RestartTest.test_last_file_missing_at_end is failing on ppc64el

[Sebastien Bacher]

(from a duplicated bug report)
Ubuntu ran into a problem recently on the ppc64el platform because of the test_last_file_missing_at_end test. It assumed that there would be more volumes than duplicity was generating on that platform. In general, the test is a bit flaky because it guesses how many volumes duplicity will generate given random data.

We should probably change it to fake the sequence of volumes like some of the other tests in that file do.

0.8.01 is failing to build on ppc64el Ubuntu builders due to test failures

https://launchpadlibrarian.net/433437044/buildlog_ubuntu-eoan-ppc64el.duplicity_0.8.01-0ubuntu1_BUILDING.txt.gz

        u"""
        Test restart when the last file being backed up is missing on restart.
        Caused when the user deletes a file after a failure. This test puts
        the file at the end of the backup, with no files following.
        """
        self.make_largefiles()
> self.backup(u"full", u"testfiles/largefiles", fail=6)

functional/test_restart.py:125:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
functional/__init__.py:186: in backup
    result = self.run_duplicity(options=options, **kwargs)
functional/__init__.py:159: in run_duplicity
    self.assertEqual(30, return_val)
E AssertionError: 30 != 0

[Kenneth Loafman (kenneth-loafman)]

rev 1409 should fix this

[Sebastien Bacher (seb128)]

Sorry but it doesn't, that's a build of the package with that revision backported
https://launchpadlibrarian.net/433853749/buildlog_ubuntu-eoan-ppc64el.duplicity_0.8.01-0ubuntu2_BUILDING.txt.gz

[Kenneth Loafman (kenneth-loafman)]

I cannot reproduce this on ubuntu:19.10. It may have to do with the ppc64 architecture.

[Sebastien Bacher (seb128)]

Yes, it's only failing on ppc64el so it's probably specific to that architecture. Let me know if you need debug informations

[Gianfranco Costamagna (costamagnagianfranco)]

Hello, ping? this bug is preventing some transition from landing in ubuntu eoan...

[Sebastien Bacher (seb128)]

The problem got workaround in Debian by Alexander by skipping the test with that comment

    @unittest.skip("Flaky test because it relies on knowing how many volumes the source files will be split into")

[Gianfranco Costamagna (costamagnagianfranco)]

I add this comment from another bug report:

"Ubuntu ran into a problem recently on the ppc64el platform because of the test_last_file_missing_at_end test. It assumed that there would be more volumes than duplicity was generating on that platform. In general, the test is a bit flaky because it guesses how many volumes duplicity will generate given random data.

We should probably change it to fake the sequence of volumes like some of the other tests in that file do."