This MR improves window management related behavior and adds support for more scenarios.
Some highlights (please see the full commit list and messages for more info):
1. Better support for tiled states. 2. Detect and handle fullscreen windows. 3. Increased robustness against state disagreements between Wine and the Wayland compositor. 4. Move top-level windows to (0,0) (using that fixed point for now, we will later generalize this mechanism for multiple monitors), in order to: a. maximize the accessible mouse input region (which may be clipped in Windows virtual screen space, but still accessible in Wayland space). b. introduce the basic mechanism to allow windows applications to (very roughly!) track which output(s) they are really on (to be continued in the full multi-monitor form).
Although (4) is not a perfect solution for the lack of absolute positioning, it has worked well in practice in its full, multi-monitor implementation (i.e., in the experimental branch). I wonder if for (4a) in particular we can do better, e.g., by being able to emit input that circumvents the normal virtual screen clipping. Is there a way to do this already? If not, do you think that such an approach would be a feasible and acceptable way forward, or is virtual screen input clipping fundamentally baked in the current core design and assumptions?
Thanks!