JBossDeveloper

Originally  treated "closeBoundStatement" "closePrepareStatement" as same, that a was a mistake

Ramesh to clarify in closeBoundStatement there is the autoCommit check and a call to closePreparedStatement with the comment about http://pgfoundry.org/tracker/?func=detail&atid=538&aid=1007690&group_id=1000125

Are you now saying that is not correct?

Steve

I think that is the issue.

But I remember during tests at that time, i never seen a "closePrepareStmt" call from client, just the "closeBoundStatement" call, thus the way it was coded. Now I am not sure if that is fixed in > 9 version of the PG driver or not and how we are going to support older drivers. Trying to think how the link in the comment is relevent?

Using the older one is probably better.  We should have enough to go on for now and will have a better baseline for you with 8.5 Beta1.

Steve

Ramesh,

I'll open a new issue for this.  You are correct that the odbc client does not issue an explicit close message of a prepared statement - instead it uses DEALLOCATE.  So I don't think we should be removing the query from the prepared map ever when just closing the portal.  If we're worried about the memory consumption, we can hold the Prepared statement with a soft reference.  But that doesn't seem necessary and is somewhat complicated.  We could also just put a hard limit on the number of open portals/prepared plans.

The pg jdbc client has different behavior.  It uses phantom references to issue specific close of the portal and statement when the corresponding objects have been garbage collected.

Steve

Since on DEALLOCATE Teiid already closes the prepare statements, that is why you are saying memory consumption on runway stuff is not of a concern?

No I'm saying that the odbc client code has the ability to close statements through DEALLOCATE.  I was assuming from your previous statement that you assumed we should cleanup on bound to prevent a leak of prepared statements.  I think for now we can just ignore system limits on the number of plans/portals as it's unlikely to be an issue any time soon.

Steve

Originally I do not think I assosiated the DELLOCATE with closing the prepare statements I belive (I do not remember), thus I was closing them close bound.

Marco,

8.5 Beta1 with TEIID-2604 TEIID-2608 and TEIID-2610 has been posted to http://www.jboss.org/teiid/downloads

This should get you further hopefully.  If there are any problems just post the org.teiid.ODBC trace logs and we'll get them resolved.

Thanks,

Steve