Re: Dropped rows on SQL INSERT

On 10-Feb-2015 06:33 -0600, Needles,Stephen J wrote:

I've got an issue I don't understand. First...we are at 7.2. I can't
find anything like this in the 'net.

We have loaded 7.2 on our development LPAR and we immediately began
to have an issue with this scenario.

We've a CTE that normally generates...say 8 rows...when writing to a
table in QTEMP. Now this results in 6 rows.

During debugging...the CTE (without the insert) will generate the
expected 8 rows. But wrap the CTE in an insert (that's all) and two
rows are dropped. Interestingly (coincidentally?), the missing rows
are also the rows that are impacted by a GROUP BY in the 2nd
expression that eliminated the duplicates. Remove this GROUP BY and
the missing rows are in the final result set.

I'll try to sanitize the SQL somewhat and post it here soon. Anyone
even heard of such a thing?

The final result set is either correct or incorrect output [kwd: INCORROUT]. If incorrect results are being achieved, then probably best to report the error as a defect to your service provider. Without a reduced\simple re-create scenario [with DDL and DML to establish the sample data], knowing whether the current-release output or the prior release output was incorrect [inferred they are different, much like the difference between the SELECT result set for the INSERT vs just the SELECT] is not possible for the readers, but surely you know.? Trying to get the community here to assist with anything but helping to establish a simplified re-create to give to the service provider is probably just delaying getting a resolution from them.