(the problem was that the scaling factor was "floored" to 0.1 precision, and not "rounded", so that with big scales it was floored to 0.... and numerical hell followed. (dividing by 0 led to "inf" values -which did not provoke the crash, but led to "nan" values, which did-)
(the problem was that the scaling factor was "floored" to 0.1 precision, and not "rounded", so that with big scales it was floored to 0.... and numerical hell followed. (dividing by 0 led to "inf" values -which did not provoke the crash, but led to "nan" values, which did-)