On Fri, Jun 18, 2004 at 05:43:36PM +0200, Jens Schmalzing wrote: > Dear Sven, > > > Ok, i have written such a one, well it is fairly basic, and does : > > > > #!/bin/sh > > mkvmlinuz -k $2 -o /boot/vmlinuz-$1 > > The existing mkvmlinuz can do this if you put the line > /boot/vmlinuz-$release in /etc/mkvmlinuz/output . We could also > deduce the locations of the compressed image from the location of the > uncompressed image, but IMHO it's better to bail out.
Well, it doesn't cost much by setting it by hand, unless we want to set /etc/mkvmlinuz/output to it by default. > > Mmm, we should probably check $1 to make sure it is a initrd kernel or > > something such, > > mkvmlinuz works for both non-initrd and 2.4 kernels. Yeah, but what would happen if i tried to install a kernel-image-2.4.25-powerpc-chrp package ? There would be no vmlinux for it, and thus the problem. > > Well, if we are not allowed to do it, then we can forget about > > automatic installation of non yaboot using subarches kernel. Also, > > this file is generated by yaboot-installer on a new install. > > The installation system is different because it's not a package and > therefore Policy doesn't apply. Sure, but we need to find a solution for the normal case, let's say when you use debootstrap to install, or something such. > > Yep, but this script could also be contained in the kernel packages > > or something such. > > It's not necessary for all users of the kernel-image packages. And it > has its uses independent of the official kernel-image packages (I use > it to flange the d-i kernel and initrd together for Newworld Powermacs > example). Ok. I would still like to have kernel-image... depend on mkvmlinuz though. > > A debconf question in the mkvmlinuz postinst would be the way to go, in > > my opinion, since mkvmlinuz would already be postinsted when the > > kernel-image is postinsted, not sure though, maybe the kernel-image > > should pre-depend on mkvmlinuz then or something such. > > > > Also, i believe that a mkvmlinuz debconfed question would be in order > > anyway. It should : > > > > 1) detect the subarch you are on. > > 2) depending on subarch, propose to run mkvmlinuz or not. > > 3) if the user chooses mkvmlinuzization, add the postinst_hook to the > > /etc/kernel-img.conf file. > > 4) maybe warn or something if kernel-img already contains a > > postinst_hook. > > Sounds good. If Policy doesn't allow 3), the postinst could still > check the file and present a warning. Why should it not ? You are asking the user if he would like to have it added, a bit like passwd or whatever asking about changing the groups and such. If this is not allowed by Policy, then the policy is buggy. > > Well, the real problem would be handling of this file by multiple > > users/packages, Manoj, could you give us your opinion on this ? > > I just looked up the relevant section (10.7.4, shared configuration > files) and have to admit that I'm at a loss. I need to think about > this a bit more, and it would help greatly if Manoj commented on the > issue, since he introduced this file and wrote the manpage. Ok, Manoj, any comment (CCing you just in case you didn't catch the discussion on debian-kernel). > > What about the kernel-source ones ? > > There is no kernel-source-2.6.7 yet. I threw together a tarball using > vanilla 2.6.7 and the patches from Christoph and ignored the missing > build-time dependency. The tarball can be found at <URL:http://www.th > eorie.physik.uni-muenchen.de/~jens/kernel-source-2.6.7.tar.bz2>, but > as I said it contains the tainted drivers and should only be used for > testing. Ok, i have seen it, but i will for now use a standard upstream tarball and drop the tg3 patch. As i understand this, we would then only need to reenable the tg3 patch, and replace the upstream tarball to be able to upload the debian package. I hope to have this in the alioth svn repo soon though, and we can then start moving the powerpc package there also. Friendly, Sven Luther
