Dear all,

I've committed the attached simple patch after discussion with
Steve (see PR).  We need to reject alternate return arguments
of FINAL subroutines.

Regtested on x86_64-pc-linux-gnu.

Thanks,
Harald