bug#17453: Isearch doesn't work properly with Follow Mode.

[Alan Mackenzie <acm@muc.de>]

Hello, Juri.

On Wed, Nov 04, 2015 at 02:28:08AM +0200, Juri Linkov wrote:
> >> >> This is complicated.  Ideally, the Follow Mode windows should be
> >> >> synchronised in FM's post-command-hook, not isearch's.  It is not
> >> >> isearch's job to realign windows.  follow-post-command-hook both realigns
> >> >> windows and choses an appropriate window to put point in.  We should let
> >> >> it.

> >> > Once again, if some code in Isearch calls the same function that is
> >> > used in follow-post-command-hook, the above is not an issue.
> >> > Moreover, saving some calls to the hook will make Emacs more
> >> > responsive.

> >> I agree with Eli and Juri on this. If there's a solution as simple as
> >> calling a follow-mode function in isearch-post-update-hook, then this
> >> sounds like a no-downside solution.

> > I'm wondering if we're still talking about the same problem.  ;-)  A
> > simpler solution is _not_ to call a FM function from that Isearch hook.
> > Unless we're talking at cross purposes, there is simply no need.  As
> > long as the Isearch command is allowed to go to completion without
> > forcibly redisplaying, FM will re-synchronise the windows (if needed)
> > and select an appropriate window for point, all on its own (in
> > follow-post-command-hook).

> It still might help to synchronise the windows from isearch-update-post-hook
> if we'll call it before calling isearch-lazy-highlight-new-loop with sit-for.

I still say, wait until we really need it before we do anything so
drastic.  As Eli noted, follow-post-command-hook is SLOW, SLOW, SLOW.
If we call it twice per command, it will be twice as slow.

Also, why is the "(sit-for 0)" there at all?  As its comment says, It
is there for one purpose, and one purpose only: it is so that
(window-start) is valid, and the check

    (not (= (window-start)
            isearch-lazy-highlight-window-start))

will work.  This check means exactly "has the window scrolled?".

> I see no problem in changing the order of calls to isearch-update-post-hook and
> isearch-lazy-highlight-new-loop in isearch-update.  Sure, follow-post-command-hook
> will be called twice but at least this simple solution for follow-mode
> doesn't require re-designing the whole lazy-highlighting machinery.

In one of my mails yesterday, I proposed removing the (sit-for 0) and
replacing this check on (window-start) with

    (redisplay-would-scroll-window-p)

.  This would fix the bug without any further changes.  It would avoid
any far reaching change in design of the lazy highlighting, avoid
calling follow-post-command-hook twice, yet would work.

> diff --git a/lisp/follow.el b/lisp/follow.el
> index 938c59e..75c2788 100644
> --- a/lisp/follow.el
> +++ b/lisp/follow.el
> @@ -420,6 +420,7 @@ (define-minor-mode follow-mode
>    (if follow-mode
>        (progn
>  	(add-hook 'compilation-filter-hook 'follow-align-compilation-windows t t)
> +	(add-hook 'isearch-update-post-hook 'follow-post-command-hook t t)
>  	(add-hook 'post-command-hook 'follow-post-command-hook t)
>  	(add-hook 'window-size-change-functions 'follow-window-size-change t))
>      ;; Remove globally-installed hook functions only if there is no
> @@ -432,6 +433,7 @@ (define-minor-mode follow-mode
>        (unless following
>  	(remove-hook 'post-command-hook 'follow-post-command-hook)
>  	(remove-hook 'window-size-change-functions 'follow-window-size-change)))
> +    (remove-hook 'isearch-update-post-hook 'follow-post-command-hook t)
>      (remove-hook 'compilation-filter-hook 'follow-align-compilation-windows t)))
 
>  (defun follow-find-file-hook ()
> diff --git a/lisp/isearch.el b/lisp/isearch.el
> index b762884..1e4a1a5 100644
> --- a/lisp/isearch.el
> +++ b/lisp/isearch.el
> @@ -1011,12 +1011,12 @@ (defun isearch-update ()
>    (setq ;; quit-flag nil  not for isearch-mode
>     isearch-adjusted nil
>     isearch-yank-flag nil)
> -  (when isearch-lazy-highlight
> -    (isearch-lazy-highlight-new-loop))
>    ;; We must prevent the point moving to the end of composition when a
>    ;; part of the composition has just been searched.
>    (setq disable-point-adjustment t)
> -  (run-hooks 'isearch-update-post-hook))
> +  (run-hooks 'isearch-update-post-hook)
> +  (when isearch-lazy-highlight
> +    (isearch-lazy-highlight-new-loop)))
 
>  (defun isearch-done (&optional nopush edit)
>    "Exit Isearch mode.

-- 
Alan Mackenzie (Nuremberg, Germany).