On Mon, Aug 18, 2003 at 11:38:22AM +0100, Paul Thomas wrote:
>
> On 17/08/2003 15:00 Oliver Jowett wrote:
> >This patch adds a finalizer to AbstractJdbc1Statement that closes the
> >statement. Without this, when server-side preparation is in use
> >statements
> >that are executed then discarded without an explicit close() will leak
> >resources on the backend while that connection remains open, as a
> >DEALLOCATE
> >never gets executed.
>
> Not a good solution IMHO. Relying on GC to clean up resource leaks is a
> poor solution.
I'm not suggesting that relying on GC for cleanup is a good idea. However,
without this patch, the driver will *always* leak backend resources with
longlived connections and leaky client code (which is possibly not under the
control of the eventual owner of the connection -- e.g. the appserver case).
> And it you simply System.exit() the JVM, GC is not called
> at all.
When the JVM exits, the physical connections go down so the backend will do
resource cleanup on its own.
-O