Re: selftests: connector: proc_filter.c:48:20: error: invalid application of 'sizeof' to an incomplete type 'struct proc_input'

[Anjali Kulkarni]

> On Jul 28, 2023, at 10:33 AM, Shuah Khan <skhan@xxxxxxxxxxxxxxxxxxx> wrote:
> 
> On 7/27/23 19:38, Anjali Kulkarni wrote:
>>> On Jul 27, 2023, at 5:43 PM, Shuah Khan <skhan@xxxxxxxxxxxxxxxxxxx> wrote:
>>> 
>>> On 7/27/23 11:34, Anjali Kulkarni wrote:
>>>>> On Jul 25, 2023, at 9:48 AM, Anjali Kulkarni <Anjali.K.Kulkarni@xxxxxxxxxx> wrote:
>>>>> 
>>>>> 
>>>>> 
>>>>>> On Jul 25, 2023, at 6:05 AM, Naresh Kamboju <naresh.kamboju@xxxxxxxxxx> wrote:
>>>>>> 
>>>>>> selftests: connector: proc_filter build failed with clang-16 due to below
>>>>>> warnings / errors on Linux next-20230725.
>>>>>> 
>>>>>> Reported-by: Linux Kernel Functional Testing <lkft@xxxxxxxxxx>
>>>>>> 
>>>>>> clang --target=aarch64-linux-gnu -fintegrated-as
>>>>>> -Werror=unknown-warning-option -Werror=ignored-optimization-argument
>>>>>> -Werror=option-ignored -Werror=unused-command-line-argument
>>>>>> --target=aarch64-linux-gnu -fintegrated-as -Wall proc_filter.c -o
>>>>>> /home/tuxbuild/.cache/tuxmake/builds/1/build/kselftest/connector/proc_filter
>>>>>> proc_filter.c:42:12: error: invalid application of 'sizeof' to an
>>>>>> incomplete type 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^~~~~~~~~~~~~~~
>>>>>> proc_filter.c:22:5: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^ ~~~~~~~~~~~~~~~~~~~
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:48:20: error: invalid application of 'sizeof' to an
>>>>>> incomplete type 'struct proc_input'
>>>>>> hdr->nlmsg_len = NL_MESSAGE_SIZE;
>>>>>> ^~~~~~~~~~~~~~~
>>>>>> proc_filter.c:22:5: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^ ~~~~~~~~~~~~~~~~~~~
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:64:14: error: invalid application of 'sizeof' to an
>>>>>> incomplete type 'struct proc_input'
>>>>>> msg->len = sizeof(struct proc_input);
>>>>>> ^ ~~~~~~~~~~~~~~~~~~~
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:65:35: error: incomplete definition of type 'struct proc_input'
>>>>>> ((struct proc_input *)msg->data)->mcast_op =
>>>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:66:31: error: incomplete definition of type 'struct proc_input'
>>>>>> ((struct proc_input *)pinp)->mcast_op;
>>>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~^
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:67:35: error: incomplete definition of type 'struct proc_input'
>>>>>> ((struct proc_input *)msg->data)->event_type =
>>>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:68:31: error: incomplete definition of type 'struct proc_input'
>>>>>> ((struct proc_input *)pinp)->event_type;
>>>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~^
>>>>>> proc_filter.c:42:12: note: forward declaration of 'struct proc_input'
>>>>>> char buff[NL_MESSAGE_SIZE];
>>>>>> ^
>>>>>> proc_filter.c:22:19: note: expanded from macro 'NL_MESSAGE_SIZE'
>>>>>> sizeof(struct proc_input))
>>>>>> ^
>>>>>> proc_filter.c:245:20: error: variable has incomplete type 'struct proc_input'
>>>>>> struct proc_input input;
>>>>>> ^
>>>>>> proc_filter.c:245:9: note: forward declaration of 'struct proc_input'
>>>>>> struct proc_input input;
>>>>>> ^
>>>>>> proc_filter.c:264:22: error: use of undeclared identifier
>>>>>> 'PROC_EVENT_NONZERO_EXIT'
>>>>>> input.event_type = PROC_EVENT_NONZERO_EXIT;
>>>>>> ^
>>>>>> 9 errors generated.
>>>>>> make[4]: Leaving directory '/builds/linux/tools/testing/selftests/connector’
>>>>>> 
>>>>>> 
>>>>> These are expected since you need to have the changes in kernel that were committed with this patch to be installed on the kernel on which this is being compiled/run on. That is what the test is for, and the check to make it run on previous kernels as well was made a runtime check. Do you expect this to compile on a kernel without the corresponding kernel changes that were committed with this patch?
>>>>> 
>>>>> Anjali
>>>> Gentle ping - could you answer above questions?
>>>>> 
>>> 
>>> I am seeing the same on linux-next next-20230727
>>> 
>>> PROC_EVENT_NONZERO_EXIT is defined and NL_MESSAGE_SIZE
>>> 
>>> Anjali,
>>> 
>>> What are the dependent commits and should they be in next?
>>> Shouldn't this test patch go with the kernel patches it depends
>>> on? Can you do some testing on next and let me know why this
>>> test is failing to build?
>> All the commits went in together - however, the kernel changes that went in this patch need to be *installed on kernel on which this is being built*. Did you do that and then try?
> 
> Building kernel and running "make headers" before building the test
> is what is needed. That is what Naresh and I did.
> 
> How are you building this test?
> 
> I sent a 3 patch series with the fix to this problem and a couple of
> others I found during testing.
> 
> Also please check the error messages - some of them are cryptic and
> could use clarity.
> 
> Another thing is argument check - arg == 2 - does this test require
> arguments? Note that without arguments the test runs - of that is
> default it is fine. This test could use an usage information. Please
> fix the above.

Yes, the test can run without any arguments. I will add usage information.

Anjali

> 
> thanks,
> -- Shuah