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

Comparing Coro/Changes (file contents):
Revision 1.531 by root, Sat Dec 12 01:30:26 2009 UTC vs.
Revision 1.623 by root, Sat Jan 25 16:25:27 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
115.21 9TODO: warnhook reports from wrong line number now, when arg is Math::GMP number.
10 - FATAL: $Coro::IDLE blocked itself will now use Carp::confess
11 to report a full stacktrace, which should help find out
12 where the actual call is.
13
146.33 Mon Nov 18 11:26:27 CET 2013
15 - do not crash when freeing padlists with holes (in 5.18).
16 - tentative SVt_BIND 5.19 port/fix.
17
186.32 Tue Nov 5 15:35:35 CET 2013
19 - use a new algorithm to derive padlists for perl 5.18. The old
20 one could lead to 0-pointer accesses inside perl (reported
21 by Darin McBride).
22
236.31 Thu May 9 07:39:48 CEST 2013
24 - Coro::AIO requests would crash if the thread was ready'd
25 while the request was ongoing.
26
276.29 Wed May 8 02:55:18 CEST 2013
28 - when an on_destroy handler destructs the coro currently being
29 destructed a perl scalar could be accessed after being freed,
30 likely causing a crash.
31
326.28 Wed Mar 6 06:58:02 CET 2013
33 - clean remnants of existing __DIE__ and __WARN__ handlers so
34 they lose their magic and will not cause segfaults later
35 (testcase by Andrey Sagulin).
36 - improved Coro::State documentation a bit.
37 - Coro::Debug::command now flushes the output.
38 - add hack detection code for x32 abi, because the braindead slugs
39 who designed that made it look exactly like x86_64 without
40 providing proper compile time symbols to test for it. as a result,
41 this detection cannot work reliably.
42 - valgrind stack registering was broken.
43 - do not rely on Time::HiRes anymore in Coro::Debug.
44
456.23 Fri Dec 7 23:36:37 CET 2012
46 - use experimental fiber implementation on native windows
47 perls.
48 - use sizeof (void *) as multiplication factor for stack sizes,
49 to accomodate the totally braindamaged microsoft 64 bit "os".
50 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
51 activeperl 5.16 crashes when PerlIO_define_layer is called due
52 to some bug in the perl dll, strawberry perl at least passes
53 the testsuite.
54 - implement Coro::Handle->peeraddr/host/port, for slightly
55 improved compatibility with LWP.
56 - implement 5.17 compatibility by almost blindly applying a
57 good-looking patch by Father Chrysostomos.
58 - move stack management functions into libcoro 3.
59 - libcoro version 3 "released".
60 - support magic values as timed_io_once args.
61 - recommend AnyEvent 7+ or EV 4+, also require EV
62 version 4 or newer for Coro::EV.
63
646.10 Tue Oct 9 01:14:27 CEST 2012
65 - updated ecb.h, it had a typo that caused it to not compile on many
66 big endian systems (reported by many people).
67 - disable memory fences in ecb.h to improve portability.
68
696.09 Sat Oct 6 23:25:02 CEST 2012
70 - Coro::EV I/O watchers were not interruptible by exceptions
71 (Coro::State::throw) (testcase by sten).
72 - ->throw now puts threads into the ready queue, as this seems to
73 be expected by existing code, and code that doesn't cope with spurious
74 wakeups needs fixing anyway.
75 - use fd -1 in mmap.
76 - cast I32 to int in error message printf.
77 - warn about broken so-called "hardened" kernels.
78
796.08 Fri Apr 13 12:05:47 CEST 2012
80 - be more aggressive about exiting like perl does - formerly,
81 exiting from the non-main thread would not execute END blocks.
82
836.07 Fri Nov 11 21:21:48 CET 2011
84 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
85 - update ecb.h.
86
876.06 Mon Aug 8 23:59:48 CEST 2011
88 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
89 backend, which is an order of magnitude slower. unfortunately, cygwins
90 pthread implementation isn't very complete either, so allocate the stack
91 twice just to be sure.
92 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
93 in its ithreaded perl - best recompile cygwin and use the 'w'indows
94 backend for much better performance. also disable ithreads for
95 even better performance...).
96
976.05 Thu Aug 4 21:36:36 CEST 2011
98 - blush, condvar values would not be propagated from send to recv anymore
99 (reported by Chip Salzenberg).
