[101] in athena10
Re: Debathena materials not imported
daemon@ATHENA.MIT.EDU (Timothy G Abbott)
Mon Mar 3 17:18:22 2008
Date: Mon, 3 Mar 2008 17:17:38 -0500 (EST)
From: Timothy G Abbott <tabbott@MIT.EDU>
To: Greg Hudson <ghudson@mit.edu>
cc: athena10@mit.edu
In-Reply-To: <200803032208.m23M8TX6024473@outgoing.mit.edu>
Message-ID: <Pine.LNX.4.64L.0803031710070.17031@vinegar-pot.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
On Mon, 3 Mar 2008, ghudson@MIT.EDU wrote:
> I've imported a couple of missing support files and adapted these
> scripts. I'm running into a couple of problems:
>
> * find-openafs-kernels produces an empty list for dapper-amd64 and
> dapper-i386. apt-file update is getting an empty Contents file.
> I haven't looked into this very much except to note that it also
> happens on debuild.mit.edu if I run the original Debathena script
> in /mit/debathena/packages/third/openafs.
Sounds like the Ubuntu dapper repository stopped generating Contents
files. I don't have a browser handy, but it's probably worth looking for
the Contents.gz in the dapper apt repository...
> * The result of find-openafs-kernels includes a whole bunch of amd64
> packages for sarge-i386, and one for lenny-i386. I don't know if
> this is wrong or merely puzzling to me. I can handwave away the
> amd64 packages for sarge-i386 because Sarge has no amd64 port, so
> maybe it had amd64-specific kernels without a full amd64 suite.
> The lenny-i386 one can't be handwaved.
I'm guessing the lenny-i386 one is intended for using an i386 userspace
with an amd64 kernel (linerva does this, for example).
> * 40 of the builds failed: one on lenny (specifically, the amd64
> package mentioned above), 21 on edgy, 3 on feisty, and 15 on
> gutsy. I notice the debathena locker also lists some failures; I
> don't know if some of them are expected or not. The one failure I
> looked into (the amd64 headers package on lenny) appeared to be
> running into an order-of-includes problem.
I think the edgy ones are the reason we use edgy-backports for our openafs
modules. I recall some gutsy failures related to weird kernels
(real-time, etc.), but 15 sounds high. I seem to recall some of the
things our find-openafs-kernels script found for Gutsy were not
really kernel headers.
I believe building openafs on an i386 system for an amd64 kernel is
generally annoying because openafs uses some questionable architecture
detection; Anders has more state on how to do this than I do.
-Tim Abbott
