Re: BUG #19108: Stack overflow duting query parse

> Could you at least provide a not-mangled-to-the-point-of-incorrectness
version of the query?

Sure, here it is:

select "public"."dtn_v2"."uuid",
"public"."dtn_v2"."first_shipment_date_route_original_departure_plan_date"
from "public"."dtn_v2"
join "public"."shipment_v2" on ("public"."shipment_v2"."uuid" =
"public"."dtn_v2"."first_shipment_uuid" and

"public"."shipment_v2"."date_route_original_departure_plan_date" =

"public"."dtn_v2"."first_shipment_date_route_original_departure_plan_date")
where ("public"."shipment_v2"."shipment_node_from_id",
"public"."shipment_v2"."date_route_original_departure_plan_date") in
((39073021, cast('2025-09-10 06:15:00+00' as timestamp)), (40493380,
cast('2025-10-23 06:15:00+00' as timestamp)),
(38609801, cast('2025-08-27 04:30:00+00' as timestamp)), (39038517,
cast('2025-09-09 06:15:00+00' as timestamp)),
...
(40788851, cast('2025-11-01 06:15:00+00' as timestamp)), (39267833,
cast('2025-09-16 04:30:00+00' as timestamp)));

BTW, there were "out of memory errors" before the segmentation fault. For
instance:

could not fork autovacuum worker process: Cannot allocate memory

But I didn't think memory exhaustion could lead to errors like
"segmentation fault".

Also "dnf install ..." command was also core dumped so maybe some shared
memory segments were corrupted.

Thanks!

On Mon, Nov 10, 2025 at 7:02 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Andrey Zhidenkov <pensnarik(at)gmail(dot)com> writes:
> >> Can you create a self-contained SQL script to reproduce this, and post
> it
> >> on this thread with reply-all, please?
>
> > Unfortunately, I cannot reproduce the issue even on an existing 15.7
> > node (one that hasn't upgraded) (don't know if it matters, but now it's a
> > hot standby node) but there are core dumps and debug info packages
> > installed so I can provide more data from core dump.
>
> Could you at least provide a not-mangled-to-the-point-of-incorrectness
> version of the query? The posted one fails with
>
> ERROR: table name "tablename_v2" specified more than once
>
> and there are other visible bugs in it. Some clarity about the column
> data types would be helpful as well.
>
> regards, tom lane
>

--
С уважением, Андрей Жиденков.