Re: [PATCH v2 3/4] nptl: Add pthread_clockjoin_np

[Adhemerval Zanella <adhemerval dot zanella at linaro dot org>]

On 28/09/2019 16:45, Mike Crowe wrote:
> On Thursday 26 September 2019 at 16:17:53 -0700, Adhemerval Zanella wrote:
>>> diff --git a/nptl/tst-join3.c b/nptl/tst-join3.c
>>> index a4ae459..c06e65a 100644
>>> --- a/nptl/tst-join3.c
>>> +++ b/nptl/tst-join3.c
>>> @@ -28,6 +28,8 @@
>>>  #include <support/xtime.h>
>>>  
>>>  
>>> +#define CLOCK_USE_TIMEDJOIN (-1)
>>> +
>>>  static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
>>>  
>>>  
>>> @@ -35,19 +37,23 @@ static void *
>>>  tf (void *arg)
>>>  {
>>>    xpthread_mutex_lock (&lock);
>>> +  xpthread_mutex_unlock (&lock);
>>>  
>>
>> Maybe it would be better to move the unlock after the loop test, 
>> to mimic how the other unlock does?
> 
> Do you mean that I should unlock the mutex after the loop inside
> do_test_clock? If so, that would mean unlocking it on a different thread
> and that doesn't sound like it's guaranteed to be safe.

Yeah, but it is not really a strong opinion.  It just make the patch
change less code in the original test.  

> 
> Or perhaps you meant something different?
> 
> (The use of a mutex for this test works, but it's non-obvious. Perhaps a
> semaphore would be clearer?)
> 
> Thanks for the rest of the review. I will make the changes you describe.
> 
> Mike.
>