On Monday 16 July 2012 01:26:50 Stefan Weil wrote: > Am 15.07.2012 23:54, schrieb Mike Frysinger: > > On Sunday 15 July 2012 15:34:33 Stefan Weil wrote: > >> Am 15.07.2012 22:25, schrieb Mike Frysinger: > >>> This file uses the define HOST_LONG_BITS, but doesn't explicitly > >>> include qemu-common.h for it leading to build warnings for some > >>> setups: In file included from qemu/target-bfin/cpu.h:17, > >>> > >>> from qemu/cputlb.c:21: > >>> qemu/cpu-defs.h:83:5: warning: "HOST_LONG_BITS" is not defined > >>> > >>> Signed-off-by: Mike Frysinger <vap...@gentoo.org> > >>> --- > >>> > >>> cpu-defs.h | 1 + > >>> 1 file changed, 1 insertion(+) > >>> > >>> diff --git a/cpu-defs.h b/cpu-defs.h > >>> index f49e950..0d6018d 100644 > >>> --- a/cpu-defs.h > >>> +++ b/cpu-defs.h > >>> @@ -28,6 +28,7 @@ > >>> > >>> #include <inttypes.h> > >>> #include <signal.h> > >>> #include "osdep.h" > >>> > >>> +#include "qemu-common.h" > >>> > >>> #include "qemu-queue.h" > >>> #include "targphys.h" > >> > >> No. Of course this works, but I don't think that it is reasonable > >> to include qemu-common.h in every *.h file. There are already too > >> many of them. > >> > >> target-bfin/cpu.h should start like all other cpu.h files with > > > >> these include statements: > > sorry, but that's fragile junk. if a header file uses defines from > > another header file, it should be including it. > > -mike > > There are different ways how things can be done. > > Normally, I agree with you that each header file should be complete, > but that's not the QEMU style. > > In your special case, it's more important to keep all */cpu.h similar. > qemu/target-bfin/cpu.h is still not part of the official QEMU code, > so it can be fixed before it is committed.
a lot of existing files in the top level pull in qemu-common.h. i don't think this is a special case considering it's the first failure i've seen since i started the Blackfin port over a year ago. -mike
signature.asc
Description: This is a digitally signed message part.
