Date: Fri, 20 Oct 2000 17:31:45 +0100
To: lwn@lwn.net, debian-devel@lists.debian.org
Subject: Re: Comments on FHS testsuite run

Just to followup on Wichert's note (that I saw posted on LWN at
http://www.lwn.net/daily/deb-testsuite.php3 ).

We should not be expecting any distributions to pass the current version
of the test suite.  Although we believe it to be a fair and accurate test of
the
LSB FHS 2.1 specification, there are issues with the specification and tests
that need to be resolved. The policy with test development is that we
test the specification "as is", and its the specification owners that
get to judge whether the spec is right or otherwise and not the test
suite developers. Going forward if we were to roll out a formal
certification process, and issues are agreed with the spec
owners then we will modify the test accordingly or issue waivers.

I attach the current issues list below that we have identified
with the FHS2.1 specification from the testing efforts.

regards
Andrew

LSB Test leader, and author of the LSB-FHS test suite


FHS2.1 Issues List (Draft 1)
------------------------------
Thu Apr 20 10:14:39 BST 2000

This file contains a list of possible issues with the FHS2.1 specification
based on assessing test results.
For each a specification reference and the associated test assertion
are quoted together with some commentary.

We will need to assess where the problems lie, in the specification
or in the test suite.


Reference 3.4-3(A)
If the implementation provides a C-shell
     The implementation provides the file /etc/csh.login

Is this a mandatory requirement that all FHS compliant systems
provide a default /etc/csh.login

Reference 3.4-4(A)
The implementation provides the file /etc/disktab

We  understand that disktab is a BSD filesystem/partition related file.
Is this meant to be /etc/fstab? or something else ?


Reference 3.4-10(A)
The implementation provides the file /etc/confissue

Should this really be /etc/issue?

Reference 3.4-15(A)
The implementation provides the file /etc/mtools

Should this be mtools.conf ?

Reference 3.4-21(A)
The implementation provides the file /etc/ttytype

This appears to be Linux specific, at best it ought to be
in the Linux annex, or perhaps removed.


-----------

These may only be present if you have configured
your machine accordingly, the FHS mandates them . Is this
correct or should these be made optional.

Note: If you look at the scripts in /etc/rc.d/init.d/ many of
them look for the existence of a file before attempting to
startup a service; therefore, creating empty configuration
files may cause problems.


Reference 3.4-22(A)
The implementation provides the file /etc/exports

Reference 3.4-23(A)
The implementation provides the file /etc/ftpusers

Reference 3.4-24(A)
The implementation provides the file /etc/gateways

Reference 3.4-29(A)
The implementation provides the file /etc/hosts.equiv


Reference 3.4-30(A)
The implementation provides the file /etc/hosts.lpd

Reference 3.4-32(A)
The implementation provides the file /etc/networks

----

The following may fail on architectures other than x86, and hence it
could be argued that the FHS2.1 specification is x86 specific:

Reference 3.1-29 (A)
The implementation provides an exec-able version of the setserial
utility in the /bin directory.

Note this also appears Linux specific and ought to be in the Linux
annex, some system might not support it.

Reference 3.4-12(A)
The implementation provides the file /etc/lilo.conf

This is only true for x86, on sparc its silo.conf, on
alphas milo.conf , so this really needs to be architecture specific.


-----
Andrew Josey
The Open Group