1 //==--- AttrDocs.td - Attribute documentation ----------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===---------------------------------------------------------------------===//
10 def GlobalDocumentation {
12 -------------------------------------------------------------------
13 NOTE: This file is automatically generated by running clang-tblgen
14 -gen-attr-docs. Do not edit this file by hand!!
15 -------------------------------------------------------------------
26 This page lists the attributes currently supported by Clang.
30 def SectionDocs : Documentation {
31 let Category = DocCatVariable;
33 The ``section`` attribute allows you to specify a specific section a
34 global variable or function should be in after translation.
36 let Heading = "section (gnu::section, __declspec(allocate))";
39 def InitSegDocs : Documentation {
40 let Category = DocCatVariable;
42 The attribute applied by ``pragma init_seg()`` controls the section into
43 which global initialization function pointers are emitted. It is only
44 available with ``-fms-extensions``. Typically, this function pointer is
45 emitted into ``.CRT$XCU`` on Windows. The user can change the order of
46 initialization by using a different section name with the same
47 ``.CRT$XC`` prefix and a suffix that sorts lexicographically before or
48 after the standard ``.CRT$XCU`` sections. See the init_seg_
49 documentation on MSDN for more information.
51 .. _init_seg: http://msdn.microsoft.com/en-us/library/7977wcck(v=vs.110).aspx
55 def TLSModelDocs : Documentation {
56 let Category = DocCatVariable;
58 The ``tls_model`` attribute allows you to specify which thread-local storage
59 model to use. It accepts the following strings:
66 TLS models are mutually exclusive.
70 def DLLExportDocs : Documentation {
71 let Category = DocCatVariable;
73 The ``__declspec(dllexport)`` attribute declares a variable, function, or
74 Objective-C interface to be exported from the module. It is available under the
75 ``-fdeclspec`` flag for compatibility with various compilers. The primary use
76 is for COFF object files which explicitly specify what interfaces are available
77 for external use. See the dllexport_ documentation on MSDN for more
80 .. _dllexport: https://msdn.microsoft.com/en-us/library/3y1sfaz2.aspx
84 def DLLImportDocs : Documentation {
85 let Category = DocCatVariable;
87 The ``__declspec(dllimport)`` attribute declares a variable, function, or
88 Objective-C interface to be imported from an external module. It is available
89 under the ``-fdeclspec`` flag for compatibility with various compilers. The
90 primary use is for COFF object files which explicitly specify what interfaces
91 are imported from external modules. See the dllimport_ documentation on MSDN
94 .. _dllimport: https://msdn.microsoft.com/en-us/library/3y1sfaz2.aspx
98 def ThreadDocs : Documentation {
99 let Category = DocCatVariable;
101 The ``__declspec(thread)`` attribute declares a variable with thread local
102 storage. It is available under the ``-fms-extensions`` flag for MSVC
103 compatibility. See the documentation for `__declspec(thread)`_ on MSDN.
105 .. _`__declspec(thread)`: http://msdn.microsoft.com/en-us/library/9w1sdazb.aspx
107 In Clang, ``__declspec(thread)`` is generally equivalent in functionality to the
108 GNU ``__thread`` keyword. The variable must not have a destructor and must have
109 a constant initializer, if any. The attribute only applies to variables
110 declared with static storage duration, such as globals, class static data
111 members, and static locals.
115 def CarriesDependencyDocs : Documentation {
116 let Category = DocCatFunction;
118 The ``carries_dependency`` attribute specifies dependency propagation into and
121 When specified on a function or Objective-C method, the ``carries_dependency``
122 attribute means that the return value carries a dependency out of the function,
123 so that the implementation need not constrain ordering upon return from that
124 function. Implementations of the function and its caller may choose to preserve
125 dependencies instead of emitting memory ordering instructions such as fences.
127 Note, this attribute does not change the meaning of the program, but may result
128 in generation of more efficient code.
132 def C11NoReturnDocs : Documentation {
133 let Category = DocCatFunction;
135 A function declared as ``_Noreturn`` shall not return to its caller. The
136 compiler will generate a diagnostic for a function declared as ``_Noreturn``
137 that appears to be capable of returning to its caller.
141 def CXX11NoReturnDocs : Documentation {
142 let Category = DocCatFunction;
144 A function declared as ``[[noreturn]]`` shall not return to its caller. The
145 compiler will generate a diagnostic for a function declared as ``[[noreturn]]``
146 that appears to be capable of returning to its caller.
150 def AssertCapabilityDocs : Documentation {
151 let Category = DocCatFunction;
152 let Heading = "assert_capability (assert_shared_capability, clang::assert_capability, clang::assert_shared_capability)";
154 Marks a function that dynamically tests whether a capability is held, and halts
155 the program if it is not held.
159 def AcquireCapabilityDocs : Documentation {
160 let Category = DocCatFunction;
161 let Heading = "acquire_capability (acquire_shared_capability, clang::acquire_capability, clang::acquire_shared_capability)";
163 Marks a function as acquiring a capability.
167 def TryAcquireCapabilityDocs : Documentation {
168 let Category = DocCatFunction;
169 let Heading = "try_acquire_capability (try_acquire_shared_capability, clang::try_acquire_capability, clang::try_acquire_shared_capability)";
171 Marks a function that attempts to acquire a capability. This function may fail to
172 actually acquire the capability; they accept a Boolean value determining
173 whether acquiring the capability means success (true), or failing to acquire
174 the capability means success (false).
178 def ReleaseCapabilityDocs : Documentation {
179 let Category = DocCatFunction;
180 let Heading = "release_capability (release_shared_capability, clang::release_capability, clang::release_shared_capability)";
182 Marks a function as releasing a capability.
186 def AssumeAlignedDocs : Documentation {
187 let Category = DocCatFunction;
189 Use ``__attribute__((assume_aligned(<alignment>[,<offset>]))`` on a function
190 declaration to specify that the return value of the function (which must be a
191 pointer type) has the specified offset, in bytes, from an address with the
192 specified alignment. The offset is taken to be zero if omitted.
196 // The returned pointer value has 32-byte alignment.
197 void *a() __attribute__((assume_aligned (32)));
199 // The returned pointer value is 4 bytes greater than an address having
200 // 32-byte alignment.
201 void *b() __attribute__((assume_aligned (32, 4)));
203 Note that this attribute provides information to the compiler regarding a
204 condition that the code already ensures is true. It does not cause the compiler
205 to enforce the provided alignment assumption.
209 def AllocSizeDocs : Documentation {
210 let Category = DocCatFunction;
212 The ``alloc_size`` attribute can be placed on functions that return pointers in
213 order to hint to the compiler how many bytes of memory will be available at the
214 returned poiner. ``alloc_size`` takes one or two arguments.
216 - ``alloc_size(N)`` implies that argument number N equals the number of
217 available bytes at the returned pointer.
218 - ``alloc_size(N, M)`` implies that the product of argument number N and
219 argument number M equals the number of available bytes at the returned
222 Argument numbers are 1-based.
224 An example of how to use ``alloc_size``
228 void *my_malloc(int a) __attribute__((alloc_size(1)));
229 void *my_calloc(int a, int b) __attribute__((alloc_size(1, 2)));
232 void *const p = my_malloc(100);
233 assert(__builtin_object_size(p, 0) == 100);
234 void *const a = my_calloc(20, 5);
235 assert(__builtin_object_size(a, 0) == 100);
238 .. Note:: This attribute works differently in clang than it does in GCC.
239 Specifically, clang will only trace ``const`` pointers (as above); we give up
240 on pointers that are not marked as ``const``. In the vast majority of cases,
241 this is unimportant, because LLVM has support for the ``alloc_size``
242 attribute. However, this may cause mildly unintuitive behavior when used with
243 other attributes, such as ``enable_if``.
247 def EnableIfDocs : Documentation {
248 let Category = DocCatFunction;
250 .. Note:: Some features of this attribute are experimental. The meaning of
251 multiple enable_if attributes on a single declaration is subject to change in
252 a future version of clang. Also, the ABI is not standardized and the name
253 mangling may change in future versions. To avoid that, use asm labels.
255 The ``enable_if`` attribute can be placed on function declarations to control
256 which overload is selected based on the values of the function's arguments.
257 When combined with the ``overloadable`` attribute, this feature is also
263 int isdigit(int c) __attribute__((enable_if(c <= -1 || c > 255, "chosen when 'c' is out of range"))) __attribute__((unavailable("'c' must have the value of an unsigned char or EOF")));
268 isdigit(-10); // results in a compile-time error.
271 The enable_if attribute takes two arguments, the first is an expression written
272 in terms of the function parameters, the second is a string explaining why this
273 overload candidate could not be selected to be displayed in diagnostics. The
274 expression is part of the function signature for the purposes of determining
275 whether it is a redeclaration (following the rules used when determining
276 whether a C++ template specialization is ODR-equivalent), but is not part of
279 The enable_if expression is evaluated as if it were the body of a
280 bool-returning constexpr function declared with the arguments of the function
281 it is being applied to, then called with the parameters at the call site. If the
282 result is false or could not be determined through constant expression
283 evaluation, then this overload will not be chosen and the provided string may
284 be used in a diagnostic if the compile fails as a result.
286 Because the enable_if expression is an unevaluated context, there are no global
287 state changes, nor the ability to pass information from the enable_if
288 expression to the function body. For example, suppose we want calls to
289 strnlen(strbuf, maxlen) to resolve to strnlen_chk(strbuf, maxlen, size of
290 strbuf) only if the size of strbuf can be determined:
294 __attribute__((always_inline))
295 static inline size_t strnlen(const char *s, size_t maxlen)
296 __attribute__((overloadable))
297 __attribute__((enable_if(__builtin_object_size(s, 0) != -1))),
298 "chosen when the buffer size is known but 'maxlen' is not")))
300 return strnlen_chk(s, maxlen, __builtin_object_size(s, 0));
303 Multiple enable_if attributes may be applied to a single declaration. In this
304 case, the enable_if expressions are evaluated from left to right in the
305 following manner. First, the candidates whose enable_if expressions evaluate to
306 false or cannot be evaluated are discarded. If the remaining candidates do not
307 share ODR-equivalent enable_if expressions, the overload resolution is
308 ambiguous. Otherwise, enable_if overload resolution continues with the next
309 enable_if attribute on the candidates that have not been discarded and have
310 remaining enable_if attributes. In this way, we pick the most specific
311 overload out of a number of viable overloads using enable_if.
315 void f() __attribute__((enable_if(true, ""))); // #1
316 void f() __attribute__((enable_if(true, ""))) __attribute__((enable_if(true, ""))); // #2
318 void g(int i, int j) __attribute__((enable_if(i, ""))); // #1
319 void g(int i, int j) __attribute__((enable_if(j, ""))) __attribute__((enable_if(true))); // #2
321 In this example, a call to f() is always resolved to #2, as the first enable_if
322 expression is ODR-equivalent for both declarations, but #1 does not have another
323 enable_if expression to continue evaluating, so the next round of evaluation has
324 only a single candidate. In a call to g(1, 1), the call is ambiguous even though
325 #2 has more enable_if attributes, because the first enable_if expressions are
328 Query for this feature with ``__has_attribute(enable_if)``.
330 Note that functions with one or more ``enable_if`` attributes may not have
331 their address taken, unless all of the conditions specified by said
332 ``enable_if`` are constants that evaluate to ``true``. For example:
336 const int TrueConstant = 1;
337 const int FalseConstant = 0;
338 int f(int a) __attribute__((enable_if(a > 0, "")));
339 int g(int a) __attribute__((enable_if(a == 0 || a != 0, "")));
340 int h(int a) __attribute__((enable_if(1, "")));
341 int i(int a) __attribute__((enable_if(TrueConstant, "")));
342 int j(int a) __attribute__((enable_if(FalseConstant, "")));
346 ptr = &f; // error: 'a > 0' is not always true
347 ptr = &g; // error: 'a == 0 || a != 0' is not a truthy constant
348 ptr = &h; // OK: 1 is a truthy constant
349 ptr = &i; // OK: 'TrueConstant' is a truthy constant
350 ptr = &j; // error: 'FalseConstant' is a constant, but not truthy
353 Because ``enable_if`` evaluation happens during overload resolution,
354 ``enable_if`` may give unintuitive results when used with templates, depending
355 on when overloads are resolved. In the example below, clang will emit a
356 diagnostic about no viable overloads for ``foo`` in ``bar``, but not in ``baz``:
360 double foo(int i) __attribute__((enable_if(i > 0, "")));
361 void *foo(int i) __attribute__((enable_if(i <= 0, "")));
363 auto bar() { return foo(I); }
365 template <typename T>
366 auto baz() { return foo(T::number); }
368 struct WithNumber { constexpr static int number = 1; };
370 bar<sizeof(WithNumber)>();
374 This is because, in ``bar``, ``foo`` is resolved prior to template
375 instantiation, so the value for ``I`` isn't known (thus, both ``enable_if``
376 conditions for ``foo`` fail). However, in ``baz``, ``foo`` is resolved during
377 template instantiation, so the value for ``T::number`` is known.
