bug#24240: 25.1.50; window-state-put, image-mode and window scrolling

[martin rudalics <rudalics@gmx.at>]

 >> But do you mean that setting NOFORCE alone will handle your bug already
 >> and we don't need ‘frame-after-make-frame’ after all?
 >
 > Yes, it would.

OK.  Pushed to master.  In any case it fixes a scenario like the following:

(let (window state point start)
   (find-file "../lisp/window.el")
   (forward-line 30)
   (setq window (selected-window))
   (setq state (window-state-get window))
   (setq start (window-start window))
   (setq point (window-point window))
   (message "Before split - %s  ... start: %s ... point: %s" window start point) (sit-for 3)
   (split-window window)
   (message "After split - %s  ... start: %s ... point: %s" window start point) (sit-for 3)
   (window-state-put state window)
   (sit-for 0)
   (message
    "Final - %s ... start: %s -> %s ... point: %s -> %s"
    window start (window-start window) point (window-point window)))

 > At least in the case of image-mode, it would actually be preferable to
 > let window-state-put override image-mode-reapply-winprops's scrolling,
 > because it has the correct value.  I don't know how familiar you are
 > with image-mode,

Not at all.

 > but if it has no record for a particular window, it
 > restores the value last set in any window.  The difference would be
 > visible in case the buffer is shown in more than one window (with
 > different scroll values).

And you mean that the situation after your fix is as expected and the
single, final call of ‘window-configuration-change-hook’ I proposed
earlier would spoil that?  Technically spoken, it would fail because
‘window-state-put’ creates windows anew and for a new window image-mode
has no record of the previous position of the image in that window?  Do
you agree with that interpretation?

martin