bug#52491: 28.0.90; Regression in window deletion with minibuffer

[Juri Linkov]

close 52491 28.0.90
thanks

>> Or maybe simply:
>>
>>       ((and dedicated (not (eq dedicated 'side))
>>             (window--delete window 'dedicated (eq bury-or-kill 'kill)))
>>        ;; If the previously selected window is still alive, select it.
>>        (when (window-live-p (nth 2 quit-restore))
>>          (select-window (nth 2 quit-restore))))
>
> Right.  Please install that.

Thanks, now pushed to emacs-28.

BTW, could you suggest where to look to fix such problem: for example,
create a grep/occur buffer, then start visiting occurrences from it
one by one, e.g. type RET that shows the occurrence in the second window,
then select the first window again, then type RET again to visit the
next file in the second window, etc.  Then invoke 'quit-window' in the
last visited buffer - it selects the first window.  Then select the
previous buffer in the second window, and again invoke 'quit-window'.
Now it removes the buffer, but doesn't select the first window.

Such inconsistency causes a lot of trouble because of this
unpredictable behavior.  It reselects another window only
on the first invocation of 'quit-window' in the same window.

Would it be possible either to always select another window
on every 'quit-window', or to never select another window
on 'quit-window'?