Re: PATH_MAX on the Hurd

[Eric Blake]

On 08/05/2011 05:19 PM, Paul Eggert wrote:
> On 08/05/2011 03:33 PM, Bruno Haible wrote:
> > > > int a = PATH_MAX;
> > > >
> > > > Shouldn't this be:
> > > >
> > > > static int a[PATH_MAX];
> > Maybe. Or possibly:
> >    enum { a = PATH_MAX };
> >
> > Are there cases when these three declarations don't all succeed and don't
> > all fail?
>
> Oh yes.  :-)
>
> "int a = PATH_MAX;" would allow a float PATH_MAX, which presumably
> you'd rather reject.  "static int a[PATH_MAX];" might warn you about
> an array too large if PATH_MAX == SIZE_MAX; I'd make it "static char
> a[PATH_MAX];".  "enum { a = PATH_MAX };" won't work if PATH_MAX
> exceeds INT_MAX, and also it doesn't check that PATH_MAX is positive;
> these problems are both fixable, but since the intended use of
> PATH_MAX is likely array sizes, the array's probably your best bet.

What if we go with the array approach, but without risking 
overallocation problems?  Would this work for a single-byte array:

static char a[PATH_MAX / PATH_MAX];

or is it worth considering the constant context in bit field width:

struct s {
  int a : PATH_MAX / PATH_MAX;
} s1;

--
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org