tomoaki/hakorune
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
2,297 Commits 1 Branch 0 Tags
9ba89bada2833c177321d745cfb7986b96e6ea54
Commit Graph

243 Commits

Author SHA1 Message Date
tomoaki
9ba89bada2 feat(pattern6): support reverse scan for last_index_of 
Extend Pattern6 (ScanWithInit) to handle both forward and reverse scans:
- Forward: i=0, loop(i < len), i=i+1 (existing)
- Reverse: i=len-1, loop(i >= 0), i=i-1 (NEW)

Implementation:
- Added ScanDirection enum (Forward/Reverse)
- Updated extract_scan_with_init_parts() to detect both patterns
- Created lower_scan_with_init_reverse() lowerer
- Pattern6 now selects appropriate lowerer based on scan direction

Files modified:
- src/mir/builder/control_flow/joinir/patterns/pattern6_scan_with_init.rs
- src/mir/join_ir/lowering/scan_with_init_reverse.rs (new)
- src/mir/join_ir/lowering/mod.rs

Known issue (pre-existing):
- PHI predecessor mismatch bug exists in Pattern6 (both forward and reverse)
- This bug existed BEFORE Phase 257 P0 implementation
- Out of scope for Phase 257 P0 - will be addressed separately

Phase 257 P0
 2025-12-20 20:28:41 +09:00
tomoaki
636b1406bc refactor(joinir): extract get_entry_function helper 
Phase 256.8.5: Eliminate duplication of entry function extraction logic
across Pattern2/4/6/7.

Changes:
- Add patterns/common/joinir_helpers.rs with get_entry_function() helper
- Refactor 4 files to use shared helper (eliminate ~40 lines of duplication)
- Consistent error messages across all patterns

Benefits:
- DRY principle: Single source of truth for entry function extraction
- Maintainability: Future changes only need to happen once
- Consistency: All patterns use identical logic and error messages
- Testability: Can be tested independently

Files refactored:
- pattern2_steps/emit_joinir_step_box.rs
- pattern4_with_continue.rs
- pattern6_scan_with_init.rs
- pattern7_split_scan.rs

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-20 20:10:48 +09:00
tomoaki
edc7355937 refactor(joinir): unify boundary join_inputs SSOT (pattern4/6/7) 
Apply Phase 256.8 SSOT fix to Pattern4/6/7:
- Use join_module.entry.params.clone() instead of hardcoded ValueIds
- Add fail-fast validation for params count mismatch
- Remove ValueId(0), ValueId(PARAM_MIN + k) patterns
- Clean up unused PARAM_MIN imports

This prevents entry_param_mismatch errors structurally and maintains
consistency with Pattern2/3.

Changes:
- pattern4_with_continue.rs: Lines 442-476 (SSOT extraction + validation)
- pattern6_scan_with_init.rs: Lines 447-471 (SSOT extraction + validation)
- pattern7_split_scan.rs: Lines 495-526 (SSOT extraction + validation)

All patterns now use the same SSOT principle:
1. Extract entry function (priority: join_module.entry → fallback "main")
2. Use params as SSOT: join_inputs = entry_func.params.clone()
3. Build host_inputs in expected order (pattern-specific)
4. Fail-fast validation: join_inputs.len() == host_inputs.len()

Verification:
- cargo build --release: ✅ PASS (no PARAM_MIN warnings)
- Quick profile: ✅ First FAIL still json_lint_vm (baseline maintained)
- Pattern6 smoke: ✅ PASS (index_of test)
- Pattern7 smoke: Pre-existing phi pred mismatch (not introduced by SSOT)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-20 20:05:11 +09:00
tomoaki
4439d64da3 refactor(joinir): make jump_args layout explicit (Phase 256) 2025-12-20 13:04:24 +09:00
tomoaki
1028bd419c fix(joinir): stabilize Phase 256 merge (jump_args, DCE, func names) 2025-12-20 11:01:48 +09:00
tomoaki
2c4268b691 wip(joinir): Phase 256 P1.10 - classify_tail_call continuation support 
## 修正内容
- classify_tail_call に is_target_continuation パラメータ追加
- k_exit 呼び出しを ExitJump として分類（BackEdge に誤分類されるのを防止）
- entry_func_name の決定ロジックを修正（loop_step を正しく識別）
- Continuation param bindings の Copy 生成ロジック追加

## 残存問題
bb8/bb9 の instructions が 0 になる問題
- add_block 時は 4 instructions ある
- printer 時は instructions=0, spans=4
- 原因調査中

🔍 WIP - ChatGPT レビュー用

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 2025-12-20 10:25:13 +09:00
tomoaki
077657b7a1 feat(joinir): Phase 256 P1.5-P1.7 contracts and naming SSOT 2025-12-20 06:38:21 +09:00
tomoaki
bfd324d7b9 feat(joinir): Phase 256 P1.5 - JoinInst::Select 命令サポート（根治） 
## 変更内容

Task 1（既完了）: boundary パラメータを bridge 経路全体に伝播
- conversion_pipeline.rs, bridge.rs（複数箇所）, meta.rs, routing_legacy_binding.rs, execute_box.rs

