Re: hash index on unlogged tables doesn't behave as expected

At Thu, 21 Sep 2017 20:35:01 -0400, Robert Haas <robertmhaas(at)gmail(dot)com> wrote in <CA+TgmobXYq1ht8R76RTvun0pY85-=Oov8EY2Fv8nhNnM7Gdzxg(at)mail(dot)gmail(dot)com>
> On Thu, Sep 21, 2017 at 8:16 PM, Kyotaro HORIGUCHI
> <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> > Though I don't see it's bug and agree that the message is not
> > proper, currently we can create hash indexes without no warning
> > on unlogged tables and it causes a problem with replication.
>
> That's true, but I don't believe it's a sufficient reason to make a change.
>
> Before 84aa8ba128a08e6fdebb2497c7a79ebf18093e12 (2014), we didn't
> issue a warning about hash indexes in any case whatsoever; we relied
> on people reading the documentation to find out about the limitations
> of hash indexes. They can still do that in any cases that the warning
> doesn't adequately cover. I really don't think it's worth kibitzing
> the cases where this message is emitted in released branches, or the
> text of the message, just as we didn't back-patch the message itself
> into older releases that are still supported. We need a compelling
> reason to change things in stable branches, and the fact that a
> warning message added in 2014 doesn't cover every limitation of a
> pre-1996 hash index implementation is not an emergency. Let's save
> back-patching for actual bugs, or we'll forever be fiddling with
> things in stable branches that would be better left alone.

Sorry for annoying you and thank you. I agree with that after
just knowing the reason is not precisely (3) (we already have
WARNING for the problematic ops).

regards,

--
Kyotaro Horiguchi
NTT Open Source Software Center