The 10GR2 documentation says about the DBMS_STATS.AUTO_INVALIDATE param. "Use DBMS_STATS.AUTO_INVALIDATE. to have Oracle decide when to invalidate dependent cursors." I did noticed it too, that's why I've explicitly used the no_invalidate => false. On 6/1/07, Jonathan Lewis <jonathan@xxxxxxxxxxxxxxxxxx> wrote:
Minor exception to worry about: The default in 10.2 (if not 10.2) is to use delayed invalidation on calculating statistics. To avoid a massive cache flush during things like the 'autotstats' job, cursors are somehow marked for invalidation over the next few hours.
-- Alexander Fatkulin -- //www.freelists.org/webpage/oracle-l