Re: [PATCH] Fix cpuset sched_relax_domain_level control file

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 7 May 2008 12:48:09 +0300 Adrian Bunk <bunk@kernel.org> wrote:

> On Tue, May 06, 2008 at 06:41:39PM -0700, Paul Menage wrote:
> > On Tue, May 6, 2008 at 6:38 PM, Andrew Morton <akpm@linux-foundation.org> wrote:
> > > On Tue, 06 May 2008 18:08:17 -0700 Paul Menage <menage@google.com> wrote:
> > >
> > >  > -static int update_relax_domain_level(struct cpuset *cs, char *buf)
> > >  > +static int update_relax_domain_level(struct cpuset *cs, s64 val)
> > >  >  {
> > >  > -     int val = simple_strtol(buf, NULL, 10);
> > >  > -
> > >  > -     if (val < 0)
> > >  > +     if ((int)val < 0)
> > >  >               val = -1;
> > >  >
> > >
> > >  Are you sure about the typecast here?  If `val' has a value of say
> > >  0x0000_ffff_ffff_ffff then I assume the casted value will be negative, only
> > >  it wasn't?
> > 
> > It seems like the simplest approach - if it's outside the range of a
> > positive int, set it to -1.
> 
> That's very hard to understand for someone who looks at the code - and 
> being able to understand the code is much more important than the 
> number of characters in the source code.
> 
> If you'd write something like
> 
>   if ((val < 0) || (val > INT_MAX))
> 
> instead it would be obvious for the reader what's happening here, and 
> that this was intended.

What he said.


Our poor reader now knows what was intended.  But he still doesn't know _why_
it was intended.