• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2          "http://www.w3.org/TR/html4/strict.dtd">
3<!-- Material used from: HTML 4.01 specs: http://www.w3.org/TR/html401/ -->
4<html>
5<head>
6  <META http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
7  <title>Clang Language Extensions</title>
8  <link type="text/css" rel="stylesheet" href="../menu.css">
9  <link type="text/css" rel="stylesheet" href="../content.css">
10  <style type="text/css">
11    td {
12            vertical-align: top;
13    }
14    th { background-color: #ffddaa; }
15  </style>
16</head>
17<body>
18
19<!--#include virtual="../menu.html.incl"-->
20
21<div id="content">
22
23<h1>Clang Language Extensions</h1>
24
25<ul>
26<li><a href="#intro">Introduction</a></li>
27<li><a href="#feature_check">Feature Checking Macros</a></li>
28<li><a href="#has_include">Include File Checking Macros</a></li>
29<li><a href="#builtinmacros">Builtin Macros</a></li>
30<li><a href="#vectors">Vectors and Extended Vectors</a></li>
31<li><a href="#deprecated">Messages on <tt>deprecated</tt> and <tt>unavailable</tt> attributes</a></li>
32<li><a href="#attributes-on-enumerators">Attributes on enumerators</a></li>
33<li><a href="#user_specified_system_framework">'User-Specified' System Frameworks</a></li>
34<li><a href="#availability">Availability attribute</a></li>
35<li><a href="#checking_language_features">Checks for Standard Language Features</a>
36  <ul>
37  <li><a href="#cxx98">C++98</a>
38    <ul>
39    <li><a href="#cxx_exceptions">C++ exceptions</a></li>
40    <li><a href="#cxx_rtti">C++ RTTI</a></li>
41  </ul></li>
42  <li><a href="#cxx11">C++11</a>
43    <ul>
44    <li><a href="#cxx_access_control_sfinae">C++11 SFINAE includes access control</a></li>
45    <li><a href="#cxx_alias_templates">C++11 alias templates</a></li>
46    <li><a href="#cxx_alignas">C++11 alignment specifiers</a></li>
47    <li><a href="#cxx_attributes">C++11 attributes</a></li>
48    <li><a href="#cxx_constexpr">C++11 generalized constant expressions</a></li>
49    <li><a href="#cxx_decltype">C++11 <tt>decltype()</tt></a></li>
50    <li><a href="#cxx_default_function_template_args">C++11 default template arguments in function templates</a></li>
51    <li><a href="#cxx_defaulted_functions">C++11 defaulted functions</a></li>
52    <li><a href="#cxx_delegating_constructor">C++11 delegating constructors</a></li>
53    <li><a href="#cxx_deleted_functions">C++11 deleted functions</a></li>
54    <li><a href="#cxx_explicit_conversions">C++11 explicit conversion functions</a></li>
55    <li><a href="#cxx_generalized_initializers">C++11 generalized initializers</a></li>
56    <li><a href="#cxx_implicit_moves">C++11 implicit move constructors/assignment operators</a></li>
57    <li><a href="#cxx_inheriting_constructors">C++11 inheriting constructors</a></li>
58    <li><a href="#cxx_inline_namespaces">C++11 inline namespaces</a></li>
59    <li><a href="#cxx_lambdas">C++11 lambdas</a></li>
60    <li><a href="#cxx_local_type_template_args">C++11 local and unnamed types as template arguments</a></li>
61    <li><a href="#cxx_noexcept">C++11 noexcept specification</a></li>
62    <li><a href="#cxx_nonstatic_member_init">C++11 in-class non-static data member initialization</a></li>
63    <li><a href="#cxx_nullptr">C++11 nullptr</a></li>
64    <li><a href="#cxx_override_control">C++11 override control</a></li>
65    <li><a href="#cxx_range_for">C++11 range-based for loop</a></li>
66    <li><a href="#cxx_raw_string_literals">C++11 raw string literals</a></li>
67    <li><a href="#cxx_rvalue_references">C++11 rvalue references</a></li>
68    <li><a href="#cxx_reference_qualified_functions">C++11 reference-qualified functions</a></li>
69    <li><a href="#cxx_static_assert">C++11 <tt>static_assert()</tt></a></li>
70    <li><a href="#cxx_auto_type">C++11 type inference</a></li>
71    <li><a href="#cxx_strong_enums">C++11 strongly-typed enumerations</a></li>
72    <li><a href="#cxx_trailing_return">C++11 trailing return type</a></li>
73    <li><a href="#cxx_unicode_literals">C++11 Unicode string literals</a></li>
74    <li><a href="#cxx_unrestricted_unions">C++11 unrestricted unions</a></li>
75    <li><a href="#cxx_user_literals">C++11 user-defined literals</a></li>
76    <li><a href="#cxx_variadic_templates">C++11 variadic templates</a></li>
77  </ul></li>
78  <li><a href="#c11">C11</a>
79    <ul>
80    <li><a href="#c_alignas">C11 alignment specifiers</a></li>
81    <li><a href="#c_atomic">C11 atomic operations</a></li>
82    <li><a href="#c_generic_selections">C11 generic selections</a></li>
83    <li><a href="#c_static_assert">C11 <tt>_Static_assert()</tt></a></li>
84  </ul></li>
85</ul></li>
86<li><a href="#checking_type_traits">Checks for Type Traits</a></li>
87<li><a href="#blocks">Blocks</a></li>
88<li><a href="#objc_features">Objective-C Features</a>
89  <ul>
90    <li><a href="#objc_instancetype">Related result types</a></li>
91    <li><a href="#objc_arc">Automatic reference counting</a></li>
92    <li><a href="#objc_fixed_enum">Enumerations with a fixed underlying type</a></li>
93    <li><a href="#objc_lambdas">Interoperability with C++11 lambdas</a></li>
94    <li><a href="#objc_object_literals_subscripting">Object Literals and Subscripting</a></li>
95  </ul>
96</li>
97<li><a href="#overloading-in-c">Function Overloading in C</a></li>
98<li><a href="#complex-list-init">Initializer lists for complex numbers in C</a></li>
99<li><a href="#builtins">Builtin Functions</a>
100  <ul>
101  <li><a href="#__builtin_shufflevector">__builtin_shufflevector</a></li>
102  <li><a href="#__builtin_unreachable">__builtin_unreachable</a></li>
103  <li><a href="#__sync_swap">__sync_swap</a></li>
104 </ul>
105</li>
106<li><a href="#targetspecific">Target-Specific Extensions</a>
107  <ul>
108  <li><a href="#x86-specific">X86/X86-64 Language Extensions</a></li>
109  </ul>
110</li>
111<li><a href="#analyzerspecific">Static Analysis-Specific Extensions</a></li>
112<li><a href="#dynamicanalyzerspecific">Dynamic Analysis-Specific Extensions</a>
113  <ul>
114  <li><a href="#address_sanitizer">AddressSanitizer</a></li>
115  </ul>
116</li>
117<li><a href="#threadsafety">Thread Safety Annotation Checking</a>
118    <ul>
119    <li><a href="#ts_noanal"><tt>no_thread_safety_analysis</tt></a></li>
120    <li><a href="#ts_lockable"><tt>lockable</tt></a></li>
121    <li><a href="#ts_scopedlockable"><tt>scoped_lockable</tt></a></li>
122    <li><a href="#ts_guardedvar"><tt>guarded_var</tt></a></li>
123    <li><a href="#ts_ptguardedvar"><tt>pt_guarded_var</tt></a></li>
124    <li><a href="#ts_guardedby"><tt>guarded_by(l)</tt></a></li>
125    <li><a href="#ts_ptguardedby"><tt>pt_guarded_by(l)</tt></a></li>
126    <li><a href="#ts_acquiredbefore"><tt>acquired_before(...)</tt></a></li>
127    <li><a href="#ts_acquiredafter"><tt>acquired_after(...)</tt></a></li>
128    <li><a href="#ts_elf"><tt>exclusive_lock_function(...)</tt></a></li>
129    <li><a href="#ts_slf"><tt>shared_lock_function(...)</tt></a></li>
130    <li><a href="#ts_etf"><tt>exclusive_trylock_function(...)</tt></a></li>
131    <li><a href="#ts_stf"><tt>shared_trylock_function(...)</tt></a></li>
132    <li><a href="#ts_uf"><tt>unlock_function(...)</tt></a></li>
133    <li><a href="#ts_lr"><tt>lock_returned(l)</tt></a></li>
134    <li><a href="#ts_le"><tt>locks_excluded(...)</tt></a></li>
135    <li><a href="#ts_elr"><tt>exclusive_locks_required(...)</tt></a></li>
136    <li><a href="#ts_slr"><tt>shared_locks_required(...)</tt></a></li>
137    </ul>
138</li>
139</ul>
140
141<!-- ======================================================================= -->
142<h2 id="intro">Introduction</h2>
143<!-- ======================================================================= -->
144
145<p>This document describes the language extensions provided by Clang.  In
146addition to the language extensions listed here, Clang aims to support a broad
147range of GCC extensions.  Please see the <a
148href="http://gcc.gnu.org/onlinedocs/gcc/C-Extensions.html">GCC manual</a> for
149more information on these extensions.</p>
150
151<!-- ======================================================================= -->
152<h2 id="feature_check">Feature Checking Macros</h2>
153<!-- ======================================================================= -->
154
155<p>Language extensions can be very useful, but only if you know you can depend
156on them.  In order to allow fine-grain features checks, we support three builtin
157function-like macros.  This allows you to directly test for a feature in your
158code without having to resort to something like autoconf or fragile "compiler
159version checks".</p>
160
161<!-- ======================================================================= -->
162<h3><a name="__has_builtin">__has_builtin</a></h3>
163<!-- ======================================================================= -->
164
165<p>This function-like macro takes a single identifier argument that is the name
166of a builtin function.  It evaluates to 1 if the builtin is supported or 0 if
167not.  It can be used like this:</p>
168
169<blockquote>
170<pre>
171#ifndef __has_builtin         // Optional of course.
172  #define __has_builtin(x) 0  // Compatibility with non-clang compilers.
173#endif
174
175...
176#if __has_builtin(__builtin_trap)
177  __builtin_trap();
178#else
179  abort();
180#endif
181...
182</pre>
183</blockquote>
184
185
186<!-- ======================================================================= -->
187<h3><a name="__has_feature_extension"> __has_feature and __has_extension</a></h3>
188<!-- ======================================================================= -->
189
190<p>These function-like macros take a single identifier argument that is the
191name of a feature.  <code>__has_feature</code> evaluates to 1 if the feature
192is both supported by Clang and standardized in the current language standard
193or 0 if not (but see <a href="#has_feature_back_compat">below</a>), while
194<code>__has_extension</code> evaluates to 1 if the feature is supported by
195Clang in the current language (either as a language extension or a standard
196language feature) or 0 if not.  They can be used like this:</p>
197
198<blockquote>
199<pre>
200#ifndef __has_feature         // Optional of course.
