From: Chad Fowler <chad@...>
Date: 2004-01-01T12:54:11+09:00
Subject: Re: sleep is broken with --enable-pthread [Was: Bad interaction between timeout.rb and --enable-pthread]

Nathaniel Talbott wrote:

> I should have reduced it more before posting...
>
> On Dec 31, 2003, at 15:34, Nathaniel Talbott wrote:
>
>> Here's a testcase that shows the problem:
>>
>>   require 'test/unit'
>>
>>   class TC < Test::Unit::TestCase
>>     def test_time
>
>
>         Thread.start{}
>
>>       t = Time.now
>>       sleep(5)
>>       assert_in_delta(5, Time.now.to_i - t.to_i, 1)
>>     end
>>   end
>
>
> So the problem is much more fundamental: sleep seems to not work at 
> all once a Ruby thread has been run if --enable-pthread is used. Any 
> ideas?
>
> Thanks,
>
>
> Nathaniel
>
> <:((><
>

Looks like the select() call in rb_thread_wait_for is being interrupted 
early with errno == EINT, but only if a Thread has been started prior to 
the call to sleep(). 

This appears to be caused by matz's change on Dec. 13 to eval.c (1.605) 
in response to [ruby-talk:87519].

That's as far as I can go I guess.  I don't know how to solve the 
problem.  I wish I understood this all better.  Anyone want to do a 
brief explanation after the bug is fixed? ;)

Happy New Year,
Chad