bug#16430: 24.3.50; When fullscreen is triggered from the window manager, I can't resize window (via M-x org-export RET)

[martin rudalics <rudalics@gmx.at>]

 > 2 windows of equal height.
 >
 >> Can you trigger it from
 >> such a configuration by simply doing M-x fit-window-to-buffer RET?
 >
 > Yes, from last emacs -Q, even outside Org Export, you need two windows
 > and M-x fit-window-to-buffer RET will fail sometimes.  But I can't
 > provide a reproducible case for now.

This hints at the org buffer having some special character heights in
it.

 > What happens is that, when `window--resizable-p' is called with the
 > pixelwise parameter set to `t' (from within `window-resize'), then
 > the delta (as pixel height) is too high

Can you tell me what delta is in both fullscreen versions?

 > and `window--resizable-p'
 > returns nil -- and window-resize an error.

Apparently `window-resize-no-error' calls `window--resizable-p' with
PIXELWISE nil yielding t while `window-resize' calls it with PIXELWISE t
yielding nil.  So rewriting `window-resize-no-error' as


(defun window-resize-no-error (window delta &optional horizontal ignore pixelwise)
   "Resize WINDOW vertically if it is resizable by DELTA lines.
This function is like `window-resize' but does not signal an
error when WINDOW cannot be resized.  For the meaning of the
optional arguments see the documentation of `window-resize'.

Optional argument PIXELWISE non-nil means interpret DELTA as
pixels."
   (unless pixelwise
     (setq delta
	  (if horizontal
	      (* delta (frame-char-width (window-frame window)))
	    (* delta (frame-char-height (window-frame window))))))

   (when (window--resizable-p
	 window delta horizontal ignore nil nil nil t)
     (window-resize window delta horizontal ignore t)))


should fix it.  Maybe I should just wrap this into a `condition-case'
instead of doing my own checking.

BTW what do (frame-pixel-height) and (frame-height) respectively give in
the two fullscreen versions?

martin