1 | Revision history for Perl extension Coro. |
1 | Revision history for Perl extension Coro. |
|
|
2 | |
|
|
3 | TODO: should explore PerlIO::coroaio (perl leaks like hell). |
|
|
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!), |
|
|
6 | but very lacking in others (higher level ipc)). |
|
|
7 | |
|
|
8 | 5.11 |
|
|
9 | - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly |
|
|
10 | working code with failed assertions, introducing more bugs than |
|
|
11 | it fixes, requiring elaborate workarounds :( |
|
|
12 | |
|
|
13 | 5.1 Mon Nov 24 08:54:59 CET 2008 |
|
|
14 | - wrote a small introductory tutorial - Coro::Intro. |
|
|
15 | - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. |
|
|
16 | - Coro::Select did errornously dup the file descriptors |
|
|
17 | and didn't work with all AnyEvent backends. |
|
|
18 | - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking |
|
|
19 | LWP data transfers. |
|
|
20 | - disassociate c contexts from coro objects - this is agruably more |
|
|
21 | correct, but mostly allows sharing of cctxs between coro and state |
|
|
22 | objects, for added memory savings and speed increases. |
|
|
23 | - bumped $Coro::POOL_RSS up to 32kb by default. |
|
|
24 | - no longer set the optype to OP_CUSTOM, as B::* understandably |
|
|
25 | doesn't like this very much (and we *are* a type of entersub). |
|
|
26 | - implement state cloning, just to prove that call/cc can be done. |
|
|
27 | - automatically load Coro::AnyEvent in Coro::Handle. |
|
|
28 | - wrap ->cancel calls in eval inside Coro::Handle as EV watchers |
|
|
29 | do not have this method (and don't need it either). |
|
|
30 | - speed up generic anyevent methods in Coro::Handle by using rouse |
|
|
31 | callbacks. |
|
|
32 | - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and |
|
|
33 | others. It also makes the debugger happier, as you can trace |
|
|
34 | through the idle threads now. |
|
|
35 | - add comppad_name* and hints ($^H, %^H) to per-thread variables. |
|
|
36 | - eg/event was pretty broken. |
|
|
37 | - better 5.8.6 compatibility. |
|
|
38 | |
|
|
39 | 5.0 Thu Nov 20 10:35:05 CET 2008 |
|
|
40 | - NEW ARCHITECTURE: use the latest 4.x version if you experience |
|
|
41 | stability issues. |
|
|
42 | - bump API version to 7 - all dependents must be recompiled. |
|
|
43 | - removed timed_* functions - they were not being used anyways |
|
|
44 | and should be replaced by a more generic mechanism - |
|
|
45 | and were annoying to support anyways :) |
|
|
46 | - removed SemaphoreSet's waiter method - use sem method instead. |
|
|
47 | - Coro::Semaphore->adjust didn't correctly wake up enough waiters. |
|
|
48 | - async_pool did free a scalar value twice |
|
|
49 | ("Attempt to unreference..."). |
|
|
50 | - fix a longstanding bug where calling terminate on a coro that |
|
|
51 | was waiting for a semaphore that was just becoming available |
|
|
52 | would cause a deadlock (semaphore would get into a state where |
|
|
53 | it was available but waiters were still blocked). |
|
|
54 | - calling throw on a coroutine that is waiting for a semaphore will |
|
|
55 | no longer make it acquire the semaphore (and thus leak a count). |
|
|
56 | - perl's process emulation is now not even theoretically supported |
|
|
57 | anymore. |
|
|
58 | - new functions Coro::rouse_cb and Coro::rouse_wait for easier |
|
|
59 | conversion of callback-style to blocking-style. |
|
|
60 | - new methods $coro->schedule_to and ->cede_to, to specifically |
|
|
61 | schedule or cede to a specific coroutine. |
|
|
62 | - new function Coro::Semaphore::wait. |
|
|
63 | - use named constants in Coro::Channel (Richard Hundt). |
|
|
64 | - directly patch the entersub opcode calling SLF functions (cede, |
|
|
65 | transfer and so on). this does speed up context switching, but |
|
|
66 | more importanly, it frees us from the hardcoded behaviour of |
|
|
67 | entersub, so we might actually be able to return something from |
|
|
68 | those functions and atcually create new ones. |
|
|
69 | - take advantage of __builtin_frame_address on gcc. |
|
|
70 | - expose THX in coroapi (not sure whether this was a wise decision, |
|
|
71 | as "threaded" perls are running at half speed anyways). |
|
|
72 | - implement execute_slf (schedule-like-function) interface that makes |
|
|
73 | it possible to implement schedule-like-functions in XS. |
|
|
74 | - use new SLF interface to massively speed up Coro::EV by roughly a |
|
|
75 | factor of two. |
|
|
76 | - used new SLF interface to massively speed up Coro::Semaphore by a |
|
|
77 | factor of three. |
|
|
78 | - used new SLF interface to speed up Coro::AIO by roughly a factor of |
|
|
79 | four and reduce its memory usage considerably. |
|
|
80 | - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore, |
|
|
81 | for a nice speedup and vastly more correct behaviour. Also implement |
|
|
82 | a new method "sem" to get at the underlying semaphore object. |
|
|
83 | - implement Coro::Channel in terms of Coro::Semaphore, for a moderate |
|
|
