Fix a crash when parsing subtitles. From the man page of sscanf:
"the next pointer must be a pointer to character array that is long
enough to hold the input sequence and the terminating null byte"
Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>

This is based on similar initialisation as is already done in the
playlist_Create() function.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Also set size only if it's > 0 (assume 0 is non-usable font-size).

Missing 0 from commit 18305c12 and change to
use CLOCK_FREQ so it's easier to understand.

SPU track works even if it doesn't have something to render at given time.
Fixes soverlay and subtitles in sout.

(Coverity report 1187490)

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

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

This is an inclusion of a patch that has been living out-of-tree for some time,
but now as Cubemap 1.0.0 has been released and the protocol has been frozen,
it would sound reasonable to get it into mainline.

Essentially it makes the VLC HTTP server support the Metacube protocol, which
is a framing format to let an external reflector know where the start and end
of the blocks are without having to understand the format itself (similar to
how the VLC HTTP server itself gets to know this information). This lets it
serve the correct header block to new clients joining in the middle of the
stream, and makes sure new clients don't start in the middle of a block,
which is a problem for non-self-synchronizing formats (such as FLV and WebM).
It includes self-synchronization and header checksumming to ensure smooth
recovery from overruns in the server-to-reflector TCP stream.

The only consumer of this protocol currently, and for the foreseeable future,
is Cubemap (http://git.sesse.net/?p=cubemap), a scalable video reflector
designed to be used with VLC.
Signed-off-by: Denis Charmet <typx@dinauz.org>

Getting the current format is unreliable, for instance if the audio
cable is plugged in while playback is running. Thus only the
first original format is trusted.

close #9957

This prevents waiting for the timeout in some situations, especially
if the format was already set to the new value before.

Indeed this can happen rarely.

close #9469

This would result in an unpleasant black or grey area for
a short time, for instance when selecting the media library.

Some OS/2 APIs are not high-memory safe. So there are needs to use wrapper
functions to make them high-memory safe.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

--stats was added at commit c24ea5fb.
So free arguments after this.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

fixes #9919