Re: linux-next: build warning after merge of the akpm-current tree

[Michal Hocko <mhocko@kernel.org>]

On Mon 13-11-17 09:09:55, Michal Hocko wrote:
> On Mon 13-11-17 16:42:06, Stephen Rothwell wrote:
> > Hi Andrew,
> > 
> > After merging the akpm-current tree, today's linux-next build (powerpc
> > ppc64_defconfig) produced this warning:
> > 
> > In file included from include/linux/mmzone.h:17:0,
> >                  from include/linux/mempolicy.h:10,
> >                  from mm/mempolicy.c:70:
> > mm/mempolicy.c: In function 'mpol_to_str':
> > include/linux/nodemask.h:107:41: warning: the address of 'nodes' will always evaluate as 'true' [-Waddress]
> >  #define nodemask_pr_args(maskp) (maskp) ? MAX_NUMNODES : 0, (maskp) ? (maskp)->bits : NULL
> >                                          ^
> > mm/mempolicy.c:2817:11: note: in expansion of macro 'nodemask_pr_args'
> >            nodemask_pr_args(&nodes));
> >            ^
> 
> Hmm, this warning is quite surprising to me. Sure in this particular
> case maskp will always be non-NULL so we always expand to
> 	MAX_NUMNODES, maskp->bits
> which is what we want. But we have other users which may be NULL. Does
> anybody understan why this warns at all?

Strange I played with the following minimal test case and it warns only
for the explicit &m use while n is clearly never null as well. This all
smells like -Waddress is just confused (at least with my gcc 7.2.0-12

#include <stdio.h>

#define MAX_NUMNODES 10
struct mask {
	void *bits;
};
#define nodemask_pr_args(maskp) (maskp) ? MAX_NUMNODES : 0, (maskp) ? (maskp)->bits : NULL

int foo(void)
{
	struct mask m;
	struct mask *n = &m;

	printf("%*p\n", nodemask_pr_args(&m));
	printf("%*p\n", nodemask_pr_args(n));

	return 0;
}
-- 
Michal Hocko
SUSE Labs