blob: b16c38c768240991fd941fb425459aa14a35039e [file] [log] [blame]
Chris Lattner5ce933f2009-02-09 08:46:11 +00001<html>
2<head>
3<title>Clang Language Extensions</title>
4<link type="text/css" rel="stylesheet" href="../menu.css" />
5<link type="text/css" rel="stylesheet" href="../content.css" />
6<style type="text/css">
7td {
8 vertical-align: top;
9}
10</style>
11</head>
12<body>
13
14<!--#include virtual="../menu.html.incl"-->
15
16<div id="content">
17
18<h1>Clang Language Extensions</h1>
19
20<ul>
21<li><a href="#intro">Introduction</a></li>
Chris Lattner148772a2009-06-13 07:13:28 +000022<li><a href="#feature_check">Feature Checking Macros</a></li>
John Thompson92bd8c72009-11-02 22:28:12 +000023<li><a href="#has_include">Include File Checking Macros</a></li>
Chris Lattner81edc9f2009-04-13 02:45:46 +000024<li><a href="#builtinmacros">Builtin Macros</a></li>
Chris Lattner5ce933f2009-02-09 08:46:11 +000025<li><a href="#vectors">Vectors and Extended Vectors</a></li>
John McCall48209082010-11-08 19:48:17 +000026<li><a href="#deprecated">Messages on <tt>deprecated</tt> and <tt>unavailable</tt> attributes</a></li>
27<li><a href="#attributes-on-enumerators">Attributes on enumerators</a></li>
Ted Kremenek87774fd2009-12-03 02:04:01 +000028<li><a href="#checking_language_features">Checks for Standard Language Features</a></li>
Ted Kremenek22c34102009-12-03 02:05:57 +000029 <ul>
30 <li><a href="#cxx_exceptions">C++ exceptions</a></li>
31 <li><a href="#cxx_rtti">C++ RTTI</a></li>
32 </ul>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000033<li><a href="#checking_upcoming_features">Checks for Upcoming Standard Language Features</a></li>
34 <ul>
35 <li><a href="#cxx_attributes">C++0x attributes</a></li>
36 <li><a href="#cxx_decltype">C++0x <tt>decltype()</tt></a></li>
Douglas Gregor07508002011-02-05 20:35:30 +000037 <li><a href="#cxx_default_function_template_args">C++0x default template arguments in function templates</a></li>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000038 <li><a href="#cxx_deleted_functions">C++0x deleted functions</a></li>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000039 <li><a href="#cxx_lambdas">C++0x lambdas</a></li>
40 <li><a href="#cxx_nullptr">C++0x nullptr</a></li>
Anders Carlssonc8b9f792011-03-25 15:04:23 +000041 <li><a href="#cxx_override_control">C++0x override control</a></li>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000042 <li><a href="#cxx_rvalue_references">C++0x rvalue references</a></li>
Douglas Gregor56209ff2011-01-26 21:25:54 +000043 <li><a href="#cxx_reference_qualified_functions">C++0x reference-qualified functions</a></li>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000044 <li><a href="#cxx_static_assert">C++0x <tt>static_assert()</tt></a></li>
45 <li><a href="#cxx_auto_type">C++0x type inference</a></li>
46 <li><a href="#cxx_variadic_templates">C++0x variadic templates</a></li>
Sebastian Redlf6c09772010-08-31 23:28:47 +000047 <li><a href="#cxx_inline_namespaces">C++0x inline namespaces</a></li>
Douglas Gregor1274ccd2010-10-08 23:50:27 +000048 <li><a href="#cxx_strong_enums">C++0x strongly-typed enumerations</a></li>
Douglas Gregordab60ad2010-10-01 18:44:50 +000049 <li><a href="#cxx_trailing_return">C++0x trailing return type</a></li>
Sebastian Redl4561ecd2011-03-15 21:17:12 +000050 <li><a href="#cxx_noexcept">C++0x noexcept specification</a></li>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +000051 </ul>
Douglas Gregorafdf1372011-02-03 21:57:35 +000052<li><a href="#checking_type_traits">Checks for Type Traits</a></li>
Chris Lattner5ce933f2009-02-09 08:46:11 +000053<li><a href="#blocks">Blocks</a></li>
Douglas Gregorcb54d432009-02-13 00:57:04 +000054<li><a href="#overloading-in-c">Function Overloading in C</a></li>
Chris Lattner5ce933f2009-02-09 08:46:11 +000055<li><a href="#builtins">Builtin Functions</a>
56 <ul>
Chris Lattner5ce933f2009-02-09 08:46:11 +000057 <li><a href="#__builtin_shufflevector">__builtin_shufflevector</a></li>
Chris Lattner21190d52009-09-21 03:09:59 +000058 <li><a href="#__builtin_unreachable">__builtin_unreachable</a></li>
Chris Lattner23aa9c82011-04-09 03:57:26 +000059 <li><a href="#__sync_swap">__sync_swap</a></li>
Douglas Gregorafdf1372011-02-03 21:57:35 +000060 </ul>
Chris Lattner5ce933f2009-02-09 08:46:11 +000061</li>
Chris Lattner1177f912009-04-09 19:58:15 +000062<li><a href="#targetspecific">Target-Specific Extensions</a>
63 <ul>
64 <li><a href="#x86-specific">X86/X86-64 Language Extensions</a></li>
65 </ul>
66</li>
John McCall87494012011-03-18 03:51:49 +000067<li><a href="#analyzerspecific">Static Analysis-Specific Extensions</a></li>
Chris Lattner5ce933f2009-02-09 08:46:11 +000068</ul>
69
Chris Lattner5ce933f2009-02-09 08:46:11 +000070<!-- ======================================================================= -->
71<h2 id="intro">Introduction</h2>
72<!-- ======================================================================= -->
73
74<p>This document describes the language extensions provided by Clang. In
Chris Lattner148772a2009-06-13 07:13:28 +000075addition to the language extensions listed here, Clang aims to support a broad
Chris Lattner5ce933f2009-02-09 08:46:11 +000076range of GCC extensions. Please see the <a
77href="http://gcc.gnu.org/onlinedocs/gcc/C-Extensions.html">GCC manual</a> for
78more information on these extensions.</p>
79
80<!-- ======================================================================= -->
Chris Lattner148772a2009-06-13 07:13:28 +000081<h2 id="feature_check">Feature Checking Macros</h2>
82<!-- ======================================================================= -->
83
84<p>Language extensions can be very useful, but only if you know you can depend
Chris Lattnerc70e1932011-03-21 16:25:11 +000085on them. In order to allow fine-grain features checks, we support three builtin
Chris Lattner148772a2009-06-13 07:13:28 +000086function-like macros. This allows you to directly test for a feature in your
87code without having to resort to something like autoconf or fragile "compiler
88version checks".</p>
89
90<!-- ======================================================================= -->
91<h3 id="__has_builtin">__has_builtin</h3>
92<!-- ======================================================================= -->
93
94<p>This function-like macro takes a single identifier argument that is the name
95of a builtin function. It evaluates to 1 if the builtin is supported or 0 if
96not. It can be used like this:</p>
97
98<blockquote>
99<pre>
100#ifndef __has_builtin // Optional of course.
