Now a regex variant;) I don't know if you'd want to use it but it does demonstrate building a regex at run time.

while ( my $line = <DATA> ) {
    my ($new, $old) = split /\s+/, $line;
    my $rx = regex($new);
    print "$new can ", ( $old =~ /$rx/? '': 'NOT ' ), "be made from $o
+ld\n";
}

sub regex {
    my %letters;
    $letters{$_}++ for split //, lc shift;
    my $regex = join '', map { "(?=(?:.*${_}){$letters{$_}})" } keys %
+letters;
    return qr/\A$regex/si;
}

__DATA__
dog good
food fodder
lot total
fuse useful
poor porridge
root rotor
__END__
dog can be made from good
food can NOT be made from fodder
lot can be made from total
fuse can be made from useful
poor can NOT be made from porridge
root can be made from rotor
[download]

I'll show the regular expression generated from 'root' in detail

my $regex = qr{
\A                  # anchor at start for efficiency
  (?=               # positive lookahead for
    (?:
      .*            # anything
      r             # with an 'r' after it
    ){1}            # at least once
  )
  (?=               # positive lookahead for
    (?:.*t){1}      # a 't'
  )
  (?=               # positive lookahead for
    (?:.*o){2}      # for two 'o's
  )
}six;               # ignore case
[download]

All the lookaheads have to succeed for the regex to match.

On a side note I benchmarked the variants (?:.*r), (?:[^r]*+r) and (?:[^r*]r) on different input strings and found the first was usually fastest.