Re: ACPICA compilation error on x86_64_defconfig (W=1 build)

["Shevchenko, Andriy" <andriy.shevchenko@intel.com>]

On Wed, Apr 09, 2025 at 04:07:09PM +0200, Rafael J. Wysocki wrote:
> On Wed, Apr 9, 2025 at 4:01 PM Shevchenko, Andriy
> <andriy.shevchenko@intel.com> wrote:
> >
> > Note, I have posted two much better solutions in that bug report thread.
> 
> Which bug report thread do you mean?

https://github.com/acpica/acpica/issues/973

> > Please, choose one and fix, it prevents kernel from building...
> 
> With Clang-19 and 'make W=1' that is.

Yes. it is not *that* critical, sure.

> > On Wed, Apr 09, 2025 at 02:24:45PM +0300, Shevchenko, Andriy wrote:
> > > GCC should ignore all clang specific pragmas as far as I know.
> > >
> > > But this seems to me the not the best solution either. What code needs is to
> > > have some refactoring to make sure the variable is used or not defined and used
> > > at all. Unfortunately, I am not familiar with ACPICA to propose anything right
> > > now, though.
> > >
> > > On Tue, Apr 08, 2025 at 09:34:36PM +0300, Dumbre, Saket wrote:
> > > > I am not sure about the Linux environment specifics like Andy described in the
> > > > previous conversation, but if GCC or some other compiler throws an error in the
> > > > code with #pragma for clang, then may I suggest using this instead:
> > > >
> > > > #pragma clang diagnostic push
> > > > #if defined(__clang__) && __clang__ >= n (replace n with the least version
> > > > number which introduces this new warning type)
> > > > #pragma clang diagnostic ignored "-Wunused-but-set-variable"
> > > > #endif
> > > > u32 num_carats;
> > > > #pragma clang diagnostic pop
> > > >
> > > > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
> > > > From: Dumbre, Saket <saket.dumbre@intel.com>
> > > > Sent: Tuesday, April 8, 2025 11:11 AM
> > > > To: Shevchenko, Andriy <andriy.shevchenko@intel.com>; Rafael J. Wysocki
> > > > <rafael@kernel.org>
> > > > Cc: Moore, Robert <robert.moore@intel.com>; Wysocki, Rafael J
> > > > <rafael.j.wysocki@intel.com>; linux-acpi@vger.kernel.org
> > > > <linux-acpi@vger.kernel.org>; acpica-devel@lists.linux.dev
> > > > <acpica-devel@lists.linux.dev>
> > > > Subject: Re: ACPICA compilation error on x86_64_defconfig (W=1 build)
> > > >
> > > > Hi Andy and Rafael,
> > > >
> > > > This issue is a duplication of an Apple ecosystem-based (ARM + M-series micro
> > > > architecture) ACPICA GitHub Issue, which I was waiting on a response from the
> > > > contributor/submitter to move ahead to resolve it:
> > > > https://github.com/acpica/acpica/issues/973
> > > >
> > > > I don't see a platform specific file for Clang/LLVM in ACPICA (only for GCC and
> > > > MSVC), so I think the worst-case untidy solution would be to locally patch this
> > > > warning-turned-error by suppressing this warning type if the compiler is Clang/
> > > > LLVM.
> > > >
> > > > It would look somewhat better to move that line all the way down after all the
> > > > other definitions/declarations just before any execution line and wrap it with
> > > > a #pragma something along the lines:
> > > >
> > > > #pragma clang diagnostic push
> > > > #pragma clang diagnostic ignored "-Wunused-but-set-variable"
> > > > u32 num_carats;
> > > > #pragma clang diagnostic pop
> > > >
> > > > I am just not sure if this is the only place in the ACPICA code that generates
> > > > this warning type and depending on the answer, this fix could be implemented at
> > > > a wider/global scale, but please give this hacky fix a try and let me know if
> > > > it resolves that. Thanks a lot!
> > > >
> > > >
> > > > Kind regards,
> > > >
> > > > Saket Dumbre
> > > >
> > > >
> > > > ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
> > > > From: Shevchenko, Andriy
> > > > Sent: Monday, April 7, 2025 11:51 AM
> > > > To: Rafael J. Wysocki
> > > > Cc: Dumbre, Saket; Moore, Robert; Wysocki, Rafael J;
> > > > linux-acpi@vger.kernel.org; acpica-devel@lists.linux.dev
> > > > Subject: Re: ACPICA compilation error on x86_64_defconfig (W=1 build)
> > > >
> > > > On Mon, Apr 07, 2025 at 08:49:41PM +0200, Rafael J. Wysocki wrote:
> > > > > On Mon, Apr 7, 2025 at 8:47 PM Andy Shevchenko
> > > > > <andriy.shevchenko@intel.com> wrote:
> > > > > >
> > > > > > v6.15-rc1 (but seems older should also be problematic) fails to compile by
> > > > > > Clang 19 with `make W=1`.
> > > > > >
> > > > > >   CC      drivers/acpi/acpica/nsaccess.o
> > > > > >   drivers/acpi/acpica/nsaccess.c:295:6: error: variable 'num_carats' set
> > > > but not used [-Werror,-Wunused-but-set-variable]
> > > > > >     295 |         u32 num_carats;
> > > > > >         |             ^
> > > > > >   1 error generated.
> > > > > >
> > > > > > I looked briefly in the code and I have no idea how to fix this, as
> > > > probably
> > > > > > __maybe_unused not an option for ACPICA. Hence this report.
> > > > > >
> > > > > > Note, the in-tree defconfig is in use:
> > > > > >
> > > > > >         git checkout v6.15-rc1
> > > > > >         make x86_64_defconfig
> > > >
> > > > > >         make W=1
> > > >
> > > > Since it's a clang, this should have also something like LLVM=1.
> > > >
> > > > > > should trigger it.
> > > >
> > > > --
> > > > With Best Regards,
> > > > Andy Shevchenko
> > > >
> > > >
> > >
> > > --
> > > With Best Regards,
> > > Andy Shevchenko
> > >
> > >
> >
> > --
> > With Best Regards,
> > Andy Shevchenko
> >
> >
> >

-- 
With Best Regards,
Andy Shevchenko