feat(joinir): Phase 132-P3 - Exit PHI collision early detection

Added verify_exit_phi_no_collision() to contract_checks.rs for early detection
of ValueId collisions between exit PHIs and other instructions.

Problem detected:
- If exit_phi_builder uses builder.value_gen.next() (module-level) instead of
  func.next_value_id() (function-level), ValueIds can collide:

  Example:
  - bb0: %1 = const 0   (counter init)
  - bb3: %1 = phi ...   (exit PHI - collision!)

Previous behavior:
- Error only detected at LLVM backend runtime
- Cryptic error: "Cannot overwrite PHI dst=1"

New behavior:
- Panic at Rust compile time (debug build)
- Clear error message with fix suggestion:
  "Exit PHI dst %1 collides with instruction in block 0
   Fix: Use func.next_value_id() in exit_phi_builder.rs"

Benefits:
- 🔥 Fail-Fast: Catch errors during Rust compilation, not LLVM execution
- 📋 Clear messages: Exact collision point + fix suggestion
- 🧪 Testable: verify_exit_phi_no_collision() can be unit tested

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

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

docs/development/current/main/investigations/phase132-case-c-llvm-exe.md

@@ -4,7 +4,7 @@
 2025-12-15
 
 ## Status
-🔴 **FAILED** - LLVM executable returns wrong result
+✅ **FIXED** - VM / LLVM EXE parity achieved
 
 ## Summary
 Testing `apps/tests/llvm_stage3_loop_only.hako` (Pattern 5: InfiniteEarlyExit) in LLVM EXE mode reveals a critical exit PHI usage bug.
@@ -32,7 +32,39 @@ static box Main {
 
 ## Actual Behavior
 - **VM execution**: `Result: 3` ✅
-- **LLVM EXE**: `Result: 0` ❌
+- **LLVM EXE**: `Result: 0` ❌（修正前）
+
+## Final Root Cause (Confirmed)
+
+**JoinIR merge 側の Exit PHI ValueId 衝突**。
+
+`src/mir/builder/control_flow/joinir/merge/exit_phi_builder.rs` が PHI dst の割り当てに
+`builder.value_gen.next()`（モジュールレベルのカウンタ）を使っていたため、
+**同一関数内で `ValueId` が衝突**し、結果として exit 側で参照すべき PHI が別の値に潰れていた。
+
+例（概念）:
+
+- `bb0: %1 = const 0`（counter 初期値）
+- `bb3: %1 = phi ...`（exit PHI）
+
+**同じ `%1`** になり得るため、exit 側で `counter` を読んでも 0 になってしまう。
+
+### Fix
+
+PHI dst の割り当てを **関数ローカルの allocator** に統一する。
+
+- Before: `builder.value_gen.next()`（module-level）
+- After: `func.next_value_id()`（function-level）
+
+コミット:
+- `bd07b7f4 fix(joinir): Phase 132-P2 - Exit PHI ValueId collision fix`
+
+### Verification
+
+| テスト | VM | LLVM |
+|---|---:|---:|
+| Pattern 1 (`/tmp/p1_return_i.hako`) | 3 ✅ | 3 ✅ |
+| Case C (`apps/tests/llvm_stage3_loop_only.hako`) | Result: 3 ✅ | Result: 3 ✅ |
 
 ## Root Cause Analysis
 
@@ -52,6 +84,10 @@ bb3:  ; Exit block
 
 **MIR is correct**: The exit block (bb3) has a PHI node that receives the counter value from bb6, and subsequent instructions correctly use `%1`.
 
+補足:
+- 上記の “MIR が正しい” は「構造として PHI がある」意味で、**ValueId 衝突があると意味論が壊れる**。
+- 本件は “JoinIR merge の ValueId allocator の選択” が原因だったため、LLVM 側で 0 に見える形で顕在化した。
+
 ### LLVM IR (BUG)
 ```llvm
 bb3:
@@ -68,14 +104,13 @@ bb3:
 
 **Bug identified**: The PHI node `%"phi_1"` is created correctly and receives the counter value from `%"add_8"`. However, **all subsequent uses of ValueId(1) are hardcoded to `i64 0` instead of using `%"phi_1"`**.
 
-### Hypothesis
-The Python LLVM builder is not correctly resolving ValueId(1) when lowering instructions in bb3. Possible causes:
+### Discarded Hypotheses (Superseded)
 
-1. **vmap issue**: The PHI node is created and stored in `self.vmap[1]` during `setup_phi_placeholders`, but when lowering instructions in bb3, `vmap_cur` may not contain the PHI.
+当初は「Python LLVM builder が ValueId(1) を解決できず 0 にフォールバックしている」と推測し、
+PHI registration / vmap filtering 周りを疑った。
 
-2. **Resolution fallback**: When `resolve_i64_strict` fails to find ValueId(1), it falls back to `ir.Constant(i64, 0)`.
-
-3. **Block-local vmap initialization**: `vmap_cur` is initialized with `dict(builder.vmap)` at the start of each block, but something may be preventing the PHI from being included.
+ただし最終的に、本件の主因は **Rust 側（JoinIR merge）の ValueId 衝突**であることが確定したため、
+ここでの Python 側の推測は “症状の説明” に留まる（根因ではない）。
 
 ## Investigation Steps
 
@@ -112,7 +147,7 @@ The Python LLVM builder is not correctly resolving ValueId(1) when lowering inst
 - Exit PHI generation: `src/mir/join_ir/lowering/simple_while_minimal.rs`
   - Pattern 5 (InfiniteEarlyExit) lowering
 
-## Root Cause Identified
+## Interim Hypothesis (Historical)
 
 ### Bug Location
 `/home/tomoaki/git/hakorune-selfhost/src/llvm_py/llvm_builder.py:342-343`
@@ -164,6 +199,5 @@ This ensures PHIs are included in `vmap_cur` when lowering their defining block.
 - ✅ STRICT mode passes without fallback warnings
 
 ## Related Documents
-- [Phase 132 Plan](/home/tomoaki/git/hakorune-selfhost/docs/development/current/main/phase132-plan.md)
 - [Phase 131-3 LLVM Lowering Inventory](/home/tomoaki/git/hakorune-selfhost/docs/development/current/main/phase131-3-llvm-lowering-inventory.md)
-- [Simple While Minimal Lowering](/home/tomoaki/git/hakorune-selfhost/src/mir/join_ir/lowering/simple_while_minimal.rs)
+- [Phase 132 Summary](/home/tomoaki/git/hakorune-selfhost/docs/development/current/main/phases/phase-132/README.md)