bug#31772: 26.1; (thing-at-point 'list) regression

[Eli Zaretskii <eliz@gnu.org>]

> From:  Leo Liu <sdl.web@gmail.com>
> Cc: 31772@debbugs.gnu.org,  tino.calancha@gmail.com
> Date: Tue, 11 Sep 2018 18:26:36 +0800
> 
> The difference is on what to return when (up-list -1) fails. They both
> try to return the sexp at point but the patched behaviour try returning
> the whole sexp while the one in <= 25.1 cut off the sexp from point.
> 
> Looking at the code in 25.1 it looks to me the original author had a
> thinko i.e. he/she meant to write:
> 
> (if (>= opoint (point))
>     (cons (point) end))
> 
> but instead write
> 
> (if (>= opoint (point))
>     (cons opoint end))
> 
> which resulted in some weird cases that you mentioned in previous email.

Right.

> > Would it be possible to modify list-at-point so that it keeps the
> > current behavior, perhaps as an option?  I'd like to find a solution
> > that doesn't just revert to the old behavior, but allows those who
> > need the new behavior to have it in some reasonable way.
> 
> Yes, this can be done. For example:
> 
> (defun list-at-point (&optional ignore-comment-or-string)
>   "Return the Lisp list at point, or nil if none is found.
> If IGNORE-COMMENT-OR-STRING is non-nil comments and strings are
> treated as white space."
>   (let ((ppss (and ignore-comment-or-string (syntax-ppss))))
>     (save-excursion
>       (goto-char (or (nth 8 ppss) (point)))
>       (form-at-point 'list 'listp))))

Would you mind submitting a patch that includes the above, and also
fixes/augments the tests accordingly?  I think these changes then
could go into the emacs-26 branch, unless someone raises objections.

Thanks.