On Sun, Aug 03, 2003 at 04:13:02PM +0200, Gaudenz Steinlin wrote: > Sven Luther schrieb: > > >On Fri, Aug 01, 2003 at 08:06:56PM +0200, Thorsten Sauter wrote: > > > > > >> > >>I guess thats the best solution to put all needed images in one udeb. > >>The makefile can copy all images into different directories on the > >>cdrom-image. > >> > We will need some additional logic in the makefile. You have to list > all the possible kernel image names in build/config/arch/linux-powerpc > and adjust the makefile logic in build/make/arch/linux-powerpc
That would be the d-i makefiles ? > >Ok, i will see if i can do that, altough the real problem is to build > >more than just the vmlinux and vmlinux.coff images. I guess we could get > >the images directly out of the arch/ppc/boot/images : > > > > > Sven, have a look at how Herbert Xu creates kernel udebs with > kernel-image-2.4.20-1-i386-udeb. He has an additional source package > for the udebs, which depens on the kernel-image. IMHO this is the best > way to package kernel udebs. You don't need an extra kernel build for the > udebs if you do it that way (even if kernel udebs change but the kernel > stays the same). Ok, i will have a look. The real problem for ppc though is that there are various subarches, which build different kind of images including the subarch specific bootloaders. I don't think this problem exists on i386. I suppose the udeb-source package gets the kernel-image package, and extracts the needed kernel image or something from it to build the udeb ? > I'm not sure, but I think it could be better if these kernel udeb source > packages would be maintained by debian-boot and not the individual kernel > maintainers. Mmm, i would have to look at the source udeb in question before i can give my opinion on this. > Like that we would have better control over the udebs and don't have to > bother > the kernel maintainers every time the content of some udeb changes. We could > also provide more unified udebs across all architectures. But I don't > want to stand > on the feets of the kernel maintainers. If they prefer to do this job, > do it. The problem is that the kernel maintainer for a given arch may have a better knowledge of what modules build or fail to build on said arch or other such problems, does he not ? Not really speaking about me, since i don't own powermac hardware, and have the impression that most powermac people use the benh kernels instead anyway. Friendly, Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
