bug#43265: 28.0.50; Inconsistent fontifying in elisp-mode

[Lars Ingebrigtsen <larsi@gnus.org>]

Mauro Aranda <maurooaranda@gmail.com> writes:

> To reproduce:
>
> emacs -Q
> In the scratch buffer type:
>
> (condition-case nil
>     (fail-badly)
>   (error (when foo t)))
>
> I don't see WHEN with font-lock-keyword-face.  However:
>
> (condition-case nil
>     (fail-badly)
>   (error (if foo t)))
>
> IF has font-lock-keyword-face.

This is the bit that determines whether to give something a keyword-face:

(defun lisp--el-match-keyword (limit)
  ;; FIXME: Move to elisp-mode.el.
  (catch 'found
    (while (re-search-forward
            (eval-when-compile
              (concat "(\\(" lisp-mode-symbol-regexp "\\)\\_>"))
            limit t)
      (let ((sym (intern-soft (match-string 1))))
	(when (or (special-form-p sym)
		  (and (macrop sym)
                       (not (get sym 'no-font-lock-keyword))
                       (not (lisp--el-non-funcall-position-p
                             (match-beginning 0)))))
	  (throw 'found t))))))

All special forms get it (`if' is a special form), but macros (like
`when') only gets it if we're in a funcall position:

(defun lisp--el-non-funcall-position-p (pos)
  "Heuristically determine whether POS is an evaluated position."

[...]

                  (and (eq parent 'condition-case)
                       (progn
                         (forward-sexp 2)
                         (< (point) pos))))))))))


And this doesn't count any of the error clauses as being in a funcall
position, so no macros in those get a keyword-face.  

I think.  I've added Stefan M to the CCs.

If that's the correct analysis, then fixing this shouldn't be too hard,
I think?  Just add some extra code in the `condition-case' bit to catch
this case?

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