Re: Recent checkins

[Roland McGrath]

> Right.  Fixing the kernel protocols is a part of fixing things for
> network transparency.

I think we're going to have a different kernel before we have a netmsgserver.

> Perhaps I misunderstand then.  How did the alpha generate these
> warnings you were fixing?  

Most of them were code written with "unsigned" or "u_int" where it was
really dealing with size_t (libc interfaces) or mach_msg_type_number_t
(stubs), or "int" and "unsigned" that should have been integer_t and
natural_t (Mach interfaces).  I've decided that
size_t==mach_msg_type_number_t is a reasonable principle, and it happens to
hold everywhere right now, so I used size_t throughout in place of
mach_msg_type_number_t.  

> I thought the alpha used 64 bit "int", but perhaps I'm remembering wrong.
> (Ah, is it that the alpha has 64 bit "void *"?)

You are.  All the "64-bit APIs" that I am aware of differ from the
canonical 32-bit machine types only in long and pointer sizes.  The other
relevant question is the portable-purpose machine-specific types like
size_t and ptrdiff_t, which are pointer size (and intptr_t of course).

> No, I did get that point, and if that is the right judgment, then we
> can just entirely punt the issue as you suggest.

Done and done!