Testing Guide

Config Race Condition Tests

Quick Test

Run the automated Go test to verify the fix:

# Test that reproduces the race condition (without fix)
go test -v ./internal/config -run TestConfigFileRaceCondition

# Test that verifies atomic writes prevent race conditions
go test -v ./internal/config -run TestAtomicConfigWrite

Manual Testing

Run the shell script for a visual demonstration:

# Run with default 50 iterations
./scripts/test-config-race.sh

# Run with more iterations for better statistical analysis
./scripts/test-config-race.sh 200

The script will:

1. Test non-atomic writes (current implementation before fix)
2. Test atomic writes (fixed implementation)
3. Show comparison of failure rates

Expected Results

Before Fix (Non-Atomic):

• Failure rate: ~1-5% (varies by disk speed and system load)
• Errors: "Partial read" or "JSON parse error"
• Demonstrates the bug reported in GitHub issue #86

After Fix (Atomic):

• Failure rate: 0%
• All reads successful
• Core never sees corrupted config files

Understanding the Test

The test simulates the real-world scenario:

1. Config file is being written (web UI saving changes)
2. Core process starts and tries to read config
3. Without fix: Core sometimes reads partial file → crash
4. With fix: Core always sees complete file → success

CI Integration

Add to CI pipeline:

- name: Test config atomicity
  run: |
    go test -v ./internal/config -run TestAtomicConfigWrite
    if [ $? -ne 0 ]; then
      echo "❌ Atomic config write test failed!"
      exit 1
    fi

Benchmarking

Compare performance of non-atomic vs atomic writes:

go test -bench=BenchmarkConfigWriteRead ./internal/config

Expected: Similar performance (fsync is negligible for config writes)

Related Files

• internal/config/loader.go - Atomic write implementation
• internal/config/race_test.go - Automated tests
• scripts/test-config-race.sh - Manual test script
• RACE_CONDITION_FIX_SUMMARY.md - Detailed analysis and results