Re: Index build temp files

On 10 January 2013 04:02, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Simon Riggs <simon(at)2ndQuadrant(dot)com> writes:
>> Having it USERSET allows different settings for different roles, which
>> is useful.
>
> That would still be possible if it were SUSET; you'd just need a
> superuser to set it for you (via ALTER ROLE SET, or perhaps a
> security-definer wrapper function if you were desperate).
>
> The real question is how necessary is it for unprivileged code to set
> temp_tablespaces *for itself*. I doubt that that's all that critical.

OK, I can accept that

> I suspect the variable is hardly used in the field at all, given that
> it's been there since 8.3 and nobody noticed these issues until
> Stephen started poking at it.

...just not that bit, since it was not being used as documented.

> Note I am certainly not suggesting that we back-patch any such change.
> This would just be for 9.3 and up. We have surely made bigger
> incompatible changes than this one, probably in just about every major
> release.

OK

--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services