[PATCH v2] c++: Implement P1009: Array size deduction in new-expressions.
Jason Merrill
jason@redhat.com
Tue Aug 25 02:49:38 GMT 2020
On 8/24/20 5:44 PM, Marek Polacek wrote:
> On Fri, Aug 21, 2020 at 05:04:43PM -0400, Jason Merrill wrote:
>> On 8/20/20 4:22 PM, Marek Polacek wrote:
>>> @@ -3917,6 +3926,47 @@ build_new (location_t loc, vec<tree, va_gc> **placement, tree type,
>>> return error_mark_node;
>>> }
>>> + /* P1009: Array size deduction in new-expressions. */
>>> + if (TREE_CODE (type) == ARRAY_TYPE
>>> + && !TYPE_DOMAIN (type)
>>> + && *init)
>>> + {
>>> + /* This means we have 'new T[]()'. */
>>> + if ((*init)->is_empty ())
>>> + {
>>> + tree ctor = build_constructor (init_list_type_node, NULL);
>>> + CONSTRUCTOR_IS_DIRECT_INIT (ctor) = true;
>>> + vec_safe_push (*init, ctor);
>>> + }
>>> + tree &elt = (**init)[0];
>>> + /* The C++20 'new T[](e_0, ..., e_k)' case allowed by P0960. */
>>> + if (!DIRECT_LIST_INIT_P (elt) && cxx_dialect >= cxx20)
>>> + {
>>> + /* Handle new char[]("foo"). */
>>> + if (vec_safe_length (*init) == 1
>>> + && char_type_p (TYPE_MAIN_VARIANT (TREE_TYPE (type)))
>>> + && TREE_CODE (tree_strip_any_location_wrapper (elt))
>>> + == STRING_CST)
>>> + /* Leave it alone: the string should not be wrapped in {}. */;
>>> + else
>>> + {
>>> + /* Create a CONSTRUCTOR from the vector INIT. */
>>> + tree list = build_tree_list_vec (*init);
>>> + tree ctor = build_constructor_from_list (init_list_type_node, list);
>>> + CONSTRUCTOR_IS_DIRECT_INIT (ctor) = true;
>>> + CONSTRUCTOR_IS_PAREN_INIT (ctor) = true;
>>
>> It seems wasteful to build a TREE_LIST only to throw it away; let's add a
>> function to build a CONSTRUCTOR directly from a vec<tree>*. And use it in
>> build_new_method_call_1 as well.
>
> Good point. Done.
>
>>> @@ -9082,14 +9084,17 @@ cp_parser_direct_new_declarator (cp_parser* parser)
>>> cp_parser_require (parser, CPP_OPEN_SQUARE, RT_OPEN_SQUARE);
>>> token = cp_lexer_peek_token (parser->lexer);
>>> - expression = cp_parser_expression (parser);
>>> + if (token->type == CPP_CLOSE_SQUARE)
>>> + expression = NULL_TREE;
>>
>> Only the first bound is optional; we need to require the expression for
>> subsequent bounds.
>
> Fixed (though the diagnostic we gave was pretty good). New test added.
>
> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
>
> -- >8 --
> This patch implements C++20 P1009, allowing code like
>
> new double[]{1,2,3}; // array bound will be deduced
>
> Since this proposal makes the initialization rules more consistent, it is
> applied to all previous versions of C++ (thus, effectively, all the way back
> to C++11).
>
> My patch is based on Jason's patch that handled the basic case. I've
> extended it to work with ()-init and also the string literal case.
> Further testing revealed that to handle stuff like
>
> new int[]{t...};
>
> in a template, we have to consider such a NEW_EXPR type-dependent.
> Obviously, we first have to expand the pack to be able to deduce the
> number of elements in the array.
>
> Curiously, while implementing this proposal, I noticed that we fail
> to accept
>
> new char[4]{"abc"};
>
> so I've assigned 77841 to self. I think the fix will depend on the
> build_new_1 hunk in this patch.
