Re: [PATCH] kselftest: Mark functions that unconditionally call exit() as __noreturn

[Nathan Chancellor <nathan@kernel.org>]

On Fri, Apr 12, 2024 at 02:05:47PM +0200, Thomas Gleixner wrote:
> On Thu, Apr 11 2024 at 11:45, Nathan Chancellor wrote:
> > I have based this change on timers/urgent, as the commit that introduces
> > this particular warning is there and it is marked for stable, even
> > though this appears to be a generic kselftest issue. I think it makes
> > the most sense for this change to go via timers/urgent with Shuah's ack.
> > While __noreturn with a return type other than 'void' does not make much
> > sense semantically, there are many places that these functions are used
> > as the return value for other functions such as main(), so I did not
> > change the return type of these functions from 'int' to 'void' to
> > minimize the necessary changes for a backport (it is an existing issue
> > anyways).
> 
> Hrmm. This really want's to be fixed once the change hits Linus tree as this:
> 
> static inline __noreturn int ksft_exit_pass(void)
> 
> looks seriously broken :)

Yeah, I only realized this morning that prior to this change, making
these functions return void instead of int would have broken

  int main(void)
  {
    <code>
    ksft_exit_pass();
  }

because without __noreturn, the compiler will complain that main() is
missing a return value. So 'int' -> '__noreturn void' would have been
the proper atomic change but the use of 'return ksft_exit_...();' made
that seem rather difficult when I was writing/testing that change on top
of this one.

However, now that I am actually sitting down and looking at it with a
fresh perspective, I am able to produce a pretty mechanical looking
change with just two sed commands:

  sed -i 's;__noreturn\(.*\)int;__noreturn\1void;g' tools/testing/selftests/kselftest.h &&
  sed -i 's/\(\s\+\)return\s\+\(.*ksft_exit_x\?\(fail\|pass\|skip\)\)/\1\2/g' $(git grep -lP 'return.*ksft_exit_x?(fail|pass|skip)' | sed s/:/-/g)

Perhaps Shuah could just run that in the kselftest tree and commit the
result once the change from Linus's tree is merged there? Otherwise, I
am happy to send a formal patch once I have something proper to base on.

Thanks for taking just the minimal change :)

Cheers,
Nathan