Re: quartz: Fix deadlock with transform filter, try 2

[Michael Karcher]

Am Montag, den 07.07.2008, 18:22 -0700 schrieb Maarten Lankhorst:
> Reworks the code so that the lock is taking in the called procedures now.
> - InputPin *pin = (InputPin *)pTransformFilter->ppPins[0];
> -
> - EnterCriticalSection(&pTransformFilter->csFilter);
> - if (pTransformFilter->state == State_Stopped)
> - {
> -
>    LeaveCriticalSection(&pTransformFilter->csFilter);
> -
>    return VFW_E_WRONG_STATE;
> - }

I have to admit that I don't know anything about quartz. Are you sure
you may access pTransformFilter->ppPins[0] before taking the lock?

Regards,
  Michael Karcher

quartz: Fix deadlock with transform filter, try 2

[Maarten Lankhorst]

Hi Michael,

2008/7/8 Michael Karcher :
> Am Montag, den 07.07.2008, 18:22 -0700 schrieb Maarten Lankhorst:
> > Reworks the code so that the lock is taking in the called procedures now.
> > - InputPin *pin = (InputPin *)pTransformFilter->ppPins[0];
> > -
> > - EnterCriticalSection(&pTransformFilter->csFilter);
> > - if (pTransformFilter->state == State_Stopped)
> > - {
> > -
> >    LeaveCriticalSection(&pTransformFilter->csFilter);
> > -
> >    return VFW_E_WRONG_STATE;
> > - }
>
> I have to admit that I don't know anything about quartz. Are you sure
> you may access pTransformFilter->ppPins[0] before taking the lock?

It's safe, pins are allocated during construction, and the pointers
themselves are not modified since. Doing operations on the pin
requires the lock though.

Cheers,
Maarten.