Hi *, Am So, den 04.07.2004 schrieb Axel Thimm um 14:25: [...] > So all is solved if > > o kernel building splits subpackages into kernel and kernel-headers > o kernel-headers are per arch and flavour and are installed under say > /usr/src/kernel-headers/`uname -r`.<arch> > o /lib/modules/`uname -r`/build (part of the kernel subpackage) > becomes a symlink to /usr/src/kernel-headers/`uname -r`.<arch> > > Benefits: > > o Users can install the kernel-header package matching their kernel and > create all modules they'd like either manually or from kernel > modules src.rpms > o ISVs/Packagers can install as many kernel-header rpms and build > kernel module rpms for the whole lot of them w/o > installing/deinstalling kernel-headers or even whole kernels. > o Vendors are happy they found a solution that work for all > > Drawbacks: > o Users will have to install kernel-header-`uname -r` for building > kernel modules. > > I think the benefits far outweigh the drawbacks. Me too, but as far a I interpret Arjan he is not willing to accept the > o Users will have to install kernel-header-`uname -r` for building > kernel modules. But couldn't that solved if the kernel package itself always Requires the install of the matching kernel-header-`uname -r` package? CU thl
