[mythtv] Transcode bug: need help from a mythplayer expert

Paul Gardiner lists at glidos.net
Sat Aug 17 14:29:22 UTC 2013


On 06/08/2013 10:02, Paul Gardiner wrote:
> I've finally tracked down the cause of some strange transcoding
> results I've noticed for some time. I wont go into the details
> of the strange result, but the cause is a problem to do with
> the use of TranscodeFrameQueue. This queues frames, holding
> the buffer pointers originating from VideoBuffers. Obviously
> it is important that buffers, while held in the queue, aren't reused.
> That is supposed to be taken care of by the main body of
> transcode calling DoneDisplayingFrame only after it has finished
> with a frame, but that seems to be failing during a seek. I've
> verified this by saving out frame data, along with frame numbers
> around a seek. I do this just before the call to DoneDisplayingFrame.
> I see images from after the seek position appearing in the frames
> numbered before the seek.
>
> My guess is that during a seek, DiscardFrames is called and that's
> allowing the frames to be reused, even though DoneDisplayingFrame
> hasn't been called. I'd like to fix this, but I could do with some
> advice on how best to do so. videobuffers.cpp mentions some special
> buffering states: perhaps they could be used.

I've fixed it. See Ticket #11759

P.


More information about the mythtv-dev mailing list