From: Tom Lane <tgl@sss.pgh.pa.us>
To: 253540651@qq.com
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #17152: ERROR: AddressSanitizer: SEGV on unknown address
In-reply-to: <17152-c7f906cc1a88e61b@postgresql.org>
References: <17152-c7f906cc1a88e61b@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Wed, 18 Aug 2021 02:56:00 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1431889.1629300960.1@sss.pgh.pa.us>
Date: Wed, 18 Aug 2021 11:36:00 -0400
Message-ID: <1431890.1629300960@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1431890.1629300960%40sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> [ unreadable example ]

This can be boiled down to

CREATE TEMP TABLE v0 ( v2 int );
INSERT INTO v0 VALUES (1);

SELECT (
 SELECT
  MODE ( ) WITHIN GROUP ( ORDER BY v2 )
    FILTER ( WHERE MODE ( ) WITHIN GROUP ( ORDER BY v2 > 0 ) )
)
FROM v0;

It seems fairly clear to me that this ought to be rejected as invalid,
because we don't allow aggregates within the arguments of aggregates.
That does happen without the sub-select:

regression=# SELECT
regression-#   MODE ( ) WITHIN GROUP ( ORDER BY v2 )
regression-#     FILTER ( WHERE MODE ( ) WITHIN GROUP ( ORDER BY v2 > 0 ) )
regression-# FROM v0;
ERROR:  aggregate functions are not allowed in FILTER
LINE 3:     FILTER ( WHERE MODE ( ) WITHIN GROUP ( ORDER BY v2 > 0 )...
                           ^

So somehow the fact that outer references are involved is misleading that
error check.  I've not traced it further than that.

			regards, tom lane