Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / clang / cccc02ac215ffc16fc4b3ebfee5ee87c40e30505 / . / docs / UsersManual.html
blob: 82c4fa27321eecbfbe03291185ee6a4e2eb67754 [file] [log] [blame]
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]1<html>
2<head>
3<title>Clang Compiler User's Manual</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 Compiler User's Manual</h1>
19
20<ul>
21<li><a href="#intro">Introduction</a>
22 <ul>
23 <li><a href="#terminology">Terminology</a></li>
24 <li><a href="#basicusage">Basic Usage</a></li>
25 </ul>
26</li>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]27<li><a href="#commandline">Command Line Options</a>
28 <ul>
29 <li><a href="#cl_diagnostics">Options to Control Error and Warning
30 Messages</a></li>
31 </ul>
32</li>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]33<li><a href="#general_features">Language and Target-Independent Features</a>
34 <ul>
35 <li><a href="#diagnostics">Controlling Errors and Warnings</a></li>
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]36 <ul>
37 <li><a href="#diagnostics_display">Controlling How Clang Displays Diagnostics</a></li>
38 <li><a href="#diagnostics_mappings">Diagnostic Mappings</a></li>
Chris Lattner67db8cd2010-05-30 23:42:51 +0000[diff] [blame]39 <li><a href="#diagnostics_categories">Diagnostic Categories</a></li>
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]40 <li><a href="#diagnostics_commandline">Controlling Diagnostics via Command Line Flags</a></li>
41 <li><a href="#diagnostics_pragmas">Controlling Diagnostics via Pragmas</a></li>
Ted Kremenek2fb11eb2010-08-24 18:12:35 +0000[diff] [blame]42 <li><a href="#analyzer_diagnositics">Controlling Static Analyzer Diagnostics</a></li>
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]43 </ul>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]44 <li><a href="#precompiledheaders">Precompiled Headers</a></li>
Mike Stump53664ca2009-12-14 23:53:10 +0000[diff] [blame]45 <li><a href="#codegen">Controlling Code Generation</a></li>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]46 </ul>
47</li>
48<li><a href="#c">C Language Features</a>
49 <ul>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]50 <li><a href="#c_ext">Extensions supported by clang</a></li>
51 <li><a href="#c_modes">Differences between various standard modes</a></li>
52 <li><a href="#c_unimpl_gcc">GCC extensions not implemented yet</a></li>
53 <li><a href="#c_unsupp_gcc">Intentionally unsupported GCC extensions</a></li>
54 <li><a href="#c_ms">Microsoft extensions</a></li>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]55 </ul>
56</li>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]57<li><a href="#target_features">Target-Specific Features and Limitations</a>
58 <ul>
59 <li><a href="#target_arch">CPU Architectures Features and Limitations</a>
60 <ul>
61 <li><a href="#target_arch_x86">X86</a></li>
Eli Friedman3b658d32009-06-08 05:12:39 +0000[diff] [blame]62 <li><a href="#target_arch_arm">ARM</a></li>
63 <li><a href="#target_arch_other">Other platforms</a></li>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]64 </ul>
65 </li>
66 <li><a href="#target_os">Operating System Features and Limitations</a>
67 <ul>
68 <li><a href="#target_os_darwin">Darwin (Mac OS/X)</a></li>
69 <li>Linux, etc.</li>
70 </ul>
71
72 </li>
73 </ul>
74</li>
75</ul>
76
77
78<!-- ======================================================================= -->
79<h2 id="intro">Introduction</h2>
80<!-- ======================================================================= -->
81
82<p>The Clang Compiler is an open-source compiler for the C family of programming
83languages, aiming to be the best in class implementation of these languages.
84Clang builds on the LLVM optimizer and code generator, allowing it to provide
85high-quality optimization and code generation support for many targets. For
86more general information, please see the <a href="http://clang.llvm.org">Clang
87Web Site</a> or the <a href="http://llvm.org">LLVM Web Site</a>.</p>
88
89<p>This document describes important notes about using Clang as a compiler for
90an end-user, documenting the supported features, command line options, etc. If
91you are interested in using Clang to build a tool that processes code, please
92see <a href="InternalsManual.html">the Clang Internals Manual</a>. If you are
93interested in the <a href="http://clang.llvm.org/StaticAnalysis.html">Clang
94Static Analyzer</a>, please see its web page.</p>
95
96<p>Clang is designed to support the C family of programming languages, which
97includes <a href="#c">C</a>, <a href="#objc">Objective-C</a>, <a
98href="#cxx">C++</a>, and <a href="#objcxx">Objective-C++</a> as well as many
99dialects of those. For language-specific information, please see the
100corresponding language specific section:</p>
101
102<ul>
103<li><a href="#c">C Language</a>: K&amp;R C, ANSI C89, ISO C90, ISO C94
104 (C89+AMD1), ISO C99 (+TC1, TC2, TC3). </li>
105<li><a href="#objc">Objective-C Language</a>: ObjC 1, ObjC 2, ObjC 2.1, plus
106 variants depending on base language.</li>
107<li><a href="#cxx">C++ Language Features</a></li>
108<li><a href="#objcxx">Objective C++ Language</a></li>
109</ul>
110
111<p>In addition to these base languages and their dialects, Clang supports a
112broad variety of language extensions, which are documented in the corresponding
113language section. These extensions are provided to be compatible with the GCC,
114Microsoft, and other popular compilers as well as to improve functionality
115through Clang-specific features. The Clang driver and language features are
116intentionally designed to be as compatible with the GNU GCC compiler as
117reasonably possible, easing migration from GCC to Clang. In most cases, code
118"just works".</p>
119
120<p>In addition to language specific features, Clang has a variety of features
121that depend on what CPU architecture or operating system is being compiled for.
