Re: BUG #15977: Inconsistent behavior in chained transactions

Hi,

On 2019-09-06 16:54:15 +0900, Michael Paquier wrote:
> On Thu, Sep 05, 2019 at 02:11:35PM -0700, Andres Freund wrote:
> > On 2019-09-05 14:16:11 +0200, Peter Eisentraut wrote:
> >> I'm content with this patch.
> >
> > Would need tests.
>
> The latest patch sends adds coverage for all the new code paths
> added. Do you have something else in mind?

Missed them somehow. But I don't think they're quite sufficient. I think
at least we also need tests that test things like multi-statement
exec_simple-query() *with* explicit transactions and chaining.

> >> Better disable questionable cases now and maybe re-enable them later
> >> if someone wants to make a case for it.
> >
> > I do think the fact that COMMIT in multi-statement implicit transaction
> > has some usecase, is an argument for just implementing it properly...
>
> Like Peter, I would also keep an ERROR for now, as we could always
> relax that later on.

I mean, I agree it's better to err that way, but it still seems
unnecessary to design things in a way that prevents legit cases, that we
then may have to allow later. Error -> no error is a behavioural change
too, even if obviously less likely to cause problems.

Greetings,

Andres Freund