fix(reviewer): fail closed on empty CodeGraph semantics - #546
Draft
seonghobae wants to merge 170 commits into
Draft
fix(reviewer): fail closed on empty CodeGraph semantics#546seonghobae wants to merge 170 commits into
seonghobae wants to merge 170 commits into
Conversation
📝 WalkthroughWalkthroughCodeGraph 탐색 출력에 wrapper 소유 provenance marker를 추가한다. 엄격한 게이트는 의미론적 탐색 컨텍스트만 검토 증거로 인정한다. Smoke test와 문서는 빈 결과, 모호한 provenance, 종료 코드 변경을 반영한다. ChangesCodeGraph 의미론적 증거
Estimated code review effort: 3 (Moderate) | ~25 minutes Merge Risk: 🟡 Moderate · up to Malformed CodeGraph output can be treated as trusted semantic evidence, potentially allowing strict reviewer runs to proceed without wrapper-owned provenance. Require standalone-marker parsing before merge. Sequence Diagram(s)sequenceDiagram
participant CLI
participant fetch_manifest
participant _semantic_codegraph_runner
participant missing_evidence
CLI->>fetch_manifest: codegraph_runner 전달
fetch_manifest->>_semantic_codegraph_runner: explore 실행
_semantic_codegraph_runner->>_semantic_codegraph_runner: raw marker 중화 및 wrapper marker 추가
_semantic_codegraph_runner-->>fetch_manifest: 라벨링된 CodeGraph 상태 반환
fetch_manifest-->>missing_evidence: ReviewManifest 전달
missing_evidence->>missing_evidence: marker와 semantic context 검증
missing_evidence-->>CLI: 누락 증거 사유 반환
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This was referenced Sep 3, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Defect
Noema strict reviewer must preserve exact current-head changed-file identity as untrusted data across the first CodeGraph explore and bounded empty-result recovery path. This lane owns reviewer semantic evidence only; provider/model routing, quarantine/security verdicts, outbound authority and central runner capacity remain with their canonical owners.
Retained RED → causal repair
a785cd4e536b84cd4bf7e4d6a0e1b9aff71d057frestores exact canonical-JSON changed-scope parsing and complete bounded symbol recovery.244a0294ab5d8fc1df0352c4b02b258a11938a39proves newline-bearing filenames and instruction-shaped symbol output remain JSON data in retry seeds; productionb6c370255e92491d4eac0cc9ec9c5c6b6f909fabserializes[{"path":...,"symbols":...}]as untrusted retrieval data.3328f7ba97bc4575665c5e539369951cb21ccd60proves a leading backslash is a legal Linux Git filename byte; production04376e279c61a7491cb311589b28dc643bf65837removes only the Windows-specific rejection while retaining POSIX absolute/traversal/NUL/bounded admission and no-follow sandbox boundaries.Hosted reviewer RED → causal repairs — 2026-09-06 KST
Exact
04376e279c61a7491cb311589b28dc643bf65837obtained a real GitHub-hosted runner inreviewer-ci 33979574209, job101342258628: 549 tests executed, 548 passed, one stale same-path/same-severity dedup assertion failed, and coverage was 99.29%.b84f0e5a99f09e54d30b2e42fb03dcb5566cc717converts that fixture to exact deterministic finding identity,6e5df50cd2c0f45dc454e849d781477de0a31b45adds fail-closed admission edge coverage, and95144d5bcf8f1cb4b9a7c552ede66737c23d6bcaremoves one logically unreachable empty-candidate branch. No threshold/exclusion/gate weakening was used.Exact
95144d5bcf8f1cb4b9a7c552ede66737c23d6bcathen obtained another real hosted runner inreviewer-ci 33986014642, job101359520394. Exact checkout, hash-pinned install, and the 100% line+branch pytest gate all succeeded. The sole failure moved todocstring coverage (100% gate): the newly introduced nestedsemantic_runnerinsidebuild_semantic_codegraph_runnerhad no docstring, so later CodeGraph/image steps correctly did not run.Minimal causal repair
7d3de5a859be96b953927201d9ba782673f4bb8eadds only the missing behavioral docstring to that nested runner. Runtime semantics, CodeGraph admission, provider routing, sandbox authority, coverage thresholds and workflow gates are unchanged.Current exact authority
main@e1ac9d50f6c646f04be8c137c8acdc7200182fcd;7d3de5a859be96b953927201d9ba782673f4bb8e;reviewer-ci 33991772403,ci 33991772454, requiredSecurity Scan 33991772437, andpatch-validator-image 33991772389are queued/non-passing.The hosted failures above remain RED evidence only; predecessor successes do not transfer. Keep Draft until this unchanged exact head has terminal current CI/reviewer/Security/coverage/package/image/SBOM/vulnerability/provenance evidence and every valid current finding is resolved. Runner/control-plane allocation remains owned by
ContextualWisdomLab/.github#712; do not use leafruns-onchanges, no-op source churn, rerun storms or required-gate suppression as substitutes.Merge discipline
After #546 reaches protected semantic-review truth, affected open heads need regenerated semantic evidence. Do not self-approve, force-push/destructively rebase, or accept raw/truncated/partial/cached-wrong-prompt evidence.