Task 2-A～2-D: Select 命令の実装
- 2-A: MirInstruction::Select バリアント追加
- 2-B: JoinIR Select → MIR Select 直接変換（branch+phi展開廃止）
- 2-C: ⭐ joinir_id_remapper に Select remap case（ValueId変換の根治）
- 2-D: value_collector に Select case

## 根本原因解決

Pattern7 の JoinInst::Select が JoinIR→MIR で未対応
→ ValueId(1002) → ValueId(57) のリマップが行われず
→ "use of undefined value ValueId(57)" エラー

## 現在地

✅ cargo check: 0 errors
⏳ Pattern7/6 VM test（Task 2-E LLVM実装後）

🧠 Generated with Claude Code
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-20 03:11:58 +09:00
tomoaki
64f679354a fix(joinir): Phase 256 P1 - Carrier PHI wiring and parameter mapping (in progress) 
**Status**: Core carrier PHI issue partially resolved, debugging loop body

**Progress**:
✅ Task 1: split_scan_minimal.rs Carriers-First ordering (6 locations)
✅ Task 2: pattern7_split_scan.rs boundary configuration (host/join inputs, exit_bindings, expr_result)
✅ Result now flows from k_exit to post-loop code (RC issue resolved)
⚠️  Loop body instruction execution needs review

**Key Fixes**:
1. Fixed host_inputs/join_inputs to match main() params Carriers-First order
2. Added result to exit_bindings (CarrierRole::LoopState)
3. Added result back to loop_invariants for variable initialization
4. Added expr_result=join_exit_value_result for loop expression return
5. Fixed jump args to k_exit to include all 4 params [i, start, result, s]

**Current Issue**:
- Loop body type errors resolved (String vs Integer fixed)
- New issue: Loop body computations (sep_len) undefined in certain blocks
- Likely cause: JoinIR→MIR conversion of local variables needs review

**Next Steps**:
- Review JoinValueSpace allocation and ValueId mapping in conversion
- Verify loop_step instruction ordering and block structure
- May need to refactor bound computation or revisit split algorithm

🤖 Generated with Claude Code
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-20 01:24:04 +09:00
tomoaki
575a5d750f refactor(joinir): Phase 255 P2 - loop_invariants + var() unification + Phase 256 prep 
## Task A: Loop Invariants Architecture (Option A - Boundary Extension)

Introduced explicit loop_invariants concept for variables that are:
- Used in loop body but never modified (e.g., substring needle in index_of)
- Need header PHI (all iterations use same value)
- Do NOT need exit PHI (not a LoopState)

Implementation:
- Added `loop_invariants: Vec<(String, ValueId)>` field to JoinInlineBoundary
- Added `with_loop_invariants()` method to JoinInlineBoundaryBuilder
- Modified Pattern 6 to use loop_invariants instead of ConditionOnly misuse
  * s (haystack) and ch (needle) now properly classified
  * exit_bindings simplified to only LoopState carriers
- Extended LoopHeaderPhiBuilder to generate invariant PHIs
  * Entry PHI: from host
  * Latch PHI: self-reference (same value every iteration)
- Updated instruction_rewriter to handle invariant latch incoming

Files Modified:
- src/mir/join_ir/lowering/inline_boundary.rs (structure + builder)
- src/mir/join_ir/lowering/inline_boundary_builder.rs (builder method)
- src/mir/builder/control_flow/joinir/patterns/pattern6_scan_with_init.rs
- src/mir/builder/control_flow/joinir/merge/loop_header_phi_builder.rs
- src/mir/builder/control_flow/joinir/merge/mod.rs
- src/mir/builder/control_flow/joinir/merge/instruction_rewriter.rs

## Task B: Code Quality - var() Helper Unification

Created common module to eliminate var() duplication:
- New module: src/mir/builder/control_flow/joinir/patterns/common/
- Centralized var() helper in ast_helpers.rs
- Updated Pattern 6 and Pattern 2 to use common var()
- Test code (5 occurrences) deferred to Phase 256+ per 80/20 rule

Result: Eliminated 2 duplicate var() functions, 5 test occurrences remain

Files Created:
- src/mir/builder/control_flow/joinir/patterns/common/ast_helpers.rs
- src/mir/builder/control_flow/joinir/patterns/common/mod.rs

Files Modified:
- src/mir/builder/control_flow/joinir/patterns/mod.rs
- src/mir/builder/control_flow/joinir/patterns/pattern6_scan_with_init.rs
- src/mir/builder/control_flow/joinir/patterns/policies/balanced_depth_scan_policy.rs

## Task C: Documentation - PostLoopEarlyReturnPlan Usage Examples

Enhanced post_loop_early_return_plan.rs with:
- Architecture explanation (exit PHI usage prevents DCE)
- Pattern 2 example (Less: balanced_depth_scan)
- Pattern 6 example (NotEqual: index_of)
- Builder defer decision: when 4+ patterns emerge, add builder

Files Modified:
- src/mir/builder/control_flow/joinir/patterns/policies/post_loop_early_return_plan.rs

## Task D: Phase 256 Preparation

Analyzed next failure from smoke tests:
- `StringUtils.split/2` with variable-step loop
- Not constant step (i += len or similar)
- Three implementation options identified

