Re: m32c support for named addr spaces branch

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

On Tue, Oct 20, 2009 at 8:09 PM, DJ Delorie <dj@redhat.com> wrote:
>
>> I don't think this observation is correct.  sizetypes are _always_
>> sign-extended.
>
> The problem on m32c is like this: size_t is HImode, pointers are
> PSImode (24 bits).  Loop optimization sees this:
>
>        char *p = x;
>        for (i=0; i<j; i++)
>          *p++ = n;
>
> It replaces the 'i' loop with a 'p'-based loop.  It pre-calculates the
> end value of 'p'.  However, it sign extends 'j' before adding it to
> 'p', or worse, casts 'p' to size_t, does the math, the re-casts it
> back to a pointer.
>
> Thus, you end up with the termination value of 'p' being truncated or
> mis-computed, and the loop overruns.
>
> My preference is this: pointers and integers are NOT interchangable.
> Pointer offsets must be sign or zero extended according to their
> underlying types before being added to pointers.  If I add -1 to a
> pointer, it had better result in a pointer that's one less than the
> previous value, not 65536 more.

Well, you are right.  sizetype originally only was present in sizetype-only
operations where the above would not have been an issue.  Now
with POINTER_PLUS_EXPR we introduced a mixed-type operation
between pointer types and sizetype - which was obviously a mistake.
Thus, the correct thing to do is to transition POINTER_PLUS_EXPR
away from using sizetype but instead use regular integer types for
the offset (I'm not sure if constraining the offset type to a single type
was a good idea - it probably was good to catch transition errors).

But I don't see anyone doing the required work.

Richard.