Leigh Scott wrote:
Your kmod idea is flawed, the new kernel builds will always be at
least 7 days behind fedora kernel. This is due to 2 pushes being required
to get the new akmods to stable repo, I haven't got enough time to do
more pushes (sign, mash and sync repo's).
That also really calls for more automation. Also, an automated rebuild
against a kernel that is already in Fedora stable updates should go directly
to stable and so require only 1 push, not 2.
But ideally, the automated rebuild would already happen when the kernel is
in Fedora updates-testing and the kmod would then end up in testing. Then,
when the Fedora update goes stable, the RPM Fusion kmod should also (ideally
automatically) go to stable. So you would always be at most 1 push behind,
and the goal is to ensure that the push happens in a timely manner through
better automation. (The push should just be a matter of you clicking on a
button or running a ./push.sh script and not consume your time as it seems
to do right now, judging from your description. And the kmod rebuilds should
really be completely automatic and not require you to do anything at all.)
As for the inevitable synchronization issue that will happen anyway (also
due to mirroring), have you seen my boolean Requires proposal? I think that
implementing that correctly should guarantee that the kernel and the kmod
are only upgraded together (without requiring special handling in DNF as in
the old experimental Yum plugin). In principle, that should even work if the
kmod is 7-14 days behind depending on schedule alignment.
Kevin Kofler