Re: BUG #14948: cost overflow

Hello Tom,

On 5 December 2017 at 22:30, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> The index cost estimation code does have some potential for
> garbage-in-garbage-out results, but nothing very promising.
> One question is whether these indexes are on a non-default
> tablespace with a non-default random_page_cost. PG should
> prevent you from putting in a negative random_page_cost, but
> it's worth checking that.

How do I check?

> I looked at some other possibilities like a corrupted tree_height
> value, but none of them seem to fit the available info. For
> instance, the tree_height is only an int, so even if it were the
> max negative value it would not explain the cost value you're getting.

I must confess that I have no clue what you are talking about.

> Don't suppose you'd like to step through btcostestimate() with
> a debugger and see where it's going off the rails?

I've never done debugging of a C program. If you give me some hints how to
do that, I might be able to (Ubuntu, PG from the apt repository).

Is there any debug log I might need to enable to get some similar information?

Best regards,

Jan
--
Jan Schulz
mail: jasc(at)gmx(dot)net