[Bug 743326] flvdemux: create pads as soon as we can.

[GStreamer (bugzilla.gnome.org)]

https://bugzilla.gnome.org/show_bug.cgi?id=743326
  GStreamer | gst-plugins-good | unspecified

--- Comment #3 from Mathieu Duponchelle <mathieu.duponchelle at epitech.eu> 2015-01-26 16:50:30 UTC ---
(In reply to comment #2)
> I see the bug and we should fix it.

Indeed :)

> How was this handled before flowcombiner?

The fields audio_linked and video_linked were used.

> What would happen if the headers advertise Video + Audio but we never get data
> for one of them? It seems like that pad would never be exposed and would
> prevent a NOT_LINKED return as it would always be on OK.

That's what I figure will happen, I don't have such broken flvs, but there's a
comment in the code saying it can happen, unfortunately that's only a comment
with no reference.

> Maybe we want to
> destroy this pad it if isn't exposed after some time?

That seems wrong, except if something in the spec allowed us to make sure there
could be no chunks of a certain type after X chunks, but I don't think it is
the case.

> Another solution would be to check if we have emitted no-more-pads after a push
> and ignore NOT_LINKED returns until we get no-more-pads but I'm ok with your
> proposal, too.

Yeah, I think I mentioned that to you the other day, can be a bit more involved
but that would definitely improve flow combiner for everyone so there's that.
In the case of flvdemux though, if the "lying header" bug happens, no-more-pads
might never get emitted. That's however a different issue that already exists
and could certainly be addressed in some way.

-- 
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.