Re: [PATCH 05/19] cfg80211: processing core regulatory hints on its own

[Johannes Berg <johannes@sipsolutions.net>]

On Thu, 2013-11-07 at 11:47 +1100, Julian Calaby wrote:

> >>> -     switch (__regulatory_hint(wiphy, reg_request)) {
> >>> +     switch (reg_request->initiator) {
> >>> +     case NL80211_REGDOM_SET_BY_CORE:
> >>> +             reg_process_hint_core(reg_request);
> >>> +             return;
> >
> > Note that for the core request we bail out early.
> >
> >>> +     case NL80211_REGDOM_SET_BY_USER:
> >>> +     case NL80211_REGDOM_SET_BY_DRIVER:
> >>> +     case NL80211_REGDOM_SET_BY_COUNTRY_IE:
> >>> +             treatment = __regulatory_hint(wiphy, reg_request);
> >>> +             break;
> >>> +     }
> >>> +
> >>> +     switch (treatment) {
> >>
> >> is used uninitialized here
> >
> > As I see it in the code we either bail early or its assigned to the
> > return value of __regulatory_hint() for the non-core case. Did I miss
> > something?
> 
> I think Johannes just wants you to assign something to it so that
> paranoid static checkers / compilers won't complain that it's unset in
> the potential case where we make it through the switch without hitting
> any of the cases. (Adding a default: case to the switch statement
> would also work, but as you've been very very careful to spell out all
> the cases here that'd tempt developers to remove all the careful
> spelling out in some crazed cleanup binge.)
> 
> Of course this will never happen if reg_request->initiator is an enum
> and all it's possible values are covered in the switch statement, but
> some compilers (and developers like me) need it spelled out for them.

Well actually I was just careless and didn't realize that it was
actually a false compiler warning :)

johannes