Methodology
About the results
Each report is a library's test run against a revision of the Python Array API test suite. The dashboard summarizes harvested reports; it does not run tests itself.
What counts as comparable?
Two adjacent runs are compared only when they come from the same registered source, the same indexed series, and the same test-suite revision. An indexed series fixes the Array API revision, Python major/minor, operating-system family, machine architecture, and stable execution-target details: kind, backend, and device model. Exact Python and platform details plus the complete execution target—including runtime and driver versions—identify a variant within that series. When adjacent runs use different variants, the dashboard lists the environment changes.
What does missing coverage mean?
A missing library, platform, or execution target means no matching report has been harvested. It is not counted as a failed test. Test totals may also change when the suite revision changes.