The borderless window is nicer IMO. Feel free to object, but this also saves some overhead with fullscreen changing.

qtcapture: Don't release objects when libvlc quit, else we may loose our main thread, and dead lock.

Don't send blocks from qtcapture Demux that we have already sent before. Avoid excessive CPU usage while there are no new blocks.

This works, but perhaps i'm loosing too much precission here. Are there better ways to change timevalue + scale into walltime ?

naturally choose.  With only an effective -O0 gcc 4.0.1 doesn't produce
code as intented for copy_u8_altivec, causing liboil to fail at runtime.

Summary, liboil really needs -O2 (always).
Signed-off-by: David Flynn <davidf@rd.bbc.co.uk>
Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

It is disabled by default, look for the two lines with
	//XXX uncomment for fullscreen controler
to enable it. Please report success/failure in order to enable it by
default (linux/win32).

Spotted by 3af47f42 and ivoire.

macosx: Fix a deadlock in the broken -[VLCPlaylist isItem: inNode: checkItemExistence:]. Fixes #1711 (Deleting a playing item's parent item result in a trivial deadlock on Mac OS X.)

Kudoz to sam and Meuuh

Note, this commit may create some assert in previously working code. That does mean that this code wasn't properly working, and that it lacks a PL_LOCK. I prefer nice assert() over races.

Should fix #1708 (Adding a folder in the Mac OS X VLC *media library* result in an assertion failure). Thanks to upcoming commit "playlist: Use PL_ASSERT_LOCKED where the playlist lock should be held.".

playlist: Hold the lock at creation time to conform to playlist_NodesPairCreate requirement to have the pl lock held.

playlist: Use PL_LOCK_IF and PL_UNLOCK_IF to perform some more checks on lock state. And fix a unlocked usage of get_current_status_item().

That may fix or help to fix some input item leaks.