Scroll position not restored correctly when aborting search about ctrlf

[tslilc]

I'd be interested in helping to solve this problem, but to be sure, is this some issue with the way emacs handles set-window-point? There's not a whole lot going on in ctrlf-cancel, so i imagine this comes from here:

(set-window-point (minibuffer-selected-window) ctrlf--starting-point)

Did you have any details on this bug?

[raxod502]

It probably has something to do with redisplay. After a display change, Emacs does not compute visual line numbers, window lines, etc. until a redisplay has happened. See e.g. code in Selectrum which works around this issue, and you can also see it right in ctrlf-cancel. As well worked around in ctrlf-minibuffer-post-command-hook. (Sheesh.)

Actually, looking at ctrlf-cancel, would moving the redisplay call to before setting the window point work?

[tslilc]

Thanks for the response. I hadn't been using CTRLF for a while, and now using Melpa version ctrlf-20200523.1558 i can't seem to reproduce this any more.

I can confirm that removing the (redisplay) call from ctrlf-cancel re-introduces the issue with every search, but even at it's present location (after set-window-point) it seems to be prohibiting the issue for me. Perhaps you have some other way of reproducing this, though.

[raxod502]

At first glance just now, I cannot reproduce the issue. Closing for now.