381 def DiagnoseIfDocs : Documentation {
382 let Category = DocCatFunction;
384 The ``diagnose_if`` attribute can be placed on function declarations to emit
385 warnings or errors at compile-time if calls to the attributed function meet
386 certain user-defined criteria. For example:
390 __attribute__((diagnose_if(a >= 0, "Redundant abs call", "warning")));
392 __attribute__((diagnose_if(a >= 0, "Redundant abs call", "error")));
394 int val = abs(1); // warning: Redundant abs call
395 int val2 = must_abs(1); // error: Redundant abs call
397 int val4 = must_abs(val); // Because run-time checks are not emitted for
398 // diagnose_if attributes, this executes without
402 ``diagnose_if`` is closely related to ``enable_if``, with a few key differences:
404 * Overload resolution is not aware of ``diagnose_if`` attributes: they're
405 considered only after we select the best candidate from a given candidate set.
406 * Function declarations that differ only in their ``diagnose_if`` attributes are
407 considered to be redeclarations of the same function (not overloads).
408 * If the condition provided to ``diagnose_if`` cannot be evaluated, no
409 diagnostic will be emitted.
411 Otherwise, ``diagnose_if`` is essentially the logical negation of ``enable_if``.
413 As a result of bullet number two, ``diagnose_if`` attributes will stack on the
414 same function. For example:
418 int foo() __attribute__((diagnose_if(1, "diag1", "warning")));
419 int foo() __attribute__((diagnose_if(1, "diag2", "warning")));
421 int bar = foo(); // warning: diag1
423 int (*fooptr)(void) = foo; // warning: diag1
426 constexpr int supportsAPILevel(int N) { return N < 5; }
428 __attribute__((diagnose_if(!supportsAPILevel(10),
429 "Upgrade to API level 10 to use baz", "error")));
431 __attribute__((diagnose_if(!a, "0 is not recommended.", "warning")));
433 int (*bazptr)(int) = baz; // error: Upgrade to API level 10 to use baz
434 int v = baz(0); // error: Upgrade to API level 10 to use baz
436 Query for this feature with ``__has_attribute(diagnose_if)``.
440 def PassObjectSizeDocs : Documentation {
441 let Category = DocCatVariable; // Technically it's a parameter doc, but eh.
443 .. Note:: The mangling of functions with parameters that are annotated with
444 ``pass_object_size`` is subject to change. You can get around this by
445 using ``__asm__("foo")`` to explicitly name your functions, thus preserving
446 your ABI; also, non-overloadable C functions with ``pass_object_size`` are
449 The ``pass_object_size(Type)`` attribute can be placed on function parameters to
450 instruct clang to call ``__builtin_object_size(param, Type)`` at each callsite
451 of said function, and implicitly pass the result of this call in as an invisible
452 argument of type ``size_t`` directly after the parameter annotated with
453 ``pass_object_size``. Clang will also replace any calls to
454 ``__builtin_object_size(param, Type)`` in the function by said implicit
461 int bzero1(char *const p __attribute__((pass_object_size(0))))
462 __attribute__((noinline)) {
464 for (/**/; i < (int)__builtin_object_size(p, 0); ++i) {
472 int n = bzero1(&chars[0]);
473 assert(n == sizeof(chars));
477 If successfully evaluating ``__builtin_object_size(param, Type)`` at the
478 callsite is not possible, then the "failed" value is passed in. So, using the
479 definition of ``bzero1`` from above, the following code would exit cleanly:
483 int main2(int argc, char *argv[]) {
484 int n = bzero1(argv);
489 ``pass_object_size`` plays a part in overload resolution. If two overload
490 candidates are otherwise equally good, then the overload with one or more
491 parameters with ``pass_object_size`` is preferred. This implies that the choice
492 between two identical overloads both with ``pass_object_size`` on one or more
493 parameters will always be ambiguous; for this reason, having two such overloads
494 is illegal. For example:
498 #define PS(N) __attribute__((pass_object_size(N)))
500 void Foo(char *a, char *b); // Overload A
501 // OK -- overload A has no parameters with pass_object_size.
502 void Foo(char *a PS(0), char *b PS(0)); // Overload B
503 // Error -- Same signature (sans pass_object_size) as overload B, and both
504 // overloads have one or more parameters with the pass_object_size attribute.
505 void Foo(void *a PS(0), void *b);
508 void Bar(void *a PS(0)); // Overload C
510 void Bar(char *c PS(1)); // Overload D
513 char known[10], *unknown;
514 Foo(unknown, unknown); // Calls overload B
515 Foo(known, unknown); // Calls overload B
516 Foo(unknown, known); // Calls overload B
517 Foo(known, known); // Calls overload B
519 Bar(known); // Calls overload D
520 Bar(unknown); // Calls overload D
523 Currently, ``pass_object_size`` is a bit restricted in terms of its usage:
525 * Only one use of ``pass_object_size`` is allowed per parameter.
527 * It is an error to take the address of a function with ``pass_object_size`` on
528 any of its parameters. If you wish to do this, you can create an overload
529 without ``pass_object_size`` on any parameters.
531 * It is an error to apply the ``pass_object_size`` attribute to parameters that
532 are not pointers. Additionally, any parameter that ``pass_object_size`` is
533 applied to must be marked ``const`` at its function's definition.
537 def OverloadableDocs : Documentation {
538 let Category = DocCatFunction;
540 Clang provides support for C++ function overloading in C. Function overloading
541 in C is introduced using the ``overloadable`` attribute. For example, one
542 might provide several overloaded versions of a ``tgsin`` function that invokes
543 the appropriate standard function computing the sine of a value with ``float``,
544 ``double``, or ``long double`` precision:
549 float __attribute__((overloadable)) tgsin(float x) { return sinf(x); }
550 double __attribute__((overloadable)) tgsin(double x) { return sin(x); }
551 long double __attribute__((overloadable)) tgsin(long double x) { return sinl(x); }
553 Given these declarations, one can call ``tgsin`` with a ``float`` value to
554 receive a ``float`` result, with a ``double`` to receive a ``double`` result,
555 etc. Function overloading in C follows the rules of C++ function overloading
556 to pick the best overload given the call arguments, with a few C-specific
559 * Conversion from ``float`` or ``double`` to ``long double`` is ranked as a
560 floating-point promotion (per C99) rather than as a floating-point conversion
563 * A conversion from a pointer of type ``T*`` to a pointer of type ``U*`` is
564 considered a pointer conversion (with conversion rank) if ``T`` and ``U`` are
567 * A conversion from type ``T`` to a value of type ``U`` is permitted if ``T``
568 and ``U`` are compatible types. This conversion is given "conversion" rank.
570 * If no viable candidates are otherwise available, we allow a conversion from a
571 pointer of type ``T*`` to a pointer of type ``U*``, where ``T`` and ``U`` are
572 incompatible. This conversion is ranked below all other types of conversions.
573 Please note: ``U`` lacking qualifiers that are present on ``T`` is sufficient
574 for ``T`` and ``U`` to be incompatible.
576 The declaration of ``overloadable`` functions is restricted to function
577 declarations and definitions. Most importantly, if any function with a given
578 name is given the ``overloadable`` attribute, then all function declarations
579 and definitions with that name (and in that scope) must have the
580 ``overloadable`` attribute. This rule even applies to redeclarations of
581 functions whose original declaration had the ``overloadable`` attribute, e.g.,
585 int f(int) __attribute__((overloadable));
586 float f(float); // error: declaration of "f" must have the "overloadable" attribute
588 int g(int) __attribute__((overloadable));
589 int g(int) { } // error: redeclaration of "g" must also have the "overloadable" attribute
591 Functions marked ``overloadable`` must have prototypes. Therefore, the
592 following code is ill-formed:
596 int h() __attribute__((overloadable)); // error: h does not have a prototype
598 However, ``overloadable`` functions are allowed to use a ellipsis even if there
599 are no named parameters (as is permitted in C++). This feature is particularly
600 useful when combined with the ``unavailable`` attribute:
604 void honeypot(...) __attribute__((overloadable, unavailable)); // calling me is an error
606 Functions declared with the ``overloadable`` attribute have their names mangled
607 according to the same rules as C++ function names. For example, the three
608 ``tgsin`` functions in our motivating example get the mangled names
609 ``_Z5tgsinf``, ``_Z5tgsind``, and ``_Z5tgsine``, respectively. There are two
610 caveats to this use of name mangling:
612 * Future versions of Clang may change the name mangling of functions overloaded
613 in C, so you should not depend on an specific mangling. To be completely
614 safe, we strongly urge the use of ``static inline`` with ``overloadable``
617 * The ``overloadable`` attribute has almost no meaning when used in C++,
618 because names will already be mangled and functions are already overloadable.
619 However, when an ``overloadable`` function occurs within an ``extern "C"``
620 linkage specification, it's name *will* be mangled in the same way as it
623 Query for this feature with ``__has_extension(attribute_overloadable)``.
627 def ObjCMethodFamilyDocs : Documentation {
628 let Category = DocCatFunction;
630 Many methods in Objective-C have conventional meanings determined by their
631 selectors. It is sometimes useful to be able to mark a method as having a
632 particular conventional meaning despite not having the right selector, or as
633 not having the conventional meaning that its selector would suggest. For these
634 use cases, we provide an attribute to specifically describe the "method family"
635 that a method belongs to.
637 **Usage**: ``__attribute__((objc_method_family(X)))``, where ``X`` is one of
638 ``none``, ``alloc``, ``copy``, ``init``, ``mutableCopy``, or ``new``. This
639 attribute can only be placed at the end of a method declaration:
643 - (NSString *)initMyStringValue __attribute__((objc_method_family(none)));
645 Users who do not wish to change the conventional meaning of a method, and who
646 merely want to document its non-standard retain and release semantics, should
647 use the retaining behavior attributes (``ns_returns_retained``,
648 ``ns_returns_not_retained``, etc).
650 Query for this feature with ``__has_attribute(objc_method_family)``.
654 def NoDebugDocs : Documentation {
655 let Category = DocCatVariable;
657 The ``nodebug`` attribute allows you to suppress debugging information for a
658 function or method, or for a variable that is not a parameter or a non-static
663 def NoDuplicateDocs : Documentation {
664 let Category = DocCatFunction;
666 The ``noduplicate`` attribute can be placed on function declarations to control
667 whether function calls to this function can be duplicated or not as a result of
668 optimizations. This is required for the implementation of functions with
669 certain special requirements, like the OpenCL "barrier" function, that might
670 need to be run concurrently by all the threads that are executing in lockstep
671 on the hardware. For example this attribute applied on the function
672 "nodupfunc" in the code below avoids that:
676 void nodupfunc() __attribute__((noduplicate));
677 // Setting it as a C++11 attribute is also valid
678 // void nodupfunc() [[clang::noduplicate]];
689 gets possibly modified by some optimizations into code similar to this:
701 where the call to "nodupfunc" is duplicated and sunk into the two branches
706 def ConvergentDocs : Documentation {
707 let Category = DocCatFunction;
709 The ``convergent`` attribute can be placed on a function declaration. It is
710 translated into the LLVM ``convergent`` attribute, which indicates that the call
711 instructions of a function with this attribute cannot be made control-dependent
712 on any additional values.
714 In languages designed for SPMD/SIMT programming model, e.g. OpenCL or CUDA,
715 the call instructions of a function with this attribute must be executed by
716 all work items or threads in a work group or sub group.
718 This attribute is different from ``noduplicate`` because it allows duplicating
719 function calls if it can be proved that the duplicated function calls are
720 not made control-dependent on any additional values, e.g., unrolling a loop
721 executed by all work items.
726 void convfunc(void) __attribute__((convergent));
727 // Setting it as a C++11 attribute is also valid in a C++ program.
728 // void convfunc(void) [[clang::convergent]];
733 def NoSplitStackDocs : Documentation {
734 let Category = DocCatFunction;
736 The ``no_split_stack`` attribute disables the emission of the split stack
737 preamble for a particular function. It has no effect if ``-fsplit-stack``
742 def ObjCRequiresSuperDocs : Documentation {
743 let Category = DocCatFunction;
745 Some Objective-C classes allow a subclass to override a particular method in a
746 parent class but expect that the overriding method also calls the overridden
747 method in the parent class. For these cases, we provide an attribute to
748 designate that a method requires a "call to ``super``" in the overriding
749 method in the subclass.
