tomoaki/hakorune
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix(llvm): Phase 131-10 - Smart console.log routing（Segfault修正）

Browse Source 
## 問題
- Integer値をi8*ポインタに変換 → Segfault（Exit 139）

## 解決策
- String literal → nyash.console.log(i8*)
- Integer/Handle → nyash.console.log_handle(i64)

## 結果
- Case B (loop_min_while): LLVM outputs `0,1,2` ✅
- VM/LLVM完全パリティ達成

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

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
nyash-codex
2025-12-14 09:19:00 +09:00
parent bfe047840b
commit 18bf35e6d4
8 changed files with 97 additions and 104 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
CURRENT_TASK.md
View File

@ -17,6 +17,8 @@
- Pattern2 の promoted carriers（DigitPos/Trim）について ExitLine 契約（ConditionOnly を exit PHI から除外）を E2E で固定済み。
- debug flag SSOT / DebugOutputBox 移行 / error tags 集約 / carrier init builder まで整備済み。
- **LLVM exe line SSOT 確立**: `tools/build_llvm.sh` を使用した .hako → executable パイプライン標準化完了。
- **LLVM AOT（Python llvmlite）ループ復旧**: `apps/tests/loop_min_while.hako` が EMIT/LINK/RUN まで到達（Phase 131-3..10）。
- 残り: `loop(true)` + `break/continue` の JoinIR パターン穴（Case C）
- **Phase 88 完了**: continue + 可変ステップ（i=i+const 差分）を dev-only fixture で固定、StepCalculator Box 抽出。
- **Phase 89 完了**: P0（ContinueReturn detector）+ P1（lowering 実装）完了。
- **Phase 90 完了**: ParseStringComposite + `Null` literal + ContinueReturn（同一値の複数 return-if）を dev-only fixture で固定。
@ -32,6 +34,7 @@
- `docs/development/current/main/phase87-selfhost-llvm-exe-line.md`
- `docs/development/current/main/phase131-2-box-resolution-map.md`
- `docs/development/current/main/phase131-3-llvm-lowering-inventory.md`
- `docs/development/current/main/phase131-5-taglink-fix-summary.md`
---

3
crates/nyash_kernel/src/plugin/console.rs
View File

@ -25,12 +25,11 @@ pub extern "C" fn print(ptr: *const i8) -> i64 {
#[export_name = "nyash.console.log_handle"]
pub extern "C" fn nyash_console_log_handle(handle: i64) -> i64 {
use nyash_rust::runtime::host_handles as handles;
eprintln!("DEBUG: handle={}", handle);
if let Some(obj) = handles::get(handle as u64) {
let s = obj.to_string_box().value;
println!("{}", s);
} else {
eprintln!("DEBUG: handle {} not found in registry", handle);
// Fallback: handle is an unboxed integer (Phase 131-10)
println!("{}", handle);
}
0

2
docs/development/current/main/01-JoinIR-Selfhost-INDEX.md
View File

@ -101,6 +101,8 @@ Phase 文書は歴史や検証ログも含むので、「JoinIR の現役設計
- → `docs/development/current/main/phase131-2-summary.md`（要点）
- LLVM（Python llvmlite）lowering の不具合切り分けで迷っているとき
- → `docs/development/current/main/phase131-3-llvm-lowering-inventory.md`（再現ケース表 + 根本原因候補）
- → `docs/development/current/main/phase131-5-taglink-fix-summary.md`（TAG-LINK: symbol 名の修正ログ）
- → `docs/development/current/main/phase131-6-ssa-dominance-diagnosis.md`（TAG-RUN の初期診断ログ・歴史）
- → `docs/development/current/main/phase87-selfhost-llvm-exe-line.md`（実行パイプラインのSSOT）
- 「この Phase 文書は現役か？」で迷ったとき
- → まず `docs/development/current/main/10-Now.md` と

4
docs/development/current/main/10-Now.md
View File

@ -73,9 +73,9 @@
の 3 層が `logging_policy.md` で整理済み。JoinIR/Loop trace も同ドキュメントに集約。
- VM backend の Box 解決（UnifiedBoxRegistry / BoxFactoryRegistry）の経路図:
- `docs/development/current/main/phase131-2-box-resolution-map.md`
- LLVM（Python llvmlite）lowering の棚卸し（Phase 131-3）:
- LLVM（Python llvmlite）lowering の棚卸し（Phase 131-3..10）:
- `docs/development/current/main/phase131-3-llvm-lowering-inventory.md`
- 現在の主要ブロッカー（要約）: PHI の配置/順序で LLVM IR が invalid になるケース、JoinIR ループパターンの未対応ケース
- 状態: Case B（Pattern1/loop_min_while）は EMIT/LINK/RUN まで復旧済み。残りは Case C の JoinIR ループパターン穴。
---

