Summary

• count JUnit tests, failures, and errors from the generated test cases instead of individual lint findings
• update unit and CLI sample expectations for grouped findings
• document the corrected counts in the changelog

Root cause

FormatJUnitXML() groups all lint findings for a file into one <testcase> with one <failure>, but the suite attributes were calculated from the raw finding lists. JUnit consumers therefore saw many more tests and failures than the XML actually contained.

The suite counts now match the emitted structure: one error testcase when process errors exist, and one failed testcase per file containing lint findings.

Fixes #431

Validation

• python -m pytest --no-cov cpplint_unittest.py -k testJUnitXML -vv
• python -m pytest --no-cov 'cpplint_clitest.py::TestNoRepoSignature::test_samples[silly-junit]' -vv
• python -m pytest — 231 passed, 96.46% coverage
• python -m pylint cpplint.py
• python -m mypy cpplint.py cpplint_clitest.py cpplint_unittest.py
• pre-commit run --all-files
• git diff --check

AI assistance disclosure

OpenAI Codex was used to investigate the issue, draft the implementation and regression updates, and run the validation commands listed above.

Summary by CodeRabbit

• Bug Fixes

  • Corrected JUnit XML reports so error counts indicate whether errors occurred.
  • Failure counts now reflect the number of affected files rather than individual failure messages.
  • Test counts now represent distinct generated test cases, preventing inflated totals when multiple messages belong to one case.
  • Updated sample reports and expectations to reflect accurate counts.

• Documentation

  • Added a changelog entry describing the corrected JUnit failure-count reporting.