harshahosur81/root-cause-tracing

Technique: Root Cause Tracing

Concept: Bugs are rarely where they explode. The explosion (Stack Trace) is just where the bad data finally violated a constraint.

The Protocol

Do not fix the line in the stack trace yet. Follow the data upstream.

Step 1: Identify the "Contraband"

What specific value caused the crash?

• Example: A null user ID.
• Example: An empty string "" where a date was expected.
• Example: A value of -1 in a price field.

Step 2: The Upstream Walk (The "5 Whys")

Open your IDE. Use "Find Usages" or "Call Hierarchy."

1. Level 0 (Crash Site): processPayment(user.id) -> Crashed because user.id is null.
   • Question: Who called processPayment?
2. Level 1: CheckoutService.submitOrder called it.
   • Question: Where did CheckoutService get the user object?
3. Level 2: It was passed in from SessionManager.getCurrentUser().
   • Question: Why did SessionManager return a user object with a null ID?
4. Level 3: SessionManager hydrated it from the Redis Cache.
   • Question: Why is the data in Redis corrupt?
5. Level 4 (Root Cause): The LoginHandler wrote the user to Redis before the database generated the ID.

Step 3: Verify the Origin

• Hypothesis: The bug is in LoginHandler (Level 4), not processPayment (Level 0).
• Test: Fix the LoginHandler ordering.
• Result: The crash at Level 0 disappears because the data is now correct.

Common Traps

• The "Null Check" Trap: Adding if (user.id == null) return; at Level 0 fixes the crash, but it creates a "Zombie Order" (an order that fails silently). Do not do this.
• The "Sanitizer" Trap: Cleaning the string at Level 1 hides the fact that the database at Level 5 is corrupt.

Rule: Fix the data generation, not the data consumption.