On 2 Nov 2015 10:09 pm, "Alan Mackenzie" <acm@muc.de> wrote:
> Of course, nothing is ever that simple.  ;-).  I've got a working
> version which doesn't use "(sit-for 0)", thus allowing point to move
> properly over Follow Mode windows.
>
> However, this meant the "old" highlighting not being erased until the
> 0.25 seconds had passed.  So I tried testing the conditions for a new
> lazy-highlight loop (apart from window-start/end changing) in the
> command loop bit.
> It struck me that instead of all this rigmarole, and instead of the
> "(sit-for 0)", we could test whether or not a scroll is about to take
> place.

Or we can cleanup the lazy highlighting _when_ scrolling takes place.

> What we could do with is an interface which will tell us whether or not
> redisplay would scroll if invoked immediately, with the current value of
> point, window-start, etc.

I don't think that's necessary. What if isearch just added
 (lazy-highlight-cleanup t) to window-scroll-functions?

I want to be very cautious about not increasing the complexity of isearch
code here, but so far these sound like good refactorings to me.