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

Comparing Coro/Changes (file contents):
Revision 1.503 by root, Mon Jun 29 04:30:25 2009 UTC vs.
Revision 1.629 by root, Thu May 1 08:00:50 2014 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::coroaio (perl leaks like hell). 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7TODO: unready_all
8TODO: myhttpd header parsing
9TODO: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6TODO: swap_sv, maybe add scope_swap_sv, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
10 8
11TODO: times per-coroutine 9 - untested port to perl 5.19 (64 bit tmps indices) (reported
10 by Andreas König).
11 - croak when cancel is called without a thread context.
12
136.37 Tue Mar 4 13:27:33 CET 2014
14 - *sigh*, removed leftover debugging code from debugging a
15 perl bug, of all things.
16
176.36 Tue Mar 4 07:11:59 CET 2014
18 - semaphores would not clear the destroy hook when interrupted
19 by ->throw, causing segfaults or worse.
20 - ->throw on a thread waiting for a semaphore did not acquire
21 the semaphore, but also didn't wake up other waiters,
22 possibly causing a deadlock.
23 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
24 to report a full stacktrace, which should help find out
25 where the actual call is.
26 - "atomically" destroy data in slf_destroy, because it is
27 the right thing to do, just in case.
28 - disable ecb.h libm dependency, to compile on ancient systems
29 or under adverse conditions.
30
316.33 Mon Nov 18 11:26:27 CET 2013
32 - do not crash when freeing padlists with holes (in 5.18).
33 - tentative SVt_BIND 5.19 port/fix.
34
356.32 Tue Nov 5 15:35:35 CET 2013
36 - use a new algorithm to derive padlists for perl 5.18. The old
37 one could lead to 0-pointer accesses inside perl (reported
38 by Darin McBride).
39
406.31 Thu May 9 07:39:48 CEST 2013
41 - Coro::AIO requests would crash if the thread was ready'd
42 while the request was ongoing.
43
446.29 Wed May 8 02:55:18 CEST 2013
45 - when an on_destroy handler destructs the coro currently being
46 destructed a perl scalar could be accessed after being freed,
47 likely causing a crash.
48
496.28 Wed Mar 6 06:58:02 CET 2013
50 - clean remnants of existing __DIE__ and __WARN__ handlers so
51 they lose their magic and will not cause segfaults later
52 (testcase by Andrey Sagulin).
53 - improved Coro::State documentation a bit.
54 - Coro::Debug::command now flushes the output.
55 - add hack detection code for x32 abi, because the braindead slugs
56 who designed that made it look exactly like x86_64 without
57 providing proper compile time symbols to test for it. as a result,
58 this detection cannot work reliably.
59 - valgrind stack registering was broken.
60 - do not rely on Time::HiRes anymore in Coro::Debug.
61
626.23 Fri Dec 7 23:36:37 CET 2012
63 - use experimental fiber implementation on native windows
64 perls.
65 - use sizeof (void *) as multiplication factor for stack sizes,
66 to accomodate the totally braindamaged microsoft 64 bit "os".
67 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
68 activeperl 5.16 crashes when PerlIO_define_layer is called due
69 to some bug in the perl dll, strawberry perl at least passes
70 the testsuite.
71 - implement Coro::Handle->peeraddr/host/port, for slightly
72 improved compatibility with LWP.
73 - implement 5.17 compatibility by almost blindly applying a
74 good-looking patch by Father Chrysostomos.
75 - move stack management functions into libcoro 3.
76 - libcoro version 3 "released".
77 - support magic values as timed_io_once args.
78 - recommend AnyEvent 7+ or EV 4+, also require EV
79 version 4 or newer for Coro::EV.
80
816.10 Tue Oct 9 01:14:27 CEST 2012
82 - updated ecb.h, it had a typo that caused it to not compile on many
83 big endian systems (reported by many people).
84 - disable memory fences in ecb.h to improve portability.
85
866.09 Sat Oct 6 23:25:02 CEST 2012
87 - Coro::EV I/O watchers were not interruptible by exceptions
88 (Coro::State::throw) (testcase by sten).
89 - ->throw now puts threads into the ready queue, as this seems to
90 be expected by existing code, and code that doesn't cope with spurious
91 wakeups needs fixing anyway.
92 - use fd -1 in mmap.
93 - cast I32 to int in error message printf.
94 - warn about broken so-called "hardened" kernels.
95
966.08 Fri Apr 13 12:05:47 CEST 2012
97 - be more aggressive about exiting like perl does - formerly,
98 exiting from the non-main thread would not execute END blocks.
