While working on changing the upgrade tests to use dnf, I found
that upgrades were failing seemingly because /etc/machine-id
was mislabelled (the upgrade process would succeed but there
were lots of errors caused by failed reads of that file, and
the upgraded system would fail to boot). restorecon'ing /etc
while we're doing stuff to the image seems to fix it.
Details
Details
- Reviewers
garretraziel jskladan
Try the upgrade tests (PR coming soon, I hope) with disk
images created without this fix, then images created with this fix.
Also you can just boot a VM with the images and check the context of
/etc/machine-id (and look out for the error messages).
Diff Detail
Diff Detail
- Repository
- rOPENQA fedora_openqa
- Branch
- etc-selinux
- Lint
No Linters Available - Unit
No Unit Test Coverage - Build Status
Buildable 181 Build 181: arc lint + arc unit
Comment Actions
seems this may actually be a consequence of a virt-builder bug, so holding fire for now.
Comment Actions
Yes, this is OK, but I don't think that this should be necessary, because whole system should be relabeled - it seems like there is a bug in virt-builder that is preventing it. As soon as https://bugzilla.redhat.com/show_bug.cgi?id=1261377 gets fixed, this won't be necessary.
Comment Actions
The problem indeed seems to be fixed by the virt-builder update, so no need for this.