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

Comparing Coro/Changes (file contents):
Revision 1.174 by root, Mon Dec 4 03:54:44 2006 UTC vs.
Revision 1.661 by root, Sat Jun 25 19:22:33 2016 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: transfer flags must be per state 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: better (earlier) error messages on broken transfer's 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?
5 8
63. 9 - simplify and speed up __DIE__ and __WARN__ handling.
10
116.51 Sat Jun 25 20:57:20 CEST 2016
12 - 6.5 release didn't compile on most older perls due to
13 a logic error enabling experimental 5.24 code for all versions.
14 - since various broken or deficient patches float around
15 to work around the vtbl API breakage in perl, let's provide
16 our own workaround, which seems to be much less invasive
17 and more compatible than the approaches seen so far. YMMV.
18 - work around assertion failure in perl_destruct on 5.24
19 (analyzed by Dave Mitchell).
20
216.5 Wed Jun 22 22:23:50 CEST 2016
22 - swap_sv swapping order was not symmetrical, causing
23 wrong swaps when swap_sv was used multiple times on the
24 same sv in the same thread.
25 - swap_sv calls can now be undone by calling it again
26 with the same variables.
27 - swap_sv calls will now be undone in async_pool threads.
28 - split Coro::Semaphore::up/adjust into separate xs functions
29 for better error reporting, at slight codesize increase.
30 - (libcoro) arm assembly support, please test and report.
31 - adjust to PL_savestack changes in perl 5.24 (adapted from
32 the debian patch, which unfortunately gets it wrong).
33
346.49 Sat Oct 17 01:40:12 CEST 2015
35 - throwing an exception to a thread waiting in
36 Coro::Handle using Coro::EV did not stop the watchers,
37 causing the next call to fail (testcase by Martin Pritchard).
38 - bump minimum perl version to 5.10.
39
406.48 Sun Oct 4 19:03:51 CEST 2015
41 - fix memory leak when Coro::Handle uses Coro::EV internally
42 (testcase by Sten Sten).
43 - update code to libev 4 API (internally, Coro still used the
44 version 3 API calls).
45 - remove unused hv_sig.
46
476.47 Sat Jul 11 03:58:20 CEST 2015
48 - swap_svs - NVs can be in the SV head beginning with 5.20.
49
506.46 Tue Jun 30 14:36:30 CEST 2015
51 - restore portability to perls with windows fork emulation
52 (patch by Petr Písař).
53
546.45 Tue Jun 30 01:40:08 CEST 2015
55 - restore portability to perl 5.16 and below.
56
576.44 Tue Jun 30 00:41:54 CEST 2015
58 - avoid segfaulting (or worse) when tracing a canceled thread.
59 - the Event module silently broke it's public hook API, causing
60 failures on perls compiled with -Duselongdouble. Requiring
61 the latest version of Event because I don't know which release
62 changed this (it's not mentioned in the ChangeLog).
63 - allow xs level enterleave hooks via CoroAPI.
64
656.43 Thu Jun 4 15:38:14 CEST 2015
66 - use stability canary.
67 - port to stableperl-5.22.0-1.001.
68 - update libecb.
69
706.42 Wed Feb 11 20:29:52 CET 2015
71 - Coro::SemaphoreSet->try did not actually work (analyzed by
72 SATO Kentaro).
73 - upgrade libecb to be C11 compliant.
74
756.41 Sat Sep 6 22:08:46 CEST 2014
76 - restore portability to perl 5.8.x.
77 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
78 up some memory, but fixes "use feature" and similar modules
79 when used inside a Coro.
80 - allow zero as argument to Coro::Channel to mean the same thing
81 as no argument. this works with older versions as well,
82 but wasn't legal until now.
83 - slightly better c header file detection.
84
856.39 Mon Jun 2 00:00:08 CEST 2014
86 - work around more incompatible changes in 5.20.
87
886.38 Sun Jun 1 21:54:23 CEST 2014
89 - check that perl slots actually have enough space to hold
90 interpreter variables.
91 - untested port to perl 5.19 (64 bit tmps indices) (reported
92 by Andreas König).
93 - croak when cancel is called without a thread context.
94
956.37 Tue Mar 4 13:27:33 CET 2014
96 - *sigh*, removed leftover debugging code from debugging a
97 perl bug, of all things.
98
996.36 Tue Mar 4 07:11:59 CET 2014
100 - semaphores would not clear the destroy hook when interrupted
101 by ->throw, causing segfaults or worse.
102 - ->throw on a thread waiting for a semaphore did not acquire
103 the semaphore, but also didn't wake up other waiters,
104 possibly causing a deadlock.
105 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
106 to report a full stacktrace, which should help find out
107 where the actual call is.
108 - "atomically" destroy data in slf_destroy, because it is
109 the right thing to do, just in case.
110 - disable ecb.h libm dependency, to compile on ancient systems
111 or under adverse conditions.
112
1136.33 Mon Nov 18 11:26:27 CET 2013
114 - do not crash when freeing padlists with holes (in 5.18).
115 - tentative SVt_BIND 5.19 port/fix.
116
1176.32 Tue Nov 5 15:35:35 CET 2013
118 - use a new algorithm to derive padlists for perl 5.18. The old
119 one could lead to 0-pointer accesses inside perl (reported
120 by Darin McBride).
121
1226.31 Thu May 9 07:39:48 CEST 2013
123 - Coro::AIO requests would crash if the thread was ready'd
124 while the request was ongoing.
125
1266.29 Wed May 8 02:55:18 CEST 2013
127 - when an on_destroy handler destructs the coro currently being
128 destructed a perl scalar could be accessed after being freed,
129 likely causing a crash.
130
1316.28 Wed Mar 6 06:58:02 CET 2013
132 - clean remnants of existing __DIE__ and __WARN__ handlers so
133 they lose their magic and will not cause segfaults later
134 (testcase by Andrey Sagulin).
