The rule I set for linting when an LLM is writing code is: Either you adhere to the rules, or you mark an exception with a valid reason.
replyhttps://github.com/kstenerud/yoloai/blob/main/docs/contribut...
Poor defaults break systems by a thousand cuts. They seem to make sense when designing the language (more convenient, less typing, etc), but then they very quickly become liabilities as project complexity increases. Go made the mistakes of mutable-by-default and silent-error-dropping, but their cyclical-import-forbidding was a good call.
It isn't entirely clear to me how that relates to what I said. errcheck prevents you from dropping errors or catching them but then overwriting them before doing anything else. There's a flag you can twiddle to throw a lint error on using underscore to ignore an error, too, if you're really perturbed about that. I have a personal rule to always have a comment explaining why it's OK to do that that predates AI coding rules. This seems to meet your criteria.
reply