Re: [PATCH] ath6kl: fix incorrect use of IEEE80211_NUM_BANDS

[Pavel Roskin <proski@gnu.org>]

On Fri, 20 Jul 2012 09:18:17 +0300
Kalle Valo <kvalo@qca.qualcomm.com> wrote:

> On 07/20/2012 12:11 AM, Pavel Roskin wrote:
> > On Thu, 19 Jul 2012 14:08:44 +0300
> > Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> >> On 07/12/2012 08:48 PM, Pavel Roskin wrote:
> >>> On Thu, 12 Jul 2012 12:13:12 +0300
> >>> Kalle Valo <kvalo@qca.qualcomm.com> wrote:
> >>>
> >>>> +	/* only check 2.4 and 5 GHz bands, skip the rest */
> >>>> +	for (band = 0; band <= IEEE80211_BAND_5GHZ; band++) {
> >>>
> >>> There is something inelegant here.  The code is mixing an integer
> >>> and an enum.  I'd rather go with one or those:
> >>>
> >>> two enums:
> >>> for (band = IEEE80211_BAND_2GHZ; band <= IEEE80211_BAND_5GHZ;
> >>> band++) {
> >>
> >> I somewhat see your point. But IMHO zero is commonly used when
> >> iterating over an enum to denote the first value and I don't see
> >> how IEEE80211_BAND_2GHZ helps here.
> > 
> > It's the lowest band we support.  What if the 900MHz band is added
> > one day?
> 
> Then that should be added to the end of the enum, not beginning. I
> think it would be bad if we change enum values on the fly.

Actually, it should be OK.  It's not like it would affect the
userspace.

I believe that properly written code should not rely on the numeric
values of enums.  Well, if 0 means something very special (like no
error), it should not be changed.  The (ab)uses of enums for bitmasks
should be exempted too.  But IEEE80211_BAND_2GHZ is just one of the
bands.

-- 
Regards,
Pavel Roskin