On Fri, Jun 20, 2025 at 5:41 AM Shuah Khan <sk...@linuxfoundation.org> wrote:
>
> On 6/19/25 13:00, John Hubbard wrote:
> > On 6/19/25 10:00 AM, Shuah Khan wrote:
> >> On 6/19/25 01:46, Chen Linxuan wrote:
> >>> On Thu, Jun 19, 2025 at 5:23 AM Shuah Khan <sk...@linuxfoundation.org>
> >>> wrote:
> >>>> On 6/9/25 20:07, Chen Linxuan wrote:
> > ...
> >>>>> diff --git a/tools/testing/selftests/filesystems/mount-notify/mount-
> >>>>> notify_test.c b/tools/testing/selftests/filesystems/mount-notify/
> >>>>> mount-notify_test.c
> >>>>> index 63ce708d93ed0..34afe27b7978f 100644
> >>>>> --- a/tools/testing/selftests/filesystems/mount-notify/mount-
> >>>>> notify_test.c
> >>>>> +++ b/tools/testing/selftests/filesystems/mount-notify/mount-
> >>>>> notify_test.c
> >>>>> @@ -465,7 +465,9 @@ TEST_F(fanotify, rmdir)
> >>>>> ASSERT_GE(ret, 0);
> >>>>>
> >>>>> if (ret == 0) {
> >>>>> - chdir("/");
> >>>>> + // Suppress -Wunused-result
> >>>>> + // Ref: https://gcc.gnu.org/bugzilla/show_bug.cgi?
> >>>>> id=66425#c34
> >>>>> + (void) !chdir("/");
> >
> > This is quite ugly. :)
I agree with you. :)
> >
> >>>>>> Why not fix the problem the right way by checking the return value.
> >>>> Suppressing the error isn't useful.
> >>>
> >>> The code is already handling cleanup in error cases,
> >>> and I don't think checking the result of chdir would be useful here.
I think I was mistaken earlier. Here we are in the child process after
a fork, not handling an error case.
I think simply calling exit(-1) here when chdir failed should be
enough to make the test fail in the parent process.
Maybe we should do the same for other similar calls as well. I will
send a v2 soon.
> >>
> >
> > Why not just fail with the appropriate test result, if chdir() fails
> > here, instead of making a bit of a mess with odd void casts to a
> > negated return value, and a reference to a compiler bug report?
> >
> > Really, Shuah is putting you on the right path here.
>
> Ha. I didn't ask to suppress the error with the cast. I asked
> to check the return and fail.
>
> >
> >> We check for chdir() in several tools in the kernel. Add a check for
> >> it instead of suppressing the [-Wunused-result] - suppressing doesn't
> >> do any good.
>
> This is what I said.
>
> thanks,
> -- Shuah
>
>
