Exact equality for floating point values seems to be processor/compiler sensitive, this causes correct tests to fails. Replacing EXPECT_EQ with EXPECT_NEAR with a tight tolerance fixes this.