Re: BUG #16122: segfault pg_detoast_datum (datum=0x0) at fmgr.c:1833 numrange query

On Mon, Nov 18, 2019 at 02:51:25PM -0800, Adam Scott wrote:
>> Did you see that after updating to 10.11.  If you used 10.10 or
>> an older version, did the problem happen?
>
> Originally this first was discovered on Centos 7, PG 10.10, then loaded the
> 40gb table on 10.11 on Ubuntu.

Which means that prior 10.9 you used this query, but did not notice
anything?  Or is 10.10 the first version in the 10.X series you used?

>> Seeing the plan of your query may help as well.  Could you run EXPLAIN
>> on it or does it crash before?  Perhaps a parallel plan is involved
>> here?
> Explain plan works fine with no crash

I was also wondering about the shape of the plan selected here.  The
crash happens when doing some selectivity on the clause when working
on a plan, but it could help.

> I've done a binary search to find out where the error occurs in the data,
> but no luck.  It seems intermittent now.  Finally, I was able to reproduce
> the error repeatably with a blank table:
> 
> Stop and start postgres from fresh, and then run this query (notice, I
> removed a sarg from the originally supplied query):
> select id from natica_hdu_test
> WHERE
> "dec_range" <@ '[88.9999998611111,90.0000001388889)';

Thanks.  I can see that you have provided a dump on the other part of
the thread.  Let's continue from that.
--
Michael