D-SafeLogger 0.2.1__tar.gz → 0.2.2__tar.gz

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/CHANGELOG.md

@@ -5,6 +5,40 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+## [0.2.2] - 2026-05-21
+
+### Added
+- Type validation as a CI quality gate: `mypy` and `pyright` are now required dev dependencies; `publication-checks` runs `uv run mypy src` and `uv run pyright src` on every push.
+- `scripts/install_built_wheel.py` — helper for installing the freshly built wheel into the uv-managed venv before running `pyright --verifytypes` against the packaged module (rather than the editable source tree).
+- `scripts/check_type_completeness.py` — threshold gate for `pyright --verifytypes dsafelogger --ignoreexternal`; CI now requires 100% public type completeness from the built wheel.
+- `scripts/check_distribution_contents.py` — standalone script replacing the inline heredoc in `publish.yml`; adds `dsafelogger/py.typed` (wheel) and `src/dsafelogger/py.typed` (sdist) to required-path checks. Also wired into `ci.yml` `publication-checks`.
+- `tests/typing_smoke/public_api_smoke.py` — pyright-only smoke test covering `ConfigureLogger`, `GetLogger`, `contextualize`, and `mp.ConfigureLogger` / `mp.GetWorkerInitializer` from a user import perspective. The directory is named `typing_smoke` (not `typing`) to avoid shadowing the stdlib `typing` module in spawn child processes. Not collected by pytest (filename does not match `test_*.py`); checked by `uv run pyright tests/typing_smoke` in `publication-checks`.
+- `dsafelogger._handler.ReopenableHandler` — `@runtime_checkable` Protocol that narrows `logging.Handler` references to those exposing `reopen()`; replaces internal `hasattr(h, 'reopen')` checks for static type checking.
+
+### Changed
+- `dsafelogger._pipeline.ResolvedConfig.sensitive_keywords` now defaults to `BUILTIN_SENSITIVE_KEYWORDS` directly (previously `field(default_factory=frozenset)`). Behaviour is preserved by the existing `or BUILTIN_SENSITIVE_KEYWORDS` consumer fallback; the API docs now show the concrete default value rather than `<factory>`.
+- `dsafelogger._mp_protocol.LogEvent` TypedDict fields `process`, `processName`, `thread`, `threadName` are now `int | None` / `str | None`, aligning with stdlib `logging.LogRecord` semantics (None is a legal value for these attributes).
+- `dsafelogger._formatter.DSafeFormatter.__init__` `style` parameter is now typed `Literal['%', '{', '$']` instead of `str`, matching `logging.Formatter`.
+
+### Fixed
+- `dsafelogger._routing.NoneStrategy.advance()` now returns the current path instead of implicitly returning `None`, satisfying its `Path` return-type contract (no observable behaviour change; `should_switch()` always returns `False` so the method is unreachable in production).
+- `dsafelogger._cli` tail follow loop asserts that the open file handle is non-`None` before reuse (defensive narrowing; the assert is unreachable in normal execution).
+- `dsafelogger._shutdown()` acquires the root logger reference before the `try` block so the `finally` block always has it available, even if an exception is raised before the original assignment site.
+
+### Internal
+- Removed 19 stale `# type: ignore[...]` comments that mypy 2.1 no longer requires.
+- Annotated 4 previously untyped functions (`_levels._make_log_method`, `_color.ColorStreamHandler.__init__`, `_transport.TransportFactory.create`).
+- `dsafelogger._mp_queue.TrackedQueue.__getstate__` / `__setstate__` are now explicitly marked `# type: ignore[override]` with a comment documenting the intentional `_QueueState` extension; cross-platform `qsize()` and `empty()` are coerced via `int(...)` / `bool(...)` to satisfy `warn_return_any`.
+- `dsafelogger._mp_attach._validate_attach_ack` and `dsafelogger.mp._validate_bootstrap_ready_ack` now accept `ControlAck` TypedDict directly (previously `dict[str, Any]` / `dict[str, object]`).
+- `pyproject.toml` adds `[tool.pyright]` (basic mode, `include = ["src"]`).
+- `TESTING.md` gains a Type Validation section describing the local-green-before-push policy, recorded versions, and the private module surface rule.
+- `README.md` / `README_ja.md` typing bullet updated to mention the CI type-check gate.
+- `CONTRIBUTING.md` adds `uv run mypy src` / `uv run pyright src` to the required PR pre-flight commands and documents the private module surface policy.
+- Public v23j design documents now describe the 0.2.2 type-validation gate, including source typing, `tests/typing_smoke`, built-wheel verifytypes, and the stdlib `typing` shadowing guard.
+- API docs regenerated for `_color`, `_formatter`, `_handler`, `_mp_attach`, `_pipeline`.
+
 ## [0.2.1] - 2026-05-20
 
 ### Fixed

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/CONTRIBUTING.md

@@ -51,6 +51,9 @@ cd D-SafeLogger
 uv sync --group dev
 uv run pytest tests/ -v
 uv run ruff check src/ tests/
+uv run mypy src
+uv run pyright src
+uv run pyright tests/typing_smoke
 ```
 
 ## 6. Code Style
@@ -58,14 +61,17 @@ uv run ruff check src/ tests/
 - **PEP 8**, max **100** characters per line.
 - **ruff** for linting (`uv run ruff check src/ tests/`).
 - **Type hints required** — use Python 3.11+ `X | Y` union syntax.
+- **Typing must stay green**: `uv run mypy src`, `uv run pyright src`, and `uv run pyright tests/typing_smoke` must report `0 errors` locally before opening a PR. CI also enforces 100% public type completeness with `pyright --verifytypes` against the built wheel.
+- Private module surface (names starting with `_`, e.g. `dsafelogger._constants.MASK_STRING`) is **not** part of the public API even though `py.typed` exposes it to user type checkers. Do not encourage external dependence on these names; rename / reshape freely as needed.
 
 ## 7. Pull Request Process
 
 1. Create a feature branch from `main`.
 2. Add tests for new functionality.
 3. Ensure **all** tests pass: `uv run pytest tests/ -v`.
-4. Update documentation if needed.
-5. Submit a PR with a clear description referencing the accepted issue.
+4. Verify type checks pass: `uv run mypy src`, `uv run pyright src`, and `uv run pyright tests/typing_smoke` (all must report `0 errors`).
+5. Update documentation if needed.
+6. Submit a PR with a clear description referencing the accepted issue.
 
 ## License
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: D-SafeLogger
-Version: 0.2.1
+Version: 0.2.2
 Summary: Zero-dependency, thread-safe, append-only logging library for Python with 3-layer config pipeline
 Author: D
 License-Expression: Apache-2.0
@@ -182,11 +182,12 @@ Notes:
 - **Centralized setup:** replace common `basicConfig()`, `dictConfig()`, formatter, handler, and rotating-file boilerplate with `ConfigureLogger()`.
 - **Fail-fast initialization:** invalid configuration and unwritable log destinations fail during setup instead of degrading silently.
 - **Append-only file routing:** the routing layer opens the next destination instead of renaming or truncating the active log file. This avoids the common Windows failure mode where active log files cannot be renamed, and it avoids the POSIX case where a writer may continue writing to the previous file after a successful rename.
+- **Retention for routed files:** routed files can be kept by `backup_count`; older files can be deleted by the purge worker or ZIP-archived with `archive_mode=True`.
