In order to make the tests more robust, we use relative error for equality comparison instead of absolute error or exact comparison. This should fix #113, but I do not have a x86_64-darwin setup to test this.
I also did some cleanup first, I can split that part off into a separate PR if necessary.
In order to make the tests more robust, we use relative error for equality comparison instead of absolute error or exact comparison. This should fix #113, but I do not have a
x86_64-darwin
setup to test this.I also did some cleanup first, I can split that part off into a separate PR if necessary.