>I noticed one more thing: pg_dumpall.c doesn't really need to prohibit
>--on-conflict-do-nothing without --insert. Its existing validation rejects
>combinations of the settings that are *not* passed on to pg_dump. It seems OK
>just pass those on and let pg_dump complain. For example, if you say
>--data-only --schema-only", it's pg_dump that complains, not pg_dumpall. I
>should do the same thing here.
Thank you for the clarification. I didn't give thought to pg_dumpall internally
>Pushed, with those changes.