Created Phase 256 README with:
- Minimal reproduction code
- Root cause analysis
- Implementation options with trade-offs
- Clear next steps

Files Created:
- docs/development/current/main/phases/phase-256/README.md

## Verification Results

✅ All tests passing:
- pattern254_p0_index_of_vm.sh: PASS
- No regression in Pattern 1-5
- Smoke test progresses past json_lint_vm to next failure

✅ Architecture achievements:
- Semantic clarity: loop_invariants vs exit_bindings properly separated
- ConditionOnly misuse eliminated
- PHI generation correct for all carrier types
- Code quality: var() duplication reduced
- Next phase clearly defined

## Summary

Phase 255 P2 achieves root-cause fix for multi-param loop support:
- Boundary concept expanded (loop_invariants field)
- Pattern 6 architecture corrected (no ConditionOnly misuse)
- Code quality improved (var() centralized)
- Next pattern (variable-step) is now the challenge

✨ Box-first principles maintained: clean separation, explicit roles, minimal coupling

🧠 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-19 23:48:49 +09:00
tomoaki
2d9c6ea3c6 feat(joinir): Phase 254-255 - Pattern 6 (ScanWithInit) + exit PHI DCE fix 
## Phase 254: Pattern 6 (ScanWithInit) Detection & JoinIR Lowering

Pattern 6 detects index_of/find/contains-style loops:
- Loop condition: i < x.length()
- Loop body: if with method call condition + early return
- Step: i = i + 1
- Post-loop: return not-found value (-1)

Key features:
- Minimal lowering: main/loop_step/k_exit functions
- substring hoisted to init-time BoxCall
- Two k_exit jumps (found: i, not found: -1)
- Tests: phase254_p0_index_of_min.hako

## Phase 255 P0: Multi-param Loop CarrierInfo

Implemented CarrierInfo architecture for Pattern 6's 3-variable loop (s, ch, i):
- i: LoopState (header PHI + exit PHI)
- s, ch: ConditionOnly (header PHI only)
- Alphabetical ordering for determinism
- All 3 PHI nodes created correctly
- Eliminates "undefined ValueId" errors

## Phase 255 P1: Exit PHI DCE Fix

Prevents exit PHI from being deleted by DCE:
- PostLoopEarlyReturnStepBox emits post-loop guard
- if (i != -1) { return i } forces exit PHI usage
- Proven pattern from Pattern 2 (balanced_depth_scan)
- VM/LLVM backends working

## Test Results

✅ pattern254_p0_index_of_vm.sh: PASS (exit code 1)
✅ pattern254_p0_index_of_llvm_exe.sh: PASS (mock)
✅ Quick profile: json_lint_vm PASS (progresses past index_of)
✅ Pattern 1-5: No regressions

## Files Added

- src/mir/builder/control_flow/joinir/patterns/pattern6_scan_with_init.rs
- src/mir/join_ir/lowering/scan_with_init_minimal.rs
- apps/tests/phase254_p0_index_of_min.hako
- docs/development/current/main/phases/phase-254/README.md
- docs/development/current/main/phases/phase-255/README.md

🧠 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-19 23:32:25 +09:00
tomoaki
6a3b6deb20 feat(anf): Phase 145 P0/P1/P2 - ANF (A-Normal Form) transformation 
Implement ANF transformation for impure expressions to fix evaluation order:

Phase 145 P0 (Skeleton):
- Add anf/ module with contract/plan/execute 3-layer separation
- AnfDiagnosticTag, AnfOutOfScopeReason, AnfPlan enums
- Stub execute_box (always returns Ok(None))
- 11 unit tests pass

Phase 145 P1 (Minimal success):
- String.length() whitelist implementation
- BinaryOp + MethodCall pattern: x + s.length() → t = s.length(); result = x + t
- Exit code 12 verification (VM + LLVM EXE)
- 17 unit tests pass

Phase 145 P2 (Generalization):
- Recursive ANF for compound expressions
- Left-to-right, depth-first evaluation order
- Patterns: x + s.length() + z, s1.length() + s2.length()
- ANF strict mode (HAKO_ANF_STRICT=1)
- Diagnostic tags (joinir/anf/*)
- 21 unit tests pass, 0 regression

Also includes Phase 143 P2 (else symmetry) completion.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 16:19:49 +09:00
tomoaki
845ae70cb7 chore: Remove unused imports in normalized_shadow modules 
Cleaned up unused imports after Phase 143 execution fix (5e662eaaf).

**Priority files (Phase 143)**:
- if_as_last_join_k.rs: removed ValueId, BTreeMap
- loop_true_break_once.rs: added #[cfg(test)] for test-only imports
- post_if_post_k.rs: removed ValueId, BTreeMap
- normalized_helpers.rs: added #[cfg(test)] for Span

**Additional cleanup**:
- contract_checks.rs: removed BasicBlockId
- joinir/mod.rs: removed Info struct re-exports (functions kept)
- patterns/mod.rs: removed Info struct re-exports (functions kept)
- ast_feature_extractor.rs: removed EscapeSkipPatternInfo
- plan_box.rs: added #[cfg(test)] for PlanKind

**Verification**:
- 0 unused import warnings (was 20+)
- All 69 normalized_shadow tests pass
- Clean build with --release

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 08:36:45 +09:00
tomoaki
5e662eaaf6 fix(normalization): Phase 143 execution fix - Param region SSOT 
Problem: normalized_helpers allocated env params as ValueId(1,2...)
in PHI Reserved region (0-99) instead of Param region (100-999)
per JoinValueSpace contract.

