I raised HARMONY-4392 for your comments on a non-bug difference in java.beans.Statement/Expression. Basically the issue is seems RI's implementation cannot handle the overloaded constructor properly, which just finds the first usable one and invoke, while spec requires the most specific method should be found and used. Harmony actually works in similar way with RI now, but because this kind of implementation depends on the order of constructors found by reflection, they show different result now. Further, even RI 5 and 6 shows difference. Please refer to JIRA for details.
My opinion is to fix Harmony following spec, and RI's behavior is almost impossible to follow even if we want :). How do you think? -- Paulex Yang China Software Development laboratory IBM
