Maxima's rat package and "quotient is not exact"

>>>>> "Raymond" == Raymond Toy <raymond.toy at ericsson.com> writes:

    Raymond> In the meantime, does anyone know the name of the function that
    Raymond> converts this internal polynomial representation into standard maxima
    Raymond> form so I can print it out in a more friendly way when debugging?
    Raymond> That would help a lot.

Thanks to Robert Dodier, Andrej Vodopivec, and Richard Fateman for the
hints.  disrep does basically what I want.

After looking a little more, I don't think it's prem that is the
problem.  It's happening higher, probably in the Risch integrator.

If I disrep the polynomials, I get:

    14*a^3-13*%i*sqrt(b-a^2)*a^2-6*b*a+%i*b*sqrt(b-a^2)

and

    2*%i*a+sqrt(b-a^2)

Note the sqrt(b-a^2).  I think prem is computing the right thing, but
as it computes it, assuming the coefficients are not functions of a,
the remainder suddenly is a function of a.  And when subresgcd checks
(I think) that everything is ok, the quotient isn't exact, because we
didn't really have a polynomial in a to begin with.

Bummer.  This makes the problem quite a bit harder.

Ray