We got a Unique Constraint error in our application. I wanted to find the sql which caused this. Since I found out about the error a bit after the fact, my thoughts turned to using logminer. When thinking about this I wondered wether the SQL which caused the error would make it to the redo stream. If it did, the error would then need to be rolled back. But this would need to be a special kind of rollback, one which wouldn't terminate the transaction. So maybe the kernal would notice the error on checking and cancel the statement before it got to the redo stream. However, even if that happened in this case, what about ROLLBACK TO SAVEPOINT. These statements defiinitely make the redo stream. I'm putting together a test case, so I'll post what I find as soon as I can. (probably with other questions) Henry -- //www.freelists.org/webpage/oracle-l