Hi
> Posting an updated set of patches to address recent feedback:
>
> - Removed conditional-locking code used in parallel-safety checking code
> (Tsunakawa-san feedback). It turns out that for the problem test case, no
> parallel-safety checking should be occurring that locks relations because
> those inserts are specifying VALUES, not an underlying SELECT, so the
> parallel-safety checking code was updated to bail out early if no underlying
> SELECT is specified for the INSERT. No change to the test code was required.
> - Added comment to better explain the reason for treating "INSERT ...
> ON CONFLICT ... DO UPDATE" as parallel-unsafe (Dilip)
> - Added assertion to heap_prepare_insert() (Amit)
> - Updated error handling for NULL index_expr_item case (Vignesh)
+
+ index_oid_list = RelationGetIndexList(rel);
...
As memtioned in the comments of RelationGetIndexList:
* we return a copy of the list palloc'd in the caller's context. The caller
* may list_free() the returned list after scanning it.
Shall we list_free(index_oid_list) at the end of function ?
Just to avoid potential memory leak.
Best regards,
houzj