201  #define __has_feature(x) 0  // Compatibility with non-clang compilers.
202#endif
203#ifndef __has_extension
204  #define __has_extension __has_feature // Compatibility with pre-3.0 compilers.
205#endif
206
207...
208#if __has_feature(cxx_rvalue_references)
209// This code will only be compiled with the -std=c++11 and -std=gnu++11
210// options, because rvalue references are only standardized in C++11.
211#endif
212
213#if __has_extension(cxx_rvalue_references)
214// This code will be compiled with the -std=c++11, -std=gnu++11, -std=c++98
215// and -std=gnu++98 options, because rvalue references are supported as a
216// language extension in C++98.
217#endif
218</pre>
219</blockquote>
220
221<p id="has_feature_back_compat">For backwards compatibility reasons,
222<code>__has_feature</code> can also be used to test for support for
223non-standardized features, i.e. features not prefixed <code>c_</code>,
224<code>cxx_</code> or <code>objc_</code>.</p>
225
226<p id="has_feature_for_non_language_features">
227Another use of <code>__has_feature</code> is to check for compiler features
228not related to the language standard, such as e.g.
229<a href="AddressSanitizer.html">AddressSanitizer</a>.
230
231<p>If the <code>-pedantic-errors</code> option is given,
232<code>__has_extension</code> is equivalent to <code>__has_feature</code>.</p>
233
234<p>The feature tag is described along with the language feature below.</p>
235
236<p>The feature name or extension name can also be specified with a preceding and
237following <code>__</code> (double underscore) to avoid interference from a macro
238with the same name. For instance, <code>__cxx_rvalue_references__</code> can be
239used instead of <code>cxx_rvalue_references</code>.</p>
240
241<!-- ======================================================================= -->
242<h3><a name="__has_attribute">__has_attribute</a></h3>
243<!-- ======================================================================= -->
244
245<p>This function-like macro takes a single identifier argument that is the name
246of an attribute.  It evaluates to 1 if the attribute is supported or 0 if not.  It
247can be used like this:</p>
248
249<blockquote>
250<pre>
251#ifndef __has_attribute         // Optional of course.
252  #define __has_attribute(x) 0  // Compatibility with non-clang compilers.
253#endif
254
255...
256#if __has_attribute(always_inline)
257#define ALWAYS_INLINE __attribute__((always_inline))
258#else
259#define ALWAYS_INLINE
260#endif
261...
262</pre>
263</blockquote>
264
265<p>The attribute name can also be specified with a preceding and
266following <code>__</code> (double underscore) to avoid interference from a macro
267with the same name. For instance, <code>__always_inline__</code> can be used
268instead of <code>always_inline</code>.</p>
269
270<!-- ======================================================================= -->
271<h2 id="has_include">Include File Checking Macros</h2>
272<!-- ======================================================================= -->
273
274<p>Not all developments systems have the same include files.
275The <a href="#__has_include">__has_include</a> and
276<a href="#__has_include_next">__has_include_next</a> macros allow you to
277check for the existence of an include file before doing
278a possibly failing #include directive.</p>
279
280<!-- ======================================================================= -->
281<h3><a name="__has_include">__has_include</a></h3>
282<!-- ======================================================================= -->
283
284<p>This function-like macro takes a single file name string argument that
285is the name of an include file.  It evaluates to 1 if the file can
286be found using the include paths, or 0 otherwise:</p>
287
288<blockquote>
289<pre>
290// Note the two possible file name string formats.
291#if __has_include("myinclude.h") &amp;&amp; __has_include(&lt;stdint.h&gt;)
292# include "myinclude.h"
293#endif
294
295// To avoid problem with non-clang compilers not having this macro.
296#if defined(__has_include) &amp;&amp; __has_include("myinclude.h")
297# include "myinclude.h"
298#endif
299</pre>
300</blockquote>
301
302<p>To test for this feature, use #if defined(__has_include).</p>
303
304<!-- ======================================================================= -->
305<h3><a name="__has_include_next">__has_include_next</a></h3>
306<!-- ======================================================================= -->
307
308<p>This function-like macro takes a single file name string argument that
309is the name of an include file.  It is like __has_include except that it
310looks for the second instance of the given file found in the include
311paths.  It evaluates to 1 if the second instance of the file can
312be found using the include paths, or 0 otherwise:</p>
313
314<blockquote>
315<pre>
316// Note the two possible file name string formats.
317#if __has_include_next("myinclude.h") &amp;&amp; __has_include_next(&lt;stdint.h&gt;)
318# include_next "myinclude.h"
319#endif
320
321// To avoid problem with non-clang compilers not having this macro.
322#if defined(__has_include_next) &amp;&amp; __has_include_next("myinclude.h")
323# include_next "myinclude.h"
324#endif
325</pre>
326</blockquote>
327
328<p>Note that __has_include_next, like the GNU extension
329#include_next directive, is intended for use in headers only,
330and will issue a warning if used in the top-level compilation
331file.  A warning will also be issued if an absolute path
332is used in the file argument.</p>
333
334
335<!-- ======================================================================= -->
336<h3><a name="__has_warning">__has_warning</a></h3>
337<!-- ======================================================================= -->
338
339<p>This function-like macro takes a string literal that represents a command
340  line option for a warning and returns true if that is a valid warning
341  option.</p>
342
343<blockquote>
344<pre>
345#if __has_warning("-Wformat")
346...
347#endif
348</pre>
349</blockquote>
350
351<!-- ======================================================================= -->
352<h2 id="builtinmacros">Builtin Macros</h2>
353<!-- ======================================================================= -->
354
355<dl>
356  <dt><code>__BASE_FILE__</code></dt>
357  <dd>Defined to a string that contains the name of the main input
358  file passed to Clang.</dd>
359
360  <dt><code>__COUNTER__</code></dt>
361  <dd>Defined to an integer value that starts at zero and is
362  incremented each time the <code>__COUNTER__</code> macro is
363  expanded.</dd>
364
365  <dt><code>__INCLUDE_LEVEL__</code></dt>
366  <dd>Defined to an integral value that is the include depth of the
367  file currently being translated. For the main file, this value is
368  zero.</dd>
369
370  <dt><code>__TIMESTAMP__</code></dt>
371  <dd>Defined to the date and time of the last modification of the
372  current source file.</dd>
373
374  <dt><code>__clang__</code></dt>
375  <dd>Defined when compiling with Clang</dd>
376
377  <dt><code>__clang_major__</code></dt>
378  <dd>Defined to the major marketing version number of Clang (e.g., the
379  2 in 2.0.1).  Note that marketing version numbers should not be used to
380  check for language features, as different vendors use different numbering
381  schemes.  Instead, use the <a href="#feature_check">feature checking
382  macros</a>.</dd>
383
384  <dt><code>__clang_minor__</code></dt>
385  <dd>Defined to the minor version number of Clang (e.g., the 0 in
386  2.0.1).  Note that marketing version numbers should not be used to
387  check for language features, as different vendors use different numbering
388  schemes.  Instead, use the <a href="#feature_check">feature checking
389  macros</a>.</dd>
390
391  <dt><code>__clang_patchlevel__</code></dt>
392  <dd>Defined to the marketing patch level of Clang (e.g., the 1 in 2.0.1).</dd>
393
394  <dt><code>__clang_version__</code></dt>
395  <dd>Defined to a string that captures the Clang marketing version, including
396  the Subversion tag or revision number, e.g., "1.5 (trunk 102332)".</dd>
397</dl>
398
399<!-- ======================================================================= -->
400<h2 id="vectors">Vectors and Extended Vectors</h2>
401<!-- ======================================================================= -->
402
403<p>Supports the GCC, OpenCL, AltiVec and NEON vector extensions.</p>
404
405<p>OpenCL vector types are created using <tt>ext_vector_type</tt> attribute. It
406support for <tt>V.xyzw</tt> syntax and other tidbits as seen in OpenCL. An
407example is:</p>
408
409<blockquote>
410<pre>
411typedef float float4 <b>__attribute__((ext_vector_type(4)))</b>;
412typedef float float2 <b>__attribute__((ext_vector_type(2)))</b>;
413
414float4 foo(float2 a, float2 b) {
415  float4 c;
416  c.xz = a;
417  c.yw = b;
418  return c;
419}
420</pre>
421</blockquote>
422
423<p>Query for this feature with
424<tt>__has_extension(attribute_ext_vector_type)</tt>.</p>
425
426<p>Giving <tt>-faltivec</tt> option to clang enables support for AltiVec vector
427syntax and functions. For example:</p>
428
429<blockquote>
430<pre>
431vector float foo(vector int a) {
432  vector int b;
433  b = vec_add(a, a) + a;
434  return (vector float)b;
435}
436</pre>
437</blockquote>
438
439<p>NEON vector types are created using <tt>neon_vector_type</tt> and
440<tt>neon_polyvector_type</tt> attributes. For example:</p>
441
442<blockquote>
443<pre>
444typedef <b>__attribute__((neon_vector_type(8)))</b> int8_t int8x8_t;
445typedef <b>__attribute__((neon_polyvector_type(16)))</b> poly8_t poly8x16_t;
446
447int8x8_t foo(int8x8_t a) {
448  int8x8_t v;
449  v = a;
450  return v;
451}
452</pre>
453</blockquote>
454
455<!-- ======================================================================= -->
456<h3><a name="vector_literals">Vector Literals</a></h3>
457<!-- ======================================================================= -->
458
459<p>Vector literals can be used to create vectors from a set of scalars, or
460vectors. Either parentheses or braces form can be used. In the parentheses form
461the number of literal values specified must be one, i.e. referring to a scalar
462value, or must match the size of the vector type being created. If a single
463scalar literal value is specified, the scalar literal value will be replicated
464to all the components of the vector type. In the brackets form any number of
465literals can be specified. For example:</p>
466
467<blockquote>
468<pre>
469typedef int v4si __attribute__((__vector_size__(16)));
470typedef float float4 __attribute__((ext_vector_type(4)));
471typedef float float2 __attribute__((ext_vector_type(2)));
472
473v4si vsi = (v4si){1, 2, 3, 4};
474float4 vf = (float4)(1.0f, 2.0f, 3.0f, 4.0f);
475vector int vi1 = (vector int)(1);    // vi1 will be (1, 1, 1, 1).
476vector int vi2 = (vector int){1};    // vi2 will be (1, 0, 0, 0).