135 - improved Coro::State documentation a bit.
136 - Coro::Debug::command now flushes the output.
137 - add hack detection code for x32 abi, because the braindead slugs
138 who designed that made it look exactly like x86_64 without
139 providing proper compile time symbols to test for it. as a result,
140 this detection cannot work reliably.
141 - valgrind stack registering was broken.
142 - do not rely on Time::HiRes anymore in Coro::Debug.
143
1446.23 Fri Dec 7 23:36:37 CET 2012
145 - use experimental fiber implementation on native windows
146 perls.
147 - use sizeof (void *) as multiplication factor for stack sizes,
148 to accomodate the totally braindamaged microsoft 64 bit "os".
149 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
150 activeperl 5.16 crashes when PerlIO_define_layer is called due
151 to some bug in the perl dll, strawberry perl at least passes
152 the testsuite.
153 - implement Coro::Handle->peeraddr/host/port, for slightly
154 improved compatibility with LWP.
155 - implement 5.17 compatibility by almost blindly applying a
156 good-looking patch by Father Chrysostomos.
157 - move stack management functions into libcoro 3.
158 - libcoro version 3 "released".
159 - support magic values as timed_io_once args.
160 - recommend AnyEvent 7+ or EV 4+, also require EV
161 version 4 or newer for Coro::EV.
162
1636.10 Tue Oct 9 01:14:27 CEST 2012
164 - updated ecb.h, it had a typo that caused it to not compile on many
165 big endian systems (reported by many people).
166 - disable memory fences in ecb.h to improve portability.
167
1686.09 Sat Oct 6 23:25:02 CEST 2012
169 - Coro::EV I/O watchers were not interruptible by exceptions
170 (Coro::State::throw) (testcase by sten).
171 - ->throw now puts threads into the ready queue, as this seems to
172 be expected by existing code, and code that doesn't cope with spurious
173 wakeups needs fixing anyway.
174 - use fd -1 in mmap.
175 - cast I32 to int in error message printf.
176 - warn about broken so-called "hardened" kernels.
177
1786.08 Fri Apr 13 12:05:47 CEST 2012
179 - be more aggressive about exiting like perl does - formerly,
180 exiting from the non-main thread would not execute END blocks.
181
1826.07 Fri Nov 11 21:21:48 CET 2011
183 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
184 - update ecb.h.
185
1866.06 Mon Aug 8 23:59:48 CEST 2011
187 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
188 backend, which is an order of magnitude slower. unfortunately, cygwins
189 pthread implementation isn't very complete either, so allocate the stack
190 twice just to be sure.
191 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
192 in its ithreaded perl - best recompile cygwin and use the 'w'indows
193 backend for much better performance. also disable ithreads for
194 even better performance...).
195
1966.05 Thu Aug 4 21:36:36 CEST 2011
197 - blush, condvar values would not be propagated from send to recv anymore
198 (reported by Chip Salzenberg).
199 - use exponential increase for the readline buffer length in
200 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
201
2026.04 Wed Aug 3 17:15:45 CEST 2011
203 - use even more efficient and more compatible condvars for
204 compatibility to AnyEvent 6.x :)
205 - more inconsequential ecb.h updates.
206
2076.03 Wed Aug 3 11:41:30 CEST 2011
208 - change how Coro patches AnyEvent condvars for compatibility to
209 AnyEvent 6.x.
210 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
211 hasn't been included.
212
2136.02 Wed Jul 13 04:35:19 CEST 2011
214 - "improve portability to Gentoo" - gentoo manages to put perl variables
215 in memory areas that are farther than 2gb apart, which the jit couldn't
216 handle and barfed. now it's just a bit slower on gentoo and similar
217 systems.
218
2196.01 Sun Jul 3 12:31:14 CEST 2011
220 - workarounds are good, but the test for whether pthreads are used
221 was not good. this one should be better.
222 - check differently whether gcc generates cfi instructions itself.
223
2246.0 Wed Jun 29 19:43:35 CEST 2011
225 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
226 destroyed and cleaned up automatically (e.g. async { schedule }).
227 - implement a JIT compiler for part of the thread switch code,
228 which gives a 50% speed improvement on threaded perls, and
229 about 4% on non-threaded perls (so threaded perls now finally
230 reach about half the speed of non-threaded perls).
231 - slightly modernise Coro::Intro, add section about rouse functions.
232 - avoid DEFSV and ERRSV, giving another 10% improvement
233 in thread switching.
234 - Coro::State->is_destroyed is now called is_zombie.
235 - implement a Coro->safe_cancel method that might fail, but
236 cancels in a "safer" way if it succeeds.
237 - add preliminary support for DEBUGGING perls.
238 - get rid of two hash-accesses when initialising a new Coro - this
239 speeds up coro creation by almost a factor of two.
240 - croak when a coro that is being cancelled tries to block
241 (e.g. while executing a guard block), instead of crashing or
242 deadlocking.
243 - use a more robust and also faster method to identify Coro::State
244 objects - speeds up everything a bit.
245 - implement Coro->cancel in XS for a 20% speed improvement, and to
246 be able to implement mutual cancellation.
247 - speed up context switches by a percent or two by more efficiently
248 allocating context stack entries.
249 - implement Coro->join and Coro->on_destroy in XS for a speedup and
250 a reduction in memory use.
251 - cancelling a coro while it itself is cancelling another coro is
252 now supported and working, instead of triggering an assertion.
253 - be a bit more crash-resistant when calling (buggy) on_destroy
254 callbacks (best effort).
255 - move on_destroy into the slf_frame, to allow extension slf
256 functions to have destructors.
257 - get rid if coro refcounting - simply crash in other interpreter
258 threads by nulling the pointers on clone.
259 - simplify warn/die hook handling when loading Coro - the convoluted
260 logic seems to be no longer neccessary.
261 - use libecb instead of our own home-grown gcc hacks.
