Integer field with values larger than can be held in a 32bit integer

[asger@…]

The problem regards attribute integer precision when using ogr to read
shapefiles having an effectively integer field with values larger than can be
held in a 32bit integer.

In the example data found at http://www.jo-informatik.dk/tmp/gdal_example.zip
there is a column named ("ID") which ArcGIS reports as "double(15,0)", and it
holds a positive integer with 15 digits (12080000240330).
Ogr2ogr maps this column to "numeric(15)" in postgres, but the value 
it actually puts into the table is negative and only 10 digits (-1742763318).