Re: BUG #17809: MERGE ... UPDATE fails with BEFORE ROW UPDATE trigger when target row updated concurrently

On Sat, 11 Mar 2023 at 18:12, Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com> wrote:
>
> Attached is an updated patch that I'm more happy with now, and a
> slightly modified one for v15, keeping the trigger API
> backwards-compatible for extensions.
>

Going over this again, I spotted another couple of things:

Passing recheckIndexes to ExecUpdateEpilogue() is pointless, since
that's something that's only ever computed and used locally within
ExecUpdateEpilogue(). It's also pointless for the caller to
list_free() it, since the caller only has a pointer to the empty list,
not the list built in ExecUpdateEpilogue(), which it should free. This
is pretty harmless, but I might as well tidy it up while hacking on
this.

If a BEFORE ROW trigger returns NULL to skip an update or delete,
ExecMergeMatched() will still update estate->es_processed, and so the
final row count in the command tag will be wrong. I'll tackle that in
a follow-on patch, since it's really a separate bug.

Regards,
Dean