(#3309) selecting a correct next window after the focused one is closed or disabled

[mattkae]

fixed #3309

What's new?

• When a focused window closes, we first try to select the next window given the previous scheme. If that doesn't work, we choose the other scheme. If that doesn't work, we select no window.
• Moved the BasicWindowManager check for is_active_window below policy->advise_delete_window. In this way, we only call BasicWindowManager::refocus if the active window remains unchaged after the advise

[RAOF]

If I understand this correctly, what this change will do is:
when the currently focused window is closed, do the equivalent of Alt-Tab or Alt-` depending on which one the user used last?

That doesn't seem like reasonable behaviour to me, so maybe I'm misunderstanding the change?

[mattkae]

If I understand this correctly, what this change will do is: when the currently focused window is closed, do the equivalent of Alt-Tab or Alt-` depending on which one the user used last?

That doesn't seem like reasonable behaviour to me, so maybe I'm misunderstanding the change?

Your understanding is correct. The reason for this is that you want to select the next app/window in the focus order. Hence, the next available one could be found by Alt + Tab/Tick depending on their last action.

[AlanGriffiths]

Your understanding is correct. The reason for this is that you want to select the next app/window in the focus order. Hence, the next available one could be found by Alt + Tab/Tick depending on their last action.

I too don't find this reasonable.

When I close a window I am not thinking about whether I last used Alt-Tab or Alt-Tick and don't expect the result to depend on it. I expect the next window in the focus order, not "the next window of another app" nor "the next window of the same app".

[AlanGriffiths]

I do agree though, that the current behaviour needs fixing

[mattkae]

I think I should revisit this later. I have a plan to do a full refactor of the focus system here, and this bug would be easily fixed by such a refactor