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 |
4 | TODO: maybe implement a default message channel, very much like Erlang's |
5 | (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 | 4.02 |
8 | - "port" to threaded perls. |
|
|
9 | |
|
|
10 | 4.91 Mon Nov 10 05:36:38 CET 2008 |
|
|
11 | - the ->throw exception object no longer leaks. |
|
|
12 | - creating a new cctx leaked a scopestack entry (memleak). |
|
|
13 | - new coroutines didn't get created with a zero flags field |
|
|
14 | (unknown impact). |
|
|
15 | - calling ->throw on a not-yet-started coroutine should now work |
|
|
16 | instead of being ignored. |
|
|
17 | - ->throw is now supported on Coro::State objects. |
|
|
18 | - clean up cctx creation code a bit. |
|
|
19 | - entersub is actually a UNOP, not a LOGOP (not a bugfix). |
|
|
20 | |
|
|
21 | 4.9 Sat Nov 8 17:45:27 CET 2008 |
|
|
22 | - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
|
|
23 | - (libcoro) no longer leak threads in the experimental pthread backend, |
|
|
24 | also speed it up considerably. |
|
|
25 | - (libcoro) do not rely on makecontext passing void *'s unscathed. |
|
|
26 | - fix compiletime dependencies on libcoro in the Makefile. |
|
|
27 | - cctx_count wasn't always updated properly. |
|
|
28 | - Coro::State::cctx_stacksize wasn't applied correctly. |
|
|
29 | - new function Coro::State::cctx_max_idle. |
|
|
30 | - the default max number of idle C contexts is now 4. |
|
|
31 | - (libcoro) try harder to get _setjmp/_longjmp. |
|
|
32 | - (libcoro) cleanup and extend the libcoro API to officially |
|
|
33 | allow the creation of empty source contexts. |
|
|
34 | - very experimental workaround for the totally broken netbsd platform. |
|
|
35 | - tried to hack around openbsd bugs. |
|
|
36 | |
|
|
37 | 4.804 Wed Nov 5 16:36:00 CET 2008 |
|
|
38 | - Coro::Debug::new_unix_server would not create a non-blocking listening |
|
|
39 | socket, sometimes causing freezes. |
|
|
40 | - (libcoro) fix misaligned stack points for setjmp and assembly |
|
|
41 | methods, which can cause crashes on x86/x86_64 with a sufficiently |
|
|
42 | aggressive compiler. |
|
|
43 | - new function: Coro::Debug::new_tcp_server. |
|
|
44 | - move ->throw into the Coro class because it only works on coro objects. |
|
|
45 | |
|
|
46 | 4.803 Mon Nov 3 17:16:12 CET 2008 |
|
|
47 | - (libcoro) use a global asm statement to become independent of gcc |
|
|
48 | otpimisations for CORO_ASM (thanks to pippijn for the idea). |
|
|
49 | - try to workaround yet another broken bsd, this time dragonfly. |
|
|
50 | |
|
|
51 | 4.802 Thu Oct 30 10:56:12 CET 2008 |
|
|
52 | - support -fno-omit-frame-pointer on x86 with the assembly method. |
|
|
53 | - tune 01_unblock.t tests a bit. |
|
|
54 | |
|
|
55 | 4.801 Wed Oct 22 18:33:37 CEST 2008 |
|
|
56 | - improve readline speed for very long "lines". |
|
|
57 | - backport to 5.8.8. |
|
|
58 | |
|
|
59 | 4.8 Thu Oct 2 13:34:40 CEST 2008 |
|
|
60 | - new function Coro::AIO::aio_wait. |
|
|
61 | - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". |
|
|
62 | - greatly speed up and reduce memory usage of Coro::AIO requests. |
|
|
63 | - implement some other µ-optimisations. |
|
|
64 | |
|
|
65 | 4.749 Mon Sep 29 14:40:12 CEST 2008 |
|
|
66 | - port to slow and broken pseudo-threaded perls. (courtesy pippijn). |
|
