There is more than one way to use SUPER than the super() function. super() happens to be nice to look at but does suffer from that drawback. Use the $self->SUPER and $self->super methods instead. Those *don't* suffer from that problem.

You're absolutely right - I misread the error message. The problem was not that it didn't look at the right @ISA but that I was using an anonymous sub.

When I called it like this:

    my ($class,$alias_for) = @_;
    install_sub(
        $class,
        'find_ids', 
        sub {
            my ($class,$params) = @_;
            do_something_with($alias_for);

            # Pass the subname to super, because 
            # this is an anonymous sub
            my $super = $class->super('find_ids'); 

            return $super->($class,$params);
        });
     ....
[download]

... then it works perfectly. And when I benchmarked the overhead of using SUPER, there was only a 1% slowdown.

. Thanks Diotalevi

UPDATE My benchmark was wrong. See Re^6: Using SUPER in dynamically generated subs for correct benchmarks

If you named your anonymous sub, you could use the SUPER module again, with something like:

local *__ANON__ = $alias_for;