101 #define __has_builtin(x) 0 // Compatibility with non-clang compilers.
102#endif
103
104...
105#if __has_builtin(__builtin_trap)
106 __builtin_trap();
107#else
108 abort();
109#endif
110...
111</pre>
112</blockquote>
113
114
115<!-- ======================================================================= -->
116<h3 id="__has_feature">__has_feature</h3>
117<!-- ======================================================================= -->
118
119<p>This function-like macro takes a single identifier argument that is the name
120of a feature. It evaluates to 1 if the feature is supported or 0 if not. It
121can be used like this:</p>
122
123<blockquote>
124<pre>
125#ifndef __has_feature // Optional of course.
126 #define __has_feature(x) 0 // Compatibility with non-clang compilers.
127#endif
128
129...
130#if __has_feature(attribute_overloadable) || \
131 __has_feature(blocks)
132...
133#endif
134...
135</pre>
136</blockquote>
137
138<p>The feature tag is described along with the language feature below.</p>
139
John Thompson92bd8c72009-11-02 22:28:12 +0000140<!-- ======================================================================= -->
Anders Carlssoncae50952010-10-20 02:31:43 +0000141<h3 id="__has_attribute">__has_attribute</h3>
142<!-- ======================================================================= -->
143
144<p>This function-like macro takes a single identifier argument that is the name
145of an attribute. It evaluates to 1 if the attribute is supported or 0 if not. It
146can be used like this:</p>
147
148<blockquote>
149<pre>
150#ifndef __has_attribute // Optional of course.
151 #define __has_attribute(x) 0 // Compatibility with non-clang compilers.
152#endif
153
154...
Anders Carlsson961003d2011-01-24 03:54:51 +0000155#if __has_attribute(always_inline)
156#define ALWAYS_INLINE __attribute__((always_inline))
Anders Carlssoncae50952010-10-20 02:31:43 +0000157#else
Anders Carlsson961003d2011-01-24 03:54:51 +0000158#define ALWAYS_INLINE
Anders Carlssoncae50952010-10-20 02:31:43 +0000159#endif
160...
161</pre>
162</blockquote>
163
164<!-- ======================================================================= -->
John Thompson92bd8c72009-11-02 22:28:12 +0000165<h2 id="has_include">Include File Checking Macros</h2>
166<!-- ======================================================================= -->
167
168<p>Not all developments systems have the same include files.
169The <a href="#__has_include">__has_include</a> and
170<a href="#__has_include_next">__has_include_next</a> macros allow you to
171check for the existence of an include file before doing
172a possibly failing #include directive.</p>
173
174<!-- ======================================================================= -->
175<h3 id="__has_include">__has_include</h3>
176<!-- ======================================================================= -->
177
178<p>This function-like macro takes a single file name string argument that
179is the name of an include file. It evaluates to 1 if the file can
180be found using the include paths, or 0 otherwise:</p>
181
182<blockquote>
183<pre>
184// Note the two possible file name string formats.
185#if __has_include("myinclude.h") && __has_include(&lt;stdint.h&gt;)
186# include "myinclude.h"
187#endif
188
189// To avoid problem with non-clang compilers not having this macro.
190#if defined(__has_include) && __has_include("myinclude.h")
191# include "myinclude.h"
192#endif
193</pre>
194</blockquote>
195
196<p>To test for this feature, use #if defined(__has_include).</p>
197
198<!-- ======================================================================= -->
199<h3 id="__has_include_next">__has_include_next</h3>
200<!-- ======================================================================= -->
201
202<p>This function-like macro takes a single file name string argument that
203is the name of an include file. It is like __has_include except that it
204looks for the second instance of the given file found in the include
205paths. It evaluates to 1 if the second instance of the file can
206be found using the include paths, or 0 otherwise:</p>
207
208<blockquote>
209<pre>
210// Note the two possible file name string formats.
211#if __has_include_next("myinclude.h") && __has_include_next(&lt;stdint.h&gt;)
212# include_next "myinclude.h"
213#endif
214
215// To avoid problem with non-clang compilers not having this macro.