100 - use exponential increase for the readline buffer length in
101 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
102
1036.04 Wed Aug 3 17:15:45 CEST 2011
104 - use even more efficient and more compatible condvars for
105 compatibility to AnyEvent 6.x :)
106 - more inconsequential ecb.h updates.
107
1086.03 Wed Aug 3 11:41:30 CEST 2011
109 - change how Coro patches AnyEvent condvars for compatibility to
110 AnyEvent 6.x.
111 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
112 hasn't been included.
113
1146.02 Wed Jul 13 04:35:19 CEST 2011
115 - "improve portability to Gentoo" - gentoo manages to put perl variables
116 in memory areas that are farther than 2gb apart, which the jit couldn't
117 handle and barfed. now it's just a bit slower on gentoo and similar
118 systems.
119
1206.01 Sun Jul 3 12:31:14 CEST 2011
121 - workarounds are good, but the test for whether pthreads are used
122 was not good. thisone should be better.
123 - check differently whether gcc generates cfi instructions itself.
124
1256.0 Wed Jun 29 19:43:35 CEST 2011
126 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
127 destroyed and cleaned up automatically (e.g. async { schedule }).
128 - implement a JIT compiler for part of the thread switch code,
129 which gives a 50% speed improvement on threaded perls, and
130 about 4% on non-thraeded perls (so threaded perls now finally
131 reach about half the speed of non-threaded perls).
132 - slightly modernise Coro::Intro, add section about rouse functions.
133 - avoid DEFSV and ERRSV, giving another 10% improvement
134 in thread switching.
135 - Coro::State->is_destroyed is now called is_zombie.
136 - implement a Coro->safe_cancel method that might fail, but
137 cancels in a "safer" way if it succeeds.
138 - add preliminary support for DEBUGGING perls.
139 - get rid of two hash-accesses when initialising a new Coro - this
140 speeds up coro creation by almost a factor of two.
141 - croak when a coro that is being cancelled tries to block
142 (e.g. while executing a guard block), instead of crashing or
143 deadlocking.
144 - use a more robust and also faster method to identify Coro::State
145 objects - speeds up everything a bit.
146 - implement Coro->cancel in XS for a 20% speed improvement, and to
147 be able to implement mutual cancellation.
148 - speed up context switches by a percent or two by more efficiently
149 allocating context stack entries.
150 - implement Coro->join and Coro->on_destroy in XS for a speedup and
151 a reduction in memory use.
152 - cancelling a coro while it itself is cancelling another coro is
153 now supported and working, instead of triggering an assertion.
154 - be a bit more crash-resistant when calling (buggy) on_destroy
155 callbacks (best effort).
156 - move on_destroy into the slf_frame, to allow extension slf
157 functions to have destructors.
158 - get rid if coro refcounting - simply crash in other interpreter
159 threads by nulling the pointers on clone.
160 - simplify warn/die hook handling when loading Coro - the convoluted
161 logic seems to be no longer neccessary.
162 - use libecb instead of our own home-grown gcc hacks.
163 - document alternatives to Coro::LWP. Please use them :)
164 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
165 sigaltstack. Really. wine suffers from it, erlang suffers from it,
166 and it's known since at least 2006.
167
1685.372 Wed Feb 23 06:14:30 CET 2011
169 - apparently mingw doesn't provide a working gettimeofday, try to
170 work around that by relying on Time::HiRes (indirectly brought to
171 my attention by Max Maischein).
172 - fix some portability issues when Time::HiRes was used.
173
1745.371 Mon Feb 21 14:36:08 CET 2011
175 - backport to windows process emulation code again.
176
1775.37 Sat Feb 19 07:49:44 CET 2011
178 - add a big "Coro thread life cycle" section to "man Coro".
179 - try a tentative workaround against the breakage that 5.13 has
180 introduced without depreciation period. sigh.
181 - no longer use Time::HiRes if gettimeofday is available, which
182 saves quite a lot of memory.
183
1845.36 Sun Feb 13 05:33:41 CET 2011
185 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
186 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
187 method call.
188 - undocument Coro::Timer::sleep and obsolete whole module.
189 - optimise Coro::Timer::timeout memory and cpu usage.
190 - slightly updated Coro::Intro for recent changes.
191 - do not initialise PL_dirty anymore.
192
1935.25 Thu Nov 11 01:08:39 CET 2010
194 - try a different approach on netbsd - netbsd 5 finally has marginally
195 working pthreads, but still broken ucontext/sigaltstack.
