On Tue, 5 Apr 2022, 17:44 Hans-Peter Nilsson via Libstdc++, <
libstd...@gcc.gnu.org> wrote:

> Ok to commit?
> -------------- 8< --------------
>
> Without this, for a target where alignment and structure-sizes are by
> default byte-aligned, such as cris-elf, you'll see, in libstdc++.log:
>
> /X/gcc/libstdc++-v3/testsuite/20_util/expected/requirements.cc:127: error:
> static assertion failed
> /X/gcc/libstdc++-v3/testsuite/20_util/expected/requirements.cc:127: note:
> the comparison reduces to '(5 == 2)'
> compiler exited with status 1
> FAIL: 20_util/expected/requirements.cc (test for excess errors)
> Excess errors:
> /X/gcc/libstdc++-v3/testsuite/20_util/expected/requirements.cc:127: error:
> static assertion failed
>
> It seems the intent is a smoke-test and that conditionals for ABI
> properties are out of scope, so best to just delete this particular
> line.
>

The idea is to ensure the object is no larger than necessary.

I think we could use == sizeof(void*)+alignof(void*) which would be correct
everywhere. Does that work for cris-elf?



> libstdc++-v3/:
>         * testsuite/20_util/expected/requirements.cc: Don't test
>         ABI-variant properties of expected<>.
> ---
>  libstdc++-v3/testsuite/20_util/expected/requirements.cc | 1 -
>  1 file changed, 1 deletion(-)
>
> diff --git a/libstdc++-v3/testsuite/20_util/expected/requirements.cc
> b/libstdc++-v3/testsuite/20_util/expected/requirements.cc
> index 485aa338679c..a51a007a4fc3 100644
> --- a/libstdc++-v3/testsuite/20_util/expected/requirements.cc
> +++ b/libstdc++-v3/testsuite/20_util/expected/requirements.cc
> @@ -124,6 +124,5 @@ static_assert( move_assignable< void, G > );
>  // QoI properties
>  static_assert( sizeof(std::expected<char, unsigned char>) == 2 );
>  static_assert( sizeof(std::expected<void, char>) == 2 );
> -static_assert( sizeof(std::expected<void*, char>) == 2 * __alignof(void*)
> );
>  static_assert( alignof(std::expected<void, char>) == 1 );
>  static_assert( alignof(std::expected<void*, char>) == alignof(void*) );
> --
> 2.30.2
>
>

Reply via email to