84 | (in comparison) 20-40% speedup. |
|
|
85 | - used new SLF interface to reimplement Coro::Signal gaining |
|
|
86 | some unknown (because I was too lazy), but certain, speedup, and also |
|
|
87 | making signals reliable for the first time. |
|
|
88 | - used new SLF interface and other optimisations to speed up async_pool |
|
|
89 | by a factor of two. It also doesn't rely on perl's exception mechanism |
|
|
90 | to exit anymore. The overhead for terminating an async_pool, coro over |
|
|
91 | a normal async is now very small. |
|
|
92 | - sped up coroutine creation/destruction by 40%. |
|
|
93 | - forgot to include Coro/libcoro/README in the dist for all these years. |
|
|
94 | - work around a freebsd pthreads bug (manual testcancel is required as |
|
|
95 | pthread_cond_wait isn't a cancellation point on freebsd). |
|
|
96 | - use new rouse functions to speed up and simplify Coro::BDB. |
|
|
97 | - make "prefer perl native functions" work with threaded perls. |
|
|
98 | - condense Coro::Debug ps output, hint at v and w flags. |
|
|
99 | - (libcoro) lots of minor cleanups and portability improvements. |
|
|
100 | |
|
|
101 | 4.914 Wed Nov 19 12:54:18 CET 2008 |
|
|
102 | - fix a disastrous bug in the readline optimisation |
|
|
103 | introduced in 4.801. |
|
|
104 | |
|
|
105 | 4.913 Sat Nov 15 07:58:28 CET 2008 |
|
|
106 | - async_pool did free a scalar value twice |
|
|
107 | ("Attempt to unreference..."). |
|
|
108 | |
|
|
109 | 4.912 Thu Nov 13 18:31:23 CET 2008 |
|
|
110 | - minor cleanups. |
|
|
111 | - use much larger stacks on linux and perl < 5.8.8. |
|
|
112 | - Coro::Debug::new_unix_server did not unlink the socket |
|
|
113 | when destroyed. |
|
|
114 | |
|
|
115 | 4.911 Tue Nov 11 04:26:17 CET 2008 |
|
|
116 | - "port" to threaded perls. |
|
|
117 | |
|
|
118 | 4.91 Mon Nov 10 05:36:38 CET 2008 |
|
|
119 | - the ->throw exception object no longer leaks. |
|
|
120 | - creating a new cctx leaked a scopestack entry (memleak). |
|
|
121 | - new coroutines didn't get created with a zero flags field |
|
|
122 | (unknown impact). |
|
|
123 | - calling ->throw on a not-yet-started coroutine should now work |
|
|
124 | instead of being ignored. |
|
|
125 | - ->throw is now supported on Coro::State objects. |
|
|
126 | - clean up cctx creation code a bit. |
|
|
127 | - entersub is actually a UNOP, not a LOGOP (not a bugfix). |
|
|
128 | |
|
|
129 | 4.9 Sat Nov 8 17:45:27 CET 2008 |
|
|
130 | - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
|
|
131 | - (libcoro) no longer leak threads in the experimental pthread backend, |
|
|
132 | also speed it up considerably. |
|
|
133 | - (libcoro) do not rely on makecontext passing void *'s unscathed. |
|
|
134 | - fix compiletime dependencies on libcoro in the Makefile. |
|
|
135 | - cctx_count wasn't always updated properly. |
|
|
136 | - Coro::State::cctx_stacksize wasn't applied correctly. |
|
|
137 | - new function Coro::State::cctx_max_idle. |
|
|
138 | - the default max number of idle C contexts is now 4. |
|
|
139 | - (libcoro) try harder to get _setjmp/_longjmp. |
|
|
140 | - (libcoro) cleanup and extend the libcoro API to officially |
|
|
141 | allow the creation of empty source contexts. |
|
|
142 | - very experimental workaround for the totally broken netbsd platform. |
|
|
143 | - tried to hack around openbsd bugs. |
|
|
144 | |
|
|
145 | 4.804 Wed Nov 5 16:36:00 CET 2008 |
|
|
146 | - Coro::Debug::new_unix_server would not create a non-blocking listening |
|
|
147 | socket, sometimes causing freezes. |
|
|
148 | - (libcoro) fix misaligned stack points for setjmp and assembly |
|
|
149 | methods, which can cause crashes on x86/x86_64 with a sufficiently |
|
|
150 | aggressive compiler. |
|
|
151 | - new function: Coro::Debug::new_tcp_server. |
|
|
152 | - move ->throw into the Coro class because it only works on coro objects. |
|
|
153 | |
|
|
154 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
|
|
155 | - (libcoro) use a global asm statement to become independent of gcc |
|
|
156 | otpimisations for CORO_ASM (thanks to pippijn for the idea). |
|
|
157 | - try to workaround yet another broken bsd, this time dragonfly. |
|
|
158 | |
|
|
159 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
|
|
160 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
|
|
161 | - tune 01_unblock.t tests a bit. |
|
|
162 | |
|
|
163 | 4.801 Wed Oct 22 18:33:37 CEST 2008 |
|
|
164 | - improve readline speed for very long "lines". |
|
|
165 | - backport to 5.8.8. |
|
|
166 | |
|
|
167 | 4.8 Thu Oct 2 13:34:40 CEST 2008 |
|
|
168 | - new function Coro::AIO::aio_wait. |
|
|
169 | - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". |
|
|
170 | - greatly speed up and reduce memory usage of Coro::AIO requests. |
|
|
171 | - implement some other µ-optimisations. |
|
|
172 | |
|
|
173 | 4.749 Mon Sep 29 14:40:12 CEST 2008 |
|
|
174 | - port to slow and broken pseudo-threaded perls. (courtesy pippijn). |
|
|
175 | |
|
|
176 | 4.748 Sat Sep 27 14:03:19 CEST 2008 |
|
|
177 | - implement, but do not document, PerlIO::cede(granularity). |
