Not sure if this is the right category. I’m experiencing sporadic test failure on Travis CI. It only occurs with Linux and not MacOS, and only a particular subtest. My tests are using the ≈ operator to compare floats. Some of the float values are not simple and so required me to find enough decimal places to match.
Is there any reason why this operation, or floating point math in general, would produce sporadically different results on Linux for the exact same operations?