where calling terminate on a coro that |
|
|
96 | was waiting for a semaphore that was just becoming available |
|
|
97 | would cause a deadlock (semaphore would get into a state where |
|
|
98 | it was available but waiters were still blocked). |
|
|
99 | - calling throw on a coroutine that is waiting for a semaphore will |
|
|
100 | no longer make it acquire the semaphore (and thus leak a count). |
|
|
101 | - perl's process emulation is now not even theoretically supported |
|
|
102 | anymore. |
|
|
103 | - new functions Coro::rouse_cb and Coro::rouse_wait for easier |
|
|
104 | conversion of callback-style to blocking-style. |
|
|
105 | - new methods $coro->schedule_to and ->cede_to, to specifically |
|
|
106 | schedule or cede to a specific coroutine. |
|
|
107 | - new function Coro::Semaphore::wait. |
|
|
108 | - use named constants in Coro::Channel (Richard Hundt). |
|
|
109 | - directly patch the entersub opcode calling SLF functions (cede, |
|
|
110 | transfer and so on). this does speed up context switching, but |
|
|
111 | more importanly, it frees us from the hardcoded behaviour of |
|
|
112 | entersub, so we might actually be able to return something from |
|
|
113 | those functions and atcually create new ones. |
|
|
114 | - take advantage of __builtin_frame_address on gcc. |
|
|
115 | - expose THX in coroapi (not sure whether this was a wise decision, |
|
|
116 | as "threaded" perls are running at half speed anyways). |
|
|
117 | - implement execute_slf (schedule-like-function) interface that makes |
|
|
118 | it possible to implement schedule-like-functions in XS. |
|
|
119 | - use new SLF interface to massively speed up Coro::EV by roughly a |
|
|
120 | factor of two. |
|
|
121 | - used new SLF interface to massively speed up Coro::Semaphore by a |
|
|
122 | factor of three. |
|
|
123 | - used new SLF interface to speed up Coro::AIO by roughly a factor of |
|
|
124 | four and reduce its memory usage considerably. |
|
|
125 | - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore, |
|
|
126 | for a nice speedup and vastly more correct behaviour. Also implement |
|
|
127 | a new method "sem" to get at the underlying semaphore object. |
|
|
128 | - implement Coro::Channel in terms of Coro::Semaphore, for a moderate |
|
|
129 | (in comparison) 20-40% speedup. |
|
|
130 | - used new SLF interface to reimplement Coro::Signal gaining |
|
|
131 | some unknown (because I was too lazy), but certain, speedup, and also |
|
|
132 | making signals reliable for the first time. |
|
|
133 | - used new SLF interface and other optimisations to speed up async_pool |
|
|
134 | by a factor of two. It also doesn't rely on perl's exception mechanism |
|
|
135 | to exit anymore. The overhead for terminating an async_pool, coro over |
|
|
136 | a normal async is now very small. |
|
|
137 | - sped up coroutine creation/destruction by 40%. |
|
|
138 | - forgot to include Coro/libcoro/README in the dist for all these years. |
|
|
139 | - work around a freebsd pthreads bug (manual testcancel is required as |
|
|
140 | pthread_cond_wait isn't a cancellation point on freebsd). |
|
|
141 | - use new rouse functions to speed up and simplify Coro::BDB. |
|
|
142 | - make "prefer perl native functions" work with threaded perls. |
|
|
143 | - condense Coro::Debug ps output, hint at v and w flags. |
|
|
