[mkgmap-dev] [PATCH v2] Remove LineClipper artefacts

Mon May 23 21:30:45 BST 2011

I am not sure if this comes from updating mkgmap-trunk from r1946 to 
r1953 or from this patch v2, but I get these additional warnings:

2011/05/23 22:51:09 WARNING (MultiPolygonRelation): 63240002.osm.gz: 
Cannot join the following ways to closed polygons. Multipolygon 
http://www.openstreetmap.org/browse/relation/36306
2011/05/23 22:53:04 WARNING (MultiPolygonRelation): 63240004.osm.gz: 
Cannot join the following ways to closed polygons. Multipolygon 
http://www.openstreetmap.org/browse/relation/38101
2011/05/23 22:54:18 WARNING (MultiPolygonRelation): 63240006.osm.gz: 
Cannot join the following ways to closed polygons. Multipolygon 
http://www.openstreetmap.org/browse/relation/900367
2011/05/23 22:54:18 WARNING (MultiPolygonRelation): 63240006.osm.gz: - 
way: http://www.openstreetmap.org/browse/way/57746082
2011/05/23 22:54:18 WARNING (MultiPolygonRelation): 63240006.osm.gz: 
Multipolygon http://www.openstreetmap.org/browse/relation/900367 does 
not contain any way tagged with role=outer or empty role.

I took a look at these relations in JOSM (ctrl-L, enter
http://api.openstreetmap.org/api/0.6/relation/900367/full
and so on).

Relations 36306 and 38101 are boundary relations, which include some 
maritime borders in the south. They might not be completely inside the 
Geofabrik extract of Finland. I do not care about them.

The outer line of 900367 (one of the multipolygons for Lake Saimaa) 
carried natural=water, which I now removed, as the multipolygon relation 
already carries that attribute.  The roles of the relation are correct: 
all member ways are either role=outer or role=inner. JOSM does not 
complain.  I guess that mkgmap is just dropping the role=outer line, 
because it is incomplete in the tile:

63240006: 2832510,1317344 to 2916352,1462272
#       : 60.779070,28.267136 to 62.578125,31.376953

The bbox of the role=outer way is 62.69143,28.0395 to 62.516,28.46425.  
That is, the outer line is slightly inside the tile. To be exact, the 
following nodes (two bays) of the outer way are inside the tile:

244270918 244272962 244270916 244272960 244270922 244270920 244272964 244270915 244270919 244272963 244272961 244270923 359093973 244270921

One island is partly at the NW corner of the tile: way 26228126. The 
other bay does not contain islands in the map data. Now, let me check 
62.578125,28.2671360 in the generated map. Sure enough, a section of the 
lake is missing SE of the island. I guess it is showing just the 
background there.

Am I right that this is a splitter problem (possibly fixable by 
increasing the split-radius)?

	Marko