|
|
178 | - Coro::Storable forgot to wrap Storable::pstore. |
|
|
179 | - work around the multitude of leaks and memory corruption |
|
|
180 | bugs in PerlIO::via by using our own C-level perliol. As a side |
|
|
181 | effect, Coro::Storable is now much, much, much faster. |
|
|
182 | |
|
|
183 | 4.747 Tue Sep 23 01:59:41 CEST 2008 |
|
|
184 | - fix a per-cv memleak (one empty array was leaked per |
|
|
185 | code reference). |
|
|
186 | - avoid a crash in coro->call|rss when the coroutine was already |
|
|
187 | destroyed (most noticably when using Coro::Debug::ps :) |
|
|
188 | - also protect *Storable::FILE. |
|
|
189 | - push up default storable granularity to 20ms. |
|
|
190 | |
|
|
191 | 4.746 Sun Sep 21 03:22:20 CEST 2008 |
|
|
192 | - be more insistent on locking Storable against reentrancy |
|
|
193 | in Coro::Storable. |
|
|
194 | - move swap_def?v and throw to Coro::State, as documented. |
|
|
195 | |
|
|
196 | 4.745 Thu Jul 24 00:14:38 CEST 2008 |
|
|
197 | - remove debugging code related to MgPV_nolen_const, also try to |
|
|
198 | make it compile with perl 5.8.6 (yes, apple apparently loves |
|
|
199 | outdated software). Reported by John S. |
|
|
200 | |
|
|
201 | 4.744 Tue Jul 8 22:06:35 CEST 2008 |
|
|
202 | - correctly provide default DIE/WARN handlers as documented. |
|
|
203 | - also overwrite PL_vtbl_sigelem.svt_clear, even though current |
|
|
204 | implementations inside perl work fine for us. |
|
|
205 | |
|
|
206 | 4.743 Mon Jun 16 00:21:57 CEST 2008 |
|
|
207 | - when using Coro::EV without running EV::loop it could |
|
|
208 | result in busy-waiting for events, this has been fixed. |
|
|
209 | - reduce codesize and improve performance by using EV_DEFAULT_UC. |
|
|
210 | |
|
|
211 | 4.742 Sat May 31 14:10:21 CEST 2008 |
|
|
212 | - implement a workaround for (some) perl <5.8.8 versions. |
|
|
213 | - require EV 3.3+. |
|
|
214 | |
|
|
215 | 4.741 Fri May 30 23:33:09 CEST 2008 |
|
|
216 | - tell netbsd how utterly broken their imitation of an OS is |
|
|
217 | and refuse to build by default if pthreads are in use. |
|
|
218 | - switch to "s" method on all bsd's by default, as their ucontext |
|
|
219 | stuff seems just too broken. |
|
|
220 | - fix a bug in Coro::Select. |
|
|
221 | |
|
|
222 | 4.74 Thu May 29 20:05:31 CEST 2008 |
|
|
223 | - do not test Coro::LWP for lack of dependencies. |
|
|
224 | - do not test Coro::Select for lack of working perls. |
|
|
225 | |
|
|
226 | 4.73 Thu May 29 2008 |
|
|
227 | - fix a bug in Coro::EV which would cause it to block despite |
|
|
228 | there being runnable coroutines. |
|
|
229 | - sprinkle "no warnings" freely over everything, also suppress |
|
|
230 | warnings for some other modules. |
|
|
231 | - fix typo in WSAEWOULDBLOCK. |
|
|
232 | |
|
|
233 | 4.72 Sun May 25 05:14:36 CEST 2008 |
|
|
234 | - tweak META.yaml a bit, unfortunately, there is no documented way |
|
|
235 | to have optional dependencies with CPAN. doh :( |
|
|
236 | - avoid running some tests on windows because they would fail due to |
|
|
237 | perl bug (broken fork, broken pipes...). |
|
|
238 | - work around perl on windows bugs where perl returns undocumented |
|
|
239 | error codes for sysread, syswrite etc. by taking advantage |
|
|
240 | of AnyEvent's workaround for that problem. |
|
|
241 | - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around |
|
|
242 | a common perl implementation bug on windows. |
|
|
243 | - use unix domain sockets in testsuite to work around |
|
|
244 | common perl implementation bugs on widows (they are emulated by |
|
|
245 | tcp sockets on windows. ugh.) |
|
|
246 | |
|
|
247 | 4.71 Sat May 24 20:01:27 CEST 2008 |
|
|
248 | - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). |
|
|
249 | - take advantage of async name resolution of AnyEvent::Util. |
|
|
250 | - work around brutal inet_aton override in Coro::LWP. |
|
|
251 | - take advantage of the readyhook in Coro::EV, for smoother |
|
|
252 | scheduling. |
|
|
253 | |
|
|
254 | 4.7 Sun May 11 00:32:19 CEST 2008 |
|
|
255 | - completely reworked the Coro manpage. |
|
|
256 | - added Coro::AnyEvent, generic event loop integration. |
|
|
257 | - implement cancel, ready and kill commands in Coro::Debug. |
|
|
258 | - document find_coro in Coro::Debug. |
|
|
259 | - incompatible change: rename has_stack to has_cctx. |
|
|
260 | - Coro::AIO and Coro::BDB no longer force event model detection, |
|
|
261 | use AnyEvent::AIO and AnyEvent::BDB. |
|
|
262 | |
|
|
263 | 4.6 Sat Apr 26 10:05:14 CEST 2008 |
|
|
264 | - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported |
|
|
265 | by the Coro module. It was a mistake to have it in the first place. |
|
|
266 | - (experimental) support for activestate perl 5.10 (method "w"). |
|
|
267 | - (experimental) support for strawberry perl (method "a"). |
|
|
268 | - coro_sigelem_set did not return a value although it had to, |
|
|
269 | actual impact unknown. |
|
|
270 | |
|
|
271 | 4.51 Mon Apr 14 13:28:27 CEST 2008 |
|
|
272 | - make it compile again on 5.8. |
