When compiled in C mode, the following test emits warnings that point at each of the problem directives:

$ cat xyz.c && /build/gcc-trunk-svn/gcc/xgcc -B /build/gcc-trunk-svn/gcc -S -Wformat xyz.c
void f (void)
{
  __builtin_printf ("%s%s", 0, 0);
}
xyz.c: In function ‘f’:
xyz.c:3:23: warning: format ‘%s’ expects argument of type ‘char *’, but argument 2 has type ‘int’ [-Wformat=]
   __builtin_printf ("%s%s", 0, 0);
                       ^
xyz.c:3:25: warning: format ‘%s’ expects argument of type ‘char *’, but argument 3 has type ‘int’ [-Wformat=]
   __builtin_printf ("%s%s", 0, 0);
                         ^

However, when compiled in C++ mode, the column information is wrong:

$ cat xyz.c && /build/gcc-trunk-svn/gcc/xgcc -B /build/gcc-trunk-svn/gcc -S -Wformat -xc++ xyz.c
void f (void)
{
  __builtin_printf ("%s%s", 0, 0);
}
xyz.c: In function ‘void f()’:
xyz.c:3:33: warning: format ‘%s’ expects argument of type ‘char*’, but argument 2 has type ‘int’ [-Wformat=]
   __builtin_printf ("%s%s", 0, 0);
                                 ^
xyz.c:3:33: warning: format ‘%s’ expects argument of type ‘char*’, but argument 3 has type ‘int’ [-Wformat=]

The same problem occurs with other directives.  I found a long discussion of some of the issues in bug 52952 but based on the patches referenced from there it looks like they have been resolved (as least for C).