196 - openbsd 4.8 finally got their act together, Coro works out of the box
197 with asm, setjmp and pthreads (no change, just informational).
198
1995.24 Sat Oct 23 11:27:12 CEST 2010
200 - port to the EV 4.0 API.
201 - work around bugs in mingw32, making strawberry perl work
202 out of the box.
203 - correctly modify Coro::AIO function prototypes
204 so that they reflect the "no optional parameters" rule.
205 - "ported" libcoro to C++.
206
2075.23 Mon May 17 18:50:42 CEST 2010
208 - be more resistant to ordering changes when initialising
209 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
210 Waldorf).
211 - document that perl 5.12 deliberately removed support for cloning.
212
2135.22 Wed Apr 14 03:55:35 CEST 2010
214 - correctly return udnef on errors in Coro::Handle::read/write
215 (testcase by Marc Mims).
216 - convert Coro::Util into a "perl compatibility wrapper" - the functions
217 are less useful now, but are drop-in replacements for existing
218 functions, listing better alternatives in the documentation. This also
219 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
220 with Coro::Util::inet_aton.
221 - do not override $Coro::idle unconditionally in Coro.pm, as other
222 modules could have provided their own idle coro already
223 (for exmaple, Coro::AnyEvent).
224 - fix Coro::Util::gethost* functions.
225 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
226
2275.21 Wed Dec 16 07:19:51 CET 2009
12 - automatically load Coro::AnyEvent when AnyEvent and Coro are used 228 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
13 together. 229 together.
14 - add some examples on how to combine other event loops with Coro in 230 - add some examples on how to combine other event loops with Coro in
15 Coro::AnyEvent, and how to run it (and not to block). Seems to be 231 Coro::AnyEvent, and how to run it (and not to block). Seems to be
16 the most common source of confusion. 232 the most common source of confusion.
… …
273 - ->throw is now supported on Coro::State objects. 489 - ->throw is now supported on Coro::State objects.
274 - clean up cctx creation code a bit. 490 - clean up cctx creation code a bit.
275 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 491 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
276 492
2774.9 Sat Nov 8 17:45:27 CET 2008 4934.9 Sat Nov 8 17:45:27 CET 2008
278 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 494 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
279 - (libcoro) no longer leak threads in the experimental pthread backend, 495 - (libcoro) no longer leak threads in the experimental pthread backend,
280 also speed it up considerably. 496 also speed it up considerably.
281 - (libcoro) do not rely on makecontext passing void *'s unscathed. 497 - (libcoro) do not rely on makecontext passing void *'s unscathed.
282 - fix compiletime dependencies on libcoro in the Makefile. 498 - fix compiletime dependencies on libcoro in the Makefile.
283 - cctx_count wasn't always updated properly. 499 - cctx_count wasn't always updated properly.
… …
345 - remove debugging code related to MgPV_nolen_const, also try to 561 - remove debugging code related to MgPV_nolen_const, also try to
346 make it compile with perl 5.8.6 (yes, apple apparently loves 562 make it compile with perl 5.8.6 (yes, apple apparently loves
347 outdated software). Reported by John S. 563 outdated software). Reported by John S.
348 564
3494.744 Tue Jul 8 22:06:35 CEST 2008 5654.744 Tue Jul 8 22:06:35 CEST 2008
350 - correctly provide default DIE/WARN handlers as documented. 566 - correctly provide default DIE/WARN handlers as documented.
351 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 567 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
352 implementations inside perl work fine for us. 568 implementations inside perl work fine for us.
353 569
3544.743 Mon Jun 16 00:21:57 CEST 2008 5704.743 Mon Jun 16 00:21:57 CEST 2008
355 - when using Coro::EV without running EV::loop it could 571 - when using Coro::EV without running EV::loop it could
… …
377 - sprinkle "no warnings" freely over everything, also suppress 593 - sprinkle "no warnings" freely over everything, also suppress
378 warnings for some other modules. 594 warnings for some other modules.
379 - fix typo in WSAEWOULDBLOCK. 595 - fix typo in WSAEWOULDBLOCK.
380 596
3814.72 Sun May 25 05:14:36 CEST 2008 5974.72 Sun May 25 05:14:36 CEST 2008
382 - tweak META.yaml a bit, unfortunately, there is no documented way 598 - tweak META.yaml a bit, unfortunately, there is no documented way
383 to have optional dependencies with CPAN. doh :( 599 to have optional dependencies with CPAN. doh :(
384 - avoid running some tests on windows because they would fail due to 600 - avoid running some tests on windows because they would fail due to
385 perl bug (broken fork, broken pipes...). 601 perl bug (broken fork, broken pipes...).
