[C++] Partial fix for a recent regression (PR c++/90947)

Hi!

The following patch is just a partial fix for a regression introduced
in the PR90947 changes, the testcase is fixed for just C++17/20.
type_initializer_zero_p has been added to the generic code, supposedly
because similar initializer_zerop is in generic code too, but that
means it has a hand-written copy of next_initializable_field which can't
do exactly what next_initializable_field does.

The following patch moves it into the C++ FE which is the only user of that
function and uses next_initializable_field in there.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

I'm afraid I'm lost on how to fix the C++11/14 case, the object
being initialized has std::atomic<int> type, which doesn't have any
direct non-static data members, and has std::__atomic_base<int> as base
class that has some field.  The only FIELD_DECL in std::atomic<int>
is DECL_ARTIFICIAL field for the base class, so next_initializable_field
or the hand-written variant thereof doesn't find any initializable fields
for 11/14.  The initializer is initializer list { {1} } and the function
just returns true if it doesn't find any initializable fields, so in the end
we misoptimize it as { {0} }.

2019-10-22  Jakub Jelinek  <jakub@redhat.com>

	PR c++/90947
	* tree.h (type_initializer_zero_p): Remove.
	* tree.c (type_initializer_zero_p): Remove.
cp/
	* cp-tree.h (type_initializer_zero_p): Declare.
	* decl.c (reshape_init_array_1): Formatting fix.
	* tree.c (type_initializer_zero_p): New function.  Moved from
	../tree.c, use next_initializable_field, formatting fix.


	Jakub

On 10/22/19 11:03 AM, Jakub Jelinek wrote:
> Hi!
> 
> The following patch is just a partial fix for a regression introduced
> in the PR90947 changes, the testcase is fixed for just C++17/20.
> type_initializer_zero_p has been added to the generic code, supposedly
> because similar initializer_zerop is in generic code too, but that
> means it has a hand-written copy of next_initializable_field which can't
> do exactly what next_initializable_field does.
> 
> The following patch moves it into the C++ FE which is the only user of that
> function and uses next_initializable_field in there.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
> 
> I'm afraid I'm lost on how to fix the C++11/14 case, the object
> being initialized has std::atomic<int> type, which doesn't have any
> direct non-static data members, and has std::__atomic_base<int> as base
> class that has some field.  The only FIELD_DECL in std::atomic<int>
> is DECL_ARTIFICIAL field for the base class, so next_initializable_field
> or the hand-written variant thereof doesn't find any initializable fields
> for 11/14.  The initializer is initializer list { {1} } and the function
> just returns true if it doesn't find any initializable fields, so in the end
> we misoptimize it as { {0} }.

I think type_initializer_zero_p should return false if 
CLASSTYPE_NON_AGGREGATE; we can't expect that value-initialization will 
have the intended effect in that case.

