Re: [LTP] [PATCH] sem_post:8-1:sleep to make sure child_3 get lock after child_2

[Wanlong G. <gao...@cn...>]

On 08/12/2011 07:59 PM, Cyril Hrubis wrote:
> Hi!
>>
>> If with no sleep, child_3 will preempt and get the lock before child_2
>> some time. It may fail the test case.
>>
>> Signed-off-by: Wanlong Gao<gao...@cn...>
>> ---
>>   .../conformance/interfaces/sem_post/8-1.c          |    3 ++-
>>   1 files changed, 2 insertions(+), 1 deletions(-)
>>
>> diff --git a/testcases/open_posix_testsuite/conformance/interfaces/sem_post/8-1.c b/testcases/open_posix_testsuite/conformance/interfaces/sem_post/8-1.c
>> index 165b094..37ed8a3 100644
>> --- a/testcases/open_posix_testsuite/conformance/interfaces/sem_post/8-1.c
>> +++ b/testcases/open_posix_testsuite/conformance/interfaces/sem_post/8-1.c
>> @@ -174,6 +174,7 @@ int main()
>>   	}
>>   	fprintf(stderr, "P: child_2: %d forked\n", c_2);
>>
>> +	sleep(1);
>>   	/* Make sure the two children has been waiting */
>>   	/*do {
>>   		sleep(1);
>> @@ -254,4 +255,4 @@ clean_up:
>>   	sem_unlink(semname);
>>   	sem_unlink(semname_1);
>>   	return retval;
>> -}
>> \ No newline at end of file
>
> While this makes the probability of preempting smaller, the test still
> would fail under some cirscumstances. The correct solution would be
> check the semafore for number of waiting processes. And the code is
> allready there but was commented out some time ago and I wonder why...
It's introduced by commit f102fd9735f01b4e2c9ae44f3c90d1a56d4d81f8
Author: Rishikesh K Rajak <ris...@li...>
Date:   Tue Mar 23 14:20:36 2010 +0530

But now it seems not working fine....

-- 
Thanks
Wanlong Gao