262 - document alternatives to Coro::LWP. Please use them :)
263 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
264 sigaltstack. Really. wine suffers from it, erlang suffers from it,
265 and it's known since at least 2006.
266
2675.372 Wed Feb 23 06:14:30 CET 2011
268 - apparently mingw doesn't provide a working gettimeofday, try to
269 work around that by relying on Time::HiRes (indirectly brought to
270 my attention by Max Maischein).
271 - fix some portability issues when Time::HiRes was used.
272
2735.371 Mon Feb 21 14:36:08 CET 2011
274 - backport to windows process emulation code again.
275
2765.37 Sat Feb 19 07:49:44 CET 2011
277 - add a big "Coro thread life cycle" section to "man Coro".
278 - try a tentative workaround against the breakage that 5.13 has
279 introduced without depreciation period. sigh.
280 - no longer use Time::HiRes if gettimeofday is available, which
281 saves quite a lot of memory.
282
2835.36 Sun Feb 13 05:33:41 CET 2011
284 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
285 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
286 method call.
287 - undocument Coro::Timer::sleep and obsolete whole module.
288 - optimise Coro::Timer::timeout memory and cpu usage.
289 - slightly updated Coro::Intro for recent changes.
290 - do not initialise PL_dirty anymore.
291
2925.25 Thu Nov 11 01:08:39 CET 2010
293 - try a different approach on netbsd - netbsd 5 finally has marginally
294 working pthreads, but still broken ucontext/sigaltstack.
295 - openbsd 4.8 finally got their act together, Coro works out of the box
296 with asm, setjmp and pthreads (no change, just informational).
297
2985.24 Sat Oct 23 11:27:12 CEST 2010
299 - port to the EV 4.0 API.
300 - work around bugs in mingw32, making strawberry perl work
301 out of the box.
302 - correctly modify Coro::AIO function prototypes
303 so that they reflect the "no optional parameters" rule.
304 - "ported" libcoro to C++.
305
3065.23 Mon May 17 18:50:42 CEST 2010
307 - be more resistant to ordering changes when initialising
308 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
309 Waldorf).
310 - document that perl 5.12 deliberately removed support for cloning.
311
3125.22 Wed Apr 14 03:55:35 CEST 2010
313 - correctly return udnef on errors in Coro::Handle::read/write
314 (testcase by Marc Mims).
315 - convert Coro::Util into a "perl compatibility wrapper" - the functions
316 are less useful now, but are drop-in replacements for existing
317 functions, listing better alternatives in the documentation. This also
318 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
319 with Coro::Util::inet_aton.
320 - do not override $Coro::idle unconditionally in Coro.pm, as other
321 modules could have provided their own idle coro already
322 (for exmaple, Coro::AnyEvent).
323 - fix Coro::Util::gethost* functions.
324 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
325
3265.21 Wed Dec 16 07:19:51 CET 2009
327 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
328 together.
329 - add some examples on how to combine other event loops with Coro in
330 Coro::AnyEvent, and how to run it (and not to block). Seems to be
331 the most common source of confusion.
332 - try to catch people naively blocking in an event callback.
333 - work around the perl filehandle bug issue in conjunction with
334 older common::sense (as indirectly pointed out by ZSystem).
335 - clarify the "not from signal handlers" section.
336
3375.2 Sun Oct 4 14:54:24 CEST 2009
338 - Coro::Storable destroyed the prototypes of the functions it wrapped.
339 - export rouse_cb and rouse_wait by default now.
340 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
341 - new method $state->swap_sv.
342 - added section on "windows process emulation" to the manpage,
343 after a not-so-fruitful (nor-friendly) "discussion" with chip
344 salzenberg (discussion implies arguments, but his only arguments
345 were ad-hominems, one wonders why he started it in the first
346 place). I hope this explains it well enough for him to understand,
347 and maybe well enough for others to understand.
348 - use common::sense everywhere now.
349 - idle callbacks are no longer supported, use idle coros instead.
350 - print a thread listing when a deadlock is detected.
351
3525.17 Sat Aug 22 23:09:31 CEST 2009
353 - work around a bug in the perl debugger causing crashes
354 when running under the debugger by marking _pool_handler as nodebug.
355 - speed up Coro::async considerably.
356 - try some hacks to get netbsd to work "more often" - their broken
357 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
358 - convert Coro to AE 5.0 API.
359
3605.162 Tue Jul 28 04:04:03 CEST 2009
361 - perl 5.8.2 is now minimum requirement.
362 - skip t/19_handle.t on broken windows perls.
363
3645.161 Wed Jul 22 04:47:38 CEST 2009
365 - Coro::AnyEvent::poll could have a different prototype when EV was
366 used as backend (analyzed by Tatsuhiko Miyagawa).
367 - Coro::AnyEvent errornously initialised the event loop when loaded,
368 not on demand.
369 - try to workaround rare */t/01_unblock.t failures.
370
3715.16 Tue Jul 21 01:44:37 CEST 2009
372 - Coro::AnyEvent failed to hook into the event loop
373 when no threads had been readied between detecting
374 the event loop and actually running it.
375 - considerably speed up Coro::Select by taking avdantage
376 of AnyEvent > 4.8 and some other optimisations.
377 - implement paragraph readline mode in Coro::Handle
378 (based on patches by Zsbán Ambrus).
379 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
380 and analyzed by Yasuhiro MATSUMOTO).
381 - clarified libcoro license and copyright.
382 - someone stole my EXTRA_META!!!
383 - implement Coro::Select::patch_pp_sselect and it's brother,
384 for hardcode select overriding.
385
3865.151 Mon Jul 6 05:41:57 CEST 2009
387 - backport to windows process emulation code again (patch by
388 Yasuhiro MATSUMOTO).
389 - slightly update Coro::MakeMaker.
