Re: [PATCH] analyzer: deal with -fshort-enums

[David Malcolm <dmalcolm@redhat.com>]

On Sun, 2023-11-19 at 04:39 -0300, Alexandre Oliva wrote:
> 
> On platforms that enable -fshort-enums by default, various switch-
> enum
> analyzer tests fail, because apply_constraints_for_gswitch doesn't
> expect the integral promotion type cast.  I've arranged for the code
> to cope with those casts.

Thanks for the patch, and sorry for the failing tests.

Which tests failed?  I'm guessing one of the ones from r13-5159-
gccd4df81aa6537.

Can you add a copy of one of those tests that needs the patch, with an
explicit -fshort-enums, to ensure that regression testing is hitting
this case on all targets in the future?

Thanks
Dave

> 
> Regstrapped on x86_64-linux-gnu, also tested on arm-eabi with default
> cpu on trunk, and with tms570 on gcc-13.  Ok to install?
> 
> 
> for  gcc/analyzer/ChangeLog
> 
>         * region-model.cc (has_nondefault_case_for_value_p): Take
>         enumerate type as a parameter.
>         (region_model::apply_constraints_for_gswitch): Cope with
>         integral promotion type casts.
> ---
>  gcc/analyzer/region-model.cc |   27 +++++++++++++++++++++++----
>  1 file changed, 23 insertions(+), 4 deletions(-)
> 
> diff --git a/gcc/analyzer/region-model.cc b/gcc/analyzer/region-
> model.cc
> index dc83440652027..05f716cb7d663 100644
> --- a/gcc/analyzer/region-model.cc
> +++ b/gcc/analyzer/region-model.cc
> @@ -5340,10 +5340,10 @@ has_nondefault_case_for_value_p (const
> gswitch *switch_stmt, tree int_cst)
>     has nondefault cases handling all values in the enum.  */
>  
>  static bool
> -has_nondefault_cases_for_all_enum_values_p (const gswitch
> *switch_stmt)
> +has_nondefault_cases_for_all_enum_values_p (const gswitch
> *switch_stmt,
> +                                           tree type)
>  {
>    gcc_assert (switch_stmt);
> -  tree type = TREE_TYPE (gimple_switch_index (switch_stmt));
>    gcc_assert (TREE_CODE (type) == ENUMERAL_TYPE);
>  
>    for (tree enum_val_iter = TYPE_VALUES (type);
> @@ -5379,6 +5379,23 @@ apply_constraints_for_gswitch (const
> switch_cfg_superedge &edge,
>  {
>    tree index  = gimple_switch_index (switch_stmt);
>    const svalue *index_sval = get_rvalue (index, ctxt);
> +  bool check_index_type = true;
> +
> +  /* With -fshort-enum, there may be a type cast.  */
> +  if (ctxt && index_sval->get_kind () == SK_UNARYOP
> +      && TREE_CODE (index_sval->get_type ()) == INTEGER_TYPE)
> +    {
> +      const unaryop_svalue *unaryop = as_a <const unaryop_svalue *>
> (index_sval);
> +      if (unaryop->get_op () == NOP_EXPR
> +         && is_a <const initial_svalue *> (unaryop->get_arg ()))
> +       if (const initial_svalue *initvalop = (as_a <const
> initial_svalue *>
> +                                              (unaryop->get_arg
> ())))
> +         if (TREE_CODE (initvalop->get_type ()) == ENUMERAL_TYPE)
> +           {
> +             index_sval = initvalop;
> +             check_index_type = false;
> +           }
> +    }
>  
>    /* If we're switching based on an enum type, assume that the user
> is only
>       working with values from the enum.  Hence if this is an
> @@ -5390,12 +5407,14 @@ apply_constraints_for_gswitch (const
> switch_cfg_superedge &edge,
>        ctxt
>        /* Must be an enum value.  */
>        && index_sval->get_type ()
> -      && TREE_CODE (TREE_TYPE (index)) == ENUMERAL_TYPE
> +      && (!check_index_type
> +         || TREE_CODE (TREE_TYPE (index)) == ENUMERAL_TYPE)
>        && TREE_CODE (index_sval->get_type ()) == ENUMERAL_TYPE
>        /* If we have a constant, then we can check it directly.  */
>        && index_sval->get_kind () != SK_CONSTANT
>        && edge.implicitly_created_default_p ()
> -      && has_nondefault_cases_for_all_enum_values_p (switch_stmt)
> +      && has_nondefault_cases_for_all_enum_values_p (switch_stmt,
> +                                                    index_sval-
> >get_type ())
>        /* Don't do this if there's a chance that the index is
>          attacker-controlled.  */
>        && !ctxt->possibly_tainted_p (index_sval))
>