After compiling all of the libraries with -O2 -m486, here are the
changes I made.
 The first change is to fix the "overflow in constant" by using
HUGE_VAL (infinity) instead of 1e300 * 1e300.
 The second change is using zero instead of tiny * tiny.

Onno van der Linden       c/o    vdlinden@fwi.uva.nl (Frank van der Linden)
*** /usr/src/lib/libm/common/pow.c.old	Sun Jan  2 17:50:35 1994
--- /usr/src/lib/libm/common/pow.c	Sun Jan  2 18:28:18 1994
***************
*** 176,180 ****
  {
  	struct Double s, t, log__D();
! 	double  exp__D(), huge = 1e300, tiny = 1e-300;
  
  	if (x == zero)
--- 176,180 ----
  {
  	struct Double s, t, log__D();
! 	double  exp__D();
  
  	if (x == zero)
***************
*** 182,186 ****
  			return (zero);
  		else if (_IEEE)
! 			return (huge*huge);
  		else
  			return (infnan(ERANGE));
--- 182,186 ----
  			return (zero);
  		else if (_IEEE)
! 			return (HUGE_VAL);
  		else
  			return (infnan(ERANGE));
***************
*** 191,202 ****
  			return (zero);
  		else if (_IEEE)
! 			return (huge*huge);
  		else
  			return (infnan(ERANGE));
  	if (y >= 7e18)		/* infinity */
  		if (x < 1)
! 			return(tiny*tiny);
  		else if (_IEEE)
! 			return (huge*huge);
  		else
  			return (infnan(ERANGE));
--- 191,202 ----
  			return (zero);
  		else if (_IEEE)
! 			return (HUGE_VAL);
  		else
  			return (infnan(ERANGE));
  	if (y >= 7e18)		/* infinity */
  		if (x < 1)
! 			return(zero);
  		else if (_IEEE)
! 			return (HUGE_VAL);
  		else
  			return (infnan(ERANGE));

------------------------------------------------------------------------------