Closed ktdreyer closed 7 years ago
@tserlin @ahills FYI, found this when I enabled the pytest suite during package builds for eng-rhel-6-candidate
Welp, of course on the final non-test build I hit this on py27 as well. It seems easier to trigger in our internal Koji instance for whatever reason.
Python does not guarantee os.walk returns results in a particular order. In some cases (discovered on RHEL 6 / py26 in Koji, exact repro steps unknown), test_nested_debian_repo could fail when repocollector is a list ordered in a way we didn't expect.
Compare the sets in the nested_debian_repo repocollector test, just like we do for the other repocollector tests.