Repository navigation
chore(release): prepare 2.0.1 - #452
Conversation
|
보류: 소유자 지시로 병합과 |
dmae97
left a comment
There was a problem hiding this comment.
Review of d77cde8
Release readiness: changes required before merging/tagging against current main (953ac02).
P2: Release notes omit user-facing changes now included by the merge
The proposed merge includes #453 (31275d8c2801ff9fdee6097ba086985165f8ae0f, skill-matching false-positive fixes) and #454 (0f4633e8c647695a6aa1ee4539d97897a55df8f8, next-task suggestions changed to a title and three labelled lines). Neither is recorded in .github/RELEASE_NOTES_v2.0.1.md or the 2.0.1 section of packages/coding-agent/CHANGELOG.md. These commits landed after this release PR was prepared. The release must describe the source actually tagged, not only the earlier base.
Before release, add the missing Fixed/Changed entries, identify OMK_NEXT_SUGGESTIONS as "flag, default off", regenerate the README release excerpt, and review the resulting head. Leave released 2.0.0 sections unchanged. This requires release metadata changes only, not uncommitted local roadmap work.
Checks performed
- Reviewed all 32 changed paths at the exact PR head; all 20 changed JSON files contain only the intended version/reference changes, and previous changelog sections are unchanged.
git diff --check HEAD^ HEAD,check-release-consistency.mjs --release,sync-readme-releases.mjs --check, andcheck-doc-links.mjs: passed.- Four focused release-script test files: 27 passed, zero failed/skipped.
- Existing PR CI: five successful checks. Current main CI: successful.
git merge-tree --write-tree origin/main HEAD: conflict-free preview, tree99188dd8a6f52b27f40f15864924028c3fd3f80e.
No blocking executable-code issue found in the release-only diff. Full clean-clone install/build/test/pack verification has not been rerun in this review because the release-documentation gate is not met. No merge, tag, publication, or global installation change was performed. This COMMENT records the direct review requested by the owner; it does not claim an independent GitHub APPROVED review.
dmae97
left a comment
There was a problem hiding this comment.
Local review of bab42c8
Target: main at 953ac02. This is a direct host-agent review, not an independent GitHub approval.
The earlier P2 release-documentation finding is resolved: #453 and #454 are recorded in the release notes and coding-agent CHANGELOG; OMK_NEXT_SUGGESTIONS is labelled "flag, default off"; the generated README excerpt is synchronized. Main was merged without rebasing or force-pushing. Released 2.0.0 changelog sections remain unchanged. No roadmap/uncommitted local work was included.
Exact-head checks
Isolated clean clone, Node 22.23.3, npm 11.14.1, isolated HOME/config/cache, no provider credentials, umask 022 and TMPDIR=/tmp:
- npm ci --ignore-scripts --no-audit --no-fund: exit 0.
- npm run build: exit 0, 12.778 seconds.
- npm run check: exit 0, 247.867 seconds.
- env VITEST_MAX_WORKERS=4 bash ./test.sh: exit 0, 561.001 seconds. All seven workspace test suites passed (14,473 passed, 1,009 skipped); final script test suite 847 passed, zero failed. Skipped live/provider/platform cases are not acceptance evidence.
- node scripts/publish.mjs --dry-run: exit 0, all eight package contents validated through npm pack --dry-run --ignore-scripts --json. No publication performed; no provenance attestation claimed.
- Release consistency, generated README excerpt, documentation links and git diff --check: passed.
- Build and verification left no tracked/untracked source changes.
Two earlier unrestricted-worker full-suite runs failed and are retained, not relabelled successful. One hit the existing 30-second extension import test limit; another failed the concurrent AgentSession integration and journal reader minimum-progress assertions. All affected focused tests passed without changing their assertions/timeouts. The complete bounded-worker run also passed without changing source, tests or safety controls. This is evidence of load-sensitive test execution, not proof that those timing-sensitive tests were fixed. Hosted CI on this head is still required.
OCR host-review coverage
OpenCodeReview v1.13.10 (7e5ce06) delegate preview/rule, isolated configuration, no external model/provider invocation. Identical pinned --from/--to refs used for preview and rules. Preview selected 15 files and excluded 17 by default. Every complete diff, including all exclusions, was reviewed manually: total 32, reviewed 32, skipped 0. OCR's successful exit is file/rule selection evidence only.
Selected files reviewed:
- package.json
- packages/adaptorch-wpl/package.json
- packages/agent/package.json
- packages/ai/package.json
- packages/book-to-skill/package.json
- packages/book-to-skill/src/metadata.ts
- packages/coding-agent/examples/extensions/custom-provider-anthropic/package.json
- packages/coding-agent/examples/extensions/custom-provider-gitlab-duo/package.json
- packages/coding-agent/examples/extensions/gondolin/package.json
- packages/coding-agent/examples/extensions/sandbox/package.json
- packages/coding-agent/examples/extensions/with-deps/package.json
- packages/coding-agent/package.json
- packages/protocol/package.json
- packages/speak-like-you-eat/package.json
- packages/tui/package.json
Default exclusions manually reviewed:
- .github/RELEASE_NOTES_v2.0.1.md (unsupported_ext)
- README.md (unsupported_ext)
- package-lock.json (default_path)
- packages/adaptorch-wpl/CHANGELOG.md (unsupported_ext)
- packages/agent/CHANGELOG.md (unsupported_ext)
- packages/ai/CHANGELOG.md (unsupported_ext)
- packages/book-to-skill/CHANGELOG.md (unsupported_ext)
- packages/coding-agent/CHANGELOG.md (unsupported_ext)
- packages/coding-agent/README.md (unsupported_ext)
- packages/coding-agent/examples/extensions/custom-provider-anthropic/package-lock.json (default_path)
- packages/coding-agent/examples/extensions/gondolin/package-lock.json (default_path)
- packages/coding-agent/examples/extensions/sandbox/package-lock.json (default_path)
- packages/coding-agent/examples/extensions/with-deps/package-lock.json (default_path)
- packages/coding-agent/npm-shrinkwrap.json (default_path)
- packages/protocol/CHANGELOG.md (unsupported_ext)
- packages/speak-like-you-eat/CHANGELOG.md (unsupported_ext)
- packages/tui/CHANGELOG.md (unsupported_ext)
Review findings: no additional blocking defect found in the release-only diff. Version and dependency-reference changes remain lockstep 2.0.1, external dependency versions are unchanged, the compiler identity matches its package version, published changelog history is preserved, and README truncation follows the existing release-excerpt generator. No new security/performance/architecture behavior is introduced by this release metadata diff.
