On Feb 11, 2009, at 13:51, Gert Münzel wrote:
Yes, r877 seems to fix it. Actually i only tested it with the simple track i sent, but i will test it with more complicated networks of trails.
Just for my interest (actually i don't no much about the structure of Garmin imgs and what is hold in the NET NOD RGN etc. sections), but how it is possible that one trackpart, which looks ok to me, causes this problem?
It was a bug in mkgmap: A rounding error in the distance calculation between two nodes in the track caused a floating point result of NaN (not a number) instead of zero, because for the particular latitude involved, it turned out that sin(lat)^2+cos(lat)^2 was greater than 1. The NaN in turn was apparently converted to the integer 0, and written to the IMG as such (in two places: the length of the road in NET, and the length of the edge in the routing graph in NOD). Cheers Robert