Hi amit,
(sorry about not cc the hacker list)
I have an issue about command id here.
It's probably not directly related to your patch, so I am sorry if it bothers
you.
+ /*
+ * Start the scan. To make the changes of the current command visible to
+ * the scan and for subsequent locking of the tuple (if any) found,
+ * increment the command counter.
+ */
+ CommandCounterIncrement();
For insert on fk relation, is it necessary to create new command id every time ?
I think it is only necessary when it modifies the referenced table.
for example: 1) has modifyingcte
2) has modifying function(trigger/domain...)
All of the above seems not supported in parallel mode(parallel unsafe).
So I was wondering if we can avoid the CommandCounterIncrement in parallel mode.
Best regards,
houzj
