| 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. |
|
|
4 | TODO: maybe implement a default message channel, very much like Erlang |
|
|
5 | (which is cool in a lot of important aspects (failures!), |
|
|
6 | but very lacking in others (higher level ipc)). |
|
|
7 | |
|
|
8 | 4.11 Thu Oct 11 02:40:24 CEST 2007 |
|
|
9 | - port to threaded perls. |
|
|
10 | |
|
|
11 | 4.1 Thu Oct 11 02:38:16 CEST 2007 |
|
|
12 | - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now |
|
|
13 | be local to each coro (see Coro::State). |
|
|
14 | - incompatible change: for very deep reasons, cede and cede_notself |
|
|
15 | cannot return anything, so nothing will be returned. |
|
|
16 | - possibly bring back 5.10 compatibility (untested). |
|
|
17 | - work around stupid (and wrong) warning on 5.10 :(. |
|
|
18 | - overlay the saved state over the context stack. This saves |
|
|
19 | a few hundred bytes per coroutine on average and also |
|
|
20 | speeds up context switching a bit. |
|
|
21 | - further tune default stack sizes. |
|
|
22 | - (more) correctly calculate stack usage in coro_rss. |
|
|
23 | - Coro::Storable::blocking_* did not properly lock |
|
|
24 | resulting in races between coroutines. |
|
|
25 | - added Coro::Storable::guard. |
|
|
26 | - stopping to trace a coroutine could destroy the cctx of |
|
|
27 | an unrelated coroutine. |
|
|
28 | - explain the relationship between Perl and C coroutines in |
|
|
29 | more detail in Coro::State. |
|
|
30 | - Coro::Util::inet_aton did not short-circuit dotted quad forms, |
|
|
31 | causing a fork per resolve. This also affected Coro::Socket. |
|
|
32 | - switch to a separate stack in $coro->call/eval to avoid |
|
|
33 | invalidating pointers. |
|
|
34 | |
|
|
35 | 4.03 Sat Oct 6 21:24:00 CEST 2007 |
|
|
36 | - added Coro::throw method. |
|
|
37 | - minor code cleanups. |
|
|
38 | |
|
|
39 | 4.02 Sat Oct 6 02:36:47 CEST 2007 |
|
|
40 | - fix a very minor per-coroutine memleak (a single codereference). |
|
|
41 | - fixed a bug where the currently in-use c context would be freed |
|
|
42 | prematurely (can happen only when programs change the stacksize |
|
|
43 | or use tracing). |
|
|
44 | - tracing can no longer keep a coro alive after it terminated. |
|
|
45 | - do static branch prediction in the common path for gcc. gives |
|
|
46 | about 2-5% speed improvement here. |
|
|
47 | |
|
|
48 | 4.01 Fri Oct 5 22:10:49 CEST 2007 |
|
|
49 | - instead of recreating *a* standard output handle we simply |
|
|
50 | use STDOUT, which is faster and hopefully more robust. |
|
|
51 | |
|
|
52 | 4.0 Fri Oct 5 12:56:00 CEST 2007 |
|
|
53 | - incompatibly changed Coro::Storable::freeze. |
|
|
54 | - major new feature: added Coro::Debug, for interactive coroutine |
|
|
55 | debugging, tracing and much more. |
|
|
56 | - major bug fix: unbelievable, but true: $_, $/ and many other |
|
|
57 | "saved" variables actually weren't being saved. This has been fixed, |
|
|
58 | of course, while increasing performance while losing all the save |
|
|
59 | flags. |
|
|
60 | - save flags are gone, and all the api functions dealing with them. |
|
|
61 | - added Coro::Semaphore::adjust. |
|
|
62 | - added Coro::Util::fork_eval. |
|
|
63 | - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}. |
|
|
64 | - added Coro::killall. |
|
|
65 | - reduce initial stack sizes to allow for "micro-coroutines". |
|
|
66 | - better async_pool resource management, moved parts of async_pool |
|
|
