[arch-dev-public] Kernel - vanilla vs patched?

[Dan McGee]

On Nov 12, 2007 6:52 PM, James Rayner <iphitus at gmail.com> wrote:
> On Tue, November 13, 2007 07:54, Aaron Griffin wrote:
> > On Nov 12, 2007 1:42 PM, Alexander Baldeck <kth5 at archlinuxppc.org> wrote:
> >> I was never even near being worried about Tobias' decisions for
> >> kernel26. Most times when I actually touched the PKGBUILD was to get a
> >> rc-kernel up on my machine to try it out. Now that I have .24-rc2
> >> running here, I have to say that if it weren't for Tobias' work, I
> >> wouldn't have a working system to this day. Even though I don't know
> >> what he fixed as there are no PKGBUILDs provided yet, a vanilla version
> >> just plainly fucked my machine even before it even tried to detect my
> >> disks.
> >
> > I agree. I didn't mean to imply (as someone else pointed out) that I
> > disliked the decisions that tpowa makes. I just wanted to point out
> > that the kernel is very important.
>
> We're not out against tpowa, we're suggesting improvements. So far mentioned:
>
> 1) Documentation of patches. Where are they from? What they do? Why are
> they included? (mostly done now)
> 2) Re-evaluating patches with each release... deally we shouldnt see
> patches in 2.6.22 that have been updated from 2.6.12 or even 2.6.21.
> 3) Proper review of patches. tpowa's done an awesome job, and his
> judgement has been great so far -- he isn't infallible as the undervolting
> patch shows.

I'd like to (re)mention one improvement, now that the fire has cooled off. :)

4) I'd love to see our kernel PKGBUILD be a bit less scary. At least
right now, I would not feel comfortable using it to build my own
kernel, replacing the existing Arch one, and having faith that it
would boot up. I think if it was a bit more transparent and made it
easier to drop things in (such as new patches, a different config,
etc) that would be great for both other developers AND the community,
without sacrificing anything technical.

-Dan