Re: [PATCH] Fix moving cursor in org-set-tags-command

[Kyle Meyer <kyle@kyleam.com>]

Matt Lundin writes:

> Commit 44ec473c199262d89b372d8a6cd35bed7672164d from Feb. 23 causes
> org-set-tags-command to move the cursor forward 1 char when situated on
> headline asterisks.
[...]
> This commit modified a previous change on Feb. 21
> (450452de4b790706d187291f9f71a286f8f62004). But that commit also had
> problems, since it would move the cursor one asterisk forward on
> headlines > 1, thus also interfering with org-speed-keys. In my view
> org-set-tags-command should not move the cursor except to fix the very
> specific thing that commit 450452de4b was meant to fix: namely the
> cursor moving when on a blank headline: i.e., from here...

Thanks for the nice description of the problem.  I wouldn't mind if at
least a condensed version made its way into the commit message :)

> I've attached a patch that corrects the problem, but it would be ideal
> if we figured out why the cursor is moving in the first place.

I looked quickly at org-set-tags (and the functions it calls).  Based on
commenting bits out, I think there are at least a couple of parts that
modify the buffer in a way that prevent save-excursion from restoring
the intended location.  While not ideal, this after-the-fact adjustment
is probably the simplest way to deal with the issue.

> Subject: [PATCH 1/1] Fix bug that placed cursor incorrectly when setting tags
>
> * lisp/org.el: (org-set-tags-command) Only fix cursor position in very

nitpick: The colon should follow (org-set-tags-command).

> diff --git a/lisp/org.el b/lisp/org.el
> index dd017e662..0e4fd7be1 100644
> --- a/lisp/org.el
> +++ b/lisp/org.el
> @@ -11846,8 +11846,9 @@ in Lisp code use `org-set-tags' instead."
>  	  (org-set-tags tags)))))
>      ;; `save-excursion' may not replace the point at the right
>      ;; position.
> -    (when (save-excursion (skip-chars-backward "*") (bolp))
> -      (forward-char))))
> +    (and (looking-at " ")
> +	 (string-match "\\*+" (buffer-substring (point-at-bol) (point)))
> +	 (forward-char))))

Looks fine to me, with the minor nit that I think looking-at-p and
string-match-p would be preferable here.

Would you mind adding a regression test?