390
3915.15 Tue Jun 30 10:28:06 CEST 2009
392 - deprecate Coro::Socket, document how to get ipv6 support via
393 AnyEvent::Socket instead.
394 - implement signal->wait ($cb) interface, similar to semaphores.
395 - work around SvOK not supporting getmagic, so we have to getmagic
396 to test for undef :/ (reported by Matthias Waldorf).
397 - load Coro::AnyEvent in all modules using AnyEvent.
398 - work around perl corrupting our internal data structures,
399 reported by Tokuhiro Matsuno.
400 - enable per-coroutine real and cpu time gathering
401 (Coro::State::enable_times).
402
4035.14 Wed Jun 24 01:37:48 CEST 2009
404 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
405 as this seems to have been a difficult concept (poll, sleep, idle,
406 idle_upto).
407 - add Coro::AnyEvent::readable/writable functions.
408 - clarify Coro::EV/Event/AnyEvent manpages.
409 - free per-thread global scalars in the thread calling ->cancel, to
410 avoid crashes when $_, $@ etc., are magical but some of those
411 had already been freed.
412 - "unexperimentalise" the callback interface for Coro::Semaphore.
413 - speed up ready queue management/context switching by using a linked
414 list instead of an array (~5-10%).
415 - implement "watch" command in Coro::Debug shells.
416 - for fun, implement time-slicing as an example in the manpage.
417 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
418 Coro::Event, use the normal AnyEvent handling instead of dieing
419 (the same is true for Coro::Handle).
420 - properly document Coro::EV::timed_io_once.
421 - avoid unneccessary ->cancel calls in Coro::Handle.
422 - maybe make it work on mingw32 with win32 backend
423 (based on patch by Yasuhiro Matsumoto).
424
4255.132 Fri May 29 09:00:39 CEST 2009
426 - do not keep a reference to the argument itself in
427 Coro::Semaphore::guard, as it could change later.
428 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
429 should support a prepare callback.
430
4315.131 Mon Mar 16 23:20:37 CET 2009
432 - implement and document Coro->suspend, Coro->resume.
433 - fix Coro::Select implementation to not (often) close
434 the passed file descriptors (testcase provided by pippijn).
435
4365.13 Mon Dec 15 21:51:42 CET 2008
437 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
438 - don't set diehook to C<undef> but instead to NULL, to avoid
439 spurious warnings.
440 - fix a lot of bugs in Coro::SemaphoreSet.
441 - Coro::SemaphoreSet will less often create a semaphore needlessly.
442 - add Coro::SemaphoreSet::count and wait methods.
443 - take advantage of the new Guard module.
444 - deprecate Coro::guard.
445 - try to fix the dreaded 01_unblock tests once more. I hate it when
446 testsuites need more fixing than the code they are supposed to test.
447 - croak in more cases when a required callback isn't resolvable.
448 - fix some minor issues in Coro::State->call/eval.
449 - use current coroutine context instead of a temporary one
450 when temporarily switching to another coroutine.
451 - do not call C-level on_destroy handlers during global destruction,
452 to avoid needless segfaults.
453
4545.12 Sun Dec 7 13:30:38 CET 2008
455 - add default config for MirOS, which seems to be a bug-to-bug
456 compatible fork of openbsd ("world domination by releasing
457 openbsd cvs before the openbsd folks do it" or so :).
458 - free_padlist did destroy the names pad, not good, but didn't
459 seem to bother perl - this could fix issues such as eval ""
460 inside a function called from multiple coroutines.
461 - use a different method to detect destruction time.
462 - be more careful when freeing padlists just before global
463 destruction.
464 - fixed and expanded the call/cc example.
465 - renamed _terminate to _coro_run.
466 - new method Coro::Channel->shutdown.
467 - try pthreads on openbsd <4.4 (broken sigaltstack, will
468 pthreads fare better?).
469 - be less picky about destroying "a" running coroutine.
470
4715.11 Tue Nov 25 21:49:05 CET 2008
472 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
473 working code with failed assertions, introducing more bugs than
474 it fixes, requiring elaborate workarounds :(
475
4765.1 Mon Nov 24 08:54:59 CET 2008
477 - wrote a small introductory tutorial - Coro::Intro.
478 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
479 - Coro::Select did errornously dup the file descriptors
480 and didn't work with all AnyEvent backends.
481 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
482 LWP data transfers.
483 - disassociate c contexts from coro objects - this is agruably more
484 correct, but mostly allows sharing of cctxs between coro and state
485 objects, for added memory savings and speed increases.
486 - bumped $Coro::POOL_RSS up to 32kb by default.
487 - no longer set the optype to OP_CUSTOM, as B::* understandably
488 doesn't like this very much (and we *are* a type of entersub).
489 - implement state cloning, just to prove that call/cc can be done.
490 - automatically load Coro::AnyEvent in Coro::Handle.
491 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
492 do not have this method (and don't need it either).
493 - speed up generic anyevent methods in Coro::Handle by using rouse
494 callbacks.
495 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
496 others. It also makes the debugger happier, as you can trace
497 through the idle threads now.
498 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
499 - eg/event was pretty broken.
500 - better 5.8.6 compatibility.
501
5025.0 Thu Nov 20 10:35:05 CET 2008
503 - NEW ARCHITECTURE: use the latest 4.x version if you experience
504 stability issues.
505 - bump API version to 7 - all dependents must be recompiled.
506 - removed timed_* functions - they were not being used anyways
507 and should be replaced by a more generic mechanism -
508 and were annoying to support anyways :)
509 - removed SemaphoreSet's waiter method - use sem method instead.
510 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
511 - async_pool did free a scalar value twice
512 ("Attempt to unreference...").
513 - fix a longstanding bug where calling terminate on a coro that
514 was waiting for a semaphore that was just becoming available
515 would cause a deadlock (semaphore would get into a state where
516 it was available but waiters were still blocked).
517 - calling throw on a coroutine that is waiting for a semaphore will
518 no longer make it acquire the semaphore (and thus leak a count).