67 | handling to XS (major speed improvement). |
|
|
68 | - actually croak before modifying important data structures. |
|
|
69 | - refuse to transfer while compiling. |
|
|
70 | - possibly support eval EXPR better now. |
|
|
71 | - enable assembly per default on linux+bsd x86+amd64. |
|
|
72 | - all internal members were renamed _something for easier subclassing. |
|
|
73 | - many minor tweaks. |
|
|
74 | |
|
|
75 | 3.63 Wed May 16 14:10:06 CEST 2007 |
|
|
76 | - implement handcoded assembly for x86/amd64 SVR ABI. |
|
|
77 | |
|
|
78 | 3.62 Fri Apr 27 21:36:06 CEST 2007 |
|
|
79 | - upgrade libcoro (which might set unwind info correctly). |
|
|
80 | - change default on linux to setjmp/longjmp. |
|
|
81 | |
|
|
82 | 3.61 Thu Apr 19 12:36:18 CEST 2007 |
|
|
83 | - Coro::Storable caused an endless loop when thawing invalid |
|
|
84 | pst's sometimes. |
|
|
85 | - use a Semaphore in Coro::Storable, as Storable doesn't |
|
|
86 | seem to be reentrant (although it is documented to |
|
|
87 | be threadsafe...). |
|
|
88 | - fix Coro::Signal to bring back the original unreliable |
|
|
89 | but stateful semantics. |
|
|
90 | - fixed a lot of typos in Coro.pm (patch submitted by David |
|
|
91 | Steinbrunner, which applied flawlessly). |
|
|
92 | |
|
|
93 | 3.6 Sat Apr 14 17:13:31 CEST 2007 |
|
|
94 | - added some bugfixes to get eg/myhttpd working again. |
|
|
95 | - added Coro::Storable for often-cede'ing freeze/thaw. |
|
|
96 | - try to do a clean exit when a coroutine calls exit |
|
|
97 | (EXPERIMENTAL). |
|
|
98 | - got rid of indirect call through _coro_init. |
|
|
99 | - updated the partly antique examples in eg/ to |
|
|
100 | work again and be a bit less magic, too. |
|
|
101 | - fixed Coro::Signal semantics to work as documented again. |
|
|
102 | |
|
|
103 | 3.55 Sun Mar 25 01:20:47 CET 2007 |
|
|
104 | - add SAVE_DEFFH to save the default filehandle and enable |
|
|
105 | it by default. |
|
|
106 | - finally move socket-operations from Coro::Socket to Coro::Handle |
|
|
107 | to be able to unblock foreign sockets. |
|
|
108 | - add Coro::State::save_also and guarded_save. |
|
|
109 | - add count accessor to Coro::Semaphore. |
|
|
110 | - add Coro::State::cctx_stacksize. |
|
|
111 | - just for the fun of it, do not rely on implicit context, |
|
|
112 | which can dramatically improve performance, but people |
|
|
113 | using windows-process-emulation perls do not care much about |
|
|
114 | performance. |
|
|
115 | |
|
|
116 | 3.51 Sun Mar 4 14:18:04 CET 2007 |
|
|
117 | - fixed a problem when you weakened references to Coro::State's |
|
|
118 | (patch by Yuval Kogman). |
|
|
119 | |
|
|
120 | 3.501 Wed Feb 28 12:44:07 CET 2007 |
|
|
121 | - rename some global symbols as macosx from hell redefines |
|
|
122 | them without asking. |
|
|
123 | |
|
|
124 | 3.5 Tue Feb 13 20:22:53 CET 2007 |
|
|
125 | - do AnyEvent model detection earlier, avoiding problems |
|
|
126 | caused by first using AnyEvent and later Coro::Event. |
|
|
127 | - implement and document Coro::Event event objects. |
|
|
128 | - fix a potential problem in Coro::Event causing crashes. |
|
|
129 | - initialise PL_comppad when creating a new coroutine, |
|
|
130 | avoids crashes on early coro destruction. |
|
|
131 | |
|
|
132 | 3.41 Mon Jan 22 19:19:49 CET 2007 |
|
|
133 | - readline on Coro::Handle did not support undefined $/, |
|
|
134 | nor did it deliver partial lines on EOF or error. |
|
|
135 | - implement malloc fallback for stack allocation because |
|
|
