[postgis-devel] Has anyone seen this before?

[Paul Ramsey]

Compile your pgsql with --enable-cassert and run your selection of
commands again. Hopefully you'll get a big bad error at some point as
a memory operation wanders off the reservation.

P

On Fri, Feb 5, 2010 at 1:55 PM, Paragon Corporation <lr at pcorp.us> wrote:
> What would result in an UNKNOWN geometry type.  One of our clients is
> getting this error when they try to vacuum analyze their table.
>
>> NOTICE:  compute_serialized_box3d called on unknown type 0NOTICE:
>> compute_serialized_box3d called on unknown type 0 ...
>
> On closer inspection -- I see that there are UNKNOWNs and those UNKNOWNS are
> not NULL but the ST_AsText looks blank viewed via psql
>
>
> select count(*),  GeometryType(the_geom) from badtable group by
> GeometryType(the_geom);
>  count | geometrytype
> -------+--------------
>  98483 | MULTIPOLYGON
>  1517 | UNKNOWN
>
>
> This is running on
>
>
> ----------------------------------------------------------------------------
> ----
>  POSTGIS="1.3.6" GEOS="3.2.0-CAPI-1.6.0" PROJ="Rel. 4.6.0, 21 Dec 2007"
> USE_STATS
>
> PostgreSQL 8.4.1 on x86_64-redhat-linux-gnu, compiled by GCC gcc (GCC) 4.1.2
> 20
> 071124 (Red Hat 4.1.2-42), 64-bit
>
>
> They say the table was created by doing a
>
> create table badtable as select * from goodtable order by addr_num_tlid
> limit 100000;
>
>
> the goodtable doesn't exhibit this bizarre behavior.  Not sure if this is a
> PostgreSQL bug or PostGIS bug.
>
> Any thoughts?
>
>
> Thanks,
> Regina
>
>
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-devel
>