Description George Harley 2003-04-23 04:44:22 EDT

I am using AspectJ 1.1rc1. Here is my simple test case...

public class Inner1
{}

public class Inner2 extends Inner1
{}

public interface NewInterface
{}

public class Outer extends Inner2 implements NewInterface
{
  public static void main(String[] args)
  {
    Outer outer = new Outer();
  }
}

public aspect InitAspect
{
  pointcut outerMatch() : initialization(new(..));
  before() : outerMatch()
  {
    System.out.println(thisJoinPoint.toLongString());
  }
}

As I understand it, the outerMatch() pointcut will match all constructors in the 
system. The above all compiles nicely but when I run Outer I get the following 
(please ignore line numbers) ...
java.lang.ExceptionInInitializerError: org.aspectj.lang.NoAspectBoundException
  at InitAspect.aspectOf(InitAspect.java)
  at InitAspect.<init>(InitAspect.java:14)
  at InitAspect.ajc$postClinit(InitAspect.java)
  at InitAspect.<clinit>(InitAspect.java:14)
  at Inner1.<init>(Inner1.java:14)
  at Inner2.<init>(Inner2.java:14)
  at Outer.<init>(Outer.java:14)
  at Outer.main(Outer.java:18)
  Exception in thread "main" 

If I compile the above system with AspectJ 1.0 before running Outer then it all 
works and I see the following output ...
initialization(public Inner1())
initialization(public Inner2())
initialization(public NewInterface())
initialization(public Outer())

It would appear that the pointcut is matching the InitAspect itself. The only 
way I can get my simple test case to work with 1.1rc1 is to modify my 
outerMatch() pointcut to explicitly remove the aspect from the scope ...

pointcut outerMatch() : initialization(new(..)) && !within(InitAspect);

When I run this I get the following output (note the subtle change in the order 
of the matches compared to what I got using AspectJ 1.0 - now the interface 
match comes after the Outer match)....
initialization(public Inner1())
initialization(public Inner2())
initialization(public Outer())
initialization(public NewInterface())

Is this behaviour a 1.1 bug or was 1.0 behaving incorrectly ?