136 | stupid broken idiotic OSX has a stupid broken |
|
|
137 | idiotic fits-the-whole-os mmap "implementation" and |
|
|
138 | my dick feels longer if Coro is portable even to |
|
|
139 | obsolete platforms. |
|
|
140 | |
|
|
141 | 3.4 Fri Jan 19 21:52:54 CET 2007 |
|
|
142 | - remove t/09_timer.t, as it isn't really testing much |
|
|
143 | but was rather flaky in practise. |
|
|
144 | - async_pool coro would keep arguments and callback alive until |
|
|
145 | it was reused. |
|
|
146 | - cancellation of a coroutine could cause spurious idle calls |
|
|
147 | in cede_notself. |
|
|
148 | |
|
|
149 | 3.3 Sat Jan 6 03:45:00 CET 2007 |
|
|
150 | - implement $coro->on_destroy. |
|
|
151 | - Coro::Event blocking semantics have been changed, |
|
|
152 | documented and - hopefully - improved. |
|
|
153 | - fix nice adding, not subtracting, from priority. |
|
|
154 | - fix ->prio and api_is_ready (patch by Mark Hinds). |
|
|
155 | - fixed an assert ("... == prev__cctx->idle_te") |
|
|
156 | that could errronously trigger. |
|
|
157 | - fix various large and small memleaks. |
|
|
158 | - use a (hopefully) more stable cancel implementation |
|
|
159 | that immediately frees the coroutine data. |
|
|
160 | - cede/cede_notself return a status now. |
|
|
161 | - added Coro::guard function. |
|
|
162 | - added a global coroutine pool for jobs (on my machine, |
|
|
163 | I can create and execute 48k simple coros/s with async, |
|
|
164 | and 128k coros with async_pool). |
|
|
165 | - Coro::AIO now uses the coroutine priority as io priority. |
|
|
166 | |
|
|
167 | 3.2 Fri Dec 22 05:07:09 CET 2006 |
|
|
168 | - improve portability to slightly older perls. |
|
|
169 | - use cleaner coroutine destruction. |
|
|
170 | - simplify configuration for users. |
|
|
171 | - optionally (unrecommended) prefer perl functions over |
|
|
172 | their coro replacements. |
|
|
173 | |
|
|
174 | 3.11 Tue Dec 5 13:11:24 CET 2006 |
|
|
175 | - fixed some bogus assert's, but as perl.h disables assert even |
|
|
176 | without NDEBUG (thank you very much), not too many people should |
|
|
177 | notice (that did include myself). Andreas König noticed, though :) |
|
|
178 | - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have |
|
|
179 | it. Also noticed by Andreas König :) |
|
|
180 | - save/restore tainted status. |
|
|
181 | - verified to pass the testsuite on my 5.9.5. |
|
|
182 | |
|
|
183 | 3.1 Mon Dec 4 23:03:40 CET 2006 |
|
|
184 | - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow). |
|
|
185 | - incompatible change: transfer flags are now per-state. |
|
|
186 | - give a better error message on deadlock. |
|
|
187 | - document Coro::nready. |
|
|
188 | - enhanced testsuite. |
|
|
189 | |
|
|
190 | 3.01 Sun Dec 3 23:47:42 CET 2006 |
|
|
191 | - forgot to include Coro::Timer. |
|
|
192 | |
|
|
193 | 3.0 Sun Dec 3 22:57:25 CET 2006 |
|
|
194 | - the "FINALLY COMPLETELY STABLE" release (coming soon: |
|
|
195 | the "FAMOUS LAST WORDS" release). |
|
|
196 | - implement a new stack sharing algorithm, which uses a stack |
|
|
197 | pool (size currently hardcoded). |
|
|
198 | - make stack sharing mandatory (it no longer uses a heuristic). |
|
|
199 | - eval/die no longer cause weird problems under heavy use. |
|
|
200 | - Coro::Event could cause livelocks if it was used but |
|
|
201 | no Coro::Event watchers were used. |
|
|
202 | - Coro::Event now uses asynccheck as prepare does not |
|
|
203 | check for changed watchers. |
|
|
