This made no sense. If the thread is detached, the handle is released
asynchronously when the thread exits. So it cannot be used in any way
(except from the thread itself).

Note that Carbon code isn't supported in 64bit applications...

lua_variables: luaL_optstring returned pointer is valid as long as the value stay on the lua stack (also avoid an strdup).

Theora support is enabled. An option is added to manually select the
payload format of unknown dynamic payload types, in pre-arranged
streaming setups, without needing SDP. The only choice for now is
Theora.

lua_variables: the pointers returned by luaL_checkstring might be invalid when the value is poped from the lua stack.

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

This is an alternative to SDP to convey the codec configuration
necessary to decode the stream. The configuration is sent only once at
the beginning of the stream; at least you won't be tempted to waste
bandwidth with retransmissions. This is a ad-hoc hack, but it's not too
bad or pervasive.

This reverts commit 193aeeff.

Conflicts:

	include/vlc/libvlc_media.h

Signed-off-by: Rafaël Carré <rafael.carre@gmail.com>

This reverts commit 2ca4cf27.

Not enabled yet

If the the link is relative, psz_uri is dynamically allocated a string of the
absolute URI (which is also returned by the function).   However, if it isn't,
NULL is returned and nothing is allocated.   segment_New performs a strdup, so
the pointer is not saved.   Thus, check if the allocation occurred and clean up
if needed.
Signed-off-by: Jean-Paul Saman <jean-paul.saman@m2x.nl>

The defined HAVE_AVCODEC_MT is used by modules/codec/avcodec/avcodec.c
and modules/codec/avcodec/video.c. It was only defined in modules/codec/avcodec/video.c.

This patch moves the define to modules/codec/avcodec/avcodec.h.

Take into account that HLS streams with bandwidth specified are not in ascending
(or descending) order. Choose really the best wrt to measurements.

Download the first couple of segments to determine the best
bandwidth match and to ensure smooth playback.

src/stream_output/stream_output.c: input_UpdateStatistics() expects an INPUT_STATISTIC_* not an SOUT_STATISTIC_*

input_UpdateStatistics() expects an INPUT_STATISTIC_* not an SOUT_STATISTIC_*

This doesn't quite follow the outdated draft-barbato-avt-rtp-theora-01,
but keeps consistant with RFC 5215 (vorbis) and with the ffmpeg
implementation