fix: keep target issues out of duplicate warnings - #19
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 978819942d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| for (const match of source.matchAll(closingReference)) { | ||
| const prefix = source.slice(Math.max(0, (match.index || 0) - 24), match.index || 0); | ||
| if (/\b(?:not|does\s+not|doesn't|won't|will\s+not)\s*$/i.test(prefix)) { |
There was a problem hiding this comment.
When a PR says does not actually fix #258, fails to fix #258, or cannot fix #258, the closing-reference regex still matches fix #258, while this negation check only recognizes a few phrases ending immediately before fix. The issue is consequently added to targetIssueRefs and removed from similarOpenIssues, so analyzeRepositoryContext returns pass without possibly-duplicate even when the PR explicitly says the defect remains; account for these common negated forms before treating the reference as a target.
Useful? React with 👍 / 👎.
1 participant
Summary
linkedTargetIssuesevidence while keeping closed-target, related-issue, and concurrent-PR warningsdoes not fix #258Problem
A contribution body that correctly said
Fixes #258could be downgraded toneeds-repairbecause PCF treated issue #258 as a duplicate of the pull request intended to close it. The target relationship is useful context, but it is not duplicate evidence by itself.Validation
npm run ci:gates— 278/278 tests, benchmark 77/77, hostile red tests 29/29, maintainer demo PASSgit diff --checkAssurance boundary
The full VERITAS pipeline passed intake, type, dependency, math, cost, incentive, security, adversary, and trace/seal gates. Its release evidence verdict remained
INCONCLUSIVEbecause all current evidence is author-side and does not satisfy the independence threshold. This PR does not claim independent validation or campaign-count progress.AI assistance
Codex helped reproduce the false positive, implement the focused repair and regressions, and run the verification. I reviewed the exact diff, signed commit, and test output before submission.