Commit Graph

5 Commits

Author SHA1 Message Date
1a8a70cc99 refactor(joinir): Phase 286C-2.1 - 3-stage pipeline scaffolding (Scan/Plan/Apply)
- Create 3-stage pipeline architecture to solve borrow checker conflicts
- Stage 1 (ScanBox): Read-only scanning, identify rewrites (170 lines)
- Stage 2 (PlanBox): Pure transformation, generate new blocks (85 lines)
- Stage 3 (ApplyBox): Builder mutation only (75 lines)

New Files:
- scan_box.rs: Stage 1 - Read-only scan for RewritePlan
- plan_box.rs: Stage 2 - Transform RewritePlan → RewrittenBlocks
- apply_box.rs: Stage 3 - Apply RewrittenBlocks to MirBuilder
- C-2.1-pipeline-refactoring.md: Implementation guide

Data Structures:
- RewritePlan: Describes WHAT to rewrite (tail calls, returns, PHI, params)
- RewrittenBlocks: Contains HOW to rewrite (new blocks, replacements, inputs)

Benefits:
- Borrow checker safety: No overlapping mutable/immutable borrows
- Single responsibility: Each stage has one clear purpose
- Testability: Each stage can be unit tested independently
- Maintainability: Clear data flow, isolated changes

Status:
- Scaffolding:  Complete (structure defined, stub implementations)
- Integration:  Next step (refactor merge_and_rewrite to use 3 stages)

Build: cargo build --release 

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

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-25 03:05:44 +09:00
d81937e744 refactor(joinir): Phase 286C-2 - RewriteContext state consolidation
- Create RewriteContext struct to consolidate 11 scattered state variables
- Add 4 helper Boxes: InstructionFilterBox, ParameterBindingBox, ReturnConverterBox, TailCallDetectorBox
- Reduce instruction_rewriter.rs cognitive overhead via centralized state API
- All state mutations now go through consistent RewriteContext methods

Changes:
- New: rewrite_context.rs (94 lines) - State consolidation SSOT
- New: 4 helper Box modules (instruction_filter, parameter_binding, return_converter, tail_call_detector)
- Modified: instruction_rewriter.rs (1454 → 1421 lines)
- Modified: rewriter/mod.rs (export RewriteContext)

Benefits:
- State tracking simplified (11 variables → 1 context object)
- Clearer intent via API methods (add_exit_phi_input vs push)
- Foundation for future refactoring (Phase 286C-2.1: 3-stage pipeline)
- Zero functionality changes (pure refactoring)

Build: cargo build --release 

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

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-25 02:58:36 +09:00
843d094598 refactor(joinir): Phase 286 P1-P3 - Boundary contract context enrichment
- P1: Add alloc_join_param()/alloc_join_local() API to JoinValueSpace
  - Prevents future API misuse (thin wrappers with explicit "JoinIR" context)
  - Updated docs with footnote-style number references

- P2: Enrich error context with host_fn for better diagnostics
  - Added context: &str parameter to verify_boundary_contract_at_creation()
  - Error format now shows: [merge_joinir_mir_blocks host=<fn> ...]

- P3: Add join-side info to error context (continuation count + boundary summary)
  - Uses boundary.continuation_func_ids.len() for join=
  - Adds [conts=X exits=Y conds=Z] suffix with fixed key names
  - Enables faster debugging with log-searchable format

Error format: [merge_joinir_mir_blocks host=X join=Y [conts=A exits=B conds=C]]

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

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-25 02:15:40 +09:00
6b64e6415c docs(phase-286): Add complete llvm.rs modularization instructions
Created comprehensive, self-contained instructions for modularizing
llvm.rs (449 lines → 10 boxes + orchestrator), following Phase 33
success pattern.

Contents:
- Background & current problems (12 responsibilities in 1 function)
- Phase 33 success case reference
- Detailed box design (10 boxes with responsibilities/inputs/outputs)
- Step-by-step implementation guide (staged migration)
- Acceptance criteria (build/test requirements)
- Important warnings (PyVM preservation, feature-gated code)

Box breakdown:
1. plugin_init.rs - Plugin initialization
2. using_resolver.rs - Using/prelude handling
3. mir_compiler.rs - AST → MIR compilation
4. method_id_injector.rs - method_id injection
5. joinir_experiment.rs - JoinIR experiment (feature-gated)
6. pyvm_executor.rs - PyVM harness (dev/test, MUST PRESERVE)
7. object_emitter.rs - LLVM object emit
8. harness_executor.rs - LLVM harness execution
9. exit_reporter.rs - Leak report (Phase 285LLVM-0)
10. fallback_executor.rs - Mock/legacy execution

Expected improvements:
- Code quality: 449 lines → 150-200 orchestrator + 10 focused boxes
- Testability: Each box independently testable
- Reusability: Boxes reusable across backends
- Maintainability: Single responsibility per box

CRITICAL: PyVM executor (pyvm_executor.rs) MUST be preserved
- Used by 8 JSON AST smoke tests
- NOT removable legacy code

Ready for handoff to other AI (ChatGPT, etc.) for implementation.

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

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-24 10:08:40 +09:00
661bbe1ab7 feat(phase284): P1 Complete - Return in Loop with Block Remap Fix
## Summary

Completed Phase 284 P1: Enable return statements in Pattern4/5 loops via
JoinInst::Ret infrastructure (100% pre-existing, no new infrastructure needed).

**Critical Bug Fix**: Block ID remap priority
- Fixed: local_block_map must take precedence over skipped_entry_redirects
- Root cause: Function-local block IDs can collide with global remap entries
  (example: loop_step:bb4 vs k_exit:bb4 after merge allocation)
- Impact: Conditional Jump else branches were incorrectly redirected to exit
- Solution: Check local_block_map FIRST, then skipped_entry_redirects

## Implementation

### New Files
- `src/mir/join_ir/lowering/return_collector.rs` - Return detection SSOT (top-level only, P1 scope)
- `apps/tests/phase284_p1_return_in_loop_min.hako` - Test fixture (exit code 7)
- Smoke test scripts (VM/LLVM)

### Modified Files
- `loop_with_continue_minimal.rs`: Return condition check + Jump generation
- `pattern4_with_continue.rs`: K_RETURN registration in continuation_funcs
- `canonical_names.rs`: K_RETURN constant
- `instruction_rewriter.rs`: Fixed Branch remap priority (P1 fix)
- `terminator.rs`: Fixed Jump/Branch remap priority (P1 fix)
- `conversion_pipeline.rs`: Return normalization support

## Testing

 VM: exit=7 PASS
 LLVM: exit=7 PASS
 Baseline: 46 PASS, 1 FAIL (pre-existing emit issue)
 Zero regression

## Design Notes

- JoinInst::Ret infrastructure was 100% complete before P1
- Bridge automatically converts JoinInst::Ret → MIR Return terminator
- Pattern4/5 now properly merge k_return as non-skippable continuation
- Correct semantics: true condition → return, false → continue loop

## Next Phase (P2+)

- Refactor: Block remap SSOT (block_remapper.rs)
- Refactor: Return jump emitter extraction
- Scope: Nested if/loop returns, multiple returns
- Design: Standardize early exit pattern (return/break/continue as Jump with cond)

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

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
2025-12-23 14:21:27 +09:00