[xmonad] Walking Fullscreen Windows

[Chris Bell]

Hi all,

I've noticed an interesting issue with some Fullscreen windows when
using SmartBorders. I apologize in advance if this has been addressed
- I couldn't find anything, but I'm not sure I was asking Google
right.

I am using the current release of Xmonad with SmartBorders. When I am
running an application fullscreen, SmartBorders removes the borders,
as it should. However, every time the window gains or loses focus, the
window will grow by a pixel or two. Specifically, the lower right
corner will extend down and to the right. All UI elements in the
window follow accordingly.

Specific example: Using Chrome to watch fullscreen HTML5 video
(Netflix). Every time the window gets/loses focus, the UI elements
will walk a pixel down, and extend a pixel right, like it's scaling to
a resized window. When focus changes again, it again creeps.
Eventually the window is far larger than the actual screen. When I
leave/re-enter fullscreen, the size is initially correct. Then it
starts growing again as I interact.

Chrome has handled this admirably - the video simply scales up a
couple pixels to compensate. Fullscreen (OpenGL) games, on the other
hand, totally freak out. The graphics won't scale, instead a border of
corrupt pixels slowly grows around the image; the new pixel
rows/columns are filled with random data.

I run 3 monitors, and if this happens on the left or center monitors,
the windows start overlapping the screen to the right. It almost seems
as if the window is getting hints from Xmonad that don't quite
correspond to the actual dims, and it keeps compensating. The other
idea is that SmartBorders is trying to cover up a border that it
thinks exists, but really doesn't. Either way, I'm not quite sure
where to start debugging.

There's one glaring exception: I have defined a window hook to force
VLC to go FullScreenFloat when FullScreen is detected (since VLC
fullscreen didn't work out of the box). VLC's window *doesn't* creep
like Chrome or any of the others do.

Has anyone else seen this? Any fixes/debug suggestions? I'm really
hoping I just screwed up something simple. Thanks!

Regards,

Chris Bell

Ph.D. Student
University of South Florida
College of Engineering
Department of Computer Science and Engineering
NarMOS Research Team