Hello!
> One other thing I just noticed in BINARY mode. I believe we're missing a
> check in ProcessCopyOptions() with ON_CONFLICT TABLE to show a proper
> error message, e.g.
It is possible to crash the current patch with binary mode, that check
is definitely needed.
+ MakeTransitionCaptureState(cstate->conflictRel->trigdesc,
+ RelationGetRelid(cstate->conflictRel),
+ CMD_INSERT);
Shouldn't this update mtstate->mt_transition_capture?
+ if (cstate->opts.on_conflict == ONCONFLICT_TABLE)
...
+ if (conflict_mstate->fireBSTriggers)
+ {
+ ExecBSInsertTriggers(conflict_mstate->ps.state,
conflict_mstate->rootResultRelInfo);
+
+ conflict_mstate->fireBSTriggers = false;
+ }
+
and
+ if (cstate->num_conflicts > 0)
+ {
...
+ /* Execute AFTER STATEMENT insertion triggers */
+ ExecASInsertTriggers(cstate->mtcontext->estate,
+ on_conflict_mtstate->rootResultRelInfo,
+ on_conflict_mtstate->mt_transition_capture);
* Doesn't statements typically fire triggers unconditionally? INSERT
ON CONFLICT DO NOTHING; fires BS+AS triggers even if it doesn't insert
any rows.
* Isn't firing a before statement trigger after some before/after row
triggers were already fired (for non conflicting rows) strange?
+ if (valid_col_count != 4)
+ errdetail_msg = _("The conflict_table is incomplete; exactly four
columns are required.");
if valid_col_count > 4, is it still incomplete, shouldn't the error
message change in that case?