|
67 | |
|
|
68 | 4.748 Sat Sep 27 14:03:19 CEST 2008 |
|
|
69 | - implement, but do not document, PerlIO::cede(granularity). |
|
|
70 | - Coro::Storable forgot to wrap Storable::pstore. |
|
|
71 | - work around the multitude of leaks and memory corruption |
|
|
72 | bugs in PerlIO::via by using our own C-level perliol. As a side |
|
|
73 | effect, Coro::Storable is now much, much, much faster. |
|
|
74 | |
|
|
75 | 4.747 Tue Sep 23 01:59:41 CEST 2008 |
|
|
76 | - fix a per-cv memleak (one empty array was leaked per |
|
|
77 | code reference). |
|
|
78 | - avoid a crash in coro->call|rss when the coroutine was already |
|
|
79 | destroyed (most noticably when using Coro::Debug::ps :) |
|
|
80 | - also protect *Storable::FILE. |
|
|
81 | - push up default storable granularity to 20ms. |
|
|
82 | |
|
|
83 | 4.746 Sun Sep 21 03:22:20 CEST 2008 |
|
|
84 | - be more insistent on locking Storable against reentrancy |
|
|
85 | in Coro::Storable. |
|
|
86 | - move swap_def?v and throw to Coro::State, as documented. |
|
|
87 | |
|
|
88 | 4.745 Thu Jul 24 00:14:38 CEST 2008 |
|
|
89 | - remove debugging code related to MgPV_nolen_const, also try to |
|
|
90 | make it compile with perl 5.8.6 (yes, apple apparently loves |
|
|
91 | outdated software). Reported by John S. |
|
|
92 | |
|
|
93 | 4.744 Tue Jul 8 22:06:35 CEST 2008 |
|
|
94 | - correctly provide default DIE/WARN handlers as documented. |
|
|
95 | - also overwrite PL_vtbl_sigelem.svt_clear, even though current |
|
|
96 | implementations inside perl work fine for us. |
|
|
97 | |
|
|
98 | 4.743 Mon Jun 16 00:21:57 CEST 2008 |
|
|
99 | - when using Coro::EV without running EV::loop it could |
|
|
100 | result in busy-waiting for events, this has been fixed. |
|
|
101 | - reduce codesize and improve performance by using EV_DEFAULT_UC. |
|
|
102 | |
|
|
103 | 4.742 Sat May 31 14:10:21 CEST 2008 |
|
|
104 | - implement a workaround for (some) perl <5.8.8 versions. |
|
|
105 | - require EV 3.3+. |
|
|
106 | |
|
|
107 | 4.741 Fri May 30 23:33:09 CEST 2008 |
|
|
108 | - tell netbsd how utterly broken their imitation of an OS is |
|
|
109 | and refuse to build by default if pthreads are in use. |
|
|
110 | - switch to "s" method on all bsd's by default, as their ucontext |
|
|
111 | stuff seems just too broken. |
|
|
112 | - fix a bug in Coro::Select. |
|
|
113 | |
|
|
114 | 4.74 Thu May 29 20:05:31 CEST 2008 |
|
|
115 | - do not test Coro::LWP for lack of dependencies. |
|
|
116 | - do not test Coro::Select for lack of working perls. |
|
|
117 | |
|
|
118 | 4.73 Thu May 29 2008 |
|
|
119 | - fix a bug in Coro::EV which would cause it to block despite |
|
|
120 | there being runnable coroutines. |
|
|
121 | - sprinkle "no warnings" freely over everything, also suppress |
|
|
122 | warnings for some other modules. |
|
|
123 | - fix typo in WSAEWOULDBLOCK. |
|
|
124 | |
|
|
125 | 4.72 Sun May 25 05:14:36 CEST 2008 |
|
|
126 | - tweak META.yaml a bit, unfortunately, there is no documented way |
|
|
127 | to have optional dependencies with CPAN. doh :( |
|
|
128 | - avoid running some tests on windows because they would fail due to |
|
|
129 | perl bug (broken fork, broken pipes...). |
|
|
130 | - work around perl on windows bugs where perl returns undocumented |
|
|
131 | error codes for sysread, syswrite etc. by taking advantage |