204 | - Coro::Event allows multiple waiting coros and will wake up one |
|
|
205 | per event. |
|
|
206 | - Coro::Event should be cleaner and more efficient now. |
|
|
207 | - new utility function Coro::unblock_sub. |
|
|
208 | - document the sad fact that Event is no longer reentrant. |
|
|
209 | - putting a coroutine into the ready queue twice could under |
|
|
210 | some circumstances lead to stack corruption. |
|
|
211 | - minor incompatibility: subclassing Coro::State is supported |
|
|
212 | directly now without going through a _coro_state member. |
|
|
213 | - state/coro switching is much faster now. |
|
|
214 | - very minor optimisations and code/documentation cleanup. |
|
|
215 | - avoid problems due to compiler inlining. |
|
|
216 | - removed timers from Coro::Timer -> use AnyEvent instead. |
|
|
217 | - replaced Coro::idle coroutine by (cleaner) idle handler. |
|
|
218 | - updated to newest libcoro. |
|
|
219 | - implement enhanced support for valgrind. |
|
|
220 | - implement is_ready and return value for ready. |
|
|
221 | - removed Coro::Cont, it was a misnomer (it's generators), |
|
|
222 | and it was somewhat annoying to get it right. Will come back |
|
|
223 | if somebody asks for it :) |
|
|
224 | - many ->wait methods and Coro::Event could return |
|
|
225 | spuriously without the event having happened. |
|
|
226 | |
|
|
227 | 2.5 Tue Nov 7 12:22:33 CET 2006 |
|
|
228 | - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket |
|
|
229 | use AnyEvent, moved them to Coro/. |
|
|
230 | - added Coro::LWP which contains all the uglyness required to |
|
|
231 | make LWP non-blocking. |
|
|
232 | - should work with perl 5.9.x now (Andreas König made me do it). |
|
|
233 | - fixed another bug in Coro::Select when the timeout was undef. |
|
|
234 | - reuse PL_start_env for all coros, saves some memory per coroutine. |
|
|
235 | - manage PL_top_env differently, hopefully to avoid panic: top_env. |
|
|
236 | - timeout argument was not properly used in Coro::Socket. |
|
|
237 | - allow limited forms of subclassing in Coro::Handle/Coro::Socket. |
|
|
238 | - emulate undocumented(!) functionality of IO::Socket required |
|
|
239 | by LWP(!!). |
|
|
240 | - updated eg/lwp to work with newer lwp's. |
|
|
241 | - remove "FATAL: uncaught exception" prefix. Coroutines that die |
|
|
242 | kill the whole process, just as exceptions in the main "coroutine" |
|
|
243 | did already. |
|
|
244 | |
|
|
245 | 2.1 Wed Nov 1 23:01:13 CET 2006 |
|
|
246 | - fix a long-standing bug in Coro::Select where select with |
|
|
247 | zero timeout would instead change the current default filehandle. |
|
|
248 | - use a simpler and hopefully more robust way to clone padlists |
|
|
249 | (uses less memory and a perl function instead of our own). |
|
|
250 | - coro can now create a stack guard area on many architectures. |
|
|
251 | - Coro::AIO properly reexports additional functions from IO::AIO. |
|
|
252 | - updated libcoro with a workaround for OS X, |
|
|
253 | pach and testing by Michael Schwern. |
|
|
254 | |
|
|
255 | 2.0 Tue Oct 24 05:47:17 CEST 2006 |
|
|
256 | - support additional aio requests in Coro::AIO. |
|
|
257 | |
|
|
258 | 1.9 (never properly released due to a glitch) |
|
|
259 | |
|
|
260 | 1.8 Thu Feb 2 00:59:06 CET 2006 |
|
|
261 | - applied suggested patch by SAMV to avoid problems during stupid |
|
|
262 | mark & sweep gc run. |
|
|
263 | - applied patch by Scott Walters for 5.9.3 compatibility. |
|
|
264 | |
|
|
