I committed improved (I hope) functions float_approx_equal and bfloat_approx_equal. I used the method given in Knuth, "The Art of Computer Programming," 3rd edition, page 233. The test suite (Clozure CL) ran OK---let me know if you all see new test failures. Barton