Root cause: All 4 normalized shadow modules started from
next_value_id=1, violating the Param region contract.

Solution:
- Add NormalizedHelperBox::alloc_env_params_param_region()
  that allocates params starting from PARAM_MIN (100)
- Update 4 normalized shadow files to use new API:
  - loop_true_if_break_continue.rs
  - loop_true_break_once.rs
  - if_as_last_join_k.rs
  - post_if_post_k.rs
- Fix instruction_rewriter.rs type mismatch
  (func.signature.params → func.params)

Verification:
- Unit tests: 69/69 PASS
- VM smoke: exit code 7 ✅
- LLVM EXE smoke: exit code 7 ✅ (timeout resolved!)

ValueId Space Contract (Phase 201):
| Region       | Range    | Purpose                      |
|--------------|----------|------------------------------|
| PHI Reserved | 0-99     | Loop header PHI dst          |
| Param        | 100-999  | env params (flag, counter)   |
| Local        | 1000+    | Const, BinOp, condition      |

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 08:23:20 +09:00
tomoaki
4082abb30c feat(normalization): Phase 142 P0 - Loop statement-level normalization 
Phase 142-loopstmt P0: Statement-level normalization

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-19 05:28:49 +09:00
tomoaki
275fe45ba4 feat(normalization): Phase 142 P0 - Statement-level normalization 
## Summary

Changed normalization unit from "block suffix" to "statement (loop only)"
to prevent pattern explosion.

## Changes

1. **PlanBox** (`plan_box.rs`):
   - Always return `loop_only()` for any `loop(true)`, regardless of what follows
   - Subsequent statements (return, assignments) handled by normal MIR lowering
   - ~70 lines reduced, 7 unit tests updated

2. **build_block** (`stmts.rs`):
   - Removed `break` after consumed=1 from suffix_router
   - Continue processing subsequent statements normally
   - Phase 142 P0 comments added

3. **Tests**:
   - Fixture: `phase142_loop_stmt_only_then_return_length_min.hako`
   - VM smoke: exit code 3 (s="abc" → s.length() → 3)

## Results

- ✅ Unit tests: 10/10 passed
- ✅ Phase 142 VM smoke: PASS
- ✅ Phase 131 regression: PASS
- ✅ Build: Success

## Design

- **Pattern Explosion Prevention**: Normalize only the loop (consumed=1)
- **Out-of-Scope Policy**: Always Ok(None) for fallback
- **Fail-Fast**: Only for "in-scope but broken" cases

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 04:41:01 +09:00
tomoaki
3ef929df53 docs(normalization): Update README for Phase 142 P0 
- Document statement-level normalization unit change
- Add Phase 142 P0 section to Pattern Detection
- Mark Phase 132-135 patterns as LEGACY
- Update suffix_router description
- Clarify LoopWithPost deprecation status

Phase 142 P0: Normalization unit changed from "block suffix" to "statement"

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 04:40:35 +09:00
tomoaki
aaba27d311 refactor(normalization): Deprecate LoopWithPost variant 
- Mark LoopWithPost enum variant as deprecated (Phase 142 P0)
- Add deprecation to loop_with_post() constructor function
- Document migration path to LoopOnly
- Keep for backward compatibility until full migration

Phase 142 P0: Statement-level normalization makes LoopWithPost obsolete

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 04:39:25 +09:00
tomoaki
21a3c6b5d4 docs(normalization): Update suffix_router comments for Phase 142 P0 
- Reflect that post statements are no longer required
- Document LoopOnly pattern acceptance
- Update responsibility description

Phase 142 P0: Normalization unit changed to "statement (loop only)"
 2025-12-19 04:31:45 +09:00
nyash-codex
1742f0412e feat(normalized): Phase 141 P1.5 - external env inputs + KnownIntrinsic SSOT 
## Task B: External env input bug fix (Priority 1)

Fix: Suffix normalization couldn't access prefix-built local variables

**Problem**: `s.length()` failed because 's' (from prefix `s = "abc"`) was
not in available_inputs during suffix normalization.

**Root cause**: `AvailableInputsCollectorBox::collect()` only collected
function params and CapturedEnv, missing `builder.variable_map`.

**Solution**: Add `prefix_variables` parameter with 3-source merge:
1. Function params (highest priority)
2. Prefix variables (medium priority - NEW)
3. CapturedEnv (lowest priority)

**Changed files**:
- src/mir/control_tree/normalized_shadow/available_inputs_collector.rs
- src/mir/builder/control_flow/normalization/execute_box.rs
- src/mir/builder/control_flow/joinir/patterns/policies/normalized_shadow_suffix_router_box.rs
- src/mir/builder/control_flow/joinir/routing.rs
- src/mir/builder/stmts.rs
- src/mir/control_tree/normalized_shadow/dev_pipeline.rs
- docs/development/current/main/design/normalized-expr-lowering.md (Available Inputs SSOT section)

**Tests**: 3 new unit tests (prefix merge, priority order)

## Task A: KnownIntrinsic SSOT化 (Priority 2)

Eliminate string literal scattered matching by centralizing to registry.

