Published result

Why two review verdicts were refused

A plain-language summary comes first, followed by the full text of the result's reproduction pack.

Result summary and pack text · reproduce/lab-review-invalid-quote-and-unread/README.md

In plain terms

One recorded verdict failed a quotation check after extra escaping; another stopped with evidence unread. The candidate fix remains unqualified.

Verdict key: Yes, The result supports the change or claim that was tested. No, The result does not support it, so nothing changes. Unresolved, The evidence is not strong enough either way. Superseded, A later test replaced this result.

Full text of the reproduction pack

The text below is the pack's README as published in the repository (a system folder name may show without its leading slash, so that the site's privacy scan passes).

Two observed invalid reviews at Kogen 2e6a8976bab3838b7db1ac4dbd9e0c51d99f1ed0 (review pin cdcbbe3e) had different causes: one of two involved extra escaped quotation marks, and one of two was correctly held on unread evidence. These are two selected failures, not an estimate of the population rate.

Onboarding Build 53563458 submitted two structured changes verdicts with blockers. The first paraphrased an unmet requirement. The retry quoted it but included literal backslashes before inner quotation marks. The requirement gate requires a blocking message to contain the exact requirement after whitespace and case normalization. The saved retry passed through the actual Go requirements validator in a model-free replay: baseline reported one problem, scratch quote normalization reported zero. That is a replay of this gate, not of all verdict checks or a claim the candidate was correct. A scratch change that normalizes backslash-escaped double quotes for the quotation comparison is plausible; it does not repair the first paraphrase or approve the Build. The existing targeted IndependentReview and ExternalReview Go tests passed on the scratch change. No negative tests for other escape patterns were run; the change is not qualified for delivery.

Documentation Build c60fc93e attempted changes without a blocker with 212 pinned parts unread, then approval with 182 unread. The unread gate correctly refused incomplete review. This calls for reviewer completion, not a weaker gate. Build a1e351e3 had a valid blocker concerning unverified run-log preservation; the blocker's substantive correctness is unverified. Build bb178205 had a valid blocker describing a Darwin regression probe where a detached child survived termination; the probe has not been independently repeated. Valid changes verdicts map to review_blocked at independent.go. No model cells were run.

Earlier work RC-B addresses a separate binding rule for requirements from refused submissions. The quote normalization here does not claim to fix that rule.

Rerun

Run python3 report.py to compare the recorded requirement and finding text with the matching rule and the proposed quote normalization. It prints false then true. This portable report checks only text matching; the Go gate replay reported above used the full saved structured output and the pinned review source, with requirements invoked on the original and scratch-modified trees. The report data contains only the relevant recorded requirement and finding, not the complete private Build log. The public source link above identifies the validator; no product source or patch is included in this pack.