Then I'll break those substrings in parts, if abc is repeated i suspect also bc is repeated, isn't it?

I personally believe that if it were only a suspect it wouldn't be enough. The nice part is that it's obviously certain it is! Anyway, I like your approach very much. For completeness I'm recasting your code in a sub with a similar behaviour to the ones in previous code:

sub oha {
    my $s=shift;
    my %saw;

    while($s =~ /(..+)(?=.*?\1)/g) {
        for my $x (0..length $1) {
            @saw{ map {substr $1, $x, $_} $x+2..length $1 } = ();
        }
    }
    $saw{$_} =()= $s =~ /\Q$_/g for keys %saw;
    \%saw;
}
[download]

Update: I see that you chaged your nodes content and that the original code is not there anymore. I recommend you to only post updates instead, and if you feel that something is wrong and needs to be "deleted", then possibly use <strike> tags. To keep the visual size of your node limited for those that do not want to read all of its details, you can also adopt <readmore> tags.