python data_error ai_generated true

FAILED tests/test_state.py::test_reads_cache - KeyError: 'user:1' (Passes with `pytest -p no:randomly`, fails with random order)

ID: python/pytest-random-order-flaky

Also available as: JSON · Markdown · 中文
80%Fix Rate
84%Confidence
0Evidence
2024-11-02First Seen

Version Compatibility

VersionStatusIntroducedDeprecatedNotes
3.x active — — —

Root Cause

pytest-randomly shuffles test order each run; a test depends on state created by another test (shared cache, module global, DB row), so it fails when it runs first.

generic

中文

pytest-randomly 每次运行打乱测试顺序；某测试依赖其他测试创建的状态（共享缓存、模块全局、数据库行），因此先运行时失败。

Workarounds

  1. 93% success 
    Make the test self-contained with a fixture that seeds the cache:
@pytest.fixture
def seeded_cache(cache):
    cache.set('user:1', {'name': 'alice'})
    return cache
def test_reads_cache(seeded_cache):
    assert seeded_cache.get('user:1')['name'] == 'alice'
  2. 90% success 
    Or reset shared state between tests:
@pytest.fixture(autouse=True)
def clean_cache(cache):
    cache.flush_all()
    yield
    cache.flush_all()

Dead Ends

Common approaches that don't work:

  1. 85% fail

    pytest-randomly ignores order markers unless `-p no:randomly` is set; the shuffle still happens.

  2. 70% fail

    Makes the order deterministic but still hides the inter-test dependency; a different seed on CI breaks it.