Douglas Gregorcd5a5052009-11-09 15:15:41 +0000[diff] [blame]122Please see the <a href="#target_features">Target-Specific Features and
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]123Limitations</a> section for more details.</p>
124
125<p>The rest of the introduction introduces some basic <a
126href="#terminology">compiler terminology</a> that is used throughout this manual
127and contains a basic <a href="#basicusage">introduction to using Clang</a>
128as a command line compiler.</p>
129
130<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
131<h3 id="terminology">Terminology</h3>
132<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
133
134<p>Front end, parser, backend, preprocessor, undefined behavior, diagnostic,
135 optimizer</p>
136
137<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
138<h3 id="basicusage">Basic Usage</h3>
139<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
140
141<p>Intro to how to use a C compiler for newbies.</p>
142<p>
143compile + link
144
145compile then link
146
147debug info
148
149enabling optimizations
150
151picking a language to use, defaults to C99 by default. Autosenses based on
152extension.
153
154using a makefile
155</p>
156
157
158<!-- ======================================================================= -->
159<h2 id="commandline">Command Line Options</h2>
160<!-- ======================================================================= -->
161
162<p>
163This section is generally an index into other sections. It does not go into
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]164depth on the ones that are covered by other sections. However, the first part
165introduces the language selection and other high level options like -c, -g, etc.
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]166</p>
167
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]168
169<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
170<h3 id="cl_diagnostics">Options to Control Error and Warning Messages</h3>
171<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
172
173<p><b>-Werror</b>: Turn warnings into errors.</p>
174<p><b>-Werror=foo</b>: Turn warning "foo" into an error.</p>
175<p><b>-Wno-error=foo</b>: Turn warning "foo" into an warning even if -Werror is
176 specified.</p>
177<p><b>-Wfoo</b>: Enable warning foo</p>
178<p><b>-Wno-foo</b>: Disable warning foo</p>
179<p><b>-w</b>: Disable all warnings.</p>
180<p><b>-pedantic</b>: Warn on language extensions.</p>
181<p><b>-pedantic-errors</b>: Error on language extensions.</p>
182<p><b>-Wsystem-headers</b>: Enable warnings from system headers.</p>
183
Chris Lattner0f0c9632010-04-07 20:49:23 +0000[diff] [blame]184<p><b>-ferror-limit=123</b>: Stop emitting diagnostics after 123 errors have
185 been produced. The default is 20, and the error limit can be disabled with
186 -ferror-limit=0.</p>
187
Douglas Gregor575cf372010-04-20 07:18:24 +0000[diff] [blame]188<p><b>-ftemplate-backtrace-limit=123</b>: Only emit up to 123 template instantiation notes within the template instantiation backtrace for a single warning or error. The default is 10, and the limit can be disabled with -ftemplate-backtrace-limit=0.</p>
189
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]190<!-- ================================================= -->
191<h4 id="cl_diag_formatting">Formatting of Diagnostics</h4>
192<!-- ================================================= -->
193
194<p>Clang aims to produce beautiful diagnostics by default, particularly for new
195users that first come to Clang. However, different people have different
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]196preferences, and sometimes Clang is driven by another program that wants to
197parse simple and consistent output, not a person. For these cases, Clang
198provides a wide range of options to control the exact output format of the
199diagnostics that it generates.</p>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]200
201<dl>
202
203<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
204<dt id="opt_fshow-column"><b>-f[no-]show-column</b>: Print column number in
205diagnostic.</dt>
206<dd>This option, which defaults to on, controls whether or not Clang prints the
207column number of a diagnostic. For example, when this is enabled, Clang will
208print something like:</p>
209
210<pre>
211 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
212 #endif bad
213 ^
214 //
215</pre>
216
217<p>When this is disabled, Clang will print "test.c:28: warning..." with no
218column number.</p>
219</dd>
220
221<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
222<dt id="opt_fshow-source-location"><b>-f[no-]show-source-location</b>: Print
223source file/line/column information in diagnostic.</dt>
224<dd>This option, which defaults to on, controls whether or not Clang prints the
225filename, line number and column number of a diagnostic. For example,
226when this is enabled, Clang will print something like:</p>
227
228<pre>
229 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
230 #endif bad
231 ^
232 //
233</pre>
234
235<p>When this is disabled, Clang will not print the "test.c:28:8: " part.</p>
236</dd>
237
238<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
239<dt id="opt_fcaret-diagnostics"><b>-f[no-]caret-diagnostics</b>: Print source
240line and ranges from source code in diagnostic.</dt>
241<dd>This option, which defaults to on, controls whether or not Clang prints the
242source line, source ranges, and caret when emitting a diagnostic. For example,
243when this is enabled, Clang will print something like:</p>
244
245<pre>
246 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
247 #endif bad
248 ^
249 //
250</pre>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]251</dd>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]252<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
Douglas Gregore8d44dd2010-07-09 16:31:58 +0000[diff] [blame]253<dt id="opt_fcolor_diagnostics"><b>-f[no-]color-diagnostics</b>: </dt>
254<dd>This option, which defaults to on when a color-capable terminal is
255 detected, controls whether or not Clang prints diagnostics in color.
