Spring AOP proxies — what is the self-invocation problem?

Deep dive

Spring AOP is usually **proxy-based**:

• JDK dynamic proxies when you proxy an interface.
• CGLIB proxies when you proxy a concrete class (subclassing).

The proxy intercepts calls **made through the proxy**. When code inside the bean calls another method on `this`, it never leaves the target object, so the proxy can’t intercept it.

Example (the issue)

@Service
class BillingService {
  @Transactional
  public void pay() { /* ... */ }

  public void run() {
    this.pay(); // bypasses proxy -> no transaction
  }
}

Fixes

• Extract the advised method to another bean and call that bean.
• Inject the proxied interface and call through it (be careful with design).
• Use AspectJ weaving only if you truly need to advise self-invocations.

Common pitfalls

• Annotating private methods with `@Transactional` and expecting it to work.
• Calling “helper” advised methods from inside the same class.
• Tests that instantiate classes directly (no proxy) and then expecting AOP behavior.