477vector int vi3 = (vector int)(1, 2); // error
478vector int vi4 = (vector int){1, 2}; // vi4 will be (1, 2, 0, 0).
479vector int vi5 = (vector int)(1, 2, 3, 4);
480float4 vf = (float4)((float2)(1.0f, 2.0f), (float2)(3.0f, 4.0f));
481</pre>
482</blockquote>
483
484<!-- ======================================================================= -->
485<h3><a name="vector_operations">Vector Operations</a></h3>
486<!-- ======================================================================= -->
487
488<p>The table below shows the support for each operation by vector extension.
489A dash indicates that an operation is not accepted according to a corresponding
490specification.</p>
491
492<table width="500" border="1" cellspacing="0">
493 <tr>
494    <th>Operator</th>
495    <th>OpenCL</th>
496    <th>AltiVec</th>
497    <th>GCC</th>
498    <th>NEON</th>
499 </tr>
500     <tr>
501      <td>[]</td>
502      <td align="center">yes</td>
503      <td align="center">yes</td>
504      <td align="center">yes</td>
505      <td align="center">-</td>
506    </tr>
507    <tr>
508      <td>unary operators +, -</td>
509      <td align="center">yes</td>
510      <td align="center">yes</td>
511      <td align="center">yes</td>
512      <td align="center">-</td>
513    </tr>
514    <tr>
515      <td>++, --</td>
516      <td align="center">yes</td>
517      <td align="center">yes</td>
518      <td align="center">-</td>
519      <td align="center">-</td>
520    </tr>
521    <tr>
522      <td>+, -, *, /, %</td>
523      <td align="center">yes</td>
524      <td align="center">yes</td>
525      <td align="center">yes</td>
526      <td align="center">-</td>
527    </tr>
528    <tr>
529      <td>bitwise operators &, |, ^, ~</td>
530      <td align="center">yes</td>
531      <td align="center">yes</td>
532      <td align="center">yes</td>
533      <td align="center">-</td>
534    </tr>
535    <tr>
536      <td>&gt&gt, &lt&lt</td>
537      <td align="center">yes</td>
538      <td align="center">yes</td>
539      <td align="center">yes</td>
540      <td align="center">-</td>
541    </tr>
542    <tr>
543      <td>!, &&,||</td>
544      <td align="center">no</td>
545      <td align="center">-</td>
546      <td align="center">-</td>
547      <td align="center">-</td>
548    </tr>
549    <tr>
550      <td>==,!=, >, <, >=, <=</td>
551      <td align="center">yes</td>
552      <td align="center">yes</td>
553      <td align="center">-</td>
554      <td align="center">-</td>
555    </tr>
556    <tr>
557      <td>=</td>
558      <td align="center">yes</td>
559      <td align="center">yes</td>
560      <td align="center">yes</td>
561      <td align="center">yes</td>
562    </tr>
563    <tr>
564      <td>:?</td>
565      <td align="center">yes</td>
566      <td align="center">-</td>
567      <td align="center">-</td>
568      <td align="center">-</td>
569    </tr>
570    <tr>
571      <td>sizeof</td>
572      <td align="center">yes</td>
573      <td align="center">yes</td>
574      <td align="center">yes</td>
575      <td align="center">yes</td>
576    </tr>
577</table>
578
579<p>See also <a href="#__builtin_shufflevector">__builtin_shufflevector</a>.</p>
580
581<!-- ======================================================================= -->
582<h2 id="deprecated">Messages on <tt>deprecated</tt> and <tt>unavailable</tt> Attributes</h2>
583<!-- ======================================================================= -->
584
585<p>An optional string message can be added to the <tt>deprecated</tt>
586and <tt>unavailable</tt> attributes.  For example:</p>
587
588<blockquote>
589<pre>void explode(void) __attribute__((deprecated("extremely unsafe, use 'combust' instead!!!")));</pre>
590</blockquote>
591
592<p>If the deprecated or unavailable declaration is used, the message
593will be incorporated into the appropriate diagnostic:</p>
594
595<blockquote>
596<pre>harmless.c:4:3: warning: 'explode' is deprecated: extremely unsafe, use 'combust' instead!!!
597      [-Wdeprecated-declarations]
598  explode();
599  ^</pre>
600</blockquote>
601
602<p>Query for this feature
603with <tt>__has_extension(attribute_deprecated_with_message)</tt>
604and <tt>__has_extension(attribute_unavailable_with_message)</tt>.</p>
605
606<!-- ======================================================================= -->
607<h2 id="attributes-on-enumerators">Attributes on Enumerators</h2>
608<!-- ======================================================================= -->
609
610<p>Clang allows attributes to be written on individual enumerators.
611This allows enumerators to be deprecated, made unavailable, etc.  The
612attribute must appear after the enumerator name and before any
613initializer, like so:</p>
614
615<blockquote>
616<pre>enum OperationMode {
617  OM_Invalid,
618  OM_Normal,
619  OM_Terrified __attribute__((deprecated)),
620  OM_AbortOnError __attribute__((deprecated)) = 4
621};</pre>
622</blockquote>
623
624<p>Attributes on the <tt>enum</tt> declaration do not apply to
625individual enumerators.</p>
626
627<p>Query for this feature with <tt>__has_extension(enumerator_attributes)</tt>.</p>
628
629<!-- ======================================================================= -->
630<h2 id="user_specified_system_framework">'User-Specified' System Frameworks</h2>
631<!-- ======================================================================= -->
632
633<p>Clang provides a mechanism by which frameworks can be built in such a way
634that they will always be treated as being 'system frameworks', even if they are
635not present in a system framework directory. This can be useful to system
636framework developers who want to be able to test building other applications
637with development builds of their framework, including the manner in which the
638compiler changes warning behavior for system headers.</p>
639
640<p>Framework developers can opt-in to this mechanism by creating a
641'.system_framework' file at the top-level of their framework. That is, the
642framework should have contents like:</p>
643
644<pre>
645 .../TestFramework.framework
646 .../TestFramework.framework/.system_framework
647 .../TestFramework.framework/Headers
648 .../TestFramework.framework/Headers/TestFramework.h
649 ...
650</pre>
651
652<p>Clang will treat the presence of this file as an indicator that the framework
653should be treated as a system framework, regardless of how it was found in the
654framework search path. For consistency, we recommend that such files never be
655included in installed versions of the framework.</p>
656
657<!-- ======================================================================= -->
658<h2 id="availability">Availability attribute</h2
659<!-- ======================================================================= -->
660
661<p>Clang introduces the <code>availability</code> attribute, which can
662be placed on declarations to describe the lifecycle of that
663declaration relative to operating system versions. Consider the function declaration for a hypothetical function <code>f</code>:</p>
664
665<pre>
666void f(void) __attribute__((availability(macosx,introduced=10.4,deprecated=10.6,obsoleted=10.7)));
667</pre>
668
669<p>The availability attribute states that <code>f</code> was introduced in Mac OS X 10.4, deprecated in Mac OS X 10.6, and obsoleted in Mac OS X 10.7. This information is used by Clang to determine when it is safe to use <code>f</code>: for example, if Clang is instructed to compile code for Mac OS X 10.5, a call to <code>f()</code> succeeds. If Clang is instructed to compile code for Mac OS X 10.6, the call succeeds but Clang emits a warning specifying that the function is deprecated. Finally, if Clang is instructed to compile code for Mac OS X 10.7, the call fails because <code>f()</code> is no longer available.</p>
670
671<p>The availablility attribute is a comma-separated list starting with the platform name and then including clauses specifying important milestones in the declaration's lifetime (in any order) along with additional information. Those clauses can be:</p>
672
673<dl>
674  <dt>introduced=<i>version</i></dt>
675  <dd>The first version in which this declaration was introduced.</dd>
676
677  <dt>deprecated=<i>version</i></dt>
678  <dd>The first version in which this declaration was deprecated, meaning that users should migrate away from this API.</dd>
679
680  <dt>obsoleted=<i>version</i></dt>
681  <dd>The first version in which this declaration was obsoleted, meaning that it was removed completely and can no longer be used.</dd>
682
683  <dt>unavailable</dt>
684  <dd>This declaration is never available on this platform.</dd>
685
686  <dt>message=<i>string-literal</i></dt>
687  <dd>Additional message text that Clang will provide when emitting a warning or error about use of a deprecated or obsoleted declaration. Useful to direct users to replacement APIs.</dd>
688</dl>
689
690<p>Multiple availability attributes can be placed on a declaration, which may correspond to different platforms. Only the availability attribute with the platform corresponding to the target platform will be used; any others will be ignored. If no availability attribute specifies availability for the current target platform, the availability attributes are ignored. Supported platforms are:</p>
691
692<dl>
693  <dt>ios</dt>
694  <dd>Apple's iOS operating system. The minimum deployment target is specified by the <code>-mios-version-min=<i>version</i></code> or <code>-miphoneos-version-min=<i>version</i></code> command-line arguments.</dd>
695
696  <dt>macosx</dt>
697  <dd>Apple's Mac OS X operating system. The minimum deployment target is specified by the <code>-mmacosx-version-min=<i>version</i></code> command-line argument.</dd>
698</dl>
699
700<p>A declaration can be used even when deploying back to a platform
701version prior to when the declaration was introduced. When this
702happens, the declaration is <a
703 href="https://developer.apple.com/library/mac/#documentation/MacOSX/Conceptual/BPFrameworks/Concepts/WeakLinking.html">weakly
704linked</a>, as if the <code>weak_import</code> attribute were added to the declaration. A weakly-linked declaration may or may not be present a run-time, and a program can determine whether the declaration is present by checking whether the address of that declaration is non-NULL.</p>
705
706<!-- ======================================================================= -->
707<h2 id="checking_language_features">Checks for Standard Language Features</h2>
708<!-- ======================================================================= -->
709
710<p>The <tt>__has_feature</tt> macro can be used to query if certain standard
711language features are enabled.  The <tt>__has_extension</tt> macro can be used
712to query if language features are available as an extension when compiling for
713a standard which does not provide them. The features which can be tested are
714listed here.</p>
715
716<h3 id="cxx98">C++98</h3>
717
718<p>The features listed below are part of the C++98 standard. These features are
719enabled by default when compiling C++ code.</p>
720
721<h4 id="cxx_exceptions">C++ exceptions</h4>
722
723<p>Use <tt>__has_feature(cxx_exceptions)</tt> to determine if C++ exceptions have been enabled. For
724example, compiling code with <tt>-fno-exceptions</tt> disables C++ exceptions.</p>
725
726<h4 id="cxx_rtti">C++ RTTI</h4>
727
728<p>Use <tt>__has_feature(cxx_rtti)</tt> to determine if C++ RTTI has been enabled. For example,
729compiling code with <tt>-fno-rtti</tt> disables the use of RTTI.</p>
730
731<h3 id="cxx11">C++11</h3>
732
733<p>The features listed below are part of the C++11 standard. As a result, all
734these features are enabled with the <tt>-std=c++11</tt> or <tt>-std=gnu++11</tt>
735option when compiling C++ code.</p>
736
737<h4 id="cxx_access_control_sfinae">C++11 SFINAE includes access control</h4>
738
739<p>Use <tt>__has_feature(cxx_access_control_sfinae)</tt> or <tt>__has_extension(cxx_access_control_sfinae)</tt> to determine whether access-control errors (e.g., calling a private constructor) are considered to be template argument deduction errors (aka SFINAE errors), per <a href="http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#1170">C++ DR1170</a>.</p>
740
741<h4 id="cxx_alias_templates">C++11 alias templates</h4>
742
743<p>Use <tt>__has_feature(cxx_alias_templates)</tt> or
744<tt>__has_extension(cxx_alias_templates)</tt> to determine if support for
745C++11's alias declarations and alias templates is enabled.</p>
746
747<h4 id="cxx_alignas">C++11 alignment specifiers</h4>
748
749<p>Use <tt>__has_feature(cxx_alignas)</tt> or
750<tt>__has_extension(cxx_alignas)</tt> to determine if support for alignment
751specifiers using <tt>alignas</tt> is enabled.</p>
752
753<h4 id="cxx_attributes">C++11 attributes</h4>
754
755<p>Use <tt>__has_feature(cxx_attributes)</tt> or
756<tt>__has_extension(cxx_attributes)</tt> to determine if support for attribute
757parsing with C++11's square bracket notation is enabled.</p>
758
759<h4 id="cxx_constexpr">C++11 generalized constant expressions</h4>
760
761<p>Use <tt>__has_feature(cxx_constexpr)</tt> to determine if support
762for generalized constant expressions (e.g., <tt>constexpr</tt>) is
763enabled.</p>
764
765<h4 id="cxx_decltype">C++11 <tt>decltype()</tt></h4>
766
767<p>Use <tt>__has_feature(cxx_decltype)</tt> or
768<tt>__has_extension(cxx_decltype)</tt> to determine if support for the
769<tt>decltype()</tt> specifier is enabled. C++11's <tt>decltype</tt>
770does not require type-completeness of a function call expression.
