docs: WP-18 の作業状況を反映し feedback 採否 19 件を登録する (ADR-073) - #390
Conversation
|
Important Review skippedAuto incremental reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
📝 WalkthroughWalkthroughADR、WP-18計画書、TODO管理文書を更新しました。夜間ループとレビュー枠の実測を記録し、WPの完了条件とTODOの追加先を整理しました。 Changes運用実測とADR更新
WP-18完了条件
TODO計画と索引
Estimated code review effort: 2 (Simple) | ~15 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 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 |
🤖 PR Monitor 分析 (GitHub Actions バックストップ)
Applicable Findings (Critical / High / Major)該当なし (レビュー指摘 0 件) Filtered (not applicable)該当なし 次のアクション
|
There was a problem hiding this comment.
Actionable comments posted: 7
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/adr/adr-072-nightly-todo-loop.md`:
- Around line 581-587: Revise the section describing the CodeRabbit rate-limit
outcome so it no longer claims that the missing review leaves no signal
anywhere. State that the “Review limit reached” comment and workflow success are
recorded, but the absence of an obtained review is not classified as an
independent failure or status; preserve the surrounding observation and
undecided-design context.
In `@docs/adr/adr-073-work-package-completion-boundary.md`:
- Around line 75-77:
WP-18の適用実績にある「残条件」を「残る完了条件」に変更し、採用率測定1件が完了条件として残っていることを明確にしてください。あわせて、非ブロッカーのスモーク未確定3件は完了条件に含まれない残作業である旨を追記してください。
In `@docs/harness-improvement-plan.md`:
- Around line 171-173: Update the smoke-uncertainty entry for Phase B and the
coderabbitai allowlist to distinguish issuing or observing a CodeRabbit review
request from actually receiving a review. State only that Decision 16 supplies
the request/observation opportunity, and keep successful review acquisition
unconfirmed, consistent with the rate-limited status of PR `#387`.
In `@docs/todo22.md`:
- Around line 13-17: docs/todo22.md
の採用候補数と採否数の母集団を分離して記載し、採用19件・却下5件が対象外の実装済み6件を含まないことを明確にしてください。冒頭の「候補24件」と、後続の「対象外6件」の数え方が矛盾しないよう、候補総数または説明文を実際の集計に合わせて修正してください。
- Around line 441-449: Update the 完了基準 section in 作業計画 to remove the requirement
that scanning cost be non-linear in the number of runs. State instead that
success is measured by reducing meta.json reads or JSON parse counts, reducing
measured execution time, or establishing a run-count limit through the retention
policy, with the selected outcome and supporting evidence recorded.
In `@docs/todo3.md`:
- Line 5:
同期対象一覧の件数と内容を更新し、docs/todo3.md(5-5)、docs/todo4.md(5-5)、docs/todo5.md(5-5)、docs/todo6.md(5-5)の各説明にdocs/todo22.mdを追加してください。各一覧が24ファイルを正確に列挙し、現在のTODO追加先を新セッションの確認対象に含める状態にしてください。
In `@docs/todo7.md`:
- Line 5:
24件の確認対象とファイル一覧を同期する。docs/todo7.md、docs/todo9.md、docs/todo10.md、docs/todo11.md
の該当記述で「todo2-21.md」を「todo2-22.md」に変更し、docs/todo8.md の24件一覧に「todo22.md」を追加する。
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: b21fedd5-a382-49b0-8992-55ed16466a79
📒 Files selected for processing (18)
CLAUDE.mddocs/adr/adr-019-coderabbit-review-hybrid-policy.mddocs/adr/adr-072-nightly-todo-loop.mddocs/adr/adr-073-work-package-completion-boundary.mddocs/harness-improvement-plan.mddocs/todo-summary.mddocs/todo-summary2.mddocs/todo10.mddocs/todo11.mddocs/todo21.mddocs/todo22.mddocs/todo3.mddocs/todo4.mddocs/todo5.mddocs/todo6.mddocs/todo7.mddocs/todo8.mddocs/todo9.md
