Hi, I saw on many packages that piupart fails for error not related to the package tested, for example from https://piuparts.debian.org/stable2sid/fail/cinnamon-desktop-data_5.6.1-1.log I spotted the same error on piupart tests of other packages, for example: https://piuparts.debian.org/stable2sid/fail/cinnamon-control-center-data_5.6.1-1.log https://piuparts.debian.org/stable2sid/fail/muffin-common_5.6.3-1.log https://piuparts.debian.org/stable2sid/fail/nemo-data_5.6.3-1.log https://piuparts.debian.org/stable2sid/fail/xapps-common_2.4.2-3.log Is possible solve this or ignore it to make piupart tests fails only with real issue with the package tested please?
I've been dealing with similar spurious error of piuparts when testing some packages that it reports that package purging left files on system: ,---- | 0m45.6s ERROR: FAIL: Package purging left files on system: | /etc/default/locale -> ../locale.conf not owned | /etc/vconsole.conf -> default/keyboard not owned | /root/.ssh/ not owned | /var/cache/private/ not owned | /var/lib/private/ not owned | /var/lib/systemd/coredump/ not owned | /var/lib/systemd/ephemeral-trees/ not owned | /var/lib/systemd/pstore/ not owned | /var/log/private/ not owned | | 0m45.6s ERROR: FAIL: Installation and purging test. `---- Full piuparts log attached. Wondering what is the rational to check not-owned files? Is it safe to just ignore them in piuparts?
not a maintainer, but id say there's a clear rationale to flag these -- installing and then purging a package should not usually create files not owned (here "not owned" means: not owned by a package, ie stuff that should have been purged). I think so. You can work round the issue by passing "--fake-essential-packages=systemd-sysv" to piuparts (if calling from sbuild use $piuparts_opts in .sbuildrc) or you can install systemd-sysv in the chroot that it uses (eg --include argument to mmdebootstrap and --basetargz argument to piuparts) (I think this works installing systemd-sysv before the package being tested, and so these are there when piuparts records the initial state of the system, and therefore don't appear to be "created" when it re-checks later. but i may be wrong. it also makes the test closer to what the user will see imo) I believe that most of these (all but the first 2) are made by systemd (or systemd-tmpfiles?) as part of normal operation -- it's arguably not a bug in systemd (but arguably systemd should tell the admin to consider removing such packages when it is purged? ) I think the first 2 are potentially bugs in whatever makes /etc/default/locale and /etc/vconsole.conf (shoulnt these be labelled conffiles / configuration files?)