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;
From:   Zhangjin Wu <falcon@tinylab.org>
To:     w@1wt.eu
Cc:     arnd@arndb.de, falcon@tinylab.org, linux-kernel@vger.kernel.org,
        linux-kselftest@vger.kernel.org, tanyuan@tinylab.org,
        thomas@t-8ch.de
Subject: Re: [PATCH v6 8/8] selftests/nolibc: customize CROSS_COMPILE for 32/64-bit powerpc
Date:   Sun,  6 Aug 2023 11:11:25 +0800
Message-Id: <20230806031125.6800-1-falcon@tinylab.org>
In-Reply-To: <20230805204845.GA7300@1wt.eu>
References: <20230805204845.GA7300@1wt.eu>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Feedback-ID: bizesmtp:tinylab.org:qybglogicsvrgz:qybglogicsvrgz5a-1
Precedence: bulk

Hi, Willy

> On Sun, Aug 06, 2023 at 02:47:09AM +0800, Zhangjin Wu wrote:
> > The little-endian powerpc64le compilers provided by Ubuntu and Fedora
> > are able to compile big endian kernel and big endian nolibc-test [1].
> 
> FWIW I'm wondering why focusing on these ones which have a different
> naming from the other ones, when I think that most users rely on the
> ones maintained by Arnd there:
> 
>    https://mirrors.edge.kernel.org/pub/tools/crosstool/
>

Arnd's toolchains may be a more distribution independent ones.

> Yours is called powerpc64le while the one above is "powerpc64", it
> requires to make an exception for this one, I find this a bit odd.
> 

Yes, one is little endian output by default, another may be big endian output
by default.

> If someone wants to use their distro's cross toolchain, that's fine,
> but I think that it will depend on distros anyway and some may not
> even be provided (like loongarch) so I think it would make more sense
> to adopt the canonical naming from Arnd's toolchains above.
>

Agree very much, let's switch to Arnd's toolchains.

> It's not critical, but as you showed below, it makes building for ppc
> a little bit cumbersome: those "export" lines could be dropped when
> using the default names, and that's what we should document as the
> recommended way to test:
> 
> > For example, it is able to build 64-bit nolibc-test with the big endian
> > powerpc64-linux-gcc crosstool from [2]:
> > 
> >     $ wget -c https://mirrors.edge.kernel.org/pub/tools/crosstool/files/bin/x86_64/13.1.0/x86_64-gcc-13.1.0-nolibc-powerpc64-linux.tar.xz
> >     $ tar xvf x86_64-gcc-13.1.0-nolibc-powerpc64-linux.tar.xz
> >     $ export PATH=$PWD/gcc-13.1.0-nolibc/powerpc64-linux/bin/:$PATH
> > 
> >     $ export CROSS_COMPILE_ppc=powerpc64-linux-
> >     $ export CROSS_COMPILE_ppc64=powerpc64-linux-
> >     $ export CROSS_COMPILE_ppc64le=powerpc64-linux-
> >     $ for arch in ppc ppc64 ppc64le; do \
> >         make run-user XARCH=$arch | grep "status: "; \
> >       done
> 
> Any opinion on this ?
>

Ok, let's go this way, if the others are ok for you, could you please
drop the last two CROSS_COMPILE patches from this v6 ppc series? Thanks.

I will send v2 CROSS_COMPILE series with them, with Arnd's toolchains, the
whole CROSS_COMPILE series will be unified and become very simple, although
they need to download the toolchains manually one by one, but it is possible to
write a common script, but that is another requirement.

BR,
Zhangjin
---
[1]: https://lore.kernel.org/lkml/cover.1691259983.git.falcon@tinylab.org/
[2]: https://lore.kernel.org/lkml/cover.1691263493.git.falcon@tinylab.org/

> Thanks,
> Willy