|
|
273 | |
|
|
274 | 4.50 Thu Apr 10 09:43:17 CEST 2008 |
|
|
275 | - I did it twice! (see 4.49). |
|
|
276 | |
|
|
277 | 4.49 Sun Apr 6 21:23:31 CEST 2008 |
|
|
278 | - grr, instead of compiling the recent changes on 5.10 only they |
|
|
279 | were compiled on 5.8 only. |
|
|
280 | |
|
|
281 | 4.48 Sun Apr 6 20:36:46 CEST 2008 |
|
|
282 | - allow coroutine switches during eval's under 5.10.x, as it seems |
|
|
283 | the parser is a per-interpreter option now, so this is safe |
|
|
284 | (this might fix the odd crashing bug). |
|
|
285 | - drop support for 5.9.x versions: they are dead, jim. |
|
|
286 | |
|
|
287 | 4.47 Sun Apr 6 00:37:52 CEST 2008 |
|
|
288 | - force cctx allocation on API calls: we know we need to force one |
|
|
289 | and gcc actually manages to confuse our heuristic nowadays, |
|
|
290 | leading to crashes and worse. |
|
|
291 | - document force_cctx. |
|
|
292 | |
|
|
293 | 4.46 Fri Apr 4 22:05:43 CEST 2008 |
|
|
294 | - upgrade libcoro, resulting in pthread-backend (which was only created |
|
|
295 | to fulfill the rules of the programming languages shootout). |
|
|
296 | |
|
|
297 | 4.45 Thu Mar 13 11:55:36 CET 2008 |
|
|
298 | - fix a file leaking bug in eg/mhyttpd that would allow |
|
|
299 | downloading of any file (reported by oesi). |
|
|
300 | - fix deadlock bug in Coro::Channel (reported by Richard Hundt) |
|
|
301 | (also add testcase). |
|
|
302 | - support Broadcast option in Coro::Socket (patch by Richard Hundt, |
|
|
303 | apparently having loads of fun with that). |
|
|
304 | - implement $state->swap_defsv, ->swap_defav and document ->throw. |
|
|
305 | |
|
|
306 | 4.4 Wed Feb 13 16:44:29 CET 2008 |
|
|
307 | - only assume .cfi pseudo isns exist on GNU/Linux. |
|
|
308 | - add get_timed method to Coro::Channels. |
|
|
309 | - fixed Coro::Debug manpage. |
|
|
310 | - perl 5.11 compatbility improvement based on patch by |
|
|
311 | Andreas König. |
|
|
312 | |
|
|
313 | 4.37 Sun Jan 20 11:25:23 CET 2008 |
|
|
314 | - soften the check for an active parser for perl 5.10 (avoiding |
|
|
315 | "Coro::State::transfer called while parsing" in many cases). |
|
|
316 | |
|
|
317 | 4.36 Sun Jan 13 10:53:56 CET 2008 |
|
|
318 | - reset diehook when terminating from an async_pool as to not |
|
|
319 | trigger any __DIE__ handlers. |
|
|
320 | |
|
|
321 | 4.35 Sun Jan 13 04:14:13 CET 2008 |
|
|
322 | - "bt" debug command now displays any exceptions |
|
|
323 | from longmess and also skips the innermost |
|
|
324 | stackframes, giving more useufl output. |
|
|
325 | - allow backtraces in coroutines blocked in Coro::EV, |
|
|
326 | at a <1% speed hit. |
|
|
327 | - handle localising of $SIG{__WARN__} and __DIE__ |
|
|
328 | properly (with a proper amount of dirty hacking). |
|
|
329 | |
|
|
330 | 4.34 Sat Dec 22 17:49:53 CET 2007 |
|
|
331 | - upgrade to EV version 2.0 API. |
|
|
332 | |
|
|
333 | 4.33 Mon Dec 17 08:36:12 CET 2007 |
|
|
334 | - make Coro::AIO etc. loadable in the absence of EV. |
|
|
335 | |
|
|
336 | 4.32 Mon Dec 17 07:46:02 CET 2007 |
|
|
337 | - majorly improved Coro::Handle's performance with Coro::EV. |
|
|
338 | - implemented the until now mythical Coro::BDB module. |
|
|
339 | - specialcase EV in Coro::AIO and Coro::BDB for extra speed. |
|
|
340 | |
|
|
341 | 4.31 Wed Dec 5 12:32:39 CET 2007 |
|
|
342 | - remove warn statement form Coro::Util. |
|
|
343 | |
|
|
344 | 4.3 Tue Dec 4 20:33:14 CET 2007 |
|
|
345 | - calls to the idle function could cause stack corruption |
|
|
346 | when the stack changed. |
|
|
347 | - do no longer rely on the presence of EV::DNS if EV is used |
|
|
348 | (because it is gone), but instead take avdantage of EV::ADNS |
|
|
349 | if available. |
|
|
350 | - add ($) prototypes to all functions in Coro::Storable. |
|
|
351 | - use a conventional (safer) idle callback in Coro::EV. |
|
|
352 | - do accept quickly in Coro::Debug to avoid endless loops. |
|
|
353 | |
|
|
354 | 4.22 Fri Nov 30 16:04:04 CET 2007 |
|
|
355 | - really use optimised versions for Event and EV in Coro::Util |
|
|
356 | and Coro::Handle. |
|
|
357 | |
|
|
358 | 4.21 Sun Nov 25 10:48:59 CET 2007 |
|
|
359 | - fix a spurious memory read. |
|
|
360 | - Coro::EV no longer keeps the eventloop "alive". |
|
|
361 | |
|
|
362 | 4.2 Fri Nov 9 20:47:05 CET 2007 |
|
|
363 | - enable/disable tracing from a new coroutine, not a pooled one. |
|
|
364 | - fix a memleak in Coro::Event. |
|
|
365 | - removed killall call from fork_eval. |
|
|
366 | - made sure store_fd is already loaded so that fork_eval |
|
|
367 | does not have to parse autoload in each subprocess. |
|
|
368 | - only use assembly method if -O switch is in $Config{optimize}. |
|
|
369 | - add (optional) Coro::EV module, so far the best event loop module |
|
|
370 | directly supported by Coro. |
|
|
371 | - if the event model is EV, use EV::DNS to resolve |
|
|
372 | stuff in Coro::Util. |
|
|
373 | - don't get confused by multiple event notifications in Coro::Handle. |
|
|
374 | - initial support for EV (libevent interface). |
|
|
