On Wed, Sep 17, 2008 at 2:58 PM, Stefanik Gábor <netrolller.3d@xxxxxxxxx> wrote:> On Wed, Sep 17, 2008 at 11:52 PM, Johannes Berg> <johannes@xxxxxxxxxxxxxxxx> wrote:>> On Wed, 2008-09-17 at 23:46 +0200, Davide Pesavento wrote:>>> $ make>>> GEN version.h>>> CC iw.o>>> CC info.o>>> info.c: In function 'print_phy_handler':>>> info.c:108: error: 'NL80211_ATTR_SUPPORTED_IFTYPES' undeclared (first>>> use in this function)>>> info.c:108: error: (Each undeclared identifier is reported only once>>> info.c:108: error: for each function it appears in.)>>> make: *** [info.o] Error 1>>>>>> I guess I have to build iw against >=2.6.27-rcX kernel headers, right?>>>> Actually this one is new in wireless-testing, I have a branch you can>> build against 2.6.27 but it's not very up-to-date.>>>> Any ideas how to handle this are welcome. Should it just ship the header>> file? I'd prefer not to, otoh doing it isn't really harmful because when>> adding a new command clearly you need to edit iw sources.>>>> johannes>>>>>> IMO that's the best thing to do - in fact, that's exactly why I> branched. Distros likely won't ship kernel headers that don't match> the kernel binaries and sources. (This is likely why Debian is> shipping my branched version, rather than the official.)
People wanting to test bleeding edge and not compile their own kernelsimply can use compat-wireless crap and that has its own nl80211.h. Wecan modify iw to let COMPAT_TREE be defined and if so look therefirst. This is what CRDA does for example.
I don't see why packages should ship their own nl80211.h unless youare replacing or providing a new cfg80211.ko.
Luis��.n��������+%������w��{.n�����{���zW����ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
