Annex C (informative) Compatibility [diff]

Subclause [diff.cpp23] lists the differences between C++ and ISO C++ 2023 (ISO/IEC 14882:2023, Programming Languages — C++), by the chapters of this document.

Change: Operations mixing a value of an enumeration type and a value of a different enumeration type or of a floating-point type are no longer valid.

Rationale: Reinforcing type safety.

Effect on original feature: A valid C++ 2023 program that performs operations mixing a value of an enumeration type and a value of a different enumeration type or of a floating-point type is ill-formed.

For example: enum E1 { e }; enum E2 { f }; bool b = e <= 3.7; // ill-formed; previously well-formed int k = f - e; // ill-formed; previously well-formed auto x = true ? e : f; // ill-formed; previously well-formed

Change: Pointer comparisons between initializer_list objects' backing arrays are unspecified.

Rationale: Permit the implementation to store backing arrays in static read-only memory.

Effect on original feature: Valid C++ 2023 code that relies on the result of pointer comparison between backing arrays may change behavior.

For example: bool ne(std::initializer_list<int> a, std::initializer_list<int> b) { return a.begin() != b.begin() + 1; } bool b = ne({2,3}, {1,2,3}); // unspecified result; previously false

Change: Previously, T...[n] would declare a pack of function parameters.

T...[n] is now a pack-index-specifier.

Rationale: Improve the handling of packs.

Effect on original feature: Valid C++ 2023 code that declares a pack of parameters without specifying a declarator-id becomes ill-formed.

For example: template <typename... T> void f(T... [1]); template <typename... T> void g(T... ptr[1]); int main() { f<int, double>(nullptr, nullptr); // ill-formed, previously void f<int, double>(int [1], double [1]) g<int, double>(nullptr, nullptr); // ok }

Change: New headers.

Rationale: New functionality.

Effect on original feature: The following C++ headers are new: <debugging>, <hazard_pointer>, <linalg>, <rcu>, and <text_encoding>.

Valid C++ 2023 code that #includes headers with these names may be invalid in this revision of C++.

Change: Output of floating-point overloads of to_string and to_wstring.

Rationale: Prevent loss of information and improve consistency with other formatting facilities.

Effect on original feature: to_string and to_wstring function calls that take floating-point arguments may produce a different output.

For example: auto s = std::to_string(1e-7); // "1e-07" // previously "0.000000" with '.' possibly // changed according to the global C locale

Change: span<const T> is constructible from initializer_list<T>.

Rationale: Permit passing a braced initializer list to a function taking span.

Effect on original feature: Valid C++ 2023 code that relies on the lack of this constructor may refuse to compile, or change behavior in this revision of C++.

For example: void one(pair<int, int>); // #1 void one(span<const int>); // #2 void t1() { one({1, 2}); } // ambiguous between #1 and #2; previously called #1 void two(span<const int, 2>); void t2() { two({{1, 2}}); } // ill-formed; previously well-formed void *a[10]; int x = span<void* const>{a, 0}.size(); // x is 2; previously 0 any b[10]; int y = span<const any>{b, b + 10}.size(); // y is 2; previously 10

Change: Remove the type alias allocator<T>​::​is_always_equal.

Rationale: Non-empty allocator classes derived from allocator needed to explicitly define an is_always_equal member type so that allocator_traits would not use the one from the allocator base class.

Effect on original feature: It is simpler to correctly define an allocator class with an allocator base class.

For example: template <class T> struct MyAlloc : allocator<T> { int tag; }; static_assert(!allocator_traits<MyAlloc<int>>::is_always_equal); // Error in C++ 2023, // OK in C++ 2026

Change: Remove the basic_string​::​reserve() overload with no parameters.

Rationale: The overload of reserve with no parameters is redundant.

The shrink_to_fit member function can be used instead.

Effect on original feature: A valid C++ 2023 program that calls reserve() on a basic_string object may fail to compile.

The old functionality can be achieved by calling shrink_to_fit() instead, or the function call can be safely eliminated with no side effects.

Change: Remove header <codecvt> and all its contents.

Rationale: The header has been deprecated for the previous three editions of this standard and no longer implements the current Unicode standard, supporting only the obsolete UCS-2 encoding.

Ongoing support is at implementer's discretion, exercising freedoms granted by [zombie.names].

Effect on original feature: A valid C++ 2023 program #include-ing the header or importing the header unit may fail to compile.

Code that uses any of the following names by importing the standard library modules may fail to compile:

• (3.1)
  codecvt_mode,
• (3.2)
  codecvt_utf16,
• (3.3)
  codecvt_utf8,
• (3.4)
  codecvt_utf8_utf16,
• (3.5)
  consume_header,
• (3.6)
  generate_header, and
• (3.7)
  little_endian.

Subclause [diff.cpp20] lists the differences between C++ and ISO C++ 2020 (ISO/IEC 14882:2020, Programming Languages — C++), by the chapters of this document.

Change: Previously valid identifiers containing characters not present in UAX #44 properties XID_Start or XID_Continue, or not in Normalization Form C, are now rejected.

Rationale: Prevent confusing characters in identifiers.

Requiring normalization of names ensures consistent linker behavior.

Effect on original feature: Some identifiers are no longer well-formed.

Change: Concatenated string-literals can no longer have conflicting encoding-prefixes.

Rationale: Removal of unimplemented conditionally-supported feature.

Effect on original feature: Concatenation of string-literals with different encoding-prefixes is now ill-formed.

For example: auto c = L"a" U"b"; // was conditionally-supported; now ill-formed

Change: Change move-eligible id-expressions from lvalues to xvalues.

Rationale: Simplify the rules for implicit move.

Effect on original feature: Valid C++ 2020 code that relies on a returned id-expression's being an lvalue may change behavior or fail to compile.

For example: decltype(auto) f(int&& x) { return (x); } // returns int&&; previously returned int& int& g(int&& x) { return x; } // ill-formed; previously well-formed

Change: Change the meaning of comma in subscript expressions.

Rationale: Enable repurposing a deprecated syntax to support multidimensional indexing.

Effect on original feature: Valid C++ 2020 code that uses a comma expression within a subscript expression may fail to compile.

For example: arr[1, 2] // was equivalent to arr[(1, 2)], // now equivalent to arr.operator[](1, 2) or ill-formed

Change: The lifetime of temporary objects in the for-range-initializer is extended until the end of the loop ([class.temporary]).

Rationale: Improve usability of the range-based for statement.

Effect on original feature: Destructors of some temporary objects are invoked later.

For example: void f() { std::vector<int> v = { 42, 17, 13 }; std::mutex m; for (int x : static_cast<void>(std::lock_guard<std::mutex>(m)), v) { // lock released in C++ 2020 std::lock_guard<std::mutex> guard(m); // OK in C++ 2020, now deadlocks } }

Change: UTF-8 string literals may initialize arrays of char or unsigned char.

Rationale: Compatibility with previously written code that conformed to previous versions of this document.

Effect on original feature: Arrays of char or unsigned char may now be initialized with a UTF-8 string literal.

This can affect initialization that includes arrays that are directly initialized within class types, typically aggregates.

For example: struct A { char8_t s[10]; }; struct B { char s[10]; }; void f(A); void f(B); int main() { f({u8""}); // ambiguous }

Change: Deducing template arguments from exception specifications.

Rationale: Facilitate generic handling of throwing and non-throwing functions.

Effect on original feature: Valid ISO C++ 2020 code may be ill-formed in this revision of C++.

For example: template<bool> struct A { }; template<bool B> void f(void (*)(A<B>) noexcept(B)); void g(A<false>) noexcept; void h() { f(g); // ill-formed; previously well-formed }

Change: New headers.

Rationale: New functionality.

Effect on original feature: The following C++ headers are new: <expected>, <flat_map>, <flat_set>, <generator>, <mdspan>, <print>, <spanstream>, <stacktrace>, <stdatomic.h>, and <stdfloat>.

