svn commit: r186252 - head/sys/kern

[John Baldwin]

On Saturday 27 December 2008 4:47:45 pm Antoine Brodin wrote:
> On Sat, Dec 27, 2008 at 10:19 PM,  <ivoras at freebsd.org> wrote:
> > On 27/12/2008, Antoine Brodin <antoine at freebsd.org> wrote:
> >> On Wed, Dec 17, 2008 at 8:57 PM, Ivan Voras <ivoras at freebsd.org> wrote:
> >
> >> Modified: head/sys/kern/subr_param.c
> >> [snip]
> >>> +enum VM_GUEST { VM_GUEST_NO, VM_GUEST_VM, VM_GUEST_XEN };
> >>> +
> >> [snip]
> >>> -               hz = detect_virtual() ? HZ_VM : HZ;
> >>> +               hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
> >> [snip]
> >>
> >> Hi Ivan,
> >>
> >> Could you change VM_GUEST_NO / VM_GUEST_VM / VM_GUEST_XEN to a #define
> >> or explicitly initialize them?
> >> The magnitude comparison between vm_guest and VM_GUEST_NO looks like
> >> gratuitous obfuscation.
> >
> > I think that the "enum" type is very well defined and its behaviour
> > widely known so it is not necessary to break it into #defines. Would
> > you be happy if I explicitly initalized the first member of the enum
> > to signify its values are important?
> >
> > enum VM_GUEST { VM_GUEST_NO = 0, VM_GUEST_VM, VM_GUEST_XEN };
> >
> 
> I would be happy if you initialized explicitly the 3 values.
> You use these values as an index in the vm_guest_sysctl_names array
> and you compare them in init_param1() so I think it's better to be
> explicit.

The definition of enum values is well defined in C.  Thus, one should not be 
explicitly initializing all the values, and using > to compare enum values is 
perfectly acceptable.

-- 
John Baldwin