CodeRabbit's pre-merge checks include a Docstring Coverage check that warns below 80 percent, scoped to the functions a diff touches. Nothing in this repository enforces docstring coverage, so the check raises an item no gate will ever confirm or deny.
Measured on the current tree:
- No CI step, composite action, or hook runs a docstring-coverage tool.
grep over .github/workflows/, .github/actions/, .husky/ and pyproject.toml finds no interrogate, no pydocstyle, and no ruff D rules.
pyproject.toml's ruff configuration sets extend-select = ["I"], isort ordering on top of the defaults. The D family is not selected..coderabbit.yaml configures no docstring rule, so the 80 percent threshold is CodeRabbit's own default rather than a fleet decision.- The threshold does not describe this repository's code.
scripts/tests/test_spec_validate.py's methods, measured before any recent change, sit at 49.6 percent, 63 with a docstring against 64 without.
It fires inconsistently rather than steadily, which is what makes it easy to merge past without deciding: warning on #1556 and #1547, passed on #1550 and #1541.
Why this belongs in the config
.coderabbit.yaml already settles this exact shape for four tools, disabling markdownlint, actionlint, shellcheck and ruff with the reason stated in the file: "CI runs these four linters and fails the pull request on them, so a review comment from the same tool is a second copy of a red check." A pre-merge check that no CI step runs is the mirror of that case. Rather than duplicating a red check it raises one that nothing can ever turn red, so a reader cannot tell a real miss from the default threshold.
What closing this takes
Decide, and record the decision in the config beside its siblings with a comment saying why:
- Disable the docstring-coverage pre-merge check, consistent with how the four tool entries are already handled.
- Or set a threshold that describes the fleet's actual convention, in which case the measured baseline above is the starting point rather than 80 percent.
Worth deciding in the same pass whether the fleet wants a docstring convention for test methods at all. CODESTYLE.md governs comments and docstrings, and if a rule is wanted then it belongs in CI as ruff D rules where a pull request fails on it, rather than in a reviewer's advisory check that every author reargues from scratch.
Raised while driving #1556, where the finding was declined with this evidence rather than acted on, since a warning nothing enforces is a class question rather than that pull request's defect.
CodeRabbit's pre-merge checks include a Docstring Coverage check that warns below 80 percent, scoped to the functions a diff touches. Nothing in this repository enforces docstring coverage, so the check raises an item no gate will ever confirm or deny.
Measured on the current tree:
grepover.github/workflows/,.github/actions/,.husky/andpyproject.tomlfinds nointerrogate, nopydocstyle, and no ruffDrules.pyproject.toml's ruff configuration setsextend-select = ["I"], isort ordering on top of the defaults. TheDfamily is not selected..coderabbit.yamlconfigures no docstring rule, so the 80 percent threshold is CodeRabbit's own default rather than a fleet decision.scripts/tests/test_spec_validate.py's methods, measured before any recent change, sit at 49.6 percent, 63 with a docstring against 64 without.It fires inconsistently rather than steadily, which is what makes it easy to merge past without deciding: warning on #1556 and #1547, passed on #1550 and #1541.
Why this belongs in the config
.coderabbit.yamlalready settles this exact shape for four tools, disabling markdownlint, actionlint, shellcheck and ruff with the reason stated in the file: "CI runs these four linters and fails the pull request on them, so a review comment from the same tool is a second copy of a red check." A pre-merge check that no CI step runs is the mirror of that case. Rather than duplicating a red check it raises one that nothing can ever turn red, so a reader cannot tell a real miss from the default threshold.What closing this takes
Decide, and record the decision in the config beside its siblings with a comment saying why:
Worth deciding in the same pass whether the fleet wants a docstring convention for test methods at all.
CODESTYLE.mdgoverns comments and docstrings, and if a rule is wanted then it belongs in CI as ruffDrules where a pull request fails on it, rather than in a reviewer's advisory check that every author reargues from scratch.Raised while driving #1556, where the finding was declined with this evidence rather than acted on, since a warning nothing enforces is a class question rather than that pull request's defect.