On 20 Oct 2022 09:32, Tsukasa OI wrote:
> Clang generates a warning if a variable may be used uninitialized on some
> cases ("-Wsometimes-uninitialized").  On the default configuration, it
> causes a build failure (unless "--disable-werror" is specified).
> 
> The cause of this error, $(builddir)/sim/sh/ppi.c is generated from
> $(srcdir)/sim/sh/gencode.c.  Clang will detect the variable res may be used
> uninitialized when used on some cases.
> 
> Likewise, GCC generates a warning if a variable may be used uninitialized
> on some cases ("-Wmaybe-uninitialized").
> 
> GCC will detect variables res, res_grd, carry, overflow and greater_equal
> may be used uninitialized on some cases.  Despite that GCC will not cause a
> build failure even when "--enable-werror" is specified, it would be better
> to fix this as well since the cause is in the same function.
> ---
>  sim/sh/gencode.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> --- a/sim/sh/gencode.c
> +++ b/sim/sh/gencode.c
> @@ -3257,8 +3257,8 @@ ppi_gensim (void)
>    printf ("  static char const u_tab[] = { 8, 10,  7,  5};\n");
>    printf ("\n");
>    printf ("  int z;\n");
> -  printf ("  int res, res_grd;\n");
> -  printf ("  int carry, overflow, greater_equal;\n");
> +  printf ("  int res = 0, res_grd = 0;\n");
> +  printf ("  int carry = 0, overflow = 0, greater_equal = 0;\n");

i've left this warning in because i don't know what the right answer is, and
i'm not inclined to dig into the ISA to find out.  simply initializing the
variables to 0 might mean we're masking real bugs where we should be loading
the state from the current CPU registers instead.
-mike