Re: [HACKERS] MERGE SQL Statement for PG11

Hi Andreas,

Sorry for the late response; I was busy at PGConf India.

On Sun, Feb 18, 2018 at 4:48 PM, Andreas Seltenreich <seltenreich(at)gmx(dot)de>
wrote:

> Pavan Deolasee writes:
>
> > Thanks for doing those tests. I've just sent v16a version of the patch
> and
> > I think it fixes the issues reported so far. Can you please recheck?
> Please
> > let me know if there are other issues detected by sqlsmith or otherwise.
>
> I re-did the testing with merge_v16a applied to master at 7923118c16
> with ad7dbee368a reverted because of conflicts. I can confirm that the
> previous testcases don't fail anymore,

Thanks for confirming.

> but sqlsmith readily triggers the
> following assertion:
>
> TRAP: FailedAssertion("!(mergeTargetRelation > 0)", File: "planner.c",
> Line: 1496)
>
>
Ah. Looks like the support and test cases for sub-partitioning were
missing. I've attached v17a of the patch which fixes this and some other
issues I noticed as part of your testing. To provide more details, while we
allow more complex un-aliased joins for the source relation, referencing
columns from the join was broken. I've now fixed that.

I've added relevant test cases to the regression. Also added some tests to
check function scans.

The patch is rebased on the current master. Please let me know how the
sqlsmith tests go with this new version.

Thanks,
Pavan

--
Pavan Deolasee http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services