Allows using tls transport and does
non persistent connection fallback.

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

Libass does the same, and I think it's a good idea. But their method is IMO overly complicated.
Or else I'm over-simplifying things. But this seems to work very well.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It should be LayoutLine(). To be consistent with LayoutParagraph() and LayoutText().
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>

FixGlyph() expects to be called immediately after FT_Load_Glyph(), which is no longer the case.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

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

Those patches prevent an infinite loop when specifying an interface name, and
avoid using a down interface as the defautl interface.
fix #14526

With waitall set to false, net_Read() is identical to recv()
outside the input thread. This avoids allocating a bogus waitpipe.

net_Read() is identical to recv() outside the input thread and if
waitall is false.

This partially reverts commit e9ffcf98.

vlc_strcasestr() is slower, locale-dependant and wrong
(e.g. httpś should not match).

This is a bad idea.

400 was clearly not enough for HD streams
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It makes more sense post move to the IEC scale
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

should appear smoother

Fixes playback with single file, simple profile

There is no actual usage of the waitpipe left in the HTTP server.

All existing call sites are in muxers and they all check that the FIFO
is not empty, and assume that it cannot be emptied asynchronously.

So simplify accordingly.

A lot of playlist probe only check the file extension that returns true for a
directory too.

We need to unselect missing PID on PMT update as some
providers add overlapping delay to the last packet of
prev pid.

reverts commit 50accb82