Using asinf may leed to nan (not a number)

[the-m-master]

Using 'asinf()' may lead to nan. In the function 'FusionQuaternionToEuler' a call is done to 'asinf', if the input of this is larger then 1 or less then -1, the result will be nan. All calculations using this pitch value (as a nan) will fail as well.

A potential solution might be something like:

float asinf_safe(float val) {
if (val >= 1.0f) { return (M_PI / 2.0f); }
if (val <= -1.0f) { return (M_PI / -2.0f); }
return asinf(val);
}

[xioTechnologies]

Thank you for raising the issue. The problem can be demonstrated using the following code. The plot ends after 45 samples because the remaining 55 samples are all NaN.

import imufusion
import matplotlib.pyplot as pyplot
import numpy

numberOfSamples = 100

ahrs = imufusion.Ahrs()
pitch = numpy.empty(numberOfSamples)

for index in range(numberOfSamples):
    ahrs.update_no_magnetometer(numpy.array([0, 0, 0]), numpy.array([1, 0, 0]), 1 / numberOfSamples)
    pitch[index] = ahrs.quaternion.to_euler()[1]

pyplot.plot(pitch)
pyplot.show()

After implementing the fix you have suggested, the plot shows all 100 samples as shown below.

As an aside, it is interesting to note that the transition from the final asinf result (-89.5) to the hardcoded -90.0f is visible in the above plot as step change. The transition becomes smoother when FUSION_USE_NORMAL_SQRT is used as shown below.

[xioTechnologies]

The fix has been applied as commit d7788de.

[the-m-master]

Thanks! We are going to use it in the next release :)