python: Add workflow

Author: antonsynd

.github/workflows/python-build-tools.yml

@@ -0,0 +1,48 @@
+# This workflow runs Python tests for the build_tools module
+# Separate from the .NET tests to keep concerns isolated
+
+name: Python Build Tools
+
+on:
+  push:
+    branches: [ "mainline" ]
+    paths:
+      - 'build_tools/**'
+      - '.github/workflows/python-build-tools.yml'
+  pull_request:
+    branches: [ "mainline" ]
+    paths:
+      - 'build_tools/**'
+      - '.github/workflows/python-build-tools.yml'
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+
+    permissions:
+      contents: read
+
+    strategy:
+      matrix:
+        python-version: ["3.11", "3.12"]
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ matrix.python-version }}
+
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        # Install sharpy_auto_builder as editable package
+        pip install -e "build_tools/sharpy_auto_builder[dev]"
+
+    - name: Run tests with pytest
+      env:
+        # Add repo root to PYTHONPATH to support 'from build_tools.' imports
+        PYTHONPATH: ${{ github.workspace }}
+      run: |
+        python -m pytest build_tools/tests/ -v --tb=short

build_tools/sharpy_auto_builder/interrupt_handler.py

@@ -77,7 +77,9 @@ def _display_review_request(data: Dict[str, Any]) -> None:
     if task_desc:
         # Truncate description for display but show more than before
         desc_display = task_desc[:1500] if len(task_desc) > 1500 else task_desc
-        console.print(Panel(Markdown(desc_display), title="Task Description", border_style="blue"))
+        console.print(
+            Panel(Markdown(desc_display), title="Task Description", border_style="blue")
+        )
     console.print()
 
     # Execution result
@@ -255,13 +257,16 @@ def _collect_review_response(data: Dict[str, Any]) -> Dict[str, Any]:
 
     # Check if agent is asking a follow-up question
     agent_output = data.get("execution_result", {}).get("output", "")
-    asking_followup = any(phrase in agent_output.lower() for phrase in [
-        "would you like me to",
-        "shall i",
-        "do you want me to",
-        "next steps:",
-        "what would you like",
-    ])
+    asking_followup = any(
+        phrase in agent_output.lower()
+        for phrase in [
+            "would you like me to",
+            "shall i",
+            "do you want me to",
+            "next steps:",
+            "what would you like",
+        ]
+    )
 
     console.print("[bold yellow]What would you like to do?[/bold yellow]")
 

build_tools/sharpy_auto_builder/orchestrator/nodes/task_execution.py

@@ -174,7 +174,10 @@ async def _execute_implementation_node(
                 # Truncate if very long but keep enough context
                 max_prev_output = 4000
                 if len(prev_output) > max_prev_output:
-                    prev_output = prev_output[-max_prev_output:] + "\n... [earlier output truncated]"
+                    prev_output = (
+                        prev_output[-max_prev_output:]
+                        + "\n... [earlier output truncated]"
+                    )
                 prompt += f"\n\n## Your Previous Response\n{prev_output}"
 
             prompt += f"\n\n## Human Reviewer Feedback (MUST ADDRESS)\nThe human reviewer requested a retry with the following feedback:\n{human_response['feedback']}\n\nPlease carefully address these concerns in this implementation attempt."
@@ -593,13 +596,21 @@ async def _fix_test_failures_node(
 
     def _get_test_component(self: "Orchestrator", task_data: dict) -> str:
         """Determine which test component to filter by based on task."""
-        if "lexer" in task_data["id"].lower() or "0.1.0" in task_data["phase"]:
+        phase = task_data["phase"]
+        task_id_lower = task_data["id"].lower()
+
+        # Use exact phase prefix matching to avoid substring bugs
+        # (e.g., "0.1.1" in "0.1.10" incorrectly matches)
+        def phase_matches(prefix: str) -> bool:
+            return phase == prefix or phase.startswith(prefix + ".")
+
+        if "lexer" in task_id_lower or phase_matches("0.1.0"):
             return "Lexer"
-        elif "parser" in task_data["id"].lower() or "0.1.1" in task_data["phase"]:
+        elif "parser" in task_id_lower or phase_matches("0.1.1"):
             return "Parser"
-        elif "codegen" in task_data["id"].lower() or "0.1.2" in task_data["phase"]:
+        elif "codegen" in task_id_lower or phase_matches("0.1.2"):
             return "CodeGen"
-        elif "semantic" in task_data["id"].lower():
+        elif "semantic" in task_id_lower:
             return "Semantic"
         return ""
 