145
docs/development/current/main/phase131-3-llvm-lowering-inventory.md
View File

@ -8,7 +8,7 @@
| Case | File | Emit | Link | Run | Notes |
|------|------|------|------|-----|-------|
| A | `apps/tests/phase87_llvm_exe_min.hako` | ✅ | ✅ | ✅ | **PASS** - Simple return 42, no BoxCall, exit code verified |
| B | `apps/tests/loop_min_while.hako` | ✅ | ✅ | ❌ | **TAG-RUN** - EMIT/LINK fixed (Phase 131-5), **infinite loop in runtime** (PHI incoming value bug - Phase 131-6) |
| B | `apps/tests/loop_min_while.hako` | ✅ | ✅ | ✅ | **PASS** - Loop/PHI path runs end-to-end (Phase 131-10): prints `0,1,2` and exits |
| B2 | `/tmp/case_b_simple.hako` | ✅ | ✅ | ✅ | **PASS** - Simple print(42) without loop works |
| C | `apps/tests/llvm_stage3_loop_only.hako` | ❌ | - | - | **TAG-EMIT** - Complex loop (break/continue) fails JoinIR pattern matching |
@ -137,7 +137,7 @@ declare i64 @nyash.console.log(i8*)
---
### 3. TAG-RUN: Loop Infinite Iteration (Case B) - ❌ CONFIRMED PHI BUG (Phase 131-6)
### 3. TAG-RUN: Loop Runtime Incorrect (Case B) - ✅ FIXED (Phase 131-10)
**File**: `apps/tests/loop_min_while.hako`
@ -150,51 +150,28 @@ $ ./target/release/hakorune apps/tests/loop_min_while.hako
RC: 0
```
**Actual Behavior** (LLVM):
```bash
$ /tmp/loop_min_while
0
0
0
... (infinite loop, prints 0 forever)
```
**Outcome** (Phase 131-10):
- LLVM AOT now matches VM behavior (prints `0,1,2`, terminates)
- No regression in Case A / B2
**Root Cause** (Phase 131-6 Diagnosis - 2025-12-14):
**What happened (timeline)**:
- **Phase 131-6**: Confirmed MIR/VM correctness; investigated as “PHI incoming wiring bug”.
- Historical diagnosis: `docs/development/current/main/phase131-6-ssa-dominance-diagnosis.md`
- Historical plan: `docs/development/current/main/phase131-6-next-steps.md`
- **Phase 131-7**: Multi-pass lowering follow-up: ensure values defined in Pass A are visible to Pass C (global `vmap` sync).
- **Phase 131-8**: ExternCall argument resolution fixed: use PHI-safe resolution (`resolve_i64_strict`) so PHI values are not treated as null pointers.
- **Phase 131-9**: MIR global PHI type inference: fix loop-carrier PHI mistakenly inferred as `String` (prevent `"0" + "1"` concatenation behavior).
- Implemented in `src/mir/builder/lifecycle.rs` (Phase 131-9 section).
- **Phase 131-10**: Console ABI routing: avoid calling `nyash.console.log(i8*)` with non-string values.
- Route string literals to `nyash.console.log(i8*)`
- Route handles/integers to `nyash.console.log_handle(i64)`
- Unboxed integer fallback handled in `crates/nyash_kernel/src/plugin/console.rs`
The MIR is **correct**. The MIR dump shows proper SSA form:
```mir
bb4:
%3 = phi [%2, bb0], [%12, bb7] // ← Should receive %12 from loop body
bb7:
extern_call env.console.log(%3) // ← Prints current value
%11 = const 1
%12 = %3 Add %11 // ← Computes i+1
br label bb4 // ← Loops back with %12
```
**Verification**:
- ✅ VM execution: Correctly outputs `0, 1, 2`
- ❌ LLVM execution: Infinite loop outputting `0`
- ✅ MIR SSA dominance: No use-before-def violations
- ❌ LLVM PHI wiring: Incoming value from bb7 not properly connected
**Bug Location**: `/home/tomoaki/git/hakorune-selfhost/src/llvm_py/llvm_builder.py`
- Function: `finalize_phis()` (lines 601-735+)
- Suspected issue: Self-carry logic (lines 679-681) or value resolution (line 683) causes %12 to be ignored