 - **Classified delivery state:** loss, reject, and drop events are not treated as invisible file gaps. When records cannot be delivered, the runtime classifies the outcome as known-rejected, known-dropped, or unexplained-lost where applicable.
 - **Bounded logging path:** D-SafeLogger uses bounded queues, explicit timeouts, and explicit rejection paths to avoid unbounded logging-side waits in the host process.
 - **Structured JSON Lines:** emit log records as JSON fields for log collectors and observability pipelines.
 - **Contextual logging:** attach request IDs, user IDs, job IDs, or other context with thread-safe and async-safe propagation. Producer-side context snapshots are taken at hand-off so listeners and Writers do not look up live `contextvars`.
-- **Integrity sidecars:** generate SHA-256 sidecars and optional manifest entries for routed log files.
+- **Integrity sidecars:** generate SHA-256 sidecars and optional manifest entries for routed log files. This is tamper-evidence for closed files, not an access-control or compliance system.
 - **Operational overrides:** change log level, module routing, console output, color, hashing, config file path, and queue/timeout parameters through environment variables, typically to raise diagnostics in production without code changes.
 - **Environment-only diagnostic mode:** opt in via `D_LOG_DIAGNOSE=1` for `f_locals` expansion of selected frames; deliberately not exposed through INI or arguments, so it cannot be enabled by an unowned configuration file.
 - **Async transport:** opt in to queue-backed logging when application threads should avoid direct sink writes.
@@ -200,6 +201,8 @@ Notes:
 
 In this mode, a parent-side Writer owns the file sinks. Workers attach to the Writer and submit log records through IPC. This centralizes file ownership and exposes delivery-state counters such as accepted, delivered, rejected, dropped, and unexplained-lost.
 
+The Writer shutdown path is bounded: it attempts to drain and join within a timeout, emits a visible warning if drain is incomplete, and avoids hanging the host process indefinitely.
+
 For setup code, the `multiprocessing` context rules, pool initializer, `ProcessPoolExecutor` integration, Windows spawn caveats, custom log levels, attach/detach lifecycle, environment-variable knobs, and shutdown handling, see [`examples/12_multiprocess_logging.md`](examples/12_multiprocess_logging.md).
 
 Public API in `dsafelogger.mp`: `ConfigureLogger`, `AttachCurrentProcess`, `DetachCurrentProcess`, `GetLogger`, `GetWorkerInitializer`, `ReopenLogFiles`.
@@ -256,11 +259,13 @@ Suggested reading paths:
 
 D-SafeLogger is competitive in the selected single-process async benchmark runs. In multiprocess benchmarks, raw throughput is not the differentiator; parent-side file output and classified delivery-state accounting are.
 
+The benchmark suite also includes multiprocess resilience profiles, such as sink-unavailable, burst backpressure, worker crash, mixed worker behavior, and shutdown behavior. These runs are not throughput claims; they check whether attempted records can be accounted for as delivered, known-rejected, known-dropped, or unexplained-lost.
+
 See [BENCHMARK.md](BENCHMARK.md) for the selected runs, methodology, and the explicit "what to claim / what not to claim" boundaries, and [`benchmarks/summary/`](benchmarks/summary/) for the published summaries.
 
 ## Testing / Quality
 
-The release gate runs the full dev test suite across Windows, macOS, and Linux on Python 3.11-3.14. Publication checks also verify generated API docs, public design documents, benchmark summaries, and package build output.
+The release gate runs the full dev test suite across Windows, macOS, and Linux on Python 3.11-3.14. CI also runs Ubuntu free-threaded CPython `3.13t` and `3.14t` compatibility jobs with `PYTHON_GIL=0`. Publication checks also verify generated API docs, public design documents, benchmark summaries, and package build output.
 
 See [TESTING.md](TESTING.md) for details.
 
@@ -269,7 +274,7 @@ See [TESTING.md](TESTING.md) for details.
 - Python: 3.11 or newer.
 - OS: Windows, macOS, and Linux.
 - Runtime dependencies: none.
-- Typing: includes `py.typed`.
+- Typing: includes `py.typed`; CI checks `mypy`, `pyright`, typing smoke tests, and 100% public type completeness with `pyright --verifytypes`. See [TESTING.md](TESTING.md).
 - API docs: [`docs/api/`](docs/api/).
 - Design docs: [`docs/design/`](docs/design/).
 - Distribution name is `d-safelogger` (with hyphen); import name is `dsafelogger` (no separator).

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/README.md

@@ -155,11 +155,12 @@ Notes:
 - **Centralized setup:** replace common `basicConfig()`, `dictConfig()`, formatter, handler, and rotating-file boilerplate with `ConfigureLogger()`.
 - **Fail-fast initialization:** invalid configuration and unwritable log destinations fail during setup instead of degrading silently.
 - **Append-only file routing:** the routing layer opens the next destination instead of renaming or truncating the active log file. This avoids the common Windows failure mode where active log files cannot be renamed, and it avoids the POSIX case where a writer may continue writing to the previous file after a successful rename.
+- **Retention for routed files:** routed files can be kept by `backup_count`; older files can be deleted by the purge worker or ZIP-archived with `archive_mode=True`.
 - **Classified delivery state:** loss, reject, and drop events are not treated as invisible file gaps. When records cannot be delivered, the runtime classifies the outcome as known-rejected, known-dropped, or unexplained-lost where applicable.
 - **Bounded logging path:** D-SafeLogger uses bounded queues, explicit timeouts, and explicit rejection paths to avoid unbounded logging-side waits in the host process.
 - **Structured JSON Lines:** emit log records as JSON fields for log collectors and observability pipelines.
 - **Contextual logging:** attach request IDs, user IDs, job IDs, or other context with thread-safe and async-safe propagation. Producer-side context snapshots are taken at hand-off so listeners and Writers do not look up live `contextvars`.
-- **Integrity sidecars:** generate SHA-256 sidecars and optional manifest entries for routed log files.
+- **Integrity sidecars:** generate SHA-256 sidecars and optional manifest entries for routed log files. This is tamper-evidence for closed files, not an access-control or compliance system.
 - **Operational overrides:** change log level, module routing, console output, color, hashing, config file path, and queue/timeout parameters through environment variables, typically to raise diagnostics in production without code changes.
 - **Environment-only diagnostic mode:** opt in via `D_LOG_DIAGNOSE=1` for `f_locals` expansion of selected frames; deliberately not exposed through INI or arguments, so it cannot be enabled by an unowned configuration file.
 - **Async transport:** opt in to queue-backed logging when application threads should avoid direct sink writes.
@@ -173,6 +174,8 @@ Notes:
 
 In this mode, a parent-side Writer owns the file sinks. Workers attach to the Writer and submit log records through IPC. This centralizes file ownership and exposes delivery-state counters such as accepted, delivered, rejected, dropped, and unexplained-lost.
 
+The Writer shutdown path is bounded: it attempts to drain and join within a timeout, emits a visible warning if drain is incomplete, and avoids hanging the host process indefinitely.
+
 For setup code, the `multiprocessing` context rules, pool initializer, `ProcessPoolExecutor` integration, Windows spawn caveats, custom log levels, attach/detach lifecycle, environment-variable knobs, and shutdown handling, see [`examples/12_multiprocess_logging.md`](examples/12_multiprocess_logging.md).
