I understand the benefits of this change. My own raspberry pi systems
failed to upgrade to resolute because of a too small partition.

But I also understand the blast radius of a regression here, and it's
not small.

What can we do to make this safer, and worthwhile? At least the test
plan needs to cover more ground.

Some cases that come to mind:
- test raspberry cases: upgrades to this package (of course), but also release 
upgrades from noble;
- install all the packages that could possibly influence the initrd generation: 
luks, iscsi, other filesystems, nfs root, something ubuntu core perhaps, 
raid-on-root, lvm-on-root, etc. And see if something breaks to a) generate the 
initrd; b) boot from it. I suppose this is a much higher bar than what was 
applied when we switched TO rust-coreutils, but this is an SRU for our latest 
Ubuntu LTS, so the bar is higher on purpose.
- while in a development release, such a change would have much more machinery 
testing it: generation of daily images, release snapshots, cloud image testing, 
and so on. Can we do any of it in the case of the SRU, before the package hits 
updates?
- boot via the rescue option in the grub menu (what is it called again?)

When did we switch to dracut, was it resolute? If yes, that means we
never had the combination of dracut + busybox before?


As for the actual release in updates, if/when that comes:
- we could leave this in proposed for much longer, on purpose -> block-proposed 
tag
- slow down phasing even more, like grub perhaps

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2150657

Title:
  Use busybox instead of rust-coreutils in initrd

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/busybox/+bug/2150657/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to