99
1006.07 Fri Nov 11 21:21:48 CET 2011
101 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
102 - update ecb.h.
103
1046.06 Mon Aug 8 23:59:48 CEST 2011
105 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
106 backend, which is an order of magnitude slower. unfortunately, cygwins
107 pthread implementation isn't very complete either, so allocate the stack
108 twice just to be sure.
109 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
110 in its ithreaded perl - best recompile cygwin and use the 'w'indows
111 backend for much better performance. also disable ithreads for
112 even better performance...).
113
1146.05 Thu Aug 4 21:36:36 CEST 2011
115 - blush, condvar values would not be propagated from send to recv anymore
116 (reported by Chip Salzenberg).
117 - use exponential increase for the readline buffer length in
118 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
119
1206.04 Wed Aug 3 17:15:45 CEST 2011
121 - use even more efficient and more compatible condvars for
122 compatibility to AnyEvent 6.x :)
123 - more inconsequential ecb.h updates.
124
1256.03 Wed Aug 3 11:41:30 CEST 2011
126 - change how Coro patches AnyEvent condvars for compatibility to
127 AnyEvent 6.x.
128 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
129 hasn't been included.
130
1316.02 Wed Jul 13 04:35:19 CEST 2011
132 - "improve portability to Gentoo" - gentoo manages to put perl variables
133 in memory areas that are farther than 2gb apart, which the jit couldn't
134 handle and barfed. now it's just a bit slower on gentoo and similar
135 systems.
136
1376.01 Sun Jul 3 12:31:14 CEST 2011
138 - workarounds are good, but the test for whether pthreads are used
139 was not good. thisone should be better.
140 - check differently whether gcc generates cfi instructions itself.
141
1426.0 Wed Jun 29 19:43:35 CEST 2011
143 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
144 destroyed and cleaned up automatically (e.g. async { schedule }).
145 - implement a JIT compiler for part of the thread switch code,
146 which gives a 50% speed improvement on threaded perls, and
147 about 4% on non-thraeded perls (so threaded perls now finally
148 reach about half the speed of non-threaded perls).
149 - slightly modernise Coro::Intro, add section about rouse functions.
150 - avoid DEFSV and ERRSV, giving another 10% improvement
151 in thread switching.
152 - Coro::State->is_destroyed is now called is_zombie.
153 - implement a Coro->safe_cancel method that might fail, but
154 cancels in a "safer" way if it succeeds.
155 - add preliminary support for DEBUGGING perls.
156 - get rid of two hash-accesses when initialising a new Coro - this
157 speeds up coro creation by almost a factor of two.
158 - croak when a coro that is being cancelled tries to block
159 (e.g. while executing a guard block), instead of crashing or
160 deadlocking.
161 - use a more robust and also faster method to identify Coro::State
162 objects - speeds up everything a bit.
163 - implement Coro->cancel in XS for a 20% speed improvement, and to
164 be able to implement mutual cancellation.
165 - speed up context switches by a percent or two by more efficiently
166 allocating context stack entries.
167 - implement Coro->join and Coro->on_destroy in XS for a speedup and
168 a reduction in memory use.
169 - cancelling a coro while it itself is cancelling another coro is
170 now supported and working, instead of triggering an assertion.
171 - be a bit more crash-resistant when calling (buggy) on_destroy
172 callbacks (best effort).
173 - move on_destroy into the slf_frame, to allow extension slf
174 functions to have destructors.
175 - get rid if coro refcounting - simply crash in other interpreter
176 threads by nulling the pointers on clone.
177 - simplify warn/die hook handling when loading Coro - the convoluted
178 logic seems to be no longer neccessary.
179 - use libecb instead of our own home-grown gcc hacks.
180 - document alternatives to Coro::LWP. Please use them :)
181 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
182 sigaltstack. Really. wine suffers from it, erlang suffers from it,
183 and it's known since at least 2006.
184
1855.372 Wed Feb 23 06:14:30 CET 2011
186 - apparently mingw doesn't provide a working gettimeofday, try to
187 work around that by relying on Time::HiRes (indirectly brought to
188 my attention by Max Maischein).
189 - fix some portability issues when Time::HiRes was used.
190
1915.371 Mon Feb 21 14:36:08 CET 2011
192 - backport to windows process emulation code again.
193
1945.37 Sat Feb 19 07:49:44 CET 2011
195 - add a big "Coro thread life cycle" section to "man Coro".