256 When this option is enabled, Clang will use colors to highlight
257 specific parts of the diagnostic, e.g.,
Chris Lattner4e1c53d2011-01-24 03:47:34 +0000[diff] [blame]258 <pre>
259 <b><font color="black">test.c:28:8: <font color="magenta">warning</font>: extra tokens at end of #endif directive [-Wextra-tokens]</font></b>
260 #endif bad
261 <font color="green">^</font>
262 <font color="green">//</font>
Douglas Gregore8d44dd2010-07-09 16:31:58 +0000[diff] [blame]263</pre>
Chris Lattner4e1c53d2011-01-24 03:47:34 +0000[diff] [blame]264
265<p>When this is disabled, Clang will just print:</p>
266
267<pre>
268 test.c:2:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
269 #endif bad
270 ^
271 //
272</pre>
273</dd>
Douglas Gregore8d44dd2010-07-09 16:31:58 +0000[diff] [blame]274<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]275<dt id="opt_fdiagnostics-show-option"><b>-f[no-]diagnostics-show-option</b>:
276Enable <tt>[-Woption]</tt> information in diagnostic line.</dt>
277<dd>This option, which defaults to on,
278controls whether or not Clang prints the associated <A
279href="#cl_diag_warning_groups">warning group</a> option name when outputting
280a warning diagnostic. For example, in this output:</p>
281
282<pre>
283 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
284 #endif bad
285 ^
286 //
287</pre>
288
289<p>Passing <b>-fno-diagnostics-show-option</b> will prevent Clang from printing
290the [<a href="#opt_Wextra-tokens">-Wextra-tokens</a>] information in the
291diagnostic. This information tells you the flag needed to enable or disable the
292diagnostic, either from the command line or through <a
293href="#pragma_GCC_diagnostic">#pragma GCC diagnostic</a>.</dd>
294
Chris Lattner28a43a42010-05-05 01:35:28 +0000[diff] [blame]295<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
296<dt id="opt_fdiagnostics-show-category"><b>-fdiagnostics-show-category=none/id/name</b>:
297Enable printing category information in diagnostic line.</dt>
298<dd>This option, which defaults to "none",
299controls whether or not Clang prints the category associated with a diagnostic
300when emitting it. Each diagnostic may or many not have an associated category,
301if it has one, it is listed in the diagnostic categorization field of the
302diagnostic line (in the []'s).</p>
303
304<p>For example, a format string warning will produce these three renditions
305based on the setting of this option:</p>
306
307<pre>
308 t.c:3:11: warning: conversion specifies type 'char *' but the argument has type 'int' [-Wformat]
309 t.c:3:11: warning: conversion specifies type 'char *' but the argument has type 'int' [-Wformat<b>,1</b>]
310 t.c:3:11: warning: conversion specifies type 'char *' but the argument has type 'int' [-Wformat<b>,Format String</b>]
311</pre>
312
313<p>This category can be used by clients that want to group diagnostics by
314category, so it should be a high level category. We want dozens of these, not
315hundreds or thousands of them.</p>
316</dd>
317
318
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]319
320<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
321<dt id="opt_fdiagnostics-fixit-info"><b>-f[no-]diagnostics-fixit-info</b>:
322Enable "FixIt" information in the diagnostics output.</dt>
323<dd>This option, which defaults to on, controls whether or not Clang prints the
324information on how to fix a specific diagnostic underneath it when it knows.
325For example, in this output:</p>
326
327<pre>
328 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
329 #endif bad
330 ^
331 //
332</pre>
333
334<p>Passing <b>-fno-diagnostics-fixit-info</b> will prevent Clang from printing
335the "//" line at the end of the message. This information is useful for users
336who may not understand what is wrong, but can be confusing for machine
337parsing.</p>
338</dd>
339
340<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
Chris Lattner2a9cc232009-04-21 05:35:32 +0000[diff] [blame]341<dt id="opt_fdiagnostics-print-source-range-info">
342<b>-f[no-]diagnostics-print-source-range-info</b>:
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]343Print machine parsable information about source ranges.</dt>
344<dd>This option, which defaults to off, controls whether or not Clang prints
345information about source ranges in a machine parsable format after the
346file/line/column number information. The information is a simple sequence of
347brace enclosed ranges, where each range lists the start and end line/column
348locations. For example, in this output:</p>
349
350<pre>
351exprs.c:47:15:{47:8-47:14}{47:17-47:24}: error: invalid operands to binary expression ('int *' and '_Complex float')
352 P = (P-42) + Gamma*4;
353 ~~~~~~ ^ ~~~~~~~
354</pre>
355
Chris Lattner2a9cc232009-04-21 05:35:32 +0000[diff] [blame]356<p>The {}'s are generated by -fdiagnostics-print-source-range-info.</p>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]357</dd>
358
Douglas Gregor4786c152010-08-19 20:24:43 +0000[diff] [blame]359<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
360<dt id="opt_fdiagnostics-parseable-fixits">
361<b>-fdiagnostics-parseable-fixits</b>:
362Print Fix-Its in a machine parseable form.</dt>
363<dd><p>This option makes Clang print available Fix-Its in a machine parseable format at the end of diagnostics. The following example illustrates the format:</p>
364
365<pre>
Douglas Gregorbf5e09d2010-08-20 03:17:33 +0000[diff] [blame]366 fix-it:"t.cpp":{7:25-7:29}:"Gamma"
Douglas Gregor4786c152010-08-19 20:24:43 +0000[diff] [blame]367</pre>
368
Douglas Gregoree45e242010-08-23 14:03:03 +0000[diff] [blame]369<p>The range printed is a half-open range, so in this example the characters at column 25 up to but not including column 29 on line 7 in t.cpp should be replaced with the string "Gamma". Either the range or the replacement string may be empty (representing strict insertions and strict erasures, respectively). Both the file name and the insertion string escape backslash (as "\\"), tabs (as "\t"), newlines (as "\n"), double quotes(as "\"") and non-printable characters (as octal "\xxx").</p>
Douglas Gregor4786c152010-08-19 20:24:43 +0000[diff] [blame]370</dd>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]371
372</dl>
373
374
375
376
377<!-- ===================================================== -->
378<h4 id="cl_diag_warning_groups">Individual Warning Groups</h4>
379<!-- ===================================================== -->
380
381<p>TODO: Generate this from tblgen. Define one anchor per warning group.</p>
382
383
384<dl>
385
386
387<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
388<dt id="opt_Wextra-tokens"><b>-Wextra-tokens</b>: Warn about excess tokens at
389 the end of a preprocessor directive.</dt>
390<dd>This option, which defaults to on, enables warnings about extra tokens at
391the end of preprocessor directives. For example:</p>
392
393<pre>
394 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
395 #endif bad
396 ^
397</pre>
398
399<p>These extra tokens are not strictly conforming, and are usually best handled
400by commenting them out.</p>
401
402<p>This option is also enabled by <a href="">-Wfoo</a>, <a href="">-Wbar</a>,
403 and <a href="">-Wbaz</a>.</p>
404</dd>
405
Jeffrey Yasskin21d07e42010-06-05 01:39:57 +0000[diff] [blame]406<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
407<dt id="opt_Wambiguous-member-template"><b>-Wambiguous-member-template</b>:
408Warn about unqualified uses of a member template whose name resolves
409to another template at the location of the use.</dt>
410<dd>This option, which defaults to on, enables a warning in the
411following code:</p>
412
413<pre>
414template&lt;typename T> struct set{};
415template&lt;typename T> struct trait { typedef const T& type; };
416struct Value {
417 template&lt;typename T> void set(typename trait&lt;T>::type value) {}
418};
419void foo() {
420 Value v;
421 v.set&lt;double>(3.2);
422}
423</pre>
424
425<p>C++ [basic.lookup.classref] requires this to be an error, but,
426because it's hard to work around, Clang downgrades it to a warning as
427an extension.</p>
428</dd>
429
Jeffrey Yasskin57d12fd2010-06-07 15:58:05 +0000[diff] [blame]430<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
431<dt id="opt_Wbind-to-temporary-copy"><b>-Wbind-to-temporary-copy</b>: Warn about
432an unusable copy constructor when binding a reference to a temporary.</dt>
433<dd>This option, which defaults to on, enables warnings about binding a
434reference to a temporary when the temporary doesn't have a usable copy
435constructor. For example:</p>
436
437<pre>
438 struct NonCopyable {
439 NonCopyable();
440 private:
441 NonCopyable(const NonCopyable&);
442 };
443 void foo(const NonCopyable&);
444 void bar() {
445 foo(NonCopyable()); // Disallowed in C++98; allowed in C++0x.