Valid C++ 2020 code that #includes headers with these names may be invalid in this revision of C++.

Change: Replace common_reference_with in three_way_comparable_with, equality_comparable_with, and totally_ordered_with with an exposition-only concept.

Rationale: Allow uncopyable, but movable, types to model these concepts.

Effect on original feature: Valid C++ 2020 code relying on subsumption with common_reference_with may fail to compile in this revision of C++.

For example: template<class T, class U> requires equality_comparable_with<T, U> bool attempted_equals(const T&, const U& u); // previously selected overload template<class T, class U> requires common_reference_with<const remove_reference_t<T>&, const remove_reference_t<U>&> bool attempted_equals(const T& t, const U& u); // ambiguous overload; previously // rejected by partial ordering bool test(shared_ptr<int> p) { return attempted_equals(p, nullptr); // ill-formed; previously well-formed }

Change: Forbid partial and explicit program-defined specializations of allocator_traits.

Rationale: Allow addition of allocate_at_least to allocator_traits, and potentially other members in the future.

Effect on original feature: Valid C++ 2020 code that partially or explicitly specializes allocator_traits is ill-formed with no diagnostic required in this revision of C++.

Change: Signature changes: format, format_to, vformat_to, format_to_n, formatted_size.

Removal of format_args_t.

Rationale: Improve safety via compile-time format string checks, avoid unnecessary template instantiations.

Effect on original feature: Valid C++ 2020 code that contained errors in format strings or relied on previous format string signatures or format_args_t may become ill-formed.

For example: auto s = std::format("{:d}", "I am not a number"); // ill-formed, // previously threw format_error

Change: Signature changes: format, format_to, format_to_n, formatted_size.

Rationale: Enable formatting of views that do not support iteration when const-qualified and that are not copyable.

Effect on original feature: Valid C++ 2020 code that passes bit-fields to formatting functions may become ill-formed.

For example: struct tiny { int bit: 1; }; auto t = tiny(); std::format("{}", t.bit); // ill-formed, previously returned "0"

Change: Restrict types of formatting arguments used as width or precision in a std-format-spec.

Rationale: Disallow types that do not have useful or portable semantics as a formatting width or precision.

Effect on original feature: Valid C++ 2020 code that passes a boolean or character type as arg-id becomes invalid.

For example: std::format("{:*^{}}", "", true); // ill-formed, previously returned "*" std::format("{:*^{}}", "", '1'); // ill-formed, previously returned an // implementation-defined number of '*' characters

Change: Removed the formatter specialization: template<size_t N> struct formatter<const charT[N], charT>;

Rationale: The specialization is inconsistent with the design of formatter, which is intended to be instantiated only with cv-unqualified object types.

Effect on original feature: Valid C++ 2020 code that instantiated the removed specialization can become ill-formed.

Change: Additional rvalue overload for the substr member function and the corresponding constructor.

Rationale: Improve efficiency of operations on rvalues.

Effect on original feature: Valid C++ 2020 code that created a substring by calling substr (or the corresponding constructor) on an xvalue expression with type S that is a specialization of basic_string may change meaning in this revision of C++.

For example: std::string s1 = "some long string that forces allocation", s2 = s1; std::move(s1).substr(10, 5); assert(s1 == s2); // unspecified, previously guaranteed to be true std::string s3(std::move(s2), 10, 5); assert(s1 == s2); // unspecified, previously guaranteed to be true

Change: Heterogeneous extract and erase overloads for associative containers.

Rationale: Improve efficiency of erasing elements from associative containers.

Effect on original feature: Valid C++ 2020 code may fail to compile in this revision of C++.

For example: struct B { auto operator<=>(const B&) const = default; }; struct D : private B { void f(std::set<B, std::less<>>& s) { s.erase(*this); // ill-formed; previously well-formed } };

Change: In this revision of C++, it is implementation-defined whether a barrier's phase completion step runs if no thread calls wait.

Previously the phase completion step was guaranteed to run on the last thread that calls arrive or arrive_and_drop during the phase.

In this revision of C++, it can run on any of the threads that arrived or waited at the barrier during the phase.

Rationale: Correct contradictory wording and improve implementation flexibility for performance.

Effect on original feature: Valid C++ 2020 code using a barrier might have different semantics in this revision of C++ if it depends on a completion function's side effects occurring exactly once, on a specific thread running the phase completion step, or on a completion function's side effects occurring without wait having been called.

For example: auto b0 = std::barrier(1); b0.arrive(); b0.arrive(); // implementation-defined; previously well-defined int data = 0; auto b1 = std::barrier(1, [&] { data++; }); b1.arrive(); assert(data == 1); // implementation-defined; previously well-defined b1.arrive(); // implementation-defined; previously well-defined

Subclause [diff.cpp17] lists the differences between C++ and ISO C++ 2017 (ISO/IEC 14882:2017, Programming Languages — C++), by the chapters of this document.

Change: New identifiers with special meaning.

Rationale: Required for new features.

Effect on original feature: Logical lines beginning with module or import may be interpreted differently in this revision of C++.

For example: class module {}; module m1; // was variable declaration; now module-declaration module *m2; // variable declaration class import {}; import j1; // was variable declaration; now module-import-declaration ::import j2; // variable declaration

Change: header-name tokens are formed in more contexts.

Rationale: Required for new features.

Effect on original feature: When the identifier import is followed by a < character, a header-name token may be formed.

For example: template<typename> class import {}; import<int> f(); // ill-formed; previously well-formed ::import<int> g(); // OK

Change: New keywords.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2017 code using char8_t, concept, consteval, constinit, co_await, co_yield, co_return, or requires as an identifier is not valid in this revision of C++.

Change: New operator <=>.

Rationale: Necessary for new functionality.

Effect on original feature: Valid C++ 2017 code that contains a <= token immediately followed by a > token may be ill-formed or have different semantics in this revision of C++.

For example: namespace N { struct X {}; bool operator<=(X, X); template<bool(X, X)> struct Y {}; Y<operator<=> y; // ill-formed; previously well-formed }

Change: Type of UTF-8 string and character literals.

Rationale: Required for new features.

The changed types enable function overloading, template specialization, and type deduction to distinguish ordinary and UTF-8 string and character literals.

Effect on original feature: Valid C++ 2017 code that depends on UTF-8 string literals having type “array of const char” and UTF-8 character literals having type “char” is not valid in this revision of C++.

For example: const auto *u8s = u8"text"; // u8s previously deduced as const char*; now deduced as const char8_t* const char *ps = u8s; // ill-formed; previously well-formed auto u8c = u8'c'; // u8c previously deduced as char; now deduced as char8_t char *pc = &u8c; // ill-formed; previously well-formed std::string s = u8"text"; // ill-formed; previously well-formed void f(const char *s); f(u8"text"); // ill-formed; previously well-formed template<typename> struct ct; template<> struct ct<char> { using type = char; }; ct<decltype(u8'c')>::type x; // ill-formed; previously well-formed.

Change: A pseudo-destructor call ends the lifetime of the object to which it is applied.

Rationale: Increase consistency of the language model.

Effect on original feature: Valid ISO C++ 2017 code may be ill-formed or have undefined behavior in this revision of C++.

For example: int f() { int a = 123; using T = int; a.~T(); return a; // undefined behavior; previously returned 123 }

Change: Except for the initial release operation, a release sequence consists solely of atomic read-modify-write operations.

Rationale: Removal of rarely used and confusing feature.

Effect on original feature: If a memory_order_release atomic store is followed by a memory_order_relaxed store to the same variable by the same thread, then reading the latter value with a memory_order_acquire load no longer provides any “happens before” guarantees, even in the absence of intervening stores by another thread.

Change: Implicit lambda capture may capture additional entities.

Rationale: Rule simplification, necessary to resolve interactions with constexpr if.

Effect on original feature: Lambdas with a capture-default may capture local entities that were not captured in C++ 2017 if those entities are only referenced in contexts that do not result in an odr-use.

