[RPL/2] still a bug in multitasking

[BERTRAND Joël]

BERTRAND Joël wrote:
> DEMAINE Benoit-Pierre wrote:
>> Playing with an old example attached:
>>
>> dhp at moon_gen_2:~/documents/programmation/rpl$ rpl -scd example_smp.rpl
>> +++RPL/2 version 4.00pre8w (Tuesday 07/10/07, 23:57:24 CEST)
>> +++Copyright (C) 1989 to 2006, 2007 BERTRAND Jol
>> 99 bottles of beer on the wall,
>> 99 bottles of beer.
>> Take one down, pass it around,
>> 98 bottles of beer on the wall.
>>
>> [...]
>>
>> 34 bottles of beer on the wall,
>> 34 bottles of beer.
>> Take one down, pass it around,
>> 33 bottles of beer on the wall.
>>
>> +++System : Process error
>>
>> +++Information : Writing rpl-core file
>> +++Information : Done
>>
>> It may fail and stop at various places: 99, 33, 34, 34, 29, 47, 41 ...
> 
> 	Thanks for report. Strange, I never saw this bug before, but I can 
> reproduce it. I think that this bug comes from the new futex management 
> in kernel. I will investigate.

	I have found this bug, due to the new implementation of POSIX threads. 
I think this bug didn't occur with old pthreads libs. Child process ends 
before it has flushed its pipes to return value to its father.

	Regards,

	JKB