Re: [abc] test run results

Quoting "Prof. Laurie HENDREN" <hendren@sable.mcgill.ca>:

> > Surely if we're aiming for ajc's "'implements' and 'extends' are
> > interchangeable in 'declare parents'", the thing to do would be to
> > change <class> implements <class> into <class> extends <class> internally?
> >
> > Or does that have an associated set of pitfalls?
> >
> > - P
> >
>
> That's a good point .... can we do that?

That is exactly what we do at the moment. The distinction is just in a flag in
the (common) AST node. This flag is (in the present version) ignored by the
ParentDeclarer. It just looks at the right-hand side to see if it is a class or
a list of interfaces. Then it changes the AST node flag to indicate this, so
that later passes (aspect info harvesting in particular) will see the correct
kind.

If <class> implements <class> is actually used out there, I think I agree with
Pavel and Ganesh that we should follow the ajc behaviour. Test case 724 would
then be skipped as a faulty (deprecated?) test case.

-Aske
Received on Fri Sep 24 20:13:00 2004