[mkgmap-dev] [PATCH v4] sea polygons

Thu Aug 20 22:29:13 BST 2009

On 09-08-20 21:42:23 CEST, Christian Gawron wrote:
> Thanks for the feedback!
> 
> One trivial bug in the patch was that a tile with no shoreline at all 
> got a sea polygon as "background".
> This can be fixed by adding
>         // don't do anything if there is no shoreline
>         if (shoreline.size() == 0)
>         return;
> at the begin of generateSeaPolygon().

great!
is was about to report this.
to facilitate testing, i've created smaller areas to investigate the
problems (now recreated with the above patch):
- brandenburg in germany: fully flooded before, now fine with above fix.
- (same with prague)
- a bbox part of the adriatic sea, the kvarner gulf:
  - there is an invisible diagonal line across the whole tile.
    - northeast of it land and water are reversed
    - southwest of it many islands are fine
      - some islands in the otherwise good half vanish in the sea, e.g.
	lastovo at about 42.75/16.85 (and its sister islands).
	i wonder why, as lastovo for example seems to have a closed
	natural=coastline made up of three ways joined at three nodes:
	  way_10435346
	  node_89314868
	  way_10435380
	  node_89319038
	  way_10435451
	  node_89314756
	  (way_10435346)
  - the tile has a triangular area of italy in the SW corner, some 20 km
    of coast line:
    - it is mostly flooded, with the exception of what seems to be a
      fake strip of beach, some 1 km wide, in front of the coast.

the bbox i out for the last part: left=14 bottom=42.5 right=18.5 top=45.5
the splitter produced one tile only.
the diagnostics from mkgmap (the three lines repeat three times):

SEVERE (Osm5XmlHandler): Non-closed coastline segment does not hit bounding box - expect strange results: 65856087 (44.80250/14.00158) 65860906 (44.88756/14.00263) http://www.openstreetmap.org/?lat=44.80250&lon=14.00158&zoom=17

SEVERE (Osm5XmlHandler): Non-closed coastline segment does not hit bounding box - expect strange results: 65865448 (44.89250/14.00023) 51492065 (42.50074/18.34468) http://www.openstreetmap.org/?lat=44.89250&lon=14.00023&zoom=17

SEVERE (Osm5XmlHandler): Non-closed coastline segment does not hit bounding box - expect strange results: 33051020 (42.50323/14.17635) 33051034 (42.70034/14.00160) http://www.openstreetmap.org/?lat=42.50323&lon=14.17635&zoom=17

44.89250/14.00023 and 42.50074/18.34468 seem to be the ending points of
the above mentioned diagonal line.

42.50323/14.17635 and 42.70034/14.00160 seem to be the ending points of
the virtual line through the triangle of italy mentioned above.

please tell me if i can help you with any further research.

thank you,
rj