this code in parse.d is responsible for computing the real number after parsing:

     eprec := max(prec, ulong(exponent(x)) + 4);
     y := toRR(x,eprec);
     if expon > 0 then y = y * pow10(expon,eprec)
     else if expon < 0 then y = y / pow10(-expon,eprec);
     if eprec != prec then y = toRR(y,prec);

Since it uses two floating point operations, it may not be computing the
correctly rounded result.  Ensure that it is.

