On Wed, Nov 25, 2015 at 01:57:13PM -0500, Jason Merrill wrote:
> The _FUN function returned by the lambda conversion operator calls the op()
> with a null object argument, which would be undefined behavior for user
> code, but we can get away with it internally since it isn't observable.
> Tell UBsan to ignore it.
>
> Tested x86_64-pc-linux-gnu, applying to trunk and 5.
Thanks. Though, perhaps we could guard those new two stmts with
if (flag_sanitize & SANITIZE_UNDEFINED)
? No need to generate a useless attribute in the common case.
> commit 003d84b026a4785cbbb2b1cfc7fda9a31792d7a4
> Author: Jason Merrill <ja...@redhat.com>
> Date: Wed Nov 25 12:37:39 2015 -0500
>
> PR c++/67941
> * lambda.c (maybe_add_lambda_conv_op): Mark _FUN as
> no_sanitize_undefined.
>
> diff --git a/gcc/cp/lambda.c b/gcc/cp/lambda.c
> index 5fe3473..f9b686b 100644
> --- a/gcc/cp/lambda.c
> +++ b/gcc/cp/lambda.c
> @@ -1053,6 +1053,12 @@ maybe_add_lambda_conv_op (tree type)
> if (generic_lambda_p)
> fn = add_inherited_template_parms (fn, DECL_TI_TEMPLATE (callop));
>
> + /* Don't UBsan this function; we're deliberately calling op() with a null
> + object argument. */
> + tree attrs = build_tree_list (get_identifier ("no_sanitize_undefined"),
> + NULL_TREE);
> + cplus_decl_attributes (&fn, attrs, 0);
> +
> add_method (type, fn, NULL_TREE);
>
> if (nested)
> diff --git a/gcc/testsuite/g++.dg/ubsan/null-6.C
> b/gcc/testsuite/g++.dg/ubsan/null-6.C
> new file mode 100644
> index 0000000..2efe5c2
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ubsan/null-6.C
> @@ -0,0 +1,5 @@
> +// PR c++/67941
> +// { dg-do run { target c++11 } }
> +// { dg-options -fsanitize=null }
> +
> +int main(){ (+[](){})(); }
Jakub
