Home Original page

Bug in big-endian ARM gas configurations

Daniel Jacobowitz drow@mvista.com
Wed Oct 30 07:47:00 GMT 2002 
On Wed, Oct 30, 2002 at 03:33:14PM +0000, Richard Earnshaw wrote:
> > I was going to just commit this, but I noticed another oddity, so I'd like a
> > second opinion.  We've got this:
> > 
> >        armeb)           cpu_type=arm endian=big ;;
> >        arm*)            cpu_type=arm endian=little ;;
> >        armb*)           cpu_type=arm endian=little ;;
> >        armv*l)          cpu_type=arm endian=little ;;
> >        armv*b)          cpu_type=arm endian=big ;;
> >        xscale*)         cpu_type=arm endian=little ;;
> > 
> > The first problem: arm* before armv*b.  So my armv5teb toolchain was
> > defaulting to little endian.  Oops!
> > 
> > Before I fix that, though, what's armb*?  And should it really be _little_
> > endian?
> 
> Ug! Where exactly is this list in the sources?

It's in gas/configure.in.

> Apart from anything else it's missing strongarm (since we seem to have 
> XScale).

Copy-pasto:
      strongarm*)       cpu_type=arm endian=little ;;
      thumb*)           cpu_type=arm endian=little ;;
right below that.

> I think the armb* entry should go unless someone can specifically point to 
> a use that indicates that the current definition is correct.

Config.sum recognizes armbe-*.  So does this hunk in
gas/configure{,.in}.  That's it, though; everything else ignores it or
has an arm*-*-linux-gnu* definition to match it.  And I've never
encountered an armb* toolchain; so let's whack it.

How's this:
      armeb)            cpu_type=arm endian=big ;;
      armv*l)           cpu_type=arm endian=little ;;
      armv*b)           cpu_type=arm endian=big ;;
      arm*)             cpu_type=arm endian=little ;;
      xscale*)          cpu_type=arm endian=little ;;
      strongarm*)       cpu_type=arm endian=little ;;
      thumb*)           cpu_type=arm endian=little ;;

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer
More information about the Binutils mailing list