Problem parsing the com.sun.java.swing.plaf.gtk.SynthButtonListener class

Most likely because it's an inner-class; inner-class resolution is slightly broken right now (see QDOX-29).

Patch attached to qdox-36 may help you.

This is most likely not parser related.

It is most likely classloader related because of some '.' not being correctly converted to '$' before attempting to load the inner class/interface.

Type.getJavaClass() (or some of the methods it calls) needs more information.

Which brings up a more general question about naming standards for inner classes.

Consider:

package a;
class B {
class C {}
class D extends C {}
}

Which one of these should pass?

a) assertTrue( d.isA("a.B.C") );
b) assertTrue( d.isA("a.B$C") );

In my opinion, a) should fail and b) should pass. The proper class name of inner classes does use '$' instead of '.' (It is only in java source imports that the '$' is replaced with '.' - for some stupid reason that causes too much confusion).

This is most likely because the inner class' name contains '.' where there should actually be '$'.

We should avoid the use of '.' for inner classes everywhere. It is evil.

Sun decided to let developers use the dotted syntax instead of $ for referring to inner classes/interfaces in source code. This must have been to please the developers eyes and keep backward compatibility on the language (lexical) level. In fact, they just introduced a new source of confusion.)

I think QDox should expect inner classes to be named with $ everywhere.

And in the cases where QDOX fails to resolve a class.inner class/interface could it be possible to tell QDOX trying again with the last "." in the class name replaced by a "$".

That is to say when QDOX fails to resolve the LazyActionMap.Loader interface, it tries again to resolve LazyActionMap$Loader...

I'm sure it is not the best way to fix, but it may work, isn't it ?

Eric, I've fixed the inner-class type-resolution issue (QDOX-29).

However, I'm not entirely confident that this solves your problem. Were you actually trying to parse the Swing sources with QDox?

In any case, might you be able to (A) try again with the CVS HEAD version of QDox, and let us know how it goes, and/or (B) provide a minimal failing test-case (e.g. one or two source-files that demonstrate the problem).

Thanks.

Thanks Mike. Your correction solves my problem

As you guessed, I am actually writing a navigator for java sources (school project :/). QDOX behaves wonderfully

I think you can close this bug now.

Thanks Eric.