bug#81584: 30.2; zero-width minibuffer

[Please always use Reply to All to reply, to keep everyone CC'ed.]

> Date: Sun, 09 Aug 2026 03:42:06 -0500
> From: Devon Sean McCullough <[email protected]>
> 
> On 2026-08-09 00:30, Eli Zaretskii wrote:
> >> Date: Sat, 08 Aug 2026 21:55:48 -0500
> >> From: Devon Sean McCullough <[email protected]>
> >> I got a broken minibuffer
> >> after deleting all but one frame.
> > How did you get it?  Can you tell how to reproduce the problem?
> Alas, (lossage-size) => 300 was insufficient by the time I noticed.
> I deleted two out of three frames, went shopping,
> when I returned to the console something strange
> happened but I have no idea what.  I doubt it
> will ever happen again.  Perhaps there is a test jig
> to spaz away at Emacs until (not (window-full-width-p 
> (minibuffer-window)))?
> 
> > And what does "zero-width" minibuffer mean?  A minibuffer's width is
> > identical to the width of its frame, so how does one get zero width
> > there? what is the width of the minibuffer's frame?
> The frame is full screen.  In the lower left of the screen shot,
> you can see a scroll bar improperly away from the right screen edge.
> Looks like a minibuffer scroll bar - by the top of that
> anomalous scroll bar, just below the mode line, appears
> a tiny left arrow and a thin green vertical stripe.
> To the right of said scroll bar is a mess of screen turds -
> perhaps redisplay knows the minibuffer is always full width,
> safely ignoring that "impossible" rectangle of screen area?
> I guess the arrow is a fringe truncation indicator.
> 
> 		Peace
> 			--Devon
> 
> P.S.  I collected this hard data before deleting the broken frame:
> (minibuffer-window) => #<window 2577 on  *Minibuf-1*>
> (active-minibuffer-window) => #<window 2577 on  *Minibuf-1*>
> resize-mini-windows => grow-only
> max-mini-window-height => 0.25
> resize-mini-frames => nil
> (minibuffer-prompt) => "Eval: "
> (window-total-width (minibuffer-window)) => 2
> (window-full-width-p (minibuffer-window)) => nil
> (window-max-chars-per-line (minibuffer-window)) => 0
> (window-resize (minibuffer-window) 1 t) ¿? (error "Cannot resize 
> minibuffer window horizontally")
> (adjust-window-trailing-edge (minibuffer-window) 1 t) ¿? (user-error "No 
> window on the right of this one")
> (fit-window-to-buffer (minibuffer-window) 4 4 64 64) => nil ;; no change
> fit-window-to-buffer-horizontally => nil
> (setq fit-window-to-buffer-horizontally t) => t
> (fit-window-to-buffer (minibuffer-window) 4 4 64 64) => nil ;; no change
> (with-current-buffer (window-buffer (minibuffer-window))
>    (erase-buffer)
>    (insert "{text}")) => nil
> (fit-window-to-buffer (minibuffer-window) 4 4 64 64) => nil ;; no change
> (balance-windows (minibuffer-window)) => t	;; no change
> (maximize-window (minibuffer-window)) ¿? (error "Cannot resize 
> minibuffer window horizontally")
> fit-window-to-buffer-horizontally => t
> (minibuffer-window) => #<window 2577 on  *Minibuf-0*>
> (delete-window (minibuffer-window)) ¿? (error "Attempt to delete 
> minibuffer or sole ordinary window")
> (with-current-buffer (window-buffer (minibuffer-window))
>    (buffer-string)) => "{text}"
> (delete-other-windows (minibuffer-window)) ¿? (error "Can’t expand 
> minibuffer to full frame")
> (window-body-width (minibuffer-window)) => 0
> (window-body-width (selected-window)) => 87
> (setf (window-body-width (minibuffer-window)) 87) ¿? (void-function 
> \(setf\ window-body-width\))
> (mapcar (lambda (d) (window-in-direction d (minibuffer-window))) '(above 
> below left right)) => (#<window 10712 on *scratch*> nil nil nil)
> 
> P.P.S.  In case it happens again, I have added
> 	(lossage-size 10000) ; was 300
> to my init file - probably still too small, because
> 99% of view-lossage output is near-endless runs of
>        ⋮
> 	<triple-wheel-down> ;; mwheel-scroll
> 	<triple-wheel-down> ;; mwheel-scroll
> 	<triple-wheel-down> ;; mwheel-scroll
> 	<triple-wheel-down> ;; mwheel-scroll
>        ⋮
> and the like - I feel I should submit a patch
> (as I did for decades after I was no longer
> paid to do so ... until I noticed my work was
> always credited to someone else in the log)
> but although Emacs bugs are fun to fix,
> I swore off that particular time sink,
> yet here I am falling off the wagon.
> 

Thanks, maybe Martin will have some ideas.