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

Comparing Coro/Changes (file contents):
Revision 1.529 by root, Tue Nov 24 06:13:01 2009 UTC vs.
Revision 1.659 by root, Wed Jun 22 22:53:49 2016 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?
7TODO: croak when async_pool tries to run canceled thread?
10 8
9 - since various broken or deficient patches float around
10 to work around the vtbl API breakage in perl, let's provide
11 out own workaround, which seems to be much less invasive
12 and more compatible than the approaches seen so far. YMMV.
13 - work around assertion failure in perl_destruct on 5.24
14 (analyzed by Dave Mitchell).
15
166.5 Wed Jun 22 22:23:50 CEST 2016
17 - swap_sv swapping order was not symmetrical, causing
18 wrong swaps when swap_sv was used multiple times on the
19 same sv in the same thread.
20 - swap_sv calls can now be undone by calling it again
21 with the same variables.
22 - swap_sv calls will now be undone in async_pool threads.
23 - split Coro::Semaphore::up/adjust into separate xs functions
24 for better error reporting, at slight codesize increase.
25 - (libcoro) arm assembly support, please test and report.
26 - adjust to PL_savestack changes in perl 5.24 (adapted from
27 the debian patch, which unfortunately gets it wrong).
28
296.49 Sat Oct 17 01:40:12 CEST 2015
30 - throwing an exception to a thread waiting in
31 Coro::Handle using Coro::EV did not stop the watchers,
32 causing the next call to fail (testcase by Martin Pritchard).
33 - bump minimum perl version to 5.10.
34
356.48 Sun Oct 4 19:03:51 CEST 2015
36 - fix memory leak when Coro::Handle uses Coro::EV internally
37 (testcase by Sten Sten).
38 - update code to libev 4 API (internally, Coro still used the
39 version 3 API calls).
40 - remove unused hv_sig.
41
426.47 Sat Jul 11 03:58:20 CEST 2015
43 - swap_svs - NVs can be in the SV head beginning with 5.20.
44
456.46 Tue Jun 30 14:36:30 CEST 2015
46 - restore portability to perls with windows fork emulation
47 (patch by Petr Písař).
48
496.45 Tue Jun 30 01:40:08 CEST 2015
50 - restore portability to perl 5.16 and below.
51
526.44 Tue Jun 30 00:41:54 CEST 2015
53 - avoid segfaulting (or worse) when tracing a canceled thread.
54 - the Event module silently broke it's public hook API, causing
55 failures on perls compiled with -Duselongdouble. Requiring
56 the latest version of Event because I don't know which release
57 changed this (it's not mentioned in the ChangeLog).
58 - allow xs level enterleave hooks via CoroAPI.
59
606.43 Thu Jun 4 15:38:14 CEST 2015
61 - use stability canary.
62 - port to stableperl-5.22.0-1.001.
63 - update libecb.
64
656.42 Wed Feb 11 20:29:52 CET 2015
66 - Coro::SemaphoreSet->try did not actually work (analyzed by
67 SATO Kentaro).
68 - upgrade libecb to be C11 compliant.
69
706.41 Sat Sep 6 22:08:46 CEST 2014
71 - restore portability to perl 5.8.x.
72 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
73 up some memory, but fixes "use feature" and similar modules
74 when used inside a Coro.
75 - allow zero as argument to Coro::Channel to mean the same thing
76 as no argument. this works with older versions as well,
77 but wasn't legal until now.
78 - slightly better c header file detection.
79
806.39 Mon Jun 2 00:00:08 CEST 2014
81 - work around more incompatible changes in 5.20.
82
836.38 Sun Jun 1 21:54:23 CEST 2014
84 - check that perl slots actually have enough space to hold
85 interpreter variables.
86 - untested port to perl 5.19 (64 bit tmps indices) (reported
87 by Andreas König).
88 - croak when cancel is called without a thread context.
89
906.37 Tue Mar 4 13:27:33 CET 2014
91 - *sigh*, removed leftover debugging code from debugging a
92 perl bug, of all things.
93
946.36 Tue Mar 4 07:11:59 CET 2014
95 - semaphores would not clear the destroy hook when interrupted
96 by ->throw, causing segfaults or worse.
97 - ->throw on a thread waiting for a semaphore did not acquire
98 the semaphore, but also didn't wake up other waiters,
99 possibly causing a deadlock.
100 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
101 to report a full stacktrace, which should help find out
102 where the actual call is.