**Problem**: Adding new intrinsics required editing if/match chains with
hard-coded string literals (`if method == KnownIntrinsic::Length0.method_name()`).

**Solution**: Create `KnownIntrinsicRegistryBox` as SSOT:
- `lookup(method, arity) -> Option<KnownIntrinsic>`
- `get_spec(intrinsic) -> KnownIntrinsicSpec`
- Adding new intrinsics now requires: (1) enum variant, (2) registry entry only

**Changed files**:
- src/mir/control_tree/normalized_shadow/common/known_intrinsics.rs (NEW)
- src/mir/control_tree/normalized_shadow/common/expr_lowerer_box.rs
- src/mir/control_tree/normalized_shadow/common/expr_lowering_contract.rs (deprecated methods removed)
- src/mir/control_tree/normalized_shadow/common/mod.rs
- docs/development/current/main/design/normalized-expr-lowering.md (Known Intrinsic SSOT section)

**Impact**: ~30% code reduction in intrinsic matching logic

## Task C: Better diagnostics (Priority 3)

Add `OutOfScopeReason::IntrinsicNotWhitelisted` for precise diagnostics.

**Changed files**:
- src/mir/control_tree/normalized_shadow/common/expr_lowering_contract.rs (enum variant)
- src/mir/control_tree/normalized_shadow/common/expr_lowerer_box.rs (diagnostic logic)

## Verification

✅ Build: `cargo build --release` - PASS
✅ Phase 97 regression: next_non_ws LLVM EXE - PASS
✅ Phase 131: loop(true) break-once VM - PASS
✅ Phase 136: return literal VM - PASS
✅ Phase 137: return x+2 VM - PASS

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 03:59:41 +09:00
nyash-codex
1264b3593d fix(normalization): Tighten loop(true) gate to prevent loop(i<n) mismatch 
Problem:
- Phase 97 LLVM EXE tests (next_non_ws, json_loader_escape) were failing
- NormalizationPlan accepted loop(i < n) but couldn't lower it
- Execute stage returned hard error instead of fallback

Root Cause:
- Plan stage only checked `ASTNode::Loop`, not condition
- loop(i < n) got Plan → Execute tried to normalize → error
- Should: loop(i < n) returns Ok(None) → fallback to Pattern2

Fix:
- Tighten gate in plan_box.rs (lines 50-71)
- Only accept loop(true) - literal Bool true
- loop(i < n) now returns Ok(None) → clean fallback

Added:
- Test: test_plan_block_suffix_no_match_loop_not_true()
- Verifies loop(i < n) returns None (not Plan)

Verification:
- Phase 97 next_non_ws LLVM EXE: PASS
- Phase 97 json_loader_escape LLVM EXE: PASS
- Phase 131/135/136/137 regression: all PASS

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-19 00:15:51 +09:00
nyash-codex
91c7dfbf0b refactor(normalization): Phase 135 P0 - Extend plan to zero post-loop assigns 
Generalize NormalizationPlan suffix detection to accept zero post-loop assignments:

Goal: Improve entry point consistency by allowing `loop + assign* + return` (N >= 0)

Implementation:
- Modified plan_box.rs detection logic (only file changed)
- Removed `post_assign_count >= 1` requirement
- Unified Phase 131 (loop + return) and Phase 132-133 (loop + assign+ + return) paths

Changes:
- src/mir/builder/control_flow/normalization/plan_box.rs:
  - Removed assignment count constraint
  - Unified pattern detection: `loop + assign* + return` (N >= 0)
- apps/tests/phase135_loop_true_break_once_post_empty_return_min.hako (new fixture)
- tools/smokes/v2/profiles/integration/apps/phase135_*.sh (new smoke tests)

Pattern support:
- Phase 131/135: loop + return only (consumed: 2, post_assign_count: 0) ✅
- Phase 132: loop + 1 assign + return (consumed: 3, post_assign_count: 1) ✅
- Phase 133: loop + N assigns + return (consumed: 2+N, post_assign_count: N) ✅

Design principles maintained:
- **Minimal change**: Only plan_box.rs modified (execute_box unchanged)
- **SSOT**: Detection logic centralized in plan_box.rs
- **Box-First**: Responsibility separation preserved (Plan/Execute)

Test results:
- Unit tests (plan_box): 9/9 PASS (2 new tests added)
- Phase 135 VM/LLVM EXE: PASS (exit code 1)
- Phase 131 regression: 2/2 PASS (path now unified)
- Phase 133 regression: 2/2 PASS
- cargo test --lib: PASS

Benefits:
- Unified entry point for all loop + post patterns
- Easier maintenance (single detection logic)
- Future extensibility (easy to add new patterns)
- Clear separation of Phase 131 and Phase 132-135 paths

