Files
hakorune/docs/development/current/main/phases/phase-275/P0-INSTRUCTIONS.md

148 lines
4.1 KiB
Markdown
Raw Normal View History

feat(llvm/phi): Phase 277 P1 - fail-fast validation for PHI strict mode ## Summary Implemented fail-fast validation for PHI ordering and value resolution in strict mode. ## Changes ### P1-1: Strict mode for "PHI after terminator" - File: `src/llvm_py/phi_wiring/wiring.py::ensure_phi` - Behavior: `NYASH_LLVM_PHI_STRICT=1` → RuntimeError if PHI created after terminator - Default: Warning only (no regression) ### P1-2: Strict mode for "fallback 0" - File: `src/llvm_py/phi_wiring/wiring.py::wire_incomings` - Behavior: Strict mode forbids silent fallback to 0 (2 locations) - Location 1: Unresolvable incoming value - Location 2: Type coercion failure - Error messages point to next debug file: `llvm_builder.py::_value_at_end_i64` ### P1-3: Connect verify_phi_ordering() to execution path - File: `src/llvm_py/builders/function_lower.py` - Behavior: Verify PHI ordering after all instructions emitted - Debug mode: Shows "✅ All N blocks have correct PHI ordering" - Strict mode: Raises RuntimeError with block list if violations found ## Testing ✅ Test 1: strict=OFF - passes without errors ✅ Test 2: strict=ON - passes without errors (no violations in test fixtures) ✅ Test 3: debug mode - verify_phi_ordering() connected and running ## Scope - LLVM harness (Python) changes only - No new environment variables (uses existing 3 from Phase 277 P2) - No JoinIR/Rust changes (root fix is Phase 279) - Default behavior unchanged (strict mode opt-in) ## Next Steps - Phase 278: Remove deprecated env var support - Phase 279: Root fix - unify "2本のコンパイラ" pipelines 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-22 14:48:37 +09:00
# Phase 275 P0 (impl): Coercion SSOT rollout
Status: planned / implementation guide
This is the “next instruction sheet” for implementing the accepted coercion SSOT (A1/B2/C2) across backends.
SSOT decisions:
- `docs/development/current/main/phases/phase-274/P3-DECISIONS.md`
---
## Scope (P0)
Implement and lock these three rule sets:
- truthiness: `Void`/`BoxRef` fail-fast (with bridge-box exceptions)
- equality: B2 (Number-only, precise Int↔Float)
- `+`: C2 (Number-only promotion; String+String only; String mixed → TypeError)
Backends in scope:
- Rust VM (primary SSOT)
- LLVM harness (llvmlite path) parity with Rust VM
Out of scope:
- adding new language features (keywords)
- expanding env var toggles
- rewriting the optimizer broadly (only touch what is required to enforce semantics)
---
## Step 0: Lock reference + plan
- Keep `docs/reference/language/types.md` as “current executable SSOT” until the implementation is complete.
- After implementation + tests land, update `types.md` to the new semantics (it becomes SSOT again).
---
## Step 1: Rust VM — truthiness (A1)
Target:
- `src/backend/abi_util.rs::to_bool_vm` (or equivalent truthiness entry)
Changes:
- `Void` in boolean context → return `TypeError`
- `BoxRef`:
- allow only explicit bridge boxes: BoolBox/IntegerBox/StringBox
- treat VoidBox as Void (→ TypeError)
- other BoxRef types → TypeError
Acceptance:
- A dedicated fixture demonstrates `if Void { ... }` is a runtime error (fail-fast).
---
## Step 2: Rust VM — equality (B2)
Target:
- `src/backend/abi_util.rs::eq_vm` (and any helpers it relies on)
Changes:
- Remove Bool↔Int coercion.
- Keep Int↔Float comparison, but make it precise:
- if Float is NaN → false
- if Float is integral and within i64 exact range → compare as Int exactly
- otherwise → false
- Mixed kinds (except Int↔Float) → false (not error).
- BoxRef equality stays identity.
Acceptance:
- Tests cover:
- `1 == 1.0` true
- `1 == 1.1` false
- `true == 1` false (or TypeError if you choose a transition rule; document explicitly)
---
## Step 3: Rust VM — `+` (C2)
Target:
- `src/backend/mir_interpreter/helpers.rs::eval_binop` for `BinaryOp::Add`
Changes:
- Numeric:
- Int+Int → Int
- Float+Float → Float
- Int+Float / Float+Int → Float (Int promoted to Float)
- String:
- String+String → concat
- String mixed → TypeError (no implicit stringify)
- Everything else → TypeError
Acceptance:
- Tests cover:
- `1 + 2.0``3.0`
- `"a" + "b"``"ab"`
- `"a" + 1` → TypeError
---
## Step 4: LLVM harness parity
Targets (likely):
- `src/llvm_py/instructions/binop.py` for `+`
- `src/llvm_py/instructions/compare.py` for `==`
- truthiness for branch conditions (inspect branch lowering):
- `src/llvm_py/instructions/controlflow/branch.py`
- and any “truthy” conversions used in control-flow lowering
Notes:
- LLVM harness uses MIR JSON metadata (`value_types`) for discriminating raw vs handle.
- Keep behavior identical to Rust VM; do not re-introduce “string mixed concat”.
Acceptance:
- VM and LLVM smokes use the same fixtures and produce identical exit codes.
---
## Step 5: Fixtures + smoke tests (SSOT lock)
Create minimal, self-contained fixtures under `apps/tests/` and add smokes under `tools/smokes/v2/profiles/integration/apps/`.
Suggested fixtures (names; adjust as needed):
- `apps/tests/phase275_p0_truthiness_void_error_min.hako`
- `apps/tests/phase275_p0_eq_number_only_min.hako`
- `apps/tests/phase275_p0_plus_number_only_min.hako`
Smoke targets:
- VM: `..._vm.sh`
- LLVM: `..._llvm.sh` (harness/EXE path consistent with current infra)
Rules:
- No reliance on hidden env toggles.
- If a test needs runtime-unknown values, avoid externs that arent supported on VM/LLVM.
---
## Step 6: Update SSOT docs
After the implementation is complete and tests pass:
- Update `docs/reference/language/types.md` to the new semantics:
- truthiness: Void/BoxRef fail-fast
- equality: B2
- `+`: C2
- Update Phase 274/275 status:
- `docs/development/current/main/phases/phase-274/README.md`
- `docs/development/current/main/10-Now.md`
- `docs/development/current/main/30-Backlog.md`