>
> The new tree.c function build_constructor_from_vec helps us morph
> a vector into a CONSTRUCTOR more efficiently.
>
> gcc/cp/ChangeLog:
>
> PR c++/93529
> * call.c (build_new_method_call_1): Use build_constructor_from_vec
> instead of build_tree_list_vec + build_constructor_from_list.
> * init.c (build_new_1): Handle new char[]{"foo"}. Use
> build_constructor_from_vec instead of build_tree_list_vec +
> build_constructor_from_list.
> (build_new): Deduce the array size in new-expression if not
> present. Handle ()-init. Handle initializing an array from
> a string literal.
> * parser.c (cp_parser_new_type_id): Leave [] alone.
> (cp_parser_direct_new_declarator): Allow [].
> * pt.c (type_dependent_expression_p): In a NEW_EXPR, consider
> array types whose dimension has to be deduced type-dependent.
>
> gcc/ChangeLog:
>
> PR c++/93529
> * tree.c (build_constructor_from_vec): New.
> * tree.h (build_constructor_from_vec): Declare.
>
> gcc/testsuite/ChangeLog:
>
> PR c++/93529
> * g++.dg/cpp0x/sfinae4.C: Adjust expected result after P1009.
> * g++.dg/cpp2a/new-array1.C: New test.
> * g++.dg/cpp2a/new-array2.C: New test.
> * g++.dg/cpp2a/new-array3.C: New test.
> * g++.dg/cpp2a/new-array4.C: New test.
>
> Co-authored-by: Jason Merrill <jason@redhat.com>
> ---
> gcc/cp/call.c | 4 +-
> gcc/cp/init.c | 55 +++++++++++++++++--
> gcc/cp/parser.c | 13 +++--
> gcc/cp/pt.c | 4 ++
> gcc/testsuite/g++.dg/cpp0x/sfinae4.C | 8 ++-
> gcc/testsuite/g++.dg/cpp2a/new-array1.C | 70 +++++++++++++++++++++++++
> gcc/testsuite/g++.dg/cpp2a/new-array2.C | 22 ++++++++
> gcc/testsuite/g++.dg/cpp2a/new-array3.C | 17 ++++++
> gcc/testsuite/g++.dg/cpp2a/new-array4.C | 10 ++++
> gcc/tree.c | 17 ++++++
> gcc/tree.h | 1 +
> 11 files changed, 211 insertions(+), 10 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/new-array1.C
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/new-array2.C
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/new-array3.C
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/new-array4.C
>
> diff --git a/gcc/cp/call.c b/gcc/cp/call.c
> index c62d9e227d3..5ebb6ba81ab 100644
> --- a/gcc/cp/call.c
> +++ b/gcc/cp/call.c
> @@ -10294,8 +10294,8 @@ build_new_method_call_1 (tree instance, tree fns, vec<tree, va_gc> **args,
> && !vec_safe_is_empty (user_args))
> {
> /* Create a CONSTRUCTOR from ARGS, e.g. {1, 2} from <1, 2>. */
> - tree list = build_tree_list_vec (user_args);
> - tree ctor = build_constructor_from_list (init_list_type_node, list);
> + tree ctor = build_constructor_from_vec (init_list_type_node,
> + user_args);
> CONSTRUCTOR_IS_DIRECT_INIT (ctor) = true;
> CONSTRUCTOR_IS_PAREN_INIT (ctor) = true;
> if (is_dummy_object (instance))
> diff --git a/gcc/cp/init.c b/gcc/cp/init.c
> index 872c23453fd..360ab8c0b52 100644
> --- a/gcc/cp/init.c
> +++ b/gcc/cp/init.c
> @@ -3559,8 +3559,8 @@ build_new_1 (vec<tree, va_gc> **placement, tree type, tree nelts,
> else if (array_p)
> {
> tree vecinit = NULL_TREE;
> - if (vec_safe_length (*init) == 1
> - && DIRECT_LIST_INIT_P ((**init)[0]))
> + const size_t len = vec_safe_length (*init);
