[ATrpms-users] Support for recent kernels
John Pilkington
J.Pilk at tesco.net
Sat Jan 10 12:34:10 CET 2009
Axel Thimm wrote:
> On Fri, Jan 09, 2009 at 01:09:00PM -0500, Brian Long wrote:
>> On Fri, Jan 9, 2009 at 10:33 AM, Paulo Cavalcanti <promac at gmail.com> wrote:
>>> Axel just keeps kmdls for the latest Fedora kernel. The number of kmdls, if
>>> you take RHEL/Centos into account, is huge.
>> Paulo et al,
>> I think the general point remains valid. Until all kmdls are
>> successfully built for the latest kernel, the previous generation of
>> kmdls should remain inside the repo. Only when new kmdls are 100%
>> built should the repo be cleaned of previous versions.
>>
>> My $0.02.
>
> True, that's what happens in theory.
>
> In practice new kernels very often break external kernel modules,
> quite often even ina way that needs upstream to resolve it. Stalling
> the new support for the latest kernel until there is a resolution to
> evey broken kmdl would last forever (note that there are 129 specfiles
> for kmdls in the repo, not all activated for every distro, but just to
> give an idea of how many kmdls would wait for a broken one).
>
> So when a new kernel comes out the builds system sweeps a build over
> it an the broken ones are examined for a uick resolution, but when it
> looks like stalling more, the new kmdls are released.
>
> Whether the old kmdls are in the repo or not is not that important as
> the new kernel is being coinstalled anyway, you just can't boot into
> the new one if you miss some kmdl, and stalling all kmdls for everyone
> wouldn't help. Also note that a new installation will not even find
> the old kernel in Fedora's repos, so the old kmlds are rather useless
> at that point.
>
Not quite true: if I have an old kernel installed I could install new
hardware using an old kmdl even if the new kmdl doesn't build. But I do
recognise that the kmdl load is very high - and that fc10, at least, is
still work-in-progress.
John P
More information about the atrpms-users
mailing list