On Aug 19, 2009, at 8:19 AM, mikeho wrote:
Hi Daivid,
Firstly thanks for the reply.
My aim is to use the Fit framework and embedded OpenEJB to test an
application.
I intend using fit fixtures to set up test data and fit fixtures to
invoke
the various session beans.
So in my test setup I beginTx() write a row to the db and commitTx()
I then insert another entity (for the same table) via the SSB (default
transaction settings ie CMP tx.required).
I then invoke a method on the SSB which should return a list of all
the
entities in the table. This is the result list I refer to earlier.
This list
only contains the rows I commited in the setup. The entity inserted
via the
SSB is not in the list.
Ok, I see. If you can create a small test case for this, I'll be
happy to take a look.
A side note on acronyms, I think you mean CMT (Container-Managed
Transactions) instead of CMP (Container-Managed Persistence). CMP was
the old javax.ejb.EntityBean persistence model that was obsoleted by
the new javax.persistence API.
Feed-back
(A) I assume then that if I do a lookup of the EntityManagerFactory
(or get
injected into my #LocalClient) that ny EMs created by it will be
able to
access the entities inserted/updated by the container?
Right. I doesn't matter what persistence technology does the
inserting/updating as long as the data is flushed to the database.
Using two at the same time in the same transaction would be a big no-
no, but it doesn't sound like you're doing that.
I could use this
mechanism to set up my test fixtures. Would I use the
UserTransaction or the
EntityTransaction API to begin/commit these transactions?
Sure, we have an example that does exactly that using the
UserTransaction:
http://svn.apache.org/repos/asf/openejb/tags/openejb-3.1.1/examples/testcase-injection/
Here the setup is done in the test case via directly using an
EntityManagerFactory. The entities inserted are later read via a
Stateful bean using an EntityManager.
(B) The containers db pool is indeed pinting to the same db and is
container
managed. You correct in your assumption that there is no active
transaction
when the methods on the SSB are invoked. And what you describe is the
behaviour that I expected.
(C) the insert and read methods are on the same Stateless Session
Bean and
reference the same EJB entity.
I am very excited by the benefits that this product offers in
testing. Both
for functional and regression testing.
There's likely something simple going on. Should be pretty clear with
a small test case to reproduce the situation.
-David
