A DESCRIPTION OF THE REQUEST :
This is from the text of JLS 3rd edition (proposed) 15.12.1:
 
"If the form is Primary . Identifier, then the name of the method is the Identifier. Let T be the type of the Primary expression; then the class or interface to be searched is T if T is a class or interface type, or the upper bound of T if T is a type variable."

This does not say what happens if T is a primitive type. We may guess that this means we don't *have* a type to search, and thus the end of 15.12.2.1 requires a compile-time error. And of course that's what happens when we compile. But the JLS is rather more vague than it needs to be on this point.

JUSTIFICATION :
This was a problem in previous editions of the JLS, however I believe it's more of an issue now thanks to autoboxing and unboxing. There are now many contexts in which a primitive may be treated as a reference type, and we need to be told explicitly that this is not one of them. Compare 15.11.1 where it's made quite clear that for field access, you can't dereference a primitive.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Add one sentence in 15.12.1 (the bullet point on Primary.Identifier invocation): "It is a compile-time error if T is a primitive type."

ACTUAL -
Vagueness. Confusion. Bad craziness.

---------- BEGIN SOURCE ----------
class Test {
    int x = Math.PI.hashCode();
}

// Obviously this results in a compile error, as it should.
// It's just the JLS explanation thats the problem
---------- END SOURCE ----------

CUSTOMER SUBMITTED WORKAROUND :
Learn how Java works from the compiler rather than from JLS.
###@###.### 2005-04-25 19:45:28 GMT