fai-mirror for non-fai-system architectures (bug)
C. V.
list-user at backenhoernchen.de
Mon Nov 1 06:47:24 CET 2010
Hi Michael,
sorry that I answer this late.
----- Ursprüngliche Mail -----
> Hi!
>
> [...]
>
> >
> > I used the ''-a'' with i386 and the mirror was build correctly, but
> > still a problem remained: Some most packages installed but some were
> > missing, eg. the kernel. I found that the kernel was there but apt
> > was looking for a slightly different version of the available one.
> > Apt wanted to install an older version but the newest was available
> > on my mirror.
> > So, again digging into ''fai-mirror'' again...
> > I found the problem in the last lines of code.
> >
> > fai-mirror original (last lines):
> >
>
> [...]
>
> >
> > You can see that ''find'' is called several times. That causes
> > problems if ''find'' finds more than one match. And it does that in
> > some cases because not all packages belong to i386. Specially the
> > kernel of the AMD64 Systems is not included in the i386 binary-tree.
> > But since it is included (as a standard) the
> > ''package_config/DEFAULT'' ''fai-mirror'' downloads it anyway.
> > So, the execution of the following statements fail (without breaking
> > the fai-mirror script) and prevent generating the new
> > Release,Package,Package.gz files. This causes the described problem.
> >
> > Now, finally after all this description, I present you my solution.
> > It works, but it might be not perfect. The benefit should also be
> > that you only need to build the mirror with the following command
> > and get a combined i386 and AMD64 mirror (but I did not yet verify
> > this due to time issues):
> >
>
> [...]
>
> Thanks a lot for looking into this problem in such detail. It seems
> that the -a
> option was not properly implemented at some point. This point,
> however, remained
> unclear to me: Which version of FAI are you using? I've looked at the
> code in
> current trunk and unfortunately it seems that the patch you described
> will no
> longer apply: These parts of the code have been rewritten a while ago.
> Could you
> hence please do us a favour: Could you give the experimental packages
> (current
> version should be 4.0~beta2+experimental28) a try? It's absolutely
> possible that
> these suffer from the same brokenness, but it would be very important
> to be sure
> about it.
I am using the lenny backports Version 3.3.5. That seems to be not the newest. I am very busy these days, so I am not able to test it right now. It will probably take at least one moth till I have time to test it.
Regards,
Corren
More information about the linux-fai
mailing list