This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: RFC: Fix an infinite loop placing sections in relocatable objects
On 2/16/06, Daniel Jacobowitz <drow@false.org> wrote:
> > > + ULONGEST align = 1 << bfd_get_section_alignment (abfd, sect);
> >
> > I think bfd_get_section_alignment can return values bigger than 31 in
> > general, and this is undefined in this case.
>
> It's only used in math with a CORE_ADDR, which will be the size of a
> bfd_vma. Can bfd_get_section_alignment return a shift greater than the
> size of a bfd_vma? What would that mean?
No, what he means is that the literal "1" is just an 'int', so the
shift will be done in the width of an int. The conversion to ULONGEST
happens later. The fix is to cast the 1 to ULONGEST before shifting
it.