Python test failures — FAILED test.py::test / AssertionError

Log signatures

FAILED tests/test_x.py::test_name
AssertionError
ModuleNotFoundError
E   assert 3 == 4

What’s happening

"FAILED path/to/test.py::test_name" is pytest's failure line — the :: makes it grep-able and gives you the exact node to rerun. The crucial split is AssertionError vs ModuleNotFoundError: an assertion means behavior drifted (a real test failure); ModuleNotFoundError means the test environment is broken — a missing dependency, a packaging/layout problem (src/ layout without installing the package), or a test importing something that was renamed. A wall of failures that are all ModuleNotFoundError for the same module is one environment bug, not N test bugs.

Fix playbook

1. 1Copy the first FAILED node and rerun exactly that: python -m pytest "tests/test_x.py::test_name" -x -q. Seconds, not minutes.
2. 2ModuleNotFoundError: fix the environment — is the package installed (pip install -e .)? Is the dependency in the right extras group that CI installs? Don't touch test logic.
3. 3AssertionError: read pytest's diff (it prints both sides). Decide which side is right: the code regressed, or the test encodes outdated expectations. Fix exactly one of them.
4. 4Passes locally, fails in CI: compare Python versions, then look for test ordering/state leakage — run the failing test alone, then the full suite with -p no:randomly (or with a fixed seed) to confirm an inter-test dependency.
5. 5Rerun the focused node, then the file, then the suite. Escalate scope only on green.