 
 Public API in `dsafelogger.mp`: `ConfigureLogger`, `AttachCurrentProcess`, `DetachCurrentProcess`, `GetLogger`, `GetWorkerInitializer`, `ReopenLogFiles`.
@@ -229,11 +232,13 @@ Suggested reading paths:
 
 D-SafeLogger is competitive in the selected single-process async benchmark runs. In multiprocess benchmarks, raw throughput is not the differentiator; parent-side file output and classified delivery-state accounting are.
 
+The benchmark suite also includes multiprocess resilience profiles, such as sink-unavailable, burst backpressure, worker crash, mixed worker behavior, and shutdown behavior. These runs are not throughput claims; they check whether attempted records can be accounted for as delivered, known-rejected, known-dropped, or unexplained-lost.
+
 See [BENCHMARK.md](BENCHMARK.md) for the selected runs, methodology, and the explicit "what to claim / what not to claim" boundaries, and [`benchmarks/summary/`](benchmarks/summary/) for the published summaries.
 
 ## Testing / Quality
 
-The release gate runs the full dev test suite across Windows, macOS, and Linux on Python 3.11-3.14. Publication checks also verify generated API docs, public design documents, benchmark summaries, and package build output.
+The release gate runs the full dev test suite across Windows, macOS, and Linux on Python 3.11-3.14. CI also runs Ubuntu free-threaded CPython `3.13t` and `3.14t` compatibility jobs with `PYTHON_GIL=0`. Publication checks also verify generated API docs, public design documents, benchmark summaries, and package build output.
 
 See [TESTING.md](TESTING.md) for details.
 
@@ -242,7 +247,7 @@ See [TESTING.md](TESTING.md) for details.
 - Python: 3.11 or newer.
 - OS: Windows, macOS, and Linux.
 - Runtime dependencies: none.
-- Typing: includes `py.typed`.
+- Typing: includes `py.typed`; CI checks `mypy`, `pyright`, typing smoke tests, and 100% public type completeness with `pyright --verifytypes`. See [TESTING.md](TESTING.md).
 - API docs: [`docs/api/`](docs/api/).
 - Design docs: [`docs/design/`](docs/design/).
 - Distribution name is `d-safelogger` (with hyphen); import name is `dsafelogger` (no separator).

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/README_ja.md

@@ -155,11 +155,12 @@ D-SafeLogger は、active file を後から動かす代わりに、書き込み
 - **中央初期化:** `basicConfig()`、`dictConfig()`、formatter、handler、rotation 設定の boilerplate を `ConfigureLogger()` に集約できます。
 - **fail-fast 初期化:** 不正な設定値や書き込めないログ出力先は初期化時に失敗させ、あとから気づきにくい形で動作が崩れることを避けます。
 - **追記専用のファイルルーティング:** ルーティング層は、使用中のログファイルを rename / truncate せず、境界到達時に次の出力先を開きます。Windows の active file rename 失敗だけでなく、POSIX 系で rename が成功したまま writer が旧世代へ書き続ける問題も避けられます。
+- **ルーティング済みファイルの保持管理:** ルーティング済みファイルは `backup_count` に基づいて保持できます。古いファイルは purge worker による削除、または `archive_mode=True` による ZIP archive 化が可能です。
 - **配送状態の分類:** 失われたレコードをファイル上の見えない欠損として扱わず、可能な範囲で known-rejected / known-dropped / unexplained-lost に分類します。
 - **境界をもった logging 経路:** 上限つきキュー、明示的な timeout、明示的な拒否経路を使うことで、logging 処理が本体プロセス側で無制限に待ち続ける設計を避けます。
 - **JSON Lines 出力:** ログ収集基盤や監視基盤へ渡しやすい JSON 形式で出力できます。
 - **コンテキスト付与:** request ID、user ID、job ID などを、スレッドや非同期境界をまたいでログへ付与できます。`contextvars` の snapshot は producer 側で取り、Writer 側では live な `contextvars` を参照しません。
-- **整合性 sidecar:** ルーティング済みログファイルに対して SHA-256 sidecar と、任意で manifest を出力できます。
+- **整合性 sidecar:** ルーティング済みログファイルに対して SHA-256 sidecar と、任意で manifest を出力できます。これは closed file の改ざんや破損を後から検出しやすくするためのもので、access-control system や compliance system ではありません。
 - **運用時の上書き:** ログレベル、モジュール別ルーティング、コンソール出力、色、ハッシュ、設定ファイルパス、キュー / timeout 関連を環境変数で変更できます。本番障害時にコードを変更せず診断レベルを上げる用途を主に想定しています。
 - **環境変数限定の診断モード:** `D_LOG_DIAGNOSE=1` を有効にすると、選択フレームの `f_locals` を展開できます。INI や引数からは意図的に有効化できないため、由来のわからない設定ファイルから ON にされることがありません。
 - **非同期出力:** アプリケーションスレッドが直接ファイル出力を待たないよう、キューを介した logging を選べます。
@@ -173,6 +174,8 @@ D-SafeLogger は、active file を後から動かす代わりに、書き込み
 
 このモードでは、親プロセス側の Writer がファイル出力先を所有します。worker は Writer に attach し、IPC 経由でログレコードを送ります。これによりファイル所有が一箇所に集約され、accepted / delivered / rejected / dropped / unexplained-lost といった配送状態のカウンターが公開されます。
 
+Writer の終了処理では上限つきの待機を行います。timeout 内で drain と join を試み、drain が完了しない場合は warning を出し、host process が無期限に hang し続けることを避けます。
+
 具体的なコード、`multiprocessing` の context ルール、Pool initializer、`ProcessPoolExecutor` 連携、Windows での spawn の注意、カスタムログレベル、attach / detach のライフサイクル、環境変数のチューニング、終了処理は [`examples/12_multiprocess_logging.md`](examples/12_multiprocess_logging.md) を参照してください。
 
 `dsafelogger.mp` の公開 API: `ConfigureLogger`, `AttachCurrentProcess`, `DetachCurrentProcess`, `GetLogger`, `GetWorkerInitializer`, `ReopenLogFiles`。
@@ -229,11 +232,13 @@ INI ファイル、dict 設定、モジュール別ルーティング、優先
 
 D-SafeLogger は、採用済みの単一プロセス async ベンチマークでは競争力があります。マルチプロセスでは、単純な throughput ではなく、親プロセス側 Writer によるファイル出力と配送状態の分類が主な価値です。
 
+ベンチマークには、sink-unavailable、burst backpressure、worker crash、mixed worker behavior、shutdown behavior などの multiprocess resilience profile も含まれます。これらは throughput claim ではなく、attempted record を delivered / known-rejected / known-dropped / unexplained-lost として説明できるかを確認するものです。
+
 採用 run、計測条件、主張できる範囲は [BENCHMARK.md](BENCHMARK.md) を、公開済みの summary は [`benchmarks/summary/`](benchmarks/summary/) を参照してください。
 
 ## テスト / 品質
 
-リリースゲートでは、Windows / macOS / Linux と Python 3.11-3.14 の組み合わせで dev 依存を含む full test suite を実行します。公開前チェックでは、生成済み API docs、公開設計書、ベンチマークサマリー、パッケージ build も検証します。
+リリースゲートでは、Windows / macOS / Linux と Python 3.11-3.14 の組み合わせで dev 依存を含む full test suite を実行します。CI では Ubuntu free-threaded CPython `3.13t` / `3.14t` の互換性ジョブも `PYTHON_GIL=0` で実行します。公開前チェックでは、生成済み API docs、公開設計書、ベンチマークサマリー、パッケージ build も検証します。
 
 詳しくは [TESTING.md](TESTING.md) を参照してください。
 
@@ -242,7 +247,7 @@ D-SafeLogger は、採用済みの単一プロセス async ベンチマークで
 - Python: 3.11 以上。
 - OS: Windows, macOS, Linux。
 - 実行時依存: なし。
-- 型情報: `py.typed` を同梱。
+- 型情報: `py.typed` を同梱。CI で `mypy` / `pyright`、typing smoke test、`pyright --verifytypes` による public type completeness 100% gate を検証。詳細は [TESTING.md](TESTING.md)。
 - API docs: [`docs/api/`](docs/api/)。
 - Design docs: [`docs/design/`](docs/design/)。
 - 配布パッケージ名は `d-safelogger` (ハイフン)、import 名は `dsafelogger` (区切りなし) です。

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/TESTING.md

@@ -129,6 +129,60 @@ Important expectations:
 - Tests must not rely on `multiprocessing.Queue.empty()` for correctness. Use timeout-based `get()` or fake/recording queues instead.
