Re: [PATCH v1] sched/topology: Make compiler happy about unused constant definitions

[Valentin Schneider]

On 25/08/20 12:32, Andy Shevchenko wrote:
> On Tue, Aug 25, 2020 at 02:26:05PM +0300, Andy Shevchenko wrote:
>> On Tue, Aug 25, 2020 at 02:24:05PM +0300, Andy Shevchenko wrote:
>> > On Tue, Aug 25, 2020 at 11:12:21AM +0100, Valentin Schneider wrote:
>> > > On 25/08/20 10:03, Valentin Schneider wrote:
>> > > > On 25/08/20 09:26, Andy Shevchenko wrote:
>> > > >> On Mon, Aug 24, 2020 at 06:09:41PM +0100, Valentin Schneider wrote:
>> > > >>> On 24/08/20 16:39, Andy Shevchenko wrote:
>> > > >>> > Compilation of almost each file ends up with
>> > > >>> >
>> > > >>> >  In file included from .../include/linux/energy_model.h:10,
>> > > >>> >                 from .../include/linux/device.h:16,
>> > > >>> >                 from .../drivers/spi/spi.c:8:
>> > > >>> >  .../include/linux/sched/topology.h:30:27: warning: ‘SD_DEGENERATE_GROUPS_MASK’ defined but not used [-Wunused-const-variable=]
>> > > >>> >     30 | static const unsigned int SD_DEGENERATE_GROUPS_MASK =
>> > > >>> >        |                           ^~~~~~~~~~~~~~~~~~~~~~~~~
>> > > >>> >  ...
>> > > >>> >
>> > > >>> > Make compiler happy by annotating the static constants with __maybwe_unused.
>> > >
>> > > > Sorry, that's what I get for trying to be too succinct; what I tried to say
>> > > > was that SD_DEGENERATE_GROUPS_MASK should very much be used for SMP. If the
>> > > > build is !SMP, it shouldn't even be defined, IOW I'm perplexed as to where
>> > > > this is coming from.
>> > >
>> > > So I see how having this as a constvar rather than a constexpr is somewhat
>> > > daft (we get an instance per compilation unit), but none of my compilers
>> > > seem to complain (even with W=1). AFAIA the kernelbot didn't catch any of
>> > > it either.
>
> And even without compiler or any other analyzer / bot I can 100% sure tell that
> spi.c does *not* use that symbol.

Aye aye, this is a daft constvar placement from my end, apologies.

For the SD_DEGENERATE_GROUPS_MASK, that one could directly be shoved into
kernel/sched/topology.c (or done via an enum, I don't care too much).

I suppose the other one causing you grief is sd_flag_debug[]; that one too
really shouldn't be in a header. I need to access that in two separate
files, so I guess I'll have to cook up some wrappers.

Let me take a jab at it, it's my own mess after all...