linux-mips
[Top] [All Lists]

Re: Success at last...

To: Rob Lembree <lembree@sgi.com>
Subject: Re: Success at last...
From: ralf@uni-koblenz.de
Date: Wed, 9 Sep 1998 16:03:00 +0200
Cc: linux@cthulhu.engr.sgi.com
In-reply-to: <35F59FCD.F7A6E459@sgi.com>; from Rob Lembree on Tue, Sep 08, 1998 at 05:21:17PM -0400
References: <Pine.LNX.3.96.980904130745.26347A-100000@lager.engsoc.carleton.ca> <35F59FCD.F7A6E459@sgi.com>
Sender: owner-linux@cthulhu.engr.sgi.com
On Tue, Sep 08, 1998 at 05:21:17PM -0400, Rob Lembree wrote:

> > This is really weird.  Can you explain in greater detail what you changed
> > to make sure that it worked?
> 
> I found that IRIX's tar wasn't very careful with modification dates,
> which I don't think is a problem, but the major and minor IDs of the
> dev files were all zero -- clearly broken.  I haven't experimented,
> but I believe this to be the failure.

This seems to indicate to me that we should replace the dev directory
with a script that creates the inodes.  Luckily we already have one which
is called ``MAKEDEV''.  As we already know the representation of the
minor / major device number isn't transparent through NFS, so the MAKEDEV
will have to detect the NFS server's OS and to corrospondingly munge the
device number used as argument for mkdev in a way that after exporting
from the NFS server the client sees what he expects to see.  If that is
at all possible in all cases.  Yuck.  But it's a longstanding problem
which before has already been reported by people using other operating
systems like HP/UX.

  Ralf

<Prev in Thread] Current Thread [Next in Thread>