[Bug 726107] Drain pipeline to support adaptive streaming scenarios and partially fix gpu resource leaks

[GStreamer (bugzilla.gnome.org)]

https://bugzilla.gnome.org/show_bug.cgi?id=726107
  GStreamer | gst-omx | git

--- Comment #10 from Nicolas Dufresne <nicolas.dufresne at collabora.co.uk> 2014-03-25 17:35:59 UTC ---
(In reply to comment #8)
> There a more complications with a new oob event to reclaim buffers.
> 
> If the new event is non-serialized, it would be very racy if we don't do
> something clever. You would need to make a start/stop event, like how the
> flushing works or you would need to try to serialize in the handler (a bit like
> how the logic of flush-stop works). It needs more thinking..

I did foresee that. Though I was to propose to retract on doing it out of band.
My idea is that the event should hold the buffer pool it's aiming at. If an
element does not know what buffer pool it's using, it would just drain, but as
soon as an element knows it would end the chain. This would return sooner and
increase the chance we don't loose frames.

Also, for threaded decoders, it would make more sense to run the reclaim (or
drain) from the thread the push buffers out. This way you can start queuing
frame for decoding sooner.

-- 
Configure bugmail: https://bugzilla.gnome.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.