Focus issue with monitor at unusual position about herbstluftwm

[t-wissmann]

The problem here is that the directional focus doesn't work well when the rectangles aren't perfectly aligned. If I remember correctly, my algorithm considers angles between the centres of the rectangles. And since HDMI-A-0 is quite wide, the angle to DVI-D-0 is too flat (flatter than 45°) to be recognized as below. The same algorithm is used for directional focus between floating windows and it seems I can reproduce the issue with a similar window layout (and the top right window being wide).

Maybe one can be a bit smarter in rectangle_is_right_of in floating.cpp. My main incentive was that every window should be reached from every other window and that it should work if the rectangles aren't perfectly aligned.

[xrun1]

I haven't deeply inspected the code yet, but how about widening the allowed range if the search fails? e.g. If nothing in a -45 to 45° range is found, then search with -89 to 89° limits, allowing any R2 that's at least slightly further than RC's center to match.

[t-wissmann]

Yes, I suggest that this would work for monitors (this fallback should only be activated for monitors, not windows, because if there is no window, then the directional focus 'escalates' to the monitor level, cf focus_crosses_monitor_boundaries :-) )