[Fedora-packaging] Updated kernel-module-packaging example with ndiswrapper (Was: example kernel-module package)

[Ville Skyttä]

On Wed, 2005-07-06 at 09:39 -0500, Tom 'spot' Callaway wrote:
> > > Personally, since the buildsystem is going to have to treat
> > > kernel-module-* packages differently, I'd like to see it build them like
> > > this:
> > > 
> > > When a make build is done in kernel-module-foo/FC-3/, the buildsystem
> > > assembles the sources and makes a SRPM. It then looks at a list (either
> > > generated at buildtime, or preexistant) of the released kernels for
> > > FC-3, and iterates through each of them, running rpmbuild --rebuild
> > > --define "kver $VERSION". At the end of the loop, we should have all the
> > > binary packages and a single SRPM.

That still doesn't answer the question, so at the risk of parroting, I'm
going to try once again because I don't know if people understand the
question and the problem:

Yes, it's a single SRPM resulting from _that_ build.  It's very likely
that there would be another one with the same NEVR from an earlier build
sitting in the repository already.  More clarification:

- in the beginning, we have only kernel X
- rebuild a kernel-module-foo package for it, results:
  - k-m-foo-1.0-1.src.rpm
  - k-m-foo-1.0-1.X...rpm (all smp, xen0 etc variants and archs)
- the above get pushed to the repository
- then, kernel Y is released, rebuild k-m-foo (which has had no changes 
  in the meantime) for it, results:
  - k-m-foo-1.0-1.src.rpm
  - k-m-foo-1.0-1.Y...rpm (all smp, xen0 etc variants and archs)
  
Do we discard the new k-m-foo-1.0-1.src.rpm or overwrite the one which
is already in the repository?

Or do we solve this with a policy (assuming the process you described
above would be implemented): "always bump the release of the "main"
package of a module package before building it for a new kernel"?
That'd result in only one srpm per module package per released kernel
"family".

> Also, if possible, it would be awesome if the buildsystem could
> automatically do a build when a new kernel goes into FC (for that kernel
> only). Barring that, it would be nice if there was some way for
> kernel-module-* maintainers to be notified of the new kernel update in
> advance.

Sure.  But I'm becoming worried about this discussion already placing
requirements on the build system, which will delay our ability to ship
any module packages even further.  We could start with something much
simpler; several suggestions have already been outlined in this thread.
The day "make build" would allow specifying the arch, we could start
building and shipping the packages, and could improve the process
incrementally.