On Fri, May 04, 2018 at 02:35:49AM +0200, Thomas Huth wrote: > On 03.05.2018 21:51, Michael S. Tsirkin wrote: > > Each target is currently built with a different set of include > > directories, this is what makes it possible to pull in a separate copy > > of cpu.h depending on the target. > > > > Replace with per-target ifdefs which are easier to understand. > > > > Signed-off-by: Michael S. Tsirkin <[email protected]> > > --- > > include/cpu.h | 2 ++ > > 1 file changed, 2 insertions(+) > > create mode 100644 include/cpu.h > > > > diff --git a/include/cpu.h b/include/cpu.h > > new file mode 100644 > > index 0000000..b18f163 > > --- /dev/null > > +++ b/include/cpu.h > > @@ -0,0 +1,2 @@ > > +#include "target-dir.h" > > +#include TARGET_DIR(cpu.h) > > That's a *bad* idea. As far as I can see, this way, cpu.h is now > suddenly available to generic, non-target specific code. The headers in > target/* are "hidden" by purpose for all code that gets compiled via > common-obj-y and if you now make it available for everybody again, we'll > end up in ugly situation where common code might have been compiled with > target specific defines from target specific headers... Please don't do > that! > > Thomas
Are you sure that is the case? TARGET_XXX defines really shouldn't be set for non-target specific code. -- MST
