Message7338

I'm interested in seeing this bug resolved.  I'm attaching a proposed patch.

My reasoning behind the patch is that if the last statement executed did *not* return a result set, then we want the rowcount to return the value of updatecount.  But if it didn't return a result set, then updatecount will be an integer >= 0, so we check for that and just return the updatecount.

Otherwise, if the last statement executed *did* return a result set, then we want to just return the rowcount value as before.  But if it did return a result set, then updatecount will be None, so in that case we return the rowcount value as before.

This patch preserves the old updatecount field behavior for now to give people relying on it time to transition.

Feedback welcome. :)