To avoid duplicates, you can sidestep the symlink issue (because any given file might also be a symlink) by stat'ing everything you process and rejecting anything where the dev/inode pair is the same. In other words:

BEGIN {
  my %seen;
  sub process {
    my $entry = shift;
    my ($dev, $ino) = stat $entry or return;
    $seen{"$dev $ino"}++ or return;
    ... rest of processing ...
  }
}
[download]