446 }
447</pre>
448<pre>
449 struct NonCopyable2 {
450 NonCopyable2();
451 NonCopyable2(NonCopyable2&);
452 };
453 void foo(const NonCopyable2&);
454 void bar() {
455 foo(NonCopyable2()); // Disallowed in C++98; allowed in C++0x.
456 }
457</pre>
458
459<p>Note that if <tt>NonCopyable2::NonCopyable2()</tt> has a default
460argument whose instantiation produces a compile error, that error will
461still be a hard error in C++98 mode even if this warning is turned
462off.</p>
463
464</dd>
465
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]466</dl>
467
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]468<!-- ======================================================================= -->
469<h2 id="general_features">Language and Target-Independent Features</h2>
470<!-- ======================================================================= -->
471
472
473<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
474<h3 id="diagnostics">Controlling Errors and Warnings</h3>
475<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
476
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]477<p>Clang provides a number of ways to control which code constructs cause it to
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]478emit errors and warning messages, and how they are displayed to the console.</p>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]479
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]480<h4 id="diagnostics_display">Controlling How Clang Displays Diagnostics</h4>
Chris Lattner65a795b2009-04-20 06:00:23 +0000[diff] [blame]481
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]482<p>When Clang emits a diagnostic, it includes rich information in the output,
483and gives you fine-grain control over which information is printed. Clang has
484the ability to print this information, and these are the options that control
485it:</p>
486
487<p>
488<ol>
489<li>A file/line/column indicator that shows exactly where the diagnostic occurs
490 in your code [<a href="#opt_fshow-column">-fshow-column</a>, <a
491 href="#opt_fshow-source-location">-fshow-source-location</a>].</li>
492<li>A categorization of the diagnostic as a note, warning, error, or fatal
493 error.</li>
494<li>A text string that describes what the problem is.</li>
495<li>An option that indicates how to control the diagnostic (for diagnostics that
496 support it) [<a
497 href="#opt_fdiagnostics-show-option">-fdiagnostics-show-option</a>].</li>
Chris Lattner3f145382010-05-24 21:35:18 +0000[diff] [blame]498<li>A <a href="#diagnostics_categories">high-level category</a> for the
499 diagnostic for clients that want to group diagnostics by class (for
500 diagnostics that support it) [<a
Chris Lattner28a43a42010-05-05 01:35:28 +0000[diff] [blame]501 href="#opt_fdiagnostics-show-category">-fdiagnostics-show-category</a>].</li>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]502<li>The line of source code that the issue occurs on, along with a caret and
503 ranges that indicate the important locations [<a
504 href="opt_fcaret-diagnostics">-fcaret-diagnostics</a>].</li>
505<li>"FixIt" information, which is a concise explanation of how to fix the
506 problem (when Clang is certain it knows) [<a
507 href="opt_fdiagnostics-fixit-info">-fdiagnostics-fixit-info</a>].</li>
508<li>A machine-parsable representation of the ranges involved (off by
509 default) [<a
Chris Lattner2a9cc232009-04-21 05:35:32 +0000[diff] [blame]510 href="opt_fdiagnostics-print-source-range-info">-fdiagnostics-print-source-range-info</a>].</li>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]511</ol></p>
512
513<p>For more information please see <a href="#cl_diag_formatting">Formatting of
514Diagnostics</a>.</p>
515
Chris Lattner3f145382010-05-24 21:35:18 +0000[diff] [blame]516
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]517<h4 id="diagnostics_mappings">Diagnostic Mappings</h4>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]518
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]519<p>All diagnostics are mapped into one of these 5 classes:</p>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]520
521<p>
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]522<ul>
523<li>Ignored</li>
524<li>Note</li>
525<li>Warning</li>
526<li>Error</li>
527<li>Fatal</li>
528</ul></p>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]529
Chris Lattner3f145382010-05-24 21:35:18 +0000[diff] [blame]530<h4 id="diagnostics_categories">Diagnostic Categories</h4>
531
532<p>Though not shown by default, diagnostics may each be associated with a
533 high-level category. This category is intended to make it possible to triage
534 builds that produce a large number of errors or warnings in a grouped way.
535</p>
536
537<p>Categories are not shown by default, but they can be turned on with the
538<a href="#opt_fdiagnostics-show-category">-fdiagnostics-show-category</a> option.
539When set to "<tt>name</tt>", the category is printed textually in the diagnostic
540output. When it is set to "<tt>id</tt>", a category number is printed. The
541mapping of category names to category id's can be obtained by running '<tt>clang
542 --print-diagnostic-categories</tt>'.
543</p>
544
545<h4 id="diagnostics_commandline">Controlling Diagnostics via Command Line
546 Flags</h4>
Chris Lattner8217f4e2009-04-20 06:26:18 +0000[diff] [blame]547
548<p>-W flags, -pedantic, etc</p>
549
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]550<h4 id="diagnostics_pragmas">Controlling Diagnostics via Pragmas</h4>
551
552<p>Clang can also control what diagnostics are enabled through the use of
553pragmas in the source code. This is useful for turning off specific warnings
554in a section of source code. Clang supports GCC's pragma for compatibility
555with existing source code, as well as several extensions. </p>
556
557<p>The pragma may control any warning that can be used from the command line.
