This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [lno-branch]: Patch to fix bootstrap problem on darwin
On Mon, Jan 12, 2004 at 10:30:36AM -0500, Daniel Berlin wrote:
>
> On Jan 12, 2004, at 10:20 AM, Pop Sébastian wrote:
>
> >A better patch, since it has to modify only the behaviour of the
> >folder.
>
> Hmmmm. I'm not sure people will like this.
>
> I modified tree-chrec under the assumption that *we* (being the scalar
> evolution code) don't care about NON_LVALUE_EXPR. That isn't to say
> that something else doesn't.
For the moment the tree-fold-const is an interface between the scalar
analyzer and the folder. If you want I can rename these functions
into scev_fold_* or whatever. I don't want to mess up the chrecs
folder with the strange actions/answers of the constant folder.
> It might have been added to the folder to resolve type casting problems
> or something.
> In fact, I think it was (there was a problem related to bool vs int
> types on powerpc that I think that resolved). I'll take a look at
> tree-ssa and see when the code was added.
> But until someone tells us *why* the folder does that, and if it is
> actually necessary, personally, I think we should just modify
> tree-chrec.
I think that all the folder issues should be kept in the
tree-fold-const files until that moment.
Thanks.