Change: Unnamed classes with a typedef name for linkage purposes can contain only C-compatible constructs.

Rationale: Necessary for implementability.

Effect on original feature: Valid C++ 2017 code may be ill-formed in this revision of C++.

For example: typedef struct { void f() {} // ill-formed; previously well-formed } S;

Change: A function cannot have different default arguments in different translation units.

Rationale: Required for modules support.

Effect on original feature: Valid C++ 2017 code may be ill-formed in this revision of C++, with no diagnostic required.

For example: // Translation unit 1 int f(int a = 42); int g() { return f(); } // Translation unit 2 int f(int a = 76) { return a; } // ill-formed, no diagnostic required; previously well-formed int g(); int main() { return g(); } // used to return 42

Change: A class that has user-declared constructors is never an aggregate.

Rationale: Remove potentially error-prone aggregate initialization which may apply notwithstanding the declared constructors of a class.

Effect on original feature: Valid C++ 2017 code that aggregate-initializes a type with a user-declared constructor may be ill-formed or have different semantics in this revision of C++.

For example: struct A { // not an aggregate; previously an aggregate A() = delete; }; struct B { // not an aggregate; previously an aggregate B() = default; int i = 0; }; struct C { // not an aggregate; previously an aggregate C(C&&) = default; int a, b; }; A a{}; // ill-formed; previously well-formed B b = {1}; // ill-formed; previously well-formed auto* c = new C{2, 3}; // ill-formed; previously well-formed struct Y; struct X { operator Y(); }; struct Y { // not an aggregate; previously an aggregate Y(const Y&) = default; X x; }; Y y{X{}}; // copy constructor call; previously aggregate-initialization

Change: Boolean conversion from a pointer or pointer-to-member type is now a narrowing conversion.

Rationale: Catches bugs.

Effect on original feature: Valid C++ 2017 code may fail to compile in this revision of C++.

For example: bool y[] = { "bc" }; // ill-formed; previously well-formed

Change: The class name can no longer be used parenthesized immediately after an explicit decl-specifier in a constructor declaration.

The conversion-function-id can no longer be used parenthesized immediately after an explicit decl-specifier in a conversion function declaration.

Rationale: Necessary for new functionality.

Effect on original feature: Valid C++ 2017 code may fail to compile in this revision of C++.

For example: struct S { explicit (S)(const S&); // ill-formed; previously well-formed explicit (operator int)(); // ill-formed; previously well-formed explicit(true) (S)(int); // OK };

Change: A simple-template-id is no longer valid as the declarator-id of a constructor or destructor.

Rationale: Remove potentially error-prone option for redundancy.

Effect on original feature: Valid C++ 2017 code may fail to compile in this revision of C++.

For example: template<class T> struct A { A<T>(); // error: simple-template-id not allowed for constructor A(int); // OK, injected-class-name used ~A<T>(); // error: simple-template-id not allowed for destructor };

Change: A function returning an implicitly movable entity may invoke a constructor taking an rvalue reference to a type different from that of the returned expression.

Function and catch-clause parameters can be thrown using move constructors.

Rationale: Side effect of making it easier to write more efficient code that takes advantage of moves.

Effect on original feature: Valid C++ 2017 code may fail to compile or have different semantics in this revision of C++.

For example: struct base { base(); base(base const &); private: base(base &&); }; struct derived : base {}; base f(base b) { throw b; // error: base(base &&) is private derived d; return d; // error: base(base &&) is private } struct S { S(const char *s) : m(s) { } S(const S&) = default; S(S&& other) : m(other.m) { other.m = nullptr; } const char * m; }; S consume(S&& s) { return s; } void g() { S s("text"); consume(static_cast<S&&>(s)); char c = *s.m; // undefined behavior; previously ok }

Change: Equality and inequality expressions can now find reversed and rewritten candidates.

Rationale: Improve consistency of equality with three-way comparison and make it easier to write the full complement of equality operations.

Effect on original feature: For certain pairs of types where one is convertible to the other, equality or inequality expressions between an object of one type and an object of the other type invoke a different operator.

Also, for certain types, equality or inequality expressions between two objects of that type become ambiguous.

For example: struct A { operator int() const; }; bool operator==(A, int); // #1 // #2 is built-in candidate: bool operator==(int, int); // #3 is built-in candidate: bool operator!=(int, int); int check(A x, A y) { return (x == y) + // ill-formed; previously well-formed (10 == x) + // calls #1, previously selected #2 (10 != x); // calls #1, previously selected #3 }

Change: Overload resolution may change for equality operators ([expr.eq]).

Rationale: Support calling operator== with reversed order of arguments.

Effect on original feature: Valid C++ 2017 code that uses equality operators with conversion functions may be ill-formed or have different semantics in this revision of C++.

For example: struct A { operator int() const { return 10; } }; bool operator==(A, int); // #1 // #2 is built-in candidate: bool operator==(int, int); bool b = 10 == A(); // calls #1 with reversed order of arguments; previously selected #2 struct B { bool operator==(const B&); // member function with no cv-qualifier }; B b1; bool eq = (b1 == b1); // ambiguous; previously well-formed

Change: An unqualified-id that is followed by a < and for which name lookup finds nothing or finds a function will be treated as a template-name in order to potentially cause argument dependent lookup to be performed.

Rationale: It was problematic to call a function template with an explicit template argument list via argument dependent lookup because of the need to have a template with the same name visible via normal lookup.

Effect on original feature: Previously valid code that uses a function name as the left operand of a < operator would become ill-formed.

For example: struct A {}; bool operator<(void (*fp)(), A); void f() {} int main() { A a; f < a; // ill-formed; previously well-formed (f) < a; // still well-formed }

Change: Remove throw() exception specification.

Rationale: Removal of obsolete feature that has been replaced by noexcept.

Effect on original feature: A valid C++ 2017 function declaration, member function declaration, function pointer declaration, or function reference declaration that uses throw() for its exception specification will be rejected as ill-formed in this revision of C++.

It should simply be replaced with noexcept for no change of meaning since C++ 2017.

Change: New headers.

Rationale: New functionality.

Effect on original feature: The following C++ headers are new: <barrier>, <bit>, <charconv>, <compare>, <concepts>, <coroutine>, <format>, <latch>, <numbers>, <ranges>, <semaphore>, <source_location>, <span>, <stop_token>, <syncstream>, and <version>.

Valid C++ 2017 code that #includes headers with these names may be invalid in this revision of C++.

Change: Remove vacuous C++ header files.

Rationale: The empty headers implied a false requirement to achieve C compatibility with the C++ headers.

Effect on original feature: A valid C++ 2017 program that #includes any of the following headers may fail to compile: <ccomplex>, <ciso646>, <cstdalign>, <cstdbool>, and <ctgmath>.

To retain the same behavior:

• (2.1)
  a #include of <ccomplex> can be replaced by a #include of <complex>,
• (2.2)
  a #include of <ctgmath> can be replaced by a #include of <cmath> and a #include of <complex>, and
• (2.3)
  a #include of <ciso646>, <cstdalign>, or <cstdbool> can simply be removed.

Change: Return types of remove, remove_if, and unique changed from void to container​::​size_type.

Rationale: Improve efficiency and convenience of finding number of removed elements.

Effect on original feature: Code that depends on the return types might have different semantics in this revision of C++.

Translation units compiled against this version of C++ may be incompatible with translation units compiled against C++ 2017, either failing to link or having undefined behavior.

Change: The specialization of iterator_traits for void* and for function pointer types no longer contains any nested typedefs.

Rationale: Corrects an issue misidentifying pointer types that are not incrementable as iterator types.

Effect on original feature: A valid C++ 2017 program that relies on the presence of the typedefs may fail to compile, or have different behavior.

Change: The number and order of deducible template parameters for algorithm declarations is now unspecified, instead of being as-declared.

Rationale: Increase implementor freedom and allow some function templates to be implemented as function objects with templated call operators.