> + if (len == 1 && DIRECT_LIST_INIT_P ((**init)[0]))
> {
> vecinit = (**init)[0];
> if (CONSTRUCTOR_NELTS (vecinit) == 0)
> @@ -3578,6 +3578,15 @@ build_new_1 (vec<tree, va_gc> **placement, tree type, tree nelts,
> vecinit = digest_init (arraytype, vecinit, complain);
> }
> }
> + /* This handles code like new char[]{"foo"}. */
> + else if (len == 1
> + && char_type_p (TYPE_MAIN_VARIANT (type))
> + && TREE_CODE (tree_strip_any_location_wrapper ((**init)[0]))
> + == STRING_CST)
> + {
> + vecinit = (**init)[0];
> + STRIP_ANY_LOCATION_WRAPPER (vecinit);
> + }
> else if (*init)
> {
> if (complain & tf_error)
> @@ -3634,8 +3643,7 @@ build_new_1 (vec<tree, va_gc> **placement, tree type, tree nelts,
> && AGGREGATE_TYPE_P (type)
> && (*init)->length () > 1)
> {
> - ie = build_tree_list_vec (*init);
> - ie = build_constructor_from_list (init_list_type_node, ie);
> + ie = build_constructor_from_vec (init_list_type_node, *init);
> CONSTRUCTOR_IS_DIRECT_INIT (ie) = true;
> CONSTRUCTOR_IS_PAREN_INIT (ie) = true;
> ie = digest_init (type, ie, complain);
> @@ -3917,6 +3925,45 @@ build_new (location_t loc, vec<tree, va_gc> **placement, tree type,
> return error_mark_node;
> }
>
> + /* P1009: Array size deduction in new-expressions. */
> + if (TREE_CODE (type) == ARRAY_TYPE
> + && !TYPE_DOMAIN (type)
> + && *init)
> + {
> + /* This means we have 'new T[]()'. */
> + if ((*init)->is_empty ())
> + {
> + tree ctor = build_constructor (init_list_type_node, NULL);
> + CONSTRUCTOR_IS_DIRECT_INIT (ctor) = true;
> + vec_safe_push (*init, ctor);
> + }
> + tree &elt = (**init)[0];
> + /* The C++20 'new T[](e_0, ..., e_k)' case allowed by P0960. */
> + if (!DIRECT_LIST_INIT_P (elt) && cxx_dialect >= cxx20)
> + {
> + /* Handle new char[]("foo"). */
> + if (vec_safe_length (*init) == 1
> + && char_type_p (TYPE_MAIN_VARIANT (TREE_TYPE (type)))
> + && TREE_CODE (tree_strip_any_location_wrapper (elt))
> + == STRING_CST)
> + /* Leave it alone: the string should not be wrapped in {}. */;
> + else
> + {
> + tree ctor = build_constructor_from_vec (init_list_type_node, *init);
> + CONSTRUCTOR_IS_DIRECT_INIT (ctor) = true;
> + CONSTRUCTOR_IS_PAREN_INIT (ctor) = true;
> + elt = ctor;
> + /* We've squashed all the vector elements into the first one;
> + truncate the rest. */
> + (*init)->truncate (1);
> + }
> + }
> + /* Otherwise we should have 'new T[]{e_0, ..., e_k}'. */
> + if (BRACE_ENCLOSED_INITIALIZER_P (elt))
> + elt = reshape_init (type, elt, complain);
> + cp_complete_array_type (&type, elt, /*do_default*/false);
> + }
> +
> /* The type allocated must be complete. If the new-type-id was
> "T[N]" then we are just checking that "T" is complete here, but
> that is equivalent, since the value of "N" doesn't matter. */
> diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
> index 7cc2dbed5fe..edc34a808fd 100644
> --- a/gcc/cp/parser.c
> +++ b/gcc/cp/parser.c
> @@ -9011,7 +9011,9 @@ cp_parser_new_type_id (cp_parser* parser, tree *nelts)
> if (*nelts == error_mark_node)
> *nelts = integer_one_node;
>
> - if (outer_declarator)
> + if (*nelts == NULL_TREE)
> + /* Leave [] in the declarator. */;
