Re: [PATCH][RFC] A GIMPLE and "GIMPLE type system" verifier

["Richard Guenther" <richard dot guenther at gmail dot com>]

On 6/2/07, Jan Hubicka <hubicka@ucw.cz> wrote:
> Hi,
> it is in nature very similar to what I tried back in tree-ssa branch
> time, but I was never so successful to get it close to bootstrapland, so
> we seem to be improving in this respect.
> >
> >  - conversions from integer to pointer types (and vice versa).
> >    I'd like to enforce extension/truncation to be done in integer
> >    type only and not allow (T *)short for example, but require
> >    (T *)(sizetype)short.  [the same would hold true for ENUMERAL_TYPE
> >    and BOOLEAN_TYPE, but I didn't try what the fallout from this would be]
>
> I just wonder what benefits are here for enforcing the sizetype
> conversion here?

The only really problematic case is conversion of a pointer to an
integer type with larger precision.  There we don't know whether
to sign- or zero-extend.  But as Andrew said - this (and the less
interesting cases) are not happening too often.

Richard.