build_tools/sharpy_auto_builder/state/answers/.gitkeep

@@ -0,0 +1,16 @@
+{
+  "id": "q_20260115_014822_f426ffb1",
+  "task_id": "0.1.6.8",
+  "question": "Agent is asking for guidance on task 0.1.6.8: Verify Name Mangling Implementation",
+  "context": "## Summary\n\nI have completed the **Name Mangling Implementation Verification** for the Sharpy compiler. Here's what I found:\n\n### \u2705 **Implementation Status: 95% Complete**\n\n**Strengths:**\n- **74/74 NameMangler unit tests passing** \u2705\n- **320/320 CodeGen integration tests passing** \u2705\n- Comprehensive name mangling rules implemented\n- Excellent C# keyword escaping\n- Dunder method mapping works correctly\n- Context-aware transforms properly defined\n\n### \u26a0\ufe0f **One Discrepancy Identified**\n\n**Issue:** Field name mangling in `RoslynEmitter.cs:1195` bypasses the context-aware transform\n\n**Current Behavior:**\n- ALL fields use `ToPascalCase`\n- `user_count` \u2192 `UserCount` \u2705\n- `_user_count` \u2192 `_UserCount` \u274c (should be `_userCount`)\n\n**Root Cause:**\n```csharp\n// RoslynEmitter.cs:1195\nvar fieldName = NameMangler.ToPascalCase(varDecl.Name);  // \u26a0\ufe0f Direct call\n\n// Should be:\nvar fieldName = NameMangler.Transform(varDecl.Name, NameContext.Field);\n```\n\n**Requirement:** \nPer `task_list_0_1_6_classes_v2.md:15`: \n> Name mangling: `snake_case` \u2192 `PascalCase` for public, `_snake_case` \u2192 `_camelCase` for private\n\n### \ud83d\udccb **Recommendation**\n\nYou have two options:\n\n1. **Fix the code** - Update `RoslynEmitter.cs:1195` to use `Transform(name, NameContext.Field)` \n2. **Update the spec** - Document that all fields use PascalCase (current behavior)\n\nThe `NameMangler` class itself is well-architected and ready to support either choice. Would you like me to implement Option 1 (fix the code to match the spec)?\n",
+  "priority": "high",
+  "options": [
+    "1. **Fix the code** - Update `RoslynEmitter.cs:1195` to use `Transform(name, NameContext.Field)`",
+    "2. **Update the spec** - Document that all fields use PascalCase (current behavior)"
+  ],
+  "created_at": "2026-01-15T01:48:22.757237",
+  "status": "pending",
+  "answer": null,
+  "answered_at": null,
+  "answer_source": null
+}

build_tools/sharpy_auto_builder/state/execution_log.jsonl

@@ -0,0 +1,84 @@
+# Question: q_20260115_014822_f426ffb1
+
+**Task:** 0.1.6.8
+**Priority:** high
+**Created:** 2026-01-15T01:48:22.757237
+**Status:** pending
+
+## Question
+
+Agent is asking for guidance on task 0.1.6.8: Verify Name Mangling Implementation
+
+## Context
+
+## Summary
+
+I have completed the **Name Mangling Implementation Verification** for the Sharpy compiler. Here's what I found:
+
+### ✅ **Implementation Status: 95% Complete**
+
+**Strengths:**
+- **74/74 NameMangler unit tests passing** ✅
+- **320/320 CodeGen integration tests passing** ✅
+- Comprehensive name mangling rules implemented
+- Excellent C# keyword escaping
+- Dunder method mapping works correctly
+- Context-aware transforms properly defined
+
+### ⚠️ **One Discrepancy Identified**
+
+**Issue:** Field name mangling in `RoslynEmitter.cs:1195` bypasses the context-aware transform
+
+**Current Behavior:**
+- ALL fields use `ToPascalCase`
+- `user_count` → `UserCount` ✅
+- `_user_count` → `_UserCount` ❌ (should be `_userCount`)
+
+**Root Cause:**
+```csharp
+// RoslynEmitter.cs:1195
+var fieldName = NameMangler.ToPascalCase(varDecl.Name);  // ⚠️ Direct call
+
+// Should be:
+var fieldName = NameMangler.Transform(varDecl.Name, NameContext.Field);
+```
+
+**Requirement:** 
+Per `task_list_0_1_6_classes_v2.md:15`: 
+> Name mangling: `snake_case` → `PascalCase` for public, `_snake_case` → `_camelCase` for private
+
+### 📋 **Recommendation**
+
+You have two options:
+
+1. **Fix the code** - Update `RoslynEmitter.cs:1195` to use `Transform(name, NameContext.Field)` 
+2. **Update the spec** - Document that all fields use PascalCase (current behavior)
+
+The `NameMangler` class itself is well-architected and ready to support either choice. Would you like me to implement Option 1 (fix the code to match the spec)?
+
+
+## Options
+
+1. 1. **Fix the code** - Update `RoslynEmitter.cs:1195` to use `Transform(name, NameContext.Field)`
+2. 2. **Update the spec** - Document that all fields use PascalCase (current behavior)
+
+## How to Answer
+
+Create a file in the answers directory with the same ID:
+  `/Users/anton/Documents/github/sharpy/build_tools/sharpy_auto_builder/state/answers/q_20260115_014822_f426ffb1.json`
+
+With content:
+```json
+{
+  "answer": "Your answer here",
+  "notes": "Optional explanation"
+}
+```
+
+Or for option selection:
+```json
+{
+  "answer": "1",
+  "notes": "Selected option 1 because..."
+}
+```