hakorune/docs/development/current/main/logging_policy.md

# Logging & Output Policy (Phase 99)

## Overview

This document establishes the clear separation of concerns between three logging/output layers in the Nyash runtime, and provides guidelines for transitioning the remaining ~1477 println!/eprintln! call sites to the appropriate mechanism.

**Status**: Phase 101-B in progress — documentation plus partial internal log migration (Ring0.log) and test-output policy fixed.

---

## Section 1: Three-Layer Log/Output Role Separation

The Nyash runtime uses three distinct layers for logging and output, each with a specific purpose:

### Layer 1: Ring0.log (OS API Layer)

**Purpose**: Runtime/OS layer internal logging

**Use Cases**:
- Memory management state
- Garbage collection events
- Thread management
- Cache information
- Internal state tracking

**Target Audience**: Developers, debugging, measurement, internal state tracking

**API**:
```rust
ring0.log.debug("message");
ring0.log.info("message");
ring0.log.warn("message");
ring0.log.error("message");
```

**Characteristics**:
- Not intended for direct user visibility
- Controlled by log levels
- Can be redirected to files or monitoring systems
- Developer-facing diagnostics

---

### Layer 2: ConsoleService (Box Layer - User-Facing)

**Purpose**: Direct CLI output for end users

**Use Cases**:
- Error messages displayed to users
- Success notifications
- Help text
- Progress information
- Command execution results

**Target Audience**: End users of the Nyash CLI

**Access Methods**:
1. Via `console_println!` macro (Rust code)
2. Via `host.core.console.println(...)` (Box code)

**API**:
```rust
// Rust side
console_println!("User-visible message");

// Box side
host.core.console.println("User-visible message");
```

**Characteristics**:
- Active after PluginHost initialization
- Falls back to `eprintln!` before initialization (Graceful Degradation)
- User-friendly messages
- Respects output redirection

---

### Layer 3: Raw println!/eprintln! (Restricted Use)

**Purpose**: Limited to tests and temporary debugging only

**Restrictions**:
- Should be removed from production paths
- Should be removed from selfhost/hack_check/VM runner paths
- Permitted in test code (isolated environment)

**Current Status**:
- Test code: ~299 instances (permitted)
- Production code: ~1477 instances (to be migrated)

**Phase 99 Stance**:
- Judgment only in Phase 99
- Implementation deferred to Phase 100+
- Test code println! remains as-is (safe in isolated environment)

---

## Section 2: Macro Policy

### console_println! Macro (Implemented in Phase 98)

**Purpose**: Safe entry point for user-facing messages

**Implementation**:
```rust
macro_rules! console_println {
    ($($arg:tt)*) => {{
        if let Some(host) = try_get_core_plugin_host() {
            host.core.console.println(&format!($($arg)*));
        } else {
            eprintln!($($arg)*);  // Graceful fallback
        }
    }};
}
```

**Usage Locations**:
- selfhost runner main output
- hack_check result display
- VM runner main output (RC, errors, etc.)

**Design Principle**:
- Follows Fail-Fast principle with exception for output destination selection
- Dynamic fallback is permitted for "where to output" decision only
- Core logic remains static and deterministic

---

### dev_eprintln! / Other dev_* Macros (Under Consideration)

**Purpose**: Temporary debug output during development

**Phase 99 Decision**:
- Evaluate necessity only
- No implementation in Phase 99
- Implementation deferred to Phase 100+

**Rationale**:
- Most use cases can be covered by Ring0.log or console_println!
- May be redundant with existing logging infrastructure
- Need to assess actual developer needs before implementing

---

## Section 3: Test Code println! Policy

### Current Status

- Test code: ~299 instances of println!/eprintln!
- Purpose: Test output and result visualization
- Location: Throughout test modules

### Policy

**Phase 99 Stance**: **Leave as-is - OK to use println! in tests**

**Rationale**:
1. Tests run in isolated environments
2. Test output is separate from production output
3. println! is safe and appropriate for test diagnostics
4. No need for special test macros (println_test!, etc.)

