Hey folks, Recently I put together per-suite and consolidated code coverage reports in advance of our upcoming releases. I’ve uploaded them to a static site on GitHub Pages so you can view them without downloading anything: https://aber.io/cassandra-coverage-reports
Disclaimers: This is just for informational purposes. I’m not advocating for any changes to the contribution process. I wanted to share this data as we go into qualification for our next major release, and hopefully start a discussion on how we can we can continue to improve our testing. I recognize that coverage is not a sole indicator of code quality or defect rate. Potential areas for discussion: - How can we improve the coverage of our fuzz suite? - Are the 5.0 upgrade risks adequately covered? In particular, things like Config compatibility, Schema compatibility, mixed-version operation, etc. These coverage reports are based on a fork of trunk with merge base 16b43e4d4bd4b49029c0fc360bae1e732a7d5aae. The branch used to produce these reports is available here: https://github.com/apache/cassandra/compare/trunk...aratno:cassandra:coverage-reports I’m looking to merge these changes into trunk since they include fixes for a few issues that prevented correct coverage collection in the past, particularly for jvm-dtest-upgrade and jvm-dtest-fuzz suites. Feedback welcome. -- Abe
