[PATCH v2] Test for clock_nanosleep with CLOCK_MONOTONIC option.

[Utkarsh Rai]

Okay, so from what I could gather the time between the two gettime calls
can exceed 1 sec if it is preempted by another process in between. Is my
line of thought correct?

On Wed, Apr 15, 2020 at 6:01 PM Sebastian Huber <
sebastian.huber at embedded-brains.de> wrote:

>
> On 15/04/2020 14:29, Utkarsh Rai wrote:
>
> On Wed, Apr 15, 2020 at 5:35 PM Sebastian Huber <
> sebastian.huber at embedded-brains.de> wrote:
>
>> On 15/04/2020 14:02, Utkarsh Rai wrote:
>>
>> > +  status = clock_gettime( CLOCK_MONOTONIC, &end_time );
>>> > +  rtems_test_assert( status == 0 );
>>> > +
>>> > +  rtems_test_assert( (end_time.tv_sec-init_time.tv_sec) == 0 );
>>>
>>> Is end_time.tv_sec - init_time.tv_sec == 0 under all circumstances?
>>>
>>
>> My idea was to check for a 1ns delay with a reasonable amount of
>> overhead, hence I checked for  end_time.tv_sec - init_time.tv_sec == 0.
>>
>> Exists there a value of init_time for which end_time.tv_sec !=
>> init_time.tv_sec and still 1ns elapsed?
>>
>
> Sorry, maybe I am confused in my concept, kidly help me out. I want to
> produce a 1ns delay, so I make a call to clock_nanosleep with flag value as
> 0 (to sleep for specified time) and the delay being 1ns. I recorded the
> time before the sleep call and after the sleep call. Now, I want to check
> if the delay produced was actually 1ns with a reasonable overhead, my
> assumption for an unreasonable overhead was that if I specify a delay of
> 1ns
>
> Up to here everything is fine.
>
> and I get a delay in seconds, it would be an error.
>
> Think about this once more.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20200415/308751d9/attachment.html>