I've just run across what appears to be a fairly serious bug in the evaluation of
a particular matrix expression involving a matrix transpose and some scalars. I'm guessing it may not be known since I can reproduce
it in 3.0.2. Here is a simple test case:

void prodTest()
{
Eigen::Matrix2Xd dNdxy(2, 3);
dNdxy <<
-0.5, 0.5, 0,
-0.333333, 0, 0.333333;
double det=6.0, wt=.5;

Eigen::MatrixXd p1 = dNdxy.transpose()*dNdxy*det*wt;
Eigen::MatrixXd p2 = det*wt*dNdxy.transpose()*dNdxy;
std::cout << p1 << std::endl;
std::cout << p2 << std::endl;
}

p2 is correct but p1 is not.

Bill

I can reproduce it, both with and without optimizations. That is quite a bad bug. It looks like the factor det in the expression for p1 is ignored (perhaps overwritten by the factor wt ?)

Thanks very much for letting us know.

Thanks a lot jitse for fixing this.... scary. This fix justify a 3.0.3 asap.