375 | - require Event and EV using configure_requires, to force their existance. |
|
|
376 | |
|
|
377 | 4.13 Wed Oct 24 07:26:45 CEST 2007 |
|
|
378 | - add Coro::Storable::blocking_thaw. |
|
|
379 | - use a vastly more complicated technique to localise |
|
|
380 | $SIG{__WARN/DIE__} that also works on perls <= 5.8.8. |
|
|
381 | - use a coroutine for the idle callback Coro::Event, |
|
|
382 | instead of running Event in the current coroutine context. |
|
|
383 | This also catches recursive invocations. |
|
|
384 | - actually report fork errors in gethostbyname and inet_aton. |
|
|
385 | |
|
|
386 | 4.11 Thu Oct 11 02:40:24 CEST 2007 |
|
|
387 | - port to threaded perls. |
|
|
388 | |
|
|
389 | 4.1 Thu Oct 11 02:38:16 CEST 2007 |
|
|
390 | - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now |
|
|
391 | be local to each coro (see Coro::State). |
|
|
392 | - incompatible change: for very deep reasons, cede and cede_notself |
|
|
393 | cannot return anything, so nothing will be returned. |
|
|
394 | - possibly bring back 5.10 compatibility (untested). |
|
|
395 | - work around stupid (and wrong) warning on 5.10 :(. |
|
|
396 | - overlay the saved state over the context stack. This saves |
|
|
397 | a few hundred bytes per coroutine on average and also |
|
|
398 | speeds up context switching a bit. |
|
|
399 | - further tune default stack sizes. |
|
|
400 | - (more) correctly calculate stack usage in coro_rss. |
|
|
401 | - Coro::Storable::blocking_* did not properly lock |
|
|
402 | resulting in races between coroutines. |
|
|
403 | - added Coro::Storable::guard. |
|
|
404 | - stopping to trace a coroutine could destroy the cctx of |
|
|
405 | an unrelated coroutine. |
|
|
406 | - explain the relationship between Perl and C coroutines in |
|
|
407 | more detail in Coro::State. |
|
|
408 | - Coro::Util::inet_aton did not short-circuit dotted quad forms, |
|
|
409 | causing a fork per resolve. This also affected Coro::Socket. |
|
|
410 | - switch to a separate stack in $coro->call/eval to avoid |
|
|
411 | invalidating pointers. |
|
|
412 | |
|
|
413 | 4.03 Sat Oct 6 21:24:00 CEST 2007 |
|
|
414 | - added Coro::throw method. |
|
|
415 | - minor code cleanups. |
|
|
416 | |
|
|
417 | 4.02 Sat Oct 6 02:36:47 CEST 2007 |
|
|
418 | - fix a very minor per-coroutine memleak (a single codereference). |
|
|
419 | - fixed a bug where the currently in-use c context would be freed |
|
|
420 | prematurely (can happen only when programs change the stacksize |
|
|
421 | or use tracing). |
|
|
422 | - tracing can no longer keep a coro alive after it terminated. |
|
|
423 | - do static branch prediction in the common path for gcc. gives |
|
|
424 | about 2-5% speed improvement here. |
|
|
425 | |
|
|
426 | 4.01 Fri Oct 5 22:10:49 CEST 2007 |
|
|
427 | - instead of recreating *a* standard output handle we simply |
|
|
428 | use STDOUT, which is faster and hopefully more robust. |
|
|
429 | |
|
|
430 | 4.0 Fri Oct 5 12:56:00 CEST 2007 |
|
|
431 | - incompatibly changed Coro::Storable::freeze. |
|
|
432 | - major new feature: added Coro::Debug, for interactive coroutine |
|
|
433 | debugging, tracing and much more. |
|
|
434 | - major bug fix: unbelievable, but true: $_, $/ and many other |
|
|
435 | "saved" variables actually weren't being saved. This has been fixed, |
|
|
436 | of course, while increasing performance while losing all the save |
|
|
437 | flags. |
|
|
438 | - save flags are gone, and all the api functions dealing with them. |
|
|
439 | - added Coro::Semaphore::adjust. |
|
|
440 | - added Coro::Util::fork_eval. |
|
|
441 | - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}. |
|
|
442 | - added Coro::killall. |
|
|
443 | - reduce initial stack sizes to allow for "micro-coroutines". |
|
|
444 | - better async_pool resource management, moved parts of async_pool |
|
|
445 | handling to XS (major speed improvement). |
|
|
446 | - actually croak before modifying important data structures. |
|
|
447 | - refuse to transfer while compiling. |
|
|
448 | - possibly support eval EXPR better now. |
|
|
449 | - enable assembly per default on linux+bsd x86+amd64. |
|
|
450 | - all internal members were renamed _something for easier subclassing. |
|
|
451 | - many minor tweaks. |
|
|
452 | |
|
|
453 | 3.63 Wed May 16 14:10:06 CEST 2007 |
|
|
454 | - implement handcoded assembly for x86/amd64 SVR ABI. |
|
|
455 | |
|
|
456 | 3.62 Fri Apr 27 21:36:06 CEST 2007 |
|
|
457 | - upgrade libcoro (which might set unwind info correctly). |
|
|
458 | - change default on linux to setjmp/longjmp. |
|
|
459 | |
|
|
460 | 3.61 Thu Apr 19 12:36:18 CEST 2007 |
|
|
461 | - Coro::Storable caused an endless loop when thawing invalid |
|
|
462 | pst's sometimes. |
|
|
463 | - use a Semaphore in Coro::Storable, as Storable doesn't |
|
|
464 | seem to be reentrant (although it is documented to |
|
|
465 | be threadsafe...). |
|
|
466 | - fix Coro::Signal to bring back the original unreliable |
|
|
467 | but stateful semantics. |
|
|
468 | - fixed a lot of typos in Coro.pm (patch submitted by David |
|
|
