Re: BUG #17434: CREATE/DROP DATABASE can be executed in the same transaction with other commands

On Tue, Jul 26, 2022 at 8:08 AM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> "David G. Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com> writes:
> > I guess I am expecting exec_execute_message to have:
>
> > if (completed && use_implicit_block)
> > {
> > EndImplicitTransactionBlock();
> > finish_xact_command();
> > } else if (completed) [existing code continues]
>
> The problem with that is "where do we get use_implicit_block from"?
> In simple query mode it's set if the simple-query message contains
> more than one statement. But the issue we face in extended mode is
> precisely that we don't know if the client will try to send another
> statement before Sync.
> [...]
> Anyway, here's an updated patch, now with docs. I was surprised
> to realize that protocol.sgml has no explicit mention of pipelining,
> even though extended query protocol was intentionally set up to make
> that possible. So I added a <sect2> about that, which provides a home
> for the caveat about immediate-commit commands.
>
>
Thanks! This added section is clear and now affirms the understanding I've
come to with this thread, mostly. I'm still of the opinion that the
definition of "cannot be executed inside a transaction block" means that we
must "auto-sync" (implicit commit) before and after the restricted command,
not just after, and that the new section should cover this - whether we do
or do not - explicitly.

David J.