David Scherfgen

Yes to both. "Sum" should be "difference", as the reproducer shows. And the forms are equal for a real x only; the request is for the case the simplifier already treats as real: the clauses in timesin that turn abs(x)/x^2 into 1/abs(x) and x^2/abs(x)^3 into 1/abs(x) test csign of the argument and leave a complex z alone, and the missing case is the same rule for the exponent 1, in the same clauses, under the same test. With declare(z, complex), abs(z)/z and z/abs(z) would stay distinct, as they ...

The sign of x has two normal forms, x/abs(x) and abs(x)/x, and their difference does not simplify

(enhancement) Let a derivative defined by a DEFGRAD lambda return a final expression

gamma_incomplete_lower(a, z) with a rational order a and a float z is not evaluated

carg, polarform and rectform of a product do not reduce the argument to (-%pi, %pi]

rectform(atan2(y, 0)) is wrong for negative y and carries a spurious imaginary part

limit(%i*x + sqrt(x), x, inf) stays noun-form

Fixed by commit [d6f593].