> Not sure.  As I said, it's not easy.  As far as I understand things,
> INTERNALIZE_SYMBOL() should look at n_type to decide whether n_value
> should be treated as a signed or an unsigned value.  The problem seems
> to be that n_value can either be interpreted as an address or as an
> offset.  Addresses should be sign-extended based on what
> bfd_get_sign_extend_vma() tells us.  Offsets should probably always be
> sign-extended.
> 
> Mark
> 

The gdb stabs docu says:

    4.1 Automatic Variables Allocated on the Stack

    The value of the stab is the offset of the variable within the
    local variables. On most machines this is an offset from the
    frame pointer and is negative.

I think we could sign extend e_value if bfd_get_sign_extend_vma() tells 
us OR if n_type == N_LSYM.

N_LSYM is also used for types. But e_value seems to be unused and = 0 in 
this case.