Re: [PATCH] kbuild: add most of Clang-specific flags unconditionally

From: Nick Desaulniers <ndesaulniers@google.com>
To: Masahiro Yamada <yamada.masahiro@socionext.com>,
	Behan Webster <behanw@converseincode.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>,
	Linux Kbuild mailing list <linux-kbuild@vger.kernel.org>,
	Nathan Chancellor <natechancellor@gmail.com>,
	Arnd Bergmann <arnd@arndb.de>,
	Clang-Built-Linux ML <clang-built-linux@googlegroups.com>,
	Michal Marek <michal.lkml@markovi.net>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Matthias Kaehlcke <mka@google.com>
Subject: Re: [PATCH] kbuild: add most of Clang-specific flags unconditionally
Date: Fri, 10 May 2019 10:57:01 -0700	[thread overview]
Message-ID: <CAKwvOdkt9eCFdy7pNer7+6ZNz4vFVX_55Fe_98W1RFFzwz3U5Q@mail.gmail.com> (raw)
In-Reply-To: <CAK7LNAS+FQqQZ0RfW8e6mxabUOq9YVk=eEEztmN-+BHnTmDa_w@mail.gmail.com>

On Fri, May 10, 2019 at 6:54 AM Masahiro Yamada
<yamada.masahiro@socionext.com> wrote:
>
> On Thu, May 9, 2019 at 4:06 PM Sedat Dilek <sedat.dilek@gmail.com> wrote:
> >
> > On Thu, May 9, 2019 at 8:45 AM Masahiro Yamada
> > <yamada.masahiro@socionext.com> wrote:
> > >
> > > We do not support old Clang versions. Upgrade your clang version
> > > if any of these flags is unsupported.
> > >
> > > Let's add flags within ifdef CONFIG_CC_IS_CLANG unconditionally,
> > > except -fcatch-undefined-behavior.
> > >
> > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> >
> > Looks good to me.
> >
> > Reviewed-by: Sedat Dilek <sedat.dilek@gmail.com>
> >
> > Just as sidenote:
> > I experimented with a snapshot version of clang-9 and lld-9 and could
> > build, link and boot on bare-metal with '-mglobal-merge' on
> > Debian/buster AMD64.
>
>
> The comment says
>  # CLANG uses a _MergedGlobals as optimization, but this breaks modpost, as the
>  # source of a reference will be _MergedGlobals and not on of the
> whitelisted names.
>  # See modpost pattern 2
>
> So, it seems it is just a matter of modpost,
> but I am not sure enough.
>
> This flag has been here since the initial support.
> (61163efae02040f66a95c8ed17f4407951ba58fa)
>
>
> Perhaps, we should review clang flags one by one again?

Yes, it's always good to re-evaluate if something is just cruft and
can be removed.

+Behan

I don't quite understand the comment about _MergedGlobals, Behan, do
happen to have more context?
-- 
Thanks,
~Nick Desaulniers