bug#21072: Brave new mark-defun (and a testing tool)

[npostavs]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> This suggests that the first arg is not "evaluated by the macro itself
> at compile time", but instead is passed unevaluated to `insert`, in
> which case it can't be instrumented.

Oh, I see, this patch works, although it surprised me a bit to start
stepping immediately when doing the C-u C-M-x.

---   i/test/lisp/emacs-lisp/lisp-tests.el
+++   w/test/lisp/emacs-lisp/lisp-tests.el
@@ -320,12 +320,10 @@ elisp-tests-with-temp-buffer
 of the form =!NAME= in CONTENTS are removed, and a for each one a
 variable called NAME is bound to the position of the word's
 start."
-  (declare (indent 1) (debug (form body)))
+  (declare (indent 1) (debug (def-form body)))
   (let* ((var-pos nil)
          (text (with-temp-buffer
-                 (insert (cond ((symbolp contents)
-                                (symbol-value contents))
-                               (t contents)))
+                 (insert (eval contents))
                  (goto-char (point-min))
                  (while (re-search-forward elisp-test-point-position-regex nil 
t)
                    (push (list (intern (match-string-no-properties 1))