> On 8 May 2023, at 22:12, PG Bug reporting form <noreply@postgresql.org> wrote:
> The query uses a 3 partitions of a table where all fields have an brin
> index.
One index per column or a multi-colum index?
> This query will segfault:
> SELECT count(*)
> FROM dw
> WHERE (dw.ts >= '2022-09-01' AND dw.ts <= '2022-10-30')
> AND dw.source='type3' AND dw.customer='123.456';
>
> The query should return 0 because the customer does not exist.
> Removing the dw.customer or dw.source constraint will make the segfault not
> occur.
> Also using 'set enable_bitmapscan to off' will not trigger the segfault.
I was unable to reproduce this, can you share the table schema with the index
creations (or a version of the schema which reproduce the problem).
Have you tried recreating the indexes, in case there is corruption, or tried it
on a table which has the same schema but no index?
--
Daniel Gustafsson