readv/writev io dumping support

[John Hughes]

Richard Kettlewell <rkettlewell at zeus.com> writes:
> "John Hughes" <john at Calva.COM> writes:
> > The "addr" argument should be a long (cf dumpstr).
>
> I'm don't find myself entirely convinced by this: by using pointers to
> represent addresses in the traced process, strace would avoid problems
> on architectures with longer-than-long pointers.

True, however that's how the rest of the code does it; and this:

> +                        dumpiov(tcp, tcp->u_arg[2], tcp->u_arg[1]);

makes an ugly warning if addr isn't a long; you should either cast
tcp->u_arg[1]
to char* or fix the addr arg.