|
|
132 | of AnyEvent's workaround for that problem. |
|
|
133 | - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around |
|
|
134 | a common perl implementation bug on windows. |
|
|
135 | - use unix domain sockets in testsuite to work around |
|
|
136 | common perl implementation bugs on widows (they are emulated by |
|
|
137 | tcp sockets on windows. ugh.) |
|
|
138 | |
|
|
139 | 4.71 Sat May 24 20:01:27 CEST 2008 |
|
|
140 | - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). |
|
|
141 | - take advantage of async name resolution of AnyEvent::Util. |
|
|
142 | - work around brutal inet_aton override in Coro::LWP. |
|
|
143 | - take advantage of the readyhook in Coro::EV, for smoother |
|
|
144 | scheduling. |
|
|
145 | |
|
|
146 | 4.7 Sun May 11 00:32:19 CEST 2008 |
|
|
147 | - completely reworked the Coro manpage. |
|
|
148 | - added Coro::AnyEvent, generic event loop integration. |
|
|
149 | - implement cancel, ready and kill commands in Coro::Debug. |
|
|
150 | - document find_coro in Coro::Debug. |
|
|
151 | - incompatible change: rename has_stack to has_cctx. |
|
|
152 | - Coro::AIO and Coro::BDB no longer force event model detection, |
|
|
153 | use AnyEvent::AIO and AnyEvent::BDB. |
|
|
154 | |
|
|
155 | 4.6 Sat Apr 26 10:05:14 CEST 2008 |
|
|
156 | - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported |
|
|
157 | by the Coro module. It was a mistake to have it in the first place. |
|
|
158 | - (experimental) support for activestate perl 5.10 (method "w"). |
|
|
159 | - (experimental) support for strawberry perl (method "a"). |
|
|
160 | - coro_sigelem_set did not return a value although it had to, |
|
|
161 | actual impact unknown. |
|
|
162 | |
|
|
163 | 4.51 Mon Apr 14 13:28:27 CEST 2008 |
|
|
164 | - make it compile again on 5.8. |
|
|
165 | |
|
|
166 | 4.50 Thu Apr 10 09:43:17 CEST 2008 |
|
|
167 | - I did it twice! (see 4.49). |
|
|
168 | |
|
|
169 | 4.49 Sun Apr 6 21:23:31 CEST 2008 |
|
|
170 | - grr, instead of compiling the recent changes on 5.10 only they |
|
|
171 | were compiled on 5.8 only. |
|
|
172 | |
|
|
173 | 4.48 Sun Apr 6 20:36:46 CEST 2008 |
|
|
174 | - allow coroutine switches during eval's under 5.10.x, as it seems |
|
|
175 | the parser is a per-interpreter option now, so this is safe |
|
|
176 | (this might fix the odd crashing bug). |
|
|
177 | - drop support for 5.9.x versions: they are dead, jim. |
|
|
178 | |
|
|
179 | 4.47 Sun Apr 6 00:37:52 CEST 2008 |
|
|
180 | - force cctx allocation on API calls: we know we need to force one |
|
|
181 | and gcc actually manages to confuse our heuristic nowadays, |
|
|
182 | leading to crashes and worse. |
|
|
183 | - document force_cctx. |
|
|
184 | |
|
|
185 | 4.46 Fri Apr 4 22:05:43 CEST 2008 |
|
|
186 | - upgrade libcoro, resulting in pthread-backend (which was only created |
|
|
187 | to fulfill the rules of the programming languages shootout). |
|
|
188 | |
|
|
189 | 4.45 Thu Mar 13 11:55:36 CET 2008 |
|
|
190 | - fix a file leaking bug in eg/mhyttpd that would allow |
|
|
191 | downloading of any file (reported by oesi). |
|
|
192 | - fix deadlock bug in Coro::Channel (reported by Richard Hundt) |
|
|
193 | (also add testcase). |
|
|
194 | - support Broadcast option in Coro::Socket (patch by Richard Hundt, |
|
|
195 | apparently having loads of fun with that). |
