Hello Marko,
It seems that the anti-island check got the piecewise defined coast line of Inarinjärvi (Lake Inari) wrong. I now defined a multipolygon for the lake and its about 700 islands, hopefully making no mistake in the process: http://www.openstreetmap.org/browse/relation/402543
When you say "wrong", what exactly do you mean?
I mean these messages from --generate-sea=multipolygon,no-sea-sectors:
2010/02/06 12:16:26 WARNING (Osm5XmlHandler): 63240003.osm.gz: Converting anti-island starting at http://www.openstreetmap.org/?mlat=68.79379&mlon=27.95413&zoom=17 into an island as it is surrounded by water 2010/02/06 12:16:26 WARNING (Osm5XmlHandler): 63240003.osm.gz: Converting anti-island starting at http://www.openstreetmap.org/?mlat=69.41525&mlon=28.71751&zoom=17 into an island as it is surrounded by water 2010/02/06 12:16:26 WARNING (Osm5XmlHandler): 63240003.osm.gz: Converting anti-island starting at http://www.openstreetmap.org/?mlat=66.56882&mlon=27.78983&zoom=17 into an island as it is surrounded by water
Today, I started from the first one and defined the Inarinjärvi multipolygon. A couple of days ago, I checked all anti-island warnings and fixed the genuine ones. These three cases look like a lake mistaken for an anti-island.
I do not blame mkgmap; I think that the right thing is to define multipolygons for all lakes.
Hmm, I wonder if the test to see if an anti-island is contained within an island is failing if the anti-island reaches a tile edge. If you still have the same data, could you try modifying containsPointsOf() in Way.java so that it looks like this: public boolean containsPointsOf(Way other) { Polygon thisPoly = new Polygon(); for(Coord p : points) thisPoly.addPoint(p.getLongitude(), p.getLatitude()); for(Coord p : other.points) if(thisPoly.contains(p.getLongitude(), p.getLatitude())) return true; return false; } i.e. make it return true if any point is inside rather than all points are inside. Mark