[1284] in Kerberos_V5_Development
Re: changing libkdb database open/lock semantics
daemon@ATHENA.MIT.EDU (Barry Jaspan)
Tue Jun 4 19:09:09 1996
Date: Tue, 4 Jun 1996 13:05:48 -0400
From: "Barry Jaspan" <bjaspan@MIT.EDU>
To: proven@cygnus.com
Cc: krbdev@MIT.EDU
In-Reply-To: <199606041654.MAA01322@qed.proven.org> (message from Christopher Provenzano on Tue, 04 Jun 1996 12:54:19 -0400)
Yes, I am looking into this bug now and would prefer to find it than
change the implementation to something that is MUCH slower.
If you can fix the bug using the current semantics, that's fine with
me; I only want something that works. Assertions about "much slower"
shouldn't be made without evidence, however. MIT has Quantify for
SunOS and Solaris, so evidence should be easy to obtain; I'm happy to
generate results for my modified libkdb.
I also know that this bug is fairly rare and that restarting the kdc
makes the problem go away. If you have a reproducable test case...
My reproducible test case the kadm5 api unit tests, conducted via
kadmind. The old KDC failed partway through the tests; with my
locking changes, the tests all pass. The tests all used to pass with
the krb5b4 KDC as well; some change introduced since then is causing
the problem.
Barry
