macosx/framework: Fix the configure ARCHS autodetection script so that we are shell independant (setenv vs export), and force to build only the selected ARCHS.

We still need the configure checks.
Only the HAL device probe plugin is gone.

It was ever dead code within VLC. On top of that, HAL is now officially
deprecated in favor of libudev (and higher level D-Bus middlewares).

This should fix a rare crash:
LibVLC fatal error destroying mutex (16) in thread 17390 at ../../src/misc/pthread.c:188 in vlc_mutex_destroy
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7fff933]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7fff9c8]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4(vlc_mutex_destroy+0x3f)[0xb7fffa39]
/home/remi/videolan/vlc/build/modules/access/rtp/.libs/librtp_plugin.so[0xb5563a41]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4(__module_unneed+0x29)[0xb7fe414f]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7f9984e]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7fa69fb]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7fa6c0c]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7faeef2]
/home/remi/videolan/vlc/build/src/.libs/libvlccore.so.4[0xb7feb16e]
/lib/i686/cmov/libpthread.so.0[0xb7f324b5]
/lib/i686/cmov/libc.so.6(clone+0x5e)[0xb7eaea5e]

Signed-off-by: Laurent Aimar <fenrir@videolan.org>

Also change comparing stream size from (<=0 && < 16384) to (<=0 && >16384)

This and earlier change to add access to prefix should fix usage of eg
http://media.meteo.it/vsat.asx playlist-file.

(cherry picked from commit b3d9ea87aa40d43f081b8fcecd7d0126e6865196)

LibVLC is not calling gpg_error functions anywhere. Linking gpg-error
causes problems with *correctly* built libgcrypt (e.g. incorrect
package dependencies).

This reverts commit f865a816.

They are used nowhere in the code, and just clutter the build system

This allow to have channelnames on saved xspf-playlist from dvb-scanning.

It is not described in the msdn documentation, but planar 420
seems to be supported by a lot of video cards.

Signed-off-by: Rémi Duraffort <ivoire@videolan.org>