Refine or-true diagnostics with corpus categories#51
Conversation
|
Fucking spectacular |
x1f9-ricon
left a comment
There was a problem hiding this comment.
Reviewed and validated. The classification buckets line up with the corpus findings, the stricter inline-ignore requirement is documented in help text and fixtures, and the tests cover arithmetic increments, grep count/probe cases, best-effort cleanup/probes, high-risk git/key suppression, and generic ignores without reasons.
Validation I ran:
bash -n .mise/tasks/lint/or-trueshellcheck --exclude=SC1091,SC2034,SC2154 .mise/tasks/lint/or-truemise run test test/lint/or-true/or-true.bats(30/30)- full
mise run test(218/218) git diff --check origin/main...HEAD
Hosted Ubuntu/macOS checks are green. Nice work.
|
Merged via merge commit Learning note / pattern worth carrying forward: this was a good example of corpus-calibrated lint work. The PR did not just make Local validation before merge: targeted |
Summary
lint:or-truefindings as arithmetic increment, count/probe, best-effort/probe, high-risk, generic ignore missing reason, or unclassified.# codebase:ignore or-true — <reason>).Closes #50.
Validation
bash -n .mise/tasks/lint/or-trueshellcheck --exclude=SC1091,SC2034,SC2154 .mise/tasks/lint/or-truemise run test test/lint/or-true/or-true.batsmise run test(218/218)Notes
codebase lint "$PWD"is not currently a normal self-validation path for this repo becausemise.tomlhas no[_.codebase].lintrules configured. Directmise run lint:or-true "$PWD"also intentionally reports fixture hits undertest/lint/or-true/fixturesand migration fixtures, so I treated the BATS suite as the self-hosted validation path.