558Warnings may be set to ignored, warning, error, or fatal. The following
559example code will tell Clang or GCC to ignore the -Wall warnings:</p>
560
561<pre>
562#pragma GCC diagnostic ignored "-Wall"
563</pre>
564
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]565<p>In addition to all of the functionality provided by GCC's pragma, Clang
Chris Lattner3401cf82009-07-12 21:22:10 +0000[diff] [blame]566also allows you to push and pop the current warning state. This is particularly
567useful when writing a header file that will be compiled by other people, because
568you don't know what warning flags they build with.</p>
569
570<p>In the below example
571-Wmultichar is ignored for only a single line of code, after which the
572diagnostics return to whatever state had previously existed.</p>
573
574<pre>
575#pragma clang diagnostic push
576#pragma clang diagnostic ignored "-Wmultichar"
577
578char b = 'df'; // no warning.
579
580#pragma clang diagnostic pop
581</pre>
582
583<p>The push and pop pragmas will save and restore the full diagnostic state of
584the compiler, regardless of how it was set. That means that it is possible to
585use push and pop around GCC compatible diagnostics and Clang will push and pop
586them appropriately, while GCC will ignore the pushes and pops as unknown
587pragmas. It should be noted that while Clang supports the GCC pragma, Clang and
588GCC do not support the exact same set of warnings, so even when using GCC
589compatible #pragmas there is no guarantee that they will have identical behaviour
590on both compilers. </p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]591
Ted Kremenek2fb11eb2010-08-24 18:12:35 +0000[diff] [blame]592<h4 id="analyzer_diagnositics">Controlling Static Analyzer Diagnostics</h4>
593
594<p>While not strictly part of the compiler, the diagnostics from Clang's <a
595href="http://clang-analyzer.llvm.org">static analyzer</a> can also be influenced
596by the user via changes to the source code. This can be done in two ways:
597
598<ul>
599
600<li id="analyzer_annotations"><b>Annotations</b>: The static analyzer recognizes various GCC-style
601attributes (e.g., <tt>__attribute__((nonnull)))</tt>) that can either suppress
602static analyzer warnings or teach the analyzer about code invariants which
603enable it to find more bugs. While many of these attributes are standard GCC
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]604attributes, additional ones have been added to Clang to specifically support the
Ted Kremenek2fb11eb2010-08-24 18:12:35 +0000[diff] [blame]605static analyzer. Detailed information on these annotations can be found in the
606<a href="http://clang-analyzer.llvm.org/annotations.html">analyzer's
607documentation</a>.</li>
608
609<li><b><tt>__clang_analyzer__</tt></b>: When the static analyzer is using Clang
610to parse source files, it implicitly defines the preprocessor macro
611<tt>__clang_analyzer__</tt>. While discouraged, code can use this macro to
612selectively exclude code the analyzer examines. Here is an example:
613
614<pre>
615#ifndef __clang_analyzer__
616// Code not to be analyzed
617#endif
618</pre>
619
620In general, this usage is discouraged. Instead, we prefer that users file bugs
621against the analyzer when it flags false positives. There is also active
622discussion of allowing users in the future to selectively silence specific
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]623analyzer warnings (some of which can already be done using <a
Ted Kremenek2fb11eb2010-08-24 18:12:35 +0000[diff] [blame]624href="analyzer_annotations">annotations</a>).</li>
625
626</ul>
627
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]628<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
629<h3 id="precompiledheaders">Precompiled Headers</h3>
630<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
631
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]632<p><a href="http://en.wikipedia.org/wiki/Precompiled_header">Precompiled
633headers</a> are a general approach employed by many compilers to reduce
634compilation time. The underlying motivation of the approach is that it is
635common for the same (and often large) header files to be included by
636multiple source files. Consequently, compile times can often be greatly improved
637by caching some of the (redundant) work done by a compiler to process headers.
638Precompiled header files, which represent one of many ways to implement
639this optimization, are literally files that represent an on-disk cache that
640contains the vital information necessary to reduce some of the work
641needed to process a corresponding header file. While details of precompiled
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]642headers vary between compilers, precompiled headers have been shown to be
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]643highly effective at speeding up program compilation on systems with very large
644system headers (e.g., Mac OS/X).</p>
645
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]646<h4>Generating a PCH File</h4>
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]647
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]648<p>To generate a PCH file using Clang, one invokes Clang with
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]649the <b><tt>-x <i>&lt;language&gt;</i>-header</tt></b> option. This mirrors the
650interface in GCC for generating PCH files:</p>
651
652<pre>
653 $ gcc -x c-header test.h -o test.h.gch
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]654 $ clang -x c-header test.h -o test.h.pch
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]655</pre>
656
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]657<h4>Using a PCH File</h4>
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]658
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]659<p>A PCH file can then be used as a prefix header when a
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]660<b><tt>-include</tt></b> option is passed to <tt>clang</tt>:</p>
661
662<pre>
663 $ clang -include test.h test.c -o test
664</pre>
665
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]666<p>The <tt>clang</tt> driver will first check if a PCH file for <tt>test.h</tt>
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]667is available; if so, the contents of <tt>test.h</tt> (and the files it includes)
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]668will be processed from the PCH file. Otherwise, Clang falls back to
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]669directly processing the content of <tt>test.h</tt>. This mirrors the behavior of
670GCC.</p>
671
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]672<p><b>NOTE:</b> Clang does <em>not</em> automatically use PCH files
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]673for headers that are directly included within a source file. For example:</p>
674
675<pre>
Chris Lattnere42ec542009-06-13 20:35:58 +0000[diff] [blame]676 $ clang -x c-header test.h -o test.h.pch
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]677 $ cat test.c
678 #include "test.h"
679 $ clang test.c -o test
680</pre>
681
Douglas Gregorf4d59532009-06-03 22:37:00 +0000[diff] [blame]682<p>In this example, <tt>clang</tt> will not automatically use the PCH file for
Chris Lattner5c3074f2009-04-20 04:37:38 +0000[diff] [blame]683<tt>test.h</tt> since <tt>test.h</tt> was included directly in the source file
684and not specified on the command line using <tt>-include</tt>.</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]685
Douglas Gregore650c8c2009-07-07 00:12:59 +0000[diff] [blame]686<h4>Relocatable PCH Files</h4>
687<p>It is sometimes necessary to build a precompiled header from headers that
688are not yet in their final, installed locations. For example, one might build a
689precompiled header within the build tree that is then meant to be installed
690alongside the headers. Clang permits the creation of "relocatable" precompiled
691headers, which are built with a given path (into the build directory) and can
692later be used from an installed location.</p>
693
694<p>To build a relocatable precompiled header, place your headers into a
695subdirectory whose structure mimics the installed location. For example, if you
696want to build a precompiled header for the header <code>mylib.h</code> that
697will be installed into <code>/usr/include</code>, create a subdirectory
698<code>build/usr/include</code> and place the header <code>mylib.h</code> into
699that subdirectory. If <code>mylib.h</code> depends on other headers, then
700they can be stored within <code>build/usr/include</code> in a way that mimics
701the installed location.</p>
702
703<p>Building a relocatable precompiled header requires two additional arguments.