 - CloseMarker drain, control-plane ACKs, backpressure, reject counters, partial delivery, and bounded shutdown warning paths are covered by MP tests.
 
+## Type Validation
+
+`D-SafeLogger` ships with `py.typed` (PEP 561), so source typing and public type completeness are part of the standard quality gate. CI runs `mypy`, `pyright`, a typing smoke test, and a 100% `pyright --verifytypes` gate on every push.
+
+### Tooling
+
+- `mypy>=2.1` — strict-ish config (`disallow_untyped_defs = true`, `warn_unused_ignores = true`, `warn_return_any = true`).
+- `pyright>=1.1.409` — `typeCheckingMode = "basic"` with `include = ["src"]`.
+
+### Required commands (PR pre-flight)
+
+```bash
+uv sync --group dev
+uv run pyright --version    # record first
+uv run mypy src             # → 0 errors
+uv run pyright src          # → 0 errors
+uv run pyright tests/typing_smoke # → 0 errors
+```
+
+Local green (`0 errors` from both tools) is the precondition for pushing. CI runs the same commands and treats any error as a hard fail.
+
+Package-level type completeness is checked from the built wheel, not the editable source tree:
+
+```bash
+uv build
+uv run python scripts/install_built_wheel.py
+uv run --no-sync python scripts/check_type_completeness.py --min-score 100
+uv sync --reinstall
+```
+
+Use `uv run --no-sync` for the completeness check so the wheel install is not replaced by the editable project before `pyright --verifytypes` runs.
+
+### Latest local validation (2026-05-21 / Python 3.14.3 / Windows)
+
+| Tool | Version | Result |
+|---|---|---|
+| `mypy` | 2.1.0 | `Success: no issues found in 27 source files` |
+| `pyright` | 1.1.409 | `0 errors, 0 warnings, 0 informations` |
+| `pyright --verifytypes dsafelogger --ignoreexternal` | 1.1.409 | `Type completeness score: 100%` (10 public symbols, 23 referenced internal symbols — all known type) |
+
+### Package name vs. type checker target
+
+D-SafeLogger uses three name forms — be careful when invoking type tools:
+
+- PyPI metadata: `D-SafeLogger`
+- Python import / `verifytypes` target: `dsafelogger`
+- Wheel filename: `d_safelogger-*.whl`
+
+For `pyright --verifytypes`, always pass `dsafelogger` (the import name).
+
+### Private module surface policy
+
+Names starting with a leading underscore (`_constants.MASK_STRING`, `_constants._resolved_sensitive_keywords`, `_pipeline.ResolvedConfig`, `_async.DSafeQueueHandler`, etc.) are **private** even though `py.typed` exposes them to user type checkers. Library consumers MUST NOT depend on private symbols; they may change without notice in PATCH or MINOR releases. Only names re-exported from `dsafelogger` / `dsafelogger.mp` `__init__.py` are public.
+
 ## Free-Threaded Python
 
 GitHub Actions runs Ubuntu free-threaded CPython `3.13t` and `3.14t` jobs with `PYTHON_GIL=0`.
@@ -166,13 +220,21 @@ See [`.github/workflows/ci.yml`](.github/workflows/ci.yml).
 The CI workflow also runs publication checks on Ubuntu/Python 3.13:
 
 ```bash
+uv run pyright --version
+uv run mypy src
+uv run pyright src
+uv run pyright tests/typing_smoke
 uv run python scripts/generate_api_docs.py --check
 uv run python scripts/check_design_docs_sync.py
 uv run python benchmarks/update_summary.py --check
 uv build
+uv run python scripts/check_distribution_contents.py
+uv run python scripts/install_built_wheel.py
+uv run --no-sync python scripts/check_type_completeness.py --min-score 100
+uv sync --reinstall --group dev
 ```
 
-The publish workflow repeats these checks, verifies the git tag matches the package version, runs the full test suite, and only then publishes.
+The publish workflow repeats release document, benchmark, build, distribution-content, metadata, and full-suite checks, verifies the git tag matches the package version, and only then publishes.
 
 ## Documentation Checks
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/api/dsafelogger__color.md

@@ -12,7 +12,7 @@ Enable ANSI escape sequences on Windows 10+.
 
 ## Classes
 
-### `ColorStreamHandler(stream=None, color_enabled: 'bool' = True, color_overrides: 'dict[str, str] | None' = None) -> 'None'`
+### `ColorStreamHandler(stream: 'IO[str] | None' = None, color_enabled: 'bool' = True, color_overrides: 'dict[str, str] | None' = None) -> 'None'`
 
 StreamHandler with ANSI color codes for log levels.
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/api/dsafelogger__formatter.md

@@ -43,7 +43,7 @@ in high-throughput multi-threaded scenarios.
 
 ## Classes
 
-### `DSafeFormatter(fmt: 'str | None' = None, datefmt: 'str | None' = None, style: 'str' = '%') -> 'None'`
+### `DSafeFormatter(fmt: 'str | None' = None, datefmt: 'str | None' = None, style: "Literal['%', '{', '$']" = '%') -> 'None'`
 
 Standard D-SafeLogger formatter with level abbreviation and context suffix.
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/api/dsafelogger__handler.md

@@ -24,3 +24,15 @@ Public methods:
 - `emit(self, record: 'logging.LogRecord') -> 'None'`
 - `flush(self) -> 'None'`
 - `reopen(self) -> 'None'`
+
+### `ReopenableHandler(*args, **kwargs)`
+
+Protocol for logging handlers that support external rotation reopen.
+
+`isinstance(h, ReopenableHandler)` narrows `logging.Handler` references
+to those that implement `reopen()`, replacing `hasattr(h, 'reopen')`
+for static type checking.
+
+Public methods:
+
+- `reopen(self) -> 'None'`

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/api/dsafelogger__mp_attach.md

@@ -39,7 +39,7 @@ Restart pump threads that were lost due to a post-fork in the parent process.