751 **Usage**: ``__attribute__((objc_requires_super))``. This attribute can only
752 be placed at the end of a method declaration:
756 - (void)foo __attribute__((objc_requires_super));
758 This attribute can only be applied the method declarations within a class, and
759 not a protocol. Currently this attribute does not enforce any placement of
760 where the call occurs in the overriding method (such as in the case of
761 ``-dealloc`` where the call must appear at the end). It checks only that it
764 Note that on both OS X and iOS that the Foundation framework provides a
765 convenience macro ``NS_REQUIRES_SUPER`` that provides syntactic sugar for this
770 - (void)foo NS_REQUIRES_SUPER;
772 This macro is conditionally defined depending on the compiler's support for
773 this attribute. If the compiler does not support the attribute the macro
776 Operationally, when a method has this annotation the compiler will warn if the
777 implementation of an override in a subclass does not call super. For example:
781 warning: method possibly missing a [super AnnotMeth] call
782 - (void) AnnotMeth{};
787 def ObjCRuntimeNameDocs : Documentation {
788 let Category = DocCatFunction;
790 By default, the Objective-C interface or protocol identifier is used
791 in the metadata name for that object. The `objc_runtime_name`
792 attribute allows annotated interfaces or protocols to use the
793 specified string argument in the object's metadata name instead of the
796 **Usage**: ``__attribute__((objc_runtime_name("MyLocalName")))``. This attribute
797 can only be placed before an @protocol or @interface declaration:
801 __attribute__((objc_runtime_name("MyLocalName")))
808 def ObjCRuntimeVisibleDocs : Documentation {
809 let Category = DocCatFunction;
811 This attribute specifies that the Objective-C class to which it applies is visible to the Objective-C runtime but not to the linker. Classes annotated with this attribute cannot be subclassed and cannot have categories defined for them.
815 def ObjCBoxableDocs : Documentation {
816 let Category = DocCatFunction;
818 Structs and unions marked with the ``objc_boxable`` attribute can be used
819 with the Objective-C boxed expression syntax, ``@(...)``.
821 **Usage**: ``__attribute__((objc_boxable))``. This attribute
822 can only be placed on a declaration of a trivially-copyable struct or union:
826 struct __attribute__((objc_boxable)) some_struct {
829 union __attribute__((objc_boxable)) some_union {
833 typedef struct __attribute__((objc_boxable)) _some_struct some_struct;
838 NSValue *boxed = @(ss);
843 def AvailabilityDocs : Documentation {
844 let Category = DocCatFunction;
846 The ``availability`` attribute can be placed on declarations to describe the
847 lifecycle of that declaration relative to operating system versions. Consider
848 the function declaration for a hypothetical function ``f``:
852 void f(void) __attribute__((availability(macos,introduced=10.4,deprecated=10.6,obsoleted=10.7)));
854 The availability attribute states that ``f`` was introduced in Mac OS X 10.4,
855 deprecated in Mac OS X 10.6, and obsoleted in Mac OS X 10.7. This information
856 is used by Clang to determine when it is safe to use ``f``: for example, if
857 Clang is instructed to compile code for Mac OS X 10.5, a call to ``f()``
858 succeeds. If Clang is instructed to compile code for Mac OS X 10.6, the call
859 succeeds but Clang emits a warning specifying that the function is deprecated.
860 Finally, if Clang is instructed to compile code for Mac OS X 10.7, the call
861 fails because ``f()`` is no longer available.
863 The availability attribute is a comma-separated list starting with the
864 platform name and then including clauses specifying important milestones in the
865 declaration's lifetime (in any order) along with additional information. Those
868 introduced=\ *version*
869 The first version in which this declaration was introduced.
871 deprecated=\ *version*
872 The first version in which this declaration was deprecated, meaning that
873 users should migrate away from this API.
875 obsoleted=\ *version*
876 The first version in which this declaration was obsoleted, meaning that it
877 was removed completely and can no longer be used.
880 This declaration is never available on this platform.
882 message=\ *string-literal*
883 Additional message text that Clang will provide when emitting a warning or
884 error about use of a deprecated or obsoleted declaration. Useful to direct
885 users to replacement APIs.
887 replacement=\ *string-literal*
888 Additional message text that Clang will use to provide Fix-It when emitting
889 a warning about use of a deprecated declaration. The Fix-It will replace
890 the deprecated declaration with the new declaration specified.
892 Multiple availability attributes can be placed on a declaration, which may
893 correspond to different platforms. Only the availability attribute with the
894 platform corresponding to the target platform will be used; any others will be
895 ignored. If no availability attribute specifies availability for the current
896 target platform, the availability attributes are ignored. Supported platforms
900 Apple's iOS operating system. The minimum deployment target is specified by
901 the ``-mios-version-min=*version*`` or ``-miphoneos-version-min=*version*``
902 command-line arguments.
905 Apple's Mac OS X operating system. The minimum deployment target is
906 specified by the ``-mmacosx-version-min=*version*`` command-line argument.
907 ``macosx`` is supported for backward-compatibility reasons, but it is
911 Apple's tvOS operating system. The minimum deployment target is specified by
912 the ``-mtvos-version-min=*version*`` command-line argument.
915 Apple's watchOS operating system. The minimum deployment target is specified by
916 the ``-mwatchos-version-min=*version*`` command-line argument.
918 A declaration can typically be used even when deploying back to a platform
919 version prior to when the declaration was introduced. When this happens, the
920 declaration is `weakly linked
921 <https://developer.apple.com/library/mac/#documentation/MacOSX/Conceptual/BPFrameworks/Concepts/WeakLinking.html>`_,
922 as if the ``weak_import`` attribute were added to the declaration. A
923 weakly-linked declaration may or may not be present a run-time, and a program
924 can determine whether the declaration is present by checking whether the
925 address of that declaration is non-NULL.
927 The flag ``strict`` disallows using API when deploying back to a
928 platform version prior to when the declaration was introduced. An
929 attempt to use such API before its introduction causes a hard error.
930 Weakly-linking is almost always a better API choice, since it allows
931 users to query availability at runtime.
933 If there are multiple declarations of the same entity, the availability
934 attributes must either match on a per-platform basis or later
935 declarations must not have availability attributes for that
936 platform. For example:
940 void g(void) __attribute__((availability(macos,introduced=10.4)));
941 void g(void) __attribute__((availability(macos,introduced=10.4))); // okay, matches
942 void g(void) __attribute__((availability(ios,introduced=4.0))); // okay, adds a new platform
943 void g(void); // okay, inherits both macos and ios availability from above.
944 void g(void) __attribute__((availability(macos,introduced=10.5))); // error: mismatch
946 When one method overrides another, the overriding method can be more widely available than the overridden method, e.g.,:
951 - (id)method __attribute__((availability(macos,introduced=10.4)));
952 - (id)method2 __attribute__((availability(macos,introduced=10.4)));
956 - (id)method __attribute__((availability(macos,introduced=10.3))); // okay: method moved into base class later
957 - (id)method __attribute__((availability(macos,introduced=10.5))); // error: this method was available via the base class in 10.4
963 def RequireConstantInitDocs : Documentation {
964 let Category = DocCatVariable;
966 This attribute specifies that the variable to which it is attached is intended
967 to have a `constant initializer <http://en.cppreference.com/w/cpp/language/constant_initialization>`_
968 according to the rules of [basic.start.static]. The variable is required to
969 have static or thread storage duration. If the initialization of the variable
970 is not a constant initializer an error will be produced. This attribute may
973 Note that in C++03 strict constant expression checking is not done. Instead
974 the attribute reports if Clang can emit the variable as a constant, even if it's
975 not technically a 'constant initializer'. This behavior is non-portable.
977 Static storage duration variables with constant initializers avoid hard-to-find
978 bugs caused by the indeterminate order of dynamic initialization. They can also
979 be safely used during dynamic initialization across translation units.
981 This attribute acts as a compile time assertion that the requirements
982 for constant initialization have been met. Since these requirements change
983 between dialects and have subtle pitfalls it's important to fail fast instead
984 of silently falling back on dynamic initialization.
989 #define SAFE_STATIC [[clang::require_constant_initialization]]
994 SAFE_STATIC T x = {42}; // Initialization OK. Doesn't check destructor.
995 SAFE_STATIC T y = 42; // error: variable does not have a constant initializer
996 // copy initialization is not a constant expression on a non-literal type.
1000 def WarnMaybeUnusedDocs : Documentation {
1001 let Category = DocCatVariable;
1002 let Heading = "maybe_unused, unused, gnu::unused";
1004 When passing the ``-Wunused`` flag to Clang, entities that are unused by the
1005 program may be diagnosed. The ``[[maybe_unused]]`` (or
1006 ``__attribute__((unused))``) attribute can be used to silence such diagnostics
1007 when the entity cannot be removed. For instance, a local variable may exist
1008 solely for use in an ``assert()`` statement, which makes the local variable
1009 unused when ``NDEBUG`` is defined.
1011 The attribute may be applied to the declaration of a class, a typedef, a
1012 variable, a function or method, a function parameter, an enumeration, an
1013 enumerator, a non-static data member, or a label.
1018 [[maybe_unused]] void f([[maybe_unused]] bool thing1,
1019 [[maybe_unused]] bool thing2) {
1020 [[maybe_unused]] bool b = thing1 && thing2;
1026 def WarnUnusedResultsDocs : Documentation {
1027 let Category = DocCatFunction;
1028 let Heading = "nodiscard, warn_unused_result, clang::warn_unused_result, gnu::warn_unused_result";
1030 Clang supports the ability to diagnose when the results of a function call
1031 expression are discarded under suspicious circumstances. A diagnostic is
1032 generated when a function or its return type is marked with ``[[nodiscard]]``
1033 (or ``__attribute__((warn_unused_result))``) and the function call appears as a
1034 potentially-evaluated discarded-value expression that is not explicitly cast to
1038 struct [[nodiscard]] error_info { /*...*/ };
1039 error_info enable_missile_safety_mode();
1041 void launch_missiles();
1042 void test_missiles() {
1043 enable_missile_safety_mode(); // diagnoses
1047 void f() { foo(); } // Does not diagnose, error_info is a reference.
1051 def FallthroughDocs : Documentation {
1052 let Category = DocCatStmt;
1053 let Heading = "fallthrough, clang::fallthrough";
1055 The ``fallthrough`` (or ``clang::fallthrough``) attribute is used
1056 to annotate intentional fall-through
1057 between switch labels. It can only be applied to a null statement placed at a
1058 point of execution between any statement and the next switch label. It is
1059 common to mark these places with a specific comment, but this attribute is
1060 meant to replace comments with a more strict annotation, which can be checked
1061 by the compiler. This attribute doesn't change semantics of the code and can
1062 be used wherever an intended fall-through occurs. It is designed to mimic
1063 control-flow statements like ``break;``, so it can be placed in most places
1064 where ``break;`` can, but only if there are no statements on the execution path
1065 between it and the next switch label.
1067 By default, Clang does not warn on unannotated fallthrough from one ``switch``
1068 case to another. Diagnostics on fallthrough without a corresponding annotation
1069 can be enabled with the ``-Wimplicit-fallthrough`` argument.
1075 // compile with -Wimplicit-fallthrough
1078 case 33: // no warning: no statements between case labels
1080 case 44: // warning: unannotated fall-through
1082 [[clang::fallthrough]];
1083 case 55: // no warning
1090 [[clang::fallthrough]];
1092 case 66: // no warning
1094 [[clang::fallthrough]]; // warning: fallthrough annotation does not
1095 // directly precede case label
1097 case 77: // warning: unannotated fall-through
1103 def ARMInterruptDocs : Documentation {
1104 let Category = DocCatFunction;
1106 Clang supports the GNU style ``__attribute__((interrupt("TYPE")))`` attribute on
1107 ARM targets. This attribute may be attached to a function definition and
1108 instructs the backend to generate appropriate function entry/exit code so that
1109 it can be used directly as an interrupt service routine.
1111 The parameter passed to the interrupt attribute is optional, but if
1112 provided it must be a string literal with one of the following values: "IRQ",
1113 "FIQ", "SWI", "ABORT", "UNDEF".
1115 The semantics are as follows:
1117 - If the function is AAPCS, Clang instructs the backend to realign the stack to
1118 8 bytes on entry. This is a general requirement of the AAPCS at public
1119 interfaces, but may not hold when an exception is taken. Doing this allows
1120 other AAPCS functions to be called.
