bolt-cpp-ml-cpp-e2e-test-gen

---

name: bolt-cpp-ml-cpp-e2e-test-gen description: Part of the bolt-cpp-ml skill: Generate comprehensive end-to-end unit tests for all functions in a C++ repository.

C++ E2E Test Generator (within bolt-cpp-ml)

Generate comprehensive E2E unit tests for every public function in a C++ repository. The process catalogs all headers, writes tests for each function, fixes compilation errors and bugs discovered along the way, stabilizes the test suite, and commits.

Workflow Overview

1. Clone & Analyze — Clone repo, map directory structure, identify test framework
2. Catalog Functions — Extract all public APIs from headers (run extract_signatures.py)
3. Fix Pre-existing Bugs — Fix header conflicts, API mismatches, segfaults found during analysis
4. Write Tests — Generate one comprehensive test file covering all modules
5. Compile & Fix — Iteratively fix compilation errors by checking actual API signatures
6. Stabilize — Run tests repeatedly, fix race conditions and intermittent failures
7. Report & Commit — Generate test report, commit, push

Phase 1: Clone & Analyze

gh repo clone <org>/<repo>
cd <repo>
find . -type f \( -name "*.hpp" -o -name "*.h" -o -name "*.cpp" \) \
  -not -path "*/vcpkg/*" -not -path "*/build/*" -not -path "*/.git/*" \
  -not -path "*/ggml/*" -not -path "*/.ccls-cache/*" | head -50

Identify:

• Test framework: Look for test/ directory, existing test macros (BOLT_TEST, TEST_F, GTEST, CATCH2)
• Build system: Check CMakeLists.txt or Makefile
• Existing test runner: Read test/test_runner.cpp or equivalent main()
• Library structure: Map include/ and src/ module directories

Phase 2: Catalog Functions

Run the signature extractor to get a JSON catalog:

python3 /home/ubuntu/skills/bolt-cpp-ml-cpp-e2e-test-gen/scripts/extract_signatures.py <repo_root> > /tmp/signatures.json

Review the summary, then read key headers manually to understand:

• Constructor signatures and required parameters
• Singleton patterns (getInstance())
• Return types (especially optional, shared_ptr, vector)
• Enum values used as parameters

Critical rule: Always grep the actual header for exact method signatures before writing any test. Never assume API names.

grep -n "methodName" include/path/to/header.hpp

Phase 3: Fix Pre-existing Bugs

Common bugs found during test creation — see references/common-bugs.md for detailed patterns:

• Duplicate struct definitions across headers (different fields, same name, same namespace)
• Double-free in RAII wrappers (unique_ptr deleter + manual release)
• Thread safety issues (deadlocks from nested lock acquisition)

Fix bugs in the source code directly. These fixes are valuable deliverables alongside the tests.

Phase 4: Write Tests

Create a single comprehensive test file: test/test_comprehensive_e2e.cpp

Structure

// 1. All includes
#include "bolt/bolt.hpp"
#include "bolt/test_framework.hpp"
// ... all module headers

// 2. Using declarations
using namespace bolt;

// 3. Tests grouped by module with section headers
// ============================================
// E2E_ModuleName: Test description
// ============================================
BOLT_TEST(E2E_ModuleName, FunctionName_Scenario) {
    // Arrange, Act, Assert
}

Test patterns by construct type

See references/test-patterns.md for complete patterns covering:

• Static utility functions
• Singleton classes (always reset state first)
• RAII / resource management (verify allocation + deallocation)
• Thread-spawning classes (add sleep_for after start/stop)
• Callback registration
• Template classes (test with int and double)
• Cross-module integration workflows

Key rules

1. Initialize all struct members — Uninitialized members cause validation exceptions
2. Reset singleton state at test start — Use forceReset(), clear(), etc.
3. Never assert is_running for short-lived thread tasks — They may halt before the check
4. Use batch cleanup instead of while(hasItems()) removeItem() — Avoids deadlocks
5. Add 50ms sleep after thread start/stop operations
6. Test error paths — Invalid input, empty containers, null pointers, out-of-range

Cross-module integration tests

Write 5-10 workflow tests that chain operations across modules:

• Editor workflow: tab → document → folding → dirty state → close
• Memory + logging: allocate → log → verify → deallocate
• Network: buffer → compress → decompress → verify
• Full session: tab → syntax highlight → cursor → find/replace → fold → close

Register in CMakeLists.txt

See templates/cmake_test_target.cmake for the template. Add the test executable target and link against all project libraries.

Phase 5: Compile & Fix

Iterative compilation loop:

cd build && cmake .. -DCMAKE_BUILD_TYPE=Debug
make <test_target> -j$(nproc) 2>&1 | tail -30

For each error:

1. Read the error message to identify the mismatched API
2. grep the actual header for the correct signature
3. Fix the test to match the real API
4. Rebuild

Common compilation fixes:

• Method name mismatch (createTab() → addTab())
• Field name mismatch (result.averageTime → result.averageDurationMs)
• Missing constructor arguments (check struct definitions)
• Template argument deduction failures

Phase 6: Stabilize

Run the test suite 5 times to detect intermittent failures:

for i in 1 2 3 4 5; do
  echo "Run $i:"
  LD_LIBRARY_PATH=build:build/lib timeout 120 build/test/<test_binary> 2>&1 \
    | grep -E "Passed:|Failed:"
done

For intermittent failures:

• Race conditions: Add sleep_for(50ms) after thread operations
• Uninitialized state: Zero-initialize all struct members
• Singleton pollution: Add reset calls between tests
• File system timing: Use unique temp paths per test

Target: 0 failures across 5 consecutive runs.

Phase 7: Report & Commit

Generate the test report:

LD_LIBRARY_PATH=build:build/lib ./build/test/<binary> > /tmp/test_output.txt 2>&1
python3 /home/ubuntu/skills/bolt-cpp-ml-cpp-e2e-test-gen/scripts/gen_test_report.py \
  /tmp/test_output.txt test/test_comprehensive_e2e.cpp -o test/TEST_REPORT.md

Commit with descriptive message listing all modules tested and bugs fixed:

git add test/ include/ src/  # include bug fixes
git commit -m "feat(test): add comprehensive E2E unit tests (N tests)

Modules: [list all modules]
Bug fixes: [list all bugs fixed]"
git push origin main

Deliverables Checklist

• test/test_comprehensive_e2e.cpp — All tests in one file
• test/TEST_REPORT.md — Summary with pass/fail counts per suite
• test/CMakeLists.txt — Updated with new test target
• Bug fixes in include/ and src/ — Documented in commit message
• 5 consecutive clean runs with 0 failures