Signed-off-by: Rémi Denis-Courmont <rdenis@simphalempin.com>

Backported from 0.9-bugfix

Patch provided by Marian Durkovic.

By the way, the exception handler claims the exception is not handled!

We don't want a different interface than the one requested (if any)

Thanks for not waiting for this announced, easy-to-backport commit.

 I have modified spu_RenderSubpictures to take a b_paused argument.
 The calls to pf_pre_render/pf_update_regions could have been moved to
spu_SortSubpicture and might have been cleaner BUT the way used keeps
all time consumming tasks inside spu_RenderSubpictures.

Buggy CacheMerge assumes this (among other wrong things)

It is possible to read from a file which is in the midst of being written.

Respect jpeg format setting as scene snapshot and respect size setting when both width and height are set.

This is a suggestion by Forum user boing (http://forum.videolan.org/viewtopic.php?f=18&t=47295&p=162690#p162690)

Freebox implements a HTTP/1.0 client, we a HTTP/1.1 server. As such we should base our determination of the keepalive vs. close behaviour on the query (client request) instead of the answer (server response).

See also: http://www.ietf.org/rfc/rfc2145.txt

Deactivate ActiveX by default and
reordering of File Asssociation to be more visible.

This fixes the matroska crash in at least our Windows and Linux version,
until a new libebml is released. Laurent Aimar had mostly worked around these in
[ce718de2] but FindNextElement was still vulnerable in

Tested on a matroska download with gaps. Many thanks to Laurent