Fix Bug

Fix Bug: Test-First Bug Fixing

Resolve software defects using a disciplined test-first approach. Write a failing test that reproduces the bug BEFORE modifying any production code.

Instructions

Phase 1: Write a Failing Test (DO THIS FIRST)

1. Choose test level:

   • Unit test (isolated component/function)
   • Integration test (component interaction)
   • Acceptance test (end-to-end behavior)

2. Write test that reproduces the bug:

   # Python example
   def test_bug_description():
       """Bug: [Describe what's wrong]"""
       # Arrange: Set up scenario that triggers bug
       input_data = create_bug_triggering_input()
   
       # Act: Execute the failing code
       result = system_under_test.process(input_data)
   
       # Assert: What SHOULD happen (test will FAIL)
       assert result == expected_correct_behavior
   

   // Java example
   @Test
   @DisplayName("Bug: [Describe what's wrong]")
   void reproduceBug() {
       // Arrange: Set up scenario that triggers bug
       var input = createBugTriggeringInput();
   
       // Act: Execute the failing code
       var result = systemUnderTest.process(input);
   
       // Assert: What SHOULD happen (test will FAIL)
       assertThat(result).isEqualTo(expectedCorrectBehavior);
   }
   

3. Run the test - MUST FAIL:

   • If test passes, it doesn't reproduce the bug
   • Adjust test until it fails

4. Verify failure reason:

   • Change assertion to expect WRONG behavior
   • Test should now PASS
   • Revert assertion back to correct behavior
   • This proves test reproduces the actual bug

5. Commit the failing test:

   git add tests/...
   git commit -m "test: reproduce bug - [description]
   
   Failing test demonstrates [bug behavior].
   Expected: [correct behavior]
   Actual: [buggy behavior]
   
   Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>"
   

Phase 2: Fix the Code

1. Minimal change:

   • Use test as guide for debugging
   • Make smallest change to pass test
   • Avoid "while I'm here" refactoring

2. Run the test - MUST PASS:

   • Test transitions from RED → GREEN
   • This proves bug is fixed

3. Run ALL tests:

   • Ensure no regressions
   • All existing tests must still pass

4. Refactor if needed:

   • Improve code quality
   • Keep tests passing

Phase 3: Commit the Fix

git add src/...
git commit -m "fix: [description of bug fix]

Resolves issue where [bug behavior] occurred.
Root cause: [explanation]
Fix: [what was changed]

Test: tests/.../test_bug_reproduction.py

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>"

Golden Rules

❌ DO NOT:

• Make changes to production code before writing test
• Run full application to debug
• Create side-scripts to investigate
• Deploy and test manually
• Guess and check
• Skip writing test because "I know what's wrong"

✅ DO:

• Write failing test FIRST (no exceptions)
• Verify test fails for the RIGHT reason
• Make minimal change to pass test
• Run all tests before committing
• Commit test and fix together
• Document root cause in commit message

Why This Works

Traditional Approach (Anti-Pattern):

1. Guess where problem is
2. Change production code
3. Deploy and test manually
4. Still broken? Go to step 1
5. Eventually works (you think)
6. Hope you didn't break anything

Problems:

• ❌ No proof bug is fixed
• ❌ No proof nothing else broke
• ❌ Wastes time on deployments
• ❌ Bug can return unnoticed

Test-First Approach:

1. Write failing test (reproduces bug)
2. Fix production code
3. Test passes
4. Run all tests
5. Done - proven fixed

Benefits:

• ✅ Proof bug existed
• ✅ Proof bug is fixed
• ✅ Proof nothing else broke
• ✅ Fast feedback (no deployment)
• ✅ Regression prevention
• ✅ Documents the issue

Example Workflow

User: Bug report - API returns 500 when input contains special characters

Agent (Backend-Dev):
  Phase 1: Reproduce bug with test

  Created: tests/test_api_special_chars.py

  def test_api_handles_special_characters():
      \"\"\"Bug: API returns 500 for input with special chars\"\"\"
      client = TestClient(app)
      response = client.post("/api/process", json={
          "text": "Hello <world> & friends!"
      })

      # Should return 200, not 500
      assert response.status_code == 200
      assert response.json()["status"] == "success"

  Running test... ❌ FAILED
  AssertionError: assert 500 == 200

  Good! Test reproduces the bug.

  Committed: git commit -m "test: reproduce API 500 with special chars"

  Phase 2: Fix the code

  Root cause: Input sanitization missing for HTML entities

  Fixed: src/api/processor.py
  - Added html.escape() before processing input

  Running test... ✅ PASSED
  Running all tests... ✅ 127/127 PASSED

  Phase 3: Commit fix

  git commit -m "fix: sanitize HTML entities in API input

  Resolves 500 error when input contains <, >, &, etc.
  Root cause: Missing input sanitization
  Fix: Added html.escape() in processor

  Test: tests/test_api_special_chars.py

  Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>"

  ✅ Bug fixed with proof (test passes)
  ✅ Regression prevented (test stays in suite)

Test Level Selection Guide

Unit Test - Choose when:

• Bug in single function/method
• Isolated component behavior wrong
• Logic error in calculation
• Example: "Function returns wrong value for edge case"

Integration Test - Choose when:

• Bug in component interaction
• Data flow between modules broken
• API contract violated
• Example: "Service A sends wrong format to Service B"

Acceptance Test - Choose when:

• End-to-end user flow broken
• Full feature behaves incorrectly
• System-level behavior wrong
• Example: "User cannot complete checkout process"

Start with smallest test that reproduces bug (usually unit test).

Common Pitfalls

Pitfall 1: "I'll write test after fixing"

• ❌ Never works - you forget or run out of time
• ✅ Test FIRST ensures it gets written

Pitfall 2: "Test passes but bug still exists"

• ❌ Test doesn't reproduce actual bug
• ✅ Verify test fails for RIGHT reason (change assertion to wrong value, should pass)

Pitfall 3: "Too hard to write test"

• ❌ Usually means code is hard to test (design smell)
• ✅ Use this as opportunity to improve testability

Pitfall 4: "I fixed it manually, works now"

• ❌ No proof, bug can return
• ✅ Capture fix in test for regression prevention

Related Skills

• /iterate - Use after fixing bug to continue with next batch
• /review - Request architect review for complex bug fixes

References

• Directive 028: .github/agents/directives/028_bugfixing_techniques.md
• Approach: .github/agents/approaches/test-first-bug-fixing.md
• Checklist: .github/agents/approaches/bug-fixing-checklist.md
• Directive 017: .github/agents/directives/017_test_driven_development.md (TDD)