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

Comparing Coro/Changes (file contents):
Revision 1.562 by root, Tue May 10 19:55:48 2011 UTC vs.
Revision 1.598 by root, Mon Oct 8 23:16:15 2012 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: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx
6TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 7
86.10 Tue Oct 9 01:14:27 CEST 2012
9 - updated ecb.h, it had a typo that caused it to not compile on many
10 big endian systems (reported by many people).
11 - disable memory fences in ecb.h to improve portability.
12
136.09 Sat Oct 6 23:25:02 CEST 2012
14 - Coro::EV I/O watchers were not interruptible by exceptions
15 (Coro::State::throw) (testcase by sten).
16 - ->throw now puts threads into the ready queue, as this seems to
17 be expected by existing code, and code that doesn't cope with spurious
18 wakeups needs fixing anyway.
19 - use fd -1 in mmap.
20 - cast I32 to int in error message printf.
21 - warn about broken so-called "hardened" kernels.
22
236.08 Fri Apr 13 12:05:47 CEST 2012
24 - be more aggressive about exiting like perl does - formerly,
25 exiting from the non-main thread would not execute END blocks.
26
276.07 Fri Nov 11 21:21:48 CET 2011
28 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
29 - update ecb.h.
30
316.06 Mon Aug 8 23:59:48 CEST 2011
32 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
33 backend, which is an order of magnitude slower. unfortunately, cygwins
34 pthread implementation isn't very complete either, so allocate the stack
35 twice just to be sure.
36 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
37 in its ithreaded perl - best recompile cygwin and use the 'w'indows
38 backend for much better performance. also disable ithreads for
39 even better performance...).
40
416.05 Thu Aug 4 21:36:36 CEST 2011
42 - blush, condvar values would not be propagated from send to recv anymore
43 (reported by Chip Salzenberg).
44 - use exponential increase for the readline buffer length in
45 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
46
476.04 Wed Aug 3 17:15:45 CEST 2011
48 - use even more efficient and more compatible condvars for
49 compatibility to AnyEvent 6.x :)
50 - more inconsequential ecb.h updates.
51
526.03 Wed Aug 3 11:41:30 CEST 2011
53 - change how Coro patches AnyEvent condvars for compatibility to
54 AnyEvent 6.x.
55 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
56 hasn't been included.
57
586.02 Wed Jul 13 04:35:19 CEST 2011
59 - "improve portability to Gentoo" - gentoo manages to put perl variables
60 in memory areas that are farther than 2gb apart, which the jit couldn't
61 handle and barfed. now it's just a bit slower on gentoo and similar
62 systems.
63
646.01 Sun Jul 3 12:31:14 CEST 2011
65 - workarounds are good, but the test for whether pthreads are used
66 was not good. thisone should be better.
67 - check differently whether gcc generates cfi instructions itself.
68
696.0 Wed Jun 29 19:43:35 CEST 2011
70 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
71 destroyed and cleaned up automatically (e.g. async { schedule }).
72 - implement a JIT compiler for part of the thread switch code,
73 which gives a 50% speed improvement on threaded perls, and
74 about 4% on non-thraeded perls (so threaded perls now finally
75 reach about half the speed of non-threaded perls).
76 - slightly modernise Coro::Intro, add section about rouse functions.
77 - avoid DEFSV and ERRSV, giving another 10% improvement
78 in thread switching.
79 - Coro::State->is_destroyed is now called is_zombie.
80 - implement a Coro->safe_cancel method that might fail, but
81 cancels in a "safer" way if it succeeds.
82 - add preliminary support for DEBUGGING perls.
83 - get rid of two hash-accesses when initialising a new Coro - this
84 speeds up coro creation by almost a factor of two.
7 - croak when a coro that is being cancelled tries to block 85 - croak when a coro that is being cancelled tries to block
8 (e.g. while executing a guard block), instead of crashing or 86 (e.g. while executing a guard block), instead of crashing or
9 deadlocking. 87 deadlocking.
10 - implement a Coro->safe_cancel method that might fail, but 88 - use a more robust and also faster method to identify Coro::State
11 cancels in a "safer" way if it succeeds. 89 objects - speeds up everything a bit.
12 - implement Coro->cancel in XS for a 20% speed improvement, and to 90 - implement Coro->cancel in XS for a 20% speed improvement, and to
13 be able to implement mutual cancellation. 91 be able to implement mutual cancellation.
14 - speed up context switches by a percent or two by more efficiently 92 - speed up context switches by a percent or two by more efficiently
15 allocating context stack entries. 93 allocating context stack entries.
94 - implement Coro->join and Coro->on_destroy in XS for a speedup and
95 a reduction in memory use.
16 - cancelling a coro while it itself is cancelling another coro is 96 - cancelling a coro while it itself is cancelling another coro is
17 now supported and working, instead of triggering an assertion. 97 now supported and working, instead of triggering an assertion.
18 - be a bit more crash-resistant when calling (buggy) on_destroy 98 - be a bit more crash-resistant when calling (buggy) on_destroy
19 callbacks (best effort). 99 callbacks (best effort).
20 - move on_destroy into the slf_frame, to allow extension slf 100 - move on_destroy into the slf_frame, to allow extension slf
21 functions to have destructors. 101 functions to have destructors.
102 - get rid if coro refcounting - simply crash in other interpreter
103 threads by nulling the pointers on clone.
104 - simplify warn/die hook handling when loading Coro - the convoluted
105 logic seems to be no longer neccessary.
106 - use libecb instead of our own home-grown gcc hacks.
107 - document alternatives to Coro::LWP. Please use them :)
108 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
109 sigaltstack. Really. wine suffers from it, erlang suffers from it,
110 and it's known since at least 2006.
22 111
235.372 Wed Feb 23 06:14:30 CET 2011 1125.372 Wed Feb 23 06:14:30 CET 2011
24 - apparently mingw doesn't provide a working gettimeofday, try to 113 - apparently mingw doesn't provide a working gettimeofday, try to
25 work around that by relying on Time::HiRes (indirectly brought to 114 work around that by relying on Time::HiRes (indirectly brought to
26 my attention by Max Maischein). 115 my attention by Max Maischein).

Diff Legend

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