I agree that it's not really worth having tests for this, and I take your point about the dependency on wal_level that we don't currently have. The problem is that the core tests include publications already, and it doesn't seem like a great idea to move the whole lot to a TAP test. Creating alternative expected files seems like a bad idea too (annoying to maintain, wouldn't compose well with the next thing like this). So... how about we just suppress WARNINGs for CREATE PUBLICATION commands that are expected to succeed? Like in the attached. This version passes installcheck with any wal_level.
All right, for me. If wal_level can not interfere with the testes result, it seems to a better approach