On Linux, this simplifies the code and improves relocability.
On Windows, OS/2 and MacOS, this should fix a small race condition.

Signed-off-by: Rafaël Carré <funman@videolan.org>

This reverts commit 3bbc2f54.
Now it exists again

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

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

Using hls_Get and segment_GetSegment instead. Those check for out of
bound access.
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>

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

parse_M3U8 uses hls_GetLast to update the HLS current hls stream, so we
have to add it to the streams array before calling parse_M3U8
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

This prevents a "bug" when we seek backward and then play again that
segment.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Change the string's text field from static to editable, so options not in the
wizard can be added. Uses the string from this field directly, rather than the
stored string.
Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

it respects the 'live-caching' setting now, is vastly more responsive and even got correct timing.

Thanks to emi and Luca for the ideas and help

this were caused by an incomplete implementation of a 2.0-next feature

This disables V4L2 completely.

This reverts commit 5740099f.
This caused autoconf errors and besides I still do not agree.

(I do not know why it was removed in the first place though - not me!
 As a consequence, we cannot scan TV channels in VLC >= 2.0...)

we could end up with 'ggcc'
forgotten in a6e2d066

Was broken since the pure virtual method getId() was renamed in
the PLModel cleanup 4a61e6a8

Those objects and shared library handle are not guaranteed to exist
if the Clean() function is called in Open() error path.

Fix the second part of #5896
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fix the first part of #5896
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

A 0 bandwidth forbid us to seek, as the position would always be 0.
This fixes one of the problems described in #5370
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

Don't stop when playing a vod file without bandwidth information

When a playlist doesn't contains any bandwidth stream or information,
initialize stream->bandwidth to 0 (it's a unsigned number). So test will
not
fail when calculating if we have the time to download it.
 >> uint64_t size = (segment->duration * hls->bandwidth);
 >> int estimated = (int)(size / s->p_sys->bandwidth);
 >> if (estimated > segment->duration)
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>

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

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

(text from the patch author)
On many systems taking couple of kilo-bytes for full sinc table is not
a problem. With a straight filter table the number of MAC operations
needed drops to one quarter. Also the inner loop of the filter becomes
more simple and easier to optimize. So the full sinc table version
runs much faster in most cases. The trade off is some kilo-bytes
bigger memory usage and more CPU needed when initializing the
resampler.