The 'Tor is ready' step fails when any systemd service fails
Extracted from #12617:
[Any systemd service failing] breaks all automated tests using Tor, since the
Tor is readystep runs
systemctl is-system-running. The placement of that check in that step is my second issue: it doesn't belong. It should be moved into a dedicated step, run in a scenario about the health of the system. A failure would then in itself immediately signal something quite descriptive about the actual failure, instead of falsely signaling that something about little-t tor is utterly broken, blocking a huge part of the test suite.
Well, IMO it's good that we run
systemctl is-system-runningas often as we can, before testing other stuff, to identify issues as early as possible. IIRC we can't run it for offline scenarios since this check will fail if Tor hasn't started or bootstrapped. But I agree the current way this is implemented gives suboptimal feedback. Perhaps renaming the "Tor is ready" step to describe more accurately what we test in it would help.
Here's a systemd journal from https://jenkins.tails.boum.org/job/test_Tails_ISO_feature-13541-save-more-data-on-htpdate-or-tor-failures/75/ that had this problem.