* H. J. Lu:
> @@ -8106,6 +8109,34 @@ grokdeclarator (const struct c_declarator *declarator,
> else
> {
> /* It's a variable. */
> +
> + const char *stack_protect_guard_name
> + = targetm.stack_protect_guard_symbol;
> + if (stack_protect_guard_name)
> + {
> + /* If the variable name is the same as the stack protect
> + guard, use the stack protect guard decl. */
> + tree guard_decl = targetm.stack_protect_guard ();
> + if (guard_decl
> + && VAR_P (guard_decl)
> + && strcmp (stack_protect_guard_name,
> + IDENTIFIER_POINTER (declarator->u.id.id)) == 0)
> + {
> + /* Allow different types with the same size. */
> + tree guard_type = TREE_TYPE (guard_decl);
> + if (TYPE_PRECISION (type)
> + != TYPE_PRECISION (guard_type))
> + {
> + error ("conflicting types for %s; have %qT, "
> + "should be %qT",
> + stack_protect_guard_name, type,
> + guard_type);
> + return error_mark_node;
> + }
> + return guard_decl;
> + }
> + }
It seems to me this goes in the wrong direction: guard_decl doesn't have
proper location information, so any subsequent diagnostics against it
will look wrong. I would expect this to be more like the re-declaration
of builtins.
But I'm not a GCC expert.
Thanks,
Florian