519 - perl's process emulation is now not even theoretically supported
520 anymore.
521 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
522 conversion of callback-style to blocking-style.
523 - new methods $coro->schedule_to and ->cede_to, to specifically
524 schedule or cede to a specific coroutine.
525 - new function Coro::Semaphore::wait.
526 - use named constants in Coro::Channel (Richard Hundt).
527 - directly patch the entersub opcode calling SLF functions (cede,
528 transfer and so on). this does speed up context switching, but
529 more importanly, it frees us from the hardcoded behaviour of
530 entersub, so we might actually be able to return something from
531 those functions and atcually create new ones.
532 - take advantage of __builtin_frame_address on gcc.
533 - expose THX in coroapi (not sure whether this was a wise decision,
534 as "threaded" perls are running at half speed anyways).
535 - implement execute_slf (schedule-like-function) interface that makes
536 it possible to implement schedule-like-functions in XS.
537 - use new SLF interface to massively speed up Coro::EV by roughly a
538 factor of two.
539 - used new SLF interface to massively speed up Coro::Semaphore by a
540 factor of three.
541 - used new SLF interface to speed up Coro::AIO by roughly a factor of
542 four and reduce its memory usage considerably.
543 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
544 for a nice speedup and vastly more correct behaviour. Also implement
545 a new method "sem" to get at the underlying semaphore object.
546 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
547 (in comparison) 20-40% speedup.
548 - used new SLF interface to reimplement Coro::Signal gaining
549 some unknown (because I was too lazy), but certain, speedup, and also
550 making signals reliable for the first time.
551 - used new SLF interface and other optimisations to speed up async_pool
552 by a factor of two. It also doesn't rely on perl's exception mechanism
553 to exit anymore. The overhead for terminating an async_pool, coro over
554 a normal async is now very small.
555 - sped up coroutine creation/destruction by 40%.
556 - forgot to include Coro/libcoro/README in the dist for all these years.
557 - work around a freebsd pthreads bug (manual testcancel is required as
558 pthread_cond_wait isn't a cancellation point on freebsd).
559 - use new rouse functions to speed up and simplify Coro::BDB.
560 - make "prefer perl native functions" work with threaded perls.
561 - condense Coro::Debug ps output, hint at v and w flags.
562 - (libcoro) lots of minor cleanups and portability improvements.
563
5644.914 Wed Nov 19 12:54:18 CET 2008
565 - fix a disastrous bug in the readline optimisation
566 introduced in 4.801.
567
5684.913 Sat Nov 15 07:58:28 CET 2008
569 - async_pool did free a scalar value twice
570 ("Attempt to unreference...").
571
5724.912 Thu Nov 13 18:31:23 CET 2008
573 - minor cleanups.
574 - use much larger stacks on linux and perl < 5.8.8.
575 - Coro::Debug::new_unix_server did not unlink the socket
576 when destroyed.
577
5784.911 Tue Nov 11 04:26:17 CET 2008
579 - "port" to threaded perls.
580
5814.91 Mon Nov 10 05:36:38 CET 2008
582 - the ->throw exception object no longer leaks.
583 - creating a new cctx leaked a scopestack entry (memleak).
584 - new coroutines didn't get created with a zero flags field
585 (unknown impact).
586 - calling ->throw on a not-yet-started coroutine should now work
587 instead of being ignored.
588 - ->throw is now supported on Coro::State objects.
589 - clean up cctx creation code a bit.
590 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
591
5924.9 Sat Nov 8 17:45:27 CET 2008
593 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
594 - (libcoro) no longer leak threads in the experimental pthread backend,
595 also speed it up considerably.
596 - (libcoro) do not rely on makecontext passing void *'s unscathed.
597 - fix compiletime dependencies on libcoro in the Makefile.
598 - cctx_count wasn't always updated properly.
599 - Coro::State::cctx_stacksize wasn't applied correctly.
600 - new function Coro::State::cctx_max_idle.
601 - the default max number of idle C contexts is now 4.
602 - (libcoro) try harder to get _setjmp/_longjmp.
603 - (libcoro) cleanup and extend the libcoro API to officially
604 allow the creation of empty source contexts.
605 - very experimental workaround for the totally broken netbsd platform.
606 - tried to hack around openbsd bugs.
607
6084.804 Wed Nov 5 16:36:00 CET 2008
609 - Coro::Debug::new_unix_server would not create a non-blocking listening
610 socket, sometimes causing freezes.
611 - (libcoro) fix misaligned stack points for setjmp and assembly
612 methods, which can cause crashes on x86/x86_64 with a sufficiently
613 aggressive compiler.
614 - new function: Coro::Debug::new_tcp_server.
615 - move ->throw into the Coro class because it only works on coro objects.
616
6174.803 Mon Nov 3 17:16:12 CET 2008
618 - (libcoro) use a global asm statement to become independent of gcc
619 otpimisations for CORO_ASM (thanks to pippijn for the idea).
620 - try to workaround yet another broken bsd, this time dragonfly.
621
6224.802 Thu Oct 30 10:56:12 CET 2008
623 - support -fno-omit-frame-pointer on x86 with the assembly method.
624 - tune 01_unblock.t tests a bit.
625
6264.801 Wed Oct 22 18:33:37 CEST 2008
627 - improve readline speed for very long "lines".
628 - backport to 5.8.8.
629
6304.8 Thu Oct 2 13:34:40 CEST 2008
631 - new function Coro::AIO::aio_wait.
632 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
633 - greatly speed up and reduce memory usage of Coro::AIO requests.
634 - implement some other µ-optimisations.
635
6364.749 Mon Sep 29 14:40:12 CEST 2008
637 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
638
6394.748 Sat Sep 27 14:03:19 CEST 2008
640 - implement, but do not document, PerlIO::cede(granularity).
641 - Coro::Storable forgot to wrap Storable::pstore.
