Hi David, Thanks. Keeping t_oid as an always NULL column makes sense to me. Also leave pageinspect behavior today make sense as well.
Let's go with your patch. I rechecked the related comments and documents, I think we need to 1. For storage.html, the paragraph after the HeapTupleHeaderData table. Its first sentence lists "an optional object ID field". The sentences from "The object ID is only present if the HEAP_HASOID_OLD bit is set" to the end of the paragraph can go. 2. htup_details.h needs some clean work. the comment still says "If an OID field is present....." 3. For pageinspect.smgl, it point users to htup_details.h, may add something like "t_oid is always NULL. Tuples no longer store an OID, and the column is kept for compatibility." Thanks, Shihao
