* patch by Pierre d'Herbemont which cleans up our locale detection code heavily and converts it to be CoreFoundation-based only. Thanks!

 I think we have a potential bug with the way subpicture_region_t->p_style
is used.
 Decoders allocate p_style memory in their context, so we don't free it in
spu_DestroyRegion. As vout object can live longer than decoder, I think
p_style might be used after the decoder has destroyed it.

 Can someone, who has worked on osd/spu internal, comment ?
 

(Use vlc_find_object(PARENT)

(Use vlc_find_object(PARENT)

(Use vlc_find_object(PARENT)

Build live555 libraries with -DUSE_SYSTEM_RANDOM=1 by default. This fixes a bug reported by Jason Gerecke (jgerecke at uoregon dot edu). The rationale from Jason:

In groupsock/inet.c: http://www.live555.com/liveMedia/doxygen/html/inet_8c-source.html a custom random generator is provided for systems which do not provide their own rand() or srand() functions. Other people have reported issues with this function segfaulting (never in conjuction with VLC however), possibly because of multithreading issues, and the fix issued by the Live555 developers was to provide an "#ifdef USE_SYSTEM_RANDOM" for builds on supporing systems (http://lists.live555.com/pipermail/live-devel/2003-November/000013.html).

the drivers get fixed

(This way it matches input_CreateThread and fixes a memleak)
Control variables are now created even for preparse, only callback are not added.
(This fix use of uninitialized variables)

Add CODEC_ID_GSM and CODEC_ID_GSM_MS to ffmpeg audio codec list. (mostly useless but it doesn't hurt. needs ffmpeg to be linked with an specific lib)

 We use vlc_UrlParse for "URL" without protocol... so it was using a part of
the URL as a protocol if it found ":/". Ensure to extract only valid protocol
at least.

- Use net_Accept()

- Fix som bugs

- Poll the system clock only once per loop (yes, it DOES matter),
- Fix rare bug with TLS and multiple listening sockets,
- Do not try to read/write to client socket with no pending events.
...and of course introduce dozens of new bugs.

This SHOULD speed httpd up a bit.