642 - work around the multitude of leaks and memory corruption
643 bugs in PerlIO::via by using our own C-level perliol. As a side
644 effect, Coro::Storable is now much, much, much faster.
645
6464.747 Tue Sep 23 01:59:41 CEST 2008
647 - fix a per-cv memleak (one empty array was leaked per
648 code reference).
649 - avoid a crash in coro->call|rss when the coroutine was already
650 destroyed (most noticably when using Coro::Debug::ps :)
651 - also protect *Storable::FILE.
652 - push up default storable granularity to 20ms.
653
6544.746 Sun Sep 21 03:22:20 CEST 2008
655 - be more insistent on locking Storable against reentrancy
656 in Coro::Storable.
657 - move swap_def?v and throw to Coro::State, as documented.
658
6594.745 Thu Jul 24 00:14:38 CEST 2008
660 - remove debugging code related to MgPV_nolen_const, also try to
661 make it compile with perl 5.8.6 (yes, apple apparently loves
662 outdated software). Reported by John S.
663
6644.744 Tue Jul 8 22:06:35 CEST 2008
665 - correctly provide default DIE/WARN handlers as documented.
666 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
667 implementations inside perl work fine for us.
668
6694.743 Mon Jun 16 00:21:57 CEST 2008
670 - when using Coro::EV without running EV::loop it could
671 result in busy-waiting for events, this has been fixed.
672 - reduce codesize and improve performance by using EV_DEFAULT_UC.
673
6744.742 Sat May 31 14:10:21 CEST 2008
675 - implement a workaround for (some) perl <5.8.8 versions.
676 - require EV 3.3+.
677
6784.741 Fri May 30 23:33:09 CEST 2008
679 - tell netbsd how utterly broken their imitation of an OS is
680 and refuse to build by default if pthreads are in use.
681 - switch to "s" method on all bsd's by default, as their ucontext
682 stuff seems just too broken.
683 - fix a bug in Coro::Select.
684
6854.74 Thu May 29 20:05:31 CEST 2008
686 - do not test Coro::LWP for lack of dependencies.
687 - do not test Coro::Select for lack of working perls.
688
6894.73 Thu May 29 2008
690 - fix a bug in Coro::EV which would cause it to block despite
691 there being runnable coroutines.
692 - sprinkle "no warnings" freely over everything, also suppress
693 warnings for some other modules.
694 - fix typo in WSAEWOULDBLOCK.
695
6964.72 Sun May 25 05:14:36 CEST 2008
697 - tweak META.yaml a bit, unfortunately, there is no documented way
698 to have optional dependencies with CPAN. doh :(
699 - avoid running some tests on windows because they would fail due to
700 perl bug (broken fork, broken pipes...).
701 - work around perl on windows bugs where perl returns undocumented
702 error codes for sysread, syswrite etc. by taking advantage
703 of AnyEvent's workaround for that problem.
704 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
705 a common perl implementation bug on windows.
706 - use unix domain sockets in testsuite to work around
707 common perl implementation bugs on widows (they are emulated by
708 tcp sockets on windows. ugh.)
709
7104.71 Sat May 24 20:01:27 CEST 2008
711 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
712 - take advantage of async name resolution of AnyEvent::Util.
713 - work around brutal inet_aton override in Coro::LWP.
714 - take advantage of the readyhook in Coro::EV, for smoother
715 scheduling.
716
7174.7 Sun May 11 00:32:19 CEST 2008
718 - completely reworked the Coro manpage.
719 - added Coro::AnyEvent, generic event loop integration.
720 - implement cancel, ready and kill commands in Coro::Debug.
721 - document find_coro in Coro::Debug.
722 - incompatible change: rename has_stack to has_cctx.
723 - Coro::AIO and Coro::BDB no longer force event model detection,
724 use AnyEvent::AIO and AnyEvent::BDB.
725
7264.6 Sat Apr 26 10:05:14 CEST 2008
727 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
728 by the Coro module. It was a mistake to have it in the first place.
729 - (experimental) support for activestate perl 5.10 (method "w").
730 - (experimental) support for strawberry perl (method "a").
731 - coro_sigelem_set did not return a value although it had to,
732 actual impact unknown.
733
7344.51 Mon Apr 14 13:28:27 CEST 2008
735 - make it compile again on 5.8.
736
7374.50 Thu Apr 10 09:43:17 CEST 2008
738 - I did it twice! (see 4.49).
739
7404.49 Sun Apr 6 21:23:31 CEST 2008
741 - grr, instead of compiling the recent changes on 5.10 only they
742 were compiled on 5.8 only.
743
7444.48 Sun Apr 6 20:36:46 CEST 2008
745 - allow coroutine switches during eval's under 5.10.x, as it seems
746 the parser is a per-interpreter option now, so this is safe
747 (this might fix the odd crashing bug).
748 - drop support for 5.9.x versions: they are dead, jim.
749
7504.47 Sun Apr 6 00:37:52 CEST 2008
751 - force cctx allocation on API calls: we know we need to force one
752 and gcc actually manages to confuse our heuristic nowadays,
753 leading to crashes and worse.
754 - document force_cctx.
755
7564.46 Fri Apr 4 22:05:43 CEST 2008
757 - upgrade libcoro, resulting in pthread-backend (which was only created
758 to fulfill the rules of the programming languages shootout).
759
7604.45 Thu Mar 13 11:55:36 CET 2008
761 - fix a file leaking bug in eg/mhyttpd that would allow
762 downloading of any file (reported by oesi).
763 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
764 (also add testcase).
765 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
766 apparently having loads of fun with that).
767 - implement $state->swap_defsv, ->swap_defav and document ->throw.
768
7694.4 Wed Feb 13 16:44:29 CET 2008
770 - only assume .cfi pseudo isns exist on GNU/Linux.
771 - add get_timed method to Coro::Channels.
772 - fixed Coro::Debug manpage.
