The implication here is that sna_pixmap_move_to_gpu() reported success but left priv->gpu_bo invalid, which in turn means that we must have marked the pixmap as being all damaged on the GPU but left the bo unset. It has been a worry in the past, such that there are lots of asserts to enforce the relationship between GPU damage and the bo, if you can reproduce this recompiling the ddx with --enable-debug will be very useful.
The implication here is that sna_pixmap_ move_to_ gpu() reported success but left priv->gpu_bo invalid, which in turn means that we must have marked the pixmap as being all damaged on the GPU but left the bo unset. It has been a worry in the past, such that there are lots of asserts to enforce the relationship between GPU damage and the bo, if you can reproduce this recompiling the ddx with --enable-debug will be very useful.