Re: kernel/workqueue.c:4435:61: sparse: expression using sizeof bool

Re: kernel/workqueue.c:4435:61: sparse: expression using sizeof bool

[Fengguang Wu]

On Fri, Jun 07, 2013 at 03:03:51PM -0700, Tejun Heo wrote:
> On Fri, Jun 07, 2013 at 02:15:21PM -0700, Andrew Morton wrote:
> > >   4434		/* copy worker description */
> > > > 4435		probe_kernel_read(&desc_valid, &worker->desc_valid, sizeof(desc_valid));
> > 
> > I don't understand the rationale for the sparse test, really.  This
> > code seems unproblematic.
> 
> Yeah, I've gotten that warning a couple times now and am just ignoring
> it at this point.  Maybe the standard says bool isn't a proper
> integeral type and its size isn't precisely defined or something?  I
> don't know.

Hmm, I'd better disable the warnings that are normally ignored anyway..

Thanks,
Fengguang

Re: kernel/workqueue.c:4435:61: sparse: expression using sizeof bool

[Fengguang Wu]

// CC sparse people.

On Fri, Jun 07, 2013 at 07:05:49PM -0700, Tejun Heo wrote:
> Hello, Andrew.
> 
> On Fri, Jun 07, 2013 at 06:46:14PM -0700, Andrew Morton wrote:
> > I believe sizeof(bool) can vary between compilers (on gcc it's 1) but
> > that doesn't matter here.
> 
> Looking at the warning, I originally suspected that maybe the standard
> was crazy enough to allow mixed usages of different sizes for bool as
> the compiler sees fit.  Just scanned the spec, nothing that insane.
> It's a proper integeral type, so I'm now more puzzled why sparse is
> warning about sizeof(bool).
> 
> > sizeof(enum) can vary between compilers as well.  In fact I've seen a
> > compiler which used 2 if the enumerated values were 0..65535, and 4
> > otherwise.  So warning about sizeof(enum) would be a bit more useful
> > than about sizeof(bool).
> 
> Oh yeah, gcc does that too.  If the enum definition contains a number
> larger than 32bit, it'll grow it to 64bit.
> 
>  #include <stdio.h>
> 
>  enum { ENUM_A = 1LU };
>  enum { ENUM_B = 1LU << 31 };
>  enum { ENUM_C = 1LU << 32 };
> 
>  int main(void)
>  {
> 	 printf("%zu %zu %zu\n", sizeof(ENUM_A), sizeof(ENUM_B), sizeof(ENUM_C));
> 	 return 0;
>  }
> 
>  $ ./a.out
>  4 4 8
> 
> Whether sparse should trigger a warning on it, I don't know.
> 
> Thanks.
> 
> -- 
> tejun