773 - perl 5.11 compatbility improvement based on patch by
774 Andreas König.
775
7764.37 Sun Jan 20 11:25:23 CET 2008
777 - soften the check for an active parser for perl 5.10 (avoiding
778 "Coro::State::transfer called while parsing" in many cases).
779
7804.36 Sun Jan 13 10:53:56 CET 2008
781 - reset diehook when terminating from an async_pool as to not
782 trigger any __DIE__ handlers.
783
7844.35 Sun Jan 13 04:14:13 CET 2008
785 - "bt" debug command now displays any exceptions
786 from longmess and also skips the innermost
787 stackframes, giving more useufl output.
788 - allow backtraces in coroutines blocked in Coro::EV,
789 at a <1% speed hit.
790 - handle localising of $SIG{__WARN__} and __DIE__
791 properly (with a proper amount of dirty hacking).
792
7934.34 Sat Dec 22 17:49:53 CET 2007
794 - upgrade to EV version 2.0 API.
795
7964.33 Mon Dec 17 08:36:12 CET 2007
797 - make Coro::AIO etc. loadable in the absence of EV.
798
7994.32 Mon Dec 17 07:46:02 CET 2007
800 - majorly improved Coro::Handle's performance with Coro::EV.
801 - implemented the until now mythical Coro::BDB module.
802 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
803
8044.31 Wed Dec 5 12:32:39 CET 2007
805 - remove warn statement form Coro::Util.
806
8074.3 Tue Dec 4 20:33:14 CET 2007
808 - calls to the idle function could cause stack corruption
809 when the stack changed.
810 - do no longer rely on the presence of EV::DNS if EV is used
811 (because it is gone), but instead take avdantage of EV::ADNS
812 if available.
813 - add ($) prototypes to all functions in Coro::Storable.
814 - use a conventional (safer) idle callback in Coro::EV.
815 - do accept quickly in Coro::Debug to avoid endless loops.
816
8174.22 Fri Nov 30 16:04:04 CET 2007
818 - really use optimised versions for Event and EV in Coro::Util
819 and Coro::Handle.
820
8214.21 Sun Nov 25 10:48:59 CET 2007
822 - fix a spurious memory read.
823 - Coro::EV no longer keeps the eventloop "alive".
824
8254.2 Fri Nov 9 20:47:05 CET 2007
826 - enable/disable tracing from a new coroutine, not a pooled one.
827 - fix a memleak in Coro::Event.
828 - removed killall call from fork_eval.
829 - made sure store_fd is already loaded so that fork_eval
830 does not have to parse autoload in each subprocess.
831 - only use assembly method if -O switch is in $Config{optimize}.
832 - add (optional) Coro::EV module, so far the best event loop module
833 directly supported by Coro.
834 - if the event model is EV, use EV::DNS to resolve
835 stuff in Coro::Util.
836 - don't get confused by multiple event notifications in Coro::Handle.
837 - initial support for EV (libevent interface).
838 - require Event and EV using configure_requires, to force their existance.
839
8404.13 Wed Oct 24 07:26:45 CEST 2007
841 - add Coro::Storable::blocking_thaw.
842 - use a vastly more complicated technique to localise
843 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
844 - use a coroutine for the idle callback Coro::Event,
845 instead of running Event in the current coroutine context.
846 This also catches recursive invocations.
847 - actually report fork errors in gethostbyname and inet_aton.
848
8494.11 Thu Oct 11 02:40:24 CEST 2007
850 - port to threaded perls.
851
8524.1 Thu Oct 11 02:38:16 CEST 2007
853 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
854 be local to each coro (see Coro::State).
855 - incompatible change: for very deep reasons, cede and cede_notself
856 cannot return anything, so nothing will be returned.
857 - possibly bring back 5.10 compatibility (untested).
858 - work around stupid (and wrong) warning on 5.10 :(.
859 - overlay the saved state over the context stack. This saves
860 a few hundred bytes per coroutine on average and also
861 speeds up context switching a bit.
862 - further tune default stack sizes.
863 - (more) correctly calculate stack usage in coro_rss.
864 - Coro::Storable::blocking_* did not properly lock
865 resulting in races between coroutines.
866 - added Coro::Storable::guard.
867 - stopping to trace a coroutine could destroy the cctx of
868 an unrelated coroutine.
869 - explain the relationship between Perl and C coroutines in
870 more detail in Coro::State.
871 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
872 causing a fork per resolve. This also affected Coro::Socket.
873 - switch to a separate stack in $coro->call/eval to avoid
874 invalidating pointers.
875
8764.03 Sat Oct 6 21:24:00 CEST 2007
877 - added Coro::throw method.
878 - minor code cleanups.
879
8804.02 Sat Oct 6 02:36:47 CEST 2007
881 - fix a very minor per-coroutine memleak (a single codereference).
882 - fixed a bug where the currently in-use c context would be freed
883 prematurely (can happen only when programs change the stacksize
884 or use tracing).
885 - tracing can no longer keep a coro alive after it terminated.
886 - do static branch prediction in the common path for gcc. gives
887 about 2-5% speed improvement here.
888
8894.01 Fri Oct 5 22:10:49 CEST 2007
890 - instead of recreating *a* standard output handle we simply
891 use STDOUT, which is faster and hopefully more robust.
892
8934.0 Fri Oct 5 12:56:00 CEST 2007
894 - incompatibly changed Coro::Storable::freeze.
895 - major new feature: added Coro::Debug, for interactive coroutine
896 debugging, tracing and much more.
897 - major bug fix: unbelievable, but true: $_, $/ and many other
898 "saved" variables actually weren't being saved. This has been fixed,
899 of course, while increasing performance while losing all the save
900 flags.
901 - save flags are gone, and all the api functions dealing with them.
902 - added Coro::Semaphore::adjust.
903 - added Coro::Util::fork_eval.
904 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
905 - added Coro::killall.