704First, pass the <code>--relocatable-pch</code> flag to indicate that the
705resulting PCH file should be relocatable. Second, pass
706<code>-isysroot /path/to/build</code>, which makes all includes for your
707library relative to the build directory. For example:</p>
708
709<pre>
710 # clang -x c-header --relocatable-pch -isysroot /path/to/build /path/to/build/mylib.h mylib.h.pch
711</pre>
712
713<p>When loading the relocatable PCH file, the various headers used in the PCH
714file are found from the system header root. For example, <code>mylib.h</code>
715can be found in <code>/usr/include/mylib.h</code>. If the headers are installed
716in some other system root, the <code>-isysroot</code> option can be used provide
717a different system root from which the headers will be based. For example,
718<code>-isysroot /Developer/SDKs/MacOSX10.4u.sdk</code> will look for
719<code>mylib.h</code> in
720<code>/Developer/SDKs/MacOSX10.4u.sdk/usr/include/mylib.h</code>.</p>
721
722<p>Relocatable precompiled headers are intended to be used in a limited number
723of cases where the compilation environment is tightly controlled and the
724precompiled header cannot be generated after headers have been installed.
725Relocatable precompiled headers also have some performance impact, because
726the difference in location between the header locations at PCH build time vs.
727at the time of PCH use requires one of the PCH optimizations,
728<code>stat()</code> caching, to be disabled. However, this change is only
729likely to affect PCH files that reference a large number of headers.</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]730
Mike Stump53664ca2009-12-14 23:53:10 +0000[diff] [blame]731<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
732<h3 id="codegen">Controlling Code Generation</h3>
733<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
734
735<p>Clang provides a number of ways to control code generation. The options are listed below.</p>
736
737<!-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
738<dt id="opt_fcatch-undefined-behavior"><b>-fcatch-undefined-behavior</b>: Turn
739on runtime code generation to check for undefined behavior.</dt>
740
741<dd>This option, which defaults to off, controls whether or not Clang
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]742adds runtime checks for undefined runtime behavior. If a check fails,
Mike Stump53664ca2009-12-14 23:53:10 +0000[diff] [blame]743<tt>__builtin_trap()</tt> is used to indicate failure.
744The checks are:
745<p>
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]746<li>Subscripting where the static type of one operand is a variable
Mike Stump88b2a172009-12-16 03:25:12 +0000[diff] [blame]747 which is decayed from an array type and the other operand is
748 greater than the size of the array or less than zero.</li>
Mike Stump53664ca2009-12-14 23:53:10 +0000[diff] [blame]749<li>Shift operators where the amount shifted is greater or equal to the
750 promoted bit-width of the left-hand-side or less than zero.</li>
Mike Stump8f6a3ed2009-12-16 03:18:14 +0000[diff] [blame]751<li>If control flow reaches __builtin_unreachable.
752<li>When llvm implements more __builtin_object_size support, reads and
753 writes for objects that __builtin_object_size indicates we aren't
754 accessing valid memory. Bit-fields and vectors are not yet checked.
Mike Stump53664ca2009-12-14 23:53:10 +0000[diff] [blame]755</p>
756</dd>
757
Nuno Lopesaa526242009-12-17 10:00:52 +0000[diff] [blame]758<dt id="opt_fno-assume-sane-operator-new"><b>-fno-assume-sane-operator-new</b>:
759Don't assume that the C++'s new operator is sane.</dt>
Nuno Lopesb23f20d2009-12-17 10:15:49 +0000[diff] [blame]760<dd>This option tells the compiler to do not assume that C++'s global new
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]761operator will always return a pointer that does not
Nuno Lopesaa526242009-12-17 10:00:52 +0000[diff] [blame]762alias any other pointer when the function returns.</dd>
763
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]764<!-- ======================================================================= -->
765<h2 id="c">C Language Features</h2>
766<!-- ======================================================================= -->
767
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]768<p>The support for standard C in clang is feature-complete except for the C99
769floating-point pragmas.</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]770
771<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]772<h3 id="c_ext">Extensions supported by clang</h3>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]773<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
774
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]775<p>See <a href="LanguageExtensions.html">clang language extensions</a>.</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]776
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]777<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
778<h3 id="c_modes">Differences between various standard modes</h3>
779<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]780
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]781<p>clang supports the -std option, which changes what language mode clang uses.
782The supported modes for C are c89, gnu89, c94, c99, gnu99 and various aliases
783for those modes. If no -std option is specified, clang defaults to gnu99 mode.
