Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200504230309.237398-1-ndesaulniers@google.com> <CAKwvOdmspKUknbzDn9kY2jMgkFw=Ktvst0ZtwambDOfybqJGWw@mail.gmail.com>
In-Reply-To: <CAKwvOdmspKUknbzDn9kY2jMgkFw=Ktvst0ZtwambDOfybqJGWw@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon, 11 May 2020 11:12:32 -0700
Message-ID: <CAHk-=wif=_ZomkWJAmQRCUAMHQ72V3NEQ-OteiPE56K7KoSjbQ@mail.gmail.com>
Subject: Re: [PATCH] x86: support i386 with Clang
To:     Nick Desaulniers <ndesaulniers@google.com>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        David Woodhouse <dwmw2@infradead.org>,
        Arnd Bergmann <arnd@arndb.de>,
        Dmitry Golovin <dima@golovin.in>,
        Dennis Zhou <dennis@kernel.org>, Tejun Heo <tj@kernel.org>,
        Christoph Lameter <cl@linux.com>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@kernel.org>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Al Viro <viro@zeniv.linux.org.uk>,
        Josh Poimboeuf <jpoimboe@redhat.com>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        LKML <linux-kernel@vger.kernel.org>,
        clang-built-linux <clang-built-linux@googlegroups.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, May 11, 2020 at 10:24 AM Nick Desaulniers
<ndesaulniers@google.com> wrote:
>
> Bumping for comment+review.
>
> On Mon, May 4, 2020 at 4:03 PM Nick Desaulniers <ndesaulniers@google.com> wrote:
> >
> > -                   : "qi" ((pto_T__)(val)));           \
> > +                   : "qi" ((unsigned char)(unsigned long)(val))); \

I find the extra casts really annoying and illogical.

I kind of see why they happen: just casting to 'unsigned char' can
cause warnings, when this case isn't even taken, and the sizeof(var)
is something else (and you cast a pointer to a char without going
through the 'unsigned long' first.

But that doesn't make this change any more sensible or readable.

Would using "__builtin_choose_expr()" be able to avoid this whole issue?

Because I'd rather re-write the whole thing with a different model
that doesn't have this issue, than make the code look insane.

And it does look insane, because that whole thing makes for a very
natural question: "why do we cast to unsigned long and then to char,
when we just checked that the size of the type is 1"?

                 Linus