Re: [PATCH] kbuild: enable unused-function warnings for W= build with Clang
From: Nathan Chancellor
Date: Mon Aug 19 2019 - 22:25:21 EST
On Tue, Aug 20, 2019 at 01:58:26AM +0900, Masahiro Yamada wrote:
> Hi Nathan,
> On Tue, Aug 20, 2019 at 1:09 AM Nathan Chancellor
> <natechancellor@xxxxxxxxx> wrote:
> > On Mon, Aug 19, 2019 at 07:51:38PM +0900, Masahiro Yamada wrote:
> > > GCC and Clang have different policy for -Wunused-function; GCC does
> > > not report unused-function warnings at all for the functions marked
> > > as 'static inline'. Clang does report unused-function warnings if they
> > > are defined in source files instead of headers.
> > >
> > > We could use Clang for detecting unused functions, but it has been
> > > suppressed since commit abb2ea7dfd82 ("compiler, clang: suppress
> > > warning for unused static inline functions").
> > >
> > > So, we never notice left-over code if functions in .c files are
> > > marked as inline.
> > >
> > > Let's remove __maybe_unused from the inline macro. As always, it is
> > > not a good idea to sprinkle warnings for the normal build. So, these
> > > warnings will be shown for the W= build.
> > >
> > > If you contribute to code clean-up, please run "make CC=clang W=1"
> > > and check -Wunused-function warnings. You will find lots of unused
> > > functions.
> > >
> > > Some of them are false-positives because the call-sites are disabled
> > > by #ifdef. I do not like to abuse the inline keyword for suppressing
> > > unused-function warnings because it might affect the compiler's
> > > optimization. When I need to fix unused-functions warnings, I prefer
> > > adding #ifdef or __maybe_unused to function definitions.
> > >
> > > Signed-off-by: Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx>
> > So if I understand everything correctly, this change allows us to start
> > finding unused static inline functions with clang at W=1 but disables
> > -Wunused-function by default... I am not sure that is a good tradeoff
> > as I am pretty sure that W=1 is fairly noisy for clang although I
> > haven't checked lately. I'd argue most regular developers do not build
> > with W=1 meaning -Wunused-function generally will not be run with clang
> > at all, missing stuff like this:
> Try "git log --grep=W=1"
> Some people are making efforts to fix W=1 warnings.
> I believe somebody will start to remove unused static inline functions.
Yes, it could be a good way to get people involved with working with
> > https://lore.kernel.org/lkml/20190523010235.GA105588@archlinux-epyc/
> > https://lore.kernel.org/lkml/1558574945-19275-1-git-send-email-skomatineni@xxxxxxxxxx/
> > Furthermore, per the documemtation , -Wno-unused-function will also
> > disable -Wunneeded-internal-declaration, which can help find bugs like
> > commit 8289c4b6f2e5 ("platform/x86: mlx-platform: Properly use
> > mlxplat_mlxcpld_msn201x_items").
> > : https://clang.llvm.org/docs/DiagnosticsReference.html#wunused-function
> How about this?
> KBUILD_CFLAGS += -Wno-unused-function
> KBUILD_CFLAGS += -Wunneeded-internal-declaration
Yes, that would work.
> > Is there a way to conditionally remove __maybe_unused from the inline
> > defintion so that we keep the current behavior but we can still
> > selectively find potentially unused functions?
> It would be possible by tweaking include/linux/compiler_types.h
> but I am not a big fan of uglifying the 'inline' replacement any more.
I agree that ugly is not ideal but I think it is even less ideal to
weaken the default set of warnings for clang... Merely food for thought