Re: Anyone understand method invocation in Java?

From:

Andreas Leitgeb <avl@gamma.logic.tuwien.ac.at>

Newsgroups:

comp.lang.java.programmer

Date:

24 Sep 2008 09:50:37 GMT

Message-ID:

<slrngdk3bd.4jb.avl@gamma.logic.tuwien.ac.at>

Joshua Cranmer <Pidgeot18@verizon.invalid> wrote:

Andreas Leitgeb wrote:

Scott <blackhole@nowhere.com> wrote:

"15.12.2.5 Choosing the Most Specific Method
If exactly one of the maximally specific methods is not declared abstract, it
is the most specific method.

Why are abstract methods "discriminated"?

I think the intent here is to prefer the more specific methods over the
automatically-generated bridge methods in the process.

Now I'm even more confused. The only candidates for "automatically-
generated bridge methods" that come to my mind are those discussed
recently, which were marked as "volatile"...ahem... ACC_BRIDGE, but
these are by no means "abstract".

Note that trying to get this section to come into play is difficult
since Java does a lot to keep you from having two override-equivalent
method signatures.

My understanding was:

interface I1 {}
interface I2 {}
interface I3 {}

class C implements I1,I2,I3 {}

abstract class MyClass {
  abstract int foo(I1 i);
  abstract int foo(I2 i);
  int foo(I3 i) { i=null; } // just dummy code
  static void bar(C c) { foo(c); }
}

According to the spec, bar(C) would always call foo(I3), not
even considering foo(I1..2) beyond that paragraph.

In practise it's different:
T.java:11: reference to foo is ambiguous, both method foo(I2) in MyClass
and method foo(I3) in MyClass match
  static void bar(C c) { foo(c); }

No automatic picking after all? Still even considering I2 ?
Where did I misunderstand the spec?