Effect on original feature: A valid C++ 2017 program that passes explicit template arguments to algorithms not explicitly specified to allow such in this version of C++ may fail to compile or have undefined behavior.

Change: Character array extraction only takes array types.

Rationale: Increase safety via preventing buffer overflow at compile time.

Effect on original feature: Valid C++ 2017 code may fail to compile in this revision of C++.

For example: auto p = new char[100]; char q[100]; std::cin >> std::setw(20) >> p; // ill-formed; previously well-formed std::cin >> std::setw(20) >> q; // OK

Change: Overload resolution for ostream inserters used with UTF-8 literals.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2017 code that passes UTF-8 literals to basic_ostream<char, ...>​::​operator<< or basic_ostream<wchar_t, ...>​::​operator<< is now ill-formed.

For example: std::cout << u8"text"; // previously called operator<<(const char*) and printed a string; // now ill-formed std::cout << u8'X'; // previously called operator<<(char) and printed a character; // now ill-formed

Change: Overload resolution for ostream inserters used with wchar_t, char16_t, or char32_t types.

Rationale: Removal of surprising behavior.

Effect on original feature: Valid C++ 2017 code that passes wchar_t, char16_t, or char32_t characters or strings to basic_ostream<char, ...>​::​operator<< or that passes char16_t or char32_t characters or strings to basic_ostream<wchar_t, ...>​::​operator<< is now ill-formed.

For example: std::cout << u"text"; // previously formatted the string as a pointer value; // now ill-formed std::cout << u'X'; // previously formatted the character as an integer value; // now ill-formed

Change: Return type of filesystem path format observer member functions.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2017 code that depends on the u8string() and generic_u8string() member functions of std​::​filesystem​::​path returning std​::​string is not valid in this revision of C++.

For example: std::filesystem::path p; std::string s1 = p.u8string(); // ill-formed; previously well-formed std::string s2 = p.generic_u8string(); // ill-formed; previously well-formed

Change: Remove uncaught_exception.

Rationale: The function did not have a clear specification when multiple exceptions were active, and has been superseded by uncaught_exceptions.

Effect on original feature: A valid C++ 2017 program that calls std​::​uncaught_exception may fail to compile.

It can be revised to use std​::​uncaught_exceptions instead, for clear and portable semantics.

Change: Remove support for adaptable function API.

Rationale: The deprecated support relied on a limited convention that could not be extended to support the general case or new language features.

It has been superseded by direct language support with decltype, and by the std​::​bind and std​::​not_fn function templates.

Effect on original feature: A valid C++ 2017 program that relies on the presence of result_type, argument_type, first_argument_type, or second_argument_type in a standard library class may fail to compile.

A valid C++ 2017 program that calls not1 or not2, or uses the class templates unary_negate or binary_negate, may fail to compile.

Change: Remove redundant members from std​::​allocator.

Rationale: std​::​allocator was overspecified, encouraging direct usage in user containers rather than relying on std​::​allocator_traits, leading to poor containers.

Effect on original feature: A valid C++ 2017 program that directly makes use of the pointer, const_pointer, reference, const_reference, rebind, address, construct, destroy, or max_size members of std​::​allocator, or that directly calls allocate with an additional hint argument, may fail to compile.

Change: Remove raw_storage_iterator.

Rationale: The iterator encouraged use of potentially-throwing algorithms, but did not return the number of elements successfully constructed, as would be necessary to destroy them.

Effect on original feature: A valid C++ 2017 program that uses this iterator class may fail to compile.

Change: Remove temporary buffers API.

Rationale: The temporary buffer facility was intended to provide an efficient optimization for small memory requests, but there is little evidence this was achieved in practice, while requiring the user to provide their own exception-safe wrappers to guard use of the facility in many cases.

Effect on original feature: A valid C++ 2017 program that calls get_temporary_buffer or return_temporary_buffer may fail to compile.

Change: Remove shared_ptr​::​unique.

Rationale: The result of a call to this member function is not reliable in the presence of multiple threads and weak pointers.

The member function use_count is similarly unreliable, but has a clearer contract in such cases, and remains available for well-defined use in single-threaded cases.

Effect on original feature: A valid C++ 2017 program that calls unique on a shared_ptr object may fail to compile.

Change: Remove deprecated type traits.

Rationale: The traits had unreliable or awkward interfaces.

The is_literal_type trait provided no way to detect which subset of constructors and member functions of a type were declared constexpr.

The result_of trait had a surprising syntax that did not directly support function types.

It has been superseded by the invoke_result trait.

Effect on original feature: A valid C++ 2017 program that relies on the is_literal_type or result_of type traits, on the is_literal_type_v variable template, or on the result_of_t alias template may fail to compile.

Subclause [diff.cpp14] lists the differences between C++ and ISO C++ 2014 (ISO/IEC 14882:2014, Programming Languages — C++), in addition to those listed above, by the chapters of this document.

Change: Removal of trigraph support as a required feature.

Rationale: Prevents accidental uses of trigraphs in non-raw string literals and comments.

Effect on original feature: Valid C++ 2014 code that uses trigraphs may not be valid or may have different semantics in this revision of C++.

Implementations may choose to translate trigraphs as specified in C++ 2014 if they appear outside of a raw string literal, as part of the implementation-defined mapping from input source file characters to the translation character set.

Change: pp-number can contain p sign and P sign.

Rationale: Necessary to enable hexadecimal-floating-point-literals.

Effect on original feature: Valid C++ 2014 code may fail to compile or produce different results in this revision of C++.

Specifically, character sequences like 0p+0 and 0e1_p+0 are three separate tokens each in C++ 2014, but one single token in this revision of C++.

For example: #define F(a) b ## a int b0p = F(0p+0); // ill-formed; equivalent to “int b0p = b0p + 0;'' in C++ 2014

Change: Remove increment operator with bool operand.

Rationale: Obsolete feature with occasionally surprising semantics.

Effect on original feature: A valid C++ 2014 expression utilizing the increment operator on a bool lvalue is ill-formed in this revision of C++.

Change: Dynamic allocation mechanism for over-aligned types.

Rationale: Simplify use of over-aligned types.

Effect on original feature: In C++ 2014 code that uses a new-expression to allocate an object with an over-aligned class type, where that class has no allocation functions of its own, ​::​operator new(std​::​size_t) is used to allocate the memory.

In this revision of C++, ​::​operator new(std​::​size_t, std​::​align_val_t) is used instead.

Change: Removal of register storage-class-specifier.

Rationale: Enable repurposing of deprecated keyword in future revisions of C++.

Effect on original feature: A valid C++ 2014 declaration utilizing the register storage-class-specifier is ill-formed in this revision of C++.

The specifier can simply be removed to retain the original meaning.

Change: auto deduction from braced-init-list.

Rationale: More intuitive deduction behavior.

Effect on original feature: Valid C++ 2014 code may fail to compile or may change meaning in this revision of C++.

For example: auto x1{1}; // was std​::​initializer_list<int>, now int auto x2{1, 2}; // was std​::​initializer_list<int>, now ill-formed

Change: Make exception specifications be part of the type system.

Rationale: Improve type-safety.

Effect on original feature: Valid C++ 2014 code may fail to compile or change meaning in this revision of C++.

For example: void g1() noexcept; void g2(); template<class T> int f(T *, T *); int x = f(g1, g2); // ill-formed; previously well-formed

Change: Definition of an aggregate is extended to apply to user-defined types with base classes.

Rationale: To increase convenience of aggregate initialization.

Effect on original feature: Valid C++ 2014 code may fail to compile or produce different results in this revision of C++; initialization from an empty initializer list will perform aggregate initialization instead of invoking a default constructor for the affected types.

For example: struct derived; struct base { friend struct derived; private: base(); }; struct derived : base {}; derived d1{}; // error; the code was well-formed in C++ 2014 derived d2; // still OK

Change: Inheriting a constructor no longer injects a constructor into the derived class.

Rationale: Better interaction with other language features.