784</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]785
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]786<p>Differences between all c* and gnu* modes:</p>
787<ul>
788<li>c* modes define "__STRICT_ANSI__".</li>
Eli Friedman26fa0ed2009-05-27 23:02:57 +0000[diff] [blame]789<li>Target-specific defines not prefixed by underscores, like "linux", are
790defined in gnu* modes.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]791<li>Trigraphs default to being off in gnu* modes; they can be enabled by the
792-trigraphs option.</li>
793<li>The parser recognizes "asm" and "typeof" as keywords in gnu* modes; the
794variants "__asm__" and "__typeof__" are recognized in all modes.</li>
Eli Friedmanb0ac5452009-05-16 23:17:30 +0000[diff] [blame]795<li>The Apple "blocks" extension is recognized by default in gnu* modes
796on some platforms; it can be enabled in any mode with the "-fblocks"
797option.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]798</ul>
799
800<p>Differences between *89 and *99 modes:</p>
801<ul>
802<li>The *99 modes default to implementing "inline" as specified in C99, while
803the *89 modes implement the GNU version. This can be overridden for individual
804functions with the __gnu_inline__ attribute.</li>
Eli Friedman26fa0ed2009-05-27 23:02:57 +0000[diff] [blame]805<li>Digraphs are not recognized in c89 mode.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]806<li>The scope of names defined inside a "for", "if", "switch", "while", or "do"
807statement is different. (example: "if ((struct x {int x;}*)0) {}".)</li>
808<li>__STDC_VERSION__ is not defined in *89 modes.</li>
Eli Friedman26fa0ed2009-05-27 23:02:57 +0000[diff] [blame]809<li>"inline" is not recognized as a keyword in c89 mode.</li>
810<li>"restrict" is not recognized as a keyword in *89 modes.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]811<li>Commas are allowed in integer constant expressions in *99 modes.</li>
812<li>Arrays which are not lvalues are not implicitly promoted to pointers in
813*89 modes.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]814<li>Some warnings are different.</li>
815</ul>
816
817<p>c94 mode is identical to c89 mode except that digraphs are enabled in
818c94 mode (FIXME: And __STDC_VERSION__ should be defined!).</p>
819
820<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
821<h3 id="c_unimpl_gcc">GCC extensions not implemented yet</h3>
822<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
823
824<p>clang tries to be compatible with gcc as much as possible, but some gcc
825extensions are not implemented yet:</p>
826
827<ul>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]828
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]829<li>clang does not support #pragma weak
Eli Friedman4da92552009-06-02 08:21:31 +0000[diff] [blame]830(<a href="http://llvm.org/bugs/show_bug.cgi?id=3679">bug 3679</a>). Due to
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]831the uses described in the bug, this is likely to be implemented at some
832point, at least partially.</li>
833
Eli Friedman4da92552009-06-02 08:21:31 +0000[diff] [blame]834<li>clang does not support code generation for local variables pinned to
835registers (<a href="http://llvm.org/bugs/show_bug.cgi?id=3933">bug 3933</a>).
836This is a relatively small feature, so it is likely to be implemented
837relatively soon.</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]838
839<li>clang does not support decimal floating point types (_Decimal32 and
840friends) or fixed-point types (_Fract and friends); nobody has expressed
841interest in these features yet, so it's hard to say when they will be
842implemented.</li>
843
844<li>clang does not support nested functions; this is a complex feature which
845is infrequently used, so it is unlikely to be implemented anytime soon.</li>
846
Daniel Dunbar05fa6292009-05-12 23:12:07 +0000[diff] [blame]847<li>clang does not support global register variables, this is unlikely
Eli Friedmanca130bd2009-06-12 20:04:25 +0000[diff] [blame]848to be implemented soon because it requires additional LLVM backend support.
849</li>
Daniel Dunbar05fa6292009-05-12 23:12:07 +0000[diff] [blame]850
851<li>clang does not support static initialization of flexible array
852members. This appears to be a rarely used extension, but could be
853implemented pending user demand.</li>
854
Eli Friedmanca130bd2009-06-12 20:04:25 +0000[diff] [blame]855<li>clang does not support __builtin_va_arg_pack/__builtin_va_arg_pack_len.
856This is used rarely, but in some potentially interesting places, like the
857glibc headers, so it may be implemented pending user demand. Note that
858because clang pretends to be like GCC 4.2, and this extension was introduced
859in 4.3, the glibc headers will not try to use this extension with clang at
860the moment.</li>
861
862<li>clang does not support the gcc extension for forward-declaring function
863parameters; this has not showed up in any real-world code yet, though, so it
864might never be implemented.</li>
865
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]866</ul>
867
868<p>This is not a complete list; if you find an unsupported extension
869missing from this list, please send an e-mail to cfe-dev. This list
870currently excludes C++; see <a href="#cxx">C++ Language Features</a>.
