On Tue, Jun 26, 2018 at 10:41 PM Philip Webb <purs...@ca.inter.net> wrote: > > 180626 Rich Freeman wrote: > > On Tue, Jun 26, 2018 at 8:58 PM Philip Webb <purs...@ca.inter.net> wrote: > >> Does anyone know why the latest stable version of Gentoo-sources is 4.9.xx > >> ? > >> I installed 4.9.16 , which I continue to use, on 2017-04-06 . > >> The tree contains versions of 4.14 4.16 4.17 , but all are still testing. > > I tend to just use my own upstream kernels. I'm following the 4.14 longterm > > and generally update within a few days of any release. > > That said, I have been burned by the odd regression. > > Thanks for the other info (snipped). All Vanilla-sources are testing, > which seems to correspond to your "upstream" kernels.
I use them directly from upstream: git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git They keep a branch for each longterm which makes updating easy. That said, sticking with gentoo-sources certainly won't hurt. > What does this say re recent kernel development or Gentoo's kernel team ? > -- very quick thanks as always to Gentoo's volunteer developers, > but something seems to be going astray here (smile). I don't really see many issues with the Gentoo kernel team. The choice of which longterm to follow is one that has pros and cons, and they're following 4.9 deliberately because of issues some have had with 4.14. MANY distros make decisions like this, and to some degree it seems to be encouraged by the stable upstream kernel maintainers as well who seem to view distros as another line of QA. Within a longterm I'm surprised they aren't a bit more up-to-date, but the reality is that the stable team has been issuing more than one stable release every week for a while now. That is a VERY fast cadence and I've been burned by just following this as their own regression testing seems a bit limited. If the Gentoo kernel team is taking its time to keyword these releases to do actual QA I certainly won't fault them for that, and presumably they push through security updates. None of this is really meant to cast blame on upstream either. Regression testing the kernel seems like a difficult prospect because of all the potential hardware-related issues. Maybe better software regression testing would be possible (filesystems, system calls, etc), but I think a monolithic kernel is always going to be problematic in this regard. (Even with a microkernel a failure of your IOMMU driver or something like that isn't exactly something you can gracefully contain...) -- Rich
