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

Comparing Coro/Changes (file contents):
Revision 1.538 by root, Fri May 7 20:28:30 2010 UTC vs.
Revision 1.596 by root, Sat Oct 6 21:25:24 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: unready_all
5TODO: myhttpd header parsing
6TODO: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
7TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx
6TODO: __GCC_HAVE_DWARF2_CFI_ASM
8 7
86.09 Sat Oct 6 23:25:02 CEST 2012
9 - Coro::EV I/O watchers were not interruptible by exceptions
10 (Coro::State::throw) (testcase by sten).
11 - ->throw now puts threads into the ready queue, as this seems to
12 be expected by existing code, and code that doesn't cope with spurious
13 wakeups needs fixing anyway.
14 - use fd -1 in mmap.
15 - cast I32 to int in error message printf.
16 - warn about broken so-called "hardened" kernels.
17
186.08 Fri Apr 13 12:05:47 CEST 2012
19 - be more aggressive about exiting like perl does - formerly,
20 exiting from the non-main thread would not execute END blocks.
21
226.07 Fri Nov 11 21:21:48 CET 2011
23 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
24 - update ecb.h.
25
266.06 Mon Aug 8 23:59:48 CEST 2011
27 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
28 backend, which is an order of magnitude slower. unfortunately, cygwins
29 pthread implementation isn't very complete either, so allocate the stack
30 twice just to be sure.
31 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
32 in its ithreaded perl - best recompile cygwin and use the 'w'indows
33 backend for much better performance. also disable ithreads for
34 even better performance...).
35
366.05 Thu Aug 4 21:36:36 CEST 2011
37 - blush, condvar values would not be propagated from send to recv anymore
38 (reported by Chip Salzenberg).
39 - use exponential increase for the readline buffer length in
40 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
41
426.04 Wed Aug 3 17:15:45 CEST 2011
43 - use even more efficient and more compatible condvars for
44 compatibility to AnyEvent 6.x :)
45 - more inconsequential ecb.h updates.
46
476.03 Wed Aug 3 11:41:30 CEST 2011
48 - change how Coro patches AnyEvent condvars for compatibility to
49 AnyEvent 6.x.
50 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
51 hasn't been included.
52
536.02 Wed Jul 13 04:35:19 CEST 2011
54 - "improve portability to Gentoo" - gentoo manages to put perl variables
55 in memory areas that are farther than 2gb apart, which the jit couldn't
56 handle and barfed. now it's just a bit slower on gentoo and similar
57 systems.
58
596.01 Sun Jul 3 12:31:14 CEST 2011
60 - workarounds are good, but the test for whether pthreads are used
61 was not good. thisone should be better.
62 - check differently whether gcc generates cfi instructions itself.
63
646.0 Wed Jun 29 19:43:35 CEST 2011
65 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
66 destroyed and cleaned up automatically (e.g. async { schedule }).
67 - implement a JIT compiler for part of the thread switch code,
68 which gives a 50% speed improvement on threaded perls, and
69 about 4% on non-thraeded perls (so threaded perls now finally
70 reach about half the speed of non-threaded perls).
71 - slightly modernise Coro::Intro, add section about rouse functions.
72 - avoid DEFSV and ERRSV, giving another 10% improvement
73 in thread switching.
74 - Coro::State->is_destroyed is now called is_zombie.
75 - implement a Coro->safe_cancel method that might fail, but
76 cancels in a "safer" way if it succeeds.
77 - add preliminary support for DEBUGGING perls.
78 - get rid of two hash-accesses when initialising a new Coro - this
79 speeds up coro creation by almost a factor of two.
80 - croak when a coro that is being cancelled tries to block
81 (e.g. while executing a guard block), instead of crashing or
82 deadlocking.
83 - use a more robust and also faster method to identify Coro::State
84 objects - speeds up everything a bit.
85 - implement Coro->cancel in XS for a 20% speed improvement, and to
86 be able to implement mutual cancellation.
87 - speed up context switches by a percent or two by more efficiently
88 allocating context stack entries.
89 - implement Coro->join and Coro->on_destroy in XS for a speedup and
90 a reduction in memory use.
91 - cancelling a coro while it itself is cancelling another coro is
92 now supported and working, instead of triggering an assertion.
93 - be a bit more crash-resistant when calling (buggy) on_destroy
94 callbacks (best effort).
95 - move on_destroy into the slf_frame, to allow extension slf
96 functions to have destructors.
97 - get rid if coro refcounting - simply crash in other interpreter
98 threads by nulling the pointers on clone.
99 - simplify warn/die hook handling when loading Coro - the convoluted
100 logic seems to be no longer neccessary.
101 - use libecb instead of our own home-grown gcc hacks.
102 - document alternatives to Coro::LWP. Please use them :)
103 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
104 sigaltstack. Really. wine suffers from it, erlang suffers from it,
105 and it's known since at least 2006.
106
1075.372 Wed Feb 23 06:14:30 CET 2011
108 - apparently mingw doesn't provide a working gettimeofday, try to
109 work around that by relying on Time::HiRes (indirectly brought to
110 my attention by Max Maischein).
111 - fix some portability issues when Time::HiRes was used.
112
1135.371 Mon Feb 21 14:36:08 CET 2011
114 - backport to windows process emulation code again.
115
1165.37 Sat Feb 19 07:49:44 CET 2011
117 - add a big "Coro thread life cycle" section to "man Coro".
118 - try a tentative workaround against the breakage that 5.13 has
119 introduced without depreciation period. sigh.
120 - no longer use Time::HiRes if gettimeofday is available, which
121 saves quite a lot of memory.
122
1235.36 Sun Feb 13 05:33:41 CET 2011
124 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
125 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
126 method call.
127 - undocument Coro::Timer::sleep and obsolete whole module.
128 - optimise Coro::Timer::timeout memory and cpu usage.
129 - slightly updated Coro::Intro for recent changes.
130 - do not initialise PL_dirty anymore.
131
1325.25 Thu Nov 11 01:08:39 CET 2010
133 - try a different approach on netbsd - netbsd 5 finally has marginally
134 working pthreads, but still broken ucontext/sigaltstack.
135 - openbsd 4.8 finally got their act together, Coro works out of the box
136 with asm, setjmp and pthreads (no change, just informational).
137
1385.24 Sat Oct 23 11:27:12 CEST 2010
139 - port to the EV 4.0 API.
140 - work around bugs in mingw32, making strawberry perl work
141 out of the box.
142 - correctly modify Coro::AIO function prototypes
143 so that they reflect the "no optional parameters" rule.
144 - "ported" libcoro to C++.
145
1465.23 Mon May 17 18:50:42 CEST 2010
9 - be more resistant to ordering changes when initialising 147 - be more resistant to ordering changes when initialising
10 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias 148 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
11 Waldorf). 149 Waldorf).
12 - document that perl 5.12 deliberately removed support for cloning. 150 - document that perl 5.12 deliberately removed support for cloning.
13 151

Diff Legend

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