Attempting to CI PR #1615 showed that before Matlab 2019b differences in numeric values of around 10^-15 caused large changes in hashing of objects, so that save/load preservation of hashes did not work.
This ticket seeks to solve the problem by truncating the mantissa of the numbers in question to approx. 6 significant figures during serialization before calculating the hash.
Attempting to CI PR #1615 showed that before Matlab 2019b differences in numeric values of around 10^-15 caused large changes in hashing of objects, so that save/load preservation of hashes did not work.
This ticket seeks to solve the problem by truncating the mantissa of the numbers in question to approx. 6 significant figures during serialization before calculating the hash.