On 2 February 2015 at 12:26, Maxim Uvarov <[email protected]> wrote:
> On 02/02/2015 02:18 PM, Ola Liljedahl wrote:
>>
>> On 2 February 2015 at 12:05, Maxim Uvarov <[email protected]> wrote:
>>>
>>> On 01/29/2015 01:39 PM, Ola Liljedahl wrote:
>>>>
>>>>          num_workers = odph_linux_cpumask_default(&cpumask,
>>>> num_workers);
>>>> -       odp_cpumask_to_str(&cpumask, cpumaskstr, sizeof(cpumaskstr));
>>>> +       size_t bufsz = sizeof(cpumaskstr);
>>>> +       (void)odp_cpumask_to_str(&cpumask, cpumaskstr, &bufsz);
>>>>
>>>
>>> why did add this (void)?
>>
>> Because odp_cpumask_to_str() has a return value that you normally
>> should check. If I think the return value for some reason doesn't
>> merit checking, then I cast the function call to "(void)".
>
> In current repo it's void:
>
> void odp_cpumask_to_str(const odp_cpumask_t *mask, char *str, int len);
Yes but the patch changed that. The formatting can fail because the
specified output buffer is too small. So the call needs a return
value.

>
>>
>> If you are closing a file/socket descriptor you have written to, you
>> should check the return value from close() because writes may have
>> been buffered and close may fail writing out your data.. But if the
>> file descriptor was e.g. only used for reading, then you don't need to
>> check the return value from close(). Ideally the compiler should warn
>> if you are not checking the return value from close() and in the
>> second case above, you would use "(void)close(fd);".
>>
>> I think the compiler can warn if return values are not used. Possibly
>> this is controlled using some GCC attribute.
>>
>> BTW: This patch series is obsolete, I am currently rebasing it.
>>
>>
>>> Maxim.
>>>
>>> _______________________________________________
>>> lng-odp mailing list
>>> [email protected]
>>> http://lists.linaro.org/mailman/listinfo/lng-odp
>
>

_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp

Reply via email to