216#if defined(__has_include_next) && __has_include_next("myinclude.h")
217# include_next "myinclude.h"
218#endif
219</pre>
220</blockquote>
221
222<p>Note that __has_include_next, like the GNU extension
223#include_next directive, is intended for use in headers only,
224and will issue a warning if used in the top-level compilation
225file. A warning will also be issued if an absolute path
226is used in the file argument.</p>
Chris Lattner148772a2009-06-13 07:13:28 +0000227
228<!-- ======================================================================= -->
Chris Lattner81edc9f2009-04-13 02:45:46 +0000229<h2 id="builtinmacros">Builtin Macros</h2>
230<!-- ======================================================================= -->
231
Douglas Gregor4290fbd2010-04-30 02:51:06 +0000232<dl>
233 <dt><code>__BASE_FILE__</code></dt>
234 <dd>Defined to a string that contains the name of the main input
235 file passed to Clang.</dd>
236
237 <dt><code>__COUNTER__</code></dt>
238 <dd>Defined to an integer value that starts at zero and is
239 incremented each time the <code>__COUNTER__</code> macro is
240 expanded.</dd>
241
242 <dt><code>__INCLUDE_LEVEL__</code></dt>
243 <dd>Defined to an integral value that is the include depth of the
244 file currently being translated. For the main file, this value is
245 zero.</dd>
246
247 <dt><code>__TIMESTAMP__</code></dt>
248 <dd>Defined to the date and time of the last modification of the
249 current source file.</dd>
250
251 <dt><code>__clang__</code></dt>
252 <dd>Defined when compiling with Clang</dd>
253
254 <dt><code>__clang_major__</code></dt>
255 <dd>Defined to the major version number of Clang (e.g., the 2 in
256 2.0.1).</dd>
257
258 <dt><code>__clang_minor__</code></dt>
259 <dd>Defined to the minor version number of Clang (e.g., the 0 in
260 2.0.1).</dd>
261
262 <dt><code>__clang_patchlevel__</code></dt>
263 <dd>Defined to the patch level of Clang (e.g., the 1 in 2.0.1).</dd>
264
265 <dt><code>__clang_version__</code></dt>
266 <dd>Defined to a string that captures the Clang version, including
267 the Subversion tag or revision number, e.g., "1.5 (trunk
268 102332)".</dd>
269</dl>
Chris Lattner81edc9f2009-04-13 02:45:46 +0000270
271<!-- ======================================================================= -->
Chris Lattner5ce933f2009-02-09 08:46:11 +0000272<h2 id="vectors">Vectors and Extended Vectors</h2>
273<!-- ======================================================================= -->
274
Owen Andersond2bf0cd2010-01-27 01:22:36 +0000275<p>Supports the GCC vector extensions, plus some stuff like V[1].</p>
276
277<p>Also supports <tt>ext_vector</tt>, which additionally support for V.xyzw
278syntax and other tidbits as seen in OpenCL. An example is:</p>
279
280<blockquote>
281<pre>
282typedef float float4 <b>__attribute__((ext_vector_type(4)))</b>;
283typedef float float2 <b>__attribute__((ext_vector_type(2)))</b>;
284
285float4 foo(float2 a, float2 b) {
286 float4 c;
287 c.xz = a;
288 c.yw = b;
289 return c;
290}
John McCall48209082010-11-08 19:48:17 +0000291</pre>
Owen Andersond2bf0cd2010-01-27 01:22:36 +0000292</blockquote>
Chris Lattner5ce933f2009-02-09 08:46:11 +0000293
Chris Lattner148772a2009-06-13 07:13:28 +0000294<p>Query for this feature with __has_feature(attribute_ext_vector_type).</p>
295
Owen Andersond2bf0cd2010-01-27 01:22:36 +0000296<p>See also <a href="#__builtin_shufflevector">__builtin_shufflevector</a>.</p>
297
Chris Lattner5ce933f2009-02-09 08:46:11 +0000298<!-- ======================================================================= -->
John McCall48209082010-11-08 19:48:17 +0000299<h2 id="deprecated">Messages on <tt>deprecated</tt> and <tt>unavailable</tt> Attributes</h2>
Fariborz Jahanianc784dc12010-10-06 23:12:32 +0000300<!-- ======================================================================= -->
301
John McCall48209082010-11-08 19:48:17 +0000302<p>An optional string message can be added to the <tt>deprecated</tt>
303and <tt>unavailable</tt> attributes. For example:</p>
Fariborz Jahanianc784dc12010-10-06 23:12:32 +0000304
John McCall48209082010-11-08 19:48:17 +0000305<blockquote>
Chris Lattner4836d6a2010-11-09 19:43:35 +0000306<pre>void explode(void) __attribute__((deprecated("extremely unsafe, use 'combust' instead!!!")));</pre>
John McCall48209082010-11-08 19:48:17 +0000307</blockquote>
308
309<p>If the deprecated or unavailable declaration is used, the message
310will be incorporated into the appropriate diagnostic:</p>
311
312<blockquote>
Chris Lattner4836d6a2010-11-09 19:43:35 +0000313<pre>harmless.c:4:3: warning: 'explode' is deprecated: extremely unsafe, use 'combust' instead!!! [-Wdeprecated-declarations]
John McCall48209082010-11-08 19:48:17 +0000314 explode();
315 ^</pre>
316</blockquote>
317
318<p>Query for this feature
319with <tt>__has_feature(attribute_deprecated_with_message)</tt>
320and <tt>__has_feature(attribute_unavailable_with_message)</tt>.</p>
321
322<!-- ======================================================================= -->
323<h2 id="attributes-on-enumerators">Attributes on Enumerators</h2>
324<!-- ======================================================================= -->
325
326<p>Clang allows attributes to be written on individual enumerators.