906 - reduce initial stack sizes to allow for "micro-coroutines".
907 - better async_pool resource management, moved parts of async_pool
908 handling to XS (major speed improvement).
909 - actually croak before modifying important data structures.
910 - refuse to transfer while compiling.
911 - possibly support eval EXPR better now.
912 - enable assembly per default on linux+bsd x86+amd64.
913 - all internal members were renamed _something for easier subclassing.
914 - many minor tweaks.
915
9163.63 Wed May 16 14:10:06 CEST 2007
917 - implement handcoded assembly for x86/amd64 SVR ABI.
918
9193.62 Fri Apr 27 21:36:06 CEST 2007
920 - upgrade libcoro (which might set unwind info correctly).
921 - change default on linux to setjmp/longjmp.
922
9233.61 Thu Apr 19 12:36:18 CEST 2007
924 - Coro::Storable caused an endless loop when thawing invalid
925 pst's sometimes.
926 - use a Semaphore in Coro::Storable, as Storable doesn't
927 seem to be reentrant (although it is documented to
928 be threadsafe...).
929 - fix Coro::Signal to bring back the original unreliable
930 but stateful semantics.
931 - fixed a lot of typos in Coro.pm (patch submitted by David
932 Steinbrunner, which applied flawlessly).
933
9343.6 Sat Apr 14 17:13:31 CEST 2007
935 - added some bugfixes to get eg/myhttpd working again.
936 - added Coro::Storable for often-cede'ing freeze/thaw.
937 - try to do a clean exit when a coroutine calls exit
938 (EXPERIMENTAL).
939 - got rid of indirect call through _coro_init.
940 - updated the partly antique examples in eg/ to
941 work again and be a bit less magic, too.
942 - fixed Coro::Signal semantics to work as documented again.
943
9443.55 Sun Mar 25 01:20:47 CET 2007
945 - add SAVE_DEFFH to save the default filehandle and enable
946 it by default.
947 - finally move socket-operations from Coro::Socket to Coro::Handle
948 to be able to unblock foreign sockets.
949 - add Coro::State::save_also and guarded_save.
950 - add count accessor to Coro::Semaphore.
951 - add Coro::State::cctx_stacksize.
952 - just for the fun of it, do not rely on implicit context,
953 which can dramatically improve performance, but people
954 using windows-process-emulation perls do not care much about
955 performance.
956
9573.51 Sun Mar 4 14:18:04 CET 2007
958 - fixed a problem when you weakened references to Coro::State's
959 (patch by Yuval Kogman).
960
9613.501 Wed Feb 28 12:44:07 CET 2007
962 - rename some global symbols as macosx from hell redefines
963 them without asking.
964
9653.5 Tue Feb 13 20:22:53 CET 2007
966 - do AnyEvent model detection earlier, avoiding problems
967 caused by first using AnyEvent and later Coro::Event.
968 - implement and document Coro::Event event objects.
969 - fix a potential problem in Coro::Event causing crashes.
970 - initialise PL_comppad when creating a new coroutine,
971 avoids crashes on early coro destruction.
972
9733.41 Mon Jan 22 19:19:49 CET 2007
974 - readline on Coro::Handle did not support undefined $/,
975 nor did it deliver partial lines on EOF or error.
976 - implement malloc fallback for stack allocation because
977 stupid broken idiotic OSX has a stupid broken
978 idiotic fits-the-whole-os mmap "implementation" and
979 my dick feels longer if Coro is portable even to
980 obsolete platforms.
981
9823.4 Fri Jan 19 21:52:54 CET 2007
983 - remove t/09_timer.t, as it isn't really testing much
984 but was rather flaky in practise.
985 - async_pool coro would keep arguments and callback alive until
986 it was reused.
987 - cancellation of a coroutine could cause spurious idle calls
988 in cede_notself.
989
9903.3 Sat Jan 6 03:45:00 CET 2007
991 - implement $coro->on_destroy.
992 - Coro::Event blocking semantics have been changed,
993 documented and - hopefully - improved.
994 - fix nice adding, not subtracting, from priority.
995 - fix ->prio and api_is_ready (patch by Mark Hinds).
996 - fixed an assert ("... == prev__cctx->idle_te")
997 that could errronously trigger.
998 - fix various large and small memleaks.
999 - use a (hopefully) more stable cancel implementation
1000 that immediately frees the coroutine data.
1001 - cede/cede_notself return a status now.
1002 - added Coro::guard function.
1003 - added a global coroutine pool for jobs (on my machine,
1004 I can create and execute 48k simple coros/s with async,
1005 and 128k coros with async_pool).
1006 - Coro::AIO now uses the coroutine priority as io priority.
1007
10083.2 Fri Dec 22 05:07:09 CET 2006
1009 - improve portability to slightly older perls.
1010 - use cleaner coroutine destruction.
1011 - simplify configuration for users.
1012 - optionally (unrecommended) prefer perl functions over
1013 their coro replacements.
1014
10153.11 Tue Dec 5 13:11:24 CET 2006
1016 - fixed some bogus assert's, but as perl.h disables assert even
1017 without NDEBUG (thank you very much), not too many people should
1018 notice (that did include myself). Andreas König noticed, though :)
1019 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
1020 it. Also noticed by Andreas König :)
1021 - save/restore tainted status.
1022 - verified to pass the testsuite on my 5.9.5.
1023
10243.1 Mon Dec 4 23:03:40 CET 2006
1025 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
1026 - incompatible change: transfer flags are now per-state.
7 - give a better error message on deadlock. 1027 - give a better error message on deadlock.
1028 - document Coro::nready.
8 - enhanced testsuite. 1029 - enhanced testsuite.
9 1030
103.01 Sun Dec 3 23:47:42 CET 2006 10313.01 Sun Dec 3 23:47:42 CET 2006
11 - forgot to include Coro::Timer. 1032 - forgot to include Coro::Timer.
12 1033

Diff Legend

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