[#45426] [ruby-trunk - Feature #6546][Open] Net::HTTP to check for HTTP_PROXY environment setting. — "dekz (Jacob Evans)" <dekzter@...>

14 messages 2012/06/04

[#45431] [ruby-trunk - Bug #6548][Open] Rake doesn't ignore arguments after -- — "rosenfeld (Rodrigo Rosenfeld Rosas)" <rr.rosas@...>

12 messages 2012/06/05

[#45474] [ANN] Request for "slide-show" of your feature proposal — Yusuke Endoh <mame@...>

(Japanese later; 日本語が後にあります)

18 messages 2012/06/07

[#45563] [ruby-trunk - Bug #6573][Open] Webrick test failures — "bkabrda (Bohuslav Kabrda)" <bkabrda@...>

19 messages 2012/06/11

[#45567] [ruby-trunk - Bug #6575][Open] Thread#kill sets rb_errinfo() to Fixnum 8 after rb_protect(function, data, &error_tag) — ibc (Iñaki Baz Castillo) <ibc@...>

9 messages 2012/06/11

[#45647] [ruby-trunk - Bug #6592][Open] test_call_double(DL::TestDL) fails on ARM HardFP — "vo.x (Vit Ondruch)" <v.ondruch@...>

15 messages 2012/06/14

[#45657] [ruby-trunk - Feature #6594][Open] Integrated Functor — "trans (Thomas Sawyer)" <transfire@...>

20 messages 2012/06/15

[#45664] [ruby-trunk - Bug #6596][Open] New method for Arrays : Array#index — "robin850 (Robin Dupret)" <robin.dupret@...>

20 messages 2012/06/15

[#45694] [ruby-trunk - Feature #6602][Open] Tail call optimization: enable by default? — "ko1 (Koichi Sasada)" <redmine@...>

12 messages 2012/06/18

[#45715] [ruby-trunk - Feature #6609][Open] Toplevel as self extended module — "trans (Thomas Sawyer)" <transfire@...>

17 messages 2012/06/19

[#45732] [ruby-trunk - Bug #6614][Open] GC doesn't collect objects bound to (collectable) proc — "rogerdpack (Roger Pack)" <rogerpack2005@...>

9 messages 2012/06/20

[#45733] [ruby-trunk - Feature #6615][Open] Release GVL in zlib when calling inflate() or deflate() — "drbrain (Eric Hodel)" <drbrain@...7.net>

12 messages 2012/06/21

[#45735] [ruby-trunk - Bug #6616][Open] MinGW: cannot build extensions or run tests due changes in exec_arg? — "luislavena (Luis Lavena)" <luislavena@...>

9 messages 2012/06/21

[#45798] [ruby-trunk - Bug #6634][Open] Deadlock with join and ConditionVariable — "meh. (meh. I don't care)" <meh@...>

20 messages 2012/06/23

[#45805] [ruby-trunk - Feature #6636][Open] Enumerable#size — "marcandre (Marc-Andre Lafortune)" <ruby-core@...>

15 messages 2012/06/23

[#45864] [ruby-trunk - Bug #6647][Open] Exceptions raised in threads should be logged — "headius (Charles Nutter)" <headius@...>

71 messages 2012/06/25

[#45902] [ruby-trunk - Bug #6653][Open] 1.9.2/1.9.3 exhibit SEGV with many threads+tcp connections — "erikh (Erik Hollensbe)" <erik@...>

11 messages 2012/06/26

[#45960] [ruby-trunk - Feature #6669][Open] A method like Hash#map but returns hash — "yhara (Yutaka HARA)" <redmine@...>

18 messages 2012/06/29

[#45963] [ruby-trunk - Feature #6670][Open] str.chars.last should be possible — "yhara (Yutaka HARA)" <redmine@...>

36 messages 2012/06/29

[#46021] [ruby-trunk - Feature #6679][Open] Default Ruby source file encoding to utf-8 — "claytrump (Clay Trump)" <clay.trump@...>

21 messages 2012/06/30

[ruby-core:45950] Re: [ruby-trunk - Feature #3620] Add Queue, SIzedQueue and ConditionVariable implementations in C in addition to ruby ones

From: SASADA Koichi <ko1@...>
Date: 2012-06-29 08:41:41 UTC
List: ruby-core #45950
Hi,

(2012/06/26 16:18), funny_falcon (Yura Sokolov) wrote:
> It seems that there is no need for mutex in a native queue implementation (considering we have GVL),
> and so that queue_synchronized wrapper:
> rb_mutex_sleep could be replaces with rb_thread_sleep_forever without semantic change.
> 
> Without mutex, native queue becomes 2 times faster.
> Modified patch is attached.

I found a bug.

### begin sample code
q = Thread::Queue.new
#q = Queue.new
th1 = Thread.new{
  begin
    p [:th1, q.pop]
  rescue RuntimeError => e
    sleep
    p e
  end
}
th2 = Thread.new{
  sleep 0.1
  p [:th2, q.pop]
}
p [th1, th2]
sleep 0.5
th1.raise "async interrupt!"
sleep 0.5
q << :s
# th1.join
p [th1, th2]
th2.join            # BLOCK forever!
### end sample code

When th1 escapes from blocking by "pop" by exception, then waiting list
of Queue should be maintained (remove th1 from waiting list).


Other comment:

* "extthread" is good name?

* variable name

static void
wakeup_all_threads(VALUE list)
{
    VALUE thread, list0 = list;
    long i;

    list = rb_ary_subseq(list, 0, LONG_MAX);
    rb_ary_clear(list0);
    for (i = 0; i < RARRAY_LEN(list); ++i) {
	thread = RARRAY_PTR(list)[i];
	rb_thread_wakeup_alive(thread);
    }
    RB_GC_GUARD(list);
}

I prefer:

static void
wakeup_all_threads(VALUE list0)
{
    VALUE thread, list = rb_ary_subseq(list0, 0, LONG_MAX);
    long i;

    rb_ary_clear(list0);
    for (i = 0; i < RARRAY_LEN(list); ++i) {
	thread = RARRAY_PTR(list)[i];
	rb_thread_wakeup_alive(thread);
    }
    RB_GC_GUARD(list);
}

But I prefer more:

static void
wakeup_all_threads(VALUE list)
{
    VALUE thread;
    long i;

    for (i = 0; i < RARRAY_LEN(list); ++i) {
	thread = RARRAY_PTR(list)[i];
	rb_thread_wakeup_alive(thread);
    }
    rb_ary_clear(list);
}

Any reason to dup array before iteration?

* T_DATA -> T_STRUCT or T_OBJECT

In this case, you don't need to use T_DATA.  You can only use T_STRUCT
or T_OBJECT (with hidden attr).  Maybe it will be simple.


Thanks,
Koichi


> ----------------------------------------
> Feature #3620: Add Queue, SIzedQueue and ConditionVariable implementations in C in addition to ruby ones
> https://bugs.ruby-lang.org/issues/3620#change-27471

-- 
// SASADA Koichi at atdot dot net



In This Thread

Prev Next