Re: delete-other-frames

[Richard Copley <rcopley@gmail.com>]

On 23 August 2016 at 15:09, Eli Zaretskii <eliz@gnu.org> wrote:
>> Date: Tue, 23 Aug 2016 10:19:59 +0200
>> From: martin rudalics <rudalics@gmx.at>
>>
>> The problematic code in ‘delete-other-frames’ is this:
>>
>>      ;; Delete mon-minibuffer-only frames first, because `delete-frame'
>>      ;; signals an error when trying to delete a mini-frame that's
>>      ;; still in use by another frame.
>>      (dolist (frame frames)
>>        (unless (eq (frame-parameter frame 'minibuffer) 'only)
>>       (delete-frame frame)))
>>
>> Here we try to delete all frames whose `minibuffer' parameter is not
>> `only'.  However, the minibuffer window of a minibuffer-less frame might
>> be on a _normal_ frame as in our example: The minibuffer window of the
>> frame ‘less’ is owned by the initial frame and ‘delete-other-frames’
>> tries to delete that frame before ‘less’, resulting in the error cited
>> above.
>>
>> Should we document the current behavior in the doc-string or should we
>> amend the code?
>
> IMO, amend the code.
>
>> (2) The doc-string says "Delete all frames on the current terminal,
>> except FRAME."  IIUC this doesn't make much sense when FRAME is not on
>> the current terminal.  In that case, ‘delete-other-frames’ will delete
>> all frames on the current terminal and leave the frames on FRAME's
>> terminal alone.  In order to delete all other frames on FRAME's terminal
>> one has to make that terminal current first.  If that is the desired
>> behavior, we should mention it in the doc-string.  Otherwise, we should
>> delete all other frames on FRAME's terminal.
>
> The latter, I think.

Possibly related: from emacs -Q,

C-x b * RET ;; create and switch to buffer "*"
M-x ediff-buffers RET RET RET ;; ediff buffers "*" and "*scratch*"
;; Now attempt to close the main Emacs frame using the window manager

This gives:

Debugger entered--Lisp error: (error "Attempt to delete a surrogate
minibuffer frame")
  delete-frame(#<frame *scratch* 00000004009c5870> t)
  handle-delete-frame((delete-frame (#<frame *scratch* 00000004009c5870>)))
  funcall-interactively(handle-delete-frame (delete-frame (#<frame
*scratch* 00000004009c5870>)))
  call-interactively(handle-delete-frame nil [(delete-frame (#<frame
*scratch* 00000004009c5870>))])
  command-execute(handle-delete-frame nil [(delete-frame (#<frame
*scratch* 00000004009c5870>))] t)
  read-event(nil t 2)
  sit-for(2)
  execute-extended-command(nil "toggle-debug-on-error" "t-d-o-e")
  funcall-interactively(execute-extended-command nil
"toggle-debug-on-error" "t-d-o-e")
  call-interactively(execute-extended-command nil nil)
  command-execute(execute-extended-command)

I've never been sure whether this deserves a bug report, or what
should be the expected behaviour. FWIW when this happens my intention
is usually to kill Emacs. In any case the error message isn't very
user friendly.