Hmm, I suppose we don't need to allow this in SFINAE context. OK.
On Wed, Feb 22, 2017 at 10:24 AM, Marek Polacek <pola...@redhat.com> wrote:
> The problem here is that we aren't properly checking what's been passed to
> __underlying_type. It needs an enum, which the code checks, but that enum
> mustn't be incomplete, otherwise we crash because ENUM_UNDERLYING_TYPE will
> be null until the definition of the enum is completed by finish_enum.
>
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>
> 2017-02-22 Marek Polacek <pola...@redhat.com>
>
> PR c++/79657
> * semantics.c (finish_underlying_type): Bail out for incomplete enums.
>
> * g++.dg/ext/underlying_type12.C: New test.
>
> diff --git gcc/cp/semantics.c gcc/cp/semantics.c
> index 6ba7c13..67f8b92 100644
> --- gcc/cp/semantics.c
> +++ gcc/cp/semantics.c
> @@ -3838,7 +3838,8 @@ finish_underlying_type (tree type)
> return underlying_type;
> }
>
> - complete_type (type);
> + if (!complete_type_or_else (type, NULL_TREE))
> + return error_mark_node;
>
> if (TREE_CODE (type) != ENUMERAL_TYPE)
> {
> diff --git gcc/testsuite/g++.dg/ext/underlying_type12.C
> gcc/testsuite/g++.dg/ext/underlying_type12.C
> index e69de29..050ecf2 100644
> --- gcc/testsuite/g++.dg/ext/underlying_type12.C
> +++ gcc/testsuite/g++.dg/ext/underlying_type12.C
> @@ -0,0 +1,6 @@
> +// PR c++/79657
> +// { dg-do compile { target c++11 } }
> +
> +enum A { x };
> +enum B { a = (__underlying_type (A)) 1 };
> +enum C { b = (__underlying_type (C)) 1 }; // { dg-error "incomplete" }
>
> Marek