327This allows enumerators to be deprecated, made unavailable, etc. The
328attribute must appear after the enumerator name and before any
329initializer, like so:</p>
330
331<blockquote>
332<pre>enum OperationMode {
333 OM_Invalid,
334 OM_Normal,
335 OM_Terrified __attribute__((deprecated)),
336 OM_AbortOnError __attribute__((deprecated)) = 4
337};</pre>
338</blockquote>
339
340<p>Attributes on the <tt>enum</tt> declaration do not apply to
341individual enumerators.</p>
342
343<p>Query for this feature with <tt>__has_feature(enumerator_attributes)</tt>.</p>
Fariborz Jahanianc784dc12010-10-06 23:12:32 +0000344
345<!-- ======================================================================= -->
Ted Kremenek87774fd2009-12-03 02:04:01 +0000346<h2 id="checking_language_features">Checks for Standard Language Features</h2>
347<!-- ======================================================================= -->
348
349<p>The <tt>__has_feature</tt> macro can be used to query if certain standard language features are
350enabled. Those features are listed here.</p>
351
Ted Kremenek22c34102009-12-03 02:05:57 +0000352<h3 id="cxx_exceptions">C++ exceptions</h3>
Ted Kremenek87774fd2009-12-03 02:04:01 +0000353
Ted Kremenek22c34102009-12-03 02:05:57 +0000354<p>Use <tt>__has_feature(cxx_exceptions)</tt> to determine if C++ exceptions have been enabled. For
355example, compiling code with <tt>-fexceptions</tt> enables C++ exceptions.</p>
Ted Kremenek87774fd2009-12-03 02:04:01 +0000356
Ted Kremenek22c34102009-12-03 02:05:57 +0000357<h3 id="cxx_rtti">C++ RTTI</h3>
Ted Kremenek87774fd2009-12-03 02:04:01 +0000358
Ted Kremenek0eb95602009-12-03 02:06:43 +0000359<p>Use <tt>__has_feature(cxx_rtti)</tt> to determine if C++ RTTI has been enabled. For example,
Ted Kremenek22c34102009-12-03 02:05:57 +0000360compiling code with <tt>-fno-rtti</tt> disables the use of RTTI.</p>
Ted Kremenek87774fd2009-12-03 02:04:01 +0000361
362<!-- ======================================================================= -->
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000363<h2 id="checking_upcoming_features">Checks for Upcoming Standard Language Features</h2>
364<!-- ======================================================================= -->
365
366<p>The <tt>__has_feature</tt> macro can be used to query if certain upcoming
367standard language features are enabled. Those features are listed here.</p>
368
369<p>Currently, all features listed here are slated for inclusion in the upcoming
370C++0x standard. As a result, all the features that clang supports are enabled
371with the <tt>-std=c++0x</tt> option when compiling C++ code. Features that are
372not yet implemented will be noted.</p>
373
374<h3 id="cxx_decltype">C++0x <tt>decltype()</tt></h3>
375
376<p>Use <tt>__has_feature(cxx_decltype)</tt> to determine if support for the
377<tt>decltype()</tt> specifier is enabled.</p>
378
379<h3 id="cxx_attributes">C++0x attributes</h3>
380
381<p>Use <tt>__has_feature(cxx_attributes)</tt> to determine if support for
Sebastian Redlf6c09772010-08-31 23:28:47 +0000382attribute parsing with C++0x's square bracket notation is enabled.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000383
Douglas Gregor07508002011-02-05 20:35:30 +0000384<h3 id="cxx_default_function_template_args">C++0x default template arguments in function templates</h3>
385
386<p>Use <tt>__has_feature(cxx_default_function_template_args)</tt> to determine if support for default template arguments in function templates is enabled.</p>
387
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000388<h3 id="cxx_deleted_functions">C++0x deleted functions</tt></h3>
389
390<p>Use <tt>__has_feature(cxx_deleted_functions)</tt> to determine if support for
Sebastian Redlf6c09772010-08-31 23:28:47 +0000391deleted function definitions (with <tt>= delete</tt>) is enabled.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000392
Douglas Gregor9cc90a32010-01-13 16:27:49 +0000393<h3 id="cxx_lambdas">C++0x lambdas</h3>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000394
395<p>Use <tt>__has_feature(cxx_lambdas)</tt> to determine if support for
Sebastian Redlf6c09772010-08-31 23:28:47 +0000396lambdas is enabled. clang does not currently implement this feature.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000397
398<h3 id="cxx_nullptr">C++0x <tt>nullptr</tt></h3>
399
400<p>Use <tt>__has_feature(cxx_nullptr)</tt> to determine if support for
Sebastian Redlf6c09772010-08-31 23:28:47 +0000401<tt>nullptr</tt> is enabled. clang does not yet fully implement this
402feature.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000403
Anders Carlssonc8b9f792011-03-25 15:04:23 +0000404<h3 id="cxx_override_control">C++0x <tt>override control</tt></h3>
405
406<p>Use <tt>__has_feature(cxx_override_control)</tt> to determine if support for
407the override control keywords is enabled.</p>
408
Douglas Gregor56209ff2011-01-26 21:25:54 +0000409<h3 id="cxx_reference_qualified_functions">C++0x reference-qualified functions</h3>
410<p>Use <tt>__has_feature(cxx_reference_qualified_functions)</tt> to determine if support for reference-qualified functions (e.g., member functions with <code>&amp;</code> or <code>&amp;&amp;</code> applied to <code>*this</code>) is enabled.</p>
411
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000412<h3 id="cxx_rvalue_references">C++0x rvalue references</tt></h3>
413
414<p>Use <tt>__has_feature(cxx_rvalue_references)</tt> to determine if support for
Douglas Gregor56209ff2011-01-26 21:25:54 +0000415rvalue references is enabled. </p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000416
417<h3 id="cxx_static_assert">C++0x <tt>static_assert()</tt></h3>
418
419<p>Use <tt>__has_feature(cxx_static_assert)</tt> to determine if support for
420compile-time assertions using <tt>static_assert</tt> is enabled.</p>
421
422<h3 id="cxx_auto_type">C++0x type inference</h3>
423
424<p>Use <tt>__has_feature(cxx_auto_type)</tt> to determine C++0x type inference
425is supported using the <tt>auto</tt> specifier. If this is disabled,