**Detailed Analysis**: See [`docs/development/current/main/phase131-6-ssa-dominance-diagnosis.md`](/home/tomoaki/git/hakorune-selfhost/docs/development/current/main/phase131-6-ssa-dominance-diagnosis.md)
**Next Steps**:
1. Inspect generated LLVM IR for store instructions after PHI
2. Check if PHI value is being used in subsequent stores
3. Verify loop increment instruction sequence
**Files to investigate**:
- `src/llvm_py/instructions/store.py` - Store instruction lowering
- `src/llvm_py/phi_wiring/wiring.py` - PHI value propagation
- `target/aot_objects/loop_min_while.ll` - Generated LLVM IR (if saved)
**Key lesson**:
- “PHI is correct but loop prints 0 forever” can be a combination of:
- multi-pass value propagation bugs,
- ExternCall resolution/ABI mismatch,
- and MIR type inference drift.
---
@ -278,46 +255,20 @@ Hint: This loop pattern is not supported. All loops must use JoinIR lowering.
---
### 🔥 Priority 3: Fix TAG-RUN (Loop Infinite Iteration) - IN PROGRESS (Phase 131-6)
### ✅ Priority 3: COMPLETED - Fix TAG-RUN (Loop Runtime Correctness)
**Target**: Case B (`loop_min_while.hako`)
**Issue**: Loop counter not updating, causes infinite loop printing `0`
**Status**: ✅ FIXED (Phase 131-10)
**Phase 131-6 Investigation Results**:
**Fixes applied (high level)**:
1. Multi-pass lowering value propagation: Pass A outputs are synced for Pass C usage
2. ExternCall value resolution: unify on PHI-safe resolver (`resolve_i64_strict`)
3. MIR global PHI type inference: correct loop-carrier PHI types from incomings
4. Console ABI routing: `console.log(i8*)` vs `console.log_handle(i64)` dispatch
#### Bug #1: MIR Copy-to-PHI (FIXED)
- **Location**: `src/mir/builder/control_flow/joinir/merge/instruction_rewriter.rs` lines 419-440
- **Problem**: Parameter binding was generating `Copy { dst: PHI_dst, src: value }` in loop latch
- **Fix**: Added check to skip Copy when `dst` is a header PHI destination
- **Status**: ✅ Copy instruction removed from block 7
#### Bug #2: Type Inference - String vs Integer (PARTIAL FIX)
- **Location**: `src/llvm_py/instructions/binop.py` lines 128-153
- **Problem**: MIR marks `i + 1` as `dst_type: StringBox` (forward-looking hint from `print(i)` usage)
- **Impact**: Python harness was doing string concatenation instead of integer addition
- **Fix Attempted**: Removed `force_string` logic that trusted `dst_type` hint
- **Status**: ⚠️ Partially fixed but infinite loop persists
#### Bug #3: Instruction Ordering Violation (DISCOVERED)
- **Location**: MIR builder (instruction scheduling)
- **Problem**: Copy instructions emitted AFTER values are used (violates SSA dominance)
- **Example**: `%6 = %4 + %5` appears before `%5 = copy %3`
- **Impact**: LLVM requires strict SSA form, Rust VM tolerates it
- **Status**: ❌ Not yet addressed
#### Additional Findings
- Simple test `/tmp/test_simple_add.hako` (`i=0; i=i+1; print(i)`) also fails (prints 0 not 1)
- Issue exists even without loops, suggesting fundamental binop/type problem
- String tagging propagation may be marking integer PHI values as strings
**Next Steps**:
1. Trace `resolver.is_stringish()` to see if integers are being marked as strings
2. Fix MIR instruction scheduling to respect SSA dominance
3. Consider runtime type checking instead of compile-time inference
**Files Modified**:
- `src/mir/builder/control_flow/joinir/merge/instruction_rewriter.rs` (Phase 131-6 fix)
- `src/llvm_py/instructions/binop.py` (Phase 131-6 partial fix)