**Future Considerations**: No changes planned

---

## Section 4: Complete Integration Criteria

The "CoreServices Log/Output Complete Integration" is considered achieved when:

### Console (Required)

- ✅ All selfhost runner user-facing output uses console_println!
- ✅ All hack_check result display uses console_println!
- ✅ All VM runner main output (RC, errors) uses console_println!

**Current Status (Phase 98)**:
- ✅ 7 locations completed
- ✅ Representative paths covered
- 🔄 ~359 locations remaining (Phase 100-101)

---

### Ring0.log (Phased)

**Existing Infrastructure**:
- debug/info/warn/error API available
- StdLog implementation outputs to stdout/stderr

**Migration Priority**:
1. Internal errors → ring0.log.error(...)
2. VM execution logs → ring0.log.info(...)
3. Memory/GC information → ring0.log.debug(...)

**Phase 99 Scope**: Planning only - document migration strategy

---

### Tests (Permitted)

- Test code println! (~299 locations): **Keep as-is**
- Production paths: Removed (migration complete)

---

## Section 5: Migration Strategy

### Phase-by-Phase Approach

**Phase 99 (Current)**:
- Document policy and categorization
- No code changes
- Establish migration framework

**Phase 100-101**:
- Migrate user-facing println! to console_println! (~366 locations)
- Prioritize src/runner/ paths
- Focus on visible user messages

**Phase 102+**:
- Evaluate Ring0.log usage for internal logging
- Migrate dev-debug println! as needed
- Leave test println! unchanged

---

### Categorization of Remaining println!/eprintln!

Total: 1776 locations (1477 excluding tests)

**Categories**:

1. **user-facing** (~366 locations, priority: high)
   - CLI messages, errors, help text
   - Target: console_println!
   - Phase 98: 7 completed (representative paths)
   - Phase 100-101: Gradual expansion

2. **dev-debug** (TBD, Ring0.log candidates)
   - Temporary debug output
   - Target: Ring0.log or dev_* macros (to be decided)
   - Priority: Medium
   - Phase 99: Assessment only

3. **test** (~299 locations, priority: low)
   - Test output and verification
   - Target: Keep as-is (isolated environment)
   - Phase 99-101: No changes

4. **internal** (~812 locations remaining)
   - Internal processing println! remnants
   - Target: TBD in Phase 99
   - Phase 99: Uncategorized

---

## Section 6: Design Principles

### Graceful Degradation

The console_println! macro implements graceful degradation:
- Primary: Use ConsoleService when available
- Fallback: Use eprintln! before PluginHost initialization
- Rationale: Output should always work, even during early initialization

This is the **only permitted exception** to the Fail-Fast principle, as it only affects output destination, not core logic.

---

### Separation of Concerns

Each layer has a clear responsibility:
- **Ring0.log**: Developer diagnostics
- **ConsoleService**: User communication
- **println!**: Test-only convenience

This separation ensures:
- Maintainability: Clear ownership of output
- Testability: Test output isolated from production
- Flexibility: Each layer can be configured independently

---

### Migration Philosophy

**80/20 Rule Applied**:
- Phase 98: 7 locations = covering representative paths
- Phase 100-101: ~366 locations = user-visible improvements
- Phase 102+: Remaining locations = diminishing returns

Focus on high-impact migrations first, defer low-priority changes.

---

## Related Documentation

- [Ring0 Inventory](ring0-inventory.md) - println! categorization and Ring0.log plans
- [CoreBoxes Design - Section 15.6-A](core_boxes_design.md#section-156-a-logsoutput-unified-design) - Architectural context
- [Phase 85 CURRENT_TASK](../../../CURRENT_TASK.md) - Implementation timeline

---

## Summary

Phase 99 establishes the **documentation foundation** for future logging/output migrations:

1. **Clear role separation**: Ring0.log (internal) vs ConsoleService (user-facing) vs println! (test-only)
2. **Macro policy**: console_println! (implemented), dev_* macros (under consideration)
3. **Test policy**: println! in tests is OK (isolated environment)
4. **Migration strategy**: Phased approach with clear priorities

