[2017] in Kerberos_V5_Development
Re: Final issues before 1.0 code freeze
daemon@ATHENA.MIT.EDU (Barry Jaspan)
Mon Nov 25 13:36:18 1996
Date: Mon, 25 Nov 96 13:36:12 -0500
From: "Barry Jaspan" <bjaspan@MIT.EDU>
To: tytso@MIT.EDU
Cc: krbdev@MIT.EDU
In-Reply-To: <9611251745.AA27194@dcl.MIT.EDU> (tytso@MIT.EDU)
(I'm sick and did not wake up until 1pm today. I'll try to be in this
afternoon.)
* make check blowing out on NetBSD and OSF/1 in tests/dejagnu
problem appears to be related to KRB5_KTNAME set to
value in src/kadmin/testing?!?
Based on the subsequent mail since my "I don't get it" message, I now
understand better what is happening. :-) My guesses of the cause of
the problem are:
o Someone changed somethning in tests/dejagnu to use env-setup.sh
instead of env.sh which Mark says does get created. This should be
simple to find and correct, if that's what happened.
o The admin unit tests create
build/kadmin/testing/krb5-test-root/setup.csh that, when sourced, sets
environment variables such that programs run outside of of
env-setup.sh can will use the testing db, profile, etc. It sets
KRB5_PROFILE, KRB5_KDC_CONFIG (or whatever it's called), KRB5_KTNAME,
and the shared-library variables. So perhaps someone sourced that
script, or otherwise set KRB5_KTNAME, before running the
tests/dejagnu.
In any event, it is clear that the source of the confusion that
KRB5_KTNAME overrides the path specified in kdc.conf; I think that
this is correct behavior, but we could decide otherwise. The admin
tests avoid this problem by (sorta by accident) always setting
KRB5_KTNAME to the value in kdc.conf anyway. Clearly, the dejagnu
tests do *not* set this variable explicitly, but depend on its not
being set or being correct. This ought to be an easily fixable bug in
the dejagnu tests, no?
* Update the README file. This should be easy, once we decide
on the disposition of the above three items.
Previous krb5 README files have been pretty utilitarian. This one
needs to be completely rewritten, pretty, and essentially advertising.
Barry
