[postgis-users] Crash when calling distance_ellipsoid with PostGIS1.0rc1

[strk at refractions.net]

Thanks for the report Stephen. 
It was a bug in the definition of the function.
It is fixed in CVS.

Here is a quick patch:

--- lwgeom_spheroid.c   5 Jan 2005 12:44:47 -0000       1.6
+++ lwgeom_spheroid.c   24 Jan 2005 18:10:45 -0000
@@ -486,10 +486,10 @@
 //distance (geometry,geometry, sphere)
 // -geometrys MUST be points
-PG_FUNCTION_INFO_V1(distance_ellipsoid);
+PG_FUNCTION_INFO_V1(LWGEOM_distance_ellipsoid_point);
 Datum LWGEOM_distance_ellipsoid_point(PG_FUNCTION_ARGS)
 {


--strk;

On Mon, Jan 24, 2005 at 12:10:22PM -0500, Stephen Marshall wrote:
> Calling the distance_ellipsoid function fails with PostGIS1.0rc1, where 
> is worked with PostGIS0.9.
> 
> In this example, the same call was made in two contexts, both using 
> PostgreSQL 8.0.0; the first with PostGIS 0.9 and the second with 
> PostGIS1.0rc1.  In the latter case, the postgres backend server terminated.
> 
> PostGIS 0.9:
> avdata_orig# select distance_spheroid(geometryfromtext('POINT(-77.122778 
> 12.008611)', 4326), geometryfromtext('POINT(-89.656111 20.938056)', 
> 4326), 'SPHEROID["WGS_1984",6378137,298.257223563]]'::spheroid);
> distance_spheroid
> -------------------
>  1662043.58935675
> (1 row)
> 
> PostGIS 1.0rc1:
> avdata=# select distance_spheroid(geometryfromtext('POINT(-77.122778 
> 12.008611)', 4326), geometryfromtext('POINT(-89.656111 20.938056)', 
> 4326), 'SPHEROID["WGS_1984",6378137,298.257223563]]'::spheroid);
> server closed the connection unexpectedly
>        This probably means the server terminated abnormally
>        before or while processing the request.
> The connection to the server was lost. Attempting reset: Failed.
> 
> I tracked the crash into the PostGIS function 
> LWGEOM_distance_ellipsoid_point().  It occurs at this call:
> 
>    PG_LWGEOM *geom1 = (PG_LWGEOM *)PG_DETOAST_DATUM(PG_GETARG_DATUM(0));
> 
> By calling this function:
> 
>    DatumGetPointer(PG_GETARG_DATUM(0));
> 
> I was able to determine that a valid (non-zero) address is being 
> returned for the Datum pointer, so I think the error occurs in the 
> PG_DETOAST_DATUM call.
> 
> My suspicion is that the data is being incorrectly identified as a 
> toasted value, and hence erroniously going through "detoasting" logic in 
> the postgres function pg_detoast_datum (found in 
> ./src/backend/utils/fmgr/fmgr.c in the postgres distribution).  However, 
> I have no inside knowledge of the postgres toasting code, so this is a 
> fairly naive guess.
> 
> Thanks for any help you can give.
> Steve Marshall
> 
> 
> 
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users