bug#69943: 30.0.50; Tabbing through widgets can signal beginning-of-buffer error

[Eli Zaretskii]

> Date: Fri, 22 Mar 2024 15:45:16 +0100
> From:  Stephen Berman via "Bug reports for GNU Emacs,
>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
> 
> 0. emacs -Q
> 
> 1. Evaluate the following sexp:
> 
> (let ((buf (get-buffer-create "*Widget Test*")))
>   (switch-to-buffer buf)
>   (dolist (el '("First" "Second" "Third"))
>     (widget-create 'push-button el))
>   (use-local-map widget-keymap)
>   (widget-setup)
>   (goto-char (point-min)))
> 
> Now the current buffer is *Widget Test* containing three push-button
> widgets labeled "First", "Second", and "Third", and point is at the
> start of the first widget, at BOB.
> 
> 2. Hit the TAB key (bound to widget-forward) three times: this moves
> point successively from "First" to "Second" to "Third" and then back to
> "First" -- but on returning to the initial position after the third TAB,
> a beginning-of-buffer error is also signaled.
> 
> 3. Likewise, hitting S-TAB (bound to widget-backward) three times moves
> backwards across the widgets, from "Third" to "Second" to "First", again
> signaling a beginning-of-buffer error after the last S-TAB.
> 
> These beginning-of-buffer errors are due to widget-move (the workhorse
> behind widget-forward and widget-backward) calling backward-char in a
> loop without checking for BOB.  The attached patch fixes this.  The
> patch also includes additions to widget-test-widget-move (from which
> most of the above sexp was taken) that test moving to a widget at BOB.
> (If the patch is acceptable, whoever commits it should use the correct
> bug# before pushing it, or I can do that myself.)

Mauro, any comments to the proposed patch?

Thanks.