Effect on original feature: Valid C++ 2014 code that uses inheriting constructors may not be valid or may have different semantics.

A using-declaration that names a constructor now makes the corresponding base class constructors visible to initializations of the derived class rather than declaring additional derived class constructors.

For example: struct A { template<typename T> A(T, typename T::type = 0); A(int); }; struct B : A { using A::A; B(int); }; B b(42L); // now calls B(int), used to call B<long>(long), // which called A(int) due to substitution failure // in A<long>(long).

Change: Allowance to deduce from the type of a non-type template argument.

Rationale: In combination with the ability to declare non-type template arguments with placeholder types, allows partial specializations to decompose from the type deduced for the non-type template argument.

Effect on original feature: Valid C++ 2014 code may fail to compile or produce different results in this revision of C++.

For example: template <int N> struct A; template <typename T, T N> int foo(A<N> *) = delete; void foo(void *); void bar(A<0> *p) { foo(p); // ill-formed; previously well-formed }

Change: Remove dynamic exception specifications.

Rationale: Dynamic exception specifications were a deprecated feature that was complex and brittle in use.

They interacted badly with the type system, which became a more significant issue in this revision of C++ where (non-dynamic) exception specifications are part of the function type.

Effect on original feature: A valid C++ 2014 function declaration, member function declaration, function pointer declaration, or function reference declaration, if it has a potentially throwing dynamic exception specification, is rejected as ill-formed in this revision of C++.

Violating a non-throwing dynamic exception specification calls terminate rather than unexpected, and it is unspecified whether stack unwinding is performed prior to such a call.

Change: New headers.

Rationale: New functionality.

Effect on original feature: The following C++ headers are new: <any>, <charconv>, <execution>, <filesystem>, <memory_resource>, <optional>,

<string_view>, and <variant>.

Valid C++ 2014 code that #includes headers with these names may be invalid in this revision of C++.

Change: New reserved namespaces.

Rationale: Reserve namespaces for future revisions of the standard library that might otherwise be incompatible with existing programs.

Effect on original feature: The global namespaces std followed by an arbitrary sequence of digits ([lex.name]) are reserved for future standardization.

Valid C++ 2014 code that uses such a top-level namespace, e.g., std2, may be invalid in this revision of C++.

Change: Constructors taking allocators removed.

Rationale: No implementation consensus.

Effect on original feature: Valid C++ 2014 code may fail to compile or may change meaning in this revision of C++.

Specifically, constructing a std​::​function with an allocator is ill-formed and uses-allocator construction will not pass an allocator to std​::​function constructors in this revision of C++.

Change: Different constraint on conversions from unique_ptr.

Rationale: Adding array support to shared_ptr, via the syntax shared_ptr<T[]> and shared_ptr<T[N]>.

Effect on original feature: Valid C++ 2014 code may fail to compile or may change meaning in this revision of C++.

For example: #include <memory> std::unique_ptr<int[]> arr(new int[1]); std::shared_ptr<int> ptr(std::move(arr)); // error: int(*)[] is not compatible with int*

Change: Non-const .data() member added.

Rationale: The lack of a non-const .data() differed from the similar member of std​::​vector.

This change regularizes behavior.

Effect on original feature: Overloaded functions which have differing code paths for char* and const char* arguments will execute differently when called with a non-const string's .data() member in this revision of C++.

For example: int f(char *) = delete; int f(const char *); string s; int x = f(s.data()); // ill-formed; previously well-formed

Change: Requirements change:

Rationale: Increase portability, clarification of associative container requirements.

Effect on original feature: Valid C++ 2014 code that attempts to use associative containers having a comparison object with non-const function call operator may fail to compile in this revision of C++.

For example: #include <set> struct compare { bool operator()(int a, int b) { return a < b; } }; int main() { const std::set<int, compare> s; s.find(0); }

Change: The class templates auto_ptr, unary_function, and binary_function, the function templates random_shuffle, and the function templates (and their return types) ptr_fun, mem_fun, mem_fun_ref, bind1st, and bind2nd are not defined.

Rationale: Superseded by new features.

Effect on original feature: Valid C++ 2014 code that uses these class templates and function templates may fail to compile in this revision of C++.

Change: Remove old iostreams members [depr.ios.members].

Rationale: Redundant feature for compatibility with pre-standard code has served its time.

Effect on original feature: A valid C++ 2014 program using these identifiers may be ill-formed in this revision of C++.

Subclause [diff.cpp11] lists the differences between C++ and ISO C++ 2011 (ISO/IEC 14882:2011, Programming Languages — C++), in addition to those listed above, by the chapters of this document.

Change: pp-number can contain one or more single quotes.

Rationale: Necessary to enable single quotes as digit separators.

Effect on original feature: Valid C++ 2011 code may fail to compile or may change meaning in this revision of C++.

For example, the following code is valid both in C++ 2011 and in this revision of C++, but the macro invocation produces different outcomes because the single quotes delimit a character-literal in C++ 2011, whereas they are digit separators in this revision of C++.

For example: #define M(x, ...) __VA_ARGS__ int x[2] = { M(1'2,3'4, 5) }; // int x[2] = { 5 };      --- C++ 2011 // int x[2] = { 3'4, 5 }; --- this revision of C++

Change: New usual (non-placement) deallocator.

Rationale: Required for sized deallocation.

Effect on original feature: Valid C++ 2011 code can declare a global placement allocation function and deallocation function as follows: void* operator new(std::size_t, std::size_t); void operator delete(void*, std::size_t) noexcept;

In this revision of C++, however, the declaration of operator delete might match a predefined usual (non-placement) operator delete ([basic.stc.dynamic]).

If so, the program is ill-formed, as it was for class member allocation functions and deallocation functions ([expr.new]).

Change: A conditional expression with a throw expression as its second or third operand keeps the type and value category of the other operand.

Rationale: Formerly mandated conversions (lvalue-to-rvalue, array-to-pointer, and function-to-pointer standard conversions), especially the creation of the temporary due to lvalue-to-rvalue conversion, were considered gratuitous and surprising.

Effect on original feature: Valid C++ 2011 code that relies on the conversions may behave differently in this revision of C++.

For example: struct S { int x = 1; void mf() { x = 2; } }; int f(bool cond) { S s; (cond ? s : throw 0).mf(); return s.x; }

In C++ 2011, f(true) returns 1.

In this revision of C++, it returns 2.

In C++ 2011, the expression yields sizeof(const char*).

In this revision of C++, it yields sizeof(const char[1]).

Change: constexpr non-static member functions are not implicitly const member functions.

Rationale: Necessary to allow constexpr member functions to mutate the object.

Effect on original feature: Valid C++ 2011 code may fail to compile in this revision of C++.

For example: struct S { constexpr const int &f(); int &f(); };

This code is valid in C++ 2011 but invalid in this revision of C++ because it declares the same member function twice with different return types.

Change: Classes with default member initializers can be aggregates.

Rationale: Necessary to allow default member initializers to be used by aggregate initialization.

Effect on original feature: Valid C++ 2011 code may fail to compile or may change meaning in this revision of C++.

For example: struct S { // Aggregate in C++ 2014 onwards. int m = 1; }; struct X { operator int(); operator S(); }; X a{}; S b{a}; // uses copy constructor in C++ 2011, // performs aggregate initialization in this revision of C++

Change: New header.

Rationale: New functionality.

Effect on original feature: The C++ header <shared_mutex> is new.

Valid C++ 2011 code that #includes a header with that name may be invalid in this revision of C++.

Change: gets is not defined.

Rationale: Use of gets is considered dangerous.

Effect on original feature: Valid C++ 2011 code that uses the gets function may fail to compile in this revision of C++.

Subclause [diff.cpp03] lists the differences between C++ and ISO C++ 2003 (ISO/IEC 14882:2003, Programming Languages — C++), in addition to those listed above, by the chapters of this document.

Change: New kinds of string-literals.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2003 code may fail to compile or produce different results in this revision of C++.

