In case there was no pcr but a splice point must be generated, then the
adaptation field generation code looked at the stuffing byte size. However
just before deciding an adaptation field was needed the adaptation field
length was properly adjusted to include a splice_countdown field.

Only generate splice points for VIDEO_ES and AUDIO_ES.

The adaptation field size was not correct in the case that there was no splicing point. This
resulted in corrupt TS packets.

Some distributions do not use x264 >= 86 yet.

get the SD video (in flv), the HD and HQ videos (in mp4) causes seeking
back and forth and the video is unplayable

It is not yet perfect as the input is not warned of the latency of the vout.

Becareful that it adds a delay of one frame, so it needs to be
flushed on discontinuity and it will return NULL one time.

 It is not yet used, but it will be needed for filter that need to delay
frames.

We now test that the selected fourcc actually works by calling vaGetImage.
Initial patch from Gwenole Beauchesne.

No functional changes.

No functional changes.

Two new functions added:

 * val = libvlc_audio_get_delay( p_mi )
 * libvlc_audio_set_delay( p_mi, i_delay )

The delay is measured in microseconds.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

As suggested by Jan Janak <jan@ryngle.com>

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Entity encoding was missing (for options) in generation of an xspf playlist.
On the reading side the entity decoding is all fine.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

This reverts commit f5b8b43a.