Default behavior unchanged: Dev-only guard maintained

Related: Phase 135 normalization pattern consistency improvement

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 22:46:32 +09:00
nyash-codex
f4ab5ca5f4 refactor(control_flow): Phase 134 P0 - Normalization entry point SSOT 
Consolidate dual entry points into unified NormalizationPlan system:

Problem:
- Dual entry points: try_normalized_shadow() + suffix_router_box
- ~220 lines of duplicated pattern detection logic
- Maintenance burden: changes required in two places

Solution:
- New normalization module with Box-First architecture
- NormalizationPlanBox: Single source of truth for pattern detection
- NormalizationExecuteBox: Single source of truth for execution

Implementation:
- src/mir/builder/control_flow/normalization/ (new module)
  - README.md: Design contract (SSOT)
  - plan.rs: NormalizationPlan data structure
  - plan_box.rs: Pattern detection (7 unit tests)
  - execute_box.rs: Execution logic
  - mod.rs: Module integration

Refactored files:
- routing.rs::try_normalized_shadow(): 165 → 87 lines (-78 lines)
- suffix_router_box::try_lower_loop_suffix(): 258 → 116 lines (-142 lines)

Pattern support maintained:
- Phase 131: loop(true) only (consumed: 1)
- Phase 132: loop + single post (consumed: 3)
- Phase 133: loop + multiple post (consumed: 2+N)

Box-First principles:
- Plan Box: Detection responsibility only
- Execute Box: Execution responsibility only
- README.md: Contract documentation (SSOT)
- Clear separation enables independent testing

Test results:
- cargo test --lib: 1186 PASS (10 new tests added)
- Phase 133 VM/LLVM EXE: PASS (exit code 6)
- Phase 132 LLVM EXE: PASS (exit code 3)
- Phase 131 LLVM EXE: PASS (exit code 1)

Benefits:
- Code duplication eliminated (~220 lines)
- Single source of truth for normalization decisions
- Improved maintainability and testability
- Future-proof extensibility (easy to add new patterns)

Default behavior unchanged: Dev-only guard maintained

Related: Phase 134 normalization infrastructure improvement

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 22:29:29 +09:00
nyash-codex
d5a36cf818 refactor(joinir): Phase 132-R0 - Continuation SSOT + legacy isolation 
**Task 1: Continuation SSOT 一本化** ✅
- Add JoinInlineBoundary::default_continuations()
- Replace all BTreeSet::from([JoinFuncId::new(2)]) hardcoding (7 locations)
- Single source of truth for continuation function IDs

**Task 2: merge 契約 docs SSOT 化** ✅
- New: src/mir/builder/control_flow/joinir/merge/README.md
- Document continuation contracts, skip conditions, forbidden behaviors
- Prohibit by-name/by-id classification

**Task 3: テスト配置正規化** ✅
- New: src/mir/builder/control_flow/joinir/merge/tests/continuation_contract.rs
- Move tests from instruction_rewriter.rs to dedicated test file
- Add 4 test cases (Case A-D)

**Task 4: legacy 導線隔離** ✅
- New: src/mir/builder/control_flow/joinir/legacy/
- Move routing_legacy_binding.rs → legacy/routing_legacy_binding.rs
- Add legacy/README.md with removal conditions
- No cfg(feature="legacy") (docs-only isolation for now)

**Task 5: ノイズ除去** ✅
- Remove unused imports (ConstValue, MirInstruction)
- Clean warnings in touched files

Changes:
- src/mir/join_ir/lowering/inline_boundary.rs: +default_continuations()
- src/mir/builder/control_flow/joinir/merge/README.md: +140 lines
- src/mir/builder/control_flow/joinir/merge/tests/: +180 lines
- src/mir/builder/control_flow/joinir/legacy/: +3 files

Test results:
- cargo test --lib: 1176 PASS
- All Phase 131/132/97 smokes: PASS

Benefits:
- Continuation definition centralized (SSOT)
- Merge contracts documented and tested
- Legacy code path clearly isolated
- Code quality improved (warnings reduced)

Related: Phase 132 infrastructure improvements

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 21:52:21 +09:00
nyash-codex
149e30cae5 fix(joinir): Phase 132 P1 - k_exit continuation classification fix 
Problem: k_exit with TailCall(post_k) was classified as skippable continuation,
causing post_k block to become unreachable.

Root cause:
- Continuation classification was name-based ("join_func_2")
- Skippable continuation check didn't examine function body
- k_exit with tail call to post_k had its block skipped during merge

Solution:
- Check function body structure before classifying as skippable
- Non-skippable pattern: continuation with TailCall to another function
- Properly emit tail call as Jump instruction

Changes:
- src/mir/builder/control_flow/joinir/merge/mod.rs:
  - Add structural check in is_skippable_continuation()
  - Detect TailCall(other_func) pattern
  - Skip only if function is truly empty (1 block, Return only)

- src/mir/builder/control_flow/joinir/merge/instruction_rewriter.rs:
  - Improve k_exit continuation handling
  - Properly rewrite TailCall → Jump for post_k

