Maybe realroots should reject polynomials with floating point coefficients, or it should first convert any floats to rationals? (%i36) realroots (x^2 + 1.0e-45, 1.0b-10); (%o36) [x=0] (%i37) realroots (x^2 + 1/10^45, 1.0b-10); [] Barton