Oracle Sequences and stored procedure

Hi,

I'm having a problem with the Ids chosen by LightSpeed for entities after calling a stored procedure.

Here is some pseudocode:

1. create some entities (let's say users)
   • after creating the user the sequence value is, for example, 35
2. call a stored procedure that creates more users. We are careful to use the LightSpeed sequence for ids
   
   • OracleCommand cmd = new OracleCommand("begin prc_create_users; end;")
   • (UnitOfWork.PrepareCommand(cmd); cmd.ExecuteNonQuery();)
   • After executing the sequence may be 40
3. create another user
   • When examing the id for this new user, it ends up as 35 as if the sequence never incremented

So of course the SaveChanges throws an exception due to a primary key violation.

When I examine the sequence in oracle it is 40, but the new entities id is 35.

My sequence had a cache size of 20 and interval of 1. I changed it to nocache to see if there was a difference.

Here is how I initialize my context

context.IdentityMethod = IdentityMethod.Sequence;
context.IdentityBlockSize = 1;

Thanks

Bert

Hi Bert,

Although I don't have a precise repro for this, I have identified a possible cause and committed a candidate fix.  This will be included in nightly builds dated 6 Nov 2009 and above, available from about 1430 GMT.  Could you give this a test in your scenario and let us know whether it improves matters?

Hi Ivan,

I grabbed the nightly dated 11/10. I believe the problem is fixed.

Is there a resource somewhere that shows all the changes for each nightly build? We are trying to decide if we can risk an update at this point of our project.

Thanks

-Bert

We don't currently publish changelogs for nightly builds.  I can tell you that there have been no changes (other than this one and a non-invasive change to add VistaDB 4 support) to the runtime since the 6 October build.  There have been a few minor fixes to the designer but the most recent of those was 19 October.  So the current nightly build is pretty stable I think -- as far as behaviour/correctness goes, the sequence fix is definitely the highest risk part of it because it's the only part that hasn't been out there and in use.

However, we don't performance test the nightlies.  Therefore, there *is* a risk around performance.  We don't believe we've had a regression in this area (and in fact there have been several improvements since 2.2 RTM), and none of our customers have reported performance issues with the nightlies, but if perf is critical for you then I would advise re-running your load tests with the updated build before making a decision.