*Completions* in minibuffer-only frames

*Completions* in minibuffer-only frames

[martin rudalics]

[-- Attachment #1: Type: text/plain, Size: 459 bytes --]

With Emacs 26 and a minibuffer-only frame, *Completions* may be
displayed in the minibuffer window.  To reproduce with emacs -Q do
(make-frame '((minibuffer . only))) and in that minibuffer-only frame
type C-h f followed by TAB.  This will display *Completions* in the
minibuffer window and it can become quite complicated to get rid of it
again.  Hence I would like to apply the attached trivial patch to the
release branch.  Any objections?

Thanks, martin

[-- Attachment #2: display-buffer-avoid-minibuffer-window.diff --]
[-- Type: text/plain, Size: 881 bytes --]

diff --git a/lisp/window.el b/lisp/window.el
index 4b30609..0a36ae7 100644
--- a/lisp/window.el
+++ b/lisp/window.el
@@ -7397,6 +7397,7 @@ display-buffer-below-selected
 	     (window--display-buffer
 	      buffer window 'window alist display-buffer-mark-dedicated))
 	(and (setq window (window-in-direction 'below))
+             (not (window-minibuffer-p window))
 	     (not (window-dedicated-p window))
 	     (window--display-buffer
 	      buffer window 'reuse alist display-buffer-mark-dedicated)))))
@@ -7432,6 +7433,7 @@ display-buffer-at-bottom
 	     (window--display-buffer
 	      buffer window 'window alist display-buffer-mark-dedicated))
 	(and (setq window bottom-window)
+             (not (window-minibuffer-p window))
 	     (not (window-dedicated-p window))
 	     (window--display-buffer
 	      buffer window 'reuse alist display-buffer-mark-dedicated)))))

Re: *Completions* in minibuffer-only frames

[Eli Zaretskii]

> Date: Tue, 19 Feb 2019 10:31:09 +0100
> From: martin rudalics <rudalics@gmx.at>
> 
> With Emacs 26 and a minibuffer-only frame, *Completions* may be
> displayed in the minibuffer window.  To reproduce with emacs -Q do
> (make-frame '((minibuffer . only))) and in that minibuffer-only frame
> type C-h f followed by TAB.  This will display *Completions* in the
> minibuffer window and it can become quite complicated to get rid of it
> again.  Hence I would like to apply the attached trivial patch to the
> release branch.  Any objections?

How old is this problem?  It doesn't sound too serious to me, FWIW
(but then I never use minibuffer-only frames).

I'd really like to release Emacs 26.2 VSN; the bug reports about
recent regressions or serious bugs all but ceased to arrive, so I
think we have a stable branch.  Installing non-trivial (by their
effect) changes is something I'd like to avoid if possible.

Re: *Completions* in minibuffer-only frames

[martin rudalics]

 > How old is this problem?

It's with us at least since Emacs 25.

 > It doesn't sound too serious to me, FWIW
 > (but then I never use minibuffer-only frames).

As a rule I don't use minibuffer-only frames either so I only recently
bumped into a number of problems while looking at them a bit harder.

The behavior I describe is a direct consequence of our idiosyncrasy to
consider the minibuffer window of a minibuffer-only frame as the root
window of its frame's window tree while we do not consider the
minibuffer window of a "normal" frame as part of that frame's window
tree.  This confuses ‘walk-window-tree’ which may not interpret its
MINIBUF argument as intended when "walking" a minibuffer-only frame.

 > I'd really like to release Emacs 26.2 VSN; the bug reports about
 > recent regressions or serious bugs all but ceased to arrive, so I
 > think we have a stable branch.  Installing non-trivial (by their
 > effect) changes is something I'd like to avoid if possible.

Since nobody ever complained I have no problems with postponing the
fix to Emacs 27.

martin

Re: *Completions* in minibuffer-only frames

[Eli Zaretskii]

> Date: Tue, 19 Feb 2019 18:46:17 +0100
> From: martin rudalics <rudalics@gmx.at>
> CC: emacs-devel@gnu.org
> 
>  > I'd really like to release Emacs 26.2 VSN; the bug reports about
>  > recent regressions or serious bugs all but ceased to arrive, so I
>  > think we have a stable branch.  Installing non-trivial (by their
>  > effect) changes is something I'd like to avoid if possible.
> 
> Since nobody ever complained I have no problems with postponing the
> fix to Emacs 27.

OK, so please install the fix on master for now.

Thanks.

Re: *Completions* in minibuffer-only frames

[martin rudalics]

 > OK, so please install the fix on master for now.

I installed a fix for 'walk-window-tree' on master which should handle
the root of the problem and not its symptoms.

martin