This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: m32c support for named addr spaces branch
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: dj at redhat dot com (DJ Delorie)
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 19 Oct 2009 16:24:20 +0200 (CEST)
- Subject: Re: m32c support for named addr spaces branch
DJ Delorie wrote:
> Found it, bug in expr.c:
>
> Index: expr.c
> ===================================================================
> --- expr.c (revision 152895)
> +++ expr.c (working copy)
> @@ -4248,13 +4248,13 @@ expand_assignment (tree to, tree from, b
> }
>
> offset_rtx = expand_expr (offset, NULL_RTX, VOIDmode, EXPAND_SUM);
> address_mode
> = targetm.addr_space.address_mode (MEM_ADDR_SPACE (to_rtx));
> if (GET_MODE (offset_rtx) != address_mode)
> - offset_rtx = convert_to_mode (address_mode, offset_rtx, 0);
> + offset_rtx = convert_to_mode (address_mode, offset_rtx, POINTERS_EXTEND_UNSIGNED);
I'm not sure POINTERS_EXTEND_UNSIGNED is the correct test here; the offset
is not itself a pointer type. What if the offset actually *is* a negative
value of a smaller, signed type?
Maybe this should take the signed-ness of the type of "offset" into
account instead.
(B.t.w. there are more instances of computing address offsets that should
probably be handled analogously. Look for other instances of offset_address
in expr.c ...)
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com