gdb 5.2.1 ppc/sh problem with newer binutils?

[Kris Warkentin]

This is just a wild stab in the dark hoping that someone might remember 
something off the top of their head.

Shared libs generated with newer binutils >= 2.12.1  seem to have a 
difference in where it thinks the lines of source code are.

***OLD***
int blah /* 0x17d8 */;
unsigned int some_global /* 0x17dc */;
int dllfunc ()
{ /* 0x5f4 */
  /* file /home/kewarken/61/test/libdll.c line 7 addr 0x5f4 */
  /* file /home/kewarken/61/test/libdll.c line 8 addr 0x608 */
  /* file /home/kewarken/61/test/libdll.c line 9 addr 0x60c */
  /* file /home/kewarken/61/test/libdll.c line 10 addr 0x614 */
} /* 0x62c */

***NEW***
int blah /* 0x17e8 */;
unsigned int some_global /* 0x17ec */;
int dllfunc ()
{ /* 0x0 */
  /* file /61/home/kewarken/test/ppc-630/blah/../../libdll.c line 7 addr 
0x0 */
  /* file /61/home/kewarken/test/ppc-630/blah/../../libdll.c line 8 addr 
0x14 */
  /* file /61/home/kewarken/test/ppc-630/blah/../../libdll.c line 9 addr 
0x18 */
  /* file /61/home/kewarken/test/ppc-630/blah/../../libdll.c line 10 
addr 0x20 */
} /* 0x38 */

I'm seeing that newer gdbs don't have a problem with the newer version 
but gdb-5.2.1 (our current shipping version) does.  Anyone remember any 
of this?  I'm wondering if it might be feasible to backport a solution.  
As it stands, I don't see source backtracing out of shared libs and gdb 
is reluctant to step into them.

cheers,

Kris