Layouts and Controls are interrelated. Whatever the ones first deallocated, it leaves pointers referencing objects already destroyed. and potentially means memory leak.

This patch adds an unsetLayout() function to pair the setLayout() function and aimed at releasing resources.
Policy should now be that things allocated in constructor are released in destructor and things allocated in setLayout are released in unsetLayout.

Remove the old way of tracking change in vout size (no longer used)

In today's implementation, notifyLayout() for a control leads to a whole
 layout rebuilt. Therefore, this function must be called _only_ if needed.

- Never needed in onResize() since we are already in the process of rebuilding
the whole layout. Calling notifyLayout() means the job is done twice.
- Never needed in onPositionChange() for the same reason as onResize().
- Not needed for a checkbox since already called in setImage() two lines above.

intf-change for playlist is never updated

When reparenting, these values must be updated to avoid unnecessary resizing
if the move() method is called.

This skins demux is not of much use, but at least, it is now functioning.
(no more crash)

Fix #3232

Otherwise, the object files will conflict

PLD was only added to I420 -> YUYV this far.

There is still too much stuff in misc/ ;-)

This should fix #3236

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

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

We cannot use SIGILL, so we have to trust CPUID.
Maybe there is a Win32 API to query the kernel support?

Sorry, I obviously wasn't compiling the right module XD

In case the LibVLC application has overriden SIGILL, we really want to
crash rather than run the handler.

Additional changes in configure.ac according to
[71fafba4].
--enable-vcdinfo was changed to --enable-libvcdinfo and some VCDINFO
were renamed to LIBVCDINFO but the checking was incomplete so pkg-config
could not correctly configure libvcdinfo
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

This reverts commit 380fc82a.

This fixes bootstrap with automake 1.9 and 1.10.

This will colorize the `make check` output if your term supports it.

As agreed the work is rather unfinished for 1.1 and Lunettes is rather promising.

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