Re: m32c support for named addr spaces branch

[Richard Guenther <richard.guenther@gmail.com>]

On Mon, Oct 19, 2009 at 4:24 PM, Ulrich Weigand <uweigand@de.ibm.com> wrote:
> 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?

If the above is expanding a POINTER_PLUS_EXPR then the offset
needs to be extended as _signed_ always, independent of the
sign of sizetype which is the type of the offset (as sizetypes
are always sign-extended).

Richard.