We do maintain a lock around vlc_object_alive() because we want to make sure we won't miss the point when a concurrent thread will call vlc_object_kill(), that is set b_die and vlc_object_signal(). If we missed b_die and the signal, next vlc_object_wait() will dead lock.

Signed-off-by: Pierre d'Herbemont <pdherbemont@videolan.org>

We can't do what is advised in the documentation of vlc_object_alive(), but we try to stay close enough of it.

Remove useless variable

Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>

The whole point of a global lock is that it is... global. Destroying it
every time it's released is severe brain damage, as we end up taking a
different lock all the time (or worse destroying the lock while another
threads is competing for it).

We should really keep the global "libvlc" lock while initializing the
modules bank. Otherwise, if a second instance is created while the
first one is initializing, it may get an incomplete bank. This commit
addresses the problem w.r.t. the "main" module - only.

libvlc: rename libvlc_media_descriptor to libvlc_media and libvlc_media_instance to libvlc_media_player.

Wrong dir

This reverts commit f11a132c.

Signed-off-by: Pierre d'Herbemont <pdherbemont@free.fr>

This reverts commit 1dfd50c2.
This is no needed of such a hack if there is no circular dependency between input and vout and if the input are properly released.

Else there is a circular dependency between the vout that yield() the input, where the vout is also an input child. This should fix properly the media_instance() tests.

stream: No use to vlc_object_yield() and object that create ourslef, and that a race could possibly corrupt in case of bad usage.

stats: Move the stat to libvlc instead of the playlist. As stated in the code it is not playlist related.