[Bug-gsl] incorrect error from gsl_cheb_eval_n_err

[John Salmon]

The calculation error in gsl_cheb_eval_n_err
uses the order stored in the cs struct, rather than the
order used in the current evaluation.  

I believe this patch fixes the problem.

Cheers,
John Salmon

address@hidden cheb]$ diff -Naur eval.c~ eval.c
--- eval.c~     2003-07-25 11:18:10.000000000 -0400
+++ eval.c      2004-12-22 07:23:51.208786196 -0500
@@ -132,14 +132,14 @@
 
   /* Estimate cumulative numerical error */
 
-  for (i = 0; i <= cs->order; i++)
+  for (i = 0; i <= eval_order; i++)
     {
       absc += fabs(cs->c[i]);
     }
 
   /* Combine truncation error and numerical error */
 
-  *abserr = fabs (cs->c[cs->order]) + absc * GSL_DBL_EPSILON;
+  *abserr = fabs (cs->c[eval_order]) + absc * GSL_DBL_EPSILON;
 
   return GSL_SUCCESS;
 }