bug#19892: 25.0.50; hideshow: hs-hide-all-non-comment-function example infloop

[Lars Ingebrigtsen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> in the header of hideshow.el, we have the following paragraph:

[...]

> ;;   (defun ttn-hs-hide-level-1 ()
> ;;     (hs-hide-level 1)
> ;;     (forward-sexp 1))
> ;;   (setq hs-hide-all-non-comment-function 'ttn-hs-hide-level-1)
>
> But this doesn't always work.  For example, eval the above example in
> emacs -Q, open "files.el", M-x hs-minor-mode, M-x hs-hide-all.  You get
> an infloop.
>
> AFAICT the definition should be (or at least it works with that)
>
> (defun ttn-hs-hide-level-1 ()
>   (when (hs-looking-at-block-start-p)
>     (hs-hide-level 1))
>   (forward-sexp 1))

I've confirmed that the example still infloops in Emacs 27, and that
your example doesn't, so I've committed it to the Emacs trunk.

> Secondly, there is this comment in `hs-hide-all' which confuses me a bit:
>
> ;; Go to end of matched data to prevent from getting stuck
> ;; with an endless loop.
>
> Which match data is meant there?  It is either match data from before
> hiding the block - then it should be documented that
> `hs-hide-all-non-comment-function' must not change match data, I guess,
> or the call should be wrapped into `save-match-data'.  Or it is even the
> case that `hs-hide-all-non-comment-function' must set the match data
> (how?), which then should probably be documented.

Here's the code:

             (progn
               (goto-char (match-beginning 1))
               (unless (if hs-hide-all-non-comment-function
                           (funcall hs-hide-all-non-comment-function)
                         (hs-hide-block-at-point t))
                 ;; Go to end of matched data to prevent from getting stuck
                 ;; with an endless loop.
                 (goto-char (match-end 0))))

`hs-hide-block-at-point' calls a lot of functions that change match
data, so presumably `hs-hide-all-non-comment-function' is allowed, too.
But this all seems nonsensical -- what `match-end' points to here seems
pretty random.  So it looks like a bug to me.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no