103 - "atomically" destroy data in slf_destroy, because it is
104 the right thing to do, just in case.
105 - disable ecb.h libm dependency, to compile on ancient systems
106 or under adverse conditions.
107
1086.33 Mon Nov 18 11:26:27 CET 2013
109 - do not crash when freeing padlists with holes (in 5.18).
110 - tentative SVt_BIND 5.19 port/fix.
111
1126.32 Tue Nov 5 15:35:35 CET 2013
113 - use a new algorithm to derive padlists for perl 5.18. The old
114 one could lead to 0-pointer accesses inside perl (reported
115 by Darin McBride).
116
1176.31 Thu May 9 07:39:48 CEST 2013
118 - Coro::AIO requests would crash if the thread was ready'd
119 while the request was ongoing.
120
1216.29 Wed May 8 02:55:18 CEST 2013
122 - when an on_destroy handler destructs the coro currently being
123 destructed a perl scalar could be accessed after being freed,
124 likely causing a crash.
125
1266.28 Wed Mar 6 06:58:02 CET 2013
127 - clean remnants of existing __DIE__ and __WARN__ handlers so
128 they lose their magic and will not cause segfaults later
129 (testcase by Andrey Sagulin).
130 - improved Coro::State documentation a bit.
131 - Coro::Debug::command now flushes the output.
132 - add hack detection code for x32 abi, because the braindead slugs
133 who designed that made it look exactly like x86_64 without
134 providing proper compile time symbols to test for it. as a result,
135 this detection cannot work reliably.
136 - valgrind stack registering was broken.
137 - do not rely on Time::HiRes anymore in Coro::Debug.
138
1396.23 Fri Dec 7 23:36:37 CET 2012
140 - use experimental fiber implementation on native windows
141 perls.
142 - use sizeof (void *) as multiplication factor for stack sizes,
143 to accomodate the totally braindamaged microsoft 64 bit "os".
144 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
145 activeperl 5.16 crashes when PerlIO_define_layer is called due
146 to some bug in the perl dll, strawberry perl at least passes
147 the testsuite.
148 - implement Coro::Handle->peeraddr/host/port, for slightly
149 improved compatibility with LWP.
150 - implement 5.17 compatibility by almost blindly applying a
151 good-looking patch by Father Chrysostomos.
152 - move stack management functions into libcoro 3.
153 - libcoro version 3 "released".
154 - support magic values as timed_io_once args.
155 - recommend AnyEvent 7+ or EV 4+, also require EV
156 version 4 or newer for Coro::EV.
157
1586.10 Tue Oct 9 01:14:27 CEST 2012
159 - updated ecb.h, it had a typo that caused it to not compile on many
160 big endian systems (reported by many people).
161 - disable memory fences in ecb.h to improve portability.
162
1636.09 Sat Oct 6 23:25:02 CEST 2012
164 - Coro::EV I/O watchers were not interruptible by exceptions
165 (Coro::State::throw) (testcase by sten).
166 - ->throw now puts threads into the ready queue, as this seems to
167 be expected by existing code, and code that doesn't cope with spurious
168 wakeups needs fixing anyway.
169 - use fd -1 in mmap.
170 - cast I32 to int in error message printf.
171 - warn about broken so-called "hardened" kernels.
172
1736.08 Fri Apr 13 12:05:47 CEST 2012
174 - be more aggressive about exiting like perl does - formerly,
175 exiting from the non-main thread would not execute END blocks.
176
1776.07 Fri Nov 11 21:21:48 CET 2011
178 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
179 - update ecb.h.
180
1816.06 Mon Aug 8 23:59:48 CEST 2011
182 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
183 backend, which is an order of magnitude slower. unfortunately, cygwins
184 pthread implementation isn't very complete either, so allocate the stack
185 twice just to be sure.
186 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
187 in its ithreaded perl - best recompile cygwin and use the 'w'indows
188 backend for much better performance. also disable ithreads for
189 even better performance...).
190
1916.05 Thu Aug 4 21:36:36 CEST 2011
192 - blush, condvar values would not be propagated from send to recv anymore
193 (reported by Chip Salzenberg).
194 - use exponential increase for the readline buffer length in
195 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
196
1976.04 Wed Aug 3 17:15:45 CEST 2011
198 - use even more efficient and more compatible condvars for
199 compatibility to AnyEvent 6.x :)
200 - more inconsequential ecb.h updates.
