> The patch looks right; I've checked it in.

Konrad has detected a bug in it:

 >> However there is a new bug now: Duplicated words that are separated by
 >> a newline-character (e.g. "bug" at the end of some line and "bug" at
 >> the beginning of the following line) are no longer detected.

Attached find a version that hopefully resolves this.  I apologize for
the inconvenience.