db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dyre.Tjeldv...@Sun.COM
Subject Re: Momory leak in JDBC driver?
Date Thu, 08 Sep 2005 09:03:50 GMT
Knut Anders Hatlen <Knut.Hatlen@Sun.COM> writes:

[snip]

>> Dyre, could you file a bug for this?  I don't think one exists for this
>> issue.
>
> FYI, I had a look in the client driver code. It seems like only
> markClosed() is called on the open ResultSets when a statement is
> re-executed. The difference between close() and markClosed() is
> basically that close() additionally checks whether it should
> auto-commit, nulls out the cursor and meta data, and calls
>
>   connection_.CommitAndRollbackListeners_.remove(this);
>
> Could this be our memory leak? Is there any situation where one
> shouldn't call close() on the result sets when re-executing?

I would like to point out, as you told me, that the leaked ResultsSets
probably do not explain the "Out of Mem" error, since the ResultSets
don't use much memory after markClosed is called.

And just to clarify: My original problem was not memory exhaustion,
but a performance degradation that I traced to the traversal of 
connection_.CommitAndRollbackListeners_ at each commit.

-- 
dt



Mime
View raw message