> That is why we have {ge,pu}t_unaligned_[bl]e*()...

Aha! I knew I was missing something. Thanks, Geert!