Re: New pg_lsn type doesn't have hash/btree opclasses

On Thu, Jun 5, 2014 at 9:54 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
>> On 2014-05-09 22:14:25 +0900, Michael Paquier wrote:
>>> [ patch ]
>
> I've committed a revised version of Andres' patch.
Thanks!

> I thought even that was kind of overkill; but a bigger problem is the
> output was sensitive to hash values which are not portable across
> different architectures.  With a bit of experimentation I found that
> a SELECT DISTINCT ... ORDER BY query would exercise both hashing and
> sorting, so that's what got committed.  (I'm not entirely sure though
> whether the plan will be stable across architectures; we might have
> to tweak the test case based on buildfarm feedback.)
Yeah this was a bit too much, and came up with some more light-weight
regression tests instead in the patch here:
http://www.postgresql.org/message-id/CAB7nPqSgZVrHRgsgUg63SCFY+AwH-=3JuDy7moq-_fo7Wi4++Q@mail.gmail.com
-- 
Michael