**Result**:
- Case B now passes EMIT ✅ LINK ✅ RUN ✅
- Output matches VM behavior (`0,1,2`), and the program terminates
---
@ -458,7 +409,7 @@ NYASH_LLVM_OBJ_OUT="$OBJ" NYASH_LLVM_USE_HARNESS=1 \
---
## Timeline Estimate
## Timeline Estimate (historical)
- **P1 (Loop PHI → LLVM IR fix)**: 1-2 hours (harness BB emission logic)
- **P2 (JoinIR pattern coverage)**: 3-4 hours (pattern design + implementation)
@ -468,19 +419,18 @@ NYASH_LLVM_OBJ_OUT="$OBJ" NYASH_LLVM_USE_HARNESS=1 \
---
## Executive Summary
## Executive Summary (updated)
### Phase 131-5 Results (TAG-LINK Fix Complete!)
### Phase 131-10 Results (Case B End-to-End PASS)
**✅ Case A (Minimal)**: PASS - Simple return works perfectly
- EMIT ✅ LINK ✅ RUN ✅
- Validates: Build pipeline, NyKernel runtime, basic MIR→LLVM lowering
**⚠️ Case B (Loop+PHI)**: EMIT ✅ LINK ✅ RUN ❌
**✅ Case B (Loop+PHI)**: PASS - Loop/PHI works in LLVM AOT
- **Phase 131-4**: Fixed TAG-EMIT (PHI after terminator) ✅
- **Phase 131-5**: Fixed TAG-LINK (symbol name mismatch) ✅
- **NEW ISSUE**: TAG-RUN (infinite loop - counter not updating) ❌
- **Progress**: 2/3 milestones achieved, runtime bug discovered
- **Phase 131-10**: Fixed TAG-RUN (value propagation + console ABI routing) ✅
**✅ Case B2 (BoxCall)**: PASS - print() without loops works
- EMIT ✅ LINK ✅ RUN ✅
@ -492,7 +442,7 @@ NYASH_LLVM_OBJ_OUT="$OBJ" NYASH_LLVM_USE_HARNESS=1 \
---
### Phase 131-5 Achievements
### Phase 131-5 Achievements (still valid)
**✅ Fixed TAG-LINK (Symbol Name Mismatch)**:
1. **Investigation**: Used `objdump` to discover NyKernel exports symbols with dots
@ -512,8 +462,8 @@ NYASH_LLVM_OBJ_OUT="$OBJ" NYASH_LLVM_USE_HARNESS=1 \
1. ✅ **Fix PHI ordering** (P1 - Phase 131-4) - DONE
2. ✅ **Fix symbol mapping** (P2 - Phase 131-5) - DONE
3. 🔥 **Fix loop runtime bug** (P3 - NEW) - IN PROGRESS
4. ⏳ **Add JoinIR Pattern 5** (P4) - PENDING
3. ✅ **Fix loop runtime correctness** (P3 - Phase 131-10) - DONE
4. ⏳ **Add JoinIR infinite-loop early-exit pattern** (P4) - PENDING
5. ⏳ **Comprehensive test** (P5) - PENDING
**Total Effort So Far**: ~3 hours (Investigation + 2 fixes)
@ -534,7 +484,7 @@ NYASH_LLVM_OBJ_OUT="$OBJ" NYASH_LLVM_USE_HARNESS=1 \
- **Recommendation**: Box-ify with interface contract (MIR JSON v1 schema)
#### 🔧 Technical Debt Found
1. **PHI emission ordering** - Architectural issue, not a quick fix
1. **Harness duplication** - Python harness vs Rust crate divergence risk
2. **Unreachable block handling** - MIR JSON marks all blocks `reachable: false` (may be stale metadata)
3. **Error logging** - Python harness errors lost after build_llvm.sh exits
@ -549,10 +499,11 @@ tmp/case_a
echo $? # Expected: 42
```
### Case B (Loop PHI - FAIL at EMIT)
### Case B (Loop PHI - PASS)
```bash
tools/build_llvm.sh apps/tests/loop_min_while.hako -o tmp/case_b
# Error: empty bb4 in LLVM IR
tmp/case_b
# Output: 0,1,2 (+ Result line), then exit
```
### Case B2 (Simple BoxCall - PASS)
@ -567,7 +518,7 @@ static box Main {
EOF
tools/build_llvm.sh /tmp/case_b_simple.hako -o tmp/case_b2
tmp/case_b2
# Output: (empty, but executes without crash)
# Output: prints `42` (+ Result line), then exit
```
### Case C (Complex Loop - FAIL at MIR)
@ -594,12 +545,12 @@ jq '.cfg.functions[] | select(.name=="main") | .blocks[] | select(.id==4)' /tmp/
## Success Criteria
**Phase 131-5 Complete** when:
**Phase 131-10 Complete** when:
1. ✅ Case A continues to pass (regression prevention) - **VERIFIED**
