You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(review): preserve bounded correction findings through run integration - #3214
Preserve correction findings and integration regressions. No matching client issue is claimed.
Integrates the already-reviewed run helper/guidance and end-to-end regressions. It also fixes typed test fixtures without any/waivers and canonicalizes three temporary fixture roots on macOS, after baseline reproduction, without production or assertion changes.
This is one bounded slice of the existing implementation and archived experiments, not a new design or a new benchmark. 375 additions plus deletions; no tests or comments are discarded to meet the 400-line limit.
Stack and dependencies
Order: decoder #3174 → optional transport → bounded result → run-only scope → integration → benchmark/reference documentation. These drafts allow review/CI in parallel. Merge sequentially only after provider support in Gentleman-Programming/gentle-ai#6107 and #6108 is delivered and each current-head check passes. Direct native calls, consent and closure authority remain unchanged.
Full exported test runner: 10,095 PASS, 0 FAIL, 94 skipped; provider-contract and runtime-harness stages PASS. Unit file concurrency was 2; no runner source, assertions or timeouts changed. This does not claim the earlier default-concurrency attempts passed.
Independent packaging probes: four worker-executed cases; four additional worker-authored public probes executed by the parent after mandatory isolated preparation. Public checks cover concurrent scope isolation, malformed rejection/no authority writes, narrow old-provider fallback/cache, direct/capture parity and correction payload behavior. The source was subsequently integrated with current main and checked again as above.
Earlier default-concurrency failures and the old-provider/static benchmark limitations remain recorded. This PR does not claim new live-model quality or performance measurements. Existing raw evidence and gentle-shell-bench archive commit 2486ec4 are preserved.
Corrected result/scope ancestors are merged. The integration adds explicit conditional disclosure for nested non_blocking_findings.statement_truncated, the 1,024 UTF-8-byte prefix, and no completeness/inferred-text claims. Incoming schema stays strict.
Independent disclosure delta at fbdf22fbf30342f7acf1dcc2f1e4d72e4b37f607: baseline 14/15 (only missing disclosure), candidate 15/15; differential covers three registration configurations and nine serialized executions per side. Outputs match after the allowed sentence and 61 SDK loader provenance paths per side are normalized, each validated against its exact synthetic wrapper. No other path/value/permission discarded; stderr empty. The original bootstrap, missing-scratch and two probe-expectation failures are preserved; 58 earlier artifacts hash-unchanged.
Retained independent suite adaptation and existing focused tests: 8/8 PASS, typecheck and diff check PASS at 1b71b4c8e049ad2918ef24dcd1d0cf87bcf9b0d7. Previous full-stack results above remain historical, not a fresh full-suite claim. No third manual retry of the old macOS job; corrected-head CI is a new run.
Current-head GitHub CI remains required before merge.
AI assistance
Material assistance: el Gentleman / Pi agents for implementation packaging, tests, independent probes, documentation and integration. Parent reviewed the partition, retained all coverage and measured its size. No actual RDD mode change or native review was performed.
Risk
Shared review-client and prompt contracts; independent verification retained. Provider permission, consent, correction selection and closure authority are not broadened.
Summary by CodeRabbit
New Features
Review results can relay correction and non-blocking findings, with bounded claims when available. Findings are informational and do not affect review routing.
Guidance now identifies truncated or omitted claims and provider statements, including when statement text is incomplete.
Documentation
Clarified which review outcomes may include findings and how to interpret them, including limits and provider compatibility behavior.
No actionable comments were generated in the recent review.
Walkthrough
The run-tool guidance and documentation describe correction and non-blocking finding relay, bounded claims, and truncated provider statements. The changes also delegate advisory claim construction to a shared helper and add tests for STATUS requests, finding validation, claim limits, and statement decoding.
Changes: Finding relay
Layer / File(s)
Summary
STATUS opt-in and correction results docs/review-coordinator.md, tests/review-coordinator-findings.test.ts, tests/review-coordinator.test.ts, tests/review-delivery-preflight.test.ts
Documentation and tests cover run-scoped STATUS opt-in, correction-result fields, and handling of absent or malformed findings. Integration tests distinguish opted-in facade STATUS requests from direct INSPECT and reviewer-group capture reads. Preflight test fixtures resolve temporary repository paths with realpathSync.