 
 ### `_same_process_identity(state: 'MPProcessState') -> 'bool'`
 
-### `_validate_attach_ack(ack: 'dict[str, Any]', *, expected_protocol_version: 'int', expected_registry_hash: 'str') -> 'None'`
+### `_validate_attach_ack(ack: 'ControlAck', *, expected_protocol_version: 'int', expected_registry_hash: 'str') -> 'None'`
 
 ### `_validate_bootstrap_context(ctx: 'BootstrapContext') -> 'None'`
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/api/dsafelogger__pipeline.md

@@ -41,6 +41,12 @@ Public methods:
 
 - `build(self, config: 'ResolvedConfig') -> 'Pipeline'`
 
-### `ResolvedConfig(pg_name: 'str', log_dir: 'Path', file_fmt: 'str | FormatterConfigDict | logging.Formatter', console_fmt: 'str | FormatterConfigDict | logging.Formatter', routing_mode: 'str', routing_kwargs: 'dict', backup_count: 'int', archive_mode: 'bool', enable_hash: 'bool', manifest_path: 'Path | None', encoding: 'str', diagnose: 'bool', max_level: 'str', console: 'bool', is_async: 'bool', queue_size: 'int', log_level: 'str', color_stream: 'bool', module_configs: 'dict[str, dict]', color_overrides: 'dict[str, str]', sensitive_keywords: 'frozenset[str]' = <factory>) -> None`
+### `ResolvedConfig(pg_name: 'str', log_dir: 'Path', file_fmt: 'str | FormatterConfigDict | logging.Formatter', console_fmt: 'str | FormatterConfigDict | logging.Formatter', routing_mode: 'str', routing_kwargs: 'dict', backup_count: 'int', archive_mode: 'bool', enable_hash: 'bool', manifest_path: 'Path | None', encoding: 'str', diagnose: 'bool', max_level: 'str', console: 'bool', is_async: 'bool', queue_size: 'int', log_level: 'str', color_stream: 'bool', module_configs: 'dict[str, dict]', color_overrides: 'dict[str, str]', sensitive_keywords: 'frozenset[str]' = frozenset({'access_key', 'api_key', 'apikey', 'auth', 'cookie', 'credential', 'passwd', 'password', 'private_key', 'secret', 'session_id', 'token'})) -> None`
 
 Resolved configuration holding 3-layer merged parameters.
+
+## Constants
+
+| Name | Type | Value |
+|---|---|---|
+| `BUILTIN_SENSITIVE_KEYWORDS` | `frozenset` | `frozenset({'access_key', 'api_key', 'apikey', 'auth', 'cookie', 'credential',...` |

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D-SafeLogger_DetailedDesign_v23j.md

@@ -4290,6 +4290,7 @@ ConfigureLogger 引数  ← デフォルト（第3層）
 |---|---|---|
 | v23j | 2026-05-05 | OSS公開前 review 対応として、実装動作は v23h から変更せず、公開運用と品質ゲートを固定した。`dsafelogger.mp` は preview / experimental ではなく正式 API として公開する判断を記録し、価値訴求を raw throughput ではなく Writer-owned sinks と classified delivery-state observability に置く。OpenTelemetry / structlog coexistence tests は `dev` dependency group の full test suite に含め、`optional_integration` marker は診断用選択 marker とする。spawn E2E は Writer IPC primitives と worker creation で同一 `multiprocessing` context を使う。`tests/` から `multiprocessing.Queue.empty()` 依存を排除した。benchmark runner は `BENCHMARK.md` を生成せず、`benchmarks/results/<session>/`、`benchmarks/summary/manifest.json`、`benchmarks/summary/*.md`、手動編集 `BENCHMARK.md` の4層 publication model を採用する。release version は `0.2.0` とし、`pyproject.toml` と `dsafelogger.__version__` を一致させる。 |
 | v23j-publication-sync | 2026-05-06 | 公開前同期として、API docs 生成、design docs sync check、examples 更新、GitHub workflow gate、coverage 再生成を追加する。これらは detailed runtime design を変更しない公開成果物同期であり、`dsafelogger.mp` formal API、dev full-test policy、benchmark publication model、release version `0.2.0` の判断を維持する。 |
+| v23j-type-validation | 2026-05-21 | 0.2.2 向け公開前品質ゲートとして、`py.typed` 配布に対応する source typing / packaged typing 検証を CI に追加する。`mypy src`、`pyright src`、`pyright tests/typing_smoke`、built wheel に対する `pyright --verifytypes dsafelogger --ignoreexternal` 100% completeness gate を publication checks に含める。verifytypes は `scripts/install_built_wheel.py` で wheel を install した直後に `uv run --no-sync` で実行し、editable install への自動復元を避ける。smoke test は標準ライブラリ `typing` shadow を避けるため `tests/typing_smoke/` に配置する。runtime behavior は変更しない。 |
 | v23 | 2026-04-25 | v22i 詳細設計書を v23 として複写。v23 系設計方針（Writer 不変条件・配送契約用語・Overload Policy）は基本設計書 §12 を参照。実装との差分棚卸しを private planning notes に記録した。挙動変更なし。 |
 | v23b | 2026-04-25 | `_log_loop` の `Queue.empty()` 依存を廃止し、client close marker 機構を実装した（差分棚卸し #4 対応）。各 client は DETACH 前に `CloseMarker` を log_queue に投入する。Writer は全 ATTACH 済み client の `CloseMarker` 受信（または `close_marker_failed` 確認）を drain 完了条件とする。`is_async=True` は pump thread join 後に CloseMarker を送信し、ordering を保証する。`close_marker_failed` は silent にならず、degraded shutdown として `_close_marker_degraded` フラグと stderr に反映する。drain deadline（`WRITER_STOP_WAIT_TIMEOUT_SEC`）超過時は outstanding marker を警告しつつ終了する。変更対象: `_mp_protocol.py`（`CloseMarker`, `_is_close_marker`）、`_mp_runtime.py`（`_drain_complete`, `_record_close_marker`, `_expected_close_markers`, `_close_markers_received`, `_close_marker_failed_clients`, `_drain_deadline`）、`_mp_attach.py`（`send_close_marker`, `_do_detach` フェーズ再構築）、`_mp_control.py`（`_make_detach_request` の `close_marker_failed` 引数）。 |
