Avoids using movi position as offset when the broken index refers
to beginning of file and the first sample size is exactly movi's
offset.

Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

NV12 is not supported in Windows 7 [1] so we need to fallback to RGBA

1: https://msdn.microsoft.com/en-us/library/windows/desktop/bb173059%28v=vs.85%29.aspx#DXGI_FORMAT_NV12Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It happens when opening the D3D11 device fails
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Because if the PCR exists and has an offset,
the first PCR update will be interpolated
and no following pcr/dts offset fixup will
work without delaying playback by the
real - interpolated pcr.

It's a great idea, but not the best implementation and broken since a
long time.

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

"end" label is only reached if psz_eltname is NULL.

If subsequent malloc or realloc failed, free would be called on a dangling
pointer.

close #14006

This is better than leaking them to the DNS resolver. This also avoids
confusing error messages if live555 is missing or failing.

refs #14193

refs #14193

refs #14193

p_owner->b_paused can only be written by the owner, so the owner thread
does not need to lock to read it. (However, locking is necessary to
write it from the owner thread, or to read it from the decoder thread.)

This removes small hacks surrounding this corner case.

Do not rely on allocating a block for draining the decoder and output.