Re: [Emacs-diffs] master fd8f724: * src/xdisp.c (overlay_arrows_changed_p): Fix last change.

[Eli Zaretskii]

> From: Stefan Monnier <address@hidden>
> Date: Tue, 28 Feb 2017 15:02:35 -0500
> 
>     (with-current-buffer A
>       (setq-local overlay-arrow-position (copy-marker 10)))
>     (with-current-buffer B
>       (setq-local overlay-arrow-position (copy-marker 5)))
> 
> after redisplay, what should be the value of
> 
>     (get 'overlay-arrow-position 'last-arrow-position)
> 
> ?  If it's 5 and we perform a redisplay, will buffer A be refreshed even
> if no changes occurred in it?

I'd expect the value to be nil, actually, since the buffer from which
you evaluate the above is most probably neither A nor B.

In any case, if you expect a buffer shown in a non-selected window to
be redisplayed just because its overlay-arrow was modified, this never
worked in Emacs.  You can see that clearly by trying the reproducer
file at the end of this message: load it into "emacs -Q" and then type
F9 -- you will see that the overlay-arrow is not updated.  But as soon
as you type "M-x", which enforces a more thorough redisplay, the
updates become visible.  And your recent changes make this worse,
because previously it was enough to type F9 from a window displaying
one of the two buffers which have overlay-arrows, whereas now the
display is not updated in that case.

> The way I look at it, this code fundamentally assumes that the variables
> on that list are not buffer-local (not only because of the use of
> global symbol properties but also because it only considers
> the value of those vars in the buffer which happens to be current).

I think it assumes that at most one buffer changes its overlay-arrows
between two redisplay cycles.

To fix this thoroughly, we probably need to add
overlay-arrow-position, overlay-arrow-string, and overlay-arrow-bitmap
to the list of variables at the end of frame.el which trigger
redisplay of their respective buffers.  But that is not enough,
because redisplay also wants to know whether a buffer whose window is
about to be redisplayed has changed one of these, in order to disable
certain redisplay optimizations.  So maybe we should add a flag to the
buffer object where this could be recorded.

>     % grep -l local.\*overlay-arrow-position **/*.el
>     calc/calc.el
>     gnus/gnus-sum.el
>     mpc.el
>     net/rcirc.el
>     progmodes/compile.el
>     progmodes/python.el
>     %
> 
> And from my understanding of how this code works, we can both end up
> redisplaying unnecessarily, as well as fail to redisplay when needed.

Changes in overlay-arrows always triggered a thorough redisplay.

Here's the reproducer I promised:

(switch-to-buffer-other-window "A")
(insert "a\na\na\na\na\na\na\na\na\na\na\na\na\na\na\na\na\na\na\na\n")
(goto-char (point-min))
(split-window-vertically)
(switch-to-buffer-other-window "B")
(insert "b\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\nb\n")
(goto-char (point-min))
(switch-to-buffer-other-window "*scratch*")

(setq cur-line 1)
(defun my-move-arrow ()
  (interactive)
  (with-current-buffer "A"
    (set (make-local-variable 'overlay-arrow-position)
         (copy-marker (1+ (line-end-position cur-line)))))
  (with-current-buffer "B"
    (set (make-local-variable 'overlay-arrow-position)
         (copy-marker (1+ (line-end-position cur-line)))))
  (setq cur-line (1+ cur-line)))
(define-key global-map [f9] 'my-move-arrow)