Type annotations on inner type that is an array components

[Alex Buckley]

Hi Werner,

On 7/18/2018 4:49 PM, Werner Dietl wrote:
> class ArrayOfInner {
>    class Inner {}
>
>    @TA(1) ArrayOfInner. @TA(2) Inner oi;
>    @TA(3) ArrayOfInner. @TA(4) Inner [] oia;
>    @TA(5) Inner i;
>    @TA(6) Inner [] ia;
> }
>
>
> Compile with javac 8 and according to javap -v field ia has the annotation:
>
>    ArrayOfInner$Inner[] ia;
>      descriptor: [LArrayOfInner$Inner;
>      flags: (0x0000)
>      RuntimeVisibleTypeAnnotations:
>        0: #10(#11=I#21): FIELD, location=[ARRAY, INNER_TYPE]
>
> Compile with javac 9, 10, or a recent 11 build and ia has the annotation:
>
>    ArrayOfInner$Inner[] ia;
>      descriptor: [LArrayOfInner$Inner;
>      flags: (0x0000)
>      RuntimeVisibleTypeAnnotations:
>        0: #10(#11=I#21): FIELD, location=[ARRAY]
>
> Note the missing INNER_TYPE location.
>
> I would like to argue that the Java 8 behavior was correct.

Sounds right. Given source code that denotes the array type "Inner[]", 
the question is whether the component type of that array type is a 
nested type (ArrayOfInner.Inner) or not (just Inner). Logically, yes, 
it's a nested type, so the target_path item deserves an INNER_TYPE entry.

This feels like the kind of detail we've been round and round on in the 
past. That said, I don't see any mails that would have led to javac 
changing how it emits target_path between 8 and 9. The closest suspect 
is a compiler-dev thread from Jan/Feb 2016 -- "Bug in encoding type 
annotations for supertype_targets" -- which spoke about the target_path 
being omitted entirely for an annotated nested type. (JDK-8148504 was 
filed against 8, but it's still open, so it can't have caused any 
collateral damage for a field type's annotation.)

Alex