771Use <tt>__has_feature(cxx_decltype_incomplete_return_types)</tt>
772or <tt>__has_extension(cxx_decltype_incomplete_return_types)</tt>
773to determine if support for this feature is enabled.</p>
774
775<h4 id="cxx_default_function_template_args">C++11 default template arguments in function templates</h4>
776
777<p>Use <tt>__has_feature(cxx_default_function_template_args)</tt> or
778<tt>__has_extension(cxx_default_function_template_args)</tt> to determine
779if support for default template arguments in function templates is enabled.</p>
780
781<h4 id="cxx_defaulted_functions">C++11 <tt>default</tt>ed functions</h4>
782
783<p>Use <tt>__has_feature(cxx_defaulted_functions)</tt> or
784<tt>__has_extension(cxx_defaulted_functions)</tt> to determine if support for
785defaulted function definitions (with <tt>= default</tt>) is enabled.</p>
786
787<h4 id="cxx_delegating_constructors">C++11 delegating constructors</h4>
788
789<p>Use <tt>__has_feature(cxx_delegating_constructors)</tt> to determine if
790support for delegating constructors is enabled.</p>
791
792<h4 id="cxx_deleted_functions">C++11 <tt>delete</tt>d functions</h4>
793
794<p>Use <tt>__has_feature(cxx_deleted_functions)</tt> or
795<tt>__has_extension(cxx_deleted_functions)</tt> to determine if support for
796deleted function definitions (with <tt>= delete</tt>) is enabled.</p>
797
798<h4 id="cxx_explicit_conversions">C++11 explicit conversion functions</h4>
799<p>Use <tt>__has_feature(cxx_explicit_conversions)</tt> to determine if support for <tt>explicit</tt> conversion functions is enabled.</p>
800
801<h4 id="cxx_generalized_initializers">C++11 generalized initializers</h4>
802
803<p>Use <tt>__has_feature(cxx_generalized_initializers)</tt> to determine if
804support for generalized initializers (using braced lists and
805<tt>std::initializer_list</tt>) is enabled.</p>
806
807<h4 id="cxx_implicit_moves">C++11 implicit move constructors/assignment operators</h4>
808
809<p>Use <tt>__has_feature(cxx_implicit_moves)</tt> to determine if Clang will
810implicitly generate move constructors and move assignment operators where needed.</p>
811
812<h4 id="cxx_inheriting_constructors">C++11 inheriting constructors</h4>
813
814<p>Use <tt>__has_feature(cxx_inheriting_constructors)</tt> to determine if support for inheriting constructors is enabled. Clang does not currently implement this feature.</p>
815
816<h4 id="cxx_inline_namespaces">C++11 inline namespaces</h4>
817
818<p>Use <tt>__has_feature(cxx_inline_namespaces)</tt> or
819<tt>__has_extension(cxx_inline_namespaces)</tt> to determine if support for
820inline namespaces is enabled.</p>
821
822<h4 id="cxx_lambdas">C++11 lambdas</h4>
823
824<p>Use <tt>__has_feature(cxx_lambdas)</tt> or
825<tt>__has_extension(cxx_lambdas)</tt> to determine if support for lambdas
826is enabled. </p>
827
828<h4 id="cxx_local_type_template_args">C++11 local and unnamed types as template arguments</h4>
829
830<p>Use <tt>__has_feature(cxx_local_type_template_args)</tt> or
831<tt>__has_extension(cxx_local_type_template_args)</tt> to determine if
832support for local and unnamed types as template arguments is enabled.</p>
833
834<h4 id="cxx_noexcept">C++11 noexcept</h4>
835
836<p>Use <tt>__has_feature(cxx_noexcept)</tt> or
837<tt>__has_extension(cxx_noexcept)</tt> to determine if support for noexcept
838exception specifications is enabled.</p>
839
840<h4 id="cxx_nonstatic_member_init">C++11 in-class non-static data member initialization</h4>
841
842<p>Use <tt>__has_feature(cxx_nonstatic_member_init)</tt> to determine whether in-class initialization of non-static data members is enabled.</p>
843
844<h4 id="cxx_nullptr">C++11 <tt>nullptr</tt></h4>
845
846<p>Use <tt>__has_feature(cxx_nullptr)</tt> or
847<tt>__has_extension(cxx_nullptr)</tt> to determine if support for
848<tt>nullptr</tt> is enabled.</p>
849
850<h4 id="cxx_override_control">C++11 <tt>override control</tt></h4>
851
852<p>Use <tt>__has_feature(cxx_override_control)</tt> or
853<tt>__has_extension(cxx_override_control)</tt> to determine if support for
854the override control keywords is enabled.</p>
855
856<h4 id="cxx_reference_qualified_functions">C++11 reference-qualified functions</h4>
857<p>Use <tt>__has_feature(cxx_reference_qualified_functions)</tt> or
858<tt>__has_extension(cxx_reference_qualified_functions)</tt> to determine
859if support for reference-qualified functions (e.g., member functions with
860<code>&amp;</code> or <code>&amp;&amp;</code> applied to <code>*this</code>)
861is enabled.</p>
862
863<h4 id="cxx_range_for">C++11 range-based <tt>for</tt> loop</h4>
864
865<p>Use <tt>__has_feature(cxx_range_for)</tt> or
866<tt>__has_extension(cxx_range_for)</tt> to determine if support for the
867range-based for loop is enabled. </p>
868
869<h4 id="cxx_raw_string_literals">C++11 raw string literals</h4>
870<p>Use <tt>__has_feature(cxx_raw_string_literals)</tt> to determine if support
871for raw string literals (e.g., <tt>R"x(foo\bar)x"</tt>) is enabled.</p>
872
873<h4 id="cxx_rvalue_references">C++11 rvalue references</h4>
874
875<p>Use <tt>__has_feature(cxx_rvalue_references)</tt> or
876<tt>__has_extension(cxx_rvalue_references)</tt> to determine if support for
877rvalue references is enabled. </p>
878
879<h4 id="cxx_static_assert">C++11 <tt>static_assert()</tt></h4>
880
881<p>Use <tt>__has_feature(cxx_static_assert)</tt> or
882<tt>__has_extension(cxx_static_assert)</tt> to determine if support for
883compile-time assertions using <tt>static_assert</tt> is enabled.</p>
884
885<h4 id="cxx_auto_type">C++11 type inference</h4>
886
887<p>Use <tt>__has_feature(cxx_auto_type)</tt> or
888<tt>__has_extension(cxx_auto_type)</tt> to determine C++11 type inference is
889supported using the <tt>auto</tt> specifier. If this is disabled, <tt>auto</tt>
890will instead be a storage class specifier, as in C or C++98.</p>
891
892<h4 id="cxx_strong_enums">C++11 strongly typed enumerations</h4>
893
894<p>Use <tt>__has_feature(cxx_strong_enums)</tt> or
895<tt>__has_extension(cxx_strong_enums)</tt> to determine if support for
896strongly typed, scoped enumerations is enabled.</p>
897
898<h4 id="cxx_trailing_return">C++11 trailing return type</h4>
899
900<p>Use <tt>__has_feature(cxx_trailing_return)</tt> or
901<tt>__has_extension(cxx_trailing_return)</tt> to determine if support for the
902alternate function declaration syntax with trailing return type is enabled.</p>
903
904<h4 id="cxx_unicode_literals">C++11 Unicode string literals</h4>
905<p>Use <tt>__has_feature(cxx_unicode_literals)</tt> to determine if
906support for Unicode string literals is enabled.</p>
907
908<h4 id="cxx_unrestricted_unions">C++11 unrestricted unions</h4>
909
910<p>Use <tt>__has_feature(cxx_unrestricted_unions)</tt> to determine if support for unrestricted unions is enabled.</p>
911
912<h4 id="cxx_user_literals">C++11 user-defined literals</h4>
913
914<p>Use <tt>__has_feature(cxx_user_literals)</tt> to determine if support for user-defined literals is enabled.</p>
915
916<h4 id="cxx_variadic_templates">C++11 variadic templates</h4>
917
918<p>Use <tt>__has_feature(cxx_variadic_templates)</tt> or
919<tt>__has_extension(cxx_variadic_templates)</tt> to determine if support
920for variadic templates is enabled.</p>
921
922<h3 id="c11">C11</h3>
923
924<p>The features listed below are part of the C11 standard. As a result, all
925these features are enabled with the <tt>-std=c11</tt> or <tt>-std=gnu11</tt>
926option when compiling C code. Additionally, because these features are all
927backward-compatible, they are available as extensions in all language modes.</p>
928
929<h4 id="c_alignas">C11 alignment specifiers</h4>
930
931<p>Use <tt>__has_feature(c_alignas)</tt> or <tt>__has_extension(c_alignas)</tt>
932to determine if support for alignment specifiers using <tt>_Alignas</tt>
933is enabled.</p>
934
935<h4 id="c_atomic">C11 atomic operations</h4>
936
937<p>Use <tt>__has_feature(c_atomic)</tt> or <tt>__has_extension(c_atomic)</tt>
938to determine if support for atomic types using <tt>_Atomic</tt> is enabled.
