It is described on http://shinnai.altervista.org/exploits/SH-008-20101026.html
Thanks to geal for helping finding the problem
(cherry picked from commit b531955c)
Signed-off-by: Rafaël Carré <rafael.carre@gmail.com>

IV attribute of tag #EXT-X-KEY was parsed for any version of the protocol.
However chapter "7. Protocol version compatibility" lists the IV tag being
protocol version 2 or higher. This patch restricts the IV attribute to version
2 or higher of the protocol.
(cherry picked from commit a07189f8)

Properly pass duration to the correct hls_stream_t instance currently in use.

Take relative URI into account. This patch relies on protocol (s->psz_access)  information being
included in the stream_t * in addition to psz_path.

In addition to p_access->psz_location also add p_access->psz_access to the contructed
stream_t *. It will allow a stream_filter to determine the protocol used in case it was
constructed with an access input.

Cleanup

Remove debug code no longer needed.

Remove unused include files.

See frame_mbs_only_flag chapter 7.4.2.1

using psz_string to initialize will leave the top 32bits uninitialized

A too short delay causes issues with late packets being dropped by the
stream output. This reverts 32e94ee0

OSX automake 1.10 can build the contribs but not VLC

Add grain in first page, move away blur
Renaming gradfun
Tab reorder

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

aout/vout were sometimes used. It was inconsistent and in some cases
those objects could be undefined (NULL).

A scaler/chroma converter is inserted as the last filter if needed.

You need a very recent libbluray for this support

As pointed out by Laurent

 It is not needed anymore as changing deinterlace settings does not
restart the vout anymore. It removes the last vlc_object_find() in
the vout code.

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

Some important things:
 * the test is deactivated by default as it needed netcat. To activate it you
just need to add check_SCRIPTS to the TESTS variable (I won't do it unless we
think that netcat is present on every hosts that run the make check?)
 * the test is really simple for the moment (that's more a POC)
 * we can write a small netcat-like tool to activate this test by default...