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

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

Removes a lot of hacks, especially around manipulating size (no more sizeHint(), very limited use of doComponentsUpdate and updateGeometry() ),
Simplify the number of starting options of the interface
Fixes crashes due to QStackWidget limitations
Still a lot of debug around, but mostly commented out, that will go in future commits
Close #3401
Close #3210
Close #3332
Close #1717
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

For skins without video elements, fullscreen mode could be launched with the inner video kept hidden. Initializing the video control to NULL fixes it.

Signed-off-by: Geoffroy Couprie <geal@videolan.org>

- make it case insensitive and sorted alphabetically
- fix a few regressions about config_Get vs. var_Inherit

In skins, slash is the preferred path separator for both Linux and Win32
yet, rather fix those separators and issue a warning than fail to open skins.

This makes drag-and-drop easier, because drag-and-drop pixmap does not
cover the area under mouse cursor.

- Make PlMimeData simply a subclass of QMimeData: destructor of indirect
subclasses of QObject is called even if their direct base class' destructor
is not virtual.
- Protection against using input_item_t* from another VLC instance:
qobject_cast<PlMimeData*>() will fail on QMimeData dropped from another process.
- Some refactoring of PLModel::dropMove() for robustness.

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

A helper object keeps reference count until drag-and-drop is finished.

This enforces the use of qtu() or whatever appropriate.