Queries about JDK-8035890

[Alex Buckley]

Hi Jan,

On 3/27/2014 4:09 PM, Jan Lahoda wrote:
> I have a few more questions on this. Considering these declarations of
> annotation types:
> @Target({ElementType.METHOD, ElementType.CONSTRUCTOR})
> @interface DA { }
> @Target(ElementType.TYPE_USE)
> @interface TA { }
>
> Do I understand correctly that these are supposed to be valid:
> public <T> @DA int foo1() { return 0;}
> public <T> @DA void foo2() { }
> public <T> @TA int foo3() { return 0;}

Yes. The @DA and @TA are part of the MethodDeclaration just like they'd 
be if <T> was not present. So, it reduces to the well-known situation of 
an annotation immediately preceding a type in a method declaration - 
where we have a mechanism for determining whether the annotation is a 
type annotation or a declaration annotation. In other words, the type 
parameter section is invisible for the purpose of ascribing meaning to 
annotations.

> These are supposed to be invalid:
> public class Foo {
>      <T> @DA Foo() { }
>      <T> @TA Foo(int i) { }
> }

Ah, constructor declarations. The EG never discussed allowing 
annotations after the type parameter section there. I guess that was an 
oversight, but as things stand your two ctors are invalid.

> I am unsure about:
> public <T> @TA void foo4() { return 0;}

Illegal. We remembered to handle this one:

"If an annotation appears before a void method declaration and is deemed 
to apply only to the type which is closest to the annotation, a 
compile-time error occurs."

Since TA is TYPE_USE, @TA is deemed to apply only to the type which is 
closest to the annotation - though it turns out to be 'void', so oops. 
All this would apply is @TA appeared before the 'public', since it's 
always been the case that an annotation and the type it means to apply 
to can be separated by other modifiers.

Alex