|
|
196 | - implement $state->swap_defsv, ->swap_defav and document ->throw. |
|
|
197 | |
|
|
198 | 4.4 Wed Feb 13 16:44:29 CET 2008 |
|
|
199 | - only assume .cfi pseudo isns exist on GNU/Linux. |
|
|
200 | - add get_timed method to Coro::Channels. |
|
|
201 | - fixed Coro::Debug manpage. |
|
|
202 | - perl 5.11 compatbility improvement based on patch by |
|
|
203 | Andreas König. |
|
|
204 | |
|
|
205 | 4.37 Sun Jan 20 11:25:23 CET 2008 |
|
|
206 | - soften the check for an active parser for perl 5.10 (avoiding |
|
|
207 | "Coro::State::transfer called while parsing" in many cases). |
|
|
208 | |
|
|
209 | 4.36 Sun Jan 13 10:53:56 CET 2008 |
|
|
210 | - reset diehook when terminating from an async_pool as to not |
|
|
211 | trigger any __DIE__ handlers. |
|
|
212 | |
|
|
213 | 4.35 Sun Jan 13 04:14:13 CET 2008 |
|
|
214 | - "bt" debug command now displays any exceptions |
|
|
215 | from longmess and also skips the innermost |
|
|
216 | stackframes, giving more useufl output. |
|
|
217 | - allow backtraces in coroutines blocked in Coro::EV, |
|
|
218 | at a <1% speed hit. |
|
|
219 | - handle localising of $SIG{__WARN__} and __DIE__ |
|
|
220 | properly (with a proper amount of dirty hacking). |
|
|
221 | |
|
|
222 | 4.34 Sat Dec 22 17:49:53 CET 2007 |
|
|
223 | - upgrade to EV version 2.0 API. |
|
|
224 | |
|
|
225 | 4.33 Mon Dec 17 08:36:12 CET 2007 |
|
|
226 | - make Coro::AIO etc. loadable in the absence of EV. |
|
|
227 | |
|
|
228 | 4.32 Mon Dec 17 07:46:02 CET 2007 |
|
|
229 | - majorly improved Coro::Handle's performance with Coro::EV. |
|
|
230 | - implemented the until now mythical Coro::BDB module. |
|
|
231 | - specialcase EV in Coro::AIO and Coro::BDB for extra speed. |
|
|
232 | |
|
|
233 | 4.31 Wed Dec 5 12:32:39 CET 2007 |
|
|
234 | - remove warn statement form Coro::Util. |
|
|
235 | |
|
|
236 | 4.3 Tue Dec 4 20:33:14 CET 2007 |
|
|
237 | - calls to the idle function could cause stack corruption |
|
|
238 | when the stack changed. |
|
|
239 | - do no longer rely on the presence of EV::DNS if EV is used |
|
|
240 | (because it is gone), but instead take avdantage of EV::ADNS |
|
|
241 | if available. |
|
|
242 | - add ($) prototypes to all functions in Coro::Storable. |
|
|
243 | - use a conventional (safer) idle callback in Coro::EV. |
|
|
244 | - do accept quickly in Coro::Debug to avoid endless loops. |
|
|
245 | |
|
|
246 | 4.22 Fri Nov 30 16:04:04 CET 2007 |
|
|
247 | - really use optimised versions for Event and EV in Coro::Util |
|
|
248 | and Coro::Handle. |
|
|
249 | |
|
|
250 | 4.21 Sun Nov 25 10:48:59 CET 2007 |
|
|
251 | - fix a spurious memory read. |
|
|
252 | - Coro::EV no longer keeps the eventloop "alive". |
|
|
253 | |
|
|
254 | 4.2 Fri Nov 9 20:47:05 CET 2007 |
|
|
255 | - enable/disable tracing from a new coroutine, not a pooled one. |
|
|
256 | - fix a memleak in Coro::Event. |
|
|
257 | - removed killall call from fork_eval. |
|
|
258 | - made sure store_fd is already loaded so that fork_eval |
|
|
259 | does not have to parse autoload in each subprocess. |
|
|
260 | - only use assembly method if -O switch is in $Config{optimize}. |
|
|
261 | - add (optional) Coro::EV module, so far the best event loop module |
|
|
262 | directly supported by Coro. |
|
|
263 | - if the event model is EV, use EV::DNS to resolve |
|
|
