On 11/19/2011 12:03 AM, Luis R. Rodriguez wrote: > On Fri, Nov 18, 2011 at 2:34 PM, John W. Linville > <linville@xxxxxxxxxxxxx> wrote: >> On Fri, Nov 18, 2011 at 02:26:21PM -0800, Joe Perches wrote: >>> On Fri, 2011-11-18 at 15:54 -0500, John W. Linville wrote: >>>> The way the compat-* header files are included causes the default >>>> pr_fmt definition from <linux/kernel.h> to be evaluated for every file. >>>> Files that define pr_fmt then generate a lot of build SPAM about >>>> pr_fmt being redefined. >>>> >>>> Eliminate the build noise by preemptively undefining pr_fmt in those >>>> files that define it. This is accomplished by adding a patch to the >>>> patches directory. >>> >>> Why not undef pr_fmt after the compat-* headers? >> >> That caused a build break. IIRC, not everyone that uses pr_* >> defines pr_fmt. If pr_fmt is undef'd then they don't compile. > > Hm, would it work if we undef but then define it? > > Luis We could restructure compat so that it will not include everything every time. Some time ago I tried to the extend every header with the things needed to backport and remove the compat-*.h files completely. The headers looked like this ( e.g. for include/linux/kernel.h): #include <linux/version.h> #include_next <linux/kernel.h> #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)) .. backport code for this kernel #endif #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,35)) .. backport code for this kernel #endif .... Then just the headers needed at that position are included and we would not have this problem with pr_fmt, but it takes some time to do this. Hauke -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
