Message12450

This continues to provide interest. I can only guess at the reasoning that led to performing this test on the .modified attributes once C3 has failed. However, it seems questionable, in that:

1. A modified attribute may conflict with a method defined in Java (e.g. proxy method __repr__ added to Map does this, and many others).

2. Since https://hg.python.org/jython/rev/f3660ced0608#l3.131, we have been changing order Java provides, placing the superclass last in the bases. :/

I think the motive for 2. is to give to give proxy methods added to interfaces Map, List, etc. precedence over implementations provided from Java, but it makes what we're doing with C3 and the MRO even less likely to match Java. I'll experiment with reversing this in circumstances where all the conflicts are reported.