> > > Note that the memory structure underlying the obj# printout is populated > by > > things like IO waits, some enqueue waits (I think) and buffer busy waits > > etc. However, these values aren't cleared, so whatever is the next wait > > event (which doesn't overwrite the obj#) - it may just report some old > > wait's obj#... > > BTW > > This same issue appear to exist in v$session tables when session is > going from waiting to on cpu state - it looks like latest wait event > is not cleared as well - sometimes can be useful but can also do some > unnecessary mess. > > Yep, it's because SQL trace picks this stuff from v$session array... For example when you have cache buffers chains latch contention, then you can't trust the obj# nor the file#,block# numbers either (not in v$sesstat, not in SQL trace) as latch wait events just don't update this information.... Tanel