[OpenAFS-devel] rxk5 branch is ready; please test

[Marcus Watts]

I managed to steal some time to work on rxk5, so I'm working
on an updated rxk5 patch.  I got something that's at least real
close to working with 1.5.28, and hope to collect some windows
fixes too, so now I'm trying to collect whatever other quick fixes I can.

I'm going to try to summarize Douglas's bug list with my thoughts
on them.

#1 - abort/assert usage.
AFS util/assert.h defines "assert()" as a forced statement.
Doug proposes to make this an expression.  That would work, but...

This looks to me like a name collision problem - there's no reason
why openafs should be defining a macro like this for solaris kerberos.  

I think the "better" answer is to either (a) use the linux (and others) assert()
and only provide the expression form if none exists, or (b) if there are
conflicting non-portable versions of assert() out there, make the AFS
symbol "afs_assert".

#2 add_error_table / remove_error_table
It's another sucky comerr problem.  Probably openafs doesn't need
to provide these at all; it's clear there's no shared api here with mit.

#3 xdr_rxk5_principal undefined.
That's history & indeed should go away.

#4 (a)klog -unwrap & HAVE_NO_KRB5_524
Probably a mistake on my part.  I want to look at what Doug did
and I think there's some other similar parallel fix for stock openafs?

#5 __FUNCTION__
Are you sure "__FILE__||"k5_get_cred"" does what you think?
Shades of PL/1.  C probably thinks that's just a short-circuit boolean And.
In any event, this can be just plain "k5_get_cred".

I do want to thank Doug in any case for what he did and hope he'll
do more - I really do appreciate it and hope to see more from him & others.

				-Marcus Watts