196 - try a tentative workaround against the breakage that 5.13 has
197 introduced without depreciation period. sigh.
198 - no longer use Time::HiRes if gettimeofday is available, which
199 saves quite a lot of memory.
200
2015.36 Sun Feb 13 05:33:41 CET 2011
202 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
203 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
204 method call.
205 - undocument Coro::Timer::sleep and obsolete whole module.
206 - optimise Coro::Timer::timeout memory and cpu usage.
207 - slightly updated Coro::Intro for recent changes.
208 - do not initialise PL_dirty anymore.
209
2105.25 Thu Nov 11 01:08:39 CET 2010
211 - try a different approach on netbsd - netbsd 5 finally has marginally
212 working pthreads, but still broken ucontext/sigaltstack.
213 - openbsd 4.8 finally got their act together, Coro works out of the box
214 with asm, setjmp and pthreads (no change, just informational).
215
2165.24 Sat Oct 23 11:27:12 CEST 2010
217 - port to the EV 4.0 API.
218 - work around bugs in mingw32, making strawberry perl work
219 out of the box.
220 - correctly modify Coro::AIO function prototypes
221 so that they reflect the "no optional parameters" rule.
222 - "ported" libcoro to C++.
223
2245.23 Mon May 17 18:50:42 CEST 2010
225 - be more resistant to ordering changes when initialising
226 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
227 Waldorf).
228 - document that perl 5.12 deliberately removed support for cloning.
229
2305.22 Wed Apr 14 03:55:35 CEST 2010
231 - correctly return udnef on errors in Coro::Handle::read/write
232 (testcase by Marc Mims).
233 - convert Coro::Util into a "perl compatibility wrapper" - the functions
234 are less useful now, but are drop-in replacements for existing
235 functions, listing better alternatives in the documentation. This also
236 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
237 with Coro::Util::inet_aton.
238 - do not override $Coro::idle unconditionally in Coro.pm, as other
239 modules could have provided their own idle coro already
240 (for exmaple, Coro::AnyEvent).
241 - fix Coro::Util::gethost* functions.
242 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
243
2445.21 Wed Dec 16 07:19:51 CET 2009
245 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
246 together.
247 - add some examples on how to combine other event loops with Coro in
248 Coro::AnyEvent, and how to run it (and not to block). Seems to be
249 the most common source of confusion.
250 - try to catch people naively blocking in an event callback.
251 - work around the perl filehandle bug issue in conjunction with
252 older common::sense (as indirectly pointed out by ZSystem).
253 - clarify the "not from signal handlers" section.
254
2555.2 Sun Oct 4 14:54:24 CEST 2009
256 - Coro::Storable destroyed the prototypes of the functions it wrapped.
257 - export rouse_cb and rouse_wait by default now.
258 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
259 - new method $state->swap_sv.
260 - added section on "windows process emulation" to the manpage,
261 after a not-so-fruitful (nor-friendly) "discussion" with chip
262 salzenberg (discussion implies arguments, but his only arguments
263 were ad-hominems, one wonders why he started it in the first
264 place). I hope this explains it well enough for him to understand,
265 and maybe well enough for others to understand.
266 - use common::sense everywhere now.
267 - idle callbacks are no longer supported, use idle coros instead.
268 - print a thread listing when a deadlock is detected.
269
2705.17 Sat Aug 22 23:09:31 CEST 2009
271 - work around a bug in the perl debugger causing crashes
272 when running under the debugger by marking _pool_handler as nodebug.
273 - speed up Coro::async considerably.
274 - try some hacks to get netbsd to work "more often" - their broken
275 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
276 - convert Coro to AE 5.0 API.
277
2785.162 Tue Jul 28 04:04:03 CEST 2009
279 - perl 5.8.2 is now minimum requirement.
280 - skip t/19_handle.t on broken windows perls.
281
2825.161 Wed Jul 22 04:47:38 CEST 2009
283 - Coro::AnyEvent::poll could have a different prototype when EV was
284 used as backend (analyzed by Tatsuhiko Miyagawa).
285 - Coro::AnyEvent errornously initialised the event loop when loaded,
286 not on demand.
287 - try to workaround rare */t/01_unblock.t failures.
288
2895.16 Tue Jul 21 01:44:37 CEST 2009
290 - Coro::AnyEvent failed to hook into the event loop
291 when no threads had been readied between detecting
292 the event loop and actually running it.
293 - considerably speed up Coro::Select by taking avdantage
294 of AnyEvent > 4.8 and some other optimisations.
