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

Comparing Coro/Changes (file contents):
Revision 1.535 by root, Wed Apr 14 01:56:02 2010 UTC vs.
Revision 1.653 by root, Fri Oct 16 23:42:56 2015 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: unready_all
5TODO: myhttpd header parsing
6TODO: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
7TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 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
8TODO: croak when async_pool tries to run canceled thread?
9
106.49 Sat Oct 17 01:40:12 CEST 2015
11 - throwing an exception to a thread waiting in
12 Coro::Handle using Coro::EV did not stop the watchers,
13 causing the next call to fail (testcase by Martin Pritchard).
14 - bump minimum perl version to 5.10.
15
166.48 Sun Oct 4 19:03:51 CEST 2015
17 - fix memory leak when Coro::Handle uses Coro::EV internally
18 (testcase by Sten Sten).
19 - update code to libev 4 API (internally, Coro still used the
20 version 3 API calls).
21 - remove unused hv_sig.
22
236.47 Sat Jul 11 03:58:20 CEST 2015
24 - swap_svs - NVs can be in the SV head beginning with 5.20.
25
266.46 Tue Jun 30 14:36:30 CEST 2015
27 - restore portability to perls with windows fork emulation
28 (patch by Petr Písař).
29
306.45 Tue Jun 30 01:40:08 CEST 2015
31 - restore portability to perl 5.16 and below.
32
336.44 Tue Jun 30 00:41:54 CEST 2015
34 - avoid segfaulting (or worse) when tracing a canceled thread.
35 - the Event module silently broke it's public hook API, causing
36 failures on perls compiled with -Duselongdouble. Requiring
37 the latest version of Event because I don't know which release
38 changed this (it's not mentioned in the ChangeLog).
39 - allow xs level enterleave hooks via CoroAPI.
40
416.43 Thu Jun 4 15:38:14 CEST 2015
42 - use stability canary.
43 - port to stableperl-5.22.0-1.001.
44 - update libecb.
45
466.42 Wed Feb 11 20:29:52 CET 2015
47 - Coro::SemaphoreSet->try did not actually work (analyzed by
48 SATO Kentaro).
49 - upgrade libecb to be C11 compliant.
50
516.41 Sat Sep 6 22:08:46 CEST 2014
52 - restore portability to perl 5.8.x.
53 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
54 up some memory, but fixes "use feature" and similar modules
55 when used inside a Coro.
56 - allow zero as argument to Coro::Channel to mean the same thing
57 as no argument. this works with older versions as well,
58 but wasn't legal until now.
59 - slightly better c header file detection.
60
616.39 Mon Jun 2 00:00:08 CEST 2014
62 - work around more incompatible changes in 5.20.
63
646.38 Sun Jun 1 21:54:23 CEST 2014
65 - check that perl slots actually have enough space to hold
66 interpreter variables.
67 - untested port to perl 5.19 (64 bit tmps indices) (reported
68 by Andreas König).
69 - croak when cancel is called without a thread context.
70
716.37 Tue Mar 4 13:27:33 CET 2014
72 - *sigh*, removed leftover debugging code from debugging a
73 perl bug, of all things.
74
756.36 Tue Mar 4 07:11:59 CET 2014
76 - semaphores would not clear the destroy hook when interrupted
77 by ->throw, causing segfaults or worse.
78 - ->throw on a thread waiting for a semaphore did not acquire
79 the semaphore, but also didn't wake up other waiters,
80 possibly causing a deadlock.
81 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
82 to report a full stacktrace, which should help find out
83 where the actual call is.
84 - "atomically" destroy data in slf_destroy, because it is
85 the right thing to do, just in case.
86 - disable ecb.h libm dependency, to compile on ancient systems
87 or under adverse conditions.
88
896.33 Mon Nov 18 11:26:27 CET 2013
90 - do not crash when freeing padlists with holes (in 5.18).
91 - tentative SVt_BIND 5.19 port/fix.
92
936.32 Tue Nov 5 15:35:35 CET 2013
94 - use a new algorithm to derive padlists for perl 5.18. The old
95 one could lead to 0-pointer accesses inside perl (reported
96 by Darin McBride).