Specifically, macros named R, u8, u8R, u, uR, U, UR, or LR will not be expanded when adjacent to a string-literal but will be interpreted as part of the string-literal.

For example: #define u8 "abc" const char* s = u8"def"; // Previously "abcdef", now "def"

Change: User-defined literal string support.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2003 code may fail to compile or produce different results in this revision of C++.

For example: #define _x "there" "hello"_x // #1

Previously, #1 would have consisted of two separate preprocessing tokens and the macro _x would have been expanded.

In this revision of C++, #1 consists of a single preprocessing token, so the macro is not expanded.

Change: New keywords.

Rationale: Required for new features.

Effect on original feature: Added to Table 5, the following identifiers are new keywords: alignas, alignof, char16_t, char32_t, constexpr, decltype, noexcept, nullptr, static_assert, and thread_local.

Valid C++ 2003 code using these identifiers is invalid in this revision of C++.

Change: Type of integer literals.

Rationale: C99 compatibility.

Effect on original feature: Certain integer literals larger than can be represented by long could change from an unsigned integer type to signed long long.

Change: Only literals are integer null pointer constants.

Rationale: Removing surprising interactions with templates and constant expressions.

Effect on original feature: Valid C++ 2003 code may fail to compile or produce different results in this revision of C++.

For example: void f(void *); // #1 void f(...); // #2 template<int N> void g() { f(0*N); // calls #2; used to call #1 }

Change: Specify rounding for results of integer / and %.

Rationale: Increase portability, C99 compatibility.

Effect on original feature: Valid C++ 2003 code that uses integer division rounds the result toward 0 or toward negative infinity, whereas this revision of C++ always rounds the result toward 0.

Change: && is valid in a type-name.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2003 code may fail to compile or produce different results in this revision of C++.

For example: bool b1 = new int && false; // previously false, now ill-formed struct S { operator int(); }; bool b2 = &S::operator int && false; // previously false, now ill-formed

Change: Remove auto as a storage class specifier.

Rationale: New feature.

Effect on original feature: Valid C++ 2003 code that uses the keyword auto as a storage class specifier may be invalid in this revision of C++.

In this revision of C++, auto indicates that the type of a variable is to be deduced from its initializer expression.

Change: Narrowing restrictions in aggregate initializers.

Rationale: Catches bugs.

Effect on original feature: Valid C++ 2003 code may fail to compile in this revision of C++.

For example: int x[] = { 2.0 };

This code is valid in C++ 2003 but invalid in this revision of C++ because double to int is a narrowing conversion.

Change: Names declared in an anonymous namespace changed from external linkage to internal linkage; language linkage applies to names with external linkage only.

Rationale: Alignment with user expectations.

Effect on original feature: Valid C++ 2003 code may violate the one-definition rule ([basic.def.odr]) in this revision of C++.

For example: namespace { extern "C" { extern int x; } } // #1, previously external linkage and C language linkage, // now internal linkage and C++ language linkage namespace A { extern "C" int x = 42; } // #2, external linkage and C language linkage int main(void) { return x; }

This code is valid in C++ 2003, but #2 is not a definition for #1 in this revision of C++, violating the one-definition rule.

Change: Implicitly-declared special member functions are defined as deleted when the implicit definition would have been ill-formed.

Rationale: Improves template argument deduction failure.

Effect on original feature: A valid C++ 2003 program that uses one of these special member functions in a context where the definition is not required (e.g., in an expression that is not potentially evaluated) becomes ill-formed.

Change: User-declared destructors have an implicit exception specification.

Rationale: Clarification of destructor requirements.

Effect on original feature: Valid C++ 2003 code may execute differently in this revision of C++.

In particular, destructors that throw exceptions will call std​::​terminate (without calling std​::​unexpected) if their exception specification is non-throwing.

Change: Repurpose export for modules ([module], [cpp.module], [cpp.import]).

Rationale: No implementation consensus for the C++ 2003 meaning of export.

Effect on original feature: A valid C++ 2003 program containing export is ill-formed in this revision of C++.

Change: Remove whitespace requirement for nested closing template right angle brackets.

Rationale: Considered a persistent but minor annoyance.

Template aliases representing non-class types would exacerbate whitespace issues.

Effect on original feature: Change to semantics of well-defined expression.

A valid C++ 2003 expression containing a right angle bracket (“>”) followed immediately by another right angle bracket may now be treated as closing two templates.

For example: template <class T> struct X { }; template <int N> struct Y { }; X< Y< 1 >> 2 > > x;

This code is valid in C++ 2003 because “>>” is a right-shift operator, but invalid in this revision of C++ because “>>” closes two templates.

Change: Allow dependent calls of functions with internal linkage.

Rationale: Overly constrained, simplify overload resolution rules.

Effect on original feature: A valid C++ 2003 program can get a different result in this revision of C++.

Affected: [library] – [thread]

Change: New reserved identifiers.

Rationale: Required by new features.

Effect on original feature: Valid C++ 2003 code that uses any identifiers added to the C++ standard library by later revisions of C++ may fail to compile or produce different results in this revision of C++.

A comprehensive list of identifiers used by the C++ standard library can be found in the Index of Library Names in this document.

Change: New headers.

Rationale: New functionality.

Effect on original feature: The following C++ headers are new: <array>, <atomic>, <chrono>, <condition_variable>, <forward_list>, <future>, <initializer_list>, <mutex>, <random>, <ratio>, <regex>, <scoped_allocator>, <system_error>, <thread>, <tuple>, <typeindex>, <type_traits>, <unordered_map>, and <unordered_set>.

In addition the following C compatibility headers are new: <cfenv>, <cinttypes>, <cstdint>, and <cuchar>.

Valid C++ 2003 code that #includes headers with these names may be invalid in this revision of C++.

Effect on original feature: Function swap moved to a different header

Rationale: Remove dependency on <algorithm> for swap.

Effect on original feature: Valid C++ 2003 code that has been compiled expecting swap to be in <algorithm> may have to instead include <utility>.

Change: New reserved namespace.

Rationale: New functionality.

Effect on original feature: The global namespace posix is now reserved for standardization.

Valid C++ 2003 code that uses a top-level namespace posix may be invalid in this revision of C++.

Change: Additional restrictions on macro names.

Rationale: Avoid hard to diagnose or non-portable constructs.

Effect on original feature: Names of attribute identifiers may not be used as macro names.

Valid C++ 2003 code that defines override, final, carries_dependency, or noreturn as macros is invalid in this revision of C++.

Change: operator new may throw exceptions other than std​::​bad_alloc.

Rationale: Consistent application of noexcept.

Effect on original feature: Valid C++ 2003 code that assumes that global operator new only throws std​::​bad_alloc may execute differently in this revision of C++.

Valid C++ 2003 code that replaces the global replaceable operator new is ill-formed in this revision of C++, because the exception specification of throw(std​::​bad_alloc) was removed.

Change: Thread-local error numbers.

Rationale: Support for new thread facilities.

Effect on original feature: Valid but implementation-specific C++ 2003 code that relies on errno being the same across threads may change behavior in this revision of C++.

Change: Standard function object types no longer derived from std​::​unary_function or std​::​binary_function.

Rationale: Superseded by new feature; unary_function and binary_function are no longer defined.

Effect on original feature: Valid C++ 2003 code that depends on function object types being derived from unary_function or binary_function may fail to compile in this revision of C++.

Change: basic_string requirements no longer allow reference-counted strings.

Rationale: Invalidation is subtly different with reference-counted strings.

This change regularizes behavior.

Effect on original feature: Valid C++ 2003 code may execute differently in this revision of C++.

Change: Loosen basic_string invalidation rules.

Rationale: Allow small-string optimization.

Effect on original feature: Valid C++ 2003 code may execute differently in this revision of C++.

Some const member functions, such as data and c_str, no longer invalidate iterators.

Change: Complexity of size() member functions now constant.

Rationale: Lack of specification of complexity of size() resulted in divergent implementations with inconsistent performance characteristics.