Remaining release gates
This COMMENT is not an independent APPROVED review. Repository policy still requires reviewer approval of the current head and successful CI before a SHA-pinned merge. After merge, successful CI on the actual main merge commit is required before tagging. Tag-triggered publication, GitHub assets, all eight npm latest versions, and isolated registry installation remain unverified. Do not manually publish or change the selected global engine.
|
소유자가 현재 작업 요청에서 head bab42c888ac4743838ba0053975c9950373b32bf를 명시적으로 승인했습니다. 이 승인과 exact-head CI 44027146376 성공을 근거로 SHA-pinned merge를 진행합니다. GitHub APPROVED 리뷰 또는 독립 리뷰가 있었다고 주장하지 않습니다. 실제 main 병합 커밋의 CI 성공 뒤에만 v2.0.1 태그를 만들고 기존 릴리스 워크플로를 사용합니다. |
|
2.0.1 릴리스 완료. 병합 789a22b02eb90a6f7a5b93be47f0bfa8da01d0b4의 main CI(44030907232) 성공 후 같은 커밋에 v2.0.1을 태그했습니다. Build Binaries 44032833852 attempt 2의 build/attest/publish-npm/release가 모두 성공했습니다. 첫 시도는 npm 발행 전 journal 동시 읽기 테스트 실패였으며, 소스/테스트/워크플로 변경 없이 실패 작업을 한 번 재실행했습니다. 실패 기록을 성공으로 바꾸지 않습니다. 공개 검증: 8개 npm 패키지 모두 명시 버전과 latest=2.0.1, gitHead=병합 SHA. 다운로드한 tarball SHA512/SHA1와 package manifest, 격리 설치 및 8개 import 검증 통과. GitHub asset 9개의 manifest/checksum과 source-bound attestation 8건 검증 통과. npm CLI와 Linux x64 릴리스 바이너리의 version/help 검사 통과. 전역 설치 전환과 라이브 모델 호출은 하지 않았으며 macOS/Windows/ARM 바이너리 실행은 검증하지 않았습니다. npm 토큰 기반 발행을 Sigstore npm provenance로 주장하지 않습니다. Release: https://github.com/dmae97/omk/releases/tag/v2.0.1 |
Why
v2.0.0 이후 main에 병합된 사용자 영향 수정(#451, #442)을 8개 workspace 패키지 2.0.1로 발행한다. 소유자가 남은 작업 진행을 지시했다.
What changed
CHANGELOG.md에[2.0.1] - 2026-11-15절. coding-agent에는 #451의 Fixed 항목 두 개를 넣었고 나머지는 빈 절이다..github/RELEASE_NOTES_v2.0.1.md신규. 이번 주기 외부 기여자 @anhnt256(test(core): add unit tests for code-point predicates (fixes #431) #437, test(core): add unit tests for tryAcquireLaneDispatch (fixes #430) #438, test(core): add unit tests for legacy resource helpers (fixes #429) #439, test(core): add unit tests for HTTP idle timeout helpers (fixes #436) #441)를 Contributors 절에 적었다.package-lock.json,npm-shrinkwrap.json이 2.0.1로 올라간다. lockfile diff는 버전 필드와 workspace 패키지 참조뿐이고 의존성 버전은 바뀌지 않는다. 소스 변경은packages/book-to-skill/src/metadata.ts의PACKAGE_VERSION한 줄이다.sync-readme-releases로 갱신했다. #445가 넣은 2.0 상태 절의.github/RELEASE_NOTES_v2.0.0.md링크를 GitHub v2.0.0 Release 페이지 링크로 바꿨다.check-release-consistency --release는 README의 첫 릴리스 노트 링크를 현재 릴리스로 읽기 때문에, 이 링크가 남아 있으면 모든 다음 릴리스가 막힌다.변경 감사 (
v2.0.0..main)node_modules누락의--allow-scripts오진 제거검증
origin/main(5f1c410)에서 새로 클론한 트리에서scripts/release.mjs의 4~5단계를 실행했다. 메인 체크아웃의 로컬main은 다른 작업의 미푸시 커밋을 가지고 있어--release워크트리 검사가 그 기준으로 실패하기 때문이다.check-release-consistency --release,shrinkwrap:coding-agent,npm run check: 통과./test.sh: umask 022,TMPDIR/SSH_*/WSL_*해제 조건에서 통과. 개발 셸 기본값(umask 0002,TMPDIR=/mnt/omkdata/tmp,SSH_TTY)에서는 verified-run 파일 모드, subagent 골든 경로, WSL 링크 테스트 33건이 실패한다. 환경 의존 테스트 문제로 이 릴리스와 무관하며 별도로 다룬다.병합 후 병합 커밋에
v2.0.1태그를 push하면 발행 워크플로가 시작된다.