1121 - If the CPU is M-class this is all that needs to be done since the architecture
1122 itself is designed in such a way that functions obeying the normal AAPCS ABI
1123 constraints are valid exception handlers.
1124 - If the CPU is not M-class, the prologue and epilogue are modified to save all
1125 non-banked registers that are used, so that upon return the user-mode state
1126 will not be corrupted. Note that to avoid unnecessary overhead, only
1127 general-purpose (integer) registers are saved in this way. If VFP operations
1128 are needed, that state must be saved manually.
1130 Specifically, interrupt kinds other than "FIQ" will save all core registers
1131 except "lr" and "sp". "FIQ" interrupts will save r0-r7.
1132 - If the CPU is not M-class, the return instruction is changed to one of the
1133 canonical sequences permitted by the architecture for exception return. Where
1134 possible the function itself will make the necessary "lr" adjustments so that
1135 the "preferred return address" is selected.
1137 Unfortunately the compiler is unable to make this guarantee for an "UNDEF"
1138 handler, where the offset from "lr" to the preferred return address depends on
1139 the execution state of the code which generated the exception. In this case
1140 a sequence equivalent to "movs pc, lr" will be used.
1144 def MipsInterruptDocs : Documentation {
1145 let Category = DocCatFunction;
1147 Clang supports the GNU style ``__attribute__((interrupt("ARGUMENT")))`` attribute on
1148 MIPS targets. This attribute may be attached to a function definition and instructs
1149 the backend to generate appropriate function entry/exit code so that it can be used
1150 directly as an interrupt service routine.
1152 By default, the compiler will produce a function prologue and epilogue suitable for
1153 an interrupt service routine that handles an External Interrupt Controller (eic)
1154 generated interrupt. This behaviour can be explicitly requested with the "eic"
1157 Otherwise, for use with vectored interrupt mode, the argument passed should be
1158 of the form "vector=LEVEL" where LEVEL is one of the following values:
1159 "sw0", "sw1", "hw0", "hw1", "hw2", "hw3", "hw4", "hw5". The compiler will
1160 then set the interrupt mask to the corresponding level which will mask all
1161 interrupts up to and including the argument.
1163 The semantics are as follows:
1165 - The prologue is modified so that the Exception Program Counter (EPC) and
1166 Status coprocessor registers are saved to the stack. The interrupt mask is
1167 set so that the function can only be interrupted by a higher priority
1168 interrupt. The epilogue will restore the previous values of EPC and Status.
1170 - The prologue and epilogue are modified to save and restore all non-kernel
1171 registers as necessary.
1173 - The FPU is disabled in the prologue, as the floating pointer registers are not
1174 spilled to the stack.
1176 - The function return sequence is changed to use an exception return instruction.
1178 - The parameter sets the interrupt mask for the function corresponding to the
1179 interrupt level specified. If no mask is specified the interrupt mask
1184 def TargetDocs : Documentation {
1185 let Category = DocCatFunction;
1187 Clang supports the GNU style ``__attribute__((target("OPTIONS")))`` attribute.
1188 This attribute may be attached to a function definition and instructs
1189 the backend to use different code generation options than were passed on the
1192 The current set of options correspond to the existing "subtarget features" for
1193 the target with or without a "-mno-" in front corresponding to the absence
1194 of the feature, as well as ``arch="CPU"`` which will change the default "CPU"
1197 Example "subtarget features" from the x86 backend include: "mmx", "sse", "sse4.2",
1198 "avx", "xop" and largely correspond to the machine specific options handled by
1203 def DocCatAMDGPUAttributes : DocumentationCategory<"AMD GPU Attributes">;
1205 def AMDGPUFlatWorkGroupSizeDocs : Documentation {
1206 let Category = DocCatAMDGPUAttributes;
1208 The flat work-group size is the number of work-items in the work-group size
1209 specified when the kernel is dispatched. It is the product of the sizes of the
1210 x, y, and z dimension of the work-group.
1213 ``__attribute__((amdgpu_flat_work_group_size(<min>, <max>)))`` attribute for the
1214 AMDGPU target. This attribute may be attached to a kernel function definition
1215 and is an optimization hint.
1217 ``<min>`` parameter specifies the minimum flat work-group size, and ``<max>``
1218 parameter specifies the maximum flat work-group size (must be greater than
1219 ``<min>``) to which all dispatches of the kernel will conform. Passing ``0, 0``
1220 as ``<min>, <max>`` implies the default behavior (``128, 256``).
1222 If specified, the AMDGPU target backend might be able to produce better machine
1223 code for barriers and perform scratch promotion by estimating available group
1226 An error will be given if:
1227 - Specified values violate subtarget specifications;
1228 - Specified values are not compatible with values provided through other
1233 def AMDGPUWavesPerEUDocs : Documentation {
1234 let Category = DocCatAMDGPUAttributes;
1236 A compute unit (CU) is responsible for executing the wavefronts of a work-group.
1237 It is composed of one or more execution units (EU), which are responsible for
1238 executing the wavefronts. An EU can have enough resources to maintain the state
1239 of more than one executing wavefront. This allows an EU to hide latency by
1240 switching between wavefronts in a similar way to symmetric multithreading on a
1241 CPU. In order to allow the state for multiple wavefronts to fit on an EU, the
1242 resources used by a single wavefront have to be limited. For example, the number
1243 of SGPRs and VGPRs. Limiting such resources can allow greater latency hiding,
1244 but can result in having to spill some register state to memory.
1246 Clang supports the ``__attribute__((amdgpu_waves_per_eu(<min>[, <max>])))``
1247 attribute for the AMDGPU target. This attribute may be attached to a kernel
1248 function definition and is an optimization hint.
1250 ``<min>`` parameter specifies the requested minimum number of waves per EU, and
1251 *optional* ``<max>`` parameter specifies the requested maximum number of waves
1252 per EU (must be greater than ``<min>`` if specified). If ``<max>`` is omitted,
1253 then there is no restriction on the maximum number of waves per EU other than
1254 the one dictated by the hardware for which the kernel is compiled. Passing
1255 ``0, 0`` as ``<min>, <max>`` implies the default behavior (no limits).
1257 If specified, this attribute allows an advanced developer to tune the number of
1258 wavefronts that are capable of fitting within the resources of an EU. The AMDGPU
1259 target backend can use this information to limit resources, such as number of
1260 SGPRs, number of VGPRs, size of available group and private memory segments, in
1261 such a way that guarantees that at least ``<min>`` wavefronts and at most
1262 ``<max>`` wavefronts are able to fit within the resources of an EU. Requesting
1263 more wavefronts can hide memory latency but limits available registers which
1264 can result in spilling. Requesting fewer wavefronts can help reduce cache
1265 thrashing, but can reduce memory latency hiding.
1267 This attribute controls the machine code generated by the AMDGPU target backend
1268 to ensure it is capable of meeting the requested values. However, when the
1269 kernel is executed, there may be other reasons that prevent meeting the request,
1270 for example, there may be wavefronts from other kernels executing on the EU.
1272 An error will be given if:
1273 - Specified values violate subtarget specifications;
1274 - Specified values are not compatible with values provided through other
1276 - The AMDGPU target backend is unable to create machine code that can meet the
1281 def AMDGPUNumSGPRNumVGPRDocs : Documentation {
1282 let Category = DocCatAMDGPUAttributes;
1284 Clang supports the ``__attribute__((amdgpu_num_sgpr(<num_sgpr>)))`` and
1285 ``__attribute__((amdgpu_num_vgpr(<num_vgpr>)))`` attributes for the AMDGPU
1286 target. These attributes may be attached to a kernel function definition and are
1287 an optimization hint.
1289 If these attributes are specified, then the AMDGPU target backend will attempt
1290 to limit the number of SGPRs and/or VGPRs used to the specified value(s). The
1291 number of used SGPRs and/or VGPRs may further be rounded up to satisfy the
1292 allocation requirements or constraints of the subtarget. Passing ``0`` as
1293 ``num_sgpr`` and/or ``num_vgpr`` implies the default behavior (no limits).
1295 These attributes can be used to test the AMDGPU target backend. It is
1296 recommended that the ``amdgpu_waves_per_eu`` attribute be used to control
1297 resources such as SGPRs and VGPRs since it is aware of the limits for different
1300 An error will be given if:
1301 - Specified values violate subtarget specifications;
1302 - Specified values are not compatible with values provided through other
1304 - The AMDGPU target backend is unable to create machine code that can meet the
1309 def DocCatCallingConvs : DocumentationCategory<"Calling Conventions"> {
1311 Clang supports several different calling conventions, depending on the target
1312 platform and architecture. The calling convention used for a function determines
1313 how parameters are passed, how results are returned to the caller, and other
1314 low-level details of calling a function.
1318 def PcsDocs : Documentation {
1319 let Category = DocCatCallingConvs;
1321 On ARM targets, this attribute can be used to select calling conventions
1322 similar to ``stdcall`` on x86. Valid parameter values are "aapcs" and
1327 def RegparmDocs : Documentation {
1328 let Category = DocCatCallingConvs;
1330 On 32-bit x86 targets, the regparm attribute causes the compiler to pass
1331 the first three integer parameters in EAX, EDX, and ECX instead of on the
1332 stack. This attribute has no effect on variadic functions, and all parameters
1333 are passed via the stack as normal.
1337 def SysVABIDocs : Documentation {
1338 let Category = DocCatCallingConvs;
1340 On Windows x86_64 targets, this attribute changes the calling convention of a
1341 function to match the default convention used on Sys V targets such as Linux,
1342 Mac, and BSD. This attribute has no effect on other targets.
1346 def MSABIDocs : Documentation {
1347 let Category = DocCatCallingConvs;
1349 On non-Windows x86_64 targets, this attribute changes the calling convention of
1350 a function to match the default convention used on Windows x86_64. This
1351 attribute has no effect on Windows targets or non-x86_64 targets.
1355 def StdCallDocs : Documentation {
1356 let Category = DocCatCallingConvs;
1358 On 32-bit x86 targets, this attribute changes the calling convention of a
1359 function to clear parameters off of the stack on return. This convention does
1360 not support variadic calls or unprototyped functions in C, and has no effect on
1361 x86_64 targets. This calling convention is used widely by the Windows API and
1362 COM applications. See the documentation for `__stdcall`_ on MSDN.
1364 .. _`__stdcall`: http://msdn.microsoft.com/en-us/library/zxk0tw93.aspx
1368 def FastCallDocs : Documentation {
1369 let Category = DocCatCallingConvs;
1371 On 32-bit x86 targets, this attribute changes the calling convention of a
1372 function to use ECX and EDX as register parameters and clear parameters off of
1373 the stack on return. This convention does not support variadic calls or
1374 unprototyped functions in C, and has no effect on x86_64 targets. This calling
1375 convention is supported primarily for compatibility with existing code. Users
1376 seeking register parameters should use the ``regparm`` attribute, which does
1377 not require callee-cleanup. See the documentation for `__fastcall`_ on MSDN.
1379 .. _`__fastcall`: http://msdn.microsoft.com/en-us/library/6xa169sk.aspx
1383 def RegCallDocs : Documentation {
1384 let Category = DocCatCallingConvs;
1386 On x86 targets, this attribute changes the calling convention to
1387 `__regcall`_ convention. This convention aims to pass as many arguments
1388 as possible in registers. It also tries to utilize registers for the
1389 return value whenever it is possible.
1391 .. _`__regcall`: https://software.intel.com/en-us/node/693069
1395 def ThisCallDocs : Documentation {
1396 let Category = DocCatCallingConvs;
1398 On 32-bit x86 targets, this attribute changes the calling convention of a
1399 function to use ECX for the first parameter (typically the implicit ``this``
1400 parameter of C++ methods) and clear parameters off of the stack on return. This
1401 convention does not support variadic calls or unprototyped functions in C, and
1402 has no effect on x86_64 targets. See the documentation for `__thiscall`_ on
1405 .. _`__thiscall`: http://msdn.microsoft.com/en-us/library/ek8tkfbw.aspx
1409 def VectorCallDocs : Documentation {
1410 let Category = DocCatCallingConvs;
1412 On 32-bit x86 *and* x86_64 targets, this attribute changes the calling
1413 convention of a function to pass vector parameters in SSE registers.