939Clang also provides <a href="#__c11_atomic">a set of builtins</a> which can be
940used to implement the <tt>&lt;stdatomic.h&gt;</tt> operations on _Atomic
941types.</p>
942
943<h4 id="c_generic_selections">C11 generic selections</h4>
944
945<p>Use <tt>__has_feature(c_generic_selections)</tt> or
946<tt>__has_extension(c_generic_selections)</tt> to determine if support for
947generic selections is enabled.</p>
948
949<p>As an extension, the C11 generic selection expression is available in all
950languages supported by Clang.  The syntax is the same as that given in the
951C11 standard.</p>
952
953<p>In C, type compatibility is decided according to the rules given in the
954appropriate standard, but in C++, which lacks the type compatibility rules
955used in C, types are considered compatible only if they are equivalent.</p>
956
957<h4 id="c_static_assert">C11 <tt>_Static_assert()</tt></h4>
958
959<p>Use <tt>__has_feature(c_static_assert)</tt> or
960<tt>__has_extension(c_static_assert)</tt> to determine if support for
961compile-time assertions using <tt>_Static_assert</tt> is enabled.</p>
962
963<!-- ======================================================================= -->
964<h2 id="checking_type_traits">Checks for Type Traits</h2>
965<!-- ======================================================================= -->
966
967<p>Clang supports the <a href="http://gcc.gnu.org/onlinedocs/gcc/Type-Traits.html">GNU C++ type traits</a> and a subset of the <a href="http://msdn.microsoft.com/en-us/library/ms177194(v=VS.100).aspx">Microsoft Visual C++ Type traits</a>. For each supported type trait <code>__X</code>, <code>__has_extension(X)</code> indicates the presence of the type trait. For example:
968<blockquote>
969<pre>
970#if __has_extension(is_convertible_to)
971template&lt;typename From, typename To&gt;
972struct is_convertible_to {
973  static const bool value = __is_convertible_to(From, To);
974};
975#else
976// Emulate type trait
977#endif
978</pre>
979</blockquote>
980
981<p>The following type traits are supported by Clang:</p>
982<ul>
983  <li><code>__has_nothrow_assign</code> (GNU, Microsoft)</li>
984  <li><code>__has_nothrow_copy</code> (GNU, Microsoft)</li>
985  <li><code>__has_nothrow_constructor</code> (GNU, Microsoft)</li>
986  <li><code>__has_trivial_assign</code> (GNU, Microsoft)</li>
987  <li><code>__has_trivial_copy</code> (GNU, Microsoft)</li>
988  <li><code>__has_trivial_constructor</code> (GNU, Microsoft)</li>
989  <li><code>__has_trivial_destructor</code> (GNU, Microsoft)</li>
990  <li><code>__has_virtual_destructor</code> (GNU, Microsoft)</li>
991  <li><code>__is_abstract</code> (GNU, Microsoft)</li>
992  <li><code>__is_base_of</code> (GNU, Microsoft)</li>
993  <li><code>__is_class</code> (GNU, Microsoft)</li>
994  <li><code>__is_convertible_to</code> (Microsoft)</li>
995  <li><code>__is_empty</code> (GNU, Microsoft)</li>
996  <li><code>__is_enum</code> (GNU, Microsoft)</li>
997  <li><code>__is_pod</code> (GNU, Microsoft)</li>
998  <li><code>__is_polymorphic</code> (GNU, Microsoft)</li>
999  <li><code>__is_union</code> (GNU, Microsoft)</li>
1000  <li><code>__is_literal(type)</code>: Determines whether the given type is a literal type</li>
1001  <li><code>__is_final</code>: Determines whether the given type is declared with a <code>final</code> class-virt-specifier.</li>
1002  <li><code>__underlying_type(type)</code>: Retrieves the underlying type for a given <code>enum</code> type. This trait is required to implement the C++11 standard library.</li>
1003  <li><code>__is_trivially_assignable(totype, fromtype)</code>: Determines whether a value of type <tt>totype</tt> can be assigned to from a value of type <tt>fromtype</tt> such that no non-trivial functions are called as part of that assignment. This trait is required to implement the C++11 standard library.</li>
1004  <li><code>__is_trivially_constructible(type, argtypes...)</code>: Determines whether a value of type <tt>type</tt> can be direct-initialized with arguments of types <tt>argtypes...</tt> such that no non-trivial functions are called as part of that initialization. This trait is required to implement the C++11 standard library.</li>
1005</ul>
1006
1007<!-- ======================================================================= -->
1008<h2 id="blocks">Blocks</h2>
1009<!-- ======================================================================= -->
1010
1011<p>The syntax and high level language feature description is in <a
1012href="BlockLanguageSpec.txt">BlockLanguageSpec.txt</a>.  Implementation and ABI
1013details for the clang implementation are in <a
1014href="Block-ABI-Apple.txt">Block-ABI-Apple.txt</a>.</p>
1015
1016
1017<p>Query for this feature with __has_extension(blocks).</p>
1018
1019<!-- ======================================================================= -->
1020<h2 id="objc_features">Objective-C Features</h2>
1021<!-- ======================================================================= -->
1022
1023<h3 id="objc_instancetype">Related result types</h3>
1024
1025<p>According to Cocoa conventions, Objective-C methods with certain names ("init", "alloc", etc.) always return objects that are an instance of the receiving class's type. Such methods are said to have a "related result type", meaning that a message send to one of these methods will have the same static type as an instance of the receiver class. For example, given the following classes:</p>
1026
1027<blockquote>
1028<pre>
1029@interface NSObject
1030+ (id)alloc;
1031- (id)init;
1032@end
1033
1034@interface NSArray : NSObject
1035@end
1036</pre>
1037</blockquote>
1038
1039<p>and this common initialization pattern</p>
1040
1041<blockquote>
1042<pre>
1043NSArray *array = [[NSArray alloc] init];
1044</pre>
1045</blockquote>
1046
1047<p>the type of the expression <code>[NSArray alloc]</code> is
1048<code>NSArray*</code> because <code>alloc</code> implicitly has a
1049related result type. Similarly, the type of the expression
1050<code>[[NSArray alloc] init]</code> is <code>NSArray*</code>, since
1051<code>init</code> has a related result type and its receiver is known
1052to have the type <code>NSArray *</code>. If neither <code>alloc</code> nor <code>init</code> had a related result type, the expressions would have had type <code>id</code>, as declared in the method signature.</p>
1053
1054<p>A method with a related result type can be declared by using the
1055type <tt>instancetype</tt> as its result type. <tt>instancetype</tt>
1056is a contextual keyword that is only permitted in the result type of
1057an Objective-C method, e.g.</p>
1058
1059<pre>
1060@interface A
1061+ (<b>instancetype</b>)constructAnA;
1062@end
1063</pre>
1064
1065<p>The related result type can also be inferred for some methods.