Effect on original feature: Some container implementations that conform to C++ 2003 may not conform to the specified size() requirements in this revision of C++.

Adjusting containers such as std​::​list to the stricter requirements may require incompatible changes.

Change: Requirements change: relaxation.

Rationale: Clarification.

Effect on original feature: Valid C++ 2003 code that attempts to meet the specified container requirements may now be over-specified.

Code that attempted to be portable across containers may need to be adjusted as follows:

• (2.1)
  not all containers provide size(); use empty() instead of size() == 0;
• (2.2)
  not all containers are empty after construction (array);
• (2.3)
  not all containers have constant complexity for swap() (array).

Change: Requirements change: default constructible.

Rationale: Clarification of container requirements.

Effect on original feature: Valid C++ 2003 code that attempts to explicitly instantiate a container using a user-defined type with no default constructor may fail to compile.

Change: Signature changes: from void return types.

Rationale: Old signature threw away useful information that may be expensive to recalculate.

Effect on original feature: The following member functions have changed:

• (4.1)
  erase(iter) for set, multiset, map, multimap
• (4.2)
  erase(begin, end) for set, multiset, map, multimap
• (4.3)
  insert(pos, num, val) for vector, deque, list, forward_list
• (4.4)
  insert(pos, beg, end) for vector, deque, list, forward_list

Valid C++ 2003 code that relies on these functions returning void (e.g., code that creates a pointer to member function that points to one of these functions) will fail to compile with this revision of C++.

Change: Signature changes: from iterator to const_iterator parameters.

Rationale: Overspecification.

Effect on original feature: The signatures of the following member functions changed from taking an iterator to taking a const_iterator:

• (5.1)
  insert(iter, val) for vector, deque, list, set, multiset, map, multimap
• (5.2)
  insert(pos, beg, end) for vector, deque, list, forward_list
• (5.3)
  erase(begin, end) for set, multiset, map, multimap
• (5.4)
  all forms of list​::​splice
• (5.5)
  all forms of list​::​merge

Valid C++ 2003 code that uses these functions may fail to compile with this revision of C++.

Change: Signature changes: resize.

Rationale: Performance, compatibility with move semantics.

Effect on original feature: For vector, deque, and list the fill value passed to resize is now passed by reference instead of by value, and an additional overload of resize has been added.

Valid C++ 2003 code that uses this function may fail to compile with this revision of C++.

Change: Result state of inputs after application of some algorithms.

Rationale: Required by new feature.

Effect on original feature: A valid C++ 2003 program may detect that an object with a valid but unspecified state has a different valid but unspecified state with this revision of C++.

For example, std​::​remove and std​::​remove_if may leave the tail of the input sequence with a different set of values than previously.

Change: Specified representation of complex numbers.

Rationale: Compatibility with C99.

Effect on original feature: Valid C++ 2003 code that uses implementation-specific knowledge about the binary representation of the required template specializations of std​::​complex may not be compatible with this revision of C++.

Change: The num_get facet recognizes hexadecimal floating point values.

Rationale: Required by new feature.

Effect on original feature: Valid C++ 2003 code may have different behavior in this revision of C++.

Change: Specify use of explicit in existing boolean conversion functions.

Rationale: Clarify intentions, avoid workarounds.

Effect on original feature: Valid C++ 2003 code that relies on implicit boolean conversions will fail to compile with this revision of C++.

Such conversions occur in the following conditions:

• (1.1)
  passing a value to a function that takes an argument of type bool;
• (1.2)
  using operator== to compare to false or true;
• (1.3)
  returning a value from a function with a return type of bool;
• (1.4)
  initializing members of type bool via aggregate initialization;
• (1.5)
  initializing a const bool& which would bind to a temporary object.

Change: Change base class of std​::​ios_base​::​failure.

Rationale: More detailed error messages.

Effect on original feature: std​::​ios_base​::​failure is no longer derived directly from std​::​exception, but is now derived from std​::​system_error, which in turn is derived from std​::​runtime_error.

Valid C++ 2003 code that assumes that std​::​ios_base​::​failure is derived directly from std​::​exception may execute differently in this revision of C++.

Change: Flag types in std​::​ios_base are now bitmasks with values defined as constexpr static members.

Rationale: Required for new features.

Effect on original feature: Valid C++ 2003 code that relies on std​::​ios_base flag types being represented as std​::​bitset or as an integer type may fail to compile with this revision of C++.

For example: #include <iostream> int main() { int flag = std::ios_base::hex; std::cout.setf(flag); // error: setf does not take argument of type int }

Subclause [diff.iso] lists the differences between C++ and ISO C, in addition to those listed above, by the chapters of this document.

Change: New Keywords

New keywords are added to C++; see [lex.key].

Rationale: These keywords were added in order to implement the new semantics of C++.

Effect on original feature: Change to semantics of well-defined feature.

Any ISO C programs that used any of these keywords as identifiers are not valid C++ programs.

Difficulty of converting: Syntactic transformation.

Converting one specific program is easy.

Converting a large collection of related programs takes more work.

How widely used: Common.

Change: Type of character-literal is changed from int to char.

Rationale: This is needed for improved overloaded function argument type matching.

For example: int function( int i ); int function( char c ); function( 'x' );

It is preferable that this call match the second version of function rather than the first.

Effect on original feature: Change to semantics of well-defined feature.

ISO C programs which depend on sizeof('x') == sizeof(int) will not work the same as C++ programs.

Difficulty of converting: Simple.

How widely used: Programs which depend upon sizeof('x') are probably rare.

Change: Concatenated string-literals can no longer have conflicting encoding-prefixes.

Rationale: Removal of non-portable feature.

Effect on original feature: Concatenation of string-literals with different encoding-prefixes is now ill-formed.

Difficulty of converting: Syntactic transformation.

How widely used: Seldom.

Change: String literals made const.

The type of a string-literal is changed from “array of char” to “array of const char”.

The type of a UTF-8 string literal is changed from “array of char” to “array of const char8_t”.

The type of a UTF-16 string literal is changed from “array of some-integer-type” to “array of const char16_t”.

The type of a UTF-32 string literal is changed from “array of some-integer-type” to “array of const char32_t”.

The type of a wide string literal is changed from “array of wchar_t” to “array of const wchar_t”.

Rationale: This avoids calling an inappropriate overloaded function, which might expect to be able to modify its argument.

Effect on original feature: Change to semantics of well-defined feature.

Difficulty of converting: Syntactic transformation.

The fix is to add a cast: char* p = "abc"; // valid in C, invalid in C++ void f(char*) { char* p = (char*)"abc"; // OK, cast added f(p); f((char*)"def"); // OK, cast added }

How widely used: Programs that have a legitimate reason to treat string literal objects as potentially modifiable memory are probably rare.

Change: C++ does not have “tentative definitions” as in C.

E.g., at file scope, int i; int i; is valid in C, invalid in C++.

This makes it impossible to define mutually referential file-local objects with static storage duration, if initializers are restricted to the syntactic forms of C.

For example, struct X { int i; struct X* next; }; static struct X a; static struct X b = { 0, &a }; static struct X a = { 1, &b };

Rationale: This avoids having different initialization rules for fundamental types and user-defined types.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

In C++, the initializer for one of a set of mutually-referential file-local objects with static storage duration must invoke a function call to achieve the initialization.

How widely used: Seldom.

Change: A struct is a scope in C++, not in C. For example, struct X { struct Y { int a; } b; }; struct Y c; is valid in C but not in C++, which would require X​::​Y c;.

Rationale: Class scope is crucial to C++, and a struct is a class.

Effect on original feature: Change to semantics of well-defined feature.

Difficulty of converting: Semantic transformation.

How widely used: C programs use struct extremely frequently, but the change is only noticeable when struct, enumeration, or enumerator names are referred to outside the struct.

The latter is probably rare.

Affected subclause: [basic.link] [also [dcl.type]]

Change: A name of file scope that is explicitly declared const, and not explicitly declared extern, has internal linkage, while in C it would have external linkage.