> + else if (outer_declarator)
> outer_declarator->declarator = declarator->declarator;
> else
> new_declarator = NULL;
> @@ -9072,6 +9074,7 @@ static cp_declarator *
> cp_parser_direct_new_declarator (cp_parser* parser)
> {
> cp_declarator *declarator = NULL;
> + bool first_p = true;
>
> while (true)
> {
> @@ -9082,14 +9085,17 @@ cp_parser_direct_new_declarator (cp_parser* parser)
> cp_parser_require (parser, CPP_OPEN_SQUARE, RT_OPEN_SQUARE);
>
> token = cp_lexer_peek_token (parser->lexer);
> - expression = cp_parser_expression (parser);
> + if (token->type == CPP_CLOSE_SQUARE && first_p)
> + expression = NULL_TREE;
> + else
> + expression = cp_parser_expression (parser);
> /* The standard requires that the expression have integral
> type. DR 74 adds enumeration types. We believe that the
> real intent is that these expressions be handled like the
> expression in a `switch' condition, which also allows
> classes with a single conversion to integral or
> enumeration type. */
> - if (!processing_template_decl)
> + if (expression && !processing_template_decl)
> {
> expression
> = build_expr_type_conversion (WANT_INT | WANT_ENUM,
> @@ -9114,6 +9120,7 @@ cp_parser_direct_new_declarator (cp_parser* parser)
> bounds. */
> if (cp_lexer_next_token_is_not (parser->lexer, CPP_OPEN_SQUARE))
> break;
> + first_p = false;
> }
>
> return declarator;
> diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
> index 5dbdd37f6e3..fe318b84385 100644
> --- a/gcc/cp/pt.c
> +++ b/gcc/cp/pt.c
> @@ -26872,6 +26872,10 @@ type_dependent_expression_p (tree expression)
> return dependent_type_p (TREE_VALUE (TREE_PURPOSE (type)))
> || value_dependent_expression_p
> (TREE_OPERAND (TREE_VALUE (type), 1));
> + /* Array type whose dimension has to be deduced. */
> + else if (TREE_CODE (type) == ARRAY_TYPE
> + && TREE_OPERAND (expression, 2) == NULL_TREE)
> + return true;
> else
> return dependent_type_p (type);
> }
> diff --git a/gcc/testsuite/g++.dg/cpp0x/sfinae4.C b/gcc/testsuite/g++.dg/cpp0x/sfinae4.C
> index 1b24966e051..f271cf1df6a 100644
> --- a/gcc/testsuite/g++.dg/cpp0x/sfinae4.C
> +++ b/gcc/testsuite/g++.dg/cpp0x/sfinae4.C
> @@ -19,5 +19,11 @@ template<typename _Tp, typename... _Args>
> static const bool value = sizeof(__test<_Tp, _Args...>(0)) == 1;
> };
>
> -static_assert( !is_constructible_mini<int[], int>::value, "");
> +// int[](...) will work with P0960 and P1009.
> +#if __cpp_aggregate_paren_init
> +constexpr bool r = true;
> +#else
> +constexpr bool r = false;
> +#endif
> +static_assert( is_constructible_mini<int[], int>::value == r, "");
> static_assert( !is_constructible_mini<void, int>::value, "");
> diff --git a/gcc/testsuite/g++.dg/cpp2a/new-array1.C b/gcc/testsuite/g++.dg/cpp2a/new-array1.C
> new file mode 100644
> index 00000000000..2353c384359
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/new-array1.C
> @@ -0,0 +1,70 @@
> +// PR c++/93529
> +// P1009: Array size deduction in new-expressions
> +// { dg-do run { target c++11 } }
> +
> +// When the array bound is deduced to 0, malloc(0) returns
> +// a non-dereferenceable pointer.
