On Mon, Dec 15, 2008 at 10:26:27PM +0100, Thomas Hellström wrote: > Hi! > > Sam Ravnborg wrote: > >On Sun, Dec 14, 2008 at 04:57:34PM +0100, Thomas Hellström wrote: > > > >>Hi! > >> > >>The intention is for via_drm.h to be self-containing when included both > >>for a kernel build and for a user-space build. In this particular case, > >>via_drmclient.h lives in the user-space clients and includes stdint.h to > >>get access to uint32_t and friends. > >> > >>Of course, the user-space clients could > >>#include "uint32.h" > >>#include "via_drm.h" > >> > >>but shouldn't really the tools be mimicing what the compiler does in > >>this case? > >> > > > >The kernel headers and thus the kernel ABI is separate and ideally > >they should not depend on any other header files to provide anything. > > > >This is why __u32, __u64 etc are preferred in the kernel ABI > >and not uint32_t as used by the drm headers. > > > >We do not adhere to this as a strict rule (yet). > >But if you do: > > > >grep -l uint32_t usr/include/linux > > > >then you will only see 7 hits. Out of 368 files. > >So we are not bad in this respect. > > > >For drm the fix seems simple - just replace all of uint32_t with __u32. > >likewise for the other 32 bit and the 64 bit variants. > > > I'm not sure that's possible, as these header files are shared with *BSD > (and possibly Solaris). I dunno about *BSD and solaris. But other linux ABI header files do: #include <linux/types.h> To get access to these basic types. And we could do the same for drm if the other systems does not prevent it. Sam -- To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
