On 08/23/16 13:42, Mikko Rapeli wrote: > On Tue, Aug 23, 2016 at 02:28:19PM +0000, Bart Van Assche wrote: >> On 08/23/16 06:57, Bart Van Assche wrote: >>> On 08/22/16 11:32, Mikko Rapeli wrote: >>>> - * uint32_t (*get_region_size)(struct dm_dirty_log *log); >>>> + * __u32 (*get_region_size)(struct dm_dirty_log *log); >>> >>> uint32_t is a type that is defined by ANSI C but __u32 not. So this >>> change looks wrong to me. Would it have been sufficient to add "#include >>> <linux/types.h>" and keep the uint32_t etc. type names? >> >> Answering my own question: adding "#include <linux/types.h>" wouldn't be >> sufficient. How about adding the following code that also occurs in a >> few other uapi header files? >> >> #ifndef __KERNEL__ >> #include <stdint.h> >> #endif >> #include <linux/types.h> > > I have tried that before but I was instructed to use the linux/types.h > versions. For example https://lkml.org/lkml/2015/6/1/160 > > But lately drm and fuse and some others have questioned this approach and > would like to use/continue using C99 stdint.h types. Hello Mikko, Thanks for the feedback. I have a minor comment though about your patch: are you aware that you have changed uint64_t into __u64 in source code comments but that int64_t has not been changed into __s64? Bart. -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
