Skip to content

Hypothesis Edge Cases Report

  • Added app/backend/tests/test_property_edge_cases.py with 19 new edge-case property tests.
  • Added app/backend/tests/test_regression_hypothesis_edge_cases.py with 5 deterministic reproductions for bugs found by the new properties.
  • Numerical edge cases: 7
  • Bayesian edge cases: 2
  • SRM edge cases: 3
  • Sequential testing edge cases: 2
  • Monte-Carlo edge cases: 5
  • ObservedResultsBinary and ObservedResultsContinuous accepted one-sample arms, producing degenerate result paths instead of validation errors. Fixed by requiring per-arm sample sizes >= 2.
  • Continuous sample-size calculations accepted near-zero standard deviation and could overflow on sys.float_info.max / 2. Fixed by rejecting numerically zero/non-finite std values and converting overflow/infinite sample-size estimates into ValueError.
  • SRM accepted zero-count variants when total observed count was positive. Fixed by requiring positive observed counts.
  • Sequential boundaries rejected n_looks=100. Fixed by extending supported looks to 100 and adding bounded inflation for n_looks > 10.
  • Acceptance property command: 19 passed, 292 deselected in 5.62s.
  • Full direct property module check, including existing *_properties.py: 61 passed in 7.34s.
  • python -m pytest -p no:schemathesis app/backend/tests/test_property_edge_cases.py -q -> 19 passed in 1.05s
  • python -m pytest -p no:schemathesis app/backend/tests/test_property_edge_cases.py app/backend/tests/test_regression_hypothesis_edge_cases.py app/backend/tests/test_binary_properties.py app/backend/tests/test_continuous_properties.py app/backend/tests/test_bayesian_properties.py app/backend/tests/test_srm_properties.py app/backend/tests/test_sequential_properties.py -q -> 61 passed in 7.34s
  • python -m pytest -p no:schemathesis app/backend/tests/ -k "property" -v -> 19 passed, 292 deselected in 5.62s
  • python -m pytest -p no:schemathesis app/backend/tests -q -> 311 passed in 139.49s
  • scripts\verify_all.cmd --with-e2e -> exit code 0
  • No posterior/prior update API exists in the current Bayesian stats module; Bayesian edge coverage was added around credible-interval precision guarantees exposed by the existing implementation.