Description Wes Isberg 2003-09-06 19:18:07 EDT

The programming guide says, 

  It is an error to try to put after returning advice on a join point 
  that does not return the correct type. For example,

    after() returning (byte b): call(int String.length()) {
        // this is an error
    }

  is not allowed. But if no return value is exposed, or the exposed 
  return value is typed to Object, then it may be applied to any 
  join point. 

The compiler emits no such error.  See

  tests/bugs/IncompatibleAfterReturningTypeCE.java

Perhaps this is the time/place to confirm and (plan to) document the correct
and/or actual behavior.  Here's what I think (users would think) is correct for
the form

  after() returning (Type t) : pc() { ... }

1) When a join point picked out by pc() is declared to return Type or a subtype
of Type, then the advice should run at the join point, even if the actual value
returned is null.  This differs from the case of args(..), et al.

2) Otherwise, when the join point is not declared to return a supertype of Type,
the compiler should emit an error stating that the join point is declared to
return a type incompatible with Type.

3) Otherwise, when the join point is declared to return a supertype of Type,
there are two options: 
(a) The advice should run if the actual return value is an instance of Type. 
This implicitly adds to the pointcut a dynamic test for the actual value of the
result.  If possible, we should write XLint messages for each such join point
shadow.  
(b) The compiler should emit an error, since (a) violates the principle that
(only) pointcuts pick out join points and might be confusing to users in cases
where the return value is null.  The user can fix the code by using a common
supertype and then doing an instanceof check.