Could go badly wrong on platforms without atomic 64-bits access

In this case, volatile was both incorrect (volatile does nothing about
thread synchronization) and not required (there is an explicit lock).

Fix blinking on redraw in ncurses interface... if anyone has an idea about why that call to clear() was needed ... (And remove a bunch of trailing spaces)

libvlc: when calling libvlc_media_player_stop() check for playing or paused state before to send a stopped event
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

Note: public key v3 signatures are not supported
Fix various bugs and potentials buffer overflows

Added the minimal-OSX class files and a couple of more related files to the project list for easier lookup

If you have a better idea, you are welcome, but I am very doutful.

Previously, we only removed links from children to dead parents.
This patch is semantically the same as Funman's earlier
vlc_object_detach_unlocked() patch. It is merely faster and less clean.

Unfortunately, this means it is impossible to detect leaked children.
That being noted, leaving dead parents "semi-attached" in objects tree
is not thread-safe anyhow. Basically, we either automatically detach
objects (and miss leaks), or require children to be detached/destroyed
before their parent is destroyed (and get leaks detection, but more
coding pain).

Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

They can only be read safely under the Big Structure Lock in
src/misc/objects.c, so it makes no sense for them to be public.

By the way, making i_children volatile wouldn't magically solve
thread-safety issues. The only correct use of volatile is in dealing
with asynchronous changes _within_ the same thread, such as signal
handling.

P.S.: I wish modules were not objects, and I wonder why they are
(they don't don't use threads, nor plugins, nor variables)

This reverts commit c85f8681.