Claim bounds and run-tool guidance extensions/gentle-ai.ts, docs/readme-reference.md, tests/review-one-call*.test.ts, tests/review-run-statement-guidance-independent.test.ts, odd/tasks/nonblocking-delivery-slices.md
Advisory claim construction delegates to boundedFindingClaims. Run-tool guidance and tests cover finding relay, claim truncation and omission, and truncated provider statements. The README describes non-blocking findings for input-required and stopped; the task log records delivery updates.
This change mainly updates guidance, documentation, and tests for relaying review findings. No concrete merge-blocking risk was found. Current-head CI should still pass before merge.
Pre-merge checks: 4
Check name
Status
Explanation
Description Check
Check skipped - CodeRabbit’s high-level summary is enabled.
Title check
The title "fix(review): preserve bounded correction findings through run integration" names the main change: the integrated run helper and guidance now keep bounded correction findings, and the PR add…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title "fix(review): preserve bounded correction findings through run integration" names the main change: the integrated run helper and guidance now keep bounded correction findings, and the PR adds regressions for that path. The summary lists the other changes (STATUS non-blocking findings, the statement-truncation disclosure, and fixture realpath handling), which the title does not name, but it does not need to cover every detail. The title is specific and uses a conventional commit format, so it is concise and clear.
Finishing Touches
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Recent review info
Run configuration
Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Repository UI
Review profile: ASSERTIVE
Plan: Team
Run ID: 28736c37-8c1a-4e4e-a589-1a83c2acc0dc
Commits
Reviewing files that changed from the base of the PR and between 3d7e5c5 and 1b71b4c.
Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 6 remain after this review.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Preserve correction findings and integration regressions. No matching client issue is claimed.
Integrates the already-reviewed run helper/guidance and end-to-end regressions. It also fixes typed test fixtures without any/waivers and canonicalizes three temporary fixture roots on macOS, after baseline reproduction, without production or assertion changes.
This is one bounded slice of the existing implementation and archived experiments, not a new design or a new benchmark. 375 additions plus deletions; no tests or comments are discarded to meet the 400-line limit.
Stack and dependencies
Order: decoder #3174 → optional transport → bounded result → run-only scope → integration → benchmark/reference documentation. These drafts allow review/CI in parallel. Merge sequentially only after provider support in Gentleman-Programming/gentle-ai#6107 and #6108 is delivered and each current-head check passes. Direct native calls, consent and closure authority remain unchanged.
Verification
gh pr createfrom a linked worktree is always denied with "Native preflight is unavailable" on Linux, whilegit -Cpush of the same range passes #3169 fixture failed on both unchanged main and candidate because /tmp aliases /private/tmp. Canonical root fixes preserve the assertions. The initial current-main test command mistakenly requested absent tsx and ran no product tests; the repository-native runner was used thereafter.non_blocking_findings.statement_truncated, the 1,024 UTF-8-byte prefix, and no completeness/inferred-text claims. Incoming schema stays strict.fbdf22fbf30342f7acf1dcc2f1e4d72e4b37f607: baseline 14/15 (only missing disclosure), candidate 15/15; differential covers three registration configurations and nine serialized executions per side. Outputs match after the allowed sentence and 61 SDK loader provenance paths per side are normalized, each validated against its exact synthetic wrapper. No other path/value/permission discarded; stderr empty. The original bootstrap, missing-scratch and two probe-expectation failures are preserved; 58 earlier artifacts hash-unchanged.1b71b4c8e049ad2918ef24dcd1d0cf87bcf9b0d7. Previous full-stack results above remain historical, not a fresh full-suite claim. No third manual retry of the old macOS job; corrected-head CI is a new run.AI assistance
Material assistance: el Gentleman / Pi agents for implementation packaging, tests, independent probes, documentation and integration. Parent reviewed the partition, retained all coverage and measured its size. No actual RDD mode change or native review was performed.
Risk
Shared review-client and prompt contracts; independent verification retained. Provider permission, consent, correction selection and closure authority are not broadened.
Summary by CodeRabbit