For false positives (tests fail on working program), the testers test themselves. Thus happens naturally - a failed test might mean app failure or a test failur

For false positives (tests fail on working program), the testers test themselves. Thus happens naturally - a failed test might mean app failure or a test failure. For false negatives (tests pass on broken program), no simple automated detection is possible, after all user cam always write "return true" in check function. Code coverage can help, but it is hard to setup and it's not panacea either. The most practical solution is false negatives are deteced by next level of tests: integration tests or manual tests or bug reports from the field. A few places I have worked on had a rule "every bugfix should be accompanied by the test for it" and it was working pretty well if code reviewrs ate attentive.