1066To determine whether a method has an inferred related result type, the first
1067word in the camel-case selector (e.g., "init" in "initWithObjects") is
1068considered, and the method will have a related result type if its return
1069type is compatible with the type of its class and if</p>
1070
1071<ul>
1072
1073  <li>the first word is "alloc" or "new", and the method is a class
1074  method, or</li>
1075
1076  <li>the first word is "autorelease", "init", "retain", or "self",
1077  and the method is an instance method.</li>
1078
1079</ul>
1080
1081<p>If a method with a related result type is overridden by a subclass
1082method, the subclass method must also return a type that is compatible
1083with the subclass type. For example:</p>
1084
1085<blockquote>
1086<pre>
1087@interface NSString : NSObject
1088- (NSUnrelated *)init; // incorrect usage: NSUnrelated is not NSString or a superclass of NSString
1089@end
1090</pre>
1091</blockquote>
1092
1093<p>Related result types only affect the type of a message send or
1094property access via the given method. In all other respects, a method
1095with a related result type is treated the same way as method that
1096returns <tt>id</tt>.</p>
1097
1098<p>Use <tt>__has_feature(objc_instancetype)</tt> to determine whether
1099the <tt>instancetype</tt> contextual keyword is available.</p>
1100
1101<!-- ======================================================================= -->
1102<h2 id="objc_arc">Automatic reference counting </h2>
1103<!-- ======================================================================= -->
1104
1105<p>Clang provides support for <a href="AutomaticReferenceCounting.html">automated reference counting</a> in Objective-C, which eliminates the need for manual retain/release/autorelease message sends. There are two feature macros associated with automatic reference counting: <code>__has_feature(objc_arc)</code> indicates the availability of automated reference counting in general, while <code>__has_feature(objc_arc_weak)</code> indicates that automated reference counting also includes support for <code>__weak</code> pointers to Objective-C objects.</p>
1106
1107<!-- ======================================================================= -->
1108<h2 id="objc_fixed_enum">Enumerations with a fixed underlying type</h2>
1109<!-- ======================================================================= -->
1110
1111<p>Clang provides support for C++11 enumerations with a fixed
1112underlying type within Objective-C. For example, one can write an
1113enumeration type as:</p>
1114
1115<pre>
1116typedef enum : unsigned char { Red, Green, Blue } Color;
1117</pre>
1118
1119<p>This specifies that the underlying type, which is used to store the
1120enumeration value, is <tt>unsigned char</tt>.</p>
1121
1122<p>Use <tt>__has_feature(objc_fixed_enum)</tt> to determine whether
1123support for fixed underlying types is available in Objective-C.</p>
1124
1125<!-- ======================================================================= -->
1126<h2 id="objc_lambdas">Interoperability with C++11 lambdas</h2>
1127<!-- ======================================================================= -->
1128
1129<p>Clang provides interoperability between C++11 lambdas and
1130blocks-based APIs, by permitting a lambda to be implicitly converted
1131to a block pointer with the corresponding signature. For example,
1132consider an API such as <code>NSArray</code>'s array-sorting
1133method:</p>
1134
1135<pre> - (NSArray *)sortedArrayUsingComparator:(NSComparator)cmptr; </pre>
1136
1137<p><code>NSComparator</code> is simply a typedef for the block pointer
1138<code>NSComparisonResult (^)(id, id)</code>, and parameters of this
1139type are generally provided with block literals as arguments. However,
1140one can also use a C++11 lambda so long as it provides the same
1141signature (in this case, accepting two parameters of type
1142<code>id</code> and returning an <code>NSComparisonResult</code>):</p>
1143
1144<pre>
1145  NSArray *array = @[@"string 1", @"string 21", @"string 12", @"String 11",
1146                     @"String 02"];
1147  const NSStringCompareOptions comparisonOptions
1148    = NSCaseInsensitiveSearch | NSNumericSearch |
1149      NSWidthInsensitiveSearch | NSForcedOrderingSearch;
1150  NSLocale *currentLocale = [NSLocale currentLocale];
1151  NSArray *sorted
1152    = [array sortedArrayUsingComparator:<b>[=](id s1, id s2) -&gt; NSComparisonResult {
1153               NSRange string1Range = NSMakeRange(0, [s1 length]);
1154               return [s1 compare:s2 options:comparisonOptions
1155                          range:string1Range locale:currentLocale];
1156       }</b>];
1157  NSLog(@"sorted: %@", sorted);
1158</pre>
1159
1160<p>This code relies on an implicit conversion from the type of the
1161lambda expression (an unnamed, local class type called the <i>closure
1162type</i>) to the corresponding block pointer type. The conversion
1163itself is expressed by a conversion operator in that closure type
1164that produces a block pointer with the same signature as the lambda
1165itself, e.g.,</p>
1166
1167<pre>
1168  operator NSComparisonResult (^)(id, id)() const;
1169</pre>
1170
1171<p>This conversion function returns a new block that simply forwards
1172the two parameters to the lambda object (which it captures by copy),
1173then returns the result. The returned block is first copied (with
1174<tt>Block_copy</tt>) and then autoreleased. As an optimization, if a
1175lambda expression is immediately converted to a block pointer (as in
1176the first example, above), then the block is not copied and
1177autoreleased: rather, it is given the same lifetime as a block literal
1178written at that point in the program, which avoids the overhead of
1179copying a block to the heap in the common case.</p>
1180
1181<p>The conversion from a lambda to a block pointer is only available
1182in Objective-C++, and not in C++ with blocks, due to its use of
1183Objective-C memory management (autorelease).</p>
1184
1185<!-- ======================================================================= -->
1186<h2 id="objc_object_literals_subscripting">Object Literals and Subscripting</h2>
1187<!-- ======================================================================= -->
1188
1189<p>Clang provides support for <a href="ObjectiveCLiterals.html">Object Literals
1190and Subscripting</a> in Objective-C, which simplifies common Objective-C
1191programming patterns, makes programs more concise, and improves the safety of
1192container creation. There are several feature macros associated with object
1193literals and subscripting: <code>__has_feature(objc_array_literals)</code>
1194tests the availability of array literals;
1195<code>__has_feature(objc_dictionary_literals)</code> tests the availability of
1196dictionary literals; <code>__has_feature(objc_subscripting)</code> tests the
1197availability of object subscripting.</p>
1198
1199<!-- ======================================================================= -->
1200<h2 id="overloading-in-c">Function Overloading in C</h2>
1201<!-- ======================================================================= -->
1202
1203<p>Clang provides support for C++ function overloading in C. Function
1204overloading in C is introduced using the <tt>overloadable</tt> attribute. For
1205example, one might provide several overloaded versions of a <tt>tgsin</tt>
1206function that invokes the appropriate standard function computing the sine of a
1207value with <tt>float</tt>, <tt>double</tt>, or <tt>long double</tt>
1208precision:</p>
1209
1210<blockquote>
1211<pre>
1212#include &lt;math.h&gt;
1213float <b>__attribute__((overloadable))</b> tgsin(float x) { return sinf(x); }
1214double <b>__attribute__((overloadable))</b> tgsin(double x) { return sin(x); }
1215long double <b>__attribute__((overloadable))</b> tgsin(long double x) { return sinl(x); }
1216</pre>
1217</blockquote>
1218
1219<p>Given these declarations, one can call <tt>tgsin</tt> with a
1220<tt>float</tt> value to receive a <tt>float</tt> result, with a
1221<tt>double</tt> to receive a <tt>double</tt> result, etc. Function
1222overloading in C follows the rules of C++ function overloading to pick
1223the best overload given the call arguments, with a few C-specific
1224semantics:</p>
1225<ul>
1226  <li>Conversion from <tt>float</tt> or <tt>double</tt> to <tt>long
1227  double</tt> is ranked as a floating-point promotion (per C99) rather
1228  than as a floating-point conversion (as in C++).</li>
1229
1230  <li>A conversion from a pointer of type <tt>T*</tt> to a pointer of type
1231  <tt>U*</tt> is considered a pointer conversion (with conversion
1232  rank) if <tt>T</tt> and <tt>U</tt> are compatible types.</li>
1233
1234  <li>A conversion from type <tt>T</tt> to a value of type <tt>U</tt>
1235  is permitted if <tt>T</tt> and <tt>U</tt> are compatible types. This
1236  conversion is given "conversion" rank.</li>
1237</ul>
1238
1239<p>The declaration of <tt>overloadable</tt> functions is restricted to
1240function declarations and definitions. Most importantly, if any
1241function with a given name is given the <tt>overloadable</tt>
1242attribute, then all function declarations and definitions with that
1243name (and in that scope) must have the <tt>overloadable</tt>
1244attribute. This rule even applies to redeclarations of functions whose original
1245declaration had the <tt>overloadable</tt> attribute, e.g.,</p>
1246
1247<blockquote>
1248<pre>
1249int f(int) __attribute__((overloadable));
1250float f(float); <i>// error: declaration of "f" must have the "overloadable" attribute</i>
1251
1252int g(int) __attribute__((overloadable));
1253int g(int) { } <i>// error: redeclaration of "g" must also have the "overloadable" attribute</i>
1254</pre>
1255</blockquote>
1256
1257<p>Functions marked <tt>overloadable</tt> must have
1258prototypes. Therefore, the following code is ill-formed:</p>
1259
1260<blockquote>
1261<pre>
1262int h() __attribute__((overloadable)); <i>// error: h does not have a prototype</i>
1263</pre>
1264</blockquote>
1265
1266<p>However, <tt>overloadable</tt> functions are allowed to use a
1267ellipsis even if there are no named parameters (as is permitted in C++). This feature is particularly useful when combined with the <tt>unavailable</tt> attribute:</p>
1268
1269<blockquote>
1270<pre>
1271void honeypot(...) __attribute__((overloadable, unavailable)); <i>// calling me is an error</i>
1272</pre>
1273</blockquote>
1274
1275<p>Functions declared with the <tt>overloadable</tt> attribute have
1276their names mangled according to the same rules as C++ function
1277names. For example, the three <tt>tgsin</tt> functions in our
1278motivating example get the mangled names <tt>_Z5tgsinf</tt>,
1279<tt>_Z5tgsind</tt>, and <tt>_Z5tgsine</tt>, respectively. There are two
1280caveats to this use of name mangling:</p>
1281
1282<ul>
1283
1284  <li>Future versions of Clang may change the name mangling of
1285  functions overloaded in C, so you should not depend on an specific
1286  mangling. To be completely safe, we strongly urge the use of
1287  <tt>static inline</tt> with <tt>overloadable</tt> functions.</li>
1288
1289  <li>The <tt>overloadable</tt> attribute has almost no meaning when
1290  used in C++, because names will already be mangled and functions are
1291  already overloadable. However, when an <tt>overloadable</tt>
1292  function occurs within an <tt>extern "C"</tt> linkage specification,
1293  it's name <i>will</i> be mangled in the same way as it would in
1294  C.</li>
1295</ul>
1296
1297<p>Query for this feature with __has_extension(attribute_overloadable).</p>
1298
1299<!-- ======================================================================= -->
1300<h2 id="complex-list-init">Initializer lists for complex numbers in C</h2>
1301<!-- ======================================================================= -->
1302
1303<p>clang supports an extension which allows the following in C:</p>
1304
1305<blockquote>
1306<pre>
1307#include &lt;math.h&gt;
1308#include &lt;complex.h&gt;
1309complex float x = { 1.0f, INFINITY }; // Init to (1, Inf)
1310</pre>
1311</blockquote>
1312
1313<p>This construct is useful because there is no way to separately
1314initialize the real and imaginary parts of a complex variable in
1315standard C, given that clang does not support <code>_Imaginary</code>.
1316(clang also supports the <code>__real__</code> and <code>__imag__</code>
1317extensions from gcc, which help in some cases, but are not usable in
1318static initializers.)
1319
1320<p>Note that this extension does not allow eliding the braces; the
1321meaning of the following two lines is different:</p>
1322
1323<blockquote>
1324<pre>
1325complex float x[] = { { 1.0f, 1.0f } }; // [0] = (1, 1)
1326complex float x[] = { 1.0f, 1.0f }; // [0] = (1, 0), [1] = (1, 0)
1327</pre>
1328</blockquote>
1329
1330<p>This extension also works in C++ mode, as far as that goes, but does not
1331    apply to the C++ <code>std::complex</code>.  (In C++11, list
1332    initialization allows the same syntax to be used with
1333    <code>std::complex</code> with the same meaning.)
1334
1335<!-- ======================================================================= -->
1336<h2 id="builtins">Builtin Functions</h2>
1337<!-- ======================================================================= -->
1338
1339<p>Clang supports a number of builtin library functions with the same syntax as
1340GCC, including things like <tt>__builtin_nan</tt>,
1341<tt>__builtin_constant_p</tt>, <tt>__builtin_choose_expr</tt>,
1342<tt>__builtin_types_compatible_p</tt>, <tt>__sync_fetch_and_add</tt>, etc.  In
1343addition to the GCC builtins, Clang supports a number of builtins that GCC does
1344not, which are listed here.</p>
1345
1346<p>Please note that Clang does not and will not support all of the GCC builtins
1347for vector operations.  Instead of using builtins, you should use the functions
1348defined in target-specific header files like <tt>&lt;xmmintrin.h&gt;</tt>, which
1349define portable wrappers for these.  Many of the Clang versions of these
1350functions are implemented directly in terms of <a href="#vectors">extended
1351vector support</a> instead of builtins, in order to reduce the number of
1352builtins that we need to implement.</p>
1353
1354<!-- ======================================================================= -->
1355<h3><a name="__builtin_shufflevector">__builtin_shufflevector</a></h3>
1356<!-- ======================================================================= -->
1357
1358<p><tt>__builtin_shufflevector</tt> is used to express generic vector
1359permutation/shuffle/swizzle operations. This builtin is also very important for
1360the implementation of various target-specific header files like
1361<tt>&lt;xmmintrin.h&gt;</tt>.
