I have a large change to the CMP over JPA code. After a tiny bit of
work with the tck, I found that we need to make a some changes to the
generation strategy for CMP beans. Specifically, the TCK likes to
reuse CMP bean implementation classes for several deployments and JPA
uses the implementation class to differentiate entities. Of course,
the two usages are in direct conflict, so when we detect reuse of
bean implementation classes for CMP1, we generate an empty subclass.
This kind of sucks, but most apps don't do this.
Unfortunately, this change caused a bunch of bugs in OpenJPA to
surface, so instead of waiting for fixes, I coded around these. Also
the generation code tends to effect lots of other stuff and I got a
bit excited and fixed more stuff, so this little fix has ballooned
into a major patch.
In the future, I'll try to keep my patches small and readable.
I hope to have it committed tomorrow,
-dain
