On Wed, Jun 13, 2007 at 02:22:37PM +0200, Mike Hommey wrote: > On Wed, Jun 13, 2007 at 02:13:50PM +0200, Pierre Habouzit <madcoder@debian.org> wrote: > > that's the point of my question. The 2.6 kernel in etch do not depends > > upon a 2.5 libc. the ones in unstable do. So in order to upgrade the > > libc to upstable one, you _have_ to install an etch 2.6.18 first. > > > > I'm asking what we can do to ensure people don't get to the point > > where they don't know how to break that chicken and egg issue like in > > #428655. > > > > IMHO we should have some kind of warning in etch ASAP, but I'd be glad > > to have the opinions of the RMs and SRMs on this. > > IANASRM, but IMHO, it would be too late to do this in a point release. > Why not a preinst script in libc 2.5 to disallow upgrade if the running > kernel is 2.4, as well as adding conflicts on etch libc to kernel packages ? ooookay, let's rephrase it again, and please read the bug I mentionned. THe libc _already_ does those check and the chicken and egg problem you guessed does exists, in lenny/sid. *BUT* it can be avoided if the user had a 2.6 kernel from etch before. This is an issue that will bit hard for the lenny upgrades, as nothing in etch _forces_ the users, neither warn them (except the release notes, but I feel it's not really enough) that continuing to work with a 2.4 kernel is stupid. The problem I'm mentioning is an etch -> lenny upgrade path issue. -- ·O· Pierre Habouzit ··O madcoder@debian.org OOO http://www.madism.org
Attachment:
pgp1erHHCuPVB.pgp
Description: PGP signature