1362</p>
1363
1364<p><b>Syntax:</b></p>
1365
1366<pre>
1367__builtin_shufflevector(vec1, vec2, index1, index2, ...)
1368</pre>
1369
1370<p><b>Examples:</b></p>
1371
1372<pre>
1373  // Identity operation - return 4-element vector V1.
1374  __builtin_shufflevector(V1, V1, 0, 1, 2, 3)
1375
1376  // "Splat" element 0 of V1 into a 4-element result.
1377  __builtin_shufflevector(V1, V1, 0, 0, 0, 0)
1378
1379  // Reverse 4-element vector V1.
1380  __builtin_shufflevector(V1, V1, 3, 2, 1, 0)
1381
1382  // Concatenate every other element of 4-element vectors V1 and V2.
1383  __builtin_shufflevector(V1, V2, 0, 2, 4, 6)
1384
1385  // Concatenate every other element of 8-element vectors V1 and V2.
1386  __builtin_shufflevector(V1, V2, 0, 2, 4, 6, 8, 10, 12, 14)
1387</pre>
1388
1389<p><b>Description:</b></p>
1390
1391<p>The first two arguments to __builtin_shufflevector are vectors that have the
1392same element type.  The remaining arguments are a list of integers that specify
1393the elements indices of the first two vectors that should be extracted and
1394returned in a new vector.  These element indices are numbered sequentially
1395starting with the first vector, continuing into the second vector.  Thus, if
1396vec1 is a 4-element vector, index 5 would refer to the second element of vec2.
1397</p>
1398
1399<p>The result of __builtin_shufflevector is a vector
1400with the same element type as vec1/vec2 but that has an element count equal to
1401the number of indices specified.
1402</p>
1403
1404<p>Query for this feature with __has_builtin(__builtin_shufflevector).</p>
1405
1406<!-- ======================================================================= -->
1407<h3><a name="__builtin_unreachable">__builtin_unreachable</a></h3>
1408<!-- ======================================================================= -->
1409
1410<p><tt>__builtin_unreachable</tt> is used to indicate that a specific point in
1411the program cannot be reached, even if the compiler might otherwise think it
1412can.  This is useful to improve optimization and eliminates certain warnings.
1413For example, without the <tt>__builtin_unreachable</tt> in the example below,
1414the compiler assumes that the inline asm can fall through and prints a "function
1415declared 'noreturn' should not return" warning.
1416</p>
1417
1418<p><b>Syntax:</b></p>
1419
1420<pre>
1421__builtin_unreachable()
1422</pre>
1423
1424<p><b>Example of Use:</b></p>
1425
1426<pre>
1427void myabort(void) __attribute__((noreturn));
1428void myabort(void) {
1429    asm("int3");
1430    __builtin_unreachable();
1431}
1432</pre>
1433
1434<p><b>Description:</b></p>
1435
1436<p>The __builtin_unreachable() builtin has completely undefined behavior.  Since
1437it has undefined behavior, it is a statement that it is never reached and the
1438optimizer can take advantage of this to produce better code.  This builtin takes
1439no arguments and produces a void result.
1440</p>
1441
1442<p>Query for this feature with __has_builtin(__builtin_unreachable).</p>
1443
1444<!-- ======================================================================= -->
1445<h3><a name="__sync_swap">__sync_swap</a></h3>
1446<!-- ======================================================================= -->
1447
1448<p><tt>__sync_swap</tt> is used to atomically swap integers or pointers in
1449memory.
1450</p>
1451
1452<p><b>Syntax:</b></p>
1453
1454<pre>
1455<i>type</i> __sync_swap(<i>type</i> *ptr, <i>type</i> value, ...)
1456</pre>
1457
1458<p><b>Example of Use:</b></p>
1459
1460<pre>
1461int old_value = __sync_swap(&amp;value, new_value);
1462</pre>
1463
1464<p><b>Description:</b></p>
1465
1466<p>The __sync_swap() builtin extends the existing __sync_*() family of atomic
1467intrinsics to allow code to atomically swap the current value with the new
1468value.  More importantly, it helps developers write more efficient and correct
1469code by avoiding expensive loops around __sync_bool_compare_and_swap() or
1470relying on the platform specific implementation details of
1471__sync_lock_test_and_set(). The __sync_swap() builtin is a full barrier.
1472</p>
1473
1474<!-- ======================================================================= -->
1475<h3><a name="__c11_atomic">__c11_atomic builtins</a></h3>
1476<!-- ======================================================================= -->
1477
1478<p>Clang provides a set of builtins which are intended to be used to implement
1479C11's <tt>&lt;stdatomic.h&gt;</tt> header. These builtins provide the semantics
1480of the <tt>_explicit</tt> form of the corresponding C11 operation, and are named
1481with a <tt>__c11_</tt> prefix. The supported operations are:</p>
1482
1483<ul>
1484  <li><tt>__c11_atomic_init</tt></li>
1485  <li><tt>__c11_atomic_thread_fence</tt></li>
1486  <li><tt>__c11_atomic_signal_fence</tt></li>
1487  <li><tt>__c11_atomic_is_lock_free</tt></li>
1488  <li><tt>__c11_atomic_store</tt></li>
1489  <li><tt>__c11_atomic_load</tt></li>
1490  <li><tt>__c11_atomic_exchange</tt></li>
1491  <li><tt>__c11_atomic_compare_exchange_strong</tt></li>
1492  <li><tt>__c11_atomic_compare_exchange_weak</tt></li>
1493  <li><tt>__c11_atomic_fetch_add</tt></li>
1494  <li><tt>__c11_atomic_fetch_sub</tt></li>
1495  <li><tt>__c11_atomic_fetch_and</tt></li>
1496  <li><tt>__c11_atomic_fetch_or</tt></li>
1497  <li><tt>__c11_atomic_fetch_xor</tt></li>
1498</ul>
1499
1500
1501<!-- ======================================================================= -->
1502<h2 id="targetspecific">Target-Specific Extensions</h2>
1503<!-- ======================================================================= -->
1504
1505<p>Clang supports some language features conditionally on some targets.</p>
1506
1507<!-- ======================================================================= -->
1508<h3 id="x86-specific">X86/X86-64 Language Extensions</h3>
1509<!-- ======================================================================= -->
1510
1511<p>The X86 backend has these language extensions:</p>
1512
1513<!-- ======================================================================= -->
1514<h4 id="x86-gs-segment">Memory references off the GS segment</h4>
1515<!-- ======================================================================= -->
1516
1517<p>Annotating a pointer with address space #256 causes it to  be code generated
1518relative to the X86 GS segment register, and address space #257 causes it to be
1519relative to the X86 FS segment.  Note that this is a very very low-level
1520feature that should only be used if you know what you're doing (for example in
1521an OS kernel).</p>
1522
1523<p>Here is an example:</p>
1524
1525<pre>
1526#define GS_RELATIVE __attribute__((address_space(256)))
1527int foo(int GS_RELATIVE *P) {
1528  return *P;
1529}
1530</pre>
1531
1532<p>Which compiles to (on X86-32):</p>
1533
1534<pre>
1535_foo:
1536	movl	4(%esp), %eax
1537	movl	%gs:(%eax), %eax
1538	ret
1539</pre>
1540
1541<!-- ======================================================================= -->
1542<h2 id="analyzerspecific">Static Analysis-Specific Extensions</h2>
1543<!-- ======================================================================= -->
1544
1545<p>Clang supports additional attributes that are useful for documenting program
1546invariants and rules for static analysis tools. The extensions documented here
1547are used by the <a
1548href="http://clang.llvm.org/StaticAnalysis.html">path-sensitive static analyzer
1549engine</a> that is part of Clang's Analysis library.</p>
1550
1551<h3 id="attr_analyzer_noreturn">The <tt>analyzer_noreturn</tt> attribute</h3>
1552
1553<p>Clang's static analysis engine understands the standard <tt>noreturn</tt>
1554attribute. This attribute, which is typically affixed to a function prototype,
1555indicates that a call to a given function never returns. Function prototypes for
1556common functions like <tt>exit</tt> are typically annotated with this attribute,
1557as well as a variety of common assertion handlers. Users can educate the static
1558analyzer about their own custom assertion handles (thus cutting down on false
1559positives due to false paths) by marking their own &quot;panic&quot; functions
1560with this attribute.</p>
1561
1562<p>While useful, <tt>noreturn</tt> is not applicable in all cases. Sometimes
1563there are special functions that for all intents and purposes should be
1564considered panic functions (i.e., they are only called when an internal program
1565error occurs) but may actually return so that the program can fail gracefully.
1566The <tt>analyzer_noreturn</tt> attribute allows one to annotate such functions
1567as being interpreted as &quot;no return&quot; functions by the analyzer (thus
1568pruning bogus paths) but will not affect compilation (as in the case of
1569<tt>noreturn</tt>).</p>
1570
1571<p><b>Usage</b>: The <tt>analyzer_noreturn</tt> attribute can be placed in the
1572same places where the <tt>noreturn</tt> attribute can be placed. It is commonly
1573placed at the end of function prototypes:</p>
1574
1575<pre>
1576  void foo() <b>__attribute__((analyzer_noreturn))</b>;
1577</pre>
1578
1579<p>Query for this feature with
1580<tt>__has_attribute(analyzer_noreturn)</tt>.</p>
1581
1582<h3 id="attr_method_family">The <tt>objc_method_family</tt> attribute</h3>
1583
1584<p>Many methods in Objective-C have conventional meanings determined
1585by their selectors.  For the purposes of static analysis, it is
1586sometimes useful to be able to mark a method as having a particular
1587conventional meaning despite not having the right selector, or as not
1588having the conventional meaning that its selector would suggest.