201
2026.03 Wed Aug 3 11:41:30 CEST 2011
203 - change how Coro patches AnyEvent condvars for compatibility to
204 AnyEvent 6.x.
205 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
206 hasn't been included.
207
2086.02 Wed Jul 13 04:35:19 CEST 2011
209 - "improve portability to Gentoo" - gentoo manages to put perl variables
210 in memory areas that are farther than 2gb apart, which the jit couldn't
211 handle and barfed. now it's just a bit slower on gentoo and similar
212 systems.
213
2146.01 Sun Jul 3 12:31:14 CEST 2011
215 - workarounds are good, but the test for whether pthreads are used
216 was not good. this one should be better.
217 - check differently whether gcc generates cfi instructions itself.
218
2196.0 Wed Jun 29 19:43:35 CEST 2011
220 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
221 destroyed and cleaned up automatically (e.g. async { schedule }).
222 - implement a JIT compiler for part of the thread switch code,
223 which gives a 50% speed improvement on threaded perls, and
224 about 4% on non-threaded perls (so threaded perls now finally
225 reach about half the speed of non-threaded perls).
226 - slightly modernise Coro::Intro, add section about rouse functions.
227 - avoid DEFSV and ERRSV, giving another 10% improvement
228 in thread switching.
229 - Coro::State->is_destroyed is now called is_zombie.
230 - implement a Coro->safe_cancel method that might fail, but
231 cancels in a "safer" way if it succeeds.
232 - add preliminary support for DEBUGGING perls.
233 - get rid of two hash-accesses when initialising a new Coro - this
234 speeds up coro creation by almost a factor of two.
235 - croak when a coro that is being cancelled tries to block
236 (e.g. while executing a guard block), instead of crashing or
237 deadlocking.
238 - use a more robust and also faster method to identify Coro::State
239 objects - speeds up everything a bit.
240 - implement Coro->cancel in XS for a 20% speed improvement, and to
241 be able to implement mutual cancellation.
242 - speed up context switches by a percent or two by more efficiently
243 allocating context stack entries.
244 - implement Coro->join and Coro->on_destroy in XS for a speedup and
245 a reduction in memory use.
246 - cancelling a coro while it itself is cancelling another coro is
247 now supported and working, instead of triggering an assertion.
248 - be a bit more crash-resistant when calling (buggy) on_destroy
249 callbacks (best effort).
250 - move on_destroy into the slf_frame, to allow extension slf
251 functions to have destructors.
252 - get rid if coro refcounting - simply crash in other interpreter
253 threads by nulling the pointers on clone.
254 - simplify warn/die hook handling when loading Coro - the convoluted
255 logic seems to be no longer neccessary.
256 - use libecb instead of our own home-grown gcc hacks.
257 - document alternatives to Coro::LWP. Please use them :)
258 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
259 sigaltstack. Really. wine suffers from it, erlang suffers from it,
260 and it's known since at least 2006.
261
2625.372 Wed Feb 23 06:14:30 CET 2011
263 - apparently mingw doesn't provide a working gettimeofday, try to
264 work around that by relying on Time::HiRes (indirectly brought to
265 my attention by Max Maischein).
266 - fix some portability issues when Time::HiRes was used.
267
2685.371 Mon Feb 21 14:36:08 CET 2011
269 - backport to windows process emulation code again.
270
2715.37 Sat Feb 19 07:49:44 CET 2011
272 - add a big "Coro thread life cycle" section to "man Coro".
273 - try a tentative workaround against the breakage that 5.13 has
274 introduced without depreciation period. sigh.
275 - no longer use Time::HiRes if gettimeofday is available, which
276 saves quite a lot of memory.
277
2785.36 Sun Feb 13 05:33:41 CET 2011
279 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
280 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
281 method call.
282 - undocument Coro::Timer::sleep and obsolete whole module.
283 - optimise Coro::Timer::timeout memory and cpu usage.
284 - slightly updated Coro::Intro for recent changes.
285 - do not initialise PL_dirty anymore.
286
2875.25 Thu Nov 11 01:08:39 CET 2010
288 - try a different approach on netbsd - netbsd 5 finally has marginally
289 working pthreads, but still broken ucontext/sigaltstack.
290 - openbsd 4.8 finally got their act together, Coro works out of the box
291 with asm, setjmp and pthreads (no change, just informational).
292
2935.24 Sat Oct 23 11:27:12 CEST 2010
294 - port to the EV 4.0 API.
295 - work around bugs in mingw32, making strawberry perl work
296 out of the box.