> +int *p0 = new int[]{};
> +int *p1 = new int[]{ 1 };
> +int *p2 = new int[]{ 1, 2, 3 };
> +char *c1 = new char[]{"foo"};
> +#if __cpp_aggregate_paren_init
> +int *q0 = new int[]();
> +int *q1 = new int[](1);
> +int *q2 = new int[](1, 2, 3);
> +char *d1 = new char[]("foo");
> +char *d2 = new char[4]("foo");
> +char *d3 = new char[]((("foo")));
> +#endif
> +
> +struct Aggr { int a; int b; int c; };
> +Aggr *a1 = new Aggr[]{};
> +Aggr *a2 = new Aggr[]{ 1, 2, 3 };
> +Aggr *a3 = new Aggr[]{ 1, 2, 3, 4 };
> +Aggr *a4 = new Aggr[]{ { 1, 2, 3 } };
> +Aggr *a5 = new Aggr[]{ { 1 }, { 6, 7 } };
> +#if __cpp_designated_initializers
> +Aggr *a9 = new Aggr[]{ { .a = 1, .b = 2, .c = 3 } };
> +#endif
> +#if __cpp_aggregate_paren_init
> +Aggr *a6 = new Aggr[]();
> +Aggr *a7 = new Aggr[]({ 1, 2, 3 });
> +Aggr *a8 = new Aggr[]({ 1 }, { 6, 7 });
> +#endif
> +
> +int
> +main ()
> +{
> + if (p1[0] != 1 || p2[0] != 1 || p2[1] != 2 || p2[2] != 3)
> + __builtin_abort ();
> + if (__builtin_strcmp (c1, "foo"))
> + __builtin_abort ();
> + if (a2->a != 1 || a2->b != 2 || a2->c != 3)
> + __builtin_abort ();
> + if (a3[0].a != 1 || a3[0].b != 2 || a3[0].c != 3
> + || a3[1].a != 4 || a3[1].b != 0 || a3[1].c != 0)
> + __builtin_abort ();
> + if (a4->a != 1 || a4->b != 2 || a4->c != 3)
> + __builtin_abort ();
> + if (a5[0].a != 1 || a5[0].b != 0 || a5[0].c != 0
> + || a5[1].a != 6 || a5[1].b != 7 || a5[1].c != 0)
> + __builtin_abort ();
> +#if __cpp_designated_initializers
> + if (a9->a != 1 || a9->b != 2 || a9->c != 3)
> + __builtin_abort ();
> +#endif
> +#if __cpp_aggregate_paren_init
> + if (q1[0] != 1)
> + __builtin_abort ();
> + if (q2[0] != 1 || q2[1] != 2 || q2[2] != 3)
> + __builtin_abort ();
> + if (__builtin_strcmp (d1, "foo") || __builtin_strcmp (d2, "foo")
> + || __builtin_strcmp (d3, "foo"))
> + __builtin_abort ();
> + if (a7[0].a != 1 || a7[0].b != 2 || a7[0].c != 3)
> + __builtin_abort ();
> + if (a8[0].a != 1 || a8[0].b != 0 || a8[0].c != 0
> + || a8[1].a != 6 || a8[1].b != 7 || a8[1].c != 0)
> + __builtin_abort ();
> +#endif
> +}
> diff --git a/gcc/testsuite/g++.dg/cpp2a/new-array2.C b/gcc/testsuite/g++.dg/cpp2a/new-array2.C
> new file mode 100644
> index 00000000000..9fa3f9ea7a3
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/new-array2.C
> @@ -0,0 +1,22 @@
> +// PR c++/93529
> +// P1009: Array size deduction in new-expressions
> +// { dg-do compile { target c++11 } }
> +
> +// Test error cases.
