Re: Debian or vanilla kernel - best of both worlds possible?
On Thu, Aug 12, 2004 at 03:06:07PM +0200, Marc Haber wrote:
> - However, Debian's new policy is to remove non-free parts of the
> kernel.
s/of the kernel//. But AFAICS there's been an exception granted for
sarge, so maybe we can add the firmware back except for those two cases
where the firmware license conflicts with the general kernel license.
> - Kernel Upstream has basically abandoned the end user by declaring
> that the 2.6 kernels being released on kernel.org are not the most
> stable kernels, and that building kernels for use on production
> systems should be the distribution's job.
That's bullshit. Upstream kernels are fine for the end user, they are
not fine and never were for the 1% or less of ultra-extrem enterprise
user that want a tested in all possible variants, certified and changed
as little as possible kernel.
> The good news is that the Debian patch only creates files in debian/,
> which will make it apply cleanly even to vanilla kernel sources. Is it
> planned (documented as a committment of the kernel team) that this
> will stay that way? If yes, one could check out the Debian
> subdirectory without checking out the "real" kernel, which would
> greatly help in reducing download volume.
Yes, it's our plan to keep it that way.
> Additional good news is that you guys use dpatch to actually change
> the kernel source. Do I see correctly that for example
> drivers-net-tg3-readd.dpatch was created because the .orig.tar.gz
> already had tg3 removed?
Yes, but that's probably going to change a little for the 2.6.8-based
kernel (Jens wrote a sed snipplet to just remove the firmware images
from tg3.c)
> The patch files themselves don't contain too much information about
> the character of the patch. Short comments inside the patch files
> themselves saying "fix local root exploit (CANxxx-yyy)" or "re-add tg3
> support which was removed in orig.tar.gz", or "fix 'does not build'
> type error on $OBSCURE_ARCH" would be very appreciated. There seems to
> be documentation in README.Debian, but I am not too sure whether that
> list is current, and I am missing the reference to the patch file. Did
> I overlook something here?
All patches have some comments like:
## DP: Description: Add 3ware 9xxx SATA-RAID driver
## DP: Patch author: Adam Radford <linuxraid@amcc.com>
## DP: Upstream status: merged after 2.6.7
or
## DP: Description: chown permission check fix for ATTR_GID
## DP: Patch author: Chris Wright
## DP: Upstream status: backport
> Are the .svn directories inside kernel-source-2.6.7_2.6.7-4.diff.gz
> left there intentionally?
I don't think so. Anyone else in the kernel team thinks keeping the
.svn dirs makes any sense? If not what's the beast way to exclude them?
> The Description of kernel-patch-debian-2.6.7 says that it should be
> applied to a pristine 2.6.7 kernel. Does that still apply now that
> some drivers have been removed from the pristine kernel sources?
No. And in fact because of that the idea of a kernel-patch-debian
doesn't make much sense at all in this new .orig.gz is different from
upstream world.
Reply to: