[kbuild-all] Re: [stable:linux-4.9.y 66/1284] drivers/iio/magnetometer/ak8974.c:511:40: sparse: sparse: incorrect type in argument 2 (different base types)

On Sat, 12 Dec 2020 15:21:10 +0100 Linus Walleij <linus.walleij(a)linaro.org&gt; wrote: > On Fri, Dec 11, 2020 at 3:27 PM kernel test robot <lkp(a)intel.com&gt; wrote: > > > "sparse warnings: (new ones prefixed by >>)" > > drivers/iio/magnetometer/ak8974.c:408:16: sparse: sparse: cast to restricted __le16 > > drivers/iio/magnetometer/ak8974.c:485:29: sparse: sparse: cast to restricted __le16 > > >> drivers/iio/magnetometer/ak8974.c:511:40: sparse: sparse: incorrect type in argument 2 (different base types) @@ expected signed short [usertype] *result @@ got restricted __le16 * @@ > > drivers/iio/magnetometer/ak8974.c:511:40: sparse: expected signed short [usertype] *result > > drivers/iio/magnetometer/ak8974.c:511:40: sparse: got restricted __le16 * > > I don't understand this, is sparse warning about implicit casting __le16 > to signed short or the other way around? > > It seems to me to reasonable to allow anyway, I don't even see how > we could avoid that except using explicit casts. Looks to me like we didn't bother backporting a tidy up of the endian types in here to 4.9. That changed the argument types to reflect that they were arrays of __le16. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit... Personally I'm not that bothered about leaving this in place. I don't think its an actual bug after all.