469 | Steinbrunner, which applied flawlessly). |
|
|
470 | |
|
|
471 | 3.6 Sat Apr 14 17:13:31 CEST 2007 |
|
|
472 | - added some bugfixes to get eg/myhttpd working again. |
|
|
473 | - added Coro::Storable for often-cede'ing freeze/thaw. |
|
|
474 | - try to do a clean exit when a coroutine calls exit |
|
|
475 | (EXPERIMENTAL). |
|
|
476 | - got rid of indirect call through _coro_init. |
|
|
477 | - updated the partly antique examples in eg/ to |
|
|
478 | work again and be a bit less magic, too. |
|
|
479 | - fixed Coro::Signal semantics to work as documented again. |
|
|
480 | |
|
|
481 | 3.55 Sun Mar 25 01:20:47 CET 2007 |
|
|
482 | - add SAVE_DEFFH to save the default filehandle and enable |
|
|
483 | it by default. |
|
|
484 | - finally move socket-operations from Coro::Socket to Coro::Handle |
|
|
485 | to be able to unblock foreign sockets. |
|
|
486 | - add Coro::State::save_also and guarded_save. |
|
|
487 | - add count accessor to Coro::Semaphore. |
|
|
488 | - add Coro::State::cctx_stacksize. |
|
|
489 | - just for the fun of it, do not rely on implicit context, |
|
|
490 | which can dramatically improve performance, but people |
|
|
491 | using windows-process-emulation perls do not care much about |
|
|
492 | performance. |
|
|
493 | |
|
|
494 | 3.51 Sun Mar 4 14:18:04 CET 2007 |
|
|
495 | - fixed a problem when you weakened references to Coro::State's |
|
|
496 | (patch by Yuval Kogman). |
|
|
497 | |
|
|
498 | 3.501 Wed Feb 28 12:44:07 CET 2007 |
|
|
499 | - rename some global symbols as macosx from hell redefines |
|
|
500 | them without asking. |
|
|
501 | |
|
|
502 | 3.5 Tue Feb 13 20:22:53 CET 2007 |
|
|
503 | - do AnyEvent model detection earlier, avoiding problems |
|
|
504 | caused by first using AnyEvent and later Coro::Event. |
|
|
505 | - implement and document Coro::Event event objects. |
|
|
506 | - fix a potential problem in Coro::Event causing crashes. |
|
|
507 | - initialise PL_comppad when creating a new coroutine, |
|
|
508 | avoids crashes on early coro destruction. |
|
|
509 | |
|
|
510 | 3.41 Mon Jan 22 19:19:49 CET 2007 |
|
|
511 | - readline on Coro::Handle did not support undefined $/, |
|
|
512 | nor did it deliver partial lines on EOF or error. |
|
|
513 | - implement malloc fallback for stack allocation because |
|
|
514 | stupid broken idiotic OSX has a stupid broken |
|
|
515 | idiotic fits-the-whole-os mmap "implementation" and |
|
|
516 | my dick feels longer if Coro is portable even to |
|
|
517 | obsolete platforms. |
|
|
518 | |
|
|
519 | 3.4 Fri Jan 19 21:52:54 CET 2007 |
|
|
520 | - remove t/09_timer.t, as it isn't really testing much |
|
|
521 | but was rather flaky in practise. |
|
|
522 | - async_pool coro would keep arguments and callback alive until |
|
|
523 | it was reused. |
|
|
524 | - cancellation of a coroutine could cause spurious idle calls |
|
|
525 | in cede_notself. |
|
|
526 | |
|
|
527 | 3.3 Sat Jan 6 03:45:00 CET 2007 |
|
|
528 | - implement $coro->on_destroy. |
|
|
529 | - Coro::Event blocking semantics have been changed, |
|
|
530 | documented and - hopefully - improved. |
|
|
531 | - fix nice adding, not subtracting, from priority. |
|
|
532 | - fix ->prio and api_is_ready (patch by Mark Hinds). |
|
|
533 | - fixed an assert ("... == prev__cctx->idle_te") |
|
|
534 | that could errronously trigger. |
|
|
535 | - fix various large and small memleaks. |
|
|
536 | - use a (hopefully) more stable cancel implementation |
|
|
537 | that immediately frees the coroutine data. |
|
|
538 | - cede/cede_notself return a status now. |
|
|
539 | - added Coro::guard function. |
|
|
540 | - added a global coroutine pool for jobs (on my machine, |
|
|
541 | I can create and execute 48k simple coros/s with async, |
|
|
542 | and 128k coros with async_pool). |
|
|
543 | - Coro::AIO now uses the coroutine priority as io priority. |
|
|
544 | |
|
|
545 | 3.2 Fri Dec 22 05:07:09 CET 2006 |
|
|
546 | - improve portability to slightly older perls. |
|
|
547 | - use cleaner coroutine destruction. |
|
|
548 | - simplify configuration for users. |
|
|
549 | - optionally (unrecommended) prefer perl functions over |
|
|
550 | their coro replacements. |
|
|
551 | |
|
|
552 | 3.11 Tue Dec 5 13:11:24 CET 2006 |
|
|
553 | - fixed some bogus assert's, but as perl.h disables assert even |
|
|
554 | without NDEBUG (thank you very much), not too many people should |
|
|
555 | notice (that did include myself). Andreas König noticed, though :) |
|
|
556 | - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have |
|
|
557 | it. Also noticed by Andreas König :) |
|
|
558 | - save/restore tainted status. |
|
|
559 | - verified to pass the testsuite on my 5.9.5. |
|
|
560 | |
|
|
561 | 3.1 Mon Dec 4 23:03:40 CET 2006 |
|
|
562 | - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow). |
|
|
563 | - incompatible change: transfer flags are now per-state. |
|
|
564 | - give a better error message on deadlock. |
|
|
565 | - document Coro::nready. |
|
|
566 | - enhanced testsuite. |
|
|
567 | |
|
|
568 | 3.01 Sun Dec 3 23:47:42 CET 2006 |