 | v23c | 2026-04-26 | queue サイズをユーザー設定可能にし、drop/reject カウンターを原因別に分離した（差分棚卸し #1・#5・#7 対応）。`mp.ConfigureLogger()` に `ipc_log_queue_maxsize: int | None` と `ipc_client_queue_maxsize: int | None` を追加。env var `{prefix}_IPC_LOG_QUEUE_MAXSIZE` / `{prefix}_IPC_CLIENT_QUEUE_MAXSIZE` で上書き可能。既定値は `ipc_log_queue_maxsize=1000`（v22i 実装値を継承）、`ipc_client_queue_maxsize` は未指定時 `ipc_log_queue_maxsize` に揃える。`BootstrapContext` に `ipc_client_queue_maxsize: int` フィールドを追加。`MPClientTransport` の process-local async queue を `ctx.ipc_client_queue_maxsize` で生成するよう変更。`MPClientTransport._drop()` に原因別カウンター（`_overload_shed`, `_transport_closed_drop`, `_writer_unavailable_drop`, `_timeout_drop`）を追加。`WriterRuntime` に `_writer_route_reject`, `_writer_event_reject` を追加し `STATUS` レスポンスに反映。制御 queue（256）は内部定数のまま維持（根拠不明を棚卸し #2 に記録済み）。 |

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D-SafeLogger_TestDesign_v23j.md

@@ -9,7 +9,7 @@
 
 v23j は v23h 完了時点の OSS 公開前指摘に対するユーザー判断に基づく修正版である。v23h の sink 分類 / bounded shutdown 契約を維持した上で、公開前のテスト契約・MP E2E・benchmark 成果物管理を確定する。
 
-現行 v23j のローカル検証 baseline は Python 3.14.3 / Windows 上の結果であり、`658 passed, 3 skipped`（`661` collected）である。fork E2E は POSIX-only、Windows spawn E2E は Windows-only であるため、OS によって skipped 数は変動し得る。
+現行 v23j のローカル検証 baseline は Python 3.14.3 / Windows 上の結果であり、`658 passed, 3 skipped`（`661` collected）である。fork E2E は POSIX-only、Windows spawn E2E は Windows-only であるため、OS によって skipped 数は変動し得る。0.2.2 向けの公開前品質ゲートでは、これに加えて `mypy src` / `pyright src` / `pyright tests/typing_smoke` / built wheel に対する `pyright --verifytypes dsafelogger --ignoreexternal` 100% completeness gate を実行する。
 
 v23h から継続する動作変更は以下:
 
@@ -34,6 +34,8 @@ v23j で追加する公開前方針は以下:
 16. merge 後の設定正規化・検証を single-process / module-specific / multiprocess で共通化する
 17. routing / 世代管理 / hash の無効組み合わせは warning 補正ではなく `ValueError` とする
 18. `structured=True` と formatter 指定、未登録 level、数値範囲不正、Python API bool 型不正を fail-fast 検証する
+19. `py.typed` 配布に対応し、source typing (`mypy src`, `pyright src`) と packaged typing (`pyright --verifytypes dsafelogger --ignoreexternal`) を公開前品質ゲートへ含める
+20. 利用者視点の public API typing smoke test は `tests/typing_smoke/` に置き、`tests/typing/` という名前は標準ライブラリ `typing` shadow を避けるため使わない
 
 ---
 
@@ -71,6 +73,10 @@ v23j で追加する公開前方針は以下:
 | DOD-v23j-28 | mp の validation は single-process と同じ共通検証を使い、`structured=True + file_fmt/console_fmt` も拒否する |
 | DOD-v23j-29 | mp module-specific `level` は worker attach 側 logger level に反映する |
 | DOD-v23j-30 | Python API bool 引数に `str` など bool 以外を渡した場合は `TypeError` |
+| DOD-v23j-type-31 | `uv run mypy src` と `uv run pyright src` が 0 errors で完了する |
+| DOD-v23j-type-32 | `uv run pyright tests/typing_smoke` が 0 errors で完了し、`tests/typing_smoke/public_api_smoke.py` は pytest default collection に含まれない |
+| DOD-v23j-type-33 | built wheel install 後、`uv run --no-sync python scripts/check_type_completeness.py --min-score 100` が `pyright --verifytypes dsafelogger --ignoreexternal` の 100% completeness を確認する |
+| DOD-v23j-type-34 | `tests/typing_smoke/` の名称により spawn worker で stdlib `typing` を shadow しない |
 
 ---
 
@@ -108,6 +114,10 @@ v23j で追加する公開前方針は以下:
 | DOD-v23j-28 | `tests/test_mp_configure.py::TestMpConfigureLogger::test_structured_and_file_fmt_mutual_exclusion`、`test_config_dict_invalid_default_level_raises` |
 | DOD-v23j-29 | `tests/test_mp_configure.py::TestMpConfigureLogger::test_module_level_propagates_to_attached_logger` |
 | DOD-v23j-30 | `tests/test_coverage_boost.py::TestConfigureAdvanced::test_enable_hash_wrong_type` と追加 bool 型検査 |
+| DOD-v23j-type-31 | `uv run mypy src` / `uv run pyright src` |
+| DOD-v23j-type-32 | `uv run pyright tests/typing_smoke` / `uv run pytest tests --collect-only -q` |
+| DOD-v23j-type-33 | `scripts/install_built_wheel.py` / `scripts/check_type_completeness.py --min-score 100` |
+| DOD-v23j-type-34 | `tests/typing_smoke/public_api_smoke.py` の配置と spawn 系 MP tests (`tests/test_mp_integration.py`, `tests/test_mp_spawn_windows.py`) |
 
 ---
 
@@ -118,6 +128,18 @@ v23j で追加する公開前方針は以下:
 ```bash
 uv sync --group dev
 uv run pytest tests -v
+uv run mypy src
+uv run pyright src
+uv run pyright tests/typing_smoke
+```
+
+packaged typing の completeness は built wheel を対象に検証する。`install_built_wheel.py` 実行後の verifytypes は editable install へ戻らないよう `uv run --no-sync` を使う。
+
+```bash
+uv build
+uv run python scripts/install_built_wheel.py
+uv run --no-sync python scripts/check_type_completeness.py --min-score 100
+uv sync --reinstall
 ```
 
 補助的な切り分け用途として以下を許可するが、release 判定の代替にはしない。
@@ -140,7 +162,7 @@ v23j では benchmark 公開成果物の管理方式を変更する。runner は
 
 ### v23j Publication Sync Addendum
 
-公開前同期の追加 DOD として、coverage 再生成、API docs `--check`、design docs sync check、GitHub CI / publish preflight、examples の formal MP / external rotation 追加を検証対象に含める。これらは runtime behavior を変更せず、release `0.2.0` の公開成果物と品質ゲートを同期するための補助検証である。
+公開前同期の追加 DOD として、coverage 再生成、API docs `--check`、design docs sync check、GitHub CI / publish preflight、examples の formal MP / external rotation 追加を検証対象に含める。0.2.2 向けにはさらに source typing、typing smoke、built wheel に対する verifytypes 100% completeness gate を追加する。これらは runtime behavior を変更せず、公開成果物と品質ゲートを同期するための補助検証である。
 
 ---
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D-SafeLogger_v23j_WhitePaper.md

@@ -243,7 +243,7 @@ D-SafeLogger は、**Python 標準ライブラリ `logging` の上に構築さ
 - 対象 Python: 3.11 以上（CPython 3.13 以上の **free-threaded build** を設計対象に含む）
 - 対象 OS: Windows / macOS / Linux
 - ランタイム依存: なし（Python 標準ライブラリのみ）
-- 型情報: `py.typed` を同梱