297 - correctly modify Coro::AIO function prototypes
298 so that they reflect the "no optional parameters" rule.
299 - "ported" libcoro to C++.
300
3015.23 Mon May 17 18:50:42 CEST 2010
302 - be more resistant to ordering changes when initialising
303 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
304 Waldorf).
305 - document that perl 5.12 deliberately removed support for cloning.
306
3075.22 Wed Apr 14 03:55:35 CEST 2010
308 - correctly return udnef on errors in Coro::Handle::read/write
309 (testcase by Marc Mims).
310 - convert Coro::Util into a "perl compatibility wrapper" - the functions
311 are less useful now, but are drop-in replacements for existing
312 functions, listing better alternatives in the documentation. This also
313 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
314 with Coro::Util::inet_aton.
315 - do not override $Coro::idle unconditionally in Coro.pm, as other
316 modules could have provided their own idle coro already
317 (for exmaple, Coro::AnyEvent).
318 - fix Coro::Util::gethost* functions.
319 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
320
3215.21 Wed Dec 16 07:19:51 CET 2009
322 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
323 together.
324 - add some examples on how to combine other event loops with Coro in
325 Coro::AnyEvent, and how to run it (and not to block). Seems to be
326 the most common source of confusion.
327 - try to catch people naively blocking in an event callback.
328 - work around the perl filehandle bug issue in conjunction with
329 older common::sense (as indirectly pointed out by ZSystem).
11 - clarify the "not from signal handlers" section. 330 - clarify the "not from signal handlers" section.
12 331
135.2 Sun Oct 4 14:54:24 CEST 2009 3325.2 Sun Oct 4 14:54:24 CEST 2009
14 - Coro::Storable destroyed the prototypes of the functions it wrapped. 333 - Coro::Storable destroyed the prototypes of the functions it wrapped.
15 - export rouse_cb and rouse_wait by default now. 334 - export rouse_cb and rouse_wait by default now.
… …
264 - ->throw is now supported on Coro::State objects. 583 - ->throw is now supported on Coro::State objects.
265 - clean up cctx creation code a bit. 584 - clean up cctx creation code a bit.
266 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 585 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
267 586
2684.9 Sat Nov 8 17:45:27 CET 2008 5874.9 Sat Nov 8 17:45:27 CET 2008
269 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 588 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
270 - (libcoro) no longer leak threads in the experimental pthread backend, 589 - (libcoro) no longer leak threads in the experimental pthread backend,
271 also speed it up considerably. 590 also speed it up considerably.
272 - (libcoro) do not rely on makecontext passing void *'s unscathed. 591 - (libcoro) do not rely on makecontext passing void *'s unscathed.
273 - fix compiletime dependencies on libcoro in the Makefile. 592 - fix compiletime dependencies on libcoro in the Makefile.
274 - cctx_count wasn't always updated properly. 593 - cctx_count wasn't always updated properly.
… …
336 - remove debugging code related to MgPV_nolen_const, also try to 655 - remove debugging code related to MgPV_nolen_const, also try to
337 make it compile with perl 5.8.6 (yes, apple apparently loves 656 make it compile with perl 5.8.6 (yes, apple apparently loves
338 outdated software). Reported by John S. 657 outdated software). Reported by John S.
339 658
3404.744 Tue Jul 8 22:06:35 CEST 2008 6594.744 Tue Jul 8 22:06:35 CEST 2008
341 - correctly provide default DIE/WARN handlers as documented. 660 - correctly provide default DIE/WARN handlers as documented.
342 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 661 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
343 implementations inside perl work fine for us. 662 implementations inside perl work fine for us.
344 663
3454.743 Mon Jun 16 00:21:57 CEST 2008 6644.743 Mon Jun 16 00:21:57 CEST 2008
346 - when using Coro::EV without running EV::loop it could 665 - when using Coro::EV without running EV::loop it could
… …
368 - sprinkle "no warnings" freely over everything, also suppress 687 - sprinkle "no warnings" freely over everything, also suppress
369 warnings for some other modules. 688 warnings for some other modules.
370 - fix typo in WSAEWOULDBLOCK. 689 - fix typo in WSAEWOULDBLOCK.