2. ⚠️ Case B (loop_min_while.hako) compiles to valid LLVM IR and links - **PARTIAL** (EMIT ✅ LINK ✅ RUN ❌)
2. ✅ Case B (loop_min_while.hako) passes end-to-end - **VERIFIED** (EMIT ✅ LINK ✅ RUN ✅)
3. ✅ Case B2 continues to pass (BoxCall regression prevention) - **VERIFIED**
4. ❌ Case C (llvm_stage3_loop_only.hako) lowers to JoinIR and runs - **NOT YET**
5. ⚠️ All 4 cases produce correct output - **PARTIAL** (2/4 passing)
5. ⚠️ All 4 cases produce correct output - **PARTIAL** (3/4 passing)
6. ⚠️ No plugin errors (or plugin errors are benign/documented) - **ACCEPTABLE** (plugin errors don't affect AOT execution)
**Definition of Done**:

7
docs/development/current/main/phase131-6-next-steps.md
View File

@ -1,5 +1,12 @@
# Phase 131-6: Next Steps - PHI Bug Fix
## Update (Phase 131-10)
Phase 131-6 時点では PHI wiring を主因として疑っていたが、実際には multi-pass lowering の値伝播・ExternCall の引数解決/ABI ルーティング・PHI 型推論など複数の要因が重なっていた。
最終的に Case B は解決済み（LLVM AOT で `0,1,2` を出して終了）。到達点と修正の全体像は次を SSOT とする：
- `docs/development/current/main/phase131-3-llvm-lowering-inventory.md`
## Summary
**Problem**: LLVM backend generates infinite loop for `loop_min_while.hako`

11
docs/development/current/main/phase131-6-ssa-dominance-diagnosis.md
View File

@ -1,10 +1,17 @@
# Phase 131-6: MIR SSA Dominance Diagnosis
## Update (Phase 131-10)
Case B (`apps/tests/loop_min_while.hako`) は **LLVM AOT でも `0,1,2` を出して終了**するところまで復旧済み。
この文書は Phase 131-6 時点の「切り分けログ（歴史）」として残し、最終的な到達点と修正の全体像は次を SSOT とする：
- `docs/development/current/main/phase131-3-llvm-lowering-inventory.md`
## Executive Summary
**Status**: ❌ LLVM Backend Bug Confirmed
**Status (Phase 131-6 時点)**: ❌ LLVM Backend Bug Confirmed
**Severity**: P0 - Breaks basic loop functionality
**Root Cause**: PHI node incoming values not properly wired in LLVM IR generation
**Root Cause (Phase 131-6 時点の仮説)**: PHI node incoming values not properly wired in LLVM IR generation
## Evidence Chain

26
src/llvm_py/instructions/externcall.py
View File

@ -80,10 +80,30 @@ def lower_externcall(
# Many call sites pass handles or pointers; we coerce below.
}
# Phase 131-10: Smart console.log routing
# If console.log is called with a non-string-literal argument, use console.log_handle instead
actual_c_symbol = c_symbol_name
use_handle_variant = False
if llvm_name in ("nyash.console.log", "nyash.console.warn", "nyash.console.error"):
# Check if first argument is a string literal
if len(args) > 0:
first_arg = args[0]
is_string_literal = False
if resolver is not None and hasattr(resolver, 'string_ptrs'):
is_string_literal = first_arg in resolver.string_ptrs
# If NOT a string literal, use the _handle variant
if not is_string_literal:
use_handle_variant = True
# Extract method name (log, warn, error)
method = llvm_name.split('.')[-1]
actual_c_symbol = f"nyash.console.{method}_handle"
# Find or declare function with appropriate prototype
func = None
for f in module.functions:
if f.name == c_symbol_name:
if f.name == actual_c_symbol:
func = f
break
if not func:
@ -91,6 +111,10 @@ def lower_externcall(
ret_ty, arg_tys = sig_map[llvm_name]
fnty = ir.FunctionType(ret_ty, arg_tys)
func = ir.Function(module, fnty, name=c_symbol_name)
elif use_handle_variant:
# console.*_handle: (i64) -> i64
fnty = ir.FunctionType(i64, [i64])
func = ir.Function(module, fnty, name=actual_c_symbol)
elif llvm_name.startswith("nyash.console."):
# console.*: (i8*) -> i64
fnty = ir.FunctionType(i64, [i8p])