97
986.31 Thu May 9 07:39:48 CEST 2013
99 - Coro::AIO requests would crash if the thread was ready'd
100 while the request was ongoing.
101
1026.29 Wed May 8 02:55:18 CEST 2013
103 - when an on_destroy handler destructs the coro currently being
104 destructed a perl scalar could be accessed after being freed,
105 likely causing a crash.
106
1076.28 Wed Mar 6 06:58:02 CET 2013
108 - clean remnants of existing __DIE__ and __WARN__ handlers so
109 they lose their magic and will not cause segfaults later
110 (testcase by Andrey Sagulin).
111 - improved Coro::State documentation a bit.
112 - Coro::Debug::command now flushes the output.
113 - add hack detection code for x32 abi, because the braindead slugs
114 who designed that made it look exactly like x86_64 without
115 providing proper compile time symbols to test for it. as a result,
116 this detection cannot work reliably.
117 - valgrind stack registering was broken.
118 - do not rely on Time::HiRes anymore in Coro::Debug.
119
1206.23 Fri Dec 7 23:36:37 CET 2012
121 - use experimental fiber implementation on native windows
122 perls.
123 - use sizeof (void *) as multiplication factor for stack sizes,
124 to accomodate the totally braindamaged microsoft 64 bit "os".
125 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
126 activeperl 5.16 crashes when PerlIO_define_layer is called due
127 to some bug in the perl dll, strawberry perl at least passes
128 the testsuite.
129 - implement Coro::Handle->peeraddr/host/port, for slightly
130 improved compatibility with LWP.
131 - implement 5.17 compatibility by almost blindly applying a
132 good-looking patch by Father Chrysostomos.
133 - move stack management functions into libcoro 3.
134 - libcoro version 3 "released".
135 - support magic values as timed_io_once args.
136 - recommend AnyEvent 7+ or EV 4+, also require EV
137 version 4 or newer for Coro::EV.
138
1396.10 Tue Oct 9 01:14:27 CEST 2012
140 - updated ecb.h, it had a typo that caused it to not compile on many
141 big endian systems (reported by many people).
142 - disable memory fences in ecb.h to improve portability.
143
1446.09 Sat Oct 6 23:25:02 CEST 2012
145 - Coro::EV I/O watchers were not interruptible by exceptions
146 (Coro::State::throw) (testcase by sten).
147 - ->throw now puts threads into the ready queue, as this seems to
148 be expected by existing code, and code that doesn't cope with spurious
149 wakeups needs fixing anyway.
150 - use fd -1 in mmap.
151 - cast I32 to int in error message printf.
152 - warn about broken so-called "hardened" kernels.
153
1546.08 Fri Apr 13 12:05:47 CEST 2012
155 - be more aggressive about exiting like perl does - formerly,
156 exiting from the non-main thread would not execute END blocks.
157
1586.07 Fri Nov 11 21:21:48 CET 2011
159 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
160 - update ecb.h.
161
1626.06 Mon Aug 8 23:59:48 CEST 2011
163 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
164 backend, which is an order of magnitude slower. unfortunately, cygwins
165 pthread implementation isn't very complete either, so allocate the stack
166 twice just to be sure.
167 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
168 in its ithreaded perl - best recompile cygwin and use the 'w'indows
169 backend for much better performance. also disable ithreads for
170 even better performance...).
171
1726.05 Thu Aug 4 21:36:36 CEST 2011
173 - blush, condvar values would not be propagated from send to recv anymore
174 (reported by Chip Salzenberg).
175 - use exponential increase for the readline buffer length in
176 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
177
1786.04 Wed Aug 3 17:15:45 CEST 2011
179 - use even more efficient and more compatible condvars for
180 compatibility to AnyEvent 6.x :)
181 - more inconsequential ecb.h updates.
182
1836.03 Wed Aug 3 11:41:30 CEST 2011
184 - change how Coro patches AnyEvent condvars for compatibility to
185 AnyEvent 6.x.
186 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
187 hasn't been included.