> +int *p = new double[] = { 1, 2, 3}; // { dg-error "invalid use of array with unspecified bounds" }
> +int *p2 = new double[] = (1, 2, 3); // { dg-error "invalid use of array with unspecified bounds" }
> +struct Aggr { int a; int b; int c; };
> +Aggr *p3 = new Aggr[]( 1, 2, 3 ); // { dg-error "could not convert|parenthesized initializer" }
> +char *p4 = new char[]("foo", "a"); // { dg-error "invalid conversion|parenthesized initializer" }
> +
> +template<typename... T>
> +int *fn(T... t)
> +{
> + return new int[]{t...}; // { dg-error "invalid conversion" }
> +}
> +
> +void
> +g ()
> +{
> + int *p = fn ("a");
> +}
> diff --git a/gcc/testsuite/g++.dg/cpp2a/new-array3.C b/gcc/testsuite/g++.dg/cpp2a/new-array3.C
> new file mode 100644
> index 00000000000..f35124e159d
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/new-array3.C
> @@ -0,0 +1,17 @@
> +// PR c++/93529
> +// P1009: Array size deduction in new-expressions
> +// { dg-do compile { target c++11 } }
> +
> +template<typename... T>
> +int *fn(T... t)
> +{
> + return new int[]{t...};
> +}
> +
> +int
> +main ()
> +{
> + int *p0 = fn ();
> + int *p1 = fn (1);
> + int *p3 = fn (1, 2, 3);
> +}
> diff --git a/gcc/testsuite/g++.dg/cpp2a/new-array4.C b/gcc/testsuite/g++.dg/cpp2a/new-array4.C
> new file mode 100644
> index 00000000000..2c327ebc853
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/new-array4.C
> @@ -0,0 +1,10 @@
> +// PR c++/93529
> +// P1009: Array size deduction in new-expressions
> +// { dg-do compile { target c++11 } }
> +
> +void
> +fn ()
> +{
> + new int[][3]{ { 1, 2, 3 } };
> + new int[][]{ { 1, 2, 3 } }; // { dg-error "expected primary-expression" }
> +}
> diff --git a/gcc/tree.c b/gcc/tree.c
> index d0202c3f785..1ff13f218ce 100644
> --- a/gcc/tree.c
> +++ b/gcc/tree.c
> @@ -2123,6 +2123,23 @@ build_constructor_from_list (tree type, tree vals)
> return build_constructor (type, v);
> }
>
> +/* Return a new CONSTRUCTOR node whose type is TYPE and whose values
> + are in a vector pointed to by VALS. Note that the TREE_PURPOSE
> + fields in the constructor remain null. */
> +
> +tree
> +build_constructor_from_vec (tree type, const vec<tree, va_gc> *vals)
> +{
> + tree ctor = build_constructor (type, NULL);
> +
> + unsigned int ix;
> + tree t;
> + FOR_EACH_VEC_SAFE_ELT (vals, ix, t)
> + CONSTRUCTOR_APPEND_ELT (CONSTRUCTOR_ELTS (ctor), NULL_TREE, t);
Either you need to build up the constructor_elt vec first and then pass
it to build_constructor, or call recompute_constructor_flags after you
add all the elements.
And perhaps
for (tree t : *vals)
Jason
> + return ctor;
> +}
> +
> /* Return a new CONSTRUCTOR node whose type is TYPE. NELTS is the number
> of elements, provided as index/value pairs. */
>
> diff --git a/gcc/tree.h b/gcc/tree.h
> index 22dd4ac0f3c..4fd51aca723 100644
> --- a/gcc/tree.h
> +++ b/gcc/tree.h
> @@ -4407,6 +4407,7 @@ extern void verify_constructor_flags (tree);
> extern tree build_constructor (tree, vec<constructor_elt, va_gc> * CXX_MEM_STAT_INFO);
> extern tree build_constructor_single (tree, tree, tree);
> extern tree build_constructor_from_list (tree, tree);
> +extern tree build_constructor_from_vec (tree, const vec<tree, va_gc> *);
> extern tree build_constructor_va (tree, int, ...);
> extern tree build_clobber (tree);
> extern tree build_real_from_int_cst (tree, const_tree);
>
> base-commit: 0d166f4a8773a43d925be006e713b7d81626ddb9
>
More information about the Gcc-patches
mailing list