Re: [PATCH 1/5] quartz: Rather than assert, continue when IAsyncReader::WaitForNext is flushing and break when it actually times out.

[Alexandre Julliard]

Erich Hoover <ehoover(a)mines.edu> writes:
> @@ -479,13 +488,27 @@ static void PullPin_Thread_Process(PullPin *This)
>          else
>          {
>              /* FIXME: This is not well handled yet! */
> -            ERR("Processing error: %x\n", hr);
>              if (hr == VFW_E_TIMEOUT)
>              {
> -                assert(!pSample);
> +                struct timeval tve;
> +
> +                /*
> +                 * Calculate the time elapsed in IAsyncReader::WaitForNext
> +                 * in order to determine whether a timeout actually occurred
> +                 * or whether the pin is flushing.
> +                 */
> +                timersub(&tv2, &tv1, &tve);
> +                if (!timercmp(&tve, &tvw, <) && pSample != NULL)
> +                {
> +                    ERR("Non-NULL sample returned with VFW_E_TIMEOUT.\n");
> +                    break;
> +                }

That's really ugly. Surely there's a better way to do this than
measuring the elapsed time.

-- 
Alexandre Julliard
julliard(a)winehq.org