Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <bde3d2dc933848bbaceeb9b7102f6f4c@AcuMS.aculab.com>
 <b4ce9dad748e489f9314a2dc95615033@AcuMS.aculab.com> <9c14002108ff49e1a54819133fe9e2a3@AcuMS.aculab.com>
In-Reply-To: <9c14002108ff49e1a54819133fe9e2a3@AcuMS.aculab.com>
From:   Nick Desaulniers <ndesaulniers@google.com>
Date:   Tue, 1 Aug 2023 09:02:06 -0700
Message-ID: <CAKwvOdm6ZVSMpzSJ0tFMR_rVb1+=REk8ETLS2SBe3s8QrbYUQw@mail.gmail.com>
Subject: Re: [PATCH next v2 5/5] minmax: Relax check to allow comparison
 between int and small unsigned constants.
To:     David Laight <David.Laight@aculab.com>
Cc:     "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        "Matthew Wilcox (Oracle)" <willy@infradead.org>,
        Christoph Hellwig <hch@infradead.org>,
        "Jason A. Donenfeld" <Jason@zx2c4.com>,
        Nathan Chancellor <nathan@kernel.org>,
        "llvm@lists.linux.dev" <llvm@lists.linux.dev>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Jul 31, 2023 at 6:27=E2=80=AFAM David Laight <David.Laight@aculab.c=
om> wrote:
>
> By the time I've done:
>
> #define __is_noneg_int(x)                                       \
>         __builtin_choose_expr(!__is_constexpr(x), false,        \
>                 ((x) >=3D (typeof(x))0 && (x) <=3D (typeof((x) + 0))(long=
)__INT_MAX__))
>
> #define __is_signed(x)                                                   =
      \
>        __builtin_choose_expr(__is_constexpr(is_signed_type(typeof(x))),  =
      \
>                is_signed_type(typeof(x)), 0)
>
> #define __types_ok(x, y)                                        \
>         (__is_signed(x) =3D=3D __is_signed(y) ||                    \
>                 __is_signed((x) + 0) =3D=3D __is_signed((y) + 0) || \
>                 __is_noneg_int(x) || __is_noneg_int(y))
>
> the error message for
>
> >       static_assert(__types_ok(x, y),                 \
> >               #op "(" #x ", " #y ") signedness error, fix types or cons=
ider " #op "_unsigned() before " #op "_t()"); \
>
> generated by clang 8.0.0 and later is similar to (see https://godbolt.org=
/z/jq613Gnsa):
>
> <source>:49:12: error: static assertion failed due to requirement '__buil=
tin_choose_expr((sizeof(int) =3D=3D sizeof (*(8 ? ((void *)((long)((((int)(=
-1)) < (int)1)) * 0L)) : (int *)8))), (((int)(-1)) < (int)1), 0) =3D=3D __b=
uiltin_choose_expr((sizeof(int) =3D=3D sizeof (*(8 ? ((void *)((long)((((un=
signed int)(-1)) < (unsigned int)1)) * 0L)) : (int *)8))), (((unsigned int)=
(-1)) < (unsigned int)1), 0) || __builtin_choose_expr((sizeof(int) =3D=3D s=
izeof (*(8 ? ((void *)((long)((((int)(-1)) < (int)1)) * 0L)) : (int *)8))),=
 (((int)(-1)) < (int)1), 0) =3D=3D __builtin_choose_expr((sizeof(int) =3D=
=3D sizeof (*(8 ? ((void *)((long)((((unsigned int)(-1)) < (unsigned int)1)=
) * 0L)) : (int *)8))), (((unsigned int)(-1)) < (unsigned int)1), 0) || (__=
builtin_choose_expr(!(sizeof(int) =3D=3D sizeof (*(8 ? ((void *)((long)(a) =
* 0L)) : (int *)8))), 0, __builtin_choose_expr(__builtin_choose_expr((sizeo=
f(int) =3D=3D sizeof (*(8 ? ((void *)((long)((((int)(-1)) < (int)1)) * 0L))=
 : (int *)8))), (((int)(-1)) < (int)1), 0), a, 0) >=3D 0 && (a) <=3D (int)(=
long)2147483647)) || (__builtin_choose_expr(!(sizeof(int) =3D=3D sizeof (*(=
8 ? ((void *)((long)(2147483648U - 0) * 0L)) : (int *)8))), 0, __builtin_ch=
oose_expr(__builtin_choose_expr((sizeof(int) =3D=3D sizeof (*(8 ? ((void *)=
((long)((((unsigned int)(-1)) < (unsigned int)1)) * 0L)) : (int *)8))), (((=
unsigned int)(-1)) < (unsigned int)1), 0), 2147483648U - 0, 0) >=3D 0 && (2=
147483648U - 0) <=3D (unsigned int)(long)2147483647))': min(a, 0x80000000u =
- 0) signedness error, fix types or consider min_unsigned() before min_t()
>
> Repeating the expression seems somewhat sub-optimal!
> Surely it shouldn't be outputting the expansion of the
> input when an error message is supplied?
>
> Is there any (sane) way to stop it being that verbose?

No, but we can probably change that in clang. Filed:
https://github.com/llvm/llvm-project/issues/64310

>
>         David
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1=
 1PT, UK
> Registration No: 1397386 (Wales)


--=20
Thanks,
~Nick Desaulniers