On Feb 10, 2010, at 11:35 PM, Kevin Menard wrote:
If your analysis is correct, the patch at the very least should have
caused a failure somewhere in the test
suite.
The test suite is not ideal. In fact it barely touches runtime
relationships, working with bi-directional relationship graphs in most
cases. We need to improve it, no question.
I still really don't feel like I mapped my STI relationship
incorrectly. It's the most natural way to do it.
The way you described it in this thread, it is indeed. And the thing
is - in my tests this type of mapping works without CAY-1009 patch. So
I wonder if the the actual failing mapping has some details not
mentioned here. I'd appreciate if you dig up that failing mapping, so
that we could inspect it.
For now I am going to commit my fix, reverting CAY-1009 + some minor
refactoring + CAY-1378 test case. Since the controversial piece is
just 3 lines of code in ObjRelationship, it should be easy to redo it
one more time if we find a reason to do so.
Andrus
