> To me there's a certain appeal to not putting (too much) logic in tests. A test with a false positive because there's a bug in the test logic is a false confidence signal that I've seen enough times to resent. I shouldn't feel the need to have tests for my tests.
I agree that there is a balance to be had here. The "logic in tests" I imply is logic which should only exist to make test suites have all of the "ables" we want in production code; extensible, readable, maintainable, understandable, etc. And when I do think "tests for tests" are applicable it is not for the tests themselves, but for the rare cases of supporting logic pervasive in non-trivial test suites.
> To be fair, it doesn't happen often, though what I'm seeing more of is just blatantly bad tests written by LLMs. It's hard to blame a philosophy when dealing with humans that have fully delegated their brains over to a random number generator.
I can't agree with you more on this.
Bad code (test or otherwise) generated by LLM's does not make clueless people more productive. It just pushes crap upstream for someone else to deal with.
I agree that there is a balance to be had here. The "logic in tests" I imply is logic which should only exist to make test suites have all of the "ables" we want in production code; extensible, readable, maintainable, understandable, etc. And when I do think "tests for tests" are applicable it is not for the tests themselves, but for the rare cases of supporting logic pervasive in non-trivial test suites.
> To be fair, it doesn't happen often, though what I'm seeing more of is just blatantly bad tests written by LLMs. It's hard to blame a philosophy when dealing with humans that have fully delegated their brains over to a random number generator.
I can't agree with you more on this.
Bad code (test or otherwise) generated by LLM's does not make clueless people more productive. It just pushes crap upstream for someone else to deal with.