On Tue, 6 Dec 2016, martin rudalics wrote:

>> Your patch has perfect sense to me.  Indeed i wrote an almost identical
>> fix but you were faster than me.
>
> Damn.  Next time I'll wait a day or two :-(
>
> But we still don't know what this ‘window-list’ stuff is all about.  I'd
> rather have asked ‘walk-window-tree-1’ on the main window to return the
> last window it sees, but consider the following emacs -Q scenario:
>
> C-x 2
> C-x o
> M-x ibuffer
> m
> m
> D
>
> The confirmation buffer appears above the ibuffer buffer.  I doubt that
> such behavior is intentional but since I'm no ibuffer user ...

I think it shows the author preference for, if possible,
display *Ibuffer confirmation* in other window that
the one showing Ibuffer.

If the current frame just has a window showing
Ibuffer, then *Ibuffer confirmation* is shown in that
same window:

emacs -Q
M-x ibuffer RET
d d x

emacs -Q
C-x b test RET
M-x ibuffer RET
H
M-x other-frame RET
d d x

If we cannot show *Ibuffer confirmation* in the side window,
because it's not allow to split it, it's more useful to chose
another window, i.e., to follow your patch, instead of signal
an error.

Tino