OK.
On Tue, Jan 3, 2017 at 5:42 PM, Jakub Jelinek <ja...@redhat.com> wrote:
> Hi!
>
> cp_lexer_new_from_tokens creates lexer that has NULL lexer->buffer,
> calling lexer->buffer->address () therefore is UB (diagnosed by
> --with-boot-config=bootstrap-ubsan).
>
> The following patch fixes this, or Markus offered
> gcc_assert (!lexer->buffer || tp != lexer->buffer->address ());
> instead. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk (or do you prefer Markus' version)?
>
> 2017-01-03 Jakub Jelinek <ja...@redhat.com>
>
> PR c++/71182
> * parser.c (cp_lexer_previous_token): Use vec_safe_address in the
> assertion, as lexer->buffer may be NULL.
>
> * g++.dg/cpp0x/pr71182.C: New test.
>
> --- gcc/cp/parser.c.jj 2017-01-03 09:58:13.000000000 +0100
> +++ gcc/cp/parser.c 2017-01-03 11:39:56.940595981 +0100
> @@ -766,7 +766,7 @@ cp_lexer_previous_token (cp_lexer *lexer
> /* Skip past purged tokens. */
> while (tp->purged_p)
> {
> - gcc_assert (tp != lexer->buffer->address ());
> + gcc_assert (tp != vec_safe_address (lexer->buffer));
> tp--;
> }
>
> --- gcc/testsuite/g++.dg/cpp0x/pr71182.C.jj 2017-01-03 11:44:19.400246340
> +0100
> +++ gcc/testsuite/g++.dg/cpp0x/pr71182.C 2017-01-03 11:44:04.795432815
> +0100
> @@ -0,0 +1,12 @@
> +// PR c++/71182
> +// { dg-do compile { target c++11 } }
> +
> +class A {
> + template <typename> void As();
> +};
> +template <typename T> class B : A {
> + void f() {
> + A *g ;
> + g ? g->As<T>() : nullptr;
> + }
> +};
>
> Jakub
