I've been looking at d-i sarge, and it seems to be rather broken. Every install I have tried today with sarge initrds and udebs has failed.
I'm seeing many instances where someone makes a change to a package, like hw-detect, that is dependent on a new version of some other package, like busybox, does not record or even notice the dependency, and it gets into testing, and things break. For example, hw-detect currently contains a call to sleep (for some reason); sleep was only added to busybox in the version in unstable. This breaks hw-detect with 2.6 kernels in sarge, but only on laptops with pcmcia. It seems rather likely to me that sid's d-i has less bugs than sarge's, at this point. I'm afraid that sid is diverging too far from sarge for it to be easy to keep maintaining sarge (we're at the point of needing relativly untested backports of some packages to be put into sarge to fix bugs), and that nobody is testing sarge's d-i, or paying much attention to it at all. It's beginning to look likely that I'll have to throw away the whole plan of incremental copying of debs into sarge, and go back to the model of taking a snapshot, hoping that it mostly works, and copying everything to sarge at one go for a release. -- see shy jo
signature.asc
Description: Digital signature
