[mkgmap-dev] Wrong multipolygon warnings

Fri Feb 12 19:10:06 GMT 2010

Am 12.02.2010 20:06, schrieb Carlos Dávila:
> WanMil escribió:
>>> Compiling Spain I get several warnings similar to the one below, that
>>> point to data that are correct (or it seems to me).
>>>
>>> |2010/02/12 09:26:39 ADVERTENCIA (MultiPolygonRelation):
>>> 63240004.osm.gz: Multipolygon
>>> http://www.openstreetmap.org/browse/relation/319017 contains errors.
>>> 2010/02/12 09:26:39 ADVERTENCIA (MultiPolygonRelation): 63240004.osm.gz:
>>> Polygon 4611686018427391078[5P : (44132028[5P]) carries role inner but
>>> is not inside any outer polygon. Potentially it does not belong to this
>>> multipolygon.
>>>
>>> |Reported way (hole in the building) lacks in the resulting map.
>>> Is it a know bug?
>>> Another question, why are polygons so deformed in MapSource compared to
>>> Mapnik?
>>>
>>>
>> Carlos,
>>
>> the way http://www.openstreetmap.org/browse/way/44132028 overlaps the
>> outer way of multipolygon
>> http://www.openstreetmap.org/browse/relation/319017. If we look keenly
>> on the definition of the multipolygon relation
>> (http://wiki.openstreetmap.org/wiki/Relation:multipolygon) this is not
>> allowed. The current implementation of mkgmap does not support this.
>>
> Way 44132028 doesn't really overlap the outer polygon. They are
> separated some 25 cm, but mkgmap seems to consider them overlapping.

~25cm is much below mkgmap resolution (I think so...)

>> But it is widely used. There is a patch
>> (http://www.mkgmap.org.uk/pipermail/mkgmap-dev/2010q1/007086.html) which
>> is not yet committed and which fixes the problem. I think it will be
>> committed within the next days.
>>
> I'll test it and give feedback

Thanks! That will speed up the commit.

>> The deformations do not look good. I think the problem is that mkgmap
>> rounds the exact OSM coordinates to lower resolution garmin coordinates.
>>
> May it be also the origin of the overlapping problem in the above case?

Yes, propably.