871Also, this list does not include bugs in mostly-implemented features; please
872see the <a href="http://llvm.org/bugs/buglist.cgi?quicksearch=product%3Aclang+component%3A-New%2BBugs%2CAST%2CBasic%2CDriver%2CHeaders%2CLLVM%2BCodeGen%2Cparser%2Cpreprocessor%2CSemantic%2BAnalyzer">
873bug tracker</a> for known existing bugs (FIXME: Is there a section for
874bug-reporting guidelines somewhere?).</p>
875
876<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
877<h3 id="c_unsupp_gcc">Intentionally unsupported GCC extensions</h3>
878<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
879
Eli Friedman0b326002009-06-12 20:11:05 +0000[diff] [blame]880<ul>
881
Eli Friedman2f3fca02009-06-12 20:08:48 +0000[diff] [blame]882<li>clang does not support the gcc extension that allows variable-length arrays
Chris Lattnerd462b6a2011-01-24 03:47:59 +0000[diff] [blame]883in structures. This is for a few reasons: one, it is tricky
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]884to implement, two, the extension is completely undocumented, and three, the
Chris Lattnerb9d511c2010-10-28 02:20:32 +0000[diff] [blame]885extension appears to be rarely used. Note that clang <em>does</em> support
886flexible array members (arrays with a zero or unspecified size at the end of
887a structure).</li>
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]888
Eli Friedman2f3fca02009-06-12 20:08:48 +0000[diff] [blame]889<li>clang does not support duplicate definitions of a function where one is
Daniel Dunbarc5f928c2009-06-04 18:37:52 +0000[diff] [blame]890inline. This complicates clients of the AST which normally can expect there is
891at most one definition for each function. Source code using this feature should
892be changed to define the inline and out-of-line definitions in separate
Eli Friedman2f3fca02009-06-12 20:08:48 +0000[diff] [blame]893translation units.</li>
Daniel Dunbarc5f928c2009-06-04 18:37:52 +0000[diff] [blame]894
Eli Friedmanca130bd2009-06-12 20:04:25 +0000[diff] [blame]895<li>clang does not have an equivalent to gcc's "fold"; this means that
896clang doesn't accept some constructs gcc might accept in contexts where a
897constant expression is required, like "x-x" where x is a variable, or calls
898to C library functions like strlen.</li>
899
Eli Friedman2f3fca02009-06-12 20:08:48 +0000[diff] [blame]900<li>clang does not support multiple alternative constraints in inline asm; this
Eli Friedmanca130bd2009-06-12 20:04:25 +0000[diff] [blame]901is an extremely obscure feature which would be complicated to implement
Eli Friedman2f3fca02009-06-12 20:08:48 +0000[diff] [blame]902correctly.</li>
Eli Friedmanca130bd2009-06-12 20:04:25 +0000[diff] [blame]903
904<li>clang does not support __builtin_apply and friends; this extension is
905extremely obscure and difficult to implement reliably.</li>
906
Eli Friedman0b326002009-06-12 20:11:05 +0000[diff] [blame]907</ul>
908
Eli Friedman660a5ac2009-04-28 18:48:34 +0000[diff] [blame]909<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
910<h3 id="c_ms">Microsoft extensions</h3>
911<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
912
913<p>clang has some experimental support for extensions from
914Microsoft Visual C++; to enable it, use the -fms-extensions command-line
Eli Friedmana1821b52009-06-08 06:21:03 +0000[diff] [blame]915option. This is the default for Windows targets. Note that the
916support is incomplete; enabling Microsoft extensions will silently drop
917certain constructs (including __declspec and Microsoft-style asm statements).
918</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]919
Douglas Gregord3b227d2010-12-14 16:21:49 +0000[diff] [blame]920<li>clang allows setting _MSC_VER with -fmsc-version=. It defaults to 1300 which
Michael J. Spencerdae4ac42010-10-21 05:21:48 +0000[diff] [blame]921is the same as Visual C/C++ 2003. Any number is supported and can greatly affect
922what Windows SDK and c++stdlib headers clang can compile. This option will be
923removed when clang supports the full set of MS extensions required for these
924headers.</li>
925
Daniel Dunbar05fa6292009-05-12 23:12:07 +0000[diff] [blame]926<li>clang does not support the Microsoft extension where anonymous
Daniel Dunbar869e0542009-05-13 00:23:53 +0000[diff] [blame]927record members can be declared using user defined typedefs.</li>
Daniel Dunbar05fa6292009-05-12 23:12:07 +0000[diff] [blame]928
Daniel Dunbar9375ed12009-05-13 21:40:49 +0000[diff] [blame]929<li>clang supports the Microsoft "#pragma pack" feature for
930controlling record layout. GCC also contains support for this feature,
931however where MSVC and GCC are incompatible clang follows the MSVC
932definition.</li>
933
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]934<!-- ======================================================================= -->
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]935<h2 id="target_features">Target-Specific Features and Limitations</h2>
936<!-- ======================================================================= -->
937
938
939<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
940<h3 id="target_arch">CPU Architectures Features and Limitations</h3>
941<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
942
943<!-- ======================== -->
944<h4 id="target_arch_x86">X86</h4>
945<!-- ======================== -->
Daniel Dunbarbcaf7aa2010-09-19 19:26:59 +0000[diff] [blame]946
947<p>The support for X86 (both 32-bit and 64-bit) is considered stable on Darwin
948(Mac OS/X), Linux, FreeBSD, and Dragonfly BSD: it has been tested to correctly
949compile many large C, C++, Objective-C, and Objective-C++ codebases.</p>
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]950
Eli Friedman3b658d32009-06-08 05:12:39 +0000[diff] [blame]951<!-- ======================== -->
952<h4 id="target_arch_arm">ARM</h4>
953<!-- ======================== -->
Daniel Dunbarbcaf7aa2010-09-19 19:26:59 +0000[diff] [blame]954
955<p>The support for ARM (specifically ARMv6 and ARMv7) is considered stable on
956Darwin (iOS): it has been tested to correctly compile many large C, C++,
Bob Wilson4ea8dfa2011-01-10 17:55:14 +0000[diff] [blame]957Objective-C, and Objective-C++ codebases. Clang only supports a limited number
958of ARM architectures. It does not yet fully support ARMv5, for example.</p>
Eli Friedman3b658d32009-06-08 05:12:39 +0000[diff] [blame]959
960<!-- ======================== -->
961<h4 id="target_arch_other">Other platforms</h4>
962<!-- ======================== -->
963clang currently contains some support for PPC and Sparc; however, significant
964pieces of code generation are still missing, and they haven't undergone
965significant testing.
966
Eli Friedman3b658d32009-06-08 05:12:39 +0000[diff] [blame]967<p>clang contains limited support for the MSP430 embedded processor, but both
968the clang support and the LLVM backend support are highly experimental.
969
970<p>Other platforms are completely unsupported at the moment. Adding the
971minimal support needed for parsing and semantic analysis on a new platform
972is quite easy; see lib/Basic/Targets.cpp in the clang source tree. This level
973of support is also sufficient for conversion to LLVM IR for simple programs.
974Proper support for conversion to LLVM IR requires adding code to
975lib/CodeGen/CGCall.cpp at the moment; this is likely to change soon, though.
976Generating assembly requires a suitable LLVM backend.
Chris Lattnercf17d9d2009-04-20 04:23:09 +0000[diff] [blame]977
978<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
979<h3 id="target_os">Operating System Features and Limitations</h3>
980<!-- = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = -->
981
982<!-- ======================================= -->
983<h4 id="target_os_darwin">Darwin (Mac OS/X)</h4>
984<!-- ======================================= -->
985
986<p>No __thread support, 64-bit ObjC support requires SL tools.</p>
987
988</div>
989</body>
990</html>