bug#19869: 25.0.50; default-frame-alist: width versus fullheight

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

Hello,

There is some inconsistent behavior when setting the with of new
frames depending on whether fullheight is specified:


(progn
  (set-face-attribute 'default nil :height 175)
  (setq default-frame-alist
        '((width . 100)
          ;; (fullscreen . fullheight)
          ))
  (frame-width (make-frame)))  ==> 100


OTOH:

(progn
  (set-face-attribute 'default nil :height 175)
  (setq default-frame-alist
        '((width . 100)
          (fullscreen . fullheight)
          ))
  (frame-width (make-frame))) ==> 63


I would expect that the results are equal.

A related bug was #1522, btw.


Thanks!

Michael.



In GNU Emacs 25.0.50.1 (x86_64-unknown-linux-gnu, GTK+ Version 3.14.5)
 of 2015-02-06 on drachen
Repository revision: 5c9ad35f1e5fd8ee5561ef48baac1f6ff20ae679
Windowing system distributor `The X.Org Foundation', version 11.0.11602901
System Description:	Debian GNU/Linux 8.0 (jessie)

Configured features:
XPM JPEG TIFF GIF PNG RSVG IMAGEMAGICK SOUND DBUS GSETTINGS NOTIFY
LIBXML2 FREETYPE XFT ZLIB

Important settings:
  value of $LC_ALL: de_DE.utf8
  value of $LC_COLLATE: C
  value of $LC_TIME: C
  value of $LANG: de_DE.utf8
  locale-coding-system: utf-8-unix

Major mode: Emacs-Lisp

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[martin rudalics]

 > There is some inconsistent behavior when setting the with of new
 > frames depending on whether fullheight is specified:
 >
 >
 > (progn
 >    (set-face-attribute 'default nil :height 175)
 >    (setq default-frame-alist
 >          '((width . 100)
 >            ;; (fullscreen . fullheight)
 >            ))
 >    (frame-width (make-frame)))  ==> 100
 >
 >
 > OTOH:
 >
 > (progn
 >    (set-face-attribute 'default nil :height 175)
 >    (setq default-frame-alist
 >          '((width . 100)
 >            (fullscreen . fullheight)
 >            ))
 >    (frame-width (make-frame))) ==> 63
 >
 >
 > I would expect that the results are equal.

They are here (Gtk+, Lucid, Motif and Windows) so probably your window
manager interferes in some way.  I tried with your second form as my
.emacs and by evaluating it in *scratch* with emacs -Q.

 > A related bug was #1522, btw.

I'm not sure.  Doesn't your behavior depend on doing
`set-face-attribute' first?.

Anyway, this gives us a good case for doing some forensics.  Please
evaluate

(setq frame-size-history '(1000))

(defun window-display-history ()
   "..."
   (interactive)
   (with-current-buffer-window
    "*history*" nil nil
    (let ((history (cdr frame-size-history))
	 element sizes rest)
      (while history
        (setq element (car history))
        (when (eq (car element) (selected-frame))
	 (when (setq sizes (nth 2 element))
	   (insert (format "%sx%s -> %sx%s "
			   (nth 0 sizes) (nth 1 sizes)
			   (nth 2 sizes) (nth 3 sizes))))
	 (insert (format "%s" (nth 1 element)))
	 (when (setq rest (nth 3 element))
	   (while rest
	     (insert (format " %s" (car rest)))
	     (setq rest (cdr rest))))
	 (insert "\n"))
        (setq history (cdr history)))))
   'history)

in *scratch* and evaluate your second form (the one where you set
fullheight).  Go to the new frame you created this way, type

M-x: window-display-history

in it and post the contents of the *history* buffer here.

If you want to do the same when that form is in your .emacs we'll have
to proceed in a slightly different way.  Let's do that later.

Thanks, martin

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

Hello Martin,

I'll do what you suggested later today, thanks.  One comment to what you wrote:

> They are here (Gtk+, Lucid, Motif and Windows) so probably your window
> manager interferes in some way.

I don't think so.  I already did some edebug yesterday, and I saw that
the width of the frame had already been changed when the frame was not
yet displayed.

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

martin rudalics <rudalics@gmx.at> writes:

> M-x: window-display-history

The buffer is empty, probably because you never push something to
`frame-size-history' (is there a part missing in your code?).

Anyway, I think I'm able to debug this myself a bit, as far as it's
Elisp and not C.


Michael.

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

Hello again,

I forgot to mention that this is all reproducible here with emacs -Q.

Apparently the frame is created with the correct width, then the width
changes while calling `face-set-after-frame-default'.  This happens
before the frame is actually displayed.

Here is what I evaluated in emacs -Q:

--8<---------------cut here---------------start------------->8---
(progn
  (set-face-attribute 'default nil :height 175)
  (setq default-frame-alist
        '((width . 100)
          (fullscreen . fullheight)))

  (advice-add
   'face-set-after-frame-default :around
   (lambda (f frame &rest args)
     (message "Width before face-set-after-frame-default: %S" (frame-width frame))
     (apply f frame args)
     (message "Width after face-set-after-frame-default: %S"  (frame-width frame))))
   
  (make-frame))
--8<---------------cut here---------------end--------------->8---

In *Messages* I get

Width before face-set-after-frame-default: 98
Width after face-set-after-frame-default: 63
#<frame emacs@drachen 0xf99350>

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> Here is what I evaluated in emacs -Q:

Next round, now with `internal-merge-in-global-face':

--8<---------------cut here---------------start------------->8---
(progn
  (set-face-attribute 'default nil :height 175)
  (setq default-frame-alist
        '((width . 100)
          (fullscreen . fullheight)))

  (advice-add
   'internal-merge-in-global-face :around
   (lambda (f face frame)
     (message "Width before merging %S: %S" face (frame-width frame))
     (funcall f face frame)
     (message "Width after  merging %S: %S" face (frame-width frame))))
   
  (make-frame))
--8<---------------cut here---------------end--------------->8---

==>

Width before merging default: 98
Width after  merging default: 63
Width before merging bold: 63
Width after  merging bold: 63
...

So it seems `internal-merge-in-global-face' is the culprit.  That's a C
function now.


Thanks,

Michael.

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[Michael Heerdegen]

Hello,

just to be sure, I have rebuilt Emacs.  My built of trunk was very few
days old, and I think I already saw the problem a very long time before
that.  That's why I didn't check with a more recent built.  Anyway, the
bug is gone in the new built, it must have been fixed in one of the last
days.

Sorry for the noise, I guess I have to be more careful in the future.


Thanks,

Michael.

bug#19869: 25.0.50; default-frame-alist: width versus fullheight

[martin rudalics]

 > just to be sure, I have rebuilt Emacs.  My built of trunk was very few
 > days old, and I think I already saw the problem a very long time before
 > that.  That's why I didn't check with a more recent built.  Anyway, the
 > bug is gone in the new built, it must have been fixed in one of the last
 > days.

Too bad.  I wanted to test whether `frame-size-history' could be used
for remote diagnostics ;-)

martin