[2727] in Release_7.7_team
Re: What's our Open AFS Validation plan?
daemon@ATHENA.MIT.EDU (Garry Zacheiss)
Mon Apr 30 14:36:48 2001
Message-Id: <200104301836.OAA00844@sweet-transvestite.mit.edu>
To: Bill Cattey <wdc@MIT.EDU>
cc: release-team@MIT.EDU, ted@MIT.EDU
In-Reply-To: Your message of "Mon, 30 Apr 2001 16:17:51 -0000."
<AuvN0jtz0001JZNu5o@mit.edu>
Date: Mon, 30 Apr 2001 14:36:25 -0400
From: Garry Zacheiss <zacheiss@MIT.EDU>
Bill,
I think you need to form this question better before we can
answer it to your satisfaction.
What sort of validation are you looking for? I don't expect us
to construct a regression test suite of the type Transarc has used in
the past. A better plan is to expose the client we wish to deploy to
the conditions it would normally see in our environment. This involves
running large software pacakges out of AFS, doing builds of software
into AFS (emacs is a good choice), etc.
I'll point out that our entire GNOME rollout has pretty much
consisted of "grab the sources and frob them until they do what we
want". I'll concede the point that AFS is a different beast from a
window manager, and is more important to our environment, but if you're
looking for iron-clad proof that we won't have any problems, I can't
provide it, and it's a myth to belive we ever had that sort of guarantee
from Transarc.
I'd recommend we proceed as planned so far, with two additional
items of work assigned:
- alexp or one of the f_l's should validate that all third party
software is working correctly on our Linux machines. I believe this
would normally be done anyway in the course of a release cycle. Note
that at least some current third party Linux software is built against
libc5, and so won't run on a 9.0.1 machine until we've installed libc5
compatability packages.
- We should validate compliation of software into AFS on these
machines. I'm willing to take care of this, although it's possible that
Greg may have done this already.
Garry
