[mkgmap-dev] Tile boundary artefacts

Tue Apr 6 11:33:41 BST 2010

On 05/04/10 18:35, WanMil wrote:
> I started to make very basic checks to get the source of the tile bounds
> problems.
>
> I created the attached bbox1.osm and bbox2.osm. Both contain one
> rectangular polygon identical to the bounds. After compiling it with the
> default style I see in MapSource that the underlying land polygon
> (yellow) does not match to the data (blue) (see land_diff.png).
>
> Any ideas why this happens? Do you have any hints how I could proceed
> and where I should continue my investigations?
> Up to now I tried to decompile the img file with the Display project but
> failed on the RGN files (will need some time to investigate this).

I looked at bbox2.  Both the background and polygon rectangles in the
map RGN are the same size at every level.

The issue is probably therefore in the overview map. As this has a
very low resolution it is not possible to accurately represent the map
bounds unless the bounds happen to fall on a low resolution
coordinate.  If the tile is produced by splitter it should be
correctly aligned, but for randomly produced tiles there is likely to
be some misalignment.

When the area in the overview map has to be rounded there are a number
of ways of doing it, each of which work in different cases.  The last
change in this area was r1532, a change made by Mark for a particular
use case.  I think that something a bit more complex is needed, for
example detecting the outside boundaries of a tile set
and rounding outward just on those, leaving the internal boundaries
rounded as they are (they have to fit without overlapping).

Once we have a proper overview map then its resolution could be
increased and this would make the problem a lot less noticeable even if
nothing else changed.

..Steve