Unlike _endthread and ExitThread, _endthreadex doesn't close the handle.
It is therefore up to the caller to close it. Hence no need for an extra
handle duplication (that mostly amounted to a memory leak when the normal
 path was used, i.e normal return from thread with implicit _endthreadex)

According to msdn doc, _endthreadex(0) is to be used when thread is
 launched with _beginthreadex().

Note that when returning from the normal path, an implicit _endthreadex is
issued anyway.  It is important to be consistent whatever the path, since
those final functions (_endthread, ExitThread, _endthreadex) don't have
the same deallocation policy.

As we use AC_PATH_PROG, if you want to force LUAC by specifying it on your command line you need to give a full path+exec_name. Which is a bit un friendly. So work-around this behaviour.

For lack of a faster detection (based on win32 api messages),
resize/move of the parent window are detected in the Manage function
on a regular basis and acted upon if need be.

Move is an issue only for directx (surface doesn't automatically follow
the associated window). Though coordinates doesn't change, an UpdateRects
is still required to update the various rectangles used by directx.

As it was coded, everything was already taken care of.
The artificial "shaking" of the window was quite useless.

drop toNativeSeparators-usage (shoudln't break win32, but then again this could
be my application for the 'stupid win32 breakage of the year' ;). This allows
for example drag youtube links to vlc playlist directly.

Parse also dropper text if it has valid url if we don't have urls on drop ( from browser addressbar for example you get text ).

This is faster as we can now alloc the right size for the shortcuts array.

There seems to be no more reason to force clipping to the display size
when zooming.  (tested successfully on WinXP with directx, gdi, opengl)

Dividing an unsigned int by 2 is different from dividing an int by 2

This division was the cause for vlc(Win32) displaying a black screen
when zooming exceeded the display size (alt-'o')

Weirdly, there was no problem for Linux !!?? and also no regression :)

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

 It removes the need of the reparentable widget (it is not needed
to reparent at all) and so simplify the code.
 It fixes win32 fullscreen support and improves win7 transitions.
 It avoids the need of an extra fullscreen widget, which is consistant
with the other media players I have tested.

LibVLC needs it.

This fixes an assertion failure whenever --programs is used.
This should also fix a memory leak in the ES output.

This needs testing and backport to 1.1-bugfix.

And patch it again and again...

This hopefully should fix multiple issues with live webm files

Also sort the scripts by file name...

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

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

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

Because now the license is ok. And update the patches on ffmpeg

This fixes #3681