Comment 32 for bug 1272291

I've noticed an additional quirk. If I slowly move the cursor from the right screen to the left (after it has become cut-up as described by others), on the right screen the cursor is as it should be. As it gets within a few pixels of the left screen (but still fully on the right screen), the first slice starts to show up on the left screen. In other words, the cursor is partially rendered on the left screen before it should be and still fully rendered on the right screen. Very curious.

The distance between the first two slices of the cursor on the left screen is apparently precisely the distance of the uncorrupted cursor from the edge of the right screen when it starts to be rendered on the left screen too early. So it seems like there is some mismatch in the rendering locations and the internal rendering of the two monitors. Since the slicing occurs when switching between the monitors (from right to left for me), it seems like this is an artifact arising between the rendering location of the cursor as it is between monitors and the boundary between the screens.

Also note that all the alternate cursors (e.g. window resizing at corners) are similarly sliced up. So perhaps there is some kind of filter applied to the cursor image before rendering to the screen that is getting mangled/corrupted when switching between screens.

I also found a way to uncorrupt the cursor. In the Catalyst Control Center for the Radeon driver, going to "Digital Monitor (2)" (I have 2 monitors, might be different for other users), selecting the tab "properties", and checking "Reduce DVI frequency on high-resolution displays" (default unchecked) restored the cursor. This resets the right screen (with the working cursor!) and somehow restores the proper cursor on the left screen. When the cursor corrupts again, just cycle the toggle.