[Bug 687183] playbin: caps negotation if VIDEO_FORMAT is unknown

[GStreamer (bugzilla.gnome.org)]

https://bugzilla.gnome.org/show_bug.cgi?id=687183
  GStreamer | gst-plugins-base | git

--- Comment #23 from Sebastian Dröge <slomo at circular-chaos.org> 2013-06-11 19:01:07 UTC ---
(In reply to comment #21)
> Hi, I think I also missed some notifications. Anyway, you should not assume
> anything about the underlying VA surface format. Never. You cannot know it (no
> API for it), and you may not be able to actually retrieve anything from it. It
> just serves as an "opaque" type. e.g. some VA surfaces can be encrypted. Once
> you decoded to it, you cannot read from it (but only the GPU for rendering).
> 
> I am not really concerned about vaapidecode + download use cases, because I
> don't actually see any sensible one but for debugging purposes (validation).
> Thus, I don't really want to expose that.

Well, it's something you should really expose to conform with how things are
supposed to work in GStreamer :) It will be slow and not always work, but it
should work if possible.

> In short, it's better to stick to the current model: you allocate a decode
> session (VA context) once, and it is live until the encoded resolution changes,
> thus requiring new VA surfaces. Though, in order to initialize that context,
> you will also need to know the maximum number of possible VA surfaces to
> allocate, i.e. parsing at least the sequence packets/layer.

Could you maybe just allocate inside the decoder? That sounds like a safer
solution with these constraints...

-- 
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.