1415 On 32-bit x86 targets, this calling convention is similar to ``__fastcall``.
1416 The first two integer parameters are passed in ECX and EDX. Subsequent integer
1417 parameters are passed in memory, and callee clears the stack. On x86_64
1418 targets, the callee does *not* clear the stack, and integer parameters are
1419 passed in RCX, RDX, R8, and R9 as is done for the default Windows x64 calling
1422 On both 32-bit x86 and x86_64 targets, vector and floating point arguments are
1423 passed in XMM0-XMM5. Homogeneous vector aggregates of up to four elements are
1424 passed in sequential SSE registers if enough are available. If AVX is enabled,
1425 256 bit vectors are passed in YMM0-YMM5. Any vector or aggregate type that
1426 cannot be passed in registers for any reason is passed by reference, which
1427 allows the caller to align the parameter memory.
1429 See the documentation for `__vectorcall`_ on MSDN for more details.
1431 .. _`__vectorcall`: http://msdn.microsoft.com/en-us/library/dn375768.aspx
1435 def DocCatConsumed : DocumentationCategory<"Consumed Annotation Checking"> {
1437 Clang supports additional attributes for checking basic resource management
1438 properties, specifically for unique objects that have a single owning reference.
1439 The following attributes are currently supported, although **the implementation
1440 for these annotations is currently in development and are subject to change.**
1444 def SetTypestateDocs : Documentation {
1445 let Category = DocCatConsumed;
1447 Annotate methods that transition an object into a new state with
1448 ``__attribute__((set_typestate(new_state)))``. The new state must be
1449 unconsumed, consumed, or unknown.
1453 def CallableWhenDocs : Documentation {
1454 let Category = DocCatConsumed;
1456 Use ``__attribute__((callable_when(...)))`` to indicate what states a method
1457 may be called in. Valid states are unconsumed, consumed, or unknown. Each
1458 argument to this attribute must be a quoted string. E.g.:
1460 ``__attribute__((callable_when("unconsumed", "unknown")))``
1464 def TestTypestateDocs : Documentation {
1465 let Category = DocCatConsumed;
1467 Use ``__attribute__((test_typestate(tested_state)))`` to indicate that a method
1468 returns true if the object is in the specified state..
1472 def ParamTypestateDocs : Documentation {
1473 let Category = DocCatConsumed;
1475 This attribute specifies expectations about function parameters. Calls to an
1476 function with annotated parameters will issue a warning if the corresponding
1477 argument isn't in the expected state. The attribute is also used to set the
1478 initial state of the parameter when analyzing the function's body.
1482 def ReturnTypestateDocs : Documentation {
1483 let Category = DocCatConsumed;
1485 The ``return_typestate`` attribute can be applied to functions or parameters.
1486 When applied to a function the attribute specifies the state of the returned
1487 value. The function's body is checked to ensure that it always returns a value
1488 in the specified state. On the caller side, values returned by the annotated
1489 function are initialized to the given state.
1491 When applied to a function parameter it modifies the state of an argument after
1492 a call to the function returns. The function's body is checked to ensure that
1493 the parameter is in the expected state before returning.
1497 def ConsumableDocs : Documentation {
1498 let Category = DocCatConsumed;
1500 Each ``class`` that uses any of the typestate annotations must first be marked
1501 using the ``consumable`` attribute. Failure to do so will result in a warning.
1503 This attribute accepts a single parameter that must be one of the following:
1504 ``unknown``, ``consumed``, or ``unconsumed``.
1508 def NoSanitizeDocs : Documentation {
1509 let Category = DocCatFunction;
1511 Use the ``no_sanitize`` attribute on a function declaration to specify
1512 that a particular instrumentation or set of instrumentations should not be
1513 applied to that function. The attribute takes a list of string literals,
1514 which have the same meaning as values accepted by the ``-fno-sanitize=``
1515 flag. For example, ``__attribute__((no_sanitize("address", "thread")))``
1516 specifies that AddressSanitizer and ThreadSanitizer should not be applied
1519 See :ref:`Controlling Code Generation <controlling-code-generation>` for a
1520 full list of supported sanitizer flags.
1524 def NoSanitizeAddressDocs : Documentation {
1525 let Category = DocCatFunction;
1526 // This function has multiple distinct spellings, and so it requires a custom
1527 // heading to be specified. The most common spelling is sufficient.
1528 let Heading = "no_sanitize_address (no_address_safety_analysis, gnu::no_address_safety_analysis, gnu::no_sanitize_address)";
1530 .. _langext-address_sanitizer:
1532 Use ``__attribute__((no_sanitize_address))`` on a function declaration to
1533 specify that address safety instrumentation (e.g. AddressSanitizer) should
1534 not be applied to that function.
1538 def NoSanitizeThreadDocs : Documentation {
1539 let Category = DocCatFunction;
1540 let Heading = "no_sanitize_thread";
1542 .. _langext-thread_sanitizer:
1544 Use ``__attribute__((no_sanitize_thread))`` on a function declaration to
1545 specify that checks for data races on plain (non-atomic) memory accesses should
1546 not be inserted by ThreadSanitizer. The function is still instrumented by the
1547 tool to avoid false positives and provide meaningful stack traces.
1551 def NoSanitizeMemoryDocs : Documentation {
1552 let Category = DocCatFunction;
1553 let Heading = "no_sanitize_memory";
1555 .. _langext-memory_sanitizer:
1557 Use ``__attribute__((no_sanitize_memory))`` on a function declaration to
1558 specify that checks for uninitialized memory should not be inserted
1559 (e.g. by MemorySanitizer). The function may still be instrumented by the tool
1560 to avoid false positives in other places.
1564 def DocCatTypeSafety : DocumentationCategory<"Type Safety Checking"> {
1566 Clang supports additional attributes to enable checking type safety properties
1567 that can't be enforced by the C type system. To see warnings produced by these
1568 checks, ensure that -Wtype-safety is enabled. Use cases include:
1570 * MPI library implementations, where these attributes enable checking that
1571 the buffer type matches the passed ``MPI_Datatype``;
1572 * for HDF5 library there is a similar use case to MPI;
1573 * checking types of variadic functions' arguments for functions like
1574 ``fcntl()`` and ``ioctl()``.
1576 You can detect support for these attributes with ``__has_attribute()``. For
1581 #if defined(__has_attribute)
1582 # if __has_attribute(argument_with_type_tag) && \
1583 __has_attribute(pointer_with_type_tag) && \
1584 __has_attribute(type_tag_for_datatype)
1585 # define ATTR_MPI_PWT(buffer_idx, type_idx) __attribute__((pointer_with_type_tag(mpi,buffer_idx,type_idx)))
1586 /* ... other macros ... */
1590 #if !defined(ATTR_MPI_PWT)
1591 # define ATTR_MPI_PWT(buffer_idx, type_idx)
1594 int MPI_Send(void *buf, int count, MPI_Datatype datatype /*, other args omitted */)
1599 def ArgumentWithTypeTagDocs : Documentation {
1600 let Category = DocCatTypeSafety;
1601 let Heading = "argument_with_type_tag";
1603 Use ``__attribute__((argument_with_type_tag(arg_kind, arg_idx,
1604 type_tag_idx)))`` on a function declaration to specify that the function
1605 accepts a type tag that determines the type of some other argument.
1607 This attribute is primarily useful for checking arguments of variadic functions
1608 (``pointer_with_type_tag`` can be used in most non-variadic cases).
1610 In the attribute prototype above:
1611 * ``arg_kind`` is an identifier that should be used when annotating all
1612 applicable type tags.
1613 * ``arg_idx`` provides the position of a function argument. The expected type of
1614 this function argument will be determined by the function argument specified
1615 by ``type_tag_idx``. In the code example below, "3" means that the type of the
1616 function's third argument will be determined by ``type_tag_idx``.
1617 * ``type_tag_idx`` provides the position of a function argument. This function
1618 argument will be a type tag. The type tag will determine the expected type of
1619 the argument specified by ``arg_idx``. In the code example below, "2" means
1620 that the type tag associated with the function's second argument should agree
1621 with the type of the argument specified by ``arg_idx``.
1627 int fcntl(int fd, int cmd, ...)
1628 __attribute__(( argument_with_type_tag(fcntl,3,2) ));
1629 // The function's second argument will be a type tag; this type tag will
1630 // determine the expected type of the function's third argument.
1634 def PointerWithTypeTagDocs : Documentation {
1635 let Category = DocCatTypeSafety;
1636 let Heading = "pointer_with_type_tag";
1638 Use ``__attribute__((pointer_with_type_tag(ptr_kind, ptr_idx, type_tag_idx)))``
1639 on a function declaration to specify that the function accepts a type tag that
1640 determines the pointee type of some other pointer argument.
1642 In the attribute prototype above:
1643 * ``ptr_kind`` is an identifier that should be used when annotating all
1644 applicable type tags.
1645 * ``ptr_idx`` provides the position of a function argument; this function
1646 argument will have a pointer type. The expected pointee type of this pointer
1647 type will be determined by the function argument specified by
1648 ``type_tag_idx``. In the code example below, "1" means that the pointee type
1649 of the function's first argument will be determined by ``type_tag_idx``.
1650 * ``type_tag_idx`` provides the position of a function argument; this function
1651 argument will be a type tag. The type tag will determine the expected pointee
1652 type of the pointer argument specified by ``ptr_idx``. In the code example
1653 below, "3" means that the type tag associated with the function's third
1654 argument should agree with the pointee type of the pointer argument specified
1661 typedef int MPI_Datatype;
1662 int MPI_Send(void *buf, int count, MPI_Datatype datatype /*, other args omitted */)
1663 __attribute__(( pointer_with_type_tag(mpi,1,3) ));
1664 // The function's 3rd argument will be a type tag; this type tag will
1665 // determine the expected pointee type of the function's 1st argument.
1669 def TypeTagForDatatypeDocs : Documentation {
1670 let Category = DocCatTypeSafety;
1672 When declaring a variable, use
1673 ``__attribute__((type_tag_for_datatype(kind, type)))`` to create a type tag that
1674 is tied to the ``type`` argument given to the attribute.
1676 In the attribute prototype above:
1677 * ``kind`` is an identifier that should be used when annotating all applicable
1679 * ``type`` indicates the name of the type.
1681 Clang supports annotating type tags of two forms.
1683 * **Type tag that is a reference to a declared identifier.**
1684 Use ``__attribute__((type_tag_for_datatype(kind, type)))`` when declaring that
1689 typedef int MPI_Datatype;
1690 extern struct mpi_datatype mpi_datatype_int
1691 __attribute__(( type_tag_for_datatype(mpi,int) ));
1692 #define MPI_INT ((MPI_Datatype) &mpi_datatype_int)
1693 // &mpi_datatype_int is a type tag. It is tied to type "int".
1695 * **Type tag that is an integral literal.**
1696 Declare a ``static const`` variable with an initializer value and attach
1697 ``__attribute__((type_tag_for_datatype(kind, type)))`` on that declaration:
1701 typedef int MPI_Datatype;
1702 static const MPI_Datatype mpi_datatype_int
1703 __attribute__(( type_tag_for_datatype(mpi,int) )) = 42;
1704 #define MPI_INT ((MPI_Datatype) 42)
1705 // The number 42 is a type tag. It is tied to type "int".
1708 The ``type_tag_for_datatype`` attribute also accepts an optional third argument
1709 that determines how the type of the function argument specified by either
1710 ``arg_idx`` or ``ptr_idx`` is compared against the type associated with the type
1711 tag. (Recall that for the ``argument_with_type_tag`` attribute, the type of the
1712 function argument specified by ``arg_idx`` is compared against the type
1713 associated with the type tag. Also recall that for the ``pointer_with_type_tag``
1714 attribute, the pointee type of the function argument specified by ``ptr_idx`` is
1715 compared against the type associated with the type tag.) There are two supported
1716 values for this optional third argument:
1718 * ``layout_compatible`` will cause types to be compared according to
1719 layout-compatibility rules (In C++11 [class.mem] p 17, 18, see the
1720 layout-compatibility rules for two standard-layout struct types and for two
1721 standard-layout union types). This is useful when creating a type tag
1722 associated with a struct or union type. For example:
1727 typedef int MPI_Datatype;
1728 struct internal_mpi_double_int { double d; int i; };
1729 extern struct mpi_datatype mpi_datatype_double_int
1730 __attribute__(( type_tag_for_datatype(mpi,
1731 struct internal_mpi_double_int, layout_compatible) ));
1733 #define MPI_DOUBLE_INT ((MPI_Datatype) &mpi_datatype_double_int)
1735 int MPI_Send(void *buf, int count, MPI_Datatype datatype, ...)
