>SunOS processes do not call setreuid(N, N). They call setuid() for
>that purpose.

On the other hand,  Ultrix does.


>You keep providing a POSIX justification. That's invalid -- POSIX
>doesn't talk about implimenting setuid in terms of setreuid; because
>POSIX doesn't include setreuid.

I'm not providing a POSIX justfication, at least not directly.
I'm trying to say this:

	* Ultrix does POSIX semantics for setuid() and setgid() at user level

	* The underlying _MECHANISM_ they use is the Ultrix setre{gid,uid}()
	   system call
	* I want NetBSD's ultrix emulation to be good enough to 
	  use Ultrix' /bin/login. This /bin/login wants POSIX
	  semantics from libc.a's setuid() and setgid(). Specifically,
	  on Ultrix, login, running as rot, is a "sufficienlty privileged"
	  process to have real, effective, and saved uids set.
	 	 
	* If NetBSD is going to emulate Ultrix setuid semantics, it has to
	   emulate the Ultrix semantics of setreuid() and setreguid().

I don't seem to  be able to communicate that. I don't understand why.
It's as if you think Ultrix setreuid() should have the same semantics
as SunOS, and it's broken because it doesn't.

>Still, it is an emulation problem. I've got a better solution now,
>which handles 3 of the 4 cases well.

Depending on which case you don't handle, that may or may not
be good enough.

------------------------------------------------------------------------------