Issue with gettimeofday() in RTEMS 4.10;

[Gedare Bloom]

I wonder if this is this related to PR 1495?:
https://www.rtems.org/bugzilla/show_bug.cgi?id=1495

The symptom is similar, but I don't think the LEON3 has the same
structure of clock driver.

-Gedare

On Fri, Jan 28, 2011 at 9:11 AM, João Rasta <freakforever at gmail.com> wrote:
> Hi,
>
> I'm using gettimeofday() to profile a function. For this i do:
>
>         struct timeval tim;
>         // *Get actual clock
>         gettimeofday(&tim, NULL);
>         double t1=tim.tv_sec+(tim.tv_usec/1000000.0);
>
>
>         // *GNC Calculations starting point
>         Execute_gnc();
>         // *Calculations termination point
>
>
>         // *Calculate elapsed time with GNC calculations.
>         gettimeofday(&tim, NULL);
>         double t2=tim.tv_sec+(tim.tv_usec/1000000.0);
>
>         procDelay = (t2 - t1 + 0.0275);
>         if (t1 > t2)
>             printf(" WARNING %f > %f\n",t1,t2);
>         else
>             printf("%f\n",procDelay);
>
> But sometimes t1 is larger than t2, which should not be possible since t2 is
> allways measured after t1. For instance, the above code produces the
> following when this issue happens:
>
> WARNING 567996584.559981 > 567996584.550001
>
> The above code runs in loop and this happens roughly once in an hour (just
> an estimate). Is there any known overflow issue with the usec counter that i
> should be aware? RTEMS is running with a LEON-3.
>
>
> Best,
> JM
>
> _______________________________________________
> rtems-users mailing list
> rtems-users at rtems.org
> http://www.rtems.org/mailman/listinfo/rtems-users
>
>