but it turns out that there were hidden test failures due to someone
changing some code on me.
As I'm coming across some of these things that worked in 2.0.3 and fail
in 3.0, I'm wondering to what degree things are supposed to be backward
compatible. Take my run-in with passing null into
ExpressionFactory#likeIgnoreCaseExp. In 3.0, this results in an NPE
(that I planned to clean up), whereas in 2.0 it results in a properly
constructed Expression. This expression ends up returning no objects
for any query, but it works. Is it more wrong for the user to pass in
null or for the framework to hide an error by degrading gracefully? I
guess without knowing why it doesn't work in 3.0, I can't really say if
it's a bug or not.
Maybe the best thing to do is open up JIRA issues and let Andrus decide
whether or not they're actually bugs?
Servprise International, Inc.