Re: GiST index build versus NaN coordinates

Andreas Seltenreich <seltenreich(at)gmx(dot)de> writes:
> I wrote:
>> Sounds like some fuzz testing with nan/infinity is in order.

> related fallout: close_ps returns a NULL pointer with NaNs around:
> select close_ps('(nan,nan)', '(nan,nan),(nan,nan)');
> -- TRAP: FailedAssertion("!(result != ((void *)0))", File: "geo_ops.c", Line: 2860)

Yeah, that Assert seems way too optimistic. Even without NaNs, I wonder
whether plain old roundoff error couldn't trigger cases where interpt_sl
fails to find an intersection point. I'm inclined to just let close_ps
return SQL NULL in such cases.

regards, tom lane