144 | - (libcoro) lots of minor cleanups and portability improvements. |
|
|
145 | |
|
|
146 | 4.914 Wed Nov 19 12:54:18 CET 2008 |
|
|
147 | - fix a disastrous bug in the readline optimisation |
|
|
148 | introduced in 4.801. |
|
|
149 | |
|
|
150 | 4.913 Sat Nov 15 07:58:28 CET 2008 |
|
|
151 | - async_pool did free a scalar value twice |
|
|
152 | ("Attempt to unreference..."). |
|
|
153 | |
|
|
154 | 4.912 Thu Nov 13 18:31:23 CET 2008 |
|
|
155 | - minor cleanups. |
|
|
156 | - use much larger stacks on linux and perl < 5.8.8. |
|
|
157 | - Coro::Debug::new_unix_server did not unlink the socket |
|
|
158 | when destroyed. |
|
|
159 | |
|
|
160 | 4.911 Tue Nov 11 04:26:17 CET 2008 |
|
|
161 | - "port" to threaded perls. |
|
|
162 | |
|
|
163 | 4.91 Mon Nov 10 05:36:38 CET 2008 |
|
|
164 | - the ->throw exception object no longer leaks. |
|
|
165 | - creating a new cctx leaked a scopestack entry (memleak). |
|
|
166 | - new coroutines didn't get created with a zero flags field |
|
|
167 | (unknown impact). |
|
|
168 | - calling ->throw on a not-yet-started coroutine should now work |
|
|
169 | instead of being ignored. |
|
|
170 | - ->throw is now supported on Coro::State objects. |
|
|
171 | - clean up cctx creation code a bit. |
|
|
172 | - entersub is actually a UNOP, not a LOGOP (not a bugfix). |
|
|
173 | |
|
|
174 | 4.9 Sat Nov 8 17:45:27 CET 2008 |
|
|
175 | - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
|
|
176 | - (libcoro) no longer leak threads in the experimental pthread backend, |
|
|
177 | also speed it up considerably. |
|
|
178 | - (libcoro) do not rely on makecontext passing void *'s unscathed. |
|
|
179 | - fix compiletime dependencies on libcoro in the Makefile. |
|
|
180 | - cctx_count wasn't always updated properly. |
|
|
181 | - Coro::State::cctx_stacksize wasn't applied correctly. |
|
|
182 | - new function Coro::State::cctx_max_idle. |
|
|
183 | - the default max number of idle C contexts is now 4. |
|
|
184 | - (libcoro) try harder to get _setjmp/_longjmp. |
|
|
185 | - (libcoro) cleanup and extend the libcoro API to officially |
|
|
186 | allow the creation of empty source contexts. |
|
|
187 | - very experimental workaround for the totally broken netbsd platform. |
|
|
188 | - tried to hack around openbsd bugs. |
|
|
189 | |
|
|
190 | 4.804 Wed Nov 5 16:36:00 CET 2008 |
|
|
191 | - Coro::Debug::new_unix_server would not create a non-blocking listening |
|
|
192 | socket, sometimes causing freezes. |
|
|
193 | - (libcoro) fix misaligned stack points for setjmp and assembly |
|
|
194 | methods, which can cause crashes on x86/x86_64 with a sufficiently |
|
|
195 | aggressive compiler. |
|
|
196 | - new function: Coro::Debug::new_tcp_server. |
|
|
197 | - move ->throw into the Coro class because it only works on coro objects. |
|
|
198 | |
|
|
199 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
|
|
200 | - (libcoro) use a global asm statement to become independent of gcc |
|
|
201 | otpimisations for CORO_ASM (thanks to pippijn for the idea). |
|
|
202 | - try to workaround yet another broken bsd, this time dragonfly. |
|
|
203 | |
|
|
204 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
|
|
205 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
|
|
206 | - tune 01_unblock.t tests a bit. |
|
|
207 | |
|
|
208 | 4.801 Wed Oct 22 18:33:37 CEST 2008 |
|
|