188
1896.02 Wed Jul 13 04:35:19 CEST 2011
190 - "improve portability to Gentoo" - gentoo manages to put perl variables
191 in memory areas that are farther than 2gb apart, which the jit couldn't
192 handle and barfed. now it's just a bit slower on gentoo and similar
193 systems.
194
1956.01 Sun Jul 3 12:31:14 CEST 2011
196 - workarounds are good, but the test for whether pthreads are used
197 was not good. this one should be better.
198 - check differently whether gcc generates cfi instructions itself.
199
2006.0 Wed Jun 29 19:43:35 CEST 2011
201 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
202 destroyed and cleaned up automatically (e.g. async { schedule }).
203 - implement a JIT compiler for part of the thread switch code,
204 which gives a 50% speed improvement on threaded perls, and
205 about 4% on non-threaded perls (so threaded perls now finally
206 reach about half the speed of non-threaded perls).
207 - slightly modernise Coro::Intro, add section about rouse functions.
208 - avoid DEFSV and ERRSV, giving another 10% improvement
209 in thread switching.
210 - Coro::State->is_destroyed is now called is_zombie.
211 - implement a Coro->safe_cancel method that might fail, but
212 cancels in a "safer" way if it succeeds.
213 - add preliminary support for DEBUGGING perls.
214 - get rid of two hash-accesses when initialising a new Coro - this
215 speeds up coro creation by almost a factor of two.
216 - croak when a coro that is being cancelled tries to block
217 (e.g. while executing a guard block), instead of crashing or
218 deadlocking.
219 - use a more robust and also faster method to identify Coro::State
220 objects - speeds up everything a bit.
221 - implement Coro->cancel in XS for a 20% speed improvement, and to
222 be able to implement mutual cancellation.
223 - speed up context switches by a percent or two by more efficiently
224 allocating context stack entries.
225 - implement Coro->join and Coro->on_destroy in XS for a speedup and
226 a reduction in memory use.
227 - cancelling a coro while it itself is cancelling another coro is
228 now supported and working, instead of triggering an assertion.
229 - be a bit more crash-resistant when calling (buggy) on_destroy
230 callbacks (best effort).
231 - move on_destroy into the slf_frame, to allow extension slf
232 functions to have destructors.
233 - get rid if coro refcounting - simply crash in other interpreter
234 threads by nulling the pointers on clone.
235 - simplify warn/die hook handling when loading Coro - the convoluted
236 logic seems to be no longer neccessary.
237 - use libecb instead of our own home-grown gcc hacks.
238 - document alternatives to Coro::LWP. Please use them :)
239 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
240 sigaltstack. Really. wine suffers from it, erlang suffers from it,
241 and it's known since at least 2006.
242
2435.372 Wed Feb 23 06:14:30 CET 2011
244 - apparently mingw doesn't provide a working gettimeofday, try to
245 work around that by relying on Time::HiRes (indirectly brought to
246 my attention by Max Maischein).
247 - fix some portability issues when Time::HiRes was used.
248
2495.371 Mon Feb 21 14:36:08 CET 2011
250 - backport to windows process emulation code again.
251
2525.37 Sat Feb 19 07:49:44 CET 2011
253 - add a big "Coro thread life cycle" section to "man Coro".
254 - try a tentative workaround against the breakage that 5.13 has
255 introduced without depreciation period. sigh.
256 - no longer use Time::HiRes if gettimeofday is available, which
257 saves quite a lot of memory.
258
2595.36 Sun Feb 13 05:33:41 CET 2011
260 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
261 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
262 method call.
263 - undocument Coro::Timer::sleep and obsolete whole module.
264 - optimise Coro::Timer::timeout memory and cpu usage.
265 - slightly updated Coro::Intro for recent changes.
266 - do not initialise PL_dirty anymore.
267
2685.25 Thu Nov 11 01:08:39 CET 2010
269 - try a different approach on netbsd - netbsd 5 finally has marginally
270 working pthreads, but still broken ucontext/sigaltstack.
271 - openbsd 4.8 finally got their act together, Coro works out of the box
272 with asm, setjmp and pthreads (no change, just informational).
273
2745.24 Sat Oct 23 11:27:12 CEST 2010
275 - port to the EV 4.0 API.