|
|
569 | - forgot to include Coro::Timer. |
|
|
570 | |
|
|
571 | 3.0 Sun Dec 3 22:57:25 CET 2006 |
|
|
572 | - the "FINALLY COMPLETELY STABLE" release (coming soon: |
|
|
573 | the "FAMOUS LAST WORDS" release). |
|
|
574 | - implement a new stack sharing algorithm, which uses a stack |
|
|
575 | pool (size currently hardcoded). |
|
|
576 | - make stack sharing mandatory (it no longer uses a heuristic). |
|
|
577 | - eval/die no longer cause weird problems under heavy use. |
|
|
578 | - Coro::Event could cause livelocks if it was used but |
|
|
579 | no Coro::Event watchers were used. |
|
|
580 | - Coro::Event now uses asynccheck as prepare does not |
|
|
581 | check for changed watchers. |
|
|
582 | - Coro::Event allows multiple waiting coros and will wake up one |
|
|
583 | per event. |
|
|
584 | - Coro::Event should be cleaner and more efficient now. |
|
|
585 | - new utility function Coro::unblock_sub. |
|
|
586 | - document the sad fact that Event is no longer reentrant. |
|
|
587 | - putting a coroutine into the ready queue twice could under |
|
|
588 | some circumstances lead to stack corruption. |
|
|
589 | - minor incompatibility: subclassing Coro::State is supported |
|
|
590 | directly now without going through a _coro_state member. |
|
|
591 | - state/coro switching is much faster now. |
|
|
592 | - very minor optimisations and code/documentation cleanup. |
|
|
593 | - avoid problems due to compiler inlining. |
|
|
594 | - removed timers from Coro::Timer -> use AnyEvent instead. |
|
|
595 | - replaced Coro::idle coroutine by (cleaner) idle handler. |
|
|
596 | - updated to newest libcoro. |
|
|
597 | - implement enhanced support for valgrind. |
|
|
598 | - implement is_ready and return value for ready. |
|
|
599 | - removed Coro::Cont, it was a misnomer (it's generators), |
|
|
600 | and it was somewhat annoying to get it right. Will come back |
|
|
601 | if somebody asks for it :) |
|
|
602 | - many ->wait methods and Coro::Event could return |
|
|
603 | spuriously without the event having happened. |
|
|
604 | |
|
|
605 | 2.5 Tue Nov 7 12:22:33 CET 2006 |
|
|
606 | - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket |
|
|
607 | use AnyEvent, moved them to Coro/. |
|
|
608 | - added Coro::LWP which contains all the uglyness required to |
|
|
609 | make LWP non-blocking. |
|
|
610 | - should work with perl 5.9.x now (Andreas König made me do it). |
|
|
611 | - fixed another bug in Coro::Select when the timeout was undef. |
|
|
612 | - reuse PL_start_env for all coros, saves some memory per coroutine. |
|
|
613 | - manage PL_top_env differently, hopefully to avoid panic: top_env. |
|
|
614 | - timeout argument was not properly used in Coro::Socket. |
|
|
615 | - allow limited forms of subclassing in Coro::Handle/Coro::Socket. |
|
|
616 | - emulate undocumented(!) functionality of IO::Socket required |
|
|
617 | by LWP(!!). |
|
|
618 | - updated eg/lwp to work with newer lwp's. |
|
|
619 | - remove "FATAL: uncaught exception" prefix. Coroutines that die |
|
|
620 | kill the whole process, just as exceptions in the main "coroutine" |
|
|
621 | did already. |
|
|
622 | |
|
|
623 | 2.1 Wed Nov 1 23:01:13 CET 2006 |
|
|
624 | - fix a long-standing bug in Coro::Select where select with |
|
|
625 | zero timeout would instead change the current default filehandle. |
|
|
626 | - use a simpler and hopefully more robust way to clone padlists |
|
|
627 | (uses less memory and a perl function instead of our own). |
|
|
628 | - coro can now create a stack guard area on many architectures. |
|
|
629 | - Coro::AIO properly reexports additional functions from IO::AIO. |
|
|
630 | - updated libcoro with a workaround for OS X, |
|
|
631 | pach and testing by Michael Schwern. |
|
|
632 | |
|
|
633 | 2.0 Tue Oct 24 05:47:17 CEST 2006 |
|
|
634 | - support additional aio requests in Coro::AIO. |
|
|
635 | |
|
|
636 | 1.9 (never properly released due to a glitch) |
|
|
637 | |
|
|
638 | 1.8 Thu Feb 2 00:59:06 CET 2006 |
|
|
639 | - applied suggested patch by SAMV to avoid problems during stupid |
|
|
640 | mark & sweep gc run. |
|
|
641 | - applied patch by Scott Walters for 5.9.3 compatibility. |
|
|
642 | |
|
|
643 | 1.7 Tue Dec 27 01:41:58 CET 2005 |
|
|
644 | - added Coro::AIO, a thin wrapper around IO::AIO. |
|
|
645 | - improved Makefile.PL explanations. |
|
|
646 | |
|
|
647 | 1.6 Mon Dec 12 21:30:05 CET 2005 |
|
|
648 | - additionally save PM_curpm might fix as-of-yet |
|
|
649 | unseen problems with regex matches being attributed to the |
|
|
650 | wrong package. |
|
|
651 | - add t/10_bugs.t, which currently checks against imho broken |
|
|
652 | perls that use 0.26MB of stack space per Perl_magic_get |
|
|
653 | invocation (newer linuxes) as opposed to the 0.0002MB perl |
|
|
654 | normally uses. |
|
|
655 | - make stacksize configurable for the ultimate debian experience. |
|
|
656 | |
|
|
657 | 1.51 Mon Dec 12 18:48:36 CET 2005 |
|
|
658 | - remove debugging warn() accidentally left in Coro::Select. |
|
|
659 | |
|
|