1736 __attribute__(( pointer_with_type_tag(mpi,1,3) ));
1739 struct my_pair { double a; int b; };
1740 struct my_pair *buffer;
1741 MPI_Send(buffer, 1, MPI_DOUBLE_INT /*, ... */); // no warning because the
1742 // layout of my_pair is
1743 // compatible with that of
1744 // internal_mpi_double_int
1746 struct my_int_pair { int a; int b; }
1747 struct my_int_pair *buffer2;
1748 MPI_Send(buffer2, 1, MPI_DOUBLE_INT /*, ... */); // warning because the
1749 // layout of my_int_pair
1750 // does not match that of
1751 // internal_mpi_double_int
1753 * ``must_be_null`` specifies that the function argument specified by either
1754 ``arg_idx`` (for the ``argument_with_type_tag`` attribute) or ``ptr_idx`` (for
1755 the ``pointer_with_type_tag`` attribute) should be a null pointer constant.
1756 The second argument to the ``type_tag_for_datatype`` attribute is ignored. For
1762 typedef int MPI_Datatype;
1763 extern struct mpi_datatype mpi_datatype_null
1764 __attribute__(( type_tag_for_datatype(mpi, void, must_be_null) ));
1766 #define MPI_DATATYPE_NULL ((MPI_Datatype) &mpi_datatype_null)
1767 int MPI_Send(void *buf, int count, MPI_Datatype datatype, ...)
1768 __attribute__(( pointer_with_type_tag(mpi,1,3) ));
1771 struct my_pair { double a; int b; };
1772 struct my_pair *buffer;
1773 MPI_Send(buffer, 1, MPI_DATATYPE_NULL /*, ... */); // warning: MPI_DATATYPE_NULL
1774 // was specified but buffer
1775 // is not a null pointer
1779 def FlattenDocs : Documentation {
1780 let Category = DocCatFunction;
1782 The ``flatten`` attribute causes calls within the attributed function to
1783 be inlined unless it is impossible to do so, for example if the body of the
1784 callee is unavailable or if the callee has the ``noinline`` attribute.
1788 def FormatDocs : Documentation {
1789 let Category = DocCatFunction;
1792 Clang supports the ``format`` attribute, which indicates that the function
1793 accepts a ``printf`` or ``scanf``-like format string and corresponding
1794 arguments or a ``va_list`` that contains these arguments.
1796 Please see `GCC documentation about format attribute
1797 <http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html>`_ to find details
1798 about attribute syntax.
1800 Clang implements two kinds of checks with this attribute.
1802 #. Clang checks that the function with the ``format`` attribute is called with
1803 a format string that uses format specifiers that are allowed, and that
1804 arguments match the format string. This is the ``-Wformat`` warning, it is
1807 #. Clang checks that the format string argument is a literal string. This is
1808 the ``-Wformat-nonliteral`` warning, it is off by default.
1810 Clang implements this mostly the same way as GCC, but there is a difference
1811 for functions that accept a ``va_list`` argument (for example, ``vprintf``).
1812 GCC does not emit ``-Wformat-nonliteral`` warning for calls to such
1813 functions. Clang does not warn if the format string comes from a function
1814 parameter, where the function is annotated with a compatible attribute,
1815 otherwise it warns. For example:
1819 __attribute__((__format__ (__scanf__, 1, 3)))
1820 void foo(const char* s, char *buf, ...) {
1824 vprintf(s, ap); // warning: format string is not a string literal
1827 In this case we warn because ``s`` contains a format string for a
1828 ``scanf``-like function, but it is passed to a ``printf``-like function.
1830 If the attribute is removed, clang still warns, because the format string is
1831 not a string literal.
1837 __attribute__((__format__ (__printf__, 1, 3)))
1838 void foo(const char* s, char *buf, ...) {
1842 vprintf(s, ap); // warning
1845 In this case Clang does not warn because the format string ``s`` and
1846 the corresponding arguments are annotated. If the arguments are
1847 incorrect, the caller of ``foo`` will receive a warning.
1851 def AlignValueDocs : Documentation {
1852 let Category = DocCatType;
1854 The align_value attribute can be added to the typedef of a pointer type or the
1855 declaration of a variable of pointer or reference type. It specifies that the
1856 pointer will point to, or the reference will bind to, only objects with at
1857 least the provided alignment. This alignment value must be some positive power
1862 typedef double * aligned_double_ptr __attribute__((align_value(64)));
1863 void foo(double & x __attribute__((align_value(128)),
1864 aligned_double_ptr y) { ... }
1866 If the pointer value does not have the specified alignment at runtime, the
1867 behavior of the program is undefined.
1871 def FlagEnumDocs : Documentation {
1872 let Category = DocCatType;
1874 This attribute can be added to an enumerator to signal to the compiler that it
1875 is intended to be used as a flag type. This will cause the compiler to assume
1876 that the range of the type includes all of the values that you can get by
1877 manipulating bits of the enumerator when issuing warnings.
1881 def EmptyBasesDocs : Documentation {
1882 let Category = DocCatType;
1884 The empty_bases attribute permits the compiler to utilize the
1885 empty-base-optimization more frequently.
1886 This attribute only applies to struct, class, and union types.
1887 It is only supported when using the Microsoft C++ ABI.
1891 def LayoutVersionDocs : Documentation {
1892 let Category = DocCatType;
1894 The layout_version attribute requests that the compiler utilize the class
1895 layout rules of a particular compiler version.
1896 This attribute only applies to struct, class, and union types.
1897 It is only supported when using the Microsoft C++ ABI.
1901 def MSInheritanceDocs : Documentation {
1902 let Category = DocCatType;
1903 let Heading = "__single_inhertiance, __multiple_inheritance, __virtual_inheritance";
1905 This collection of keywords is enabled under ``-fms-extensions`` and controls
1906 the pointer-to-member representation used on ``*-*-win32`` targets.
1908 The ``*-*-win32`` targets utilize a pointer-to-member representation which
1909 varies in size and alignment depending on the definition of the underlying
1912 However, this is problematic when a forward declaration is only available and
1913 no definition has been made yet. In such cases, Clang is forced to utilize the
1914 most general representation that is available to it.
1916 These keywords make it possible to use a pointer-to-member representation other
1917 than the most general one regardless of whether or not the definition will ever
1918 be present in the current translation unit.
1920 This family of keywords belong between the ``class-key`` and ``class-name``:
1924 struct __single_inheritance S;
1928 This keyword can be applied to class templates but only has an effect when used
1929 on full specializations:
1933 template <typename T, typename U> struct __single_inheritance A; // warning: inheritance model ignored on primary template
1934 template <typename T> struct __multiple_inheritance A<T, T>; // warning: inheritance model ignored on partial specialization
1935 template <> struct __single_inheritance A<int, float>;
1937 Note that choosing an inheritance model less general than strictly necessary is
1942 struct __multiple_inheritance S; // error: inheritance model does not match definition
1948 def MSNoVTableDocs : Documentation {
1949 let Category = DocCatType;
1951 This attribute can be added to a class declaration or definition to signal to
1952 the compiler that constructors and destructors will not reference the virtual
1953 function table. It is only supported when using the Microsoft C++ ABI.
1957 def OptnoneDocs : Documentation {
1958 let Category = DocCatFunction;
1960 The ``optnone`` attribute suppresses essentially all optimizations
1961 on a function or method, regardless of the optimization level applied to
1962 the compilation unit as a whole. This is particularly useful when you
1963 need to debug a particular function, but it is infeasible to build the
1964 entire application without optimization. Avoiding optimization on the
1965 specified function can improve the quality of the debugging information
1968 This attribute is incompatible with the ``always_inline`` and ``minsize``
1973 def LoopHintDocs : Documentation {
1974 let Category = DocCatStmt;
1975 let Heading = "#pragma clang loop";
1977 The ``#pragma clang loop`` directive allows loop optimization hints to be
1978 specified for the subsequent loop. The directive allows vectorization,
1979 interleaving, and unrolling to be enabled or disabled. Vector width as well
1980 as interleave and unrolling count can be manually specified. See
1981 `language extensions
1982 <http://clang.llvm.org/docs/LanguageExtensions.html#extensions-for-loop-hint-optimizations>`_
1987 def UnrollHintDocs : Documentation {
1988 let Category = DocCatStmt;
1989 let Heading = "#pragma unroll, #pragma nounroll";
1991 Loop unrolling optimization hints can be specified with ``#pragma unroll`` and
1992 ``#pragma nounroll``. The pragma is placed immediately before a for, while,
1993 do-while, or c++11 range-based for loop.
1995 Specifying ``#pragma unroll`` without a parameter directs the loop unroller to
1996 attempt to fully unroll the loop if the trip count is known at compile time and
1997 attempt to partially unroll the loop if the trip count is not known at compile
2007 Specifying the optional parameter, ``#pragma unroll _value_``, directs the
2008 unroller to unroll the loop ``_value_`` times. The parameter may optionally be
2009 enclosed in parentheses:
2023 Specifying ``#pragma nounroll`` indicates that the loop should not be unrolled:
2032 ``#pragma unroll`` and ``#pragma unroll _value_`` have identical semantics to
2033 ``#pragma clang loop unroll(full)`` and
2034 ``#pragma clang loop unroll_count(_value_)`` respectively. ``#pragma nounroll``
2035 is equivalent to ``#pragma clang loop unroll(disable)``. See
2036 `language extensions
2037 <http://clang.llvm.org/docs/LanguageExtensions.html#extensions-for-loop-hint-optimizations>`_
2038 for further details including limitations of the unroll hints.
2042 def OpenCLUnrollHintDocs : Documentation {
2043 let Category = DocCatStmt;
2044 let Heading = "__attribute__((opencl_unroll_hint))";
2046 The opencl_unroll_hint attribute qualifier can be used to specify that a loop
2047 (for, while and do loops) can be unrolled. This attribute qualifier can be
2048 used to specify full unrolling or partial unrolling by a specified amount.
2049 This is a compiler hint and the compiler may ignore this directive. See
2050 `OpenCL v2.0 <https://www.khronos.org/registry/cl/specs/opencl-2.0.pdf>`_
2051 s6.11.5 for details.
2055 def OpenCLAccessDocs : Documentation {
2056 let Category = DocCatStmt;
2057 let Heading = "__read_only, __write_only, __read_write (read_only, write_only, read_write)";
2059 The access qualifiers must be used with image object arguments or pipe arguments
2060 to declare if they are being read or written by a kernel or function.
2062 The read_only/__read_only, write_only/__write_only and read_write/__read_write
2063 names are reserved for use as access qualifiers and shall not be used otherwise.
2068 foo (read_only image2d_t imageA,
2069 write_only image2d_t imageB) {
2073 In the above example imageA is a read-only 2D image object, and imageB is a
2074 write-only 2D image object.
2076 The read_write (or __read_write) qualifier can not be used with pipe.
2078 More details can be found in the OpenCL C language Spec v2.0, Section 6.6.
2082 def DocOpenCLAddressSpaces : DocumentationCategory<"OpenCL Address Spaces"> {
2084 The address space qualifier may be used to specify the region of memory that is
2085 used to allocate the object. OpenCL supports the following address spaces:
2086 __generic(generic), __global(global), __local(local), __private(private),
2087 __constant(constant).
2091 __constant int c = ...;
2093 __generic int* foo(global int* g) {
2100 More details can be found in the OpenCL C language Spec v2.0, Section 6.5.
2104 def OpenCLAddressSpaceGenericDocs : Documentation {
2105 let Category = DocOpenCLAddressSpaces;
2107 The generic address space attribute is only available with OpenCL v2.0 and later.
2108 It can be used with pointer types. Variables in global and local scope and
2109 function parameters in non-kernel functions can have the generic address space
2110 type attribute. It is intended to be a placeholder for any other address space
2111 except for '__constant' in OpenCL code which can be used with multiple address
2116 def OpenCLAddressSpaceConstantDocs : Documentation {
2117 let Category = DocOpenCLAddressSpaces;
2119 The constant address space attribute signals that an object is located in
2120 a constant (non-modifiable) memory region. It is available to all work items.
2121 Any type can be annotated with the constant address space attribute. Objects
2122 with the constant address space qualifier can be declared in any scope and must
2123 have an initializer.
2127 def OpenCLAddressSpaceGlobalDocs : Documentation {
2128 let Category = DocOpenCLAddressSpaces;
2130 The global address space attribute specifies that an object is allocated in
2131 global memory, which is accessible by all work items. The content stored in this
2132 memory area persists between kernel executions. Pointer types to the global
2133 address space are allowed as function parameters or local variables. Starting
2134 with OpenCL v2.0, the global address space can be used with global (program
2135 scope) variables and static local variable as well.