209 | - improve readline speed for very long "lines". |
|
|
210 | - backport to 5.8.8. |
|
|
211 | |
|
|
212 | 4.8 Thu Oct 2 13:34:40 CEST 2008 |
|
|
213 | - new function Coro::AIO::aio_wait. |
|
|
214 | - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". |
|
|
215 | - greatly speed up and reduce memory usage of Coro::AIO requests. |
|
|
216 | - implement some other ยต-optimisations. |
|
|
217 | |
|
|
218 | 4.749 Mon Sep 29 14:40:12 CEST 2008 |
|
|
219 | - port to slow and broken pseudo-threaded perls. (courtesy pippijn). |
|
|
220 | |
|
|
221 | 4.748 Sat Sep 27 14:03:19 CEST 2008 |
|
|
222 | - implement, but do not document, PerlIO::cede(granularity). |
|
|
223 | - Coro::Storable forgot to wrap Storable::pstore. |
|
|
224 | - work around the multitude of leaks and memory corruption |
|
|
225 | bugs in PerlIO::via by using our own C-level perliol. As a side |
|
|
226 | effect, Coro::Storable is now much, much, much faster. |
|
|
227 | |
|
|
228 | 4.747 Tue Sep 23 01:59:41 CEST 2008 |
|
|
229 | - fix a per-cv memleak (one empty array was leaked per |
|
|
230 | code reference). |
|
|
231 | - avoid a crash in coro->call|rss when the coroutine was already |
|
|
232 | destroyed (most noticably when using Coro::Debug::ps :) |
|
|
233 | - also protect *Storable::FILE. |
|
|
234 | - push up default storable granularity to 20ms. |
|
|
235 | |
|
|
236 | 4.746 Sun Sep 21 03:22:20 CEST 2008 |
|
|
237 | - be more insistent on locking Storable against reentrancy |
|
|
238 | in Coro::Storable. |
|
|
239 | - move swap_def?v and throw to Coro::State, as documented. |
|
|
240 | |
|
|
241 | 4.745 Thu Jul 24 00:14:38 CEST 2008 |
|
|
242 | - remove debugging code related to MgPV_nolen_const, also try to |
|
|
243 | make it compile with perl 5.8.6 (yes, apple apparently loves |
|
|
244 | outdated software). Reported by John S. |
|
|
245 | |
|
|
246 | 4.744 Tue Jul 8 22:06:35 CEST 2008 |
|
|
247 | - correctly provide default DIE/WARN handlers as documented. |
|
|
248 | - also overwrite PL_vtbl_sigelem.svt_clear, even though current |
|
|
249 | implementations inside perl work fine for us. |
|
|
250 | |
|
|
251 | 4.743 Mon Jun 16 00:21:57 CEST 2008 |
|
|
252 | - when using Coro::EV without running EV::loop it could |
|
|
253 | result in busy-waiting for events, this has been fixed. |
|
|
254 | - reduce codesize and improve performance by using EV_DEFAULT_UC. |
|
|
255 | |
|
|
256 | 4.742 Sat May 31 14:10:21 CEST 2008 |
|
|
257 | - implement a workaround for (some) perl <5.8.8 versions. |
|
|
258 | - require EV 3.3+. |
|
|
259 | |
|
|
260 | 4.741 Fri May 30 23:33:09 CEST 2008 |
|
|
261 | - tell netbsd how utterly broken their imitation of an OS is |
|
|
262 | and refuse to build by default if pthreads are in use. |
|
|
263 | - switch to "s" method on all bsd's by default, as their ucontext |
|
|
264 | stuff seems just too broken. |
|
|
265 | - fix a bug in Coro::Select. |
|
|
266 | |
|
|
267 | 4.74 Thu May 29 20:05:31 CEST 2008 |
|
|
268 | - do not test Coro::LWP for lack of dependencies. |
|
|
269 | - do not test Coro::Select for lack of working perls. |
|
|
270 | |
|
|
271 | 4.73 Thu May 29 2008 |
|
|
272 | - fix a bug in Coro::EV which would cause it to block despite |
|
|
273 | there being runnable coroutines. |
|
|
274 | - sprinkle "no warnings" freely over everything, also suppress |
|
|
275 | warnings for some other modules. |
|
|
