Bug#771705: osm2pgsql: Import fails in PolygonBuilder.cpp:261

[Sebastiaan Couwenberg]

Hi Daniel,

On 12/01/2014 08:22 PM, Daniel Baur wrote:
> I try to import the last planet-dump into the database. The import fails
> with
> 
> --snip----
> 
> Reading in file: /data/dump/planet-latest.osm.pbf
> Processing: Node(2614096k 177.8k/s) Way(261025k 27.43k/s) Relation(46730
> 40.49/s)osm2pgsql: PolygonBuilder.cpp:261: geos::geomgraph::EdgeRing*
> geos::operation::overlay::PolygonBuilder::findShell(std::vector<geos::operation::overlay::MinimalEdgeRing*,
> std::allocator<geos::operation::overlay::MinimalEdgeRing*> >*): Assertion
> `shellCount <= 1' failed.
> 
> --snap---
> 
> after a few hours. If you need more data, please tell.

This looks like an issue in GEOS, not osm2pgsql.

Searching for the error message turns up several threads marking this
issue as fixed in GEOS >= 3.3.3. E.g.

https://github.com/openstreetmap/osm2pgsql/issues/124

This change in GEOS removes the `shellCount <= 1' assert, that is
triggered on your system:

http://trac.osgeo.org/geos/changeset/3286

This leads me to think that the mix of squeeze, wheezy and jessie
packages on your system is the actual cause.

To rule out your esoteric mix of package from different Debian releases,
having a smaller test case than importing a full planet dump would be
most helpful. We can't expect anybody to perform full planet imports to
troubleshoot this issue. I for instance don't have the disk space
available for a full planet import.

If you can pinpoint which OSM objects were being processed at the time
of the failure, reproducing the failure with a smaller extract for the
area should be doable.

Reproducing the issue on a proper jessie system would be even more helpful.

Kind Regards,

Bas

-- 
 GPG Key ID: 4096R/E88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146  50D1 6750 F10A E88D 4AF1