# Re: \[PATCH\] selftests/powerpc: Fix set-but-not-used errors with GCC v16.2

- From: Thomas Huth \<hidden\>
- Date: 2026-11-16 12:27:14 
- List: linuxppc-dev
- Page: https://ratatoskr.run/linuxppc-dev/2026/11/17944727
- Thread: https://ratatoskr.run/linuxppc-dev/2026/11/17943872/t.md

```text
Am 16.11.26 um 12:07 schrieb Christophe Leroy (CS GROUP):
>
>
> Le 16/11/2026 à 11:37, Thomas Huth a écrit :
>> When compiling the powerpc selftests with GCC v16.2, the compilation
>> fails with:
>>
>>   ../pmu/lib.c: In function ‘eat_cpu_child’:
>>   ../pmu/lib.c:86:22: error: variable ‘i’ set but not used [- Werror=unused-but-set-variable=]
>>      86 |         volatile int i = 0;
>>         |                      ^
>>
>> and:
>>
>>   ebb.c: In function ‘ebb_child’:
>>   ebb.c:352:18: error: variable ‘val’ set but not used [-Werror=unused- but-set-variable=]
>>     352 |         uint64_t val;
>>         |                  ^~~
>>   cc1: all warnings being treated as errors
>>
>> Add some "asm volatile" statements here to make to mark the variables
>> as used and to avoid that the compiler might optimize the code away.
>
> This looks really uggly, a cleaner way is likely possible, see below
>
>>
>> Signed-off-by: Thomas Huth <redacted>
>> ---
>>   tools/testing/selftests/powerpc/pmu/ebb/ebb.c | 1 +
>>   tools/testing/selftests/powerpc/pmu/lib.c     | 5 ++++-
>>   2 files changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/powerpc/pmu/ebb/ebb.c b/tools/ testing/selftests/powerpc/pmu/ebb/ebb.c
>> index 21537d6eb6b7d..f4ef92bacd9ab 100644
>> --- a/tools/testing/selftests/powerpc/pmu/ebb/ebb.c
>> +++ b/tools/testing/selftests/powerpc/pmu/ebb/ebb.c
>> @@ -391,6 +391,7 @@ int ebb_child(union pipe read_pipe, union pipe write_pipe)
>>           val  = mfspr(SPRN_MMCRA);
>>           val |= mfspr(SPRN_MMCR2);
>>           val |= mfspr(SPRN_MMCR0);
>> +        asm volatile("" :: "r"(val) : "memory");
>
> If val is not used, why not just do:
>
>          mfspr(SPRN_MMCRA);
>          mfspr(SPRN_MMCR2);
>          mfspr(SPRN_MMCR0);
>
> mfspr() is already an 'asm volatile' so it will be executed even if the result is not used.

Ok, agreed, that makes sense.

>>       }
>>       ebb_global_disable();
>> diff --git a/tools/testing/selftests/powerpc/pmu/lib.c b/tools/testing/ selftests/powerpc/pmu/lib.c
>> index 3213579874084..32cafd40d2c56 100644
>> --- a/tools/testing/selftests/powerpc/pmu/lib.c
>> +++ b/tools/testing/selftests/powerpc/pmu/lib.c
>> @@ -95,7 +95,10 @@ static int eat_cpu_child(union pipe read_pipe, union pipe write_pipe)
>>       wait_for_parent(read_pipe);
>>       /* Soak up cpu forever */
>> -    while (1) i++;
>> +    while (1) {
>> +        i++;
>> +        asm volatile("" :: "r"(i) : "memory");
>> +    }
>
> If 'i' is not used, why keep it ?
>
> Just do
>
>      while (1)
>          ;
I assume the original intention was to make the CPU do something beside a "branch to self" instruction ... ? Not sure whether it makes a difference for these pmu tests, though?

 Thomas


```
