The macosx drawable variable that used to hold the drawable window reference is splitted up in two variables "drawable-agl" (used by macosx safari/mozilla plugin) and "drawable-gl" (used by macosx interface).

Patch by Erwan10, <brezhoneg1@yahoo.fr>, simplified by /me

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

In one case (noted by Erwan Tulou), this also fixes a large leak.

There is _absolutely_ no point in holding an object and releasing it
immediately. Holding an object only makes sense if:
 - the object cannot vanish while calling vlc_object_hold(), and
 - either:
   - the object could have otherwise vanished before
     vlc_object_release(), or
   - there is no way to avoid the (otherwise not needed)
     vlc_object_hold() because it comes from another component.

Conversely, it makes no sense to hold an object, if:
 - it could vanish already while calling hold (race condition)
   -> this is *invalid* and might crash,
 - there is already a valid reference to the object throughout.

I don't know in which case that code was, but the popup menu definitely
does invalid object access in some cases. For instance:
 - start playing a single video,
 - seek to a few seconds before the end,
 - open the popup menu, go to video / deinterlace,
 - wait for the video to finish and the playlist to stop,
 - click on "X" from the deinterlace submenu.
Oops.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Renamed:
 npunix.c to npunix_old.c
 npwin.cpp to npwin_old.cpp

Added new files from mozilla source code
 npunix.c
 npwin.cpp

macosx: playlistlock is held when playlistinfo panel is updated, so call PreparseEnqueue with pl_Locked.

updatePanelWithItem is called from modules/gui/macosx/intf.m line 1565 with playlist lock held.
If this solution is wrong then revert and fix properly.

dialog provider for skins and 1 as the regular qt4). The problems for
skins are real (lead to crashes when using menus). The one for regular
qt4 was obviously a typo, but sometimes even removing a typo may have
side effects. If so, please find attached a patch that reverts this part
of the previous patch.

Erwan10
Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>

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

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

...before calling release function of the window provider
(only way to prevent possible BadWindow xlib error)
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>