> Sorry for the delay. V2 looks good to me. Hi, I applied v2 on master with asserts and measured the output. Please commit 0003 separately and soon. It is an unrelated five-line fix for a v19 feature, and v19 has branched.
Do we need the following line in 003? pubform = GETSTRUCT(tup) line; heap_modify_tuple() does not change the OID. 0001/0002 collect when the catalog changed, but ddl_command_end means the command ran. Measured, all three applied: ADD TABLE t2 membership row only SET TABLE t1,t2,t3 (same set) zero rows DROP TABLE t1 publication row SET (publish='insert') twice publication row both times SET ALL TABLES (already all tables) zero rows But one command now has two rules. And 0002's claim that it fixes "no entry" for SET ALL TABLES is untrue for the no-op case I think weneed to fix the commit message. Thanks, Shihao