276 - work around bugs in mingw32, making strawberry perl work
277 out of the box.
278 - correctly modify Coro::AIO function prototypes
279 so that they reflect the "no optional parameters" rule.
280 - "ported" libcoro to C++.
281
2825.23 Mon May 17 18:50:42 CEST 2010
283 - be more resistant to ordering changes when initialising
284 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
285 Waldorf).
286 - document that perl 5.12 deliberately removed support for cloning.
8 287
95.22 Wed Apr 14 03:55:35 CEST 2010 2885.22 Wed Apr 14 03:55:35 CEST 2010
10 - correctly return udnef on errors in Coro::Handle::read/write 289 - correctly return udnef on errors in Coro::Handle::read/write
11 (testcase by Marc Wims). 290 (testcase by Marc Mims).
12 - convert Coro::Util into a "perl compatibility wrapper" - the functions 291 - convert Coro::Util into a "perl compatibility wrapper" - the functions
13 are less useful now, but are drop-in replacements for existing 292 are less useful now, but are drop-in replacements for existing
14 functions, listing better alternatives in the documentation. This also 293 functions, listing better alternatives in the documentation. This also
15 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton 294 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
16 with Coro::Util::inet_aton. 295 with Coro::Util::inet_aton.
… …
285 - ->throw is now supported on Coro::State objects. 564 - ->throw is now supported on Coro::State objects.
286 - clean up cctx creation code a bit. 565 - clean up cctx creation code a bit.
287 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 566 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
288 567
2894.9 Sat Nov 8 17:45:27 CET 2008 5684.9 Sat Nov 8 17:45:27 CET 2008
290 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 569 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
291 - (libcoro) no longer leak threads in the experimental pthread backend, 570 - (libcoro) no longer leak threads in the experimental pthread backend,
292 also speed it up considerably. 571 also speed it up considerably.
293 - (libcoro) do not rely on makecontext passing void *'s unscathed. 572 - (libcoro) do not rely on makecontext passing void *'s unscathed.
294 - fix compiletime dependencies on libcoro in the Makefile. 573 - fix compiletime dependencies on libcoro in the Makefile.
295 - cctx_count wasn't always updated properly. 574 - cctx_count wasn't always updated properly.
… …
357 - remove debugging code related to MgPV_nolen_const, also try to 636 - remove debugging code related to MgPV_nolen_const, also try to
358 make it compile with perl 5.8.6 (yes, apple apparently loves 637 make it compile with perl 5.8.6 (yes, apple apparently loves
359 outdated software). Reported by John S. 638 outdated software). Reported by John S.
360 639
3614.744 Tue Jul 8 22:06:35 CEST 2008 6404.744 Tue Jul 8 22:06:35 CEST 2008
362 - correctly provide default DIE/WARN handlers as documented. 641 - correctly provide default DIE/WARN handlers as documented.
363 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 642 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
364 implementations inside perl work fine for us. 643 implementations inside perl work fine for us.
365 644
3664.743 Mon Jun 16 00:21:57 CEST 2008 6454.743 Mon Jun 16 00:21:57 CEST 2008
367 - when using Coro::EV without running EV::loop it could 646 - when using Coro::EV without running EV::loop it could
… …
389 - sprinkle "no warnings" freely over everything, also suppress 668 - sprinkle "no warnings" freely over everything, also suppress
390 warnings for some other modules. 669 warnings for some other modules.
391 - fix typo in WSAEWOULDBLOCK. 670 - fix typo in WSAEWOULDBLOCK.
392 671
3934.72 Sun May 25 05:14:36 CEST 2008 6724.72 Sun May 25 05:14:36 CEST 2008
394 - tweak META.yaml a bit, unfortunately, there is no documented way 673 - tweak META.yaml a bit, unfortunately, there is no documented way
395 to have optional dependencies with CPAN. doh :( 674 to have optional dependencies with CPAN. doh :(
396 - avoid running some tests on windows because they would fail due to 675 - avoid running some tests on windows because they would fail due to
397 perl bug (broken fork, broken pipes...). 676 perl bug (broken fork, broken pipes...).
