Re: [linux-next:master 2008/10581] include/linux/unaligned/be_byteshift.h:41:19: error: redefinition of 'get_unaligned_be16'

From: Harvey Harrison <harvey.harrison@gmail.com>
To: Linus Walleij <linus.walleij@linaro.org>
Cc: kernel test robot <lkp@intel.com>,
	Mark Brown <broonie@kernel.org>,
	 Bart Van Assche <bvanassche@acm.org>,
	kbuild-all@lists.01.org,
	 Linux Memory Management List <linux-mm@kvack.org>,
	Jonathan Cameron <Jonathan.Cameron@huawei.com>,
	 Andy Shevchenko <andy.shevchenko@gmail.com>
Subject: Re: [linux-next:master 2008/10581] include/linux/unaligned/be_byteshift.h:41:19: error: redefinition of 'get_unaligned_be16'
Date: Fri, 12 Feb 2021 09:37:58 -0800	[thread overview]
Message-ID: <CAHLZz10+ouAa+r_7WKzEUKq16BsigsAY8_FezCq-A08c6UOgsw@mail.gmail.com> (raw)
In-Reply-To: <CACRpkdYsrz1NjjdisXa7C+jQEppT6AJUDPyX=MC4QbU_omXcmA@mail.gmail.com>

On Fri, Feb 12, 2021 at 8:43 AM Linus Walleij <linus.walleij@linaro.org> wrote:
>
> On Fri, Feb 12, 2021 at 12:53 AM kernel test robot <lkp@intel.com> wrote:
>
> But what kind of syntactic trainwreck is this!
>
> (...)
> > commit: de8860b1ed4701ea7e6f760f02d79ca6a3b656a1 [2008/10581] iio: magnetometer: Add driver for Yamaha YAS530
> (...)
> >    In file included from drivers/iio/magnetometer/yamaha-yas530.c:35:
> >    include/linux/unaligned/be_byteshift.h: At top level:
> > >> include/linux/unaligned/be_byteshift.h:41:19: error: redefinition of 'get_unaligned_be16'
> >       41 | static inline u16 get_unaligned_be16(const void *p)
> >          |                   ^~~~~~~~~~~~~~~~~~
> (...)
> >                     from include/linux/regulator/consumer.h:35,
> >                     from drivers/iio/magnetometer/yamaha-yas530.c:33:
> >    include/linux/unaligned/be_struct.h:7:19: note: previous definition of 'get_unaligned_be16' was here
> >        7 | static inline u16 get_unaligned_be16(const void *p)
> >          |                   ^~~~~~~~~~~~~~~~~~
>
> Inspecting be_byteshift.h and be_struct.h I do see that both of them
> define get_unaligned_be16/32/64.
>
> They both end up calling the different implementations of
> __get_unaligned_be16/32/64 on top of that.
>
> So include one or the other and never both at the same time?
> Well that is hard to avoid if one of them is in turn included by other
> headers such as the very generic <linux/regulator/consumer.h>
> in this case.
>
> Harvey, anyone: ideas on how to fix this?
>

It's been a long time since I looked at this, it seems to have had a
few more changes in the 10 years since I was helping consolidate the
implementations. The intention at the time was never to include the
specific implementations directly (be_struct.h be_byteshift.h etc) and
only to include asm/unaligned.h in the places where needed.

Harvey

> One of them have to change name in the whole world I suppose?
>
> I suppose my code is working either because they both do exactly
> the same thing or thanks to the order I include the files :P
>
> I have the feeling there is something I don't understand about how
> this was thought out, like I must have missed something. All
> the collisions are even in the same include directory :(
>
> Yours,
> Linus Walleij