The Fennel sources for the executable live in test-runner/.fnl, with the entry point at __testrunner__/main.fnl. test-runner/.lua hosts vendored runtime dependencies produced during builds; avoid editing it by hand. Packaged assets and the redbean bootstrap files ship under redbean-fennel/. Repository-level tests reside in test/ and must export tables of test-* functions, matching the runner’s discovery logic. Build outputs are created inside artifacts/; the folder is disposable and should not be committed.
Run make (default) or make test to build artifacts/test-runner.com and execute the suite; pass extra args with make test ARGS='--help'. Use make lint to run fennel-ls against every .fnl file, and make clean to remove artifacts/. After a build you can invoke the tool directly via ./artifacts/test-runner.com path/to/tests to scope runs.
Write Fennel with two-space indentation and hang closing parens on their own line when forms span clauses. Prefer keyword tables ({:key val}) and dotted module namespaces (__testrunner__.future). Provide a short string literal doc immediately inside functions that surface user-facing help, mirroring the pattern in main.fnl. Tests should expose functions named test-*, while helpers use dash-delimited names (test.deep-equal).
Target the bundled runner: make test or ./artifacts/test-runner.com from the repo root. Keep files in test/ named *_test.fnl; the runner loads them automatically and executes groups in parallel, so avoid shared mutable globals. When reproducing regressions add focused assertions and consider the test.testing helper for grouping. There is no enforced coverage threshold, but add regression tests for every bug.
Commit history favors short, descriptive subject lines (namespace test-runner internal modules). Use the imperative mood, stick to lowercase unless naming proper nouns, and keep subjects under 60 characters. Squash locally to keep related changes together. Pull requests should outline the motivation, list functional changes, note any new commands or flags, and include before/after output snippets when adjusting runner reporting. Link issues when available and confirm tests or lint ran.