Re: Free indexed_tlist memory explicitly within set_plan_refs()

On Thu, May 28, 2015 at 2:37 PM, Peter Geoghegan <pg(at)heroku(dot)com> wrote:
> Attached, revised version incorporates this small fix, while adding an
> additional big fix, and a number of small style tweaks.
>
> This is mainly concerned with fixing the bug I was trying to fix when
> I spotted the minor pfree() issue:
>
> postgres=# insert into upsert (key, val) values('Foo', 'Bar') on
> conflict (key) do update set val = excluded.val where excluded.* is
> not null;
> ERROR: XX000: variable not found in subplan target lists
> LOCATION: fix_join_expr_mutator, setrefs.c:2003

My fix for this issue
(0001-Fix-bug-with-whole-row-Vars-in-excluded-targetlist.patch) still
missed something. There needs to be additional handling in
ruleutils.c:

postgres=# explain insert into upsert as u
values (1, 'fooz') on conflict (key)
do update set val = excluded.val where excluded.* is not null;
ERROR: XX000: bogus varattno for INNER_VAR var: 0
LOCATION: get_variable, ruleutils.c:5904

I'll look for a fix for this additional issue tomorrow.
--
Peter Geoghegan