If byterange request is done and reply does not provide
content-length reply, read will block due to keepalive.

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

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

Regression from 890f8e0f

ReleasePicture is already locked since it's now called from
AndroidOpaquePicture_DetachDecoder.

Since it was not really working for 2.4.
Handle it the same way for Tizen 2.3 and 2.4.

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

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

Fixes chapters in titles that play multiple playlists
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This matches all the other networking access modules and solves performance issues when playing high bitrate content

This avoids spurious errors in the decoders (and logs).

This function will cancel the decoder picture pool.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

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

This reverts commit 09bb0203.

This improves the compatibility with vlc_poll().
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

New profiles that might break packetization.
(ITU-T H264 2014/02)

Old profile, out of spec
ref: https://ffmpeg.org/pipermail/ffmpeg-devel/2012-October/132400.html

new sample:
samples/V-codecs/h264/lossless_timecode_HIGH444_IDC144.h264

chroma/scaling/matrix is on HIGH profile SPS
(as in 14496-10:2012)

Against further dynamic linker namespace pollution.

Also optimize it away where applicable.