Re: Uploading linux-2.6
On Thu, Jan 28, 2010 at 10:50:16PM +0000, Ben Hutchings wrote:
> On Mon, 2010-01-25 at 20:19 +0100, Andreas Barth wrote:
> > * Ben Hutchings (ben@decadent.org.uk) [100125 20:14]:
> > > On Mon, Jan 25, 2010 at 08:02:31PM +0100, Andreas Barth wrote:
> > > > * Julien Cristau (jcristau@debian.org) [100125 19:27]:
> > > > > On Mon, Jan 25, 2010 at 18:56:47 +0100, Luk Claes wrote:
> > > > >
> > > > > > I guess this means that the next version is no candidate for the release
> > > > > > unless it gets a stable ABI (versioning) and should block the kernel
> > > > > > from migrating for the time being?
> > > > > >
> > > > > The 2.6.30 kernel and the current 2.6.32 one aren't candidates either,
> > > > > so I'm not sure what difference blocking the next one makes.
> > > >
> > > > That our testing users don't have to life with strange error messages
> > > > they wouldn't get if the abi would be bumped properly.
> > >
> > > OK, maybe we should start numbering ABIs with this next version.
> >
> > I'd appreciate that very much.
>
> There has been some argument over whether we should do this or not.
> Bastian and Max seem to think that many more ABI-changing config changes
> will still be needed
seems likely
> and do not want to start numbering yet. I'm not
> convinced that changing the ABI number repeatedly in testing/unstable
> will even hurt that much since we no longer have to worry about
> dependent module packages in the archive. It may add to cruft on users'
> systems but this can be fixed after a reboot with 'apt-get autoremove'.
yeah, i think people running sid/testing who do use oot modules
shouldn't have much problem w/ frequent module rebuilds - esp if the
abi is silently changing and they may really need to rebuild anyway.
[Like with a module on my mythbox recently :)]
> I do want to see this argument resolved before building and uploading,
> but I also don't want to hold back security and stability fixes.
Reply to: