[postgis-users] Contradiction between ST_IsValid() and ST_IsValidReason() ?

[Sandro Santilli]

On Wed, Dec 23, 2020 at 04:51:07PM +0100, Andreas Neumann wrote:
> Hi Strk and Nicolas,
> 
> Thanks for your replies.
> 
> I'm using GEOS 3.9, PostgreSQL 13.1 and Postgis 3.1 (self-compiled).
> 
> The situation is really weird: if I run the same SQL query a second
> time, the contradiction is gone and all geometries are labeled as
> invalid by ST_IsValidReason(). If I run it many times, the results
> change even more, suddenly, the invalid geometries are not even detected
> correctly by ST_IsValid().
> 
> @strk: these invalid geometries contain "NaN" coordinates.
> 
> I will open an issue and submit the geometries.

Can you first try to rebuild PostGIS from current master branch ?
I've fixed a memory issue somewhat related to "NaN" coordinates

--strk;