On Oct 4, 2007, at 1:45 PM, Martin Thelian wrote:
Yes, but this would not help to avoid a possible race-condition
between
thread1 and another thread2. Because there is no guarantee that
thread2
has not already send an SQL statement via the connection when
thread1 is
reseting the transaction isolation level.
Ok, looks like you are right, and there is a possibility of a race
condition between tx.commit/rollback and the user code resetting the
connection policy. A workaround would be a TransactionDelegate:
* willAddConnection may set the isolation level
* didCommit/didRollback may reset it back before the connection is
returned to the pool.
Do you think this will work?
Andrus