+- 型情報: `py.typed` を同梱。CI では `mypy` / `pyright` / typing smoke test / `pyright --verifytypes` 100% public type completeness gate を検証する
 
 #### 1.1.2 ポジショニング
 
@@ -4770,7 +4770,7 @@ D-SafeLogger の 3 層パイプラインは次の点で観測可能な差別化
 
 | 項目 | D-SafeLogger v23j | stdlib `logging` | Loguru 0.7.3 | structlog 25.5.0 | picologging 0.9.3 | Eliot 1.18.0 | Logbook 1.9.2 | logfire 4.32.1 | OTel SDK 1.41.1 |
 |---|---|---|---|---|---|---|---|---|---|
-| 最新 release 日 | （v23j / 0.2.0、本レポート時点） | Python 3.14 | 2024-12-06 | 2025-10-27 | 2024-09-13（GitHub） | 2026-05-07 | — | （4.32.1） | （1.41.1） |
+| 最新 release 日 | （v23j / 0.2.1、本レポート時点） | Python 3.14 | 2024-12-06 | 2025-10-27 | 2024-09-13（GitHub） | 2026-05-07 | — | （4.32.1） | （1.41.1） |
 | ライセンス | Apache 2.0 | PSF | MIT | MIT or Apache-2.0 | MIT | Apache 2.0 | BSD-3-Clause | MIT | Apache-2.0 |
 | Python 要件 | >=3.11 | （Python 自体） | >=3.5,<4 | >=3.8 | >=3.7 | >=3.10.0 | >=3.9 | >=3.9 | >=3.9 |
 | ランタイム依存数 | **0** | 0 | 3（条件付き） | 1（条件付き） | 0 | 4 | 1 | 8 | 3 |
@@ -5342,6 +5342,7 @@ OSS 公開時には国内・海外双方からの評価軸が想定されるが
 - skipped 数は OS 依存。fork E2E は POSIX-only、Windows spawn E2E は Windows-only であるため、CI matrix では OS により skipped 数が変動し得る。
 - カバレッジ: terminal total **87%**, XML line-rate **88.97%**, branch-rate **81.46%**
 - multiprocess tests / OTel・structlog coexistence tests は公式品質ゲートに含まれる
+- 型検証: `mypy src` / `pyright src` / `pyright tests/typing_smoke` / built wheel に対する `pyright --verifytypes dsafelogger --ignoreexternal` 100% completeness gate を公開前検証に含める。smoke test ディレクトリは標準ライブラリ `typing` shadow を避けるため `tests/typing_smoke/` とする。
 - free-threaded build テスト: `PYTHON_GIL=0 uvx --python cpython-3.13+freethreaded --from pytest pytest tests -v`
 
 #### 7.10.4 リリース管理
@@ -5709,6 +5710,7 @@ structlog（2 パターンで共存）/ OpenTelemetry Python（`LoggingHandler`
 - 公式テスト baseline: 658 passed / 3 skipped（収集 661、`uv run pytest tests -v`、Python 3.14.3 / Windows）。fork E2E は POSIX-only、Windows spawn E2E は Windows-only のため、OS によって skipped 数は変動し得る。
 - カバレッジ: terminal total 87%, XML line-rate 88.97%, branch-rate 81.46%
 - multiprocess tests / OTel・structlog coexistence tests は公式品質ゲートに含まれる
+- 型検証: `mypy src` / `pyright src` / `pyright tests/typing_smoke` / built wheel に対する `pyright --verifytypes dsafelogger --ignoreexternal` 100% completeness gate を公開前検証に含める
 - free-threaded build テスト手順あり（`PYTHON_GIL=0 uvx ...`）
 - `scripts/check_design_docs_sync.py` および `scripts/generate_api_docs.py --check` で内部同期検証
 - `benchmarks/summary/manifest.json` で公開代表セッションを固定
@@ -5760,7 +5762,7 @@ structlog（2 パターンで共存）/ OpenTelemetry Python（`LoggingHandler`
 
 #### 8.8.5 ドキュメント・品質運用
 
-10. **品質ゲートと内部同期検証スクリプトが整備されている**: Python 3.14.3 / Windows で 658 passed / 3 skipped（収集 661）、カバレッジ 87%、`scripts/check_design_docs_sync.py` と `scripts/generate_api_docs.py --check` による内部同期検証、`benchmarks/summary/manifest.json` による公開代表セッション固定。skipped 数は OS 依存で変動し得る。これらは導入候補ライブラリの評価時に観測可能な品質指標として記録される。
+10. **品質ゲートと内部同期検証スクリプトが整備されている**: Python 3.14.3 / Windows で 658 passed / 3 skipped（収集 661）、カバレッジ 87%、`mypy` / `pyright` / typing smoke / `pyright --verifytypes` 100% completeness gate、`scripts/check_design_docs_sync.py` と `scripts/generate_api_docs.py --check` による内部同期検証、`benchmarks/summary/manifest.json` による公開代表セッション固定。skipped 数は OS 依存で変動し得る。これらは導入候補ライブラリの評価時に観測可能な品質指標として記録される。
 
 ---
 
@@ -5772,7 +5774,7 @@ D-SafeLogger v23j は次の総合像として整理できる:
 2. **8 軸の設計姿勢がライブラリ全体を貫いて一貫している**: 構造的に排除 / 責務分離 / 失敗分類 / 絶対防衛線 / opt-in 境界の明示 / 境界の能動的明示 / standardness 維持 / 責務委譲。
 3. **エコシステム上の位置は 4 命題に集約される**: 直接競合は観測されない / stdlib 拡張型 × 機能差別化のニッチを占める / stdlib 公式の既知制約への直接応答 / 競合ではなく共存を志向。
 4. **ベンチマーク観測事実は raw throughput ではなく観測性に主張軸が置かれる**: 単一プロセス async は競争力ある数値、multiprocess は raw throughput では stdlib に劣るが配送状態の説明可能性で固有の観測性を提供。
-5. **ドキュメント・運用構造は内部同期検証と境界の能動的明示で整備されている**: 公開設計書 + 自動生成 API + 多言語 README + manifest 固定 benchmark + 品質ゲートのテストマトリックス。
+5. **ドキュメント・運用構造は内部同期検証と境界の能動的明示で整備されている**: 公開設計書 + 自動生成 API + 多言語 README + manifest 固定 benchmark + 品質ゲートのテストマトリックス + 型検証 gate。
 
 これらは設計書 §1 の絶対条件「標準ライブラリへの完全な準拠を絶対条件としつつ、サードパーティ製ライブラリ（Loguru 等）を凌駕する診断能力と、Windows 環境での致命的なファイルロック問題を回避する堅牢性を外部依存ゼロで実現する」が、v23j 時点で**仕様・設計・実装・テスト・ドキュメント・ベンチマークの全層で一貫した形に到達している**ことを示す観測事実の集約である。
 

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D-SafeLogger_v23j_WhitePaper_en.md

@@ -224,7 +224,7 @@ D-SafeLogger is a logging platform built on the Python standard library `logging
 - Target Python: 3.11 or higher (including **free-threaded build** of CPython 3.13 or higher)
 - Target OS: Windows / macOS / Linux
 - Runtime dependencies: None (Python standard library only)
-- Type information: Includes `py.typed`
+- Type information: Includes `py.typed`; CI verifies `mypy`, `pyright`, a typing smoke test, and a 100% public type-completeness gate with `pyright --verifytypes`
 #### 1.1.2 Positioning
 Official design document §1 defines the position of this module as follows.
 > This module is a lightweight, fast, and highly functional logging platform that is commonly used by all projects in the various Python ecosystems (D-Settings, DPySide, D-MessageRouter, etc.) provided by `D`. **The premise is that it will be released as a standalone OSS, but the top priority will be to operate it as a common platform for the "D Ecosystem" rather than to widely disseminate it**.
