On Thu, Aug 29, 2024 at 09:37:34PM +0300, Jani Nikula wrote:
> On Thu, 29 Aug 2024, Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> > Since commit 6863f5643dd7 ("kbuild: allow Clang to find unused static
> > inline functions for W=1 build"). clang warns about unused static inline
> > functions in .c files, unlike GCC (they both do not warn for functions
> > coming from .h files). This difference is worked around for the normal
> > build by adding '__maybe_unused' to the definition of 'inline' but
> > Masahiro wanted to disable it for W=1 to allow this difference to find
> > unused/dead code. There have not been too many complaints as far as I am
> > aware but I can see how it is surprising.
>
> Heh, I was just going to reply citing the same commit.
>
> I occasionally build with clang myself, and we do enable most W=1 by
> default in the drm subsystem, so I was wondering why I hadn't hit
> this. The crucial difference is that we lack -DKBUILD_EXTRA_WARN1 which
> W=1 adds.
>
> I see there's no subdir-cppflags-y, but I don't see any harm in us
> adding -Wundef and -DKBUILD_EXTRA_WARN1 to subdir-ccflags-y. After we
> fix the fallout, of course. Do you?
No, that seems entirely reasonable when your goal is to enable W=1 for
your subsystem.
> I don't much like the __maybe_unused stuff, but I guess it's fine as a
> stopgap measure, and then we can grep for that when running out of
> things to do. :p
Perhaps worth a TODO or something? Maybe a kernel newbie could work on
that at some point if it is not high enough priority.
Cheers,
Nathan