264 | stuff in Coro::Util. |
|
|
265 | - don't get confused by multiple event notifications in Coro::Handle. |
|
|
266 | - initial support for EV (libevent interface). |
|
|
267 | - require Event and EV using configure_requires, to force their existance. |
|
|
268 | |
|
|
269 | 4.13 Wed Oct 24 07:26:45 CEST 2007 |
|
|
270 | - add Coro::Storable::blocking_thaw. |
|
|
271 | - use a vastly more complicated technique to localise |
|
|
272 | $SIG{__WARN/DIE__} that also works on perls <= 5.8.8. |
|
|
273 | - use a coroutine for the idle callback Coro::Event, |
|
|
274 | instead of running Event in the current coroutine context. |
|
|
275 | This also catches recursive invocations. |
|
|
276 | - actually report fork errors in gethostbyname and inet_aton. |
|
|
277 | |
|
|
278 | 4.11 Thu Oct 11 02:40:24 CEST 2007 |
|
|
279 | - port to threaded perls. |
|
|
280 | |
|
|
281 | 4.1 Thu Oct 11 02:38:16 CEST 2007 |
|
|
282 | - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now |
|
|
283 | be local to each coro (see Coro::State). |
|
|
284 | - incompatible change: for very deep reasons, cede and cede_notself |
|
|
285 | cannot return anything, so nothing will be returned. |
|
|
286 | - possibly bring back 5.10 compatibility (untested). |
|
|
287 | - work around stupid (and wrong) warning on 5.10 :(. |
|
|
288 | - overlay the saved state over the context stack. This saves |
|
|
289 | a few hundred bytes per coroutine on average and also |
|
|
290 | speeds up context switching a bit. |
|
|
291 | - further tune default stack sizes. |
|
|
292 | - (more) correctly calculate stack usage in coro_rss. |
|
|
293 | - Coro::Storable::blocking_* did not properly lock |
|
|
294 | resulting in races between coroutines. |
|
|
295 | - added Coro::Storable::guard. |
|
|
296 | - stopping to trace a coroutine could destroy the cctx of |
|
|
297 | an unrelated coroutine. |
|
|
298 | - explain the relationship between Perl and C coroutines in |
|
|
299 | more detail in Coro::State. |
|
|
300 | - Coro::Util::inet_aton did not short-circuit dotted quad forms, |
|
|
301 | causing a fork per resolve. This also affected Coro::Socket. |
|
|
302 | - switch to a separate stack in $coro->call/eval to avoid |
|
|
303 | invalidating pointers. |
|
|
304 | |
|
|
305 | 4.03 Sat Oct 6 21:24:00 CEST 2007 |
|
|
306 | - added Coro::throw method. |
|
|
307 | - minor code cleanups. |
|
|
308 | |
|
|
309 | 4.02 Sat Oct 6 02:36:47 CEST 2007 |
9 | - fix a very minor per-coroutine memleak (a single codereference). |
310 | - fix a very minor per-coroutine memleak (a single codereference). |
|
|
311 | - fixed a bug where the currently in-use c context would be freed |
|
|
312 | prematurely (can happen only when programs change the stacksize |
|
|
313 | or use tracing). |
|
|
314 | - tracing can no longer keep a coro alive after it terminated. |
|
|
315 | - do static branch prediction in the common path for gcc. gives |
|
|
316 | about 2-5% speed improvement here. |
10 | |
317 | |
11 | 4.01 Fri Oct 5 22:10:49 CEST 2007 |
318 | 4.01 Fri Oct 5 22:10:49 CEST 2007 |
12 | - instead of recreating *a* standard output handle we simply |
319 | - instead of recreating *a* standard output handle we simply |
13 | use STDOUT, which is faster and hopefully more robust. |
320 | use STDOUT, which is faster and hopefully more robust. |
14 | |
321 | |