386 - work around perl on windows bugs where perl returns undocumented 602 - work around perl on windows bugs where perl returns undocumented
387 error codes for sysread, syswrite etc. by taking advantage 603 error codes for sysread, syswrite etc. by taking advantage
… …
391 - use unix domain sockets in testsuite to work around 607 - use unix domain sockets in testsuite to work around
392 common perl implementation bugs on widows (they are emulated by 608 common perl implementation bugs on widows (they are emulated by
393 tcp sockets on windows. ugh.) 609 tcp sockets on windows. ugh.)
394 610
3954.71 Sat May 24 20:01:27 CEST 2008 6114.71 Sat May 24 20:01:27 CEST 2008
396 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 612 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
397 - take advantage of async name resolution of AnyEvent::Util. 613 - take advantage of async name resolution of AnyEvent::Util.
398 - work around brutal inet_aton override in Coro::LWP. 614 - work around brutal inet_aton override in Coro::LWP.
399 - take advantage of the readyhook in Coro::EV, for smoother 615 - take advantage of the readyhook in Coro::EV, for smoother
400 scheduling. 616 scheduling.
401 617
4024.7 Sun May 11 00:32:19 CEST 2008 6184.7 Sun May 11 00:32:19 CEST 2008
403 - completely reworked the Coro manpage. 619 - completely reworked the Coro manpage.
404 - added Coro::AnyEvent, generic event loop integration. 620 - added Coro::AnyEvent, generic event loop integration.
405 - implement cancel, ready and kill commands in Coro::Debug. 621 - implement cancel, ready and kill commands in Coro::Debug.
406 - document find_coro in Coro::Debug. 622 - document find_coro in Coro::Debug.
407 - incompatible change: rename has_stack to has_cctx. 623 - incompatible change: rename has_stack to has_cctx.
408 - Coro::AIO and Coro::BDB no longer force event model detection, 624 - Coro::AIO and Coro::BDB no longer force event model detection,
409 use AnyEvent::AIO and AnyEvent::BDB. 625 use AnyEvent::AIO and AnyEvent::BDB.
… …
533 749
5344.11 Thu Oct 11 02:40:24 CEST 2007 7504.11 Thu Oct 11 02:40:24 CEST 2007
535 - port to threaded perls. 751 - port to threaded perls.
536 752
5374.1 Thu Oct 11 02:38:16 CEST 2007 7534.1 Thu Oct 11 02:38:16 CEST 2007
538 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 754 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
539 be local to each coro (see Coro::State). 755 be local to each coro (see Coro::State).
540 - incompatible change: for very deep reasons, cede and cede_notself 756 - incompatible change: for very deep reasons, cede and cede_notself
541 cannot return anything, so nothing will be returned. 757 cannot return anything, so nothing will be returned.
542 - possibly bring back 5.10 compatibility (untested). 758 - possibly bring back 5.10 compatibility (untested).
543 - work around stupid (and wrong) warning on 5.10 :(. 759 - work around stupid (and wrong) warning on 5.10 :(.
… …
615 but stateful semantics. 831 but stateful semantics.
616 - fixed a lot of typos in Coro.pm (patch submitted by David 832 - fixed a lot of typos in Coro.pm (patch submitted by David
617 Steinbrunner, which applied flawlessly). 833 Steinbrunner, which applied flawlessly).
618 834
6193.6 Sat Apr 14 17:13:31 CEST 2007 8353.6 Sat Apr 14 17:13:31 CEST 2007
620 - added some bugfixes to get eg/myhttpd working again. 836 - added some bugfixes to get eg/myhttpd working again.
621 - added Coro::Storable for often-cede'ing freeze/thaw. 837 - added Coro::Storable for often-cede'ing freeze/thaw.
622 - try to do a clean exit when a coroutine calls exit 838 - try to do a clean exit when a coroutine calls exit
623 (EXPERIMENTAL). 839 (EXPERIMENTAL).
624 - got rid of indirect call through _coro_init. 840 - got rid of indirect call through _coro_init.
625 - updated the partly antique examples in eg/ to 841 - updated the partly antique examples in eg/ to

Diff Legend

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