Re: linux-next: build warnings from Linus' tree

[Theodore Y. Ts'o]

On Mon, Aug 20, 2018 at 08:13:23AM +1000, Stephen Rothwell wrote:
> fs/ext4/super.c: In function '__save_error_info':
> fs/ext4/super.c:344:2: warning: 'strncpy' specified bound 32 equals destination size [-Wstringop-truncation]
>   strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
>   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> fs/ext4/super.c:349:3: warning: 'strncpy' specified bound 32 equals destination size [-Wstringop-truncation]
>    strncpy(es->s_first_error_func, func,
>    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>     sizeof(es->s_first_error_func));
>     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

All of ext4 superblock char[] fields are not necessarily null
terminated, so this is a false positive.  I suppose we could do
something like this:

inline char *
strncpy_I_solemnly_swear_I_know_what_I_am_doing(char *dest,
			const char *src, size_t n)
{
#if __GNUC_PREREQ (8, 2)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wstringop-truncate"
#endif
	return strncpy(dest, src, n);
#if __GNUC_PREREQ (8, 2)
#pragma GCC diagnostic pop
#endif
}

(if we really think this warning is worthwhile enough that we don't
just want to globally disable it, of course)

       		      	       - Ted

P.S.  It's really, really too bad there isn't a simpler way to shut up
gcc.  You need the #ifdef __GNUC_PREREQ nonsense because otherwise
older versions of gcc that don't understand the particular warning
you're trying to suppress will complain loudly.  (Ask me how I
know....)