1589For these use cases, we provide an attribute to specifically describe
1590the <q>method family</q> that a method belongs to.</p>
1591
1592<p><b>Usage</b>: <tt>__attribute__((objc_method_family(X)))</tt>,
1593where <tt>X</tt> is one of <tt>none</tt>, <tt>alloc</tt>, <tt>copy</tt>,
1594<tt>init</tt>, <tt>mutableCopy</tt>, or <tt>new</tt>.  This attribute
1595can only be placed at the end of a method declaration:</p>
1596
1597<pre>
1598  - (NSString*) initMyStringValue <b>__attribute__((objc_method_family(none)))</b>;
1599</pre>
1600
1601<p>Users who do not wish to change the conventional meaning of a
1602method, and who merely want to document its non-standard retain and
1603release semantics, should use the
1604<a href="#attr_retain_release">retaining behavior attributes</a>
1605described below.</p>
1606
1607<p>Query for this feature with
1608<tt>__has_attribute(objc_method_family)</tt>.</p>
1609
1610<h3 id="attr_retain_release">Objective-C retaining behavior attributes</h3>
1611
1612<p>In Objective-C, functions and methods are generally assumed to take
1613and return objects with +0 retain counts, with some exceptions for
1614special methods like <tt>+alloc</tt> and <tt>init</tt>.  However,
1615there are exceptions, and so Clang provides attributes to allow these
1616exceptions to be documented, which helps the analyzer find leaks (and
1617ignore non-leaks).  Some exceptions may be better described using
1618the <a href="#attr_method_family"><tt>objc_method_family</tt></a>
1619attribute instead.</p>
1620
1621<p><b>Usage</b>: The <tt>ns_returns_retained</tt>, <tt>ns_returns_not_retained</tt>,
1622<tt>ns_returns_autoreleased</tt>, <tt>cf_returns_retained</tt>,
1623and <tt>cf_returns_not_retained</tt> attributes can be placed on
1624methods and functions that return Objective-C or CoreFoundation
1625objects.  They are commonly placed at the end of a function prototype
1626or method declaration:</p>
1627
1628<pre>
1629  id foo() <b>__attribute__((ns_returns_retained))</b>;
1630
1631  - (NSString*) bar: (int) x <b>__attribute__((ns_returns_retained))</b>;
1632</pre>
1633
1634<p>The <tt>*_returns_retained</tt> attributes specify that the
1635returned object has a +1 retain count.
1636The <tt>*_returns_not_retained</tt> attributes specify that the return
1637object has a +0 retain count, even if the normal convention for its
1638selector would be +1.  <tt>ns_returns_autoreleased</tt> specifies that the
1639returned object is +0, but is guaranteed to live at least as long as the
1640next flush of an autorelease pool.</p>
1641
1642<p><b>Usage</b>: The <tt>ns_consumed</tt> and <tt>cf_consumed</tt>
1643attributes can be placed on an parameter declaration; they specify
1644that the argument is expected to have a +1 retain count, which will be
1645balanced in some way by the function or method.
1646The <tt>ns_consumes_self</tt> attribute can only be placed on an
1647Objective-C method; it specifies that the method expects
1648its <tt>self</tt> parameter to have a +1 retain count, which it will
1649balance in some way.</p>
1650
1651<pre>
1652  void <b>foo(__attribute__((ns_consumed))</b> NSString *string);
1653
1654  - (void) bar <b>__attribute__((ns_consumes_self))</b>;
1655  - (void) baz: (id) <b>__attribute__((ns_consumed))</b> x;
1656</pre>
1657
1658<p>Query for these features with <tt>__has_attribute(ns_consumed)</tt>,
1659<tt>__has_attribute(ns_returns_retained)</tt>, etc.</p>
1660
1661<!-- ======================================================================= -->
1662<h2 id="dynamicanalyzerspecific">Dynamic Analysis-Specific Extensions</h2>
1663<!-- ======================================================================= -->
1664<h3 id="address_sanitizer">AddressSanitizer</h3>
1665<p> Use <code>__has_feature(address_sanitizer)</code>
1666to check if the code is being built with <a
1667  href="AddressSanitizer.html">AddressSanitizer</a>.
1668</p>
1669<p>Use <tt>__attribute__((no_address_safety_analysis))</tt> on a function
1670declaration to specify that address safety instrumentation (e.g.
1671AddressSanitizer) should not be applied to that function.
1672</p>
1673
1674<!-- ======================================================================= -->
1675<h2 id="threadsafety">Thread-Safety Annotation Checking</h2>
1676<!-- ======================================================================= -->
1677
1678<p>Clang supports additional attributes for checking basic locking policies in
1679multithreaded programs.
1680Clang currently parses the following list of attributes, although
1681<b>the implementation for these annotations is currently in development.</b>
1682For more details, see the
1683<a href="http://gcc.gnu.org/wiki/ThreadSafetyAnnotation">GCC implementation</a>.
1684</p>
1685
1686<h4 id="ts_noanal">no_thread_safety_analysis</h4>
1687
1688<p>Use <tt>__attribute__((no_thread_safety_analysis))</tt> on a function
1689declaration to specify that the thread safety analysis should not be run on that
1690function. This attribute provides an escape hatch (e.g. for situations when it
1691is difficult to annotate the locking policy). </p>
1692
1693<h4 id="ts_lockable">lockable</h4>
1694
1695<p>Use <tt>__attribute__((lockable))</tt> on a class definition to specify
1696that it has a lockable type (e.g. a Mutex class). This annotation is primarily
1697used to check consistency.</p>
1698
1699<h4 id="ts_scopedlockable">scoped_lockable</h4>
1700
1701<p>Use <tt>__attribute__((scoped_lockable))</tt> on a class definition to
1702specify that it has a "scoped" lockable type. Objects of this type will acquire
1703the lock upon construction and release it upon going out of scope.
1704 This annotation is primarily used to check
1705consistency.</p>
1706
1707<h4 id="ts_guardedvar">guarded_var</h4>
1708
1709<p>Use <tt>__attribute__((guarded_var))</tt> on a variable declaration to
1710specify that the variable must be accessed while holding some lock.</p>
1711
1712<h4 id="ts_ptguardedvar">pt_guarded_var</h4>
1713
1714<p>Use <tt>__attribute__((pt_guarded_var))</tt> on a pointer declaration to
1715specify that the pointer must be dereferenced while holding some lock.</p>
1716
1717<h4 id="ts_guardedby">guarded_by(l)</h4>
1718
1719<p>Use <tt>__attribute__((guarded_by(l)))</tt> on a variable declaration to
1720specify that the variable must be accessed while holding lock <tt>l</tt>.</p>
1721
1722<h4 id="ts_ptguardedby">pt_guarded_by(l)</h4>
1723
1724<p>Use <tt>__attribute__((pt_guarded_by(l)))</tt> on a pointer declaration to
1725specify that the pointer must be dereferenced while holding lock <tt>l</tt>.</p>
1726
1727<h4 id="ts_acquiredbefore">acquired_before(...)</h4>
1728
1729<p>Use <tt>__attribute__((acquired_before(...)))</tt> on a declaration
1730of a lockable variable to specify that the lock must be acquired before all
1731attribute arguments. Arguments must be lockable type, and there must be at
1732least one argument.</p>
1733
1734<h4 id="ts_acquiredafter">acquired_after(...)</h4>
1735
1736<p>Use <tt>__attribute__((acquired_after(...)))</tt> on a declaration
1737of a lockable variable to specify that the lock must be acquired after all
1738attribute arguments. Arguments must be lockable type, and there must be at
1739least one argument.</p>
1740
1741<h4 id="ts_elf">exclusive_lock_function(...)</h4>
1742
1743<p>Use <tt>__attribute__((exclusive_lock_function(...)))</tt> on a function
1744declaration to specify that the function acquires all listed locks
1745exclusively. This attribute takes zero or more arguments: either of lockable
1746type or integers indexing into function parameters of lockable type. If no
1747arguments are given, the acquired lock is implicitly <tt>this</tt> of the
1748enclosing object.</p>
1749
1750<h4 id="ts_slf">shared_lock_function(...)</h4>
1751
1752<p>Use <tt>__attribute__((shared_lock_function(...)))</tt> on a function
1753declaration to specify that the function acquires all listed locks, although
1754 the locks may be shared (e.g. read locks). This attribute takes zero or more
1755arguments: either of lockable type or integers indexing into function
1756parameters of lockable type. If no arguments are given, the acquired lock is
1757implicitly <tt>this</tt> of the enclosing object.</p>
1758
1759<h4 id="ts_etf">exclusive_trylock_function(...)</h4>
1760
1761<p>Use <tt>__attribute__((exclusive_lock_function(...)))</tt> on a function
1762declaration to specify that the function will try (without blocking) to acquire
1763all listed locks exclusively. This attribute takes one or more arguments. The
1764first argument is an integer or boolean value specifying the return value of a
1765successful lock acquisition. The remaining arugments are either of lockable type
1766or integers indexing into function parameters of lockable type. If only one
1767argument is given, the acquired lock is implicitly <tt>this</tt> of the
1768enclosing object.</p>
1769
1770<h4 id="ts_stf">shared_trylock_function(...)</h4>
1771
1772<p>Use <tt>__attribute__((shared_lock_function(...)))</tt> on a function
1773declaration to specify that the function will try (without blocking) to acquire
1774all listed locks, although the locks may be shared (e.g. read locks). This
1775attribute takes one or more arguments. The first argument is an integer or
1776boolean value specifying the return value of a successful lock acquisition. The
1777remaining arugments are either of lockable type or integers indexing into
1778function parameters of lockable type. If only one argument is given, the
1779acquired lock is implicitly <tt>this</tt> of the enclosing object.</p>
1780
1781<h4 id="ts_uf">unlock_function(...)</h4>
1782
1783<p>Use <tt>__attribute__((unlock_function(...)))</tt> on a function
1784declaration to specify that the function release all listed locks. This
1785attribute takes zero or more arguments: either of lockable type or integers
1786indexing into function parameters of lockable type. If no arguments are given,
1787the acquired lock is implicitly <tt>this</tt> of the enclosing object.</p>
1788
1789<h4 id="ts_lr">lock_returned(l)</h4>
1790
1791<p>Use <tt>__attribute__((lock_returned(l)))</tt> on a function
1792declaration to specify that the function returns lock <tt>l</tt> (<tt>l</tt>
1793must be of lockable type). This annotation is used to aid in resolving lock
1794expressions.</p>
1795
1796<h4 id="ts_le">locks_excluded(...)</h4>
1797
1798<p>Use <tt>__attribute__((locks_excluded(...)))</tt> on a function declaration
1799to specify that the function must not be called with the listed locks. Arguments
1800must be lockable type, and there must be at least one argument.</p>
1801
1802<h4 id="ts_elr">exclusive_locks_required(...)</h4>
1803
1804<p>Use <tt>__attribute__((exclusive_locks_required(...)))</tt> on a function
1805declaration to specify that the function must be called while holding the listed
1806exclusive locks. Arguments must be lockable type, and there must be at
1807least one argument.</p>
1808
1809<h4 id="ts_slr">shared_locks_required(...)</h4>
1810
1811<p>Use <tt>__attribute__((shared_locks_required(...)))</tt> on a function
1812declaration to specify that the function must be called while holding the listed
1813shared locks. Arguments must be lockable type, and there must be at
1814least one argument.</p>
1815
1816</div>
1817</body>
1818</html>
1819