Eigen SVD strange failure in Google test

Question

I'm executing next code in google test

TEST(LinearALgebra, SVD) {
    Eigen::Matrix3d m;

    m << -0.0793311, 0.999997, -1.17221e-07,
         1.74, 0.00249557, 0.000445095,
         9.88131e-324, 0.000191222, -0.000284459;

    Eigen::Matrix3d m_inv = m.inverse();

    //or

    auto svd = m.jacobiSvd(Eigen::ComputeFullU | Eigen::ComputeFullV);
}

It failed in BinaryFunctors.h in scalar_product_op::result_type (line 86) The same operation in simpe main application, python numpy or opencv works correct without failure.

Answer 1

So, the problem was with subnormal number 9.88131e-324. It leads to floating point exception. Changing it to zero solved problem.