Testing Shared (all languages) active any

Resource-dependent tests are explicitly gated outside default runs

test-004

Intent

Default test runs should not fail nondeterministically because a test silently assumes GPUs, large datasets, unusual hardware, or other special execution environments.

Applicability

Applies when changed tests require special hardware, large external datasets, unusually long runtimes, or environment-specific resources not present in routine developer and CI runs. Return unknown when the suite classification is not visible.

What to inspect

Check test markers, skip logic, suite configuration, resource assumptions, and whether the changed tests are excluded from the default path unless the required environment is present.

Pass criteria

Resource-dependent tests are clearly marked, skipped, or isolated into a non-default suite with explicit environment requirements.

Fail criteria

Default test runs unconditionally require special hardware, large datasets, or similarly scarce resources.

Do not flag

Fast CPU-only tests, small repo-local fixtures, or specialized suites already isolated outside normal PR validation.

Confidence guidance

HIGH when the test directly assumes a missing special resource with no guard. MEDIUM when suite wiring is partly hidden. LOW when default-path membership is unclear.

Remediation

Add explicit markers or skip guards and move the test behind a dedicated opt-in gate.

Pass example

@pytest.mark.gpu
def test_trains_on_cuda():
    if not torch.cuda.is_available():
        pytest.skip('CUDA required')
    train_on_gpu()

Fail example

def test_trains_on_cuda():
    train_on_gpu()