Re: failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

On Tue, Mar 3, 2015 at 7:14 PM, Peter Geoghegan <pg(at)heroku(dot)com> wrote:
> My patch actually matches Andrew Gierth's datumsort patch, in that it
> also uses this convention, as I believe it should. For that reason,
> I'd prefer to make the comment added in November true, rather than
> changing the comment...I feel it ought to be true anyway (which is
> what I meant about a pre-existing issue). You'll still need the
> defenses you've added for the the hash case, of course. You might want
> to also put a comment specifying why it's necessary, since the hash
> tuple case is the oddball cases that doesn't always have "sortKeys"
> set.
>
> In other words, I suggest that you commit the union of our two
> patches, less your changes to the comments around "sortKeys'.

I think we should commit my patch, and if a future patch needs
sortKeys set in more places, it can make that change itself. There's
no reason why it's needed with the code as it is today, and no reason
to let bits of future changes leak into a bug-fix patch.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company