Richard Smithfd405ef2011-02-23 00:41:16 +0000426<tt>auto</tt> will instead be a storage class specifier, as in C or C++98.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000427
Sebastian Redlf6c09772010-08-31 23:28:47 +0000428<h3 id="cxx_variadic_templates">C++0x variadic templates</h3>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000429
430<p>Use <tt>__has_feature(cxx_variadic_templates)</tt> to determine if support
Douglas Gregor83d77812011-01-19 23:15:20 +0000431for variadic templates is enabled.</p>
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000432
Sebastian Redlf6c09772010-08-31 23:28:47 +0000433<h3 id="cxx_inline_namespaces">C++0x inline namespaces</h3>
434
435<p>Use <tt>__has_feature(cxx_inline_namespaces)</tt> to determine if support for
436inline namespaces is enabled.</p>
437
Douglas Gregordab60ad2010-10-01 18:44:50 +0000438<h3 id="cxx_trailing_return">C++0x trailing return type</h3>
439
440<p>Use <tt>__has_feature(cxx_trailing_return)</tt> to determine if support for
441the alternate function declaration syntax with trailing return type is enabled.</p>
442
Sebastian Redl4561ecd2011-03-15 21:17:12 +0000443<h3 id="cxx_noexcept">C++0x noexcept</h3>
444
445<p>Use <tt>__has_feature(cxx_noexcept)</tt> to determine if support for
446noexcept exception specifications is enabled.</p>
447
Douglas Gregor1274ccd2010-10-08 23:50:27 +0000448<h3 id="cxx_strong_enums">C++0x strongly typed enumerations</h3>
449
450<p>Use <tt>__has_feature(cxx_strong_enums)</tt> to determine if support for
451strongly typed, scoped enumerations is enabled.</p>
452
Sean Hunt4ef4c6b2010-01-13 08:31:49 +0000453<!-- ======================================================================= -->
Douglas Gregorafdf1372011-02-03 21:57:35 +0000454<h2 id="checking_type_traits">Checks for Type Traits</h2>
455<!-- ======================================================================= -->
456
457<p>Clang supports the <a hef="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_feature(X)</code> indicates the presence of the type trait. For example:
458<blockquote>
459<pre>
460#if __has_feature(is_convertible_to)
461template&lt;typename From, typename To&gt;
462struct is_convertible_to {
463 static const bool value = __is_convertible_to(From, To);
464};
465#else
466// Emulate type trait
467#endif
468</pre>
469</blockquote>
470
471<p>The following type traits are supported by Clang:</p>
472<ul>
473 <li><code>__has_nothrow_assign</code> (GNU, Microsoft)</li>
474 <li><code>__has_nothrow_copy</code> (GNU, Microsoft)</li>
475 <li><code>__has_nothrow_constructor</code> (GNU, Microsoft)</li>
476 <li><code>__has_trivial_assign</code> (GNU, Microsoft)</li>
477 <li><code>__has_trivial_copy</code> (GNU, Microsoft)</li>
478 <li><code>__has_trivial_constructor</code> (GNU, Microsoft)</li>
479 <li><code>__has_trivial_destructor</code> (GNU, Microsoft)</li>
480 <li><code>__has_virtual_destructor</code> (GNU, Microsoft)</li>
481 <li><code>__is_abstract</code> (GNU, Microsoft)</li>
482 <li><code>__is_base_of</code> (GNU, Microsoft)</li>
483 <li><code>__is_class</code> (GNU, Microsoft)</li>
484 <li><code>__is_convertible_to</code> (Microsoft)</li>
485 <li><code>__is_empty</code> (GNU, Microsoft)</li>
486 <li><code>__is_enum</code> (GNU, Microsoft)</li>
487 <li><code>__is_pod</code> (GNU, Microsoft)</li>
488 <li><code>__is_polymorphic</code> (GNU, Microsoft)</li>
489 <li><code>__is_union</code> (GNU, Microsoft)</li>
490 <li><code>__is_literal(type)</code>: Determines whether the given type is a literal type</li>
491</ul>
492
493<!-- ======================================================================= -->
Chris Lattner5ce933f2009-02-09 08:46:11 +0000494<h2 id="blocks">Blocks</h2>
495<!-- ======================================================================= -->
496
Chris Lattnera7dbdf52009-03-09 07:03:22 +0000497<p>The syntax and high level language feature description is in <a
498href="BlockLanguageSpec.txt">BlockLanguageSpec.txt</a>. Implementation and ABI
499details for the clang implementation are in <a
Chris Lattner5d7650b2010-03-16 21:43:03 +0000500href="Block-ABI-Apple.txt">Block-ABI-Apple.txt</a>.</p>
Chris Lattner5ce933f2009-02-09 08:46:11 +0000501
Chris Lattner148772a2009-06-13 07:13:28 +0000502
503<p>Query for this feature with __has_feature(blocks).</p>
504
Chris Lattner5ce933f2009-02-09 08:46:11 +0000505<!-- ======================================================================= -->
Douglas Gregorcb54d432009-02-13 00:57:04 +0000506<h2 id="overloading-in-c">Function Overloading in C</h2>
507<!-- ======================================================================= -->
508
Chris Lattnerf161d412009-02-13 21:51:45 +0000509<p>Clang provides support for C++ function overloading in C. Function
510overloading in C is introduced using the <tt>overloadable</tt> attribute. For
511example, one might provide several overloaded versions of a <tt>tgsin</tt>
512function that invokes the appropriate standard function computing the sine of a
513value with <tt>float</tt>, <tt>double</tt>, or <tt>long double</tt>
514precision:</p>
Douglas Gregorcb54d432009-02-13 00:57:04 +0000515
516<blockquote>
517<pre>
518#include &lt;math.h&gt;
519float <b>__attribute__((overloadable))</b> tgsin(float x) { return sinf(x); }
520double <b>__attribute__((overloadable))</b> tgsin(double x) { return sin(x); }
521long double <b>__attribute__((overloadable))</b> tgsin(long double x) { return sinl(x); }
522</pre>
523</blockquote>
524
525<p>Given these declarations, one can call <tt>tgsin</tt> with a
526<tt>float</tt> value to receive a <tt>float</tt> result, with a
527<tt>double</tt> to receive a <tt>double</tt> result, etc. Function
528overloading in C follows the rules of C++ function overloading to pick
529the best overload given the call arguments, with a few C-specific
530semantics:</p>
531<ul>
532 <li>Conversion from <tt>float</tt> or <tt>double</tt> to <tt>long
533 double</tt> is ranked as a floating-point promotion (per C99) rather
534 than as a floating-point conversion (as in C++).</li>
535
536 <li>A conversion from a pointer of type <tt>T*</tt> to a pointer of type
