Re: BUG #19099: Conditional DELETE from partitioned table with non-updatable partition raises internal error

On Wed, Jan 14, 2026 at 10:38 PM Amit Langote <amitlangote09(at)gmail(dot)com> wrote:
> On Wed, Jan 14, 2026 at 9:30 PM Amit Langote <amitlangote09(at)gmail(dot)com> wrote:
> > Attached is an updated version with improved comments and simplified test cases.
> >
> > Regarding back-patch safety (to v14 where the bug was introduced):
> >
> > * EXPLAIN VERBOSE output order changes (ctid now appears before tableoid)
> > * AddForeignUpdateTargets is no longer called when the FDW doesn't
> > support the command
>
> Hit send too soon re the 2nd bit I guess. Actually we can just return
> false when AddForeignUpdateTargets is missing, so the callback is
> still called when present.
>
> Updated patch attached.

To summarize, the two approaches we've thought about:

1. Executor-side fix
(v4-0001-Fix-bogus-ctid-requirement-for-dummy-root-partiti.patch
posted with my Nov 8 email):

Make ExecInitModifyTable() not require ctid when the only result
relation is a dummy partitioned root. This is minimally invasive but
leaves EXPLAIN VERBOSE output inconsistent depending on
enable_partition_pruning -- with pruning off, you see tableoid but no
ctid, while with pruning on, you see ctid. That's confusing for users
as mentioned upthread.

2. Planner-side fix
(v4-0001-Fix-row-identity-handling-for-dummy-partitioned-r.patch
posted with my last email):

Don't add tableoid for child relations that don't contribute
row-identity columns. This keeps root->row_identity_vars empty when
there exists only one such child relation, so
distribute_row_identity_vars() can add ctid for the dummy root.
EXPLAIN output is consistent regardless of pruning setting. (Some may
notice in the patch that there's still a minor change, but that's due
to how explain.c decides whether to print the table name before the
column name, which is unrelated to this.)

I'm inclined to go with the second approach. The only back-patching
concern is that EXPLAIN VERBOSE output order changes (ctid now appears
before tableoid). This is cosmetic -- junk columns are looked up by
name, not position -- but could affect tests or tools that parse
EXPLAIN output by position.

If there are no objections, I'll commit patch #2 next week.

--
Thanks, Amit Langote