Re: Dbms_stats giving OA-06512 error

At 10:01 PM 1/5/2007, Mladen Gogala wrote:

>Hmmmm, you live and learn. I still like to specify my histogram size.
>I stand corrected, it's not wrong, but it's wrong from the aesthetic
>point of view. William, I didn't mislead you on purpose.

I agree. You should say what you mean, not rely on defaults. Just as with date conversions.

> >
> > William,
> > I am convinced your error has nothing to do with your change to the
> > gathering procedure. It has more likely to do with a change to table
> > "EDRS"."AMENDMENT". Did someone create a function-based index?
>
>
>Now, I'm really interested. Why would you suspect a function based index?

because of the sql which William sent in his 2nd post. Ordinarily you only see column names in the sql generated by dbms_stats for histogram collection. My best - and apparently correct - guess was that this is the hidden column for an FB index.

>To me, William's case looks like a case of bug 5558878, described in the note
>5558878.8. Fortunately, there is a patch for that. Of course, my
>recommendation
>to contact Oracle Support, just to verify, still stands. They'll probably love
>the trace you produced by setting the event 979.

Don't you just love that: we're told over and over - especially with the hash group by in 10.2 - not to rely on group by for the order but to always also include an order by. And then Oracle creates a bug for precicely the "group by order by" combination. That scenario doesn't seem to have it made into Oracle's regression tests.

Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com

--
http://www.freelists.org/webpage/oracle-l