On Thu, Oct 10, 2019 at 06:06:40PM +0000, Bernd Edlinger wrote:
> On 10/10/19 7:49 PM, Jason Merrill wrote:
> > On 10/10/19 10:42 AM, Bernd Edlinger wrote:
> >> Hi,
> >>
> >> this fixes a crash when -Wshadow=compatible-local is
> >> enabled in the testcase g++.dg/parse/crash68.C
> >
> > Why does that flag cause this crash?
> >
>
> gcc/cp/name-lookup.c:
>
> if (warn_shadow)
> warning_code = OPT_Wshadow;
> else if (warn_shadow_local)
> warning_code = OPT_Wshadow_local;
> else if (warn_shadow_compatible_local
> && (same_type_p (TREE_TYPE (old), TREE_TYPE (decl))
> || (!dependent_type_p (TREE_TYPE (decl))
> && !dependent_type_p (TREE_TYPE (old))
> /* If the new decl uses auto, we don't yet know
> its type (the old type cannot be using auto
> at this point, without also being
> dependent). This is an indication we're
> (now) doing the shadow checking too
> early. */
> && !type_uses_auto (TREE_TYPE (decl))
> && can_convert (TREE_TYPE (old), TREE_TYPE (decl),
> tf_none))))
> warning_code = OPT_Wshadow_compatible_local;
>
> if -Wshadow=compatible-local is used, the can_convert function crashes
> in instantiate_class_template_1.
>
> The problem there is that CLASSTYPE_TI_TEMPLATE (type)
> uses CLASSTYPE_TEMPLATE_INFO (type) but that is NULL.
... because the warning is called from pushtag which ran before building the
template info for 'c':
9926 // Build template info for the new specialization.
9927 SET_TYPE_TEMPLATE_INFO (t, build_template_info (found, arglist));
Looks like another indication that the shadow checking is done too early.
> Since other errors may return error_mark_node as well
> and the template is totally erroneous anyway, I figured
> we might get away with that simple fix.
But it doesn't have to be broken; this valid test crashes with that option too
template<typename>
struct S {
S () {
struct c;
{
struct c {};
}
}
};
S<int> s;
--
Marek Polacek • Red Hat, Inc. • 300 A St, Boston, MA
