bug#63204: 29.0.90; vc-annotate does not properly test whether a face exits

[Philip Kaludercic]

Tobias Bading <tbading@web.de> writes:

> 1. emacs -Q
>
> 2. Imagine that in a previous Emacs session you have copied a string with
>    text properties from a buffer in vc-annotate-mode, maybe also searched
>    for it somewhere. IOW, your ~/.emacs.d/.emacs.desktop file may contain
>    something like this:
>
>    (setq some-search-ring-or-whatever
>          '(#("heisenbug" 0 9 (face vc-annotate-face-CCCCFF))))
>
>    Evaluate this expression to simulate a desktop-read of such a previous
>    session.
>
> 3. Open lisp/vc/vc-annotate.el from the Git branch emacs-29.
>
> 4. C-x v g (vc-annotate)
>
> 5. There are lines with white background caused by the use of undefined face
>    vc-annotate-face-CCCCFF. C-h e should contain:
>
>    Invalid face reference: vc-annotate-face-CCCCFF
>
>    If for whatever reason you don’t see these bad lines, use C-u C-x = to
>    see what face a line uses and repeat these steps with this face instead
>    of vc-annotate-face-CCCCFF.
>
>
> This can be fixed with:
>
> diff --git a/lisp/vc/vc-annotate.el b/lisp/vc/vc-annotate.el
> index 70057a6aac..ae8db021ad 100644
> --- a/lisp/vc/vc-annotate.el
> +++ b/lisp/vc/vc-annotate.el
> @@ -724,7 +724,9 @@ vc-annotate-lines
>                                        (substring (cdr color) 1)
>                                      (cdr color))))
>                 ;; Make the face if not done.
> -               (face (or (intern-soft face-name)
> +               (face (or (let ((sym (intern-soft face-name)))
> +                           (if (facep sym)
> +                               sym))

I'd write this as (and (facep sym) sym).

>                           (let ((tmp-face (make-face (intern face-name))))
>                             (set-face-extend tmp-face t)
>                             (cond
>
> My Emacs Lisp isn’t that great so please feel free to improve this code
> if needed. The important part is that (intern-soft face-name) is not
> sufficient to check whether a face exists.

I think a better solution would be to revise general approach
vc-annotate takes, and instead of hard-coding faces/colours to use (that
break when switching between light and dark themes), to have a list of
faces that could be used independently of vc-annotate.