On Wed, 2016-06-01 at 11:07 -0400, George Spelvin wrote: > On Wed, 01 Jun 2016 at 15:32:47, Andy Shevchenko wrote: > > On Tue, 2016-05-31 at 14:36 -0700, Joe Perches wrote: > > > On Tue, 2016-05-31 at 16:31 -0400, George Spelvin wrote: > > > > -static int __uuid_to_bin(const char *uuid, __u8 b[16], const u8 > > > > ei[16]) > > > > +static int __uuid_to_bin(const char uuid[36], __u8 b[16], const u8 > > > > si[16]) > > > Functions with sized array arguments are generally undesired. > > That function follows existing UUID API. Since we have now it > > consolidated in one place someone may fix it eventually. > Just to clarify: > > int foo(char *); > int foo(char *floccinaucinihilipilifcation); > int foo(char *p); > int foo(char p[]); > int foo(char []); > int foo(char p[1]); > int foo(char p[999999999]); > > are all the exact same declaration. There is no API change; the only > difference is stylistic. (Try it! Copy them to a .c file and > compile it. Observe the lack of conflicting declaration warnings.) > > Although the compiler doesn't care, I happen to prefer to include > parameter names for the benefit of someone reading the header files. > > For the same reason, if the pointer is to the start of an array with > definite length (that's not 1), I prefer to use the array form showing > that length. > > But it's a matter of style, not substance, either way.
I believe the substantive bit of the argument is the ability to misunderstand that sizeof(argument with size char[999]) is not 999 but is sizeof(*).