2139 def OpenCLAddressSpaceLocalDocs : Documentation {
2140 let Category = DocOpenCLAddressSpaces;
2142 The local address space specifies that an object is allocated in the local (work
2143 group) memory area, which is accessible to all work items in the same work
2144 group. The content stored in this memory region is not accessible after
2145 the kernel execution ends. In a kernel function scope, any variable can be in
2146 the local address space. In other scopes, only pointer types to the local address
2147 space are allowed. Local address space variables cannot have an initializer.
2151 def OpenCLAddressSpacePrivateDocs : Documentation {
2152 let Category = DocOpenCLAddressSpaces;
2154 The private address space specifies that an object is allocated in the private
2155 (work item) memory. Other work items cannot access the same memory area and its
2156 content is destroyed after work item execution ends. Local variables can be
2157 declared in the private address space. Function arguments are always in the
2158 private address space. Kernel function arguments of a pointer or an array type
2159 cannot point to the private address space.
2163 def OpenCLNoSVMDocs : Documentation {
2164 let Category = DocCatVariable;
2166 OpenCL 2.0 supports the optional ``__attribute__((nosvm))`` qualifier for
2167 pointer variable. It informs the compiler that the pointer does not refer
2168 to a shared virtual memory region. See OpenCL v2.0 s6.7.2 for details.
2170 Since it is not widely used and has been removed from OpenCL 2.1, it is ignored
2174 def NullabilityDocs : DocumentationCategory<"Nullability Attributes"> {
2176 Whether a particular pointer may be "null" is an important concern when working with pointers in the C family of languages. The various nullability attributes indicate whether a particular pointer can be null or not, which makes APIs more expressive and can help static analysis tools identify bugs involving null pointers. Clang supports several kinds of nullability attributes: the ``nonnull`` and ``returns_nonnull`` attributes indicate which function or method parameters and result types can never be null, while nullability type qualifiers indicate which pointer types can be null (``_Nullable``) or cannot be null (``_Nonnull``).
2178 The nullability (type) qualifiers express whether a value of a given pointer type can be null (the ``_Nullable`` qualifier), doesn't have a defined meaning for null (the ``_Nonnull`` qualifier), or for which the purpose of null is unclear (the ``_Null_unspecified`` qualifier). Because nullability qualifiers are expressed within the type system, they are more general than the ``nonnull`` and ``returns_nonnull`` attributes, allowing one to express (for example) a nullable pointer to an array of nonnull pointers. Nullability qualifiers are written to the right of the pointer to which they apply. For example:
2182 // No meaningful result when 'ptr' is null (here, it happens to be undefined behavior).
2183 int fetch(int * _Nonnull ptr) { return *ptr; }
2185 // 'ptr' may be null.
2186 int fetch_or_zero(int * _Nullable ptr) {
2187 return ptr ? *ptr : 0;
2190 // A nullable pointer to non-null pointers to const characters.
2191 const char *join_strings(const char * _Nonnull * _Nullable strings, unsigned n);
2193 In Objective-C, there is an alternate spelling for the nullability qualifiers that can be used in Objective-C methods and properties using context-sensitive, non-underscored keywords. For example:
2195 .. code-block:: objective-c
2197 @interface NSView : NSResponder
2198 - (nullable NSView *)ancestorSharedWithView:(nonnull NSView *)aView;
2199 @property (assign, nullable) NSView *superview;
2200 @property (readonly, nonnull) NSArray *subviews;
2205 def TypeNonNullDocs : Documentation {
2206 let Category = NullabilityDocs;
2208 The ``_Nonnull`` nullability qualifier indicates that null is not a meaningful value for a value of the ``_Nonnull`` pointer type. For example, given a declaration such as:
2212 int fetch(int * _Nonnull ptr);
2214 a caller of ``fetch`` should not provide a null value, and the compiler will produce a warning if it sees a literal null value passed to ``fetch``. Note that, unlike the declaration attribute ``nonnull``, the presence of ``_Nonnull`` does not imply that passing null is undefined behavior: ``fetch`` is free to consider null undefined behavior or (perhaps for backward-compatibility reasons) defensively handle null.
2218 def TypeNullableDocs : Documentation {
2219 let Category = NullabilityDocs;
2221 The ``_Nullable`` nullability qualifier indicates that a value of the ``_Nullable`` pointer type can be null. For example, given:
2225 int fetch_or_zero(int * _Nullable ptr);
2227 a caller of ``fetch_or_zero`` can provide null.
2231 def TypeNullUnspecifiedDocs : Documentation {
2232 let Category = NullabilityDocs;
2234 The ``_Null_unspecified`` nullability qualifier indicates that neither the ``_Nonnull`` nor ``_Nullable`` qualifiers make sense for a particular pointer type. It is used primarily to indicate that the role of null with specific pointers in a nullability-annotated header is unclear, e.g., due to overly-complex implementations or historical factors with a long-lived API.
2238 def NonNullDocs : Documentation {
2239 let Category = NullabilityDocs;
2241 The ``nonnull`` attribute indicates that some function parameters must not be null, and can be used in several different ways. It's original usage (`from GCC <https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes>`_) is as a function (or Objective-C method) attribute that specifies which parameters of the function are nonnull in a comma-separated list. For example:
2245 extern void * my_memcpy (void *dest, const void *src, size_t len)
2246 __attribute__((nonnull (1, 2)));
2248 Here, the ``nonnull`` attribute indicates that parameters 1 and 2
2249 cannot have a null value. Omitting the parenthesized list of parameter indices means that all parameters of pointer type cannot be null:
2253 extern void * my_memcpy (void *dest, const void *src, size_t len)
2254 __attribute__((nonnull));
2256 Clang also allows the ``nonnull`` attribute to be placed directly on a function (or Objective-C method) parameter, eliminating the need to specify the parameter index ahead of type. For example:
2260 extern void * my_memcpy (void *dest __attribute__((nonnull)),
2261 const void *src __attribute__((nonnull)), size_t len);
2263 Note that the ``nonnull`` attribute indicates that passing null to a non-null parameter is undefined behavior, which the optimizer may take advantage of to, e.g., remove null checks. The ``_Nonnull`` type qualifier indicates that a pointer cannot be null in a more general manner (because it is part of the type system) and does not imply undefined behavior, making it more widely applicable.
2267 def ReturnsNonNullDocs : Documentation {
2268 let Category = NullabilityDocs;
2270 The ``returns_nonnull`` attribute indicates that a particular function (or Objective-C method) always returns a non-null pointer. For example, a particular system ``malloc`` might be defined to terminate a process when memory is not available rather than returning a null pointer:
2274 extern void * malloc (size_t size) __attribute__((returns_nonnull));
2276 The ``returns_nonnull`` attribute implies that returning a null pointer is undefined behavior, which the optimizer may take advantage of. The ``_Nonnull`` type qualifier indicates that a pointer cannot be null in a more general manner (because it is part of the type system) and does not imply undefined behavior, making it more widely applicable
2280 def NoAliasDocs : Documentation {
2281 let Category = DocCatFunction;
2283 The ``noalias`` attribute indicates that the only memory accesses inside
2284 function are loads and stores from objects pointed to by its pointer-typed
2285 arguments, with arbitrary offsets.
2289 def OMPDeclareSimdDocs : Documentation {
2290 let Category = DocCatFunction;
2291 let Heading = "#pragma omp declare simd";
2293 The `declare simd` construct can be applied to a function to enable the creation
2294 of one or more versions that can process multiple arguments using SIMD
2295 instructions from a single invocation in a SIMD loop. The `declare simd`
2296 directive is a declarative directive. There may be multiple `declare simd`
2297 directives for a function. The use of a `declare simd` construct on a function
2298 enables the creation of SIMD versions of the associated function that can be
2299 used to process multiple arguments from a single invocation from a SIMD loop
2301 The syntax of the `declare simd` construct is as follows:
2305 #pragma omp declare simd [clause[[,] clause] ...] new-line
2306 [#pragma omp declare simd [clause[[,] clause] ...] new-line]
2308 function definition or declaration
2310 where clause is one of the following:
2315 linear(argument-list[:constant-linear-step])
2316 aligned(argument-list[:alignment])
2317 uniform(argument-list)
2324 def OMPDeclareTargetDocs : Documentation {
2325 let Category = DocCatFunction;
2326 let Heading = "#pragma omp declare target";
2328 The `declare target` directive specifies that variables and functions are mapped
2329 to a device for OpenMP offload mechanism.
2331 The syntax of the declare target directive is as follows:
2335 #pragma omp declare target new-line
2336 declarations-definition-seq
2337 #pragma omp end declare target new-line
2341 def NotTailCalledDocs : Documentation {
2342 let Category = DocCatFunction;
2344 The ``not_tail_called`` attribute prevents tail-call optimization on statically bound calls. It has no effect on indirect calls. Virtual functions, objective-c methods, and functions marked as ``always_inline`` cannot be marked as ``not_tail_called``.
2346 For example, it prevents tail-call optimization in the following case:
2350 int __attribute__((not_tail_called)) foo1(int);
2353 return foo1(a); // No tail-call optimization on direct calls.
2356 However, it doesn't prevent tail-call optimization in this case:
2360 int __attribute__((not_tail_called)) foo1(int);
2363 int (*fn)(int) = &foo1;
2365 // not_tail_called has no effect on an indirect call even if the call can be
2366 // resolved at compile time.
2370 Marking virtual functions as ``not_tail_called`` is an error:
2376 // not_tail_called on a virtual function is an error.
2377 [[clang::not_tail_called]] virtual int foo1();
2381 // Non-virtual functions can be marked ``not_tail_called``.
2382 [[clang::not_tail_called]] int foo3();
2385 class Derived1 : public Base {
2387 int foo1() override;
2389 // not_tail_called on a virtual function is an error.
2390 [[clang::not_tail_called]] int foo2() override;
2395 def InternalLinkageDocs : Documentation {
2396 let Category = DocCatFunction;
2398 The ``internal_linkage`` attribute changes the linkage type of the declaration to internal.
2399 This is similar to C-style ``static``, but can be used on classes and class methods. When applied to a class definition,
2400 this attribute affects all methods and static data members of that class.
2401 This can be used to contain the ABI of a C++ library by excluding unwanted class methods from the export tables.
2405 def DisableTailCallsDocs : Documentation {
2406 let Category = DocCatFunction;
2408 The ``disable_tail_calls`` attribute instructs the backend to not perform tail call optimization inside the marked function.
2416 int foo(int a) __attribute__((disable_tail_calls)) {
2417 return callee(a); // This call is not tail-call optimized.
2420 Marking virtual functions as ``disable_tail_calls`` is legal.
2428 [[clang::disable_tail_calls]] virtual int foo1() {
2429 return callee(); // This call is not tail-call optimized.
2433 class Derived1 : public Base {
2435 int foo1() override {
2436 return callee(); // This call is tail-call optimized.
2443 def AnyX86InterruptDocs : Documentation {
2444 let Category = DocCatFunction;
2446 Clang supports the GNU style ``__attribute__((interrupt))`` attribute on
2447 x86/x86-64 targets.The compiler generates function entry and exit sequences
2448 suitable for use in an interrupt handler when this attribute is present.
2449 The 'IRET' instruction, instead of the 'RET' instruction, is used to return
2450 from interrupt or exception handlers. All registers, except for the EFLAGS
2451 register which is restored by the 'IRET' instruction, are preserved by the
2454 Any interruptible-without-stack-switch code must be compiled with
2455 -mno-red-zone since interrupt handlers can and will, because of the
2456 hardware design, touch the red zone.
2458 1. interrupt handler must be declared with a mandatory pointer argument:
2462 struct interrupt_frame
2471 __attribute__ ((interrupt))
2472 void f (struct interrupt_frame *frame) {
2476 2. exception handler:
2478 The exception handler is very similar to the interrupt handler with
2479 a different mandatory function signature:
2483 __attribute__ ((interrupt))
2484 void f (struct interrupt_frame *frame, uword_t error_code) {
2488 and compiler pops 'ERROR_CODE' off stack before the 'IRET' instruction.
2490 The exception handler should only be used for exceptions which push an
2491 error code and all other exceptions must use the interrupt handler.
2492 The system will crash if the wrong handler is used.
2496 def SwiftCallDocs : Documentation {
2497 let Category = DocCatVariable;
2499 The ``swiftcall`` attribute indicates that a function should be called
2500 using the Swift calling convention for a function or function pointer.