398 - work around perl on windows bugs where perl returns undocumented 677 - work around perl on windows bugs where perl returns undocumented
399 error codes for sysread, syswrite etc. by taking advantage 678 error codes for sysread, syswrite etc. by taking advantage
… …
403 - use unix domain sockets in testsuite to work around 682 - use unix domain sockets in testsuite to work around
404 common perl implementation bugs on widows (they are emulated by 683 common perl implementation bugs on widows (they are emulated by
405 tcp sockets on windows. ugh.) 684 tcp sockets on windows. ugh.)
406 685
4074.71 Sat May 24 20:01:27 CEST 2008 6864.71 Sat May 24 20:01:27 CEST 2008
408 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 687 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
409 - take advantage of async name resolution of AnyEvent::Util. 688 - take advantage of async name resolution of AnyEvent::Util.
410 - work around brutal inet_aton override in Coro::LWP. 689 - work around brutal inet_aton override in Coro::LWP.
411 - take advantage of the readyhook in Coro::EV, for smoother 690 - take advantage of the readyhook in Coro::EV, for smoother
412 scheduling. 691 scheduling.
413 692
4144.7 Sun May 11 00:32:19 CEST 2008 6934.7 Sun May 11 00:32:19 CEST 2008
415 - completely reworked the Coro manpage. 694 - completely reworked the Coro manpage.
416 - added Coro::AnyEvent, generic event loop integration. 695 - added Coro::AnyEvent, generic event loop integration.
417 - implement cancel, ready and kill commands in Coro::Debug. 696 - implement cancel, ready and kill commands in Coro::Debug.
418 - document find_coro in Coro::Debug. 697 - document find_coro in Coro::Debug.
419 - incompatible change: rename has_stack to has_cctx. 698 - incompatible change: rename has_stack to has_cctx.
420 - Coro::AIO and Coro::BDB no longer force event model detection, 699 - Coro::AIO and Coro::BDB no longer force event model detection,
421 use AnyEvent::AIO and AnyEvent::BDB. 700 use AnyEvent::AIO and AnyEvent::BDB.
… …
545 824
5464.11 Thu Oct 11 02:40:24 CEST 2007 8254.11 Thu Oct 11 02:40:24 CEST 2007
547 - port to threaded perls. 826 - port to threaded perls.
548 827
5494.1 Thu Oct 11 02:38:16 CEST 2007 8284.1 Thu Oct 11 02:38:16 CEST 2007
550 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 829 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
551 be local to each coro (see Coro::State). 830 be local to each coro (see Coro::State).
552 - incompatible change: for very deep reasons, cede and cede_notself 831 - incompatible change: for very deep reasons, cede and cede_notself
553 cannot return anything, so nothing will be returned. 832 cannot return anything, so nothing will be returned.
554 - possibly bring back 5.10 compatibility (untested). 833 - possibly bring back 5.10 compatibility (untested).
555 - work around stupid (and wrong) warning on 5.10 :(. 834 - work around stupid (and wrong) warning on 5.10 :(.
… …
627 but stateful semantics. 906 but stateful semantics.
628 - fixed a lot of typos in Coro.pm (patch submitted by David 907 - fixed a lot of typos in Coro.pm (patch submitted by David
629 Steinbrunner, which applied flawlessly). 908 Steinbrunner, which applied flawlessly).
630 909
6313.6 Sat Apr 14 17:13:31 CEST 2007 9103.6 Sat Apr 14 17:13:31 CEST 2007
632 - added some bugfixes to get eg/myhttpd working again. 911 - added some bugfixes to get eg/myhttpd working again.
633 - added Coro::Storable for often-cede'ing freeze/thaw. 912 - added Coro::Storable for often-cede'ing freeze/thaw.
634 - try to do a clean exit when a coroutine calls exit 913 - try to do a clean exit when a coroutine calls exit
635 (EXPERIMENTAL). 914 (EXPERIMENTAL).
636 - got rid of indirect call through _coro_init. 915 - got rid of indirect call through _coro_init.
637 - updated the partly antique examples in eg/ to 916 - updated the partly antique examples in eg/ to

Diff Legend

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