**Next Steps**: Phase 100+ will implement gradual migrations based on this policy framework.

---

## Section 7: Phase 100 Implementation Complete (2025-12-03)

### user-facing 出力の CoreServices 化完了

**実装概要**: selfhost と LLVM runner の主要なユーザー向け出力を ConsoleService (console_println!) 経由に統一

**完了箇所**:
- **selfhost.rs**: 6箇所 → console_println!
  - Line 57: CoreInitError 出力
  - Line 194, 363, 418, 519, 570: Result 出力
- **llvm.rs**: 23箇所（ユーザー向けメッセージ） → console_println!
  - Line 26, 44, 53, 60, 116: エラーメッセージ（❌）
  - Line 121-122: 成功メッセージ（📊）
  - Line 215, 230, 239, 275, 287, 295: LLVM/harness エラー
  - Line 324, 328, 334-335, 353-354, 357-358, 362: 実行結果
  - Line 369-370, 379, 383, 391: Mock LLVM メッセージ
- **vm.rs**: 1箇所（Phase 98 で完了済み）
- **core_bridge.rs**: 2箇所（Phase 98 で完了済み）
- **selfhost 関連**: 5箇所（Phase 98 で完了済み）

**合計**: Phase 98 (7箇所) + Phase 100 (29箇所) = **36箇所完了**

**除外箇所**（意図的に残した）:
- llvm.rs の `[joinir/llvm]`, `[parse/context]` デバッグログ（Phase 101 対象）
- hack_check: .hako アプリ（Nyash言語の ConsoleBox 経由、別フェーズ）

**テスト結果**:
- ✅ cargo build --release 成功
- ✅ core_services テスト: 11 passed
- ✅ plugin_host テスト: 7 passed
- ✅ 代表ケース動作確認:
  - loop_min_while.hako: "📊 MIR Module compiled successfully!" 等が console_println! 経由で出力
  - エラーケース: "❌ Error reading file..." が console_println! 経由で出力

**残りの user-facing 出力**:
- 推定: ~330箇所（その他の runner/modes/*）
- 優先度: HIGH → Phase 101-102 で段階的拡張

**技術的成果**:
- selfhost/LLVM runner のユーザー向けメッセージが ConsoleService に統一
- Phase 99 で確立したログ/出力ポリシーが実装レベルで実現
- デバッグログとユーザー向け出力の明確な分離

---

## Section 7-A: Phase 101-A dev-debug ログの Ring0.log 統一（2025-12-03）

### dev-debug ログ ~34箇所を Ring0.log に統一

**実装概要**: llvm.rs, loop_form.rs, phi_core モジュールの代表的なデバッグログを Ring0.log に移行

**完了箇所**:
- **llvm.rs**: 13箇所 → Ring0.log
  - `[joinir/llvm]` デバッグログ: 12箇所（JoinIR 実験パス関連）
  - `[parse/context]` デバッグログ: 1箇所（プリロードファイル一覧）

- **loop_form.rs**: 全 [loopform] ログ → Ring0.log
  - `[loopform]` 基本ログ: 変数マップ、ブロックID等
  - `[loopform/condition]` ログ: 条件式処理関連
  - `[loopform/writes]` ログ: 変数書き込み収集
  - `[loopform/27.4-C]` ログ: JoinIR ヘッダーφバイパス

- **phi_core モジュール**: 21箇所 → Ring0.log
  - **loopform_builder.rs**: 16箇所
    - `[loopform/prepare]` ログ: 構造準備、変数分類、サマリー
    - `[loopform/seal_phis]` ログ: PHI シール処理
  - **loop_snapshot_merge.rs**: 5箇所
    - `[Option C]` ログ: Exit PHI 分類、変数解析

**実装パターン**:
```rust
// Before
eprintln!("[loopform] variable_map size={}", size);

// After
crate::runtime::get_global_ring0().log.debug(&format!(
    "[loopform] variable_map size={}", size
));
```

**合計**: Phase 101-A で **34箇所**のデバッグログを Ring0.log に統一

**テスト結果**:
- ✅ cargo build --release 成功（警告のみ、エラーなし）
- ✅ VM実行テスト: loop_min_while.hako 正常動作
- ✅ デバッグログが stderr に出なくなることを確認

**環境変数制御**:
- `NYASH_LOOPFORM_DEBUG=1`: LoopForm 関連デバッグログ有効化
- `NYASH_OPTION_C_DEBUG=1`: Option C 関連デバッグログ有効化
- Ring0.log のログレベル設定で出力制御可能

