>And why do you think this is a bug? Immediately before the part of code
>you copied we have this:
>
> /*
> * The ScanTupleSlot holds the (copied) first tuple of each group.
> */
> firsttupleslot = node->ss.ss_ScanTupleSlot;
>And node->ss.ss_ScanTupleSlot is expected to be non-NULL. So the initial
>assumption that firsttupleslot is NULL is incorrect.
IMHO, the test could be improved, this way it silences the scan tool.
--- \dll\postgresql-12.0\a\backend\executor\nodeGroup.c Mon Sep 30 17:06:55 2019
+++ nodeGroup.c Sat Nov 23 00:23:27 2019
@@ -64,7 +64,7 @@
* If first time through, acquire first input tuple and determine
whether
* to return it or not.
*/
- if (TupIsNull(firsttupleslot))
+ if ((firsttupleslot != NULL) && TTS_EMPTY(firsttupleslot))
{
outerslot = ExecProcNode(outerPlanState(node));
if (TupIsNull(outerslot))
best regards.
Ranier Vilela
