On 30.05.2015 18:46, Russell King - ARM Linux wrote: > On Sat, May 30, 2015 at 05:37:57PM +0200, Mikko Rapeli wrote: >> Fixes userspace compilation error: >> >> drm/exynos_drm.h:30:2: error: unknown type name âuint64_tâ >> >> Signed-off-by: Mikko Rapeli <mikko.rapeli at iki.fi> > This is another thing which we need to address. We should not be using > unsigned int/unsigned long/uintXX_t/etc in here, but the __uXX and __sXX > types. > > The lesson learned from other DRM developers is that using these types > simplifies the API especially when it comes to the differences between > 32-bit and 64-bit machines, and compat applications. > > Note that drm/drm.h is all that should need to be included - drm/drm.h > takes care of including linux/types.h when building on Linux platforms. > (note: if your compiler doesn't set __linux__ then you're probably not > using the proper compiler...) >
Using types that differs on 32-bit and 64-bit machines for a kernel interface is indeed a rather bad idea. This not only includes longs, but pointers as well. But the int8_t, int16_t int32_t, int64_t and their unsigned counterparts are defined in stdint.h which is part of the ISO/IEC 9899:1999 standard, similar is true for size_t. The __uXX, __sXX and _kernel_size_t types are linux specific as far as I know. For best interoperability and standard conformance I think that definitions from the C standard we use should out-rule linux specific definitions. Additional to that "linux/types.h" is not part of the uapi as far as I know, so including it in a header which is part of the uapi should be forbidden. So this is a NAK from my side for the whole series, userspace programs should include <stdint.h> for the definition of the ISO/IEC 9899:1999 standard types if necessary. Regards, Christian.