537 <tt>U*</tt> is considered a pointer conversion (with conversion
538 rank) if <tt>T</tt> and <tt>U</tt> are compatible types.</li>
539
540 <li>A conversion from type <tt>T</tt> to a value of type <tt>U</tt>
541 is permitted if <tt>T</tt> and <tt>U</tt> are compatible types. This
542 conversion is given "conversion" rank.</li>
543</ul>
544
545<p>The declaration of <tt>overloadable</tt> functions is restricted to
546function declarations and definitions. Most importantly, if any
547function with a given name is given the <tt>overloadable</tt>
548attribute, then all function declarations and definitions with that
549name (and in that scope) must have the <tt>overloadable</tt>
Chris Lattnerf161d412009-02-13 21:51:45 +0000550attribute. This rule even applies to redeclarations of functions whose original
551declaration had the <tt>overloadable</tt> attribute, e.g.,</p>
Douglas Gregorcb54d432009-02-13 00:57:04 +0000552
553<blockquote>
554<pre>
555int f(int) __attribute__((overloadable));
556float f(float); <i>// error: declaration of "f" must have the "overloadable" attribute</i>
557
558int g(int) __attribute__((overloadable));
559int g(int) { } <i>// error: redeclaration of "g" must also have the "overloadable" attribute</i>
560</pre>
561</blockquote>
562
Douglas Gregor965acbb2009-02-18 07:07:28 +0000563<p>Functions marked <tt>overloadable</tt> must have
564prototypes. Therefore, the following code is ill-formed:</p>
565
566<blockquote>
567<pre>
568int h() __attribute__((overloadable)); <i>// error: h does not have a prototype</i>
569</pre>
570</blockquote>
571
572<p>However, <tt>overloadable</tt> functions are allowed to use a
573ellipsis 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>
574
575<blockquote>
576<pre>
Chris Lattner02246802009-02-18 22:27:46 +0000577void honeypot(...) __attribute__((overloadable, unavailable)); <i>// calling me is an error</i>
Douglas Gregor965acbb2009-02-18 07:07:28 +0000578</pre>
579</blockquote>
580
Douglas Gregorcb54d432009-02-13 00:57:04 +0000581<p>Functions declared with the <tt>overloadable</tt> attribute have
582their names mangled according to the same rules as C++ function
583names. For example, the three <tt>tgsin</tt> functions in our
584motivating example get the mangled names <tt>_Z5tgsinf</tt>,
Chris Lattner71b48d62010-11-28 18:19:13 +0000585<tt>_Z5tgsind</tt>, and <tt>_Z5tgsine</tt>, respectively. There are two
Douglas Gregorcb54d432009-02-13 00:57:04 +0000586caveats to this use of name mangling:</p>
587
588<ul>
589
590 <li>Future versions of Clang may change the name mangling of
591 functions overloaded in C, so you should not depend on an specific
592 mangling. To be completely safe, we strongly urge the use of
593 <tt>static inline</tt> with <tt>overloadable</tt> functions.</li>
594
595 <li>The <tt>overloadable</tt> attribute has almost no meaning when
596 used in C++, because names will already be mangled and functions are
597 already overloadable. However, when an <tt>overloadable</tt>
598 function occurs within an <tt>extern "C"</tt> linkage specification,
599 it's name <i>will</i> be mangled in the same way as it would in
600 C.</li>
601</ul>
602
Chris Lattner148772a2009-06-13 07:13:28 +0000603<p>Query for this feature with __has_feature(attribute_overloadable).</p>
604
605
Douglas Gregorcb54d432009-02-13 00:57:04 +0000606<!-- ======================================================================= -->
Chris Lattner5ce933f2009-02-09 08:46:11 +0000607<h2 id="builtins">Builtin Functions</h2>
608<!-- ======================================================================= -->
609
610<p>Clang supports a number of builtin library functions with the same syntax as
611GCC, including things like <tt>__builtin_nan</tt>,
612<tt>__builtin_constant_p</tt>, <tt>__builtin_choose_expr</tt>,
613<tt>__builtin_types_compatible_p</tt>, <tt>__sync_fetch_and_add</tt>, etc. In
614addition to the GCC builtins, Clang supports a number of builtins that GCC does
615not, which are listed here.</p>
616
617<p>Please note that Clang does not and will not support all of the GCC builtins
618for vector operations. Instead of using builtins, you should use the functions
619defined in target-specific header files like <tt>&lt;xmmintrin.h&gt;</tt>, which
620define portable wrappers for these. Many of the Clang versions of these
621functions are implemented directly in terms of <a href="#vectors">extended
622vector support</a> instead of builtins, in order to reduce the number of
623builtins that we need to implement.</p>
624
Chris Lattner5ce933f2009-02-09 08:46:11 +0000625<!-- ======================================================================= -->
Chris Lattner6f72da52009-02-13 20:00:20 +0000626<h3 id="__builtin_shufflevector">__builtin_shufflevector</h3>
Chris Lattner5ce933f2009-02-09 08:46:11 +0000627<!-- ======================================================================= -->
628
Chris Lattneraad826b2009-09-16 18:56:12 +0000629<p><tt>__builtin_shufflevector</tt> is used to express generic vector
Chris Lattner6f72da52009-02-13 20:00:20 +0000630permutation/shuffle/swizzle operations. This builtin is also very important for
631the implementation of various target-specific header files like
632<tt>&lt;xmmintrin.h&gt;</tt>.
Chris Lattner5ce933f2009-02-09 08:46:11 +0000633</p>
634
635<p><b>Syntax:</b></p>
636
637<pre>
Chris Lattner6f72da52009-02-13 20:00:20 +0000638__builtin_shufflevector(vec1, vec2, index1, index2, ...)
Chris Lattner5ce933f2009-02-09 08:46:11 +0000639</pre>
640
641<p><b>Examples:</b></p>
642
643<pre>
Chris Lattner6f72da52009-02-13 20:00:20 +0000644 // Identity operation - return 4-element vector V1.
645 __builtin_shufflevector(V1, V1, 0, 1, 2, 3)
646
647 // "Splat" element 0 of V1 into a 4-element result.
648 __builtin_shufflevector(V1, V1, 0, 0, 0, 0)
649
650 // Reverse 4-element vector V1.
651 __builtin_shufflevector(V1, V1, 3, 2, 1, 0)
652
653 // Concatenate every other element of 4-element vectors V1 and V2.
654 __builtin_shufflevector(V1, V2, 0, 2, 4, 6)
655
656 // Concatenate every other element of 8-element vectors V1 and V2.
