[2101] in Kerberos_V5_Development
Re: Krb5-1.0 and t_kdb - It works on HP
daemon@ATHENA.MIT.EDU (tytso@MIT.EDU)
Tue Dec 10 01:25:12 1996
Date: Tue, 10 Dec 1996 01:22:49 -0500
To: deengert@anl.gov
Cc: krbdev@MIT.EDU, Ken Hornstein <kenh@cmf.nrl.navy.mil>,
Paul A Vixie <paul@vix.com>, Doug Engert <deengert@anl.gov>,
Andrew Hobson <ahobson@mindspring.com>,
Dave McGuire <mcguire@neurotica.com>, Joe Gross <jgross@uiuc.edu>
In-Reply-To: <32A9E552.1646@anl.gov> (message from Doug Engert on Sat, 07 Dec
1996 15:44:50 -0600)
From: tytso@MIT.EDU
Date: Sat, 07 Dec 1996 15:44:50 -0600
From: Doug Engert <deengert@anl.gov>
I got the source on Friday, and built it from home on the HPUX 10.10.
The t_kdb will run correctly if the test_db is is a real unix
directory. The problem appears to be a file locking problem.
The build environment I am using on all the platforms is AFS, but
on the HP it is via the NFS/AFS translator. The t_kdb builds the
test_db in the current directory, then tries to do some file locking
which appear to fail. It works if the directory is on the local system.
You may which in some future version to change the test to put the
test DB in some other directory.
I'll put a mention in the build instructions stating that you should do
the build on local disk, because there may be some problems with the
test suite if you are building via some remote filesystems.
The problem is that there is no good place to put the test DB. On other
operating systems, /tmp or /usr/tmp is a tmpfs filesystem which also
violates the POSIX standard. So there is no good guarantee, in general,
how to guarantee that the test DB is on a valid filesystem. So it seems
that the only thing which makes sense is to require that you put the
build directory on a real filesystem. (At least for the alpha testers
where it's really important that the test suites be run successfully.)
Note that AFS really does not qualify in almost any way as a real
filesystem.....
- Ted
