which says:

    We need the up-cast here, since my_decimal has sign() member functions,
    which conflicts with decimal_t::size

But decimal_t does not provide `size` field.