db-jdo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michelle Caisse <Michelle.Cai...@Sun.COM>
Subject Re: pmf not closed
Date Tue, 03 May 2005 20:29:30 GMT
Hi, Michael,

Responses in-line...

Michael Bouschen wrote:

> Hi Michelle,
>
> thanks for testing.
>
>> Hi, Michael,
>>
>> The tests still don't pass because you are attempting to delete the 
>> PCPoints before the PCRects.  Even though it is one transaction, you 
>> have to change the order as I did below.
>
>
> I agree we need to change the order and delete PCRect instances first 
> as you suggest. The cleanup code should succeed in any case. But maybe 
> this scenario is an interesting test for the TCK delete test. I think 
> the JDO implementation should reorder the SQL DELETE statements 
> according to the foreign keys in the database such that the 
> transaction succeeds no matter which instances the user deleted first 
> in the JDO transaction.

Interesting point.  I hadn't considered that.  Is it in the spec?

>
>>
>> The cleanup appears to happen correctly now so the PMF is closed.  
>> But I don't see a failure message, just "FAIL".  I don't know why 
>> this is.
>
>
> Yes, just getting "FAIL" is strange. I need to run the test cases you 
> suggest below and figure out the details.

In general, the tests are not giving as much feedback when they fail as 
I would expect.

>
> Does this hold you up or does the current changes of JDO_Test and 
> PersistenceManagerTest allow you to continue with the TCK tests?

This is not holding me up.  You can check in now or later, whatever 
works for you.

Thanks!!

-- Michelle

>
> Regards Michael
>
>>
>> I am using CurrentTransaction as the test case. I believe that 
>> ConcurrentPersistenceManagersSameClasses leaves behind the PCRects 
>> that PersistenceManagerTest then tries to delete when 
>> CurrentTransaction is run.  These two run in sequence when you run 
>> maven runtck.jdori, so you can just watch for them to scroll by.  Or 
>> you can be more clever and just run the two tests in sequence.
>>
>> -- Michelle
>


Mime
View raw message