bug#59862: quit-restore per window buffer

[martin rudalics via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

 > I suppose it will affect only interactive uses of 'kill-buffer'.

Why?  It affects all uses of 'kill-buffer'.

 > PS: I noticed one difference between handling of frames and tabs:
 >
 > C-x 5 5   ;; other-frame-prefix
 > C-h i     ;; info
 > C-h e     ;; view-echo-area-messages
 > q         ;; quit-window
 > This deletes the Info window, and doesn't delete the frame.
 > Everything is correct.
 >
 > C-x t t   ;; other-tab-prefix
 > C-h i     ;; info
 > C-h e     ;; view-echo-area-messages
 > q         ;; quit-window
 > This doesn't close the tab (correct),
 > but doesn't delete the Info window.
 > This differs from the frame case that keeps
 > only the Messages window on the frame.

Suppose you do

C-h i     ;; info
C-h e     ;; view-echo-area-messages
q         ;; quit-window

Then the difference is obvious.  The C-h i reuses the one existing
window showing *scratch*.  'display-buffer-record-window' cannot tell
this case apart from the one where a user does want to show *scratch*
again in that window after being done with *info*.  If we wanted a
behavior that deleted the *info* window in this case (it could do that
only if there is another window on that frame), the caller could add an
alist entry, say 'delete-reused-window-if-possible', and add something
like

     (when (and (eq type 'reuse)
	       (assq 'delete-reused-window-if-possible alist))
       (setq type 'window))

immediately before the

     (display-buffer-record-window type window buffer)

call in 'window--display-buffer'.

But you pass 'tab' as TYPE argument to 'window--display-buffer' and that
would have to be handled in a maybe similar but still different way.

martin