For me this fixes "make check" with a newer gcc and Eigen.
Instead of taking relative tolerances w.r.t. exact reaction rates, we
use tol as a minimum denominator. This prevents false positives when
a reaction rate of 1e-1600 is off by an order of magnitude.
Slightly bumping up the tolerance in kooij_rate_vec_unit is also
necessary to get float to pass for me.
For me this fixes "make check" with a newer gcc and Eigen.
Instead of taking relative tolerances w.r.t. exact reaction rates, we use tol as a minimum denominator. This prevents false positives when a reaction rate of 1e-1600 is off by an order of magnitude.
Slightly bumping up the tolerance in kooij_rate_vec_unit is also necessary to get float to pass for me.