On 27.06.2017 10:53, Laurent Vivier wrote:
> On 27/06/2017 10:33, Thomas Huth wrote:
>> On 27.06.2017 10:24, David Hildenbrand wrote:
>>> On 27.06.2017 06:18, Thomas Huth wrote:
>>>> The intercept test currently can not be compiled with GCC 4.8 anymore.
>>>> It generates the following warning (which is fatal due to -Werror):
>>>>
>>>> s390x/intercept.c: In function ‘test_stidp’:
>>>> s390x/intercept.c:111:9: error: missing initializer for field ‘version’ of ‘struct cpuid’ [-Werror=missing-field-initializers]
>>>> struct cpuid id = {};
>>>> ^
>>>> Fix it by using a "0" as intializer here.
>>>>
>>>> Signed-off-by: Thomas Huth <thuth@xxxxxxxxxx>
>>>> ---
>>>> NB: We could also remove the -Wextra from the CFLAGS instead. IMHO
>>>> using -Wextra together with -Werror is just like playing Russian roulette.
>>>> Since -Wextra is some kind of "compiler warning playground" for the GCC
>>>> folks, you never know which compiler version will trigger an unexpected
>>>> (and often also unfounded) warning here, so using this together with -Werror
>>>> is just a nuisance.
>>>
>>> I agree, this is really not deterministic.
>>>
>>>>
>>>> s390x/intercept.c | 2 +-
>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/s390x/intercept.c b/s390x/intercept.c
>>>> index 9766289..9fe86cf 100644
>>>> --- a/s390x/intercept.c
>>>> +++ b/s390x/intercept.c
>>>> @@ -108,7 +108,7 @@ static void test_stap(void)
>>>> /* Test the STORE CPU ID instruction */
>>>> static void test_stidp(void)
>>>> {
>>>> - struct cpuid id = {};
>>>> + struct cpuid id = { 0 };
>>>>
>>>> asm volatile ("stidp %0\n" : "+Q"(id));
>>>> report("type set", id.type);
>>>>
>>>
>>> arm and powerpc also use -Wextra, maybe we should remove this then for all.
>>>
>>> Whatever you prefer.
>>
>> True ... maybe Drew and Laurent can also comment on whether they like
>> -Wextra or not ... if we all agree, then we can remove it, otherwise
>> let's try to go with this patch first (in the hope that we won't hit the
>> next problem too soon).
>
> I like -Wextra :)
>
> My opinion is more checking we have, less error we have.
The problem is that -Wextra often produces unfounded or even wrong
warnings, like in this case. I'd prefer to add the individual parameters
that rather always make sense instead, like -Wtype-limits for example.
Thomas