🤖 PR Monitor 分析 (GitHub Actions バックストップ)
Applicable Findings (Critical / High / Major)
Applicable Findings (Medium 以下)
Filtered (not applicable)
次のアクション
|
8ae9b79 to
26cbfca
Compare
🤖 PR Monitor 分析 (GitHub Actions バックストップ)
重要な新情報: 「修正しました」返信は実コミットを伴っていない4 件の返信スレッド(
Applicable Findings (Critical / High / Major)
Applicable Findings (Medium 以下)
Filtered (not applicable)
次のアクション
|
本セッションで実施した WP-18 (2) 運用問題 5 件の対処 (#385/#386/#388/#389) について、 実走観測の記録・計画書の整理・feedback 採否の登録をまとめて行う。 ## 実走観測の記録 - ADR-072 へ定常運用 2 巡目 (PR #387) の観測を追加する。決定 15-17 投入後の 9 項目が設計どおり動いたことと、review-request の成功判定が「反応の有無」で 止まっている (拒否も success になる) ことを事実として記録する - ADR-019 へレート制限の競合が記録の翌日に実地で再現したことを追加する - 順位 386 の観測を 7 回 → 9 回へ更新する。うち 1 件は空コミットではなく 「近い revset を優先する規則」そのものが原因で、本命の対処案だけでは 解決しない可能性がある点を併記する ## 計画書の整理 (ADR-073 新設) - 完了条件の切り方 (残作業を 3 区分に分け、その WP が生んだ問題は完了条件に 含め、WP 外の派生は含めない) を ADR-073 として切り出す - WP-18 節を 71 行 → 35 行へ整理し、完了記録を削除して残作業のみにする - ローカル実行時の jj workspace 注記を ADR-072 へ移す ## post-merge feedback 採否 (順位 414-432) - 採用候補 24 件のうち 6 件は当該 PR 内で実装済みのため対象外とした (実物と照合して確認) - 採用 19 件を系統 A-G + セッション由来へ分類して登録する - SIGPIPE resilience は却下する。レポートが「実測証拠」とした recovery が 実際には発生しておらず (run 1 回・completed・marker の痕跡なし)、提案内容も ADR-030 §L1 で実装済みだった。feedback レポート自身が根拠を誤った初の実例 として記録し、順位 403 の対象へ含めるよう申し送る ## 付随 - todo21.md が 57KB (50KB 閾値超過) のため todo22.md を新設する - todo-summary.md の「現行の追加先」が todo14.md のまま stale だったので直す - cli-docs-lint が検出した preamble の数詞ずれ (23 → 24) を 9 ファイルで更新する ADR-073 / 順位 414-432
26cbfca to
b428d78
Compare
Summary
docs/todo21.mdが 50KB 閾値を超えたため todo22.md を新設したContext
Why(実走観測): 2026-08-11 の夜間ループ実走(PR #387)で、決定 15-17 を入れた後の経路を通しで観測した。設計どおりに動いた 9 項目の対照表がどこにも残っていなかったため ADR-072 へ記録する。あわせて 順位 401 に記録した「自律 PR が人間の作業と CodeRabbit の枠を奪い合う」構造が、記録の翌日に実地で再現したことを ADR-019 へ追加する。
Why(計画書の整理): WP-18 節に完了記録が積み上がり、「この WP を閉じるのに何が要るか」が読み取れなくなっていた。ephemeral な計画書の役割は残作業の索引であり、完了記録は永続成果物側の責務である(順位 117 の 3 ステップ原則)。
Why(ADR-073 の新設): 「実装が終わっても運用は回らない」という WP-18 の経験は他の WP でも再利用できる判断規則である。計画書に置いたままだと退役時に失われるため、ADR として切り出した。
Scope decision:
feedback レポートが根拠を誤った初の実例
却下した #386 T1-2 は、「本セッション自体が
386.md.failedmarker からの再実行であり、この不具合クラスが現在進行形で運用摩擦を生んでいる実測証拠」を採用根拠にしていた。しかしこの事実が成立していない。.failedmarker からの再実行だったstatus: completed)PASS提案内容(marker 書き込みを Drop guard の外で pre-emptive に行う)も ADR-030 §L1 で実装済みで、提案先の
src/lib-post-merge-markercrate は存在しない。順位 403(AI レビューの主張は仮説として扱い実測で二重検証する)が扱う型だが、feedback レポート自身がその対象になった初の実例として todo22.md へ記録し、順位 403 の着手時に対象へ含めるよう申し送った。
登録した 19 件(順位 414-432)
#[cfg(unix)]テストの実行保証RemoteOnlyの変異操作検出 /--pr 0の検出Modeenum パターン.takt/runs全走査コスト🚀 Tier 1 は 4 件(414 / 417 / 419 / 420)。
Validation
pnpm lint:docs(preamble + cross-ref + priority-inversion): 0 errorverdict=APPROVErust-lint-testgroup を skip(ADR-035 / ADR-057)。quality gate は 3.2 秒lint が検出した追従漏れ
todo22.md の新設でファイル数が 23 → 24 になり、
cli-docs-lintが 9 ファイルの preamble の数詞ずれを検出した。まとめて更新済み。ファイル追加時の追従漏れが機械的に止まった実例。付随して直したもの
docs/todo-summary.mdの stale ポインタ: 「現行の追加先ファイル」がtodo14.mdのままで、実際の追加先(todo21.md)と乖離していた。放置すると次のセッションが誤ったファイルへ追加する。todo22.md を指すよう更新し、13 → 14 → 20 → 21 → 22 の移行履歴も残したReferences
Summary by CodeRabbit