I think I may have found a bug introduced in 3.0.2.
My application's regression tests are failing with 3.0.2 even though they pass on 3.0.1.

I think I tracked down the problem to the following change:

http://cvs.sourceforge.net/viewcvs.py/h ... 42&r2=1.43

Shouldn't loadByUniqueKey contain the following?

Object result = persistenceContext.getEntity(euk);
if ( result==null ) {
result = persister.loadByUniqueKey(uniqueKeyPropertyName, key, session);
}

Let me know if I need to create a JIRA issue and/or submit a test case for this.

Thanks.

Shit, you're right. I commented too much.

Damn, obviously there is a hole in the test suite!

Would you please create a patch to our test suite to reproduce this? We have plenty of existing property-ref examples.

That would be a great help, thanks.

Ok, I'll see what I can do about creating the test-case.
Meanwhile, it can still be fixed in CVS, right?
Wouldn't want other users to have to wait for my newbie self to get a proper test-case together, now would we? :-)

P.S. Thanks for the prompt reply!

of course, i already fixed it

but, a fix without a test is a non-fix

Do your test-cases have a mechanism for evaluating the number of SQLs executed on a particular load operation?
I ran the org.hibernate.test.propertyref.PropertyRefTest with and without the bug, and here are the results of "s.get(Person.class, p.getId() )" for each:

//correct (one SQL)
Hibernate: select person0_.id as id1_, person0_.name as name0_1_, person0_.userId as userId0_1_, case when person0_1_.id is not null then 1 when person0_.id is not null then 0 end as clazz_1_, address1_.id as id0_, address1_.address as address1_0_, address1_.zip as zip1_0_, address1_.country as country1_0_, address1_.person as person1_0_ from Person person0_ left outer join Employee person0_1_ on person0_.id=person0_1_.id left outer join Address address1_ on person0_.id=address1_.person where person0_.id=?

//bug (duplicate/unnecessary DB fetch)
Hibernate: select person0_.id as id1_, person0_.name as name0_1_, person0_.userId as userId0_1_, case when person0_1_.id is not null then 1 when person0_.id is not null then 0 end as clazz_1_, address1_.id as id0_, address1_.address as address1_0_, address1_.zip as zip1_0_, address1_.country as country1_0_, address1_.person as person1_0_ from Person person0_ left outer join Employee person0_1_ on person0_.id=person0_1_.id left outer join Address address1_ on person0_.id=address1_.person where person0_.id=?

Hibernate: select address0_.id as id0_, address0_.address as address1_0_, address0_.zip as zip1_0_, address0_.country as country1_0_, address0_.person as person1_0_ from Address address0_ where address0_.person=?

Also, it turns out that my own application regression test failure was due to the second fetch of the associated object not returning any rows due to a problem with my test data/mappings.

Yes, we can use Statistics to see any fetches that happen.

Hi Gavin,

It pains me to be the bearer of so much bad news, but in trying to write up a test case for this issue, it looks like I stumbled onto another bug introduced in 3.0.2.

The newly added StatisticsImpl prepareStatementCount and closeStatementCount aren't being reset to 0 in the clear() method:

http://cvs.sourceforge.net/viewcvs.py/h ... 11&r2=1.12

On the bright side, I think I'm on my way to completing a test case for the propertyRef regression... should a Jira task be created so I can name the test method with the Jira reference number? Or should I just add a javadoc comment explaining what the test is for?

Note: I was thinking of adding a seperate test method to org.hibernate.test.propertyref.PropertyRefTest... is that the right place for it?

Thanks.

Thanks, fixed in CVS.

Yes, please add to JIRA.

And yes, that sounds like the right place :)

Should a JIRA bug for the EntityType.loadByUniqueKey() regression be created so that the testcase can reference its number?
Or should I simply create a patch with a description of the bug and have the testcase reference that number instead?

Sorry, kinda new to JIRA and not sure what your procedures are...

For the bug :)

Ok, here's the JIRA issue:

http://opensource.atlassian.com/project ... se/HHH-414

I'll submit the modified org.hibernate.test.propertyref.PropertyRefTest once I'm done with it.

Ok, I think I've got the regression test completed and have attached it to the JIRA issue.
Please have a look and let me know if looks alright.

Thanks.