I should point out.. adding the additional named classes to the PAR also
leads to the test passing.
On Mon, Apr 28, 2014 at 4:56 PM, Steve Ebersole <steve(a)hibernate.org> wrote:
With my latest Scanner changes as outlined in the earlier email there
one failure in the HEM "packaged" tests that I simply cannot explain why it
was previously considered valid.
The test builds a PAR (via ShrinkWrap) and then tries to use it. The
persistence.xml simply points to a cfg.xml. The cfg.xml names 4 classes.
But the PAR only bundles 2 of them. The test fails because Binder cannot
find the other 2 classes. Should this be valid? Should the cfg.xml inside
a PAR be allowed to name classes not contained in the PAR? Obviously if
the test is allowed to pick up all classes on the "real test classpath" the
test will pass.