> 2019-10-22  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR c++/90947
> 	* tree.h (type_initializer_zero_p): Remove.
> 	* tree.c (type_initializer_zero_p): Remove.
> cp/
> 	* cp-tree.h (type_initializer_zero_p): Declare.
> 	* decl.c (reshape_init_array_1): Formatting fix.
> 	* tree.c (type_initializer_zero_p): New function.  Moved from
> 	../tree.c, use next_initializable_field, formatting fix.
> 
> --- gcc/tree.h.jj	2019-09-20 12:25:13.737920929 +0200
> +++ gcc/tree.h	2019-10-22 10:07:26.411826804 +0200
> @@ -4690,12 +4690,6 @@ extern tree first_field (const_tree);
>   extern bool initializer_zerop (const_tree, bool * = NULL);
>   extern bool initializer_each_zero_or_onep (const_tree);
>   
> -/* Analogous to initializer_zerop but also examines the type for
> -   which the initializer is being used.  Unlike initializer_zerop,
> -   considers empty strings to be zero initializers for arrays and
> -   non-zero for pointers.  */
> -extern bool type_initializer_zero_p (tree, tree);
> -
>   extern wide_int vector_cst_int_elt (const_tree, unsigned int);
>   extern tree vector_cst_elt (const_tree, unsigned int);
>   
> --- gcc/tree.c.jj	2019-10-19 09:22:14.830893404 +0200
> +++ gcc/tree.c	2019-10-22 10:08:36.501748481 +0200
> @@ -11396,73 +11396,6 @@ initializer_each_zero_or_onep (const_tre
>       }
>   }
>   
> -/* Given an initializer INIT for a TYPE, return true if INIT is zero
> -   so that it can be replaced by value initialization.  This function
> -   distinguishes betwen empty strings as initializers for arrays and
> -   for pointers (which make it return false).  */
> -
> -bool
> -type_initializer_zero_p (tree type, tree init)
> -{
> -  if (type  == error_mark_node || init == error_mark_node)
> -    return false;
> -
> -  STRIP_NOPS (init);
> -
> -  if (POINTER_TYPE_P (type))
> -    return TREE_CODE (init) != STRING_CST && initializer_zerop (init);
> -
> -  if (TREE_CODE (init) != CONSTRUCTOR)
> -    return initializer_zerop (init);
> -
> -  if (TREE_CODE (type) == ARRAY_TYPE)
> -    {
> -      tree elt_type = TREE_TYPE (type);
> -      elt_type = TYPE_MAIN_VARIANT (elt_type);
> -      if (elt_type == char_type_node)
> -	return initializer_zerop (init);
> -
> -      tree elt_init;
> -      unsigned HOST_WIDE_INT i;
> -      FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (init), i, elt_init)
> -	if (!type_initializer_zero_p (elt_type, elt_init))
> -	  return false;
> -      return true;
> -    }
> -
> -  if (TREE_CODE (type) != RECORD_TYPE)
> -    return initializer_zerop (init);
> -
> -  tree fld = TYPE_FIELDS (type);
> -
> -  tree fld_init;
> -  unsigned HOST_WIDE_INT i;
> -  FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (init), i, fld_init)
> -    {
> -      /* Advance to the next member, skipping over everything that
> -	 canot be initialized (including unnamed bit-fields).  */
> -      while (TREE_CODE (fld) != FIELD_DECL
> -	     || DECL_ARTIFICIAL (fld)
> -	     || (DECL_BIT_FIELD (fld) && !DECL_NAME (fld)))
> -	{
> -	  fld = DECL_CHAIN (fld);
> -	  if (!fld)
> -	    return true;
> -	  continue;
> -	}
> -
> -      tree fldtype = TREE_TYPE (fld);
> -      if (!type_initializer_zero_p (fldtype, fld_init))
> -	return false;
> -
> -      fld = DECL_CHAIN (fld);
> -      if (!fld)
> -	break;
> -    }
> -
> -  return true;
> -}
> -
>   /* Check if vector VEC consists of all the equal elements and
>      that the number of elements corresponds to the type of VEC.
>      The function returns first element of the vector
> --- gcc/cp/cp-tree.h.jj	2019-10-22 08:15:53.810775827 +0200
> +++ gcc/cp/cp-tree.h	2019-10-22 10:10:57.265582861 +0200
> @@ -7379,6 +7379,11 @@ extern tree cxx_copy_lang_qualifiers		(c
>   
>   extern void cxx_print_statistics		(void);
>   extern bool maybe_warn_zero_as_null_pointer_constant (tree, location_t);
> +/* Analogous to initializer_zerop but also examines the type for
> +   which the initializer is being used.  Unlike initializer_zerop,
> +   considers empty strings to be zero initializers for arrays and
> +   non-zero for pointers.  */
> +extern bool type_initializer_zero_p		(tree, tree);
>   
>   /* in ptree.c */
>   extern void cxx_print_xnode			(FILE *, tree, int);
> --- gcc/cp/decl.c.jj	2019-10-22 08:57:12.913654620 +0200
> +++ gcc/cp/decl.c	2019-10-22 10:13:39.038094034 +0200
> @@ -5982,9 +5982,8 @@ reshape_init_array_1 (tree elt_type, tre
>         /* Pointers initialized to strings must be treated as non-zero
>   	 even if the string is empty.  */
>         tree init_type = TREE_TYPE (elt_init);
> -      if ((POINTER_TYPE_P (elt_type) != POINTER_TYPE_P (init_type)))
> -	last_nonzero = index;
> -      else if (!type_initializer_zero_p (elt_type, elt_init))
> +      if (POINTER_TYPE_P (elt_type) != POINTER_TYPE_P (init_type)
> +	  || !type_initializer_zero_p (elt_type, elt_init))
>   	last_nonzero = index;
>   
>         /* This can happen with an invalid initializer (c++/54501).  */
> --- gcc/cp/tree.c.jj	2019-10-19 09:22:16.594866462 +0200
> +++ gcc/cp/tree.c	2019-10-22 10:10:29.151015397 +0200
> @@ -5527,6 +5527,65 @@ maybe_warn_zero_as_null_pointer_constant
>     return false;
>   }
>   
> +/* Given an initializer INIT for a TYPE, return true if INIT is zero
> +   so that it can be replaced by value initialization.  This function
> +   distinguishes betwen empty strings as initializers for arrays and
> +   for pointers (which make it return false).  */
> +
> +bool
> +type_initializer_zero_p (tree type, tree init)
> +{
> +  if (type == error_mark_node || init == error_mark_node)
> +    return false;
> +
> +  STRIP_NOPS (init);
> +
> +  if (POINTER_TYPE_P (type))
> +    return TREE_CODE (init) != STRING_CST && initializer_zerop (init);
> +
> +  if (TREE_CODE (init) != CONSTRUCTOR)
> +    return initializer_zerop (init);
> +
> +  if (TREE_CODE (type) == ARRAY_TYPE)
> +    {
> +      tree elt_type = TREE_TYPE (type);
> +      elt_type = TYPE_MAIN_VARIANT (elt_type);
> +      if (elt_type == char_type_node)
> +	return initializer_zerop (init);
> +
> +      tree elt_init;
> +      unsigned HOST_WIDE_INT i;
> +      FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (init), i, elt_init)
> +	if (!type_initializer_zero_p (elt_type, elt_init))
  > +	  return false;
> +      return true;
> +    }
> +
> +  if (TREE_CODE (type) != RECORD_TYPE)
> +    return initializer_zerop (init);
> +
> +  tree fld = TYPE_FIELDS (type);
> +
> +  tree fld_init;
> +  unsigned HOST_WIDE_INT i;
> +  FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (init), i, fld_init)
> +    {
> +      fld = next_initializable_field (fld);
> +      if (!fld)
> +	return true;
> +
> +      tree fldtype = TREE_TYPE (fld);
> +      if (!type_initializer_zero_p (fldtype, fld_init))
> +	return false;
> +
> +      fld = DECL_CHAIN (fld);
> +      if (!fld)
> +	break;
> +    }
> +
> +  return true;
> +}
> +
>   #if defined ENABLE_TREE_CHECKING && (GCC_VERSION >= 2007)
>   /* Complain that some language-specific thing hanging off a tree
>      node has been accessed improperly.  */
> 
> 	Jakub
>