**残りの dev-debug ログ**:
- 推定: ~585箇所（全体 ~615箇所から Phase 101-A の 34箇所を除く）
- 対象外: test 出力（~347箇所、Phase 101-B で別途検討）
- 対象外: internal 出力（Phase 101-B で別途検討）

**技術的成果**:
- Ring0.log で dev-debug ログを一元管理
- 環境に応じた出力制御が可能（将来の活用に向けて）
- stderr の cleanness 向上（ユーザー向けメッセージのみになる）
- Phase 99-100 で確立した 3層設計を実装レベルで完成

---

## Section 7-B: Phase 101-B internal/test ログ整理（2025-12-04）

### 内部ログを Ring0.log に寄せ、テスト出力ポリシーを固定

**実装概要**:
- internal/dev ログ 26 箇所を Ring0.log に移行（stderr 汚染を削減） → 第1バッチ
  - 対象: provider_lock.rs, plugin_loader_unified.rs, type_meta.rs, deprecations.rs, leak_tracker.rs
  - Plugin loader v2 系: loader/config.rs, loader/library.rs, loader/metadata.rs, instance_manager.rs, ffi_bridge.rs
- internal/dev ログ 21 箇所を追加で Ring0.log 化 → 第2バッチ
  - 対象: provider_verify.rs, scheduler.rs, gc_controller.rs, box_registry.rs
  - Plugin loader v2 specs: loader/specs.rs（TypeBox ABI/trace）
  - Runner trace: runner/trace.rs（cli_verbose トレース）
  - MIR verifier dev-trace: mir/verification.rs（NYASH_BREAKFINDER_SSA_TRACE/NYASH_DEBUG_VERIFIER）
- internal/dev ログ 20 箇所を追加で Ring0.log 化 → 第3バッチ
  - MIR core: basic_block.rs, control_form.rs, hints.rs, effect.rs, printer.rs, optimizer.rs
- internal/dev ログ 26 箇所を追加で Ring0.log 化 → 第4バッチ
  - MIR builder/region: loop_builder/phi_ops.rs, builder/type_registry.rs, region/observer.rs
  - Plugin loader v2: enabled/extern_functions.rs（trace）/types.rs（finalize trace）
- internal/dev ログ 20 箇所を追加で Ring0.log 化 → 第5バッチ
  - MIR loop_builder JoinIR: joinir_if_phi_selector.rs（dry-run trace）, control.rs（LoopForm debug）
  - MIR builder observe: observe/types.rs（NYASH_MIR_TYPE_TRACE）, observe/resolve.rs（NYASH_DEBUG_KPI_KNOWN）
  - joinir VM bridge: join_ir_vm_bridge_dispatch/exec_routes.rs（run_generic_joinir_route trace）
  - Plugin loader v2: enabled/extern_functions.rs（NYASH_DEBUG_TRACE / runtime_checkpoint_trace / NYASH_BOX_INTROSPECT_TRACE）
- ログレベル整理: init/loader 失敗は error、warn-once 系は warn、トレースは debug/info に整理
- ログレベル整理: init/loader 失敗は error、warn-once 系は warn、トレースは debug/info に整理

**テスト出力方針**:
- Rust テスト内（src/tests/, tests/）の println!/eprintln! は原則許容（大きな出力のみ将来検討）
- 本フェーズではテストコードは無変更、ポリシーを docs に明文化

**残件**:
- internal/dev ログ残量: 概算で ~475–495 箇所（引き続き段階的に Ring0.log へ移行/削除）
- user-facing: console_println! 移行は別ラインで継続
- .hako/hack_check: Rust とは別フェーズで整理

**成果**:
- Ring0/Ring1/Core の責務分離を保ったまま internal ログを OS 抽象層に集約
- 環境変数ベースのデバッグトレース（PLUGIN_DEBUG, HAKO_*）も Ring0.log 経由に統一
- stderr のノイズ低減とログ観測の一元化を達成