re PR fortran/24096 (huge() returns infinity for long doubles)

	PR fortran/24096
	* trans-types.c (gfc_init_kinds): Use one less for max_exponent
	of IBM extended double format.

From-SVN: r107041

gcc/fortran/ChangeLog

+2005-11-16  Alan Modra  <amodra@bigpond.net.au>
+
+	PR fortran/24096
+	* trans-types.c (gfc_init_kinds): Use one less for max_exponent
+	of IBM extended double format.
+
 2005-11-13  Francois-Xavier Coudert  <coudert@clipper.ens.fr>
 
 	* intrinsic.c (add_functions): Add COMPLEX, FTELL, FGETC, FGET,

gcc/fortran/trans-types.c

@@ -192,6 +192,15 @@ gfc_init_kinds (void)
       gfc_real_kinds[r_index].digits = fmt->p;
       gfc_real_kinds[r_index].min_exponent = fmt->emin;
       gfc_real_kinds[r_index].max_exponent = fmt->emax;
+      if (fmt->pnan < fmt->p)
+	/* This is an IBM extended double format (or the MIPS variant)
+	   made up of two IEEE doubles.  The value of the long double is
+	   the sum of the values of the two parts.  The most significant
+	   part is required to be the value of the long double rounded
+	   to the nearest double.  If we use emax of 1024 then we can't
+	   represent huge(x) = (1 - b**(-p)) * b**(emax-1) * b, because
+	   rounding will make the most significant part overflow.  */
+	gfc_real_kinds[r_index].max_exponent = fmt->emax - 1;
       gfc_real_kinds[r_index].mode_precision = GET_MODE_PRECISION (mode);
       r_index += 1;
     }