660 | 1.5 Tue Nov 29 13:32:44 CET 2005 |
|
|
661 | - use Coro::Event inside Coro::Select to avoid spurious deadlocks. |
|
|
662 | - fix Coro::Select 'select'. |
|
|
663 | - strict'ify some modules. |
|
|
664 | - libcoro errornously restored the SIGUSR2 handler to SIGUSR1. |
|
|
665 | - use XSLoader in selected modules. |
|
|
666 | - remove some 5.6 compatibility cruft. |
|
|
667 | |
|
|
668 | 1.4 Tue Sep 6 00:11:05 CEST 2005 |
|
|
669 | - libcoro did not take into account the trampoline on amd64, |
|
|
670 | when 'l'inux method was used. |
|
|
671 | |
|
|
672 | 1.31 Tue Aug 30 23:31:33 CEST 2005 |
|
|
673 | - some portability fixes/workarounds. |
|
|
674 | |
|
|
675 | 1.3 Sat Aug 20 03:08:56 CEST 2005 |
|
|
676 | - no code changes. module seems to work fine. |
|
|
677 | - improve Coro::State docs, remove reference to nonexisting |
|
|
678 | Coro::State::flush method. |
|
|
679 | - no longer autodetect windows, present it as an option instead. |
|
|
680 | |
|
|
681 | 1.2 Mon May 16 02:00:55 CEST 2005 |
|
|
682 | - included libcoro.c earlier in State.xs, might improve portability. |
|
|
683 | - use a faster and possibly more stable Coro::Cont implementation. |
|
|
684 | - accept x86_64 in addition to amd64 for optimized linux-amd64 support. |
|
|
685 | - fix bugs Coro::SemaphoreSet that could cause locks never to be freed. |
|
|
686 | - fix bugs in CoroAPI.h and document it in Coro::MakeMaker. |
|
|
687 | |
|
|
688 | 1.11 Thu Mar 3 18:00:52 CET 2005 |
|
|
689 | - change of contact address. |
|
|
690 | |
|
|
691 | 1.1 Tue Feb 22 20:51:16 CET 2005 |
|
|
692 | - support [l]inux method on amd64. |
|
|
693 | - allow some context switches while compiling/eval'ing. |
|
|
694 | |
|
|
695 | 1.0 Tue Aug 10 03:47:27 CEST 2004 |
|
|
696 | - Event 0.89 fixes the bug again. |
|
|
697 | - fix the remaining(??) memleak. I didn't free the cache-AV for |
|
|
698 | padlists, so one AV per created anon-sub leaked. |
|
|
699 | |
|
|
700 | 0.97 Fri May 14 15:23:32 CEST 2004 |
|
|
701 | - remove stupid gcc extension that was an experiment only anyway. |
|
|
702 | |
|
|
703 | 0.96 Thu May 13 18:09:29 CEST 2004 |
|
|
704 | - tell people if they have yet another broken Event variant. |
|
|
705 | - made Coro to work _better_ with threads (it seems that, as long |
|
|
706 | as you use Coros in one thread only everything will work fine. |
|
|
707 | Looking at the perl source the problem might be missing locking |
|
|
708 | between Coro and some perl internal routines. That's not easily |
|
|
709 | fixable). |
|
|
710 | - (hopefully) got rid of the static xnext inside transfer. |
|
|
711 | |
|
|
712 | 0.95 Sun Feb 15 03:20:28 CET 2004 |
|
|
713 | - removed Coro::State::flush, implemented a new and much |
|
|
714 | faster caching mechanism by attaching magic to the CV. |
|
|
715 | - a memleak with (real) closures remains. |
|
|
716 | - dramatically reduced size of stacks to 65536 (linux/x86), after |
|
|
717 | observing that even myhttpd never had more than a kilobyte |
|
|
718 | of stack in use (stack is still only physically allocated in |
|
|
719 | pagesize increments for systems suporting mmap). |
|
|
720 | - fix for a "cygwin-64" architecture added, proposed by Stefan Traby. |
|
|
721 | - fix for threaded perl. |
|
|
722 | |
|
|
723 | 0.9 Sun Nov 23 04:32:39 CET 2003 |
|
|
724 | - fix a few very hard to track down but (of course) stupid |
|
|
725 | memory leaks. |
|
|
726 | - stringify version number to avoid locale problems :(. |
|
|
727 | - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters, |
|
|
728 | hopefully ignoring it "just works". |
|
|
729 | |
|
|
730 | 0.8 Wed Nov 5 19:38:40 CET 2003 |
|
|
731 | - port to perl5.005_03, but only 5.8.x+ is supported! |
|
|
732 | - honor LocalAddr even without LocalPort in Coro::Socket. |
|
|
733 | |
|
|
734 | 0.7 Tue May 27 03:12:38 CEST 2003 |
|
|
735 | - the version jump indicates some level of testing, not gobs |
|
|
736 | of new features. |
|
|
737 | - uh... I found the prompt function in ExtUtils::MakeMaker. |
|
|
738 | Highly correct stuff, that is... |
|
|
739 | - fixed(?) a bug with die's in coroutines causing "panic: top_env". |
|
|
740 | the fix is not well understood by the author, so beware :(. |
|
|
741 | |
|
|
742 | 0.652 Thu May 8 02:54:46 CEST 2003 |
|
|
743 | - Applied patch by Slaven Rezic to set default to "s" on FreeBSD, |
|
|
744 | cause version 4 doesn't have ucontext. |
|
|
745 | - Benjamin Reed reported that setjmp works fine on darwin, |
|
|
746 | so preselect it. |
|
|
747 | |
|
|
748 | 0.651 Sat Mar 29 15:00:23 CET 2003 |
|
|
749 | - fix a bug in Coro::Handle where some bytes could get lost |
|
|
750 | on reading, reported by jason@nichego.net. |
2 | |
751 | |
3 | 0.65 Sun Mar 23 00:08:26 CET 2003 |
752 | 0.65 Sun Mar 23 00:08:26 CET 2003 |
4 | - added a README.linux-glibc. |
753 | - added a README.linux-glibc. |
5 | - new module Coro::Select. |
754 | - new module Coro::Select. |
6 | - also save/restore PL_comppad, fixes segfaults with 5.8.1. |
755 | - also save/restore PL_comppad, fixes segfaults with 5.8.1. |