657 __builtin_shufflevector(V1, V2, 0, 2, 4, 6, 8, 10, 12, 14)
Chris Lattner5ce933f2009-02-09 08:46:11 +0000658</pre>
659
660<p><b>Description:</b></p>
661
Chris Lattner6f72da52009-02-13 20:00:20 +0000662<p>The first two arguments to __builtin_shufflevector are vectors that have the
663same element type. The remaining arguments are a list of integers that specify
664the elements indices of the first two vectors that should be extracted and
665returned in a new vector. These element indices are numbered sequentially
666starting with the first vector, continuing into the second vector. Thus, if
667vec1 is a 4-element vector, index 5 would refer to the second element of vec2.
Chris Lattner5ce933f2009-02-09 08:46:11 +0000668</p>
669
Chris Lattner6f72da52009-02-13 20:00:20 +0000670<p>The result of __builtin_shufflevector is a vector
671with the same element type as vec1/vec2 but that has an element count equal to
672the number of indices specified.
673</p>
Chris Lattner5ce933f2009-02-09 08:46:11 +0000674
Chris Lattner21190d52009-09-21 03:09:59 +0000675<p>Query for this feature with __has_builtin(__builtin_shufflevector).</p>
676
677<!-- ======================================================================= -->
678<h3 id="__builtin_unreachable">__builtin_unreachable</h3>
679<!-- ======================================================================= -->
680
681<p><tt>__builtin_unreachable</tt> is used to indicate that a specific point in
682the program cannot be reached, even if the compiler might otherwise think it
683can. This is useful to improve optimization and eliminates certain warnings.
684For example, without the <tt>__builtin_unreachable</tt> in the example below,
685the compiler assumes that the inline asm can fall through and prints a "function
686declared 'noreturn' should not return" warning.
687</p>
688
689<p><b>Syntax:</b></p>
690
691<pre>
692__builtin_unreachable()
693</pre>
694
695<p><b>Example of Use:</b></p>
696
697<pre>
698void myabort(void) __attribute__((noreturn));
699void myabort(void) {
700 asm("int3");
701 __builtin_unreachable();
702}
703</pre>
704
705<p><b>Description:</b></p>
706
707<p>The __builtin_unreachable() builtin has completely undefined behavior. Since
708it has undefined behavior, it is a statement that it is never reached and the
709optimizer can take advantage of this to produce better code. This builtin takes
710no arguments and produces a void result.
711</p>
712
713<p>Query for this feature with __has_builtin(__builtin_unreachable).</p>
714
Chris Lattner23aa9c82011-04-09 03:57:26 +0000715<!-- ======================================================================= -->
716<h3 id="__sync_swap">__sync_swap</h3>
717<!-- ======================================================================= -->
718
719<p><tt>__sync_swap</tt> is used to atomically swap integers or pointers in
720memory.
721</p>
722
723<p><b>Syntax:</b></p>
724
725<pre>
726<i>type</i> __sync_swap(<i>type</i> *ptr, <i>type</i> value, ...)
727</pre>
728
729<p><b>Example of Use:</b></p>
730
731<pre>
732int old_value = __sync_swap(&value, new_value);
733</pre>
734
735<p><b>Description:</b></p>
736
737<p>The __sync_swap() builtin extends the existing __sync_*() family of atomic
738intrinsics to allow code to atomically swap the current value with the new
739value. More importantly, it helps developers write more efficient and correct
740code by avoiding expensive loops around __sync_bool_compare_and_swap() or
741relying on the platform specific implementation details of
742__sync_lock_test_and_set(). The __sync_swap() builtin is a full barrier.
743</p>
744
Chris Lattner21190d52009-09-21 03:09:59 +0000745
Chris Lattner1177f912009-04-09 19:58:15 +0000746<!-- ======================================================================= -->
747<h2 id="targetspecific">Target-Specific Extensions</h2>
748<!-- ======================================================================= -->
749
750<p>Clang supports some language features conditionally on some targets.</p>
751
752<!-- ======================================================================= -->
753<h3 id="x86-specific">X86/X86-64 Language Extensions</h3>
754<!-- ======================================================================= -->
755
756<p>The X86 backend has these language extensions:</p>
757
758<!-- ======================================================================= -->
759<h4 id="x86-gs-segment">Memory references off the GS segment</h4>
760<!-- ======================================================================= -->
761
762<p>Annotating a pointer with address space #256 causes it to be code generated
Chris Lattnera021e7c2009-05-05 18:54:47 +0000763relative to the X86 GS segment register, and address space #257 causes it to be
764relative to the X86 FS segment. Note that this is a very very low-level
765feature that should only be used if you know what you're doing (for example in
766an OS kernel).</p>
Chris Lattner1177f912009-04-09 19:58:15 +0000767
768<p>Here is an example:</p>
769
770<pre>
771#define GS_RELATIVE __attribute__((address_space(256)))
772int foo(int GS_RELATIVE *P) {
773 return *P;
774}
775</pre>
776
777<p>Which compiles to (on X86-32):</p>
778
779<pre>
780_foo:
781 movl 4(%esp), %eax
782 movl %gs:(%eax), %eax
783 ret
784</pre>
785
Ted Kremeneked869312009-04-10 05:03:33 +0000786<!-- ======================================================================= -->
787<h2 id="analyzerspecific">Static Analysis-Specific Extensions</h2>
788<!-- ======================================================================= -->
789
790<p>Clang supports additional attributes that are useful for documenting program
791invariants and rules for static analysis tools. The extensions documented here
792are used by the <a
793href="http://clang.llvm.org/StaticAnalysis.html">path-sensitive static analyzer
794engine</a> that is part of Clang's Analysis library.</p>
795
John McCall87494012011-03-18 03:51:49 +0000796<h3 id="attr_analyzer_noreturn">The <tt>analyzer_noreturn</tt> attribute</h3>
Ted Kremeneked869312009-04-10 05:03:33 +0000797
798<p>Clang's static analysis engine understands the standard <tt>noreturn</tt>
Ted Kremenek4df21142009-04-10 05:04:22 +0000799attribute. This attribute, which is typically affixed to a function prototype,
800indicates that a call to a given function never returns. Function prototypes for
801common functions like <tt>exit</tt> are typically annotated with this attribute,
802as well as a variety of common assertion handlers. Users can educate the static
803analyzer about their own custom assertion handles (thus cutting down on false
804positives due to false paths) by marking their own &quot;panic&quot; functions
805with this attribute.</p>
Ted Kremeneked869312009-04-10 05:03:33 +0000806
807<p>While useful, <tt>noreturn</tt> is not applicable in all cases. Sometimes
Nick Lewycky625b5862009-06-14 04:08:08 +0000808there are special functions that for all intents and purposes should be
809considered panic functions (i.e., they are only called when an internal program
810error occurs) but may actually return so that the program can fail gracefully.