371 690
3724.72 Sun May 25 05:14:36 CEST 2008 6914.72 Sun May 25 05:14:36 CEST 2008
373 - tweak META.yaml a bit, unfortunately, there is no documented way 692 - tweak META.yaml a bit, unfortunately, there is no documented way
374 to have optional dependencies with CPAN. doh :( 693 to have optional dependencies with CPAN. doh :(
375 - avoid running some tests on windows because they would fail due to 694 - avoid running some tests on windows because they would fail due to
376 perl bug (broken fork, broken pipes...). 695 perl bug (broken fork, broken pipes...).
377 - work around perl on windows bugs where perl returns undocumented 696 - work around perl on windows bugs where perl returns undocumented
378 error codes for sysread, syswrite etc. by taking advantage 697 error codes for sysread, syswrite etc. by taking advantage
… …
382 - use unix domain sockets in testsuite to work around 701 - use unix domain sockets in testsuite to work around
383 common perl implementation bugs on widows (they are emulated by 702 common perl implementation bugs on widows (they are emulated by
384 tcp sockets on windows. ugh.) 703 tcp sockets on windows. ugh.)
385 704
3864.71 Sat May 24 20:01:27 CEST 2008 7054.71 Sat May 24 20:01:27 CEST 2008
387 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 706 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
388 - take advantage of async name resolution of AnyEvent::Util. 707 - take advantage of async name resolution of AnyEvent::Util.
389 - work around brutal inet_aton override in Coro::LWP. 708 - work around brutal inet_aton override in Coro::LWP.
390 - take advantage of the readyhook in Coro::EV, for smoother 709 - take advantage of the readyhook in Coro::EV, for smoother
391 scheduling. 710 scheduling.
392 711
3934.7 Sun May 11 00:32:19 CEST 2008 7124.7 Sun May 11 00:32:19 CEST 2008
394 - completely reworked the Coro manpage. 713 - completely reworked the Coro manpage.
395 - added Coro::AnyEvent, generic event loop integration. 714 - added Coro::AnyEvent, generic event loop integration.
396 - implement cancel, ready and kill commands in Coro::Debug. 715 - implement cancel, ready and kill commands in Coro::Debug.
397 - document find_coro in Coro::Debug. 716 - document find_coro in Coro::Debug.
398 - incompatible change: rename has_stack to has_cctx. 717 - incompatible change: rename has_stack to has_cctx.
399 - Coro::AIO and Coro::BDB no longer force event model detection, 718 - Coro::AIO and Coro::BDB no longer force event model detection,
400 use AnyEvent::AIO and AnyEvent::BDB. 719 use AnyEvent::AIO and AnyEvent::BDB.
… …
524 843
5254.11 Thu Oct 11 02:40:24 CEST 2007 8444.11 Thu Oct 11 02:40:24 CEST 2007
526 - port to threaded perls. 845 - port to threaded perls.
527 846
5284.1 Thu Oct 11 02:38:16 CEST 2007 8474.1 Thu Oct 11 02:38:16 CEST 2007
529 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 848 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
530 be local to each coro (see Coro::State). 849 be local to each coro (see Coro::State).
531 - incompatible change: for very deep reasons, cede and cede_notself 850 - incompatible change: for very deep reasons, cede and cede_notself
532 cannot return anything, so nothing will be returned. 851 cannot return anything, so nothing will be returned.
533 - possibly bring back 5.10 compatibility (untested). 852 - possibly bring back 5.10 compatibility (untested).
534 - work around stupid (and wrong) warning on 5.10 :(. 853 - work around stupid (and wrong) warning on 5.10 :(.
… …
606 but stateful semantics. 925 but stateful semantics.
607 - fixed a lot of typos in Coro.pm (patch submitted by David 926 - fixed a lot of typos in Coro.pm (patch submitted by David
608 Steinbrunner, which applied flawlessly). 927 Steinbrunner, which applied flawlessly).
609 928
6103.6 Sat Apr 14 17:13:31 CEST 2007 9293.6 Sat Apr 14 17:13:31 CEST 2007
611 - added some bugfixes to get eg/myhttpd working again. 930 - added some bugfixes to get eg/myhttpd working again.
612 - added Coro::Storable for often-cede'ing freeze/thaw. 931 - added Coro::Storable for often-cede'ing freeze/thaw.
613 - try to do a clean exit when a coroutine calls exit 932 - try to do a clean exit when a coroutine calls exit
614 (EXPERIMENTAL). 933 (EXPERIMENTAL).
615 - got rid of indirect call through _coro_init. 934 - got rid of indirect call through _coro_init.
616 - updated the partly antique examples in eg/ to 935 - updated the partly antique examples in eg/ to

Diff Legend

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