Rationale: Because const objects may be used as values during translation in C++, this feature urges programmers to provide an explicit initializer for each const object.

This feature allows the user to put const objects in source files that are included in more than one translation unit.

Effect on original feature: Change to semantics of well-defined feature.

Difficulty of converting: Semantic transformation.

How widely used: Seldom.

Change: The main function cannot be called recursively and cannot have its address taken.

Rationale: The main function may require special actions.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Trivial: create an intermediary function such as mymain(argc, argv).

How widely used: Seldom.

Change: C allows “compatible types” in several places, C++ does not.

For example, otherwise-identical struct types with different tag names are “compatible” in C but are distinctly different types in C++.

Rationale: Stricter type checking is essential for C++.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

The “typesafe linkage” mechanism will find many, but not all, of such problems.

Those problems not found by typesafe linkage will continue to function properly, according to the “layout compatibility rules” of this document.

How widely used: Common.

Change: Converting void* to a pointer-to-object type requires casting.

ISO C accepts this usage of pointer to void being assigned to a pointer to object type.

C++ does not.

Rationale: C++ tries harder than C to enforce compile-time type safety.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Can be automated.

Violations will be diagnosed by the C++ translator.

The fix is to add a cast.

For example: char* c = (char*) b;

How widely used: This is fairly widely used but it is good programming practice to add the cast when assigning pointer-to-void to pointer-to-object.

Some ISO C translators will give a warning if the cast is not used.

Change: Operations mixing a value of an enumeration type and a value of a different enumeration type or of a floating-point type are not valid.

For example: enum E1 { e }; enum E2 { f }; int b = e <= 3.7; // valid in C; ill-formed in C++ int k = f - e; // valid in C; ill-formed in C++ int x = 1 ? e : f; // valid in C; ill-formed in C++

Rationale: Reinforcing type safety in C++.

Effect on original feature: Well-formed C code will not compile with this International Standard.

Difficulty of converting: Violations will be diagnosed by the C++ translator.

The original behavior can be restored with a cast or integral promotion.

For example: enum E1 { e }; enum E2 { f }; int b = (int)e <= 3.7; int k = +f - e;

How widely used: Uncommon.

Change: Decrement operator is not allowed with bool operand.

Rationale: Feature with surprising semantics.

Effect on original feature: A valid ISO C expression utilizing the decrement operator on a bool lvalue (for instance, via the C typedef in <stdbool.h>) is ill-formed in C++.

Change: In C++, types can only be defined in declarations, not in expressions.

In C, a sizeof expression or cast expression may define a new type.

For example, p = (void*)(struct x {int i;} *)0; defines a new type, struct x.

Rationale: This prohibition helps to clarify the location of definitions in the source code.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Syntactic transformation.

How widely used: Seldom.

Change: The result of a conditional expression, an assignment expression, or a comma expression may be an lvalue.

Rationale: C++ is an object-oriented language, placing relatively more emphasis on lvalues.

For example, function calls may yield lvalues.

Effect on original feature: Change to semantics of well-defined feature.

Some C expressions that implicitly rely on lvalue-to-rvalue conversions will yield different results.

For example, char arr[100]; sizeof(0, arr) yields 100 in C++ and sizeof(char*) in C.

Difficulty of converting: Programs must add explicit casts to the appropriate rvalue.

How widely used: Rare.

Change: It is now invalid to jump past a declaration with explicit or implicit initializer (except across entire block not entered).

Rationale: Constructors used in initializers may allocate resources which need to be de-allocated upon leaving the block.

Allowing jump past initializers would require complicated runtime determination of allocation.

Furthermore, many operations on such an uninitialized object have undefined behavior.

With this simple compile-time rule, C++ assures that if an initialized variable is in scope, then it has assuredly been initialized.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

How widely used: Seldom.

Change: It is now invalid to return (explicitly or implicitly) from a function which is declared to return a value without actually returning a value.

Rationale: The caller and callee may assume fairly elaborate return-value mechanisms for the return of class objects.

If some flow paths execute a return without specifying any value, the implementation must embody many more complications.

Besides, promising to return a value of a given type, and then not returning such a value, has always been recognized to be a questionable practice, tolerated only because very-old C had no distinction between functions with void and int return types.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

Add an appropriate return value to the source code, such as zero.

How widely used: Seldom.

For several years, many existing C implementations have produced warnings in this case.

Change: In C++, the static or extern specifiers can only be applied to names of objects or functions.

Using these specifiers with type declarations is illegal in C++.

In C, these specifiers are ignored when used on type declarations.

Example: static struct S { // valid C, invalid in C++ int i; };

Rationale: Storage class specifiers don't have any meaning when associated with a type.

In C++, class members can be declared with the static storage class specifier.

Storage class specifiers on type declarations can be confusing for users.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Syntactic transformation.

How widely used: Seldom.

Change: In C++, register is not a storage class specifier.

Rationale: The storage class specifier had no effect in C++.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Syntactic transformation.

How widely used: Common.

Change: A C++ typedef-name must be different from any class type name declared in the same scope (except if the typedef is a synonym of the class name with the same name).

In C, a typedef-name and a struct tag name declared in the same scope can have the same name (because they have different name spaces).

Example: typedef struct name1 { /* ... */ } name1; // valid C and C++ struct name { /* ... */ }; typedef int name; // valid C, invalid C++

Rationale: For ease of use, C++ doesn't require that a type name be prefixed with the keywords class, struct or union when used in object declarations or type casts.

Example: class name { /* ... */ }; name i; // i has type class name

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

One of the 2 types has to be renamed.

How widely used: Seldom.

Affected subclause: [dcl.type] [see also [basic.link]]

Change: Const objects must be initialized in C++ but can be left uninitialized in C.

Rationale: A const object cannot be assigned to so it must be initialized to hold a useful value.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

How widely used: Seldom.

Change: The keyword auto cannot be used as a storage class specifier.

Example: void f() { auto int x; // valid C, invalid C++ }

Rationale: Allowing the use of auto to deduce the type of a variable from its initializer results in undesired interpretations of auto as a storage class specifier in certain contexts.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Syntactic transformation.

How widely used: Rare.

Change: In C++, a function declared with an empty parameter list takes no arguments.

In C, an empty parameter list means that the number and type of the function arguments are unknown.

Example: int f(); // means int f(void) in C++ // int f( unknown ) in C

Rationale: This is to avoid erroneous function calls (i.e., function calls with the wrong number or type of arguments).

Effect on original feature: Change to semantics of well-defined feature.

This feature was marked as “obsolescent” in C.

Difficulty of converting: Syntactic transformation.

The function declarations using C incomplete declaration style must be completed to become full prototype declarations.

A program may need to be updated further if different calls to the same (non-prototype) function have different numbers of arguments or if the type of corresponding arguments differed.

How widely used: Common.

Affected subclause: [dcl.fct] [see [expr.sizeof]]

Change: In C++, types may not be defined in return or parameter types.

In C, these type definitions are allowed.

Example: void f( struct S { int a; } arg ) {} // valid C, invalid C++ enum E { A, B, C } f() {} // valid C, invalid C++

Rationale: When comparing types in different translation units, C++ relies on name equivalence when C relies on structural equivalence.

Regarding parameter types: since the type defined in a parameter list would be in the scope of the function, the only legal calls in C++ would be from within the function itself.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Semantic transformation.

The type definitions must be moved to file scope, or in header files.

How widely used: Seldom.

This style of type definition is seen as poor coding style.

Change: In C++, the syntax for function definition excludes the “old-style” C function.

In C, “old-style” syntax is allowed, but deprecated as “obsolescent”.

Rationale: Prototypes are essential to type safety.

Effect on original feature: Deletion of semantically well-defined feature.

Difficulty of converting: Syntactic transformation.

How widely used: Common in old programs, but already known to be obsolescent.

Change: In C++, designated initialization support is restricted compared to the corresponding functionality in C.