@@ -3473,7 +3473,7 @@ D-SafeLogger's three-layer pipeline has observable differentiation in the follow
 ### 6.12 Latest status summary by library
 | Item | D-SafeLogger v23j | stdlib `logging` | Loguru 0.7.3 | structlog 25.5.0 | picologging 0.9.3 | Eliot 1.18.0 | Logbook 1.9.2 | logfire 4.32.1 | OTel SDK 1.41.1 |
 |---|---|---|---|---|---|---|---|---|---|
-| Latest release date | (v23j/0.2.0, as of this report) | Python 3.14 | 2024-12-06 | 2025-10-27 | 2024-09-13 (GitHub) | 2026-05-07 | — | (4.32.1) | (1.41.1) |
+| Latest release date | (v23j/0.2.1, as of this report) | Python 3.14 | 2024-12-06 | 2025-10-27 | 2024-09-13 (GitHub) | 2026-05-07 | — | (4.32.1) | (1.41.1) |
 | License | Apache 2.0 | PSF | MIT | MIT or Apache-2.0 | MIT | Apache 2.0 | BSD-3-Clause | MIT | Apache-2.0 |
 | Python requirements | >=3.11 | (Python itself) | >=3.5,<4 | >=3.8 | >=3.7 | >=3.10.0 | >=3.9 | >=3.9 | >=3.9 |
 | Number of runtime dependencies | **0** | 0 | 3 (conditional) | 1 (conditional) | 0 | 4 | 1 | 8 | 3 |
@@ -3872,6 +3872,7 @@ Facts that can be confirmed from `pyproject.toml` and `MANIFEST.in`:
 - The skipped count is platform-dependent because fork E2E tests are POSIX-only and Windows spawn E2E tests are Windows-only.
 - Coverage: terminal total **87%**, XML line-rate **88.97%**, branch-rate **81.46%**
 - multiprocess tests / OTel/structlog coexistence tests are included in the official quality gate
+- Type validation: public validation includes `mypy src`, `pyright src`, `pyright tests/typing_smoke`, and a 100% `pyright --verifytypes dsafelogger --ignoreexternal` completeness gate against the built wheel. The smoke-test directory is named `tests/typing_smoke/` to avoid shadowing the standard-library `typing` module.
 - free-threaded build test: `PYTHON_GIL=0 uvx --python cpython-3.13+freethreaded --from pytest pytest tests -v`
 #### 7.10.4 Release Management
 Public validation procedures:
@@ -4134,6 +4135,7 @@ Boundaries that public bench analysis actively enumerates:
 - Official test baseline: 658 passed / 3 skipped (661 collected, `uv run pytest tests -v`, Python 3.14.3 / Windows). The skipped count can vary by OS because fork E2E tests are POSIX-only and Windows spawn E2E tests are Windows-only.
 - Coverage: terminal total 87%, XML line-rate 88.97%, branch-rate 81.46%
 - multiprocess tests / OTel/structlog coexistence tests are included in the official quality gate
+- Type validation: public validation includes `mypy src`, `pyright src`, `pyright tests/typing_smoke`, and a 100% `pyright --verifytypes dsafelogger --ignoreexternal` completeness gate against the built wheel
 - free-threaded build test procedure included (`PYTHON_GIL=0 uvx ...`)
 - Internal synchronization verification on `scripts/check_design_docs_sync.py` and `scripts/generate_api_docs.py --check`
 - Fixed public representative session with `benchmarks/summary/manifest.json`
@@ -4168,7 +4170,7 @@ Based on the observed facts in this report as a whole, we have summarized the ob
 8. **Multiprocess raw throughput is led by stdlib logging**: In `root_p8`, D-SafeLogger reaches 63-75% of stdlib throughput. `BENCHMARK.md` states this clearly as a design tradeoff reflecting fixed costs in the specification (IPC + Writer dispatch). The multiprocess value of this library is not raw throughput, but delivery-state observability.
 9. **Classifies and explains 12/12 rows in the multiprocess resilience profile**: stdlib / loguru rows are marked with `observability_gap`. Delivery-state explainability is recorded as observability specific to this library.
 #### 8.8.5 Documentation/quality operations
-10. **Quality gate and internal synchronization verification scripts are in place**: 658 passed / 3 skipped on Python 3.14.3 / Windows (661 collected), coverage 87%, internal synchronization verification by `scripts/check_design_docs_sync.py` and `scripts/generate_api_docs.py --check`, public representative session fixation by `benchmarks/summary/manifest.json`. The skipped count can vary by OS. These are recorded as observable quality indicators when evaluating candidate libraries for introduction.
+10. **Quality gate and internal synchronization verification scripts are in place**: 658 passed / 3 skipped on Python 3.14.3 / Windows (661 collected), coverage 87%, `mypy` / `pyright` / typing smoke / `pyright --verifytypes` 100% completeness gate, internal synchronization verification by `scripts/check_design_docs_sync.py` and `scripts/generate_api_docs.py --check`, public representative session fixation by `benchmarks/summary/manifest.json`. The skipped count can vary by OS. These are recorded as observable quality indicators when evaluating candidate libraries for introduction.
 ---
 
 ### 8.9 Summary of this chapter

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D_SafeLogger_Specification_v23j_full.md

@@ -2496,3 +2496,7 @@ bounded wait (≤ timeout) -> visible warning (drain incomplete を可視化) ->
 ### v23j Publication Sync Addendum
 
 公開前同期として、coverage 再生成、API docs 再生成、examples の formal MP / external rotation 追加、GitHub workflow gate 強化、`docs/design/` 公開設計書の readiness check を対象に含める。これらは runtime behavior を変更しない公開成果物同期であり、release version は `0.2.0` のまま維持する。
+
+### v23j Type Validation CI Addendum
+
+0.2.2 向けの公開前品質ゲートとして、`py.typed` 配布に対応する型検証を CI に追加する。source typing は `mypy src` と `pyright src`、利用者視点の smoke test は `pyright tests/typing_smoke`、packaged typing は built wheel を install した上で `pyright --verifytypes dsafelogger --ignoreexternal` を 100% completeness threshold で検証する。verifytypes 実行時は editable install へ戻らないよう `uv run --no-sync` を使う。smoke test ディレクトリは spawn worker で標準ライブラリ `typing` を shadow しないよう `tests/typing_smoke/` とし、`tests/typing/` は使わない。これらは runtime behavior を変更しない公開品質ゲートの追加であり、release version bump は release readiness 確定時に行う。

{d_safelogger-0.2.1 → d_safelogger-0.2.2}/docs/design/D_SafeLogger_Specification_v23j_full_en.md

@@ -2479,3 +2479,7 @@ This design prevents the host process from permanently blocking even if an unkno
 ### v23j Publication Sync Addendum
 
 Pre-publication synchronization includes coverage regeneration, API docs regeneration, addition of formal MP/external rotation for examples, enhancement of GitHub workflow gate, and readiness check of `docs/design/` publication design document. These are public artifact synchronizations that do not change the runtime behavior and keep the release version as `0.2.0`.
+
+### v23j Type Validation CI Addendum
+
+For the 0.2.2 pre-publication quality gate, CI adds type validation for the `py.typed` distribution. Source typing is checked with `mypy src` and `pyright src`, user-perspective smoke typing is checked with `pyright tests/typing_smoke`, and packaged typing is checked from the installed built wheel with `pyright --verifytypes dsafelogger --ignoreexternal` at a 100% completeness threshold. The verifytypes step uses `uv run --no-sync` so the wheel install is not replaced by the editable install before verification. The smoke-test directory is named `tests/typing_smoke/`, not `tests/typing/`, to avoid shadowing the standard-library `typing` module in spawn workers. These additions change only the public quality gate, not runtime behavior; the release-version bump is deferred until release readiness is confirmed.