Test results:
- Phase 132 LLVM EXE: PASS (exit code 3)
- Phase 132 VM: PASS
- Phase 131 regression: PASS
- Phase 97 regression: PASS

Related: Phase 132 loop(true) + post-loop VM/LLVM parity

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 21:51:52 +09:00
nyash-codex
b5d8ace6ab feat(control_tree): Phase 132 P0 + P0.5 - loop(true) + post-loop support 
**Phase 132 P0**: Extend loop(true) break-once to support post-loop statements

Goal: Support `loop(true) { x = 1; break }; x = x + 2; return x` → exit code 3

Implementation:
- loop_true_break_once.rs: Add post_k continuation generation
- Reuse Phase 130's lower_assign_stmt for post statements
- ExitMeta uses DirectValue mode (PHI-free)

**Phase 132 P0.5**: Fix StepTree post-loop statement visibility

Root cause: routing.rs created StepTree from Loop node only, losing post statements

Solution:
- New: normalized_shadow_suffix_router_box.rs
- Detects block suffix: Loop + Assign* + Return
- Creates StepTree from entire suffix (Block([Loop, Assign, Return]))
- Modified build_block() to call suffix router (dev-only)

Changes:
- apps/tests/phase132_loop_true_break_once_post_add_min.hako (new fixture)
- tools/smokes/v2/profiles/integration/apps/phase132_loop_true_break_once_post_add_*.sh
- src/mir/control_tree/normalized_shadow/loop_true_break_once.rs (+150 lines)
- src/mir/builder/control_flow/joinir/patterns/policies/normalized_shadow_suffix_router_box.rs (+380 lines)
- src/mir/builder/stmts.rs (build_block modified to support suffix skipping)

Design principles:
- StepTree unchanged: Block is SSOT for statement order
- No data duplication: Loop doesn't hold post_nodes
- Suffix router handles detection + conversion
- build_block() handles wiring only

Test results:
- Phase 132 VM: PASS (exit code 3)
- Phase 131 regression: PASS
- Phase 97 regression: PASS

Related: Phase 132 loop(true) + post-loop minimal support

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 21:51:33 +09:00
nyash-codex
d6225a88d0 refactor(joinir): Task 6 & 7 - MergeConfig unification + contract_checks tests 
Task 6: MergeConfig 一構造体化
- Consolidate dispersed parameters (dev_log, strict_mode, exit_reconnect_mode)
- Reduce function argument clutter in merge/mod.rs
- Single source of truth for merge configuration
- Added MergeConfig struct with factory methods (default, strict, with_debug)

Task 7: contract_checks 単体テスト追加
- Add 4 test cases for verify_all_terminator_targets_exist
- Test coverage: all_present, missing_allowed, missing_disallowed, MergeContracts creation
- Validate fail-fast behavior with freeze_with_hint
- Enable regression detection during future refactoring

Changes:
- src/mir/builder/control_flow/joinir/merge/mod.rs (+54 -6 lines)
- src/mir/builder/control_flow/joinir/merge/contract_checks.rs (+116 lines)

Test Results:
- New tests: 4 PASS
- Regression tests: 1162 PASS
- No breaking changes

Related: Phase 131 refactoring to improve code organization and maintainability

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2025-12-18 18:28:24 +09:00
nyash-codex
02c4c313e5 feat(control_tree): Phase 131 P1.5-P2 DirectValue exit reconnection 
Implement DirectValue mode for Normalized shadow exit handling:

**P1.5 Changes**:
- Add ExitReconnectMode::DirectValue (skip exit PHI generation)
- Carry remapped_exit_values through merge result
- Update host variable_map directly with exit values
- Fix loop(true) { x = 1; break }; return x to return 1 correctly

**P2 Changes**:
- Normalize k_exit continuation entry/exit edges
- Rewrite TailCall(k_exit) → Jump(exit_block) for proper merge
- Add verify_all_terminator_targets_exist contract check
- Extend ExitLineReconnector to handle DirectValue mode

**Infrastructure**:
- tools/build_llvm.sh: Force TMPDIR under target/ (EXDEV mitigation)
- llvm_exe_runner.sh: Add exit_code verification support
- Phase 131 smokes: Update for dev-only + exit code validation

**Contracts**:
- PHI-free: Normalized path uses continuations only
- Exit values reconnect via remapped ValueIds
- Existing patterns unaffected (既定挙動不変)

Related: Phase 131 loop(true) break-once Normalized support

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
 2025-12-18 17:48:05 +09:00
nyash-codex
052040e59c refactor(joinir): extract ExitArgsCollectorBox for jump_args SSOT 
Phase 118 refactoring: Box modularization for exit value collection.

