[SLOF] [PATCH slof 08/13] libnet: Compile with -Wextra

Alexey Kardashevskiy aik at ozlabs.ru
Sun Jan 31 14:54:08 AEDT 2021



On 29/01/2021 17:42, Thomas Huth wrote:
> On 29/01/2021 02.59, Alexey Kardashevskiy wrote:
>>
>>
>> On 29/01/2021 02:19, Thomas Huth wrote:
>>> On 27/01/2021 09.57, Alexey Kardashevskiy wrote:
>>>> -Wextra enables a bunch of rather useful checks which this fixes.
> [...]
>>>> diff --git a/slof/ppc64.c b/slof/ppc64.c
>>>> index 83a8e82cfb42..ca6cafffc35d 100644
>>>> --- a/slof/ppc64.c
>>>> +++ b/slof/ppc64.c
>>>> @@ -144,6 +144,12 @@ int socket(int domain, int type, int proto, 
>>>> char *mac_addr)
>>>>       int prop_len;
>>>>       int fd;
>>>> +    if (!(domain == AF_INET || domain == AF_INET6))
>>>
>>> Better:
>>>
>>>         if (domain != AF_INET && domain != AF_INET6))
>>
>> No it is not :)
>>
>>>
>>>> +        return -1;
>>>> +
>>>> +    if (type != SOCK_DGRAM || proto != 0)
>>>> +        return -1;
>>>
>>> I think these changes are not necessary anymore since you're 
>>> compiling with -Wno-unused-parameter ... so either drop these or put 
>>> them into a separate patch?
>>
>>
>> Well, it is also self documenting what we do implement in slof. And a 
>> little bit less work when I remove -Wno-unused-parameter later. I'll 
>> make it a separate patch. Thanks,
> 
> I'd rather vote for keeping -Wno-unused-parameter. Otherwise you're 
> always forced to write silly code or use __attribute__((unused)) when a 
> function has to match a certain parameter list, e.g. because it is used 
> as a function pointer later, and that's rather annoying.

Then why pass them around? I understand if it is a callback of some sort 
and not all instances may need additional parameters but this is not the 
case here.

And passing around 3 zeroes - domain/type/proto is as stupid, I am 
basically commenting on what is supported in socket(). Thanks,



-- 
Alexey


More information about the SLOF mailing list