On Sat, May 13, 2023 at 6:07 AM Greg Stark <stark@mit.edu> wrote:
> On Thu, 11 May 2023 at 05:37, Peter Eisentraut
> <peter.eisentraut@enterprisedb.com> wrote:
> > Maybe it was never meant that way and only works accidentally? Maybe
> > hash indexes are broken?
>
> It's explicitly documented to be this way. And I think it has to work
> this way for recovery to work.
>
> I think the reason you and Bharath and Andres are talking past each
> other is that they're thinking about how the implementation works and
> you're talking about the API definition.
>
> If you read the API definition and treat the functions as a black box
> I think you're right -- those two definitions sound pretty much
> equivalent to me. They both extend the file, possibly multiple blocks,
> and zero fill. The only difference is that smgrextend() additionally
> allows you to provide data.
Just a thought: should RelationCopyStorageUsingBuffer(), the new code
used by CREATE DATABASE with the default strategy WAL_LOG, use the
newer interface so that it creates fully allocated files instead of
sparse ones?