I don't use the redhat kernel rpms and instead build my own kernels for
various reasons from the official source taballs. This causes me to
continually have to force rpms to install because they've got dependencies
on the redhat kernel rpms, none of which are installed on my machines. If
the rpms were built to check for the kernel version using 'uname' instead
of querying installed rpms, or at least had a fall back somehow to do this
if no kernel rpms were found, it seems like they would be more robust. Is
there anyway you could do this in your rpms?
Perhaps this isn't so much a bug as a wish list item, and it certainly
isn't only knfsd that has this problem...still it seems buggy
that a rpm that requires functionality in a 2.2.x kernel would say it can't
install cause I don't have that kernel, when I in fact do. In the meantime
I guess I can always force the installation, but I worry then if I'll miss
other real error messages.
Try adding the missing dependencies to /etc/rpmrc, as in
If you need to satisfy specific versions, you can also create a virtual
package that has, aside from the minimal name, vesrsion, ... required by
rpm, in a spec file
Provides: kernel = x.y.z-abc # substititue for x.y.z-abc
Build and install the virtual package to satisfy rpm dependencies w/o
actually installing Red Hat kernel rpm's.