Adrian Bunk wrote:
> Control: affects -1 src:beaker
[…]
> beaker has a FTBFS that looks similar, without fakeredis installed:
> https://buildd.debian.org/status/logs.php?pkg=beaker&ver=1.12.1-1
Putting aside the question of the beaker FTBFS for a second, this
issue (ie. #1030600, ie. preventing redis from migrating…) is, as I
now believe, caused by the python-fakeredis testsuite being flaky.
I can reproduce this fairly easily:
$ PYTHONPATH=. python3.11 -Wd -m pytest -v
test/test_hypothesis.py::TestString::test
[…]
test/test_hypothesis.py::TestString::test PASSED
======== 1 passed in 6.20s =========
$ PYTHONPATH=. python3.11 -Wd -m pytest -v
test/test_hypothesis.py::TestString::test
[…]
test/test_hypothesis.py::TestString::test PASSED
======== 1 passed in 6.20s =========
$ PYTHONPATH=. python3.11 -Wd -m pytest -v
test/test_hypothesis.py::TestString::test
[…]
test/test_hypothesis.py::TestString::test FAILED
[…]
In fact, Hypothesis is actually detecting this flakiness:
E hypothesis.errors.Flaky: Inconsistent data generation! Data
generation behaved differently between different runs. Is
your data generation depending on external state?
There might be other issues with redis (eg. the beaker FTBFS perhaps),
but given that the fakeredis testsuite is currently nondeterministic,
it's difficult to have something solid to reason from. :)
(For the avoidance of doubt, I don't maintain python-fakeredis.)
Best wishes,
--
,''`.
: :' : Chris Lamb
`. `'` la...@debian.org 🍥 chris-lamb.co.uk
`-
