On 9/11/06, Manish Khettry <[EMAIL PROTECTED]> wrote:
DERBY-883 did change how the SQL layer modifies the ResultSetNode tree for
group by's so its possible that it has altered the runtime behavior in some
cases, although this was not the intent!
Its worth verifying first, that the new results are indeed correct and then
which checkin fixed/regressed this query.
Thanks,
Manish
On 9/11/06, Kristian Waagan <[EMAIL PROTECTED]> wrote:
> Kathey Marsden wrote:
> > I am looking at a wrong results case and it would appear that it has
> > already been fixed. It appears to be wrong results when RTRIM is added
> > in to the query with 10.1.3, and 10.2. In the trunk I think it is
ok.
> >
> > I am kind of guessing it was DERBY-883 that fixed it, but the query is
> > not group by expression actually.
>
> Hello Kathey,
>
> Not an answer to your question, but DERBY-883 also fixed a bug with a
> query using COALESCE. Maybe the patches fixed a more general bug
> somewhere in the sql/optimizer layer.
>
> At least one of the patches were merged into 10.2. Do you still see the
> bug there? If so, then there must be another patch affecting these
> queries as well.
>
I did a quick check with some 10.2 beta builds I had lying about, and
the problem (trim trims off 12 to 1) occurs in 10.2.1.1 (436991) but
not in 10.2.1.3 (441574).
Of course, that doesn't narrow it down much. :-/
Myrna