276 | - fix typo in WSAEWOULDBLOCK. |
|
|
277 | |
|
|
278 | 4.72 Sun May 25 05:14:36 CEST 2008 |
|
|
279 | - tweak META.yaml a bit, unfortunately, there is no documented way |
|
|
280 | to have optional dependencies with CPAN. doh :( |
|
|
281 | - avoid running some tests on windows because they would fail due to |
|
|
282 | perl bug (broken fork, broken pipes...). |
|
|
283 | - work around perl on windows bugs where perl returns undocumented |
|
|
284 | error codes for sysread, syswrite etc. by taking advantage |
|
|
285 | of AnyEvent's workaround for that problem. |
|
|
286 | - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around |
|
|
287 | a common perl implementation bug on windows. |
|
|
288 | - use unix domain sockets in testsuite to work around |
|
|
289 | common perl implementation bugs on widows (they are emulated by |
|
|
290 | tcp sockets on windows. ugh.) |
|
|
291 | |
|
|
292 | 4.71 Sat May 24 20:01:27 CEST 2008 |
|
|
293 | - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). |
|
|
294 | - take advantage of async name resolution of AnyEvent::Util. |
|
|
295 | - work around brutal inet_aton override in Coro::LWP. |
|
|
296 | - take advantage of the readyhook in Coro::EV, for smoother |
|
|
297 | scheduling. |
|
|
298 | |
|
|
299 | 4.7 Sun May 11 00:32:19 CEST 2008 |
|
|
300 | - completely reworked the Coro manpage. |
|
|
301 | - added Coro::AnyEvent, generic event loop integration. |
|
|
302 | - implement cancel, ready and kill commands in Coro::Debug. |
|
|
303 | - document find_coro in Coro::Debug. |
|
|
304 | - incompatible change: rename has_stack to has_cctx. |
|
|
305 | - Coro::AIO and Coro::BDB no longer force event model detection, |
|
|
306 | use AnyEvent::AIO and AnyEvent::BDB. |
|
|
307 | |
|
|
308 | 4.6 Sat Apr 26 10:05:14 CEST 2008 |
|
|
309 | - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported |
|
|
310 | by the Coro module. It was a mistake to have it in the first place. |
|
|
311 | - (experimental) support for activestate perl 5.10 (method "w"). |
|
|
312 | - (experimental) support for strawberry perl (method "a"). |
|
|
313 | - coro_sigelem_set did not return a value although it had to, |
|
|
314 | actual impact unknown. |
|
|
315 | |
|
|
316 | 4.51 Mon Apr 14 13:28:27 CEST 2008 |
|
|
317 | - make it compile again on 5.8. |
|
|
318 | |
|
|
319 | 4.50 Thu Apr 10 09:43:17 CEST 2008 |
|
|
320 | - I did it twice! (see 4.49). |
|
|
321 | |
|
|
322 | 4.49 Sun Apr 6 21:23:31 CEST 2008 |
|
|
323 | - grr, instead of compiling the recent changes on 5.10 only they |
|
|
324 | were compiled on 5.8 only. |
|
|
325 | |
|
|
326 | 4.48 Sun Apr 6 20:36:46 CEST 2008 |
|
|
327 | - allow coroutine switches during eval's under 5.10.x, as it seems |
|
|
328 | the parser is a per-interpreter option now, so this is safe |
|
|
329 | (this might fix the odd crashing bug). |
|
|
330 | - drop support for 5.9.x versions: they are dead, jim. |
7 | |
331 | |
8 | 4.47 Sun Apr 6 00:37:52 CEST 2008 |
332 | 4.47 Sun Apr 6 00:37:52 CEST 2008 |
9 | - force cctx allocation on API calls: we know we need to force one |
333 | - force cctx allocation on API calls: we know we need to force one |
10 | and gcc actually manages to confuse our heuristic nowadays, |
334 | and gcc actually manages to confuse our heuristic nowadays, |
11 | leading to crashes and worse. |
335 | leading to crashes and worse. |