PG18 crashing on master branch (3.6) on tickets regress

[robe]

I was seeing reds on Debbie and Github, but woodie is fine.

On further inspection these appear to only be in PG18.

debbie

​https://debbie.postgis.net/job/PostGIS_Regress/32179/consoleFull

16:25:03 psql:tickets.sql:1019: NOTICE:  Hole lies outside shell at or near point 25495368.044100001 6671726.9312000005
16:25:03 #3719a|f
16:25:03 #3719b|t
16:25:03 #3774|t
16:25:03 #1014a|POINT(0 0)
16:25:03 #1014a|POINT(0 0)
16:25:03 #1014b|POINT(0 1)
16:25:03 psql:tickets.sql:1047: server closed the connection unexpectedly
16:25:03 	This probably means the server terminated abnormally
16:25:03 	before or while processing the request.
16:25:03 psql:tickets.sql:1047: error: connection to server was lost
16:25:03 ------------------------------------------------------------

github ​https://github.com/postgis/postgis/actions/runs/12323216874/job/34398402435

 psql:tickets.sql:1019: NOTICE:  Hole lies outside shell at or near point 25495368.044100001 6671726.9312000005
#3719a|f
#3719b|t
#3774|t
#1014a|POINT(0 0)
#1014a|POINT(0 0)
#1014b|POINT(0 1)
psql:tickets.sql:1047: server closed the connection unexpectedly
	This probably means the server terminated abnormally
	before or while processing the request.
psql:tickets.sql:1047: error: connection to server was lost

So they are failing on same place in 1047 which has this query

SELECT '#1014c', id, st_astext(g) FROM path;

[robe]

Oops sorry didn't add the full query should be:

CREATE TABLE rec (id integer, g geometry);
INSERT INTO rec VALUES (1, 'POINT(0 1)');
INSERT INTO rec VALUES (2, 'POINT(1 2)');
INSERT INTO rec VALUES (3, 'POINT(2 3)');
WITH RECURSIVE path (id, g) AS (
    SELECT id, g FROM rec WHERE id = 1
    UNION
    SELECT rec.id, rec.g
    FROM path, rec
    WHERE ST_Y(path.g) = ST_X(rec.g)
)
SELECT '#1014c', id, st_astext(g) FROM path;
SELECT '#1014d', ST_AsEWKT(g) as t FROM (
	SELECT 'SRID=1;POINT(0 1)'::geometry AS g
	UNION
	SELECT 'SRID=2;POINT(0 1)'::geometry AS g
) a ORDER BY t;

[robe]

Paul confirmed it's an upstream issue - reported here - ​https://www.postgresql.org/message-id/99F064C1-B3EB-4BE7-97D2-D2A0AA487A71%40cleverelephant.ca