Re: Regexp error scan (March 26)

["Mattias Engdegård" <mattiase@acm.org>]

27 mars 2019 kl. 03.10 skrev Paul Eggert <eggert@cs.ucla.edu>:
> 
> Thanks, I installed the attached patch to try to fix those issues.

Thank you, that was quick!

-  (concat (nth 2 (or (assq footnote-style footnote-style-alist)
-		     (nth 0 footnote-style-alist)))
-	  "*"))
+  (let ((regexp (nth 2 (or (assq footnote-style footnote-style-alist)
+			   (nth 0 footnote-style-alist)))))
+    (concat
+     ;; Hack to avoid repetition of repetition.
+     (if (string-match "[^\\]\\\\\\{2\\}*[*+?]\\'" regexp)
+	 (substring regexp 0 -1)
+       regexp)
+     "*")))
 
The repetition-of-repetition check is useful and has caught several
actual errors; this may be the first true false positive that has to
be hacked around. We could suppress the complaint if the inner
repetition is wrapped in brackets, like \(?:a+\)*, but it would reduce
the error-finding power of the check.

In this case I'm not sure the added * makes sense at all; there is
always a single number within the footnote-{start,end}-tag pairs.
Worse, the code goes on and adds yet another + later on, in
footnote-refresh-footnotes, just in case. That makes even less sense.
Likely, both the * and the extra + should go away.

-    ("^[ \t]*\\(:root\\)\\(?:[\n \t]*\\)*{" (1 'css-selector keep))
+    ("^[ \t]*\\(:root\\)\\(?:[\n \t]*\\){" (1 'css-selector keep))

Those brackets around the [\n \t] could go away, too.