2502 The lowering for the Swift calling convention, as described by the Swift
2503 ABI documentation, occurs in multiple phases. The first, "high-level"
2504 phase breaks down the formal parameters and results into innately direct
2505 and indirect components, adds implicit paraameters for the generic
2506 signature, and assigns the context and error ABI treatments to parameters
2507 where applicable. The second phase breaks down the direct parameters
2508 and results from the first phase and assigns them to registers or the
2509 stack. The ``swiftcall`` convention only handles this second phase of
2510 lowering; the C function type must accurately reflect the results
2511 of the first phase, as follows:
2513 - Results classified as indirect by high-level lowering should be
2514 represented as parameters with the ``swift_indirect_result`` attribute.
2516 - Results classified as direct by high-level lowering should be represented
2519 - First, remove any empty direct results.
2521 - If there are no direct results, the C result type should be ``void``.
2523 - If there is one direct result, the C result type should be a type with
2524 the exact layout of that result type.
2526 - If there are a multiple direct results, the C result type should be
2527 a struct type with the exact layout of a tuple of those results.
2529 - Parameters classified as indirect by high-level lowering should be
2530 represented as parameters of pointer type.
2532 - Parameters classified as direct by high-level lowering should be
2533 omitted if they are empty types; otherwise, they should be represented
2534 as a parameter type with a layout exactly matching the layout of the
2535 Swift parameter type.
2537 - The context parameter, if present, should be represented as a trailing
2538 parameter with the ``swift_context`` attribute.
2540 - The error result parameter, if present, should be represented as a
2541 trailing parameter (always following a context parameter) with the
2542 ``swift_error_result`` attribute.
2544 ``swiftcall`` does not support variadic arguments or unprototyped functions.
2546 The parameter ABI treatment attributes are aspects of the function type.
2547 A function type which which applies an ABI treatment attribute to a
2548 parameter is a different type from an otherwise-identical function type
2549 that does not. A single parameter may not have multiple ABI treatment
2552 Support for this feature is target-dependent, although it should be
2553 supported on every target that Swift supports. Query for this support
2554 with ``__has_attribute(swiftcall)``. This implies support for the
2555 ``swift_context``, ``swift_error_result``, and ``swift_indirect_result``
2560 def SwiftContextDocs : Documentation {
2561 let Category = DocCatVariable;
2563 The ``swift_context`` attribute marks a parameter of a ``swiftcall``
2564 function as having the special context-parameter ABI treatment.
2566 This treatment generally passes the context value in a special register
2567 which is normally callee-preserved.
2569 A ``swift_context`` parameter must either be the last parameter or must be
2570 followed by a ``swift_error_result`` parameter (which itself must always be
2571 the last parameter).
2573 A context parameter must have pointer or reference type.
2577 def SwiftErrorResultDocs : Documentation {
2578 let Category = DocCatVariable;
2580 The ``swift_error_result`` attribute marks a parameter of a ``swiftcall``
2581 function as having the special error-result ABI treatment.
2583 This treatment generally passes the underlying error value in and out of
2584 the function through a special register which is normally callee-preserved.
2585 This is modeled in C by pretending that the register is addressable memory:
2587 - The caller appears to pass the address of a variable of pointer type.
2588 The current value of this variable is copied into the register before
2589 the call; if the call returns normally, the value is copied back into the
2592 - The callee appears to receive the address of a variable. This address
2593 is actually a hidden location in its own stack, initialized with the
2594 value of the register upon entry. When the function returns normally,
2595 the value in that hidden location is written back to the register.
2597 A ``swift_error_result`` parameter must be the last parameter, and it must be
2598 preceded by a ``swift_context`` parameter.
2600 A ``swift_error_result`` parameter must have type ``T**`` or ``T*&`` for some
2601 type T. Note that no qualifiers are permitted on the intermediate level.
2603 It is undefined behavior if the caller does not pass a pointer or
2604 reference to a valid object.
2606 The standard convention is that the error value itself (that is, the
2607 value stored in the apparent argument) will be null upon function entry,
2608 but this is not enforced by the ABI.
2612 def SwiftIndirectResultDocs : Documentation {
2613 let Category = DocCatVariable;
2615 The ``swift_indirect_result`` attribute marks a parameter of a ``swiftcall``
2616 function as having the special indirect-result ABI treatment.
2618 This treatment gives the parameter the target's normal indirect-result
2619 ABI treatment, which may involve passing it differently from an ordinary
2620 parameter. However, only the first indirect result will receive this
2621 treatment. Furthermore, low-level lowering may decide that a direct result
2622 must be returned indirectly; if so, this will take priority over the
2623 ``swift_indirect_result`` parameters.
2625 A ``swift_indirect_result`` parameter must either be the first parameter or
2626 follow another ``swift_indirect_result`` parameter.
2628 A ``swift_indirect_result`` parameter must have type ``T*`` or ``T&`` for
2629 some object type ``T``. If ``T`` is a complete type at the point of
2630 definition of a function, it is undefined behavior if the argument
2631 value does not point to storage of adequate size and alignment for a
2632 value of type ``T``.
2634 Making indirect results explicit in the signature allows C functions to
2635 directly construct objects into them without relying on language
2636 optimizations like C++'s named return value optimization (NRVO).
2640 def AbiTagsDocs : Documentation {
2641 let Category = DocCatFunction;
2643 The ``abi_tag`` attribute can be applied to a function, variable, class or
2644 inline namespace declaration to modify the mangled name of the entity. It gives
2645 the ability to distinguish between different versions of the same entity but
2646 with different ABI versions supported. For example, a newer version of a class
2647 could have a different set of data members and thus have a different size. Using
2648 the ``abi_tag`` attribute, it is possible to have different mangled names for
2649 a global variable of the class type. Therefor, the old code could keep using
2650 the old manged name and the new code will use the new mangled name with tags.
2654 def PreserveMostDocs : Documentation {
2655 let Category = DocCatCallingConvs;
2657 On X86-64 and AArch64 targets, this attribute changes the calling convention of
2658 a function. The ``preserve_most`` calling convention attempts to make the code
2659 in the caller as unintrusive as possible. This convention behaves identically
2660 to the ``C`` calling convention on how arguments and return values are passed,
2661 but it uses a different set of caller/callee-saved registers. This alleviates
2662 the burden of saving and recovering a large register set before and after the
2663 call in the caller. If the arguments are passed in callee-saved registers,
2664 then they will be preserved by the callee across the call. This doesn't
2665 apply for values returned in callee-saved registers.
2667 - On X86-64 the callee preserves all general purpose registers, except for
2668 R11. R11 can be used as a scratch register. Floating-point registers
2669 (XMMs/YMMs) are not preserved and need to be saved by the caller.
2671 The idea behind this convention is to support calls to runtime functions
2672 that have a hot path and a cold path. The hot path is usually a small piece
2673 of code that doesn't use many registers. The cold path might need to call out to
2674 another function and therefore only needs to preserve the caller-saved
2675 registers, which haven't already been saved by the caller. The
2676 `preserve_most` calling convention is very similar to the ``cold`` calling
2677 convention in terms of caller/callee-saved registers, but they are used for
2678 different types of function calls. ``coldcc`` is for function calls that are
2679 rarely executed, whereas `preserve_most` function calls are intended to be
2680 on the hot path and definitely executed a lot. Furthermore ``preserve_most``
2681 doesn't prevent the inliner from inlining the function call.
2683 This calling convention will be used by a future version of the Objective-C
2684 runtime and should therefore still be considered experimental at this time.
2685 Although this convention was created to optimize certain runtime calls to
2686 the Objective-C runtime, it is not limited to this runtime and might be used
2687 by other runtimes in the future too. The current implementation only
2688 supports X86-64 and AArch64, but the intention is to support more architectures
2693 def PreserveAllDocs : Documentation {
2694 let Category = DocCatCallingConvs;
2696 On X86-64 and AArch64 targets, this attribute changes the calling convention of
2697 a function. The ``preserve_all`` calling convention attempts to make the code
2698 in the caller even less intrusive than the ``preserve_most`` calling convention.
2699 This calling convention also behaves identical to the ``C`` calling convention
2700 on how arguments and return values are passed, but it uses a different set of
2701 caller/callee-saved registers. This removes the burden of saving and
2702 recovering a large register set before and after the call in the caller. If
2703 the arguments are passed in callee-saved registers, then they will be
2704 preserved by the callee across the call. This doesn't apply for values
2705 returned in callee-saved registers.
2707 - On X86-64 the callee preserves all general purpose registers, except for
2708 R11. R11 can be used as a scratch register. Furthermore it also preserves
2709 all floating-point registers (XMMs/YMMs).
2711 The idea behind this convention is to support calls to runtime functions
2712 that don't need to call out to any other functions.
2714 This calling convention, like the ``preserve_most`` calling convention, will be
2715 used by a future version of the Objective-C runtime and should be considered
2716 experimental at this time.
2720 def DeprecatedDocs : Documentation {
2721 let Category = DocCatFunction;
2723 The ``deprecated`` attribute can be applied to a function, a variable, or a
2724 type. This is useful when identifying functions, variables, or types that are
2725 expected to be removed in a future version of a program.
2727 Consider the function declaration for a hypothetical function ``f``:
2731 void f(void) __attribute__((deprecated("message", "replacement")));
2733 When spelled as `__attribute__((deprecated))`, the deprecated attribute can have
2734 two optional string arguments. The first one is the message to display when
2735 emitting the warning; the second one enables the compiler to provide a Fix-It
2736 to replace the deprecated name with a new name. Otherwise, when spelled as
2737 `[[gnu::deprecated]] or [[deprecated]]`, the attribute can have one optional
2738 string argument which is the message to display when emitting the warning.
2742 def IFuncDocs : Documentation {
2743 let Category = DocCatFunction;
2745 ``__attribute__((ifunc("resolver")))`` is used to mark that the address of a declaration should be resolved at runtime by calling a resolver function.
2747 The symbol name of the resolver function is given in quotes. A function with this name (after mangling) must be defined in the current translation unit; it may be ``static``. The resolver function should take no arguments and return a pointer.
2749 The ``ifunc`` attribute may only be used on a function declaration. A function declaration with an ``ifunc`` attribute is considered to be a definition of the declared entity. The entity must not have weak linkage; for example, in C++, it cannot be applied to a declaration if a definition at that location would be considered inline.
2751 Not all targets support this attribute. ELF targets support this attribute when using binutils v2.20.1 or higher and glibc v2.11.1 or higher. Non-ELF targets currently do not support this attribute.
2755 def LTOVisibilityDocs : Documentation {
2756 let Category = DocCatType;
2758 See :doc:`LTOVisibility`.
2762 def RenderScriptKernelAttributeDocs : Documentation {
2763 let Category = DocCatFunction;
2765 ``__attribute__((kernel))`` is used to mark a ``kernel`` function in
2768 In RenderScript, ``kernel`` functions are used to express data-parallel
2769 computations. The RenderScript runtime efficiently parallelizes ``kernel``
2770 functions to run on computational resources such as multi-core CPUs and GPUs.
2771 See the RenderScript_ documentation for more information.
2773 .. _RenderScript: https://developer.android.com/guide/topics/renderscript/compute.html
2777 def XRayDocs : Documentation {
2778 let Category = DocCatFunction;
2779 let Heading = "xray_always_instrument (clang::xray_always_instrument), xray_never_instrument (clang::xray_never_instrument)";
2781 ``__attribute__((xray_always_instrument))`` or ``[[clang::xray_always_instrument]]`` is used to mark member functions (in C++), methods (in Objective C), and free functions (in C, C++, and Objective C) to be instrumented with XRay. This will cause the function to always have space at the beginning and exit points to allow for runtime patching.
2783 Conversely, ``__attribute__((xray_never_instrument))`` or ``[[clang::xray_never_instrument]]`` will inhibit the insertion of these instrumentation points.
2785 If a function has neither of these attributes, they become subject to the XRay heuristics used to determine whether a function should be instrumented or otherwise.
2789 def TransparentUnionDocs : Documentation {
2790 let Category = DocCatType;
2792 This attribute can be applied to a union to change the behaviour of calls to
2793 functions that have an argument with a transparent union type. The compiler
2794 behaviour is changed in the following manner:
2796 - A value whose type is any member of the transparent union can be passed as an
2797 argument without the need to cast that value.
2799 - The argument is passed to the function using the calling convention of the
2800 first member of the transparent union. Consequently, all the members of the
2801 transparent union should have the same calling convention as its first member.
2803 Transparent unions are not supported in C++.
2807 def ObjCSubclassingRestrictedDocs : Documentation {
2808 let Category = DocCatType;
2810 This attribute can be added to an Objective-C ``@interface`` declaration to
2811 ensure that this class cannot be subclassed.