VincentyGeodesy.distanceInMeters can result in NaN for valid input coordinates

[satorg]

Sometimes VincentyGeodesy.distanceInMeters produces NaN for pretty valid input points.
There are examples (not a complete list!) of arguments for distanceInMeters below that result it to generate NaN:

foo=(0.0,0.0), bar=(0.0,-180.0)
foo=(1.0,180.0), bar=(0.0,1.0)
foo=(0.0,0.0), bar=(0.0,180.0)
foo=(-1.0,1.0), bar=(0.0,-180.0)
foo=(-1.0,-103.0), bar=(0.0,78.0)
foo=(-1.0,-106.0), bar=(0.0,75.0)
foo=(-1.0,-142.0), bar=(0.0,39.0)
foo=(-1.0,-155.0), bar=(0.0,26.0)
foo=(-1.0,-161.0), bar=(0.0,18.0)
foo=(-1.0,-164.0), bar=(0.0,17.0)
foo=(-1.0,-176.0), bar=(1.0,4.0)
foo=(-1.0,-178.0), bar=(0.0,1.0)
foo=(-1.0,-179.0), bar=(0.0,0.0)
foo=(-1.0,-30.0), bar=(0.0,149.0)
foo=(-1.0,-33.0), bar=(0.0,148.0)
foo=(-1.0,-50.0), bar=(1.0,130.0)
foo=(-1.0,-74.0), bar=(0.0,105.0)
foo=(-1.0,-79.0), bar=(1.0,101.0)
foo=(-1.0,-93.0), bar=(0.0,86.0)
foo=(-1.0,0.0), bar=(0.0,-179.0)
foo=(-1.0,0.0), bar=(0.0,179.0)
foo=(-1.0,1.0), bar=(0.0,-178.0)
foo=(-1.0,1.0), bar=(1.0,-179.0)
foo=(-1.0,110.0), bar=(0.0,-69.0)
foo=(-1.0,144.0), bar=(0.0,-35.0)
foo=(-1.0,158.0), bar=(1.0,-22.0)
foo=(-1.0,167.0), bar=(1.0,-13.0)
foo=(-1.0,174.0), bar=(0.0,-5.0)
foo=(-1.0,176.0), bar=(0.0,-3.0)
foo=(-1.0,179.0), bar=(0.0,0.0)
foo=(-1.0,33.0), bar=(0.0,-148.0)
foo=(-1.0,46.0), bar=(0.0,-135.0)
foo=(-1.0,47.0), bar=(0.0,-134.0)
foo=(-1.0,5.0), bar=(1.0,-175.0)

The list above is generated with help of ScalaCheck.

Version: 1.4.0

[mrjonjonjon]

This probably happens because "Between two nearly antipodal points, the iterative formula may fail to converge". Perhaps in such cases, half the circumference of earth should be used.

[satorg]

@mrjonjonjon yes, looks like so... But seems there's no clear and ready-to-use solution yet.