Hi,
On 2017-03-15 14:46:22 -0400, Robert Haas wrote:
> On Wed, Mar 15, 2017 at 2:40 PM, Andres Freund <andres@anarazel.de> wrote:
> Yeah, I don't think that would hurt anything.
>
> (I'm not sure it'll help anything either - the overhead of an extra
> AcceptInvalidationMessages() call is quite minimal - but, as you say,
> maybe it's worth doing just to discourage future code authors from
> including unnecessary fluff.)
I don't think there's an actual runtime advantage either - but it's
indeed confusing for others, because it doesn't square with what's
needed. It's not like the AcceptInvalidationMessages() would actually
make things race-free if used without RangeVarGetRelidExtended()...
- Andres