295 - implement paragraph readline mode in Coro::Handle
296 (based on patches by Zsbán Ambrus).
297 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
298 and analyzed by Yasuhiro MATSUMOTO).
299 - clarified libcoro license and copyright.
300 - someone stole my EXTRA_META!!!
301 - implement Coro::Select::patch_pp_sselect and it's brother,
302 for hardcode select overriding.
303
3045.151 Mon Jul 6 05:41:57 CEST 2009
305 - backport to windows process emulation code again (patch by
306 Yasuhiro MATSUMOTO).
307 - slightly update Coro::MakeMaker.
308
3095.15 Tue Jun 30 10:28:06 CEST 2009
12 - deprecate Coro::Socket, document how to get ipv6 support via 310 - deprecate Coro::Socket, document how to get ipv6 support via
13 AnyEvent::Socket instead. 311 AnyEvent::Socket instead.
14 - implement signal->wait ($cb) interface, similar to semaphores. 312 - implement signal->wait ($cb) interface, similar to semaphores.
15 - work around SvOK not supporting getmagic, so we have to getmagic 313 - work around SvOK not supporting getmagic, so we have to getmagic
16 to test for undef :/ (reported by Matthias Waldorf). 314 to test for undef :/ (reported by Matthias Waldorf).
17 - load Coro::AnyEvent in all modules using AnyEvent. 315 - load Coro::AnyEvent in all modules using AnyEvent.
18 - work around perl corrupting our internal data structures, 316 - work around perl corrupting our internal data structures,
19 reported by Tokuhiro Matsuno. 317 reported by Tokuhiro Matsuno.
318 - enable per-coroutine real and cpu time gathering
319 (Coro::State::enable_times).
20 320
215.14 Wed Jun 24 01:37:48 CEST 2009 3215.14 Wed Jun 24 01:37:48 CEST 2009
22 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent, 322 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
23 as this seems to have been a difficult concept (poll, sleep, idle, 323 as this seems to have been a difficult concept (poll, sleep, idle,
24 idle_upto). 324 idle_upto).
… …
206 - ->throw is now supported on Coro::State objects. 506 - ->throw is now supported on Coro::State objects.
207 - clean up cctx creation code a bit. 507 - clean up cctx creation code a bit.
208 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 508 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
209 509
2104.9 Sat Nov 8 17:45:27 CET 2008 5104.9 Sat Nov 8 17:45:27 CET 2008
211 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 511 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
212 - (libcoro) no longer leak threads in the experimental pthread backend, 512 - (libcoro) no longer leak threads in the experimental pthread backend,
213 also speed it up considerably. 513 also speed it up considerably.
214 - (libcoro) do not rely on makecontext passing void *'s unscathed. 514 - (libcoro) do not rely on makecontext passing void *'s unscathed.
215 - fix compiletime dependencies on libcoro in the Makefile. 515 - fix compiletime dependencies on libcoro in the Makefile.
216 - cctx_count wasn't always updated properly. 516 - cctx_count wasn't always updated properly.
… …
278 - remove debugging code related to MgPV_nolen_const, also try to 578 - remove debugging code related to MgPV_nolen_const, also try to
279 make it compile with perl 5.8.6 (yes, apple apparently loves 579 make it compile with perl 5.8.6 (yes, apple apparently loves
280 outdated software). Reported by John S. 580 outdated software). Reported by John S.
281 581
2824.744 Tue Jul 8 22:06:35 CEST 2008 5824.744 Tue Jul 8 22:06:35 CEST 2008
283 - correctly provide default DIE/WARN handlers as documented. 583 - correctly provide default DIE/WARN handlers as documented.
284 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 584 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
285 implementations inside perl work fine for us. 585 implementations inside perl work fine for us.
286 586
2874.743 Mon Jun 16 00:21:57 CEST 2008 5874.743 Mon Jun 16 00:21:57 CEST 2008
288 - when using Coro::EV without running EV::loop it could 588 - when using Coro::EV without running EV::loop it could
… …
310 - sprinkle "no warnings" freely over everything, also suppress 610 - sprinkle "no warnings" freely over everything, also suppress
311 warnings for some other modules. 611 warnings for some other modules.
312 - fix typo in WSAEWOULDBLOCK. 612 - fix typo in WSAEWOULDBLOCK.
