Solve 'lost synchro' warnings when using a file input. The warning was
triggered by the functions GetFirstPCR(), CheckPCR() and GetLastPCR().
They searched for a TS packet which contains a PCR but did not take
the TS packet size boundarys into account. By rounding to the nearest
p_sys->i_packet_size this warning was resolved.

And when it was successfully added

Close #8395

Delegates can't request updates themselves.
Repaint needs to be triggered at model or view level.

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

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

Solve a bug where the display stays black when the dimension of the
video is not supported by MediaCodec (either too small or too large).
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

Also don't care input fps and only calculate what should be the
next output pts. We anyway define output fps to be either 25fps
or input fps if it's available.

This should help to eliminate skips/visual artefacts on video if
there is small skips in input. And output constans fps from transcode.

Fixes #10453 for me

Prints dts, segment number, length, difference to previous dts and md5
of block.

with prefix-option you can give prefix to print in stats-line, so you can
add stats module multiple times with different prefixes. Output to file
is given with output parameter.

Skeleton handling does early write of some headers,
and then skipped flagging them.

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

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

See CVE-2013-6954

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

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

This patch introduces FFT windowing routines to reduce spectral leakage in
spectrum-based visualizations.  It supports four types of windows that are
commonly used in spectrum analyzers (besides the pre-existing rectangular
window):  Hann, flat top, Blackman-Harris, and Kaiser.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

For instance changing from 10:10 to 1:1 should not generate an event
if we are not in display fill mode.

Avoid signalling a change of aspect-ratio when it only changed from
1:1 to 0:0.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

BluRay streams have 4-byte header before TS sync byte. Current code handles 192-byte ts packets,
but it assumes extra bytes are at the end of packet. This results in re-sync at stream start:
  1) first packet is dropped
     First packet is usually PAT, and losing it means losing the first GOP.
     This is fatal with still-image based menus.
  2) 4-byte header of the second packet is skipped
     This results in read of the last packet returning only 188 bytes,
     and it is also skipped.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

A maximum display time does not make sense for real teletext pages. Instead
they should be displayed until the user disables teletext or the page is
updated. Therefore do not set a stop timestamp on normal pages, but only when
rendering in text mode, which is meant for subtitles, where the timeout makes
sense.
Signed-off-by: Dennis Hamester <dhamester@jusst.de>
Signed-off-by: Julian Scheel <julian@jusst.de>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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