Package: autopkgtest
Version: 5.24
Severity: important
X-Debbugs-Cc: [email protected], [email protected], [email protected]
(Cc'd: recent contributors to the unshare backend who might have an
insight into what is going on, or at least might be using this backend
more regularly than the autopkgtest maintainers)
On 2022-08-13 17:53Z, autopkgtest commit 960ae522 was passing tests on
Gitlab-CI, and so were my branches as recently as 2022-08-15 22:44Z.
On 2022-08-21 10:18Z, one of my branches failed CI for a reason that was
apparently unrelated to the change in the branch. I retried the pipeline
for known-good version 960ae522 at 2022-08-21 15:54Z and that also failed.
The failure is that in test-unshare (which tests the unshare backend),
several test-cases run autopkgtest against a mock package in which a
test script "debian/tests/$test_name" emits known strings on its stdout,
and assert that autopkgtest's stdout contains those strings.
For example, test_tree_allow_stderr_nobuild_success runs this test script:
#!/bin/sh
echo I am fine >&2
echo babble
and expects to see "I am fine" in autopkgtest's stderr, and "babble"
in autopkgtest's stdout. In fact "babble" does not always appear, causing
a test failure. This appears to be intermittent: sometimes this test-case
succeeds. stderr seems to be reliable, as far as I can see?
Similarly, test_tree_allow_stderr_nobuild_fail_on_exit and
test_tree_norestrictions_nobuild_fail_on_exit assert that test stdout
appears on autopkgtest's stdout, and these assertions are sometimes failing.
Any ideas?
Thanks,
smcv
