in reply to How do I match the closest of repeated strings?

Looks like the kind of game where 'paragraph-mode' input would be a good idea: 
$/ = "";  # input record separator set to empty string

my $last_rec;
while (<>) {  # each record is terminated by /\n{2,}/
    if ( /^def/ ) {
        print $last_rec.$_ if ( $last_rec );
    }
    else {
        $last_rec = $_;
    }
}

[download]
Check perlvar for more info on $/.

UPDATE: Sorry, I probably misunderstood the problem. If blank lines are actually distributed as shown in the OP, then something like this would be needed in order to use "paragraph-based" input: 

$/ = "";
my $last_prefix = my $last_target = '';
while (<>) {
    if ( /^abc/ ) {
        print $last_prefix.$last_target if ( $last_target );
        $last_prefix = $_;
        $last_target = '';
    }
    elsif ( /^def/ ) {
        $last_target = $_;
    }
    elsif ( $last_target ) {
        $last_target .= $_;
    }
    elsif ( $last_prefix ) {
        $last_prefix .= $_;
    }
}

[download]
That's pretty icky, really -- very sorry. I suggest you just slurp the whole file and use one of the regex solutions from an earlier reply.

Another update: I still did not get the OP's intent -- but no point fixing this code, since other solutions are up there (I think). Never mind me.

In Section Seekers of Perl Wisdom