811The <tt>analyzer_noreturn</tt> attribute allows one to annotate such functions
812as being interpreted as &quot;no return&quot; functions by the analyzer (thus
Chris Lattner28935892009-04-10 05:54:56 +0000813pruning bogus paths) but will not affect compilation (as in the case of
Ted Kremeneked869312009-04-10 05:03:33 +0000814<tt>noreturn</tt>).</p>
815
816<p><b>Usage</b>: The <tt>analyzer_noreturn</tt> attribute can be placed in the
Chris Lattner28935892009-04-10 05:54:56 +0000817same places where the <tt>noreturn</tt> attribute can be placed. It is commonly
Ted Kremeneked869312009-04-10 05:03:33 +0000818placed at the end of function prototypes:</p>
819
820<pre>
821 void foo() <b>__attribute__((analyzer_noreturn))</b>;
Chris Lattner148772a2009-06-13 07:13:28 +0000822</pre>
823
John McCall87494012011-03-18 03:51:49 +0000824<p>Query for this feature with
825<tt>__has_attribute(analyzer_noreturn)</tt>.</p>
Chris Lattner148772a2009-06-13 07:13:28 +0000826
John McCall87494012011-03-18 03:51:49 +0000827<h3 id="attr_method_family">The <tt>objc_method_family</tt> attribute</h3>
828
829<p>Many methods in Objective-C have conventional meanings determined
830by their selectors. For the purposes of static analysis, it is
831sometimes useful to be able to mark a method as having a particular
832conventional meaning despite not having the right selector, or as not
833having the conventional meaning that its selector would suggest.
834For these use cases, we provide an attribute to specifically describe
835the <q>method family</q> that a method belongs to.</p>
836
837<p><b>Usage</b>: <tt>__attribute__((objc_method_family(X)))</tt>,
838where <tt>X</tt> is one of <tt>none</tt>, <tt>alloc</tt>, <tt>copy</tt>,
839<tt>init</tt>, <tt>mutableCopy</tt>, or <tt>new</tt>. This attribute
840can only be placed at the end of a method declaration:</p>
841
842<pre>
843 - (NSString*) initMyStringValue <b>__attribute__((objc_method_family(none)))</b>;
844</pre>
845
846<p>Users who do not wish to change the conventional meaning of a
847method, and who merely want to document its non-standard retain and
848release semantics, should use the
849<a href="#attr_retain_release">retaining behavior attributes</a>
850described below.</p>
851
852<p>Query for this feature with
853<tt>__has_attribute(objc_method_family)</tt>.</p>
854
855<h3 id="attr_retain_release">Objective-C retaining behavior attributes</h3>
John McCall630b7ae2011-01-25 04:26:21 +0000856
857<p>In Objective-C, functions and methods are generally assumed to take
858and return objects with +0 retain counts, with some exceptions for
859special methods like <tt>+alloc</tt> and <tt>init</tt>. However,
860there are exceptions, and so Clang provides attributes to allow these
861exceptions to be documented, which helps the analyzer find leaks (and
John McCall87494012011-03-18 03:51:49 +0000862ignore non-leaks). Some exceptions may be better described using
863the <a href="#attr_method_family"><tt>objc_method_family</tt></a>
864attribute instead.</p>
John McCall630b7ae2011-01-25 04:26:21 +0000865
866<p><b>Usage</b>: The <tt>ns_returns_retained</tt>, <tt>ns_returns_not_retained</tt>,
867<tt>ns_returns_autoreleased</tt>, <tt>cf_returns_retained</tt>,
868and <tt>cf_returns_not_retained</tt> attributes can be placed on
869methods and functions that return Objective-C or CoreFoundation
870objects. They are commonly placed at the end of a function prototype
871or method declaration:</p>
872
873<pre>
874 id foo() <b>__attribute__((ns_returns_retained))</b>;
875
876 - (NSString*) bar: (int) x <b>__attribute__((ns_returns_retained))</b>;
877</pre>
878
879<p>The <tt>*_returns_retained</tt> attributes specify that the
880returned object has a +1 retain count.
881The <tt>*_returns_not_retained</tt> attributes specify that the return
882object has a +0 retain count, even if the normal convention for its
883selector would be +1. <tt>ns_returns_autoreleased</tt> specifies that the
884returned object is +0, but is guaranteed to live at least as long as the
885next flush of an autorelease pool.</p>
886
887<p><b>Usage</b>: The <tt>ns_consumed</tt> and <tt>cf_consumed</tt>
888attributes can be placed on an parameter declaration; they specify
889that the argument is expected to have a +1 retain count, which will be
890balanced in some way by the function or method.
891The <tt>ns_consumes_self</tt> attribute can only be placed on an
892Objective-C method; it specifies that the method expects
893its <tt>self</tt> parameter to have a +1 retain count, which it will
894balance in some way.</p>
895
896<pre>
897 void <b>foo(__attribute__((ns_consumed))</b> NSString *string);
898
899 - (void) bar <b>__attribute__((ns_consumes_self))</b>;
900 - (void) baz: (id) <b>__attribute__((ns_consumed))</b> x;
901</pre>
Ted Kremeneked869312009-04-10 05:03:33 +0000902
John McCall87494012011-03-18 03:51:49 +0000903<p>Query for these features with <tt>__has_attribute(ns_consumed)</tt>,
904<tt>__has_attribute(ns_returns_retained)</tt>, etc.</p>
905
Chris Lattner5ce933f2009-02-09 08:46:11 +0000906</div>
907</body>
908</html>