Tighten CREATE AGGREGATE's checks on validity of a SORTOP option.

A sort-operator option only makes sense for an aggregate that behaves
like MIN or MAX.  We can't really verify that the proposed aggregate
does, but for our purposes the most important expectation is that
the aggregate returns the same type as its input.  Without this, the
planner's conversion to a MinMaxAggPath plan risks type violations.
So check it during CREATE AGGREGATE.

While we're at it, try to make sure that the specified SORTOP operator
is sane.  Ideally we'd insist that the sortop be an ordering operator
in some index opclass.  But checking that during CREATE AGGREGATE
risks dump/reload ordering problems, since the opclass might not exist
yet; and it's not essential since the planner will discover the lack
of a suitable opclass when trying to plan usage of the aggregate.
We can at least check that the operator returns boolean.

This is a sufficient fix for the master branch, but it wouldn't
prevent exploitation of pre-installed dangerous aggregates.
So the back branches will use a different fix.

Reported-by: Zheng Yu <[email protected]>
Author: Tom Lane <[email protected]>
Security: CVE-2026-116331

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/fffd8a41f3ceaa200bcbd422006306db49c9ba23
Author: Tom Lane <[email protected]>

Modified Files
--------------
src/backend/catalog/pg_aggregate.c | 26 ++++++++++++++++++++++++++
1 file changed, 26 insertions(+)

Reply via email to