After feedback from Nick Roberts -- it doesn't seem as if there is
any reason for the strdup.

I left the test for NULL in place, because I can't convince myself
that it is un-necessary.  Therefore the conservative approach is to
leave it in.  I've moved it to earlier so that if it's goint to error,
we won't do unnecessary work.

Version 1.1 of the file has this NULL check in it, right after the
xstrdup.  I would presume that Andrew Cagney would know that xstrdup
won't return if it fails, so I can't explain the test away by saying
he was just checking for xstrdup failure.