I believe that 0.3 improves the situation, because it now passes `-cpu
max` rather than `-cpu host` and thus improves the fallback behavior.
Additionally, I think this is a bug in Ubuntu's infrastructure, because
the `/dev/kvm` device is writable to the autopkgtest user and not
working. Earlier
https://salsa.debian.org/helmutg/bootstrap-usrmerge-
demo/-/commit/49a35aa03472277b80406e152ee96989d8081d45 is my change for
the Debian side (which has not reached unstable yet) and adds a safe-
guard to libc6.preinst. I think that would be reasonable to include in
Ubuntu too.
--
You received
Thank you. Can you please verify which of /bin, /lib, /lib64, /sbin are
symlinks on that source system?
If they are all symlinks there, your rsync copy somehow managed to
unmerge the system. If they are all directories there, please try to go
into more detail how that baremetal 22.04 system was
Thank you. Quite clearly, that system is not /usr-merged, but should be.
Not being /usr-merged is unsupported at this time. The question arises
how you got there. Can you explain what kind of system that is? Is it a
regular installation, a container, a chroot or something else? Was it
created
I don't fully understand the situation at hand, but let me add some
background.
From the log it seems fairly clear that `lib6.preinst` is the thing that
fails here. It adds two diversions:
dpkg-divert --quiet --add --no-rename --package base-files --divert
"/.${rtlddir#/}.usr-is-merged"