Knowledge check Choose an answer to see why it is right or wrong.
JavaScript is off, so every explanation is shown at once.
01 Your PR changes a Rust function whose new behavior is fully observable in a single pure-function unit test. A reviewer asks you to also add an end-to-end test for it. What does the Testing & Quality Standard say?
Test at the lowest tier that can observe the behavior; a bug provable by a unit test must not require an e2e test. The unit test is the right tier. Correct. This is the rule that overrides the pyramid/trophy ratios - ratios are a smell detector, not a target, and forcing a higher tier for a unit-observable behavior is tier confusion. Add the e2e test, because every changed behavior needs coverage at all three tiers to be safe. Incorrect. The standard explicitly rejects this: e2e is reserved for few, critical cross-boundary flows, and a bug provable at the unit tier must not require an e2e test. Add the e2e test to raise the repo's e2e share toward the ~10% guideline. Incorrect. The ~10% e2e figure is an order-of-magnitude guide, not a quota, and the standard forbids writing tests to hit a percentage. 02 A change touches a large legacy file with low overall coverage. Which coverage check is the blocking gate on the PR?
Patch (diff) coverage - at least 90% of the lines the PR changed - which is what makes the gate fair on legacy code and strict on new code. Correct. Diff coverage is the primary gate; the repo total is a trend guarded by a no-regression ratchet, not the blocking number. Repo overall coverage must reach the 90% target before the PR can merge. Incorrect. The repo total is a trend (90% target, 85% hard floor) with a ratchet, not the blocking check - gating legacy PRs on the whole-repo number would be unfair and is not the policy. Per-file coverage of every file in the repo must be at least 70%. Incorrect. The 70% per-file floor warns generally and only blocks on a new or changed file; it is not the primary gate, and it does not apply to untouched files. 03 CI reports that one of your tests passed on one run and failed on a rerun of the same commit. What does the standard require?
Treat it as a bug: quarantine it immediately so it reports but does not block merge, and file a tracked defect with an owner and a deadline (default 7 days). Correct. A flaky test is a bug, not 'just retry'. Quarantine is a holding cell with a timer; past the deadline it blocks the owning team's merges. Add an automatic retry so the test passes on the second attempt and move on. Incorrect. Retries may detect flakiness but may never be the remedy - a green-after-retry result is recorded as flaky, not fixed. Comment the test out until someone has time to look at it. Incorrect. Quarantining is automatic or one-click and keeps the test running and reporting - 'comment it out' removes the evidence and is explicitly not the mechanism. 04 In review, a teammate insists you extract an interface (a port) for a component that has exactly one implementation and no test-double need, citing Dependency Inversion. How should this be resolved under the Code Craftsmanship Standard?
Do not add the seam yet - YAGNI wins over a speculative port; introduce the abstraction when a real second implementation or a genuine test-double need appears. Dogmatic principle application is itself an anti-pattern. Correct. The standard's arbitration table resolves OCP/DIP vs YAGNI in YAGNI's favor for a single implementation, and demanding an interface for one implementation is named as causing the damage the principles exist to prevent. Add the interface, because Dependency Inversion is a SOLID principle and SOLID is mandatory on every component. Incorrect. SOLID letters are heuristics with costs, introduced when a real second case forces them - not laws to apply preemptively to every component. Add the interface to be safe, since more abstraction is always easier to change later. Incorrect. Speculative abstraction costs complexity now for flexibility you may never use, and a premature abstraction is more expensive to unwind than the duplication it was meant to prevent.