For anyone coming at this from a driver debugging point of view.. from what I could tell of it (whilst looking at the 855 bug), the xrandr rotate code works via the rendering system to put a picture on screen from a "shadow" buffer, transformed by a rotation matrix.
The symptom here might be due to the damage not being updated correctly, or possibly a bug in the render acceleration of the graphics driver.
For anyone coming at this from a driver debugging point of view.. from what I could tell of it (whilst looking at the 855 bug), the xrandr rotate code works via the rendering system to put a picture on screen from a "shadow" buffer, transformed by a rotation matrix.
The symptom here might be due to the damage not being updated correctly, or possibly a bug in the render acceleration of the graphics driver.