Closed matt-bernhardt closed 8 years ago
An intrinsic part of this change, if adopted, is a commitment to not forget about the "ignored" problems - that there would need to be work done to resolve (or pardon) the problems on this list. Some things may never be resolvable - one example might be the switch_to_blog checks, which may be necessary given our goals. But we should make sure we know what those problems are, and not just forget them / turn a blind eye.
This branch implements an approach that @JPrevost recommended to me some time ago, but I've resisted until now. Specifically, when analyzing projecs with large amounts of technical debt, the approach is to whitelist the checks that we know will fail, in order to focus on finding new problems.
To walk through this approach, check out the result of one of our earliest analyses of this theme:
Compare that with a more recent run:
Through our efforts, we've caused a large number of sniffers to pass - but the suite as a whole still fails. This is somewhat risky, because new PRs can introduce new problems that might get lost in the noise of failures we already know about. By explicitly whitelisting the tests above, we can cause a passing test - which then causes new problems to stand out.