This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix IVOPTs problem, PR23177
On Mon, 2005-08-01 at 14:04 -0400, Daniel Berlin wrote:
> On Mon, 2005-08-01 at 11:31 -0600, Jeffrey A Law wrote:
> > On Mon, 2005-08-01 at 16:01 +0200, Richard Guenther wrote:
> > > This teaches get_tmr_operands to eventually cope with tags that
> > > were created from bases that have subvars.
> > >
> > > This patch was bootstrapped and regtested on x86_64-unknown-linux-gnu
> > > as part of the array aliasing patch. Bootstrapping seems to be broken
> > > at the moment.
> > >
> > > Obvious enough? Ok for mainline?
> > I don't think it's obvious, far from it.
> >
> > At the core, I don't see why you want to treat _DECL nodes
> > differently in this code.
>
> We should probably be *always* calling get_expr_operands, unless we know
> for a fact that nothing that may access subvars gets into the tag
> variable.
>
> Definitely, the current add_stmt_operand is wrong for the case when you
> have something with subvars, because it won't add those subvars.
>
> IE the patch posted is more correct than what we do now, but it may not
> go far enough.
I'd be a lot more inclined to approve a tested patch which called
get_expr_operands unconditionally -- unless someone can explain
why that would be a bad idea.
jeff