That build failure looks accurate, see inline.

I'm not really sure that it's valuable to split tests like the define() test apart? I feel like when a test breaks, I often need to look at both positive and negative test cases to fix it. This seems to spread relevant information out without providing an obvious benefit -- it's not like this is breaking 1000-line ultra-tests apart, just splitting tiny tests of the same code (which already produce tailored, relevant errors on failure) apart? That said, I don't have particularly strong feelings either way.