313 613
3144.72 Sun May 25 05:14:36 CEST 2008 6144.72 Sun May 25 05:14:36 CEST 2008
315 - tweak META.yaml a bit, unfortunately, there is no documented way 615 - tweak META.yaml a bit, unfortunately, there is no documented way
316 to have optional dependencies with CPAN. doh :( 616 to have optional dependencies with CPAN. doh :(
317 - avoid running some tests on windows because they would fail due to 617 - avoid running some tests on windows because they would fail due to
318 perl bug (broken fork, broken pipes...). 618 perl bug (broken fork, broken pipes...).
319 - work around perl on windows bugs where perl returns undocumented 619 - work around perl on windows bugs where perl returns undocumented
320 error codes for sysread, syswrite etc. by taking advantage 620 error codes for sysread, syswrite etc. by taking advantage
… …
324 - use unix domain sockets in testsuite to work around 624 - use unix domain sockets in testsuite to work around
325 common perl implementation bugs on widows (they are emulated by 625 common perl implementation bugs on widows (they are emulated by
326 tcp sockets on windows. ugh.) 626 tcp sockets on windows. ugh.)
327 627
3284.71 Sat May 24 20:01:27 CEST 2008 6284.71 Sat May 24 20:01:27 CEST 2008
329 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 629 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
330 - take advantage of async name resolution of AnyEvent::Util. 630 - take advantage of async name resolution of AnyEvent::Util.
331 - work around brutal inet_aton override in Coro::LWP. 631 - work around brutal inet_aton override in Coro::LWP.
332 - take advantage of the readyhook in Coro::EV, for smoother 632 - take advantage of the readyhook in Coro::EV, for smoother
333 scheduling. 633 scheduling.
334 634
3354.7 Sun May 11 00:32:19 CEST 2008 6354.7 Sun May 11 00:32:19 CEST 2008
336 - completely reworked the Coro manpage. 636 - completely reworked the Coro manpage.
337 - added Coro::AnyEvent, generic event loop integration. 637 - added Coro::AnyEvent, generic event loop integration.
338 - implement cancel, ready and kill commands in Coro::Debug. 638 - implement cancel, ready and kill commands in Coro::Debug.
339 - document find_coro in Coro::Debug. 639 - document find_coro in Coro::Debug.
340 - incompatible change: rename has_stack to has_cctx. 640 - incompatible change: rename has_stack to has_cctx.
341 - Coro::AIO and Coro::BDB no longer force event model detection, 641 - Coro::AIO and Coro::BDB no longer force event model detection,
342 use AnyEvent::AIO and AnyEvent::BDB. 642 use AnyEvent::AIO and AnyEvent::BDB.
… …
466 766
4674.11 Thu Oct 11 02:40:24 CEST 2007 7674.11 Thu Oct 11 02:40:24 CEST 2007
468 - port to threaded perls. 768 - port to threaded perls.
469 769
4704.1 Thu Oct 11 02:38:16 CEST 2007 7704.1 Thu Oct 11 02:38:16 CEST 2007
471 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 771 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
472 be local to each coro (see Coro::State). 772 be local to each coro (see Coro::State).
473 - incompatible change: for very deep reasons, cede and cede_notself 773 - incompatible change: for very deep reasons, cede and cede_notself
474 cannot return anything, so nothing will be returned. 774 cannot return anything, so nothing will be returned.
475 - possibly bring back 5.10 compatibility (untested). 775 - possibly bring back 5.10 compatibility (untested).
476 - work around stupid (and wrong) warning on 5.10 :(. 776 - work around stupid (and wrong) warning on 5.10 :(.
… …
548 but stateful semantics. 848 but stateful semantics.
549 - fixed a lot of typos in Coro.pm (patch submitted by David 849 - fixed a lot of typos in Coro.pm (patch submitted by David
550 Steinbrunner, which applied flawlessly). 850 Steinbrunner, which applied flawlessly).
551 851
5523.6 Sat Apr 14 17:13:31 CEST 2007 8523.6 Sat Apr 14 17:13:31 CEST 2007
553 - added some bugfixes to get eg/myhttpd working again. 853 - added some bugfixes to get eg/myhttpd working again.
554 - added Coro::Storable for often-cede'ing freeze/thaw. 854 - added Coro::Storable for often-cede'ing freeze/thaw.
555 - try to do a clean exit when a coroutine calls exit 855 - try to do a clean exit when a coroutine calls exit
556 (EXPERIMENTAL). 856 (EXPERIMENTAL).
557 - got rid of indirect call through _coro_init. 857 - got rid of indirect call through _coro_init.
558 - updated the partly antique examples in eg/ to 858 - updated the partly antique examples in eg/ to

Diff Legend

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