# Changes
- Created `exit_args_collector.rs` with ExitArgsCollectorBox
- Extracted jump_args processing logic from instruction_rewriter.rs
- Implemented SSOT for offset calculation (0 vs 1 legacy layout)
- Added comprehensive unit tests (6 tests, all passing)

# Design
- **Responsibility**: Collects exit PHI inputs + carrier inputs from jump_args
- **SSOT**: Centralized offset calculation logic
- **Fail-Fast**: Validates jump_args length against exit_bindings
- **Phase 118 P2 Contract**: Uses boundary.exit_bindings as SSOT

# Key Features
- Filters ConditionOnly carriers (no exit PHI needed)
- Handles both direct mapping (offset=0) and legacy layout (offset=1)
- Strict mode: Fail-Fast on validation errors
- Non-strict mode: Warns and continues with best effort

# Test Results
- Unit tests: 6/6 PASS
- Phase 118 smoke tests: PASS (VM + LLVM)
- Phase 117 regression: PASS

# Reduced Code Complexity
- instruction_rewriter.rs: ~90 lines → ~25 lines (box call)
- Offset logic: Centralized in ExitArgsCollectorBox

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 2025-12-18 03:56:21 +09:00
nyash-codex
1080dee58f fix(joinir): Phase 118 Pattern3 exit carrier PHI SSOT 2025-12-18 03:43:00 +09:00
nyash-codex
ce7e2c1b91 refactor(env): centralize NYASH_JOINIR_STRUCTURE_ONLY flag 
Moved NYASH_JOINIR_STRUCTURE_ONLY environment variable handling
from inline std::env::var() in routing.rs to a centralized helper
function in src/config/env/joinir_flags.rs.

Changes:
- Added joinir_structure_only_enabled() helper function
- Replaced direct env::var() call in routing.rs with helper
- Maintains existing behavior: default ON, NYASH_JOINIR_STRUCTURE_ONLY=0/off to disable
- Follows env module centralization pattern (Box Theory: separation of concerns)

Testing:
- cargo test --lib: 1126 passed ✅
- Regression: Phase 107 VM smoke ✅
- Phase 113/114 maintained ✅

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 2025-12-18 02:19:14 +09:00
nyash-codex
09ce24e132 feat(joinir): Phase 112 strict guard for StepTree required_caps 
- Add control_tree_capability_guard.rs with check(tree, func_name, strict, dev)
- Allowlist: If, NestedIf, Loop, Return, Break, Continue
- Deny (strict): NestedLoop, TryCatch, Throw, Lambda, While, ForRange, Match, Arrow
- Wire into lower_function_body() (strict-only check)
- Error format: [joinir/control_tree/cap_missing/<Cap>] with 1-line Hint
- Unit tests: nested_loop_rejects, if_only_passes, strict_false_passes
- Default behavior unchanged (strict=false always Ok)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 2025-12-18 01:38:25 +09:00
nyash-codex
14730c227f feat(control_tree): add StepTreeContract and signature (dev-only) 2025-12-18 00:57:58 +09:00
nyash-codex
9bcda215f8 refactor(joinir): split Pattern2 orchestrator into smaller steps 2025-12-18 00:44:31 +09:00
nyash-codex
3987aa5b06 refactor(joinir): drop unused break helper wrappers 2025-12-18 00:30:01 +09:00
nyash-codex
2b5c141e22 feat(control_tree): add StepTree builder (dev-only) 2025-12-18 00:22:21 +09:00
nyash-codex
a9f92f8f1a refactor(joinir): add hints for Phase107/104/100 policy rejects 
- balanced_depth_scan: missing_tail_inc, missing_return_i hints
- read_digits: missing_eos_guard, digit_set_mismatch hints
- pinned: missing_host_id hint
- Gradual migration (representative cases only)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
 2025-12-17 23:45:49 +09:00
nyash-codex
27e8e0f16a refactor(joinir): Phase 108 unify Pattern2 policy routing 2025-12-17 23:30:08 +09:00
nyash-codex
3c1d8a824d feat(joinir): extend balanced depth-scan policy to object family 2025-12-17 23:09:19 +09:00
nyash-codex
09dd10349f refactor(joinir): make Pattern2 body-local handling policy-driven 2025-12-17 23:00:26 +09:00
nyash-codex
10a2f3b48b refactor(joinir): centralize balanced depth-scan policy decision 2025-12-17 22:59:27 +09:00
nyash-codex
6036627920 fix(joinir): avoid false rejects in balanced depth-scan policy 2025-12-17 22:52:29 +09:00
nyash-codex
d8ce9fdb99 fix(joinir): wire balanced depth-scan policy through Pattern2 2025-12-17 22:47:36 +09:00
nyash-codex
9ec2e28b6a feat(joinir): add Pattern2 post-loop early return step 2025-12-17 22:32:55 +09:00
nyash-codex
aa29dc8085 feat(joinir): emit balanced depth-scan derived vars 2025-12-17 22:32:50 +09:00
nyash-codex
bf80789757 feat(joinir): route balanced depth-scan via Pattern2 policy 2025-12-17 22:32:43 +09:00
nyash-codex
f1a570fd45 feat(joinir): Phase 107 balanced depth-scan policy (analysis-only) 2025-12-17 22:25:34 +09:00
nyash-codex
5f4d8ba112 refactor(joinir): Phase 106 Pattern2 step boxes 2025-12-17 22:01:19 +09:00
nyash-codex
368b363694 refactor(joinir): split Pattern2 facts from lowering orchestration 2025-12-17 21:34:11 +09:00
nyash-codex
ae2a1d4339 refactor(joinir): boxify Pattern2 routing and schedule facts 2025-12-17 21:24:46 +09:00