dead code in div_q.c?

[paul zimmermann]

       Hi,

together with Raphaël Rieu-Hleft (in cc), we believe we have found some dead code in
mpn/generic/div_q.c around lines 173-182:

          else if (UNLIKELY (qh != 0))
            {
              /* This happens only when the quotient is close to B^n and        
                 mpn_*_divappr_q returned B^n.  */
              mp_size_t i, n;
              n = new_nn - dn;
              for (i = 0; i < n; i++)
		qp[i] = GMP_NUMB_MAX;
              qh = 0;           /* currently ignored */
            }

Indeed, in the case cy <> 0, since cnt < GMP_NUMB_BITS, the most significant limb
new_np[nn] of the dividend has at least one 0 leading bit, and thus is smaller
than new_dp[dn-1] (which is normalized). Therefore qh is always 0 (no approximate
division is performed here).

Raphaël and Paul