Hi John, On Tue, 27 Apr 2010 10:07:12 -0700 Steve deRosier <steve@xxxxxxxxxxx> wrote: > > On Tue, Apr 27, 2010 at 8:40 AM, John W. Linville > <linville@xxxxxxxxxxxxx> wrote: > > > > Right. So in wireless-testing I did the includes in the other order > > (i.e. "deb_defs.h" first), but that is a bit ugly. Any suggestions > > on alternatives? I have done that as well for today in linux-next. > > "#undef pr_fmt" just before the "#define pr_fmt(fmt)..." line in > > db_defs.h seems to eliminate the warning even with the more normal > > ordering of the #include lines. I'm not familiar with the usage of > > pr_fmt -- will doing the above preserve the desired effect? > > I thought about that particular strategy (doing the #undef) instead of > the non-traditional include mess. But not being familiar enough with > the pr_fmt stuff, I didn't want to do it. > > My goal was to get the '#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt' > line in the deb_defs.h header so it was only in one place. But to > build, that define must be before kernel.h gets included anywhere. > Hence the current mess. > > I'm OK with the #undef strategy and moving the deb_defs.h include to a > better position if that's the correct way to do this. Please let me > know if I have any action items on this. The only problem I could see with that is is there are some inline functions in the headers files between where the first and second #define pr_fmt s get done. -- Cheers, Stephen Rothwell sfr@xxxxxxxxxxxxxxxx http://www.canb.auug.org.au/~sfr/
Attachment:
pgpBanJmkx1S8.pgp
Description: PGP signature