Re: [PATCH v2] sched: Fix print format for u64

[Paul Turner]

On Fri, Jan 25, 2013 at 12:06 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
> * Paul Turner <pjt@xxxxxxxxxx> wrote:
>
>> On Thu, Jan 24, 2013 at 11:19 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>> >
>> > * Fabio Estevam <festevam@xxxxxxxxx> wrote:
>> >
>> >> On Thu, Jan 24, 2013 at 12:19 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>> >>
>> >> > I suppose - is this patch warning-free both on 64-bit and 32-bit
>> >> > systems?
>> >>
>> >> Yes, just confirmed that this patch is warning-free on both 64-bit and
>> >> 32-bit machines.
>> >
>> > Apparently it's not all good, see the warning attached below.
>>
>> Yeah this patch is broken; this is not properly fixable as is without
>> #ifdefs (or fixing the insanity that is atomic64_read).
>>
>> Specifically:
>>
>> On some architectures (e.g. x86_64) atomic64_read is typed long
>> On some others (e.g. x86-32) it's typed long long
>> On yet others (e.g. arm) it's typed u64
>
> Hm, cannot we type atomic64_read() to u64 on x86-64 as well, and
> fix other architectures? How widespread is this problem, have
> you checked that perhaps?

I have not looked at how much fall-out this would cause.  Ideally
someone with a cross-compiing tool-chain or from the specific arches
could take ownership of cleaning up their atomic64 implementations to
avoid introducing more build pain.

Theoretically, given the list above x86-64 is actually the outlier in
assuming that it's just a long type; perhaps we can gain sufficient
mileage by changing only the x86-64 implementation; which is much
easier for us to validate :-).

>
> I'm all for clean, consistent types instead of #ifdef or
> temporary variable uglies.

Yes I agree this would be the best solution.

>
> Thanks,
>
>         Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/