265 | 1.7 Tue Dec 27 01:41:58 CET 2005 |
|
|
266 | - added Coro::AIO, a thin wrapper around IO::AIO. |
|
|
267 | - improved Makefile.PL explanations. |
|
|
268 | |
|
|
269 | 1.6 Mon Dec 12 21:30:05 CET 2005 |
|
|
270 | - additionally save PM_curpm might fix as-of-yet |
|
|
271 | unseen problems with regex matches being attributed to the |
|
|
272 | wrong package. |
|
|
273 | - add t/10_bugs.t, which currently checks against imho broken |
|
|
274 | perls that use 0.26MB of stack space per Perl_magic_get |
|
|
275 | invocation (newer linuxes) as opposed to the 0.0002MB perl |
|
|
276 | normally uses. |
|
|
277 | - make stacksize configurable for the ultimate debian experience. |
|
|
278 | |
|
|
279 | 1.51 Mon Dec 12 18:48:36 CET 2005 |
|
|
280 | - remove debugging warn() accidentally left in Coro::Select. |
|
|
281 | |
|
|
282 | 1.5 Tue Nov 29 13:32:44 CET 2005 |
|
|
283 | - use Coro::Event inside Coro::Select to avoid spurious deadlocks. |
|
|
284 | - fix Coro::Select 'select'. |
|
|
285 | - strict'ify some modules. |
|
|
286 | - libcoro errornously restored the SIGUSR2 handler to SIGUSR1. |
|
|
287 | - use XSLoader in selected modules. |
|
|
288 | - remove some 5.6 compatibility cruft. |
|
|
289 | |
|
|
290 | 1.4 Tue Sep 6 00:11:05 CEST 2005 |
|
|
291 | - libcoro did not take into account the trampoline on amd64, |
|
|
292 | when 'l'inux method was used. |
|
|
293 | |
|
|
294 | 1.31 Tue Aug 30 23:31:33 CEST 2005 |
|
|
295 | - some portability fixes/workarounds. |
|
|
296 | |
|
|
297 | 1.3 Sat Aug 20 03:08:56 CEST 2005 |
|
|
298 | - no code changes. module seems to work fine. |
|
|
299 | - improve Coro::State docs, remove reference to nonexisting |
|
|
300 | Coro::State::flush method. |
|
|
301 | - no longer autodetect windows, present it as an option instead. |
|
|
302 | |
|
|
303 | 1.2 Mon May 16 02:00:55 CEST 2005 |
| 3 | - included libcoro.c earlier in State.xs, might improve portability. |
304 | - included libcoro.c earlier in State.xs, might improve portability. |
| 4 | - use a faster and possibly more stable Coro::Cont implementation. |
305 | - use a faster and possibly more stable Coro::Cont implementation. |
| 5 | - accept x86_64 in addition to amd64 for optimized linux-amd64 support. |
306 | - accept x86_64 in addition to amd64 for optimized linux-amd64 support. |
| 6 | - fix bugs Coro::SemaphoreSet that could cause locks never to be freed. |
307 | - fix bugs Coro::SemaphoreSet that could cause locks never to be freed. |
|
|
308 | - fix bugs in CoroAPI.h and document it in Coro::MakeMaker. |
| 7 | |
309 | |
| 8 | 1.11 Thu Mar 3 18:00:52 CET 2005 |
310 | 1.11 Thu Mar 3 18:00:52 CET 2005 |
| 9 | - change of contact address. |
311 | - change of contact address. |
| 10 | |
312 | |
| 11 | 1.1 Tue Feb 22 20:51:16 CET 2005 |
313 | 1.1 Tue Feb 22 20:51:16 CET 2005 |
| 12 | - support [l]inux method on amd64. |
314 | - support [l]inux method on amd64. |
| 13 | - make some context switches while compiling/eval'ing work. |
315 | - allow some context switches while compiling/eval'ing. |
| 14 | |
316 | |
| 15 | 1.0 Tue Aug 10 03:47:27 CEST 2004 |
317 | 1.0 Tue Aug 10 03:47:27 CEST 2004 |
| 16 | - Event 0.89 fixes the bug again. |
318 | - Event 0.89 fixes the bug again. |
| 17 | - fix the remaining(??) memleak. I didn't free the cache-AV for |
319 | - fix the remaining(??) memleak. I didn't free the cache-AV for |
| 18 | padlists, so one AV per created anon-sub leaked. |
320 | padlists, so one AV per created anon-sub leaked. |
