Repository navigation
Conversation
ApprovabilityVerdict: Approved at Macroscope's review found this PR approvable — PR search now preserves inherited property names as literal text rather than interpreting them as review or checks filters. Recognized qualifiers and product defaults remain unchanged. You can add or adjust custom eligibility rules. Learn more. |
|
No actionable comments were generated in the recent review.
|
| Layer / File(s) | Summary |
|---|---|
Own-property checks and regression tests packages/client-runtime/src/pullRequestQuery.ts, packages/client-runtime/src/pullRequestQuery.test.ts |
Review and checks qualifier parsing accepts only own lookup-table properties. Tests verify that prototype-like values remain query text and produce no filters. |
Estimated code review effort: 2 (Simple) | ~10 minutes
Change: Bug fix
Suggested labels: bugfix, size:XS
Suggested reviewers: juliusmarminge
Merge Risk: 
Minimal · up to fe5f7
-
No merge-blocking risk was identified; the parser change is ready to merge.
Pre-merge checks: 
4
Finishing Touches
-
Generate unit tests (beta)
-
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Problem
Entering
review:constructororchecks:__proto__in PR search removes the literal token and produces a filter containing an inherited object property. That filter violates the existing request contract. Web and mobile use the same parser.Change
Require each review/checks value to be an own entry in its lookup table. Unknown values retain their original text, while valid spellings, aliases, negation and precedence keep their existing behavior. This is a small correction to the existing unknown-value fallback.
Scope and approval
This is a small, directly reproducible correction to the existing unknown-qualifier fallback, submitted under the focused obvious-bug exception in CONTRIBUTING.md. It adds no filter, setting or product default.
Verification
Found by source inspection and reproduced on base
9b1d351ba0f6, using macOS arm64 and Node 26.12.1. Eight new parser cases failed before the fix while all 17 existing cases passed. All 25 pass afterward.vp run check:changed --concurrency 1passes on the two changed files: formatting, lint, both knip steps, typechecks of seven affected packages, and the focused test file. Independent QA also exercised 595 behavior cases and 12 schema rejection controls. Full repository suites and build were not run, following the repository's scoped-check policy. Real browser, native-client and live RPC execution were not exercised.Fresh validation on
fe5f7f159ff4037fdea1e4fc69a56788cc46f132against main2359643cffa6ba9ee84d6f2694b935de54a8013e: source review found no issues; parser, contract and mobile PR-list tests and base/target diagnostics passed. Fresh formatting, lint, both Knip scans and typechecks for seven affected packages passed with no tracked changes. The no-mistakes run completed with an explicit exception for unavailable live-client validation; this is not a clean live Test pass.Models and harnesses: Claude Sonnet 5.5 through T3 Code's Claude Agent harness; GPT-6.1-sol through T3 Code's Codex harness.