1 ============================
2 Clang Compiler User's Manual
3 ============================
11 The Clang Compiler is an open-source compiler for the C family of
12 programming languages, aiming to be the best in class implementation of
13 these languages. Clang builds on the LLVM optimizer and code generator,
14 allowing it to provide high-quality optimization and code generation
15 support for many targets. For more general information, please see the
16 `Clang Web Site <http://clang.llvm.org>`_ or the `LLVM Web
17 Site <http://llvm.org>`_.
19 This document describes important notes about using Clang as a compiler
20 for an end-user, documenting the supported features, command line
21 options, etc. If you are interested in using Clang to build a tool that
22 processes code, please see :doc:`InternalsManual`. If you are interested in the
23 `Clang Static Analyzer <http://clang-analyzer.llvm.org>`_, please see its web
26 Clang is designed to support the C family of programming languages,
27 which includes :ref:`C <c>`, :ref:`Objective-C <objc>`, :ref:`C++ <cxx>`, and
28 :ref:`Objective-C++ <objcxx>` as well as many dialects of those. For
29 language-specific information, please see the corresponding language
32 - :ref:`C Language <c>`: K&R C, ANSI C89, ISO C90, ISO C94 (C89+AMD1), ISO
34 - :ref:`Objective-C Language <objc>`: ObjC 1, ObjC 2, ObjC 2.1, plus
35 variants depending on base language.
36 - :ref:`C++ Language <cxx>`
37 - :ref:`Objective C++ Language <objcxx>`
39 In addition to these base languages and their dialects, Clang supports a
40 broad variety of language extensions, which are documented in the
41 corresponding language section. These extensions are provided to be
42 compatible with the GCC, Microsoft, and other popular compilers as well
43 as to improve functionality through Clang-specific features. The Clang
44 driver and language features are intentionally designed to be as
45 compatible with the GNU GCC compiler as reasonably possible, easing
46 migration from GCC to Clang. In most cases, code "just works".
47 Clang also provides an alternative driver, :ref:`clang-cl`, that is designed
48 to be compatible with the Visual C++ compiler, cl.exe.
50 In addition to language specific features, Clang has a variety of
51 features that depend on what CPU architecture or operating system is
52 being compiled for. Please see the :ref:`Target-Specific Features and
53 Limitations <target_features>` section for more details.
55 The rest of the introduction introduces some basic :ref:`compiler
56 terminology <terminology>` that is used throughout this manual and
57 contains a basic :ref:`introduction to using Clang <basicusage>` as a
58 command line compiler.
65 Front end, parser, backend, preprocessor, undefined behavior,
73 Intro to how to use a C compiler for newbies.
75 compile + link compile then link debug info enabling optimizations
76 picking a language to use, defaults to C11 by default. Autosenses based
77 on extension. using a makefile
82 This section is generally an index into other sections. It does not go
83 into depth on the ones that are covered by other sections. However, the
84 first part introduces the language selection and other high level
85 options like :option:`-c`, :option:`-g`, etc.
87 Options to Control Error and Warning Messages
88 ---------------------------------------------
92 Turn warnings into errors.
94 .. This is in plain monospaced font because it generates the same label as
95 .. -Werror, and Sphinx complains.
99 Turn warning "foo" into an error.
101 .. option:: -Wno-error=foo
103 Turn warning "foo" into an warning even if :option:`-Werror` is specified.
107 Enable warning "foo".
111 Disable warning "foo".
115 Disable all diagnostics.
117 .. option:: -Weverything
119 :ref:`Enable all diagnostics. <diagnostics_enable_everything>`
121 .. option:: -pedantic
123 Warn on language extensions.
125 .. option:: -pedantic-errors
127 Error on language extensions.
129 .. option:: -Wsystem-headers
131 Enable warnings from system headers.
133 .. option:: -ferror-limit=123
135 Stop emitting diagnostics after 123 errors have been produced. The default is
136 20, and the error limit can be disabled with :option:`-ferror-limit=0`.
138 .. option:: -ftemplate-backtrace-limit=123
140 Only emit up to 123 template instantiation notes within the template
141 instantiation backtrace for a single warning or error. The default is 10, and
142 the limit can be disabled with :option:`-ftemplate-backtrace-limit=0`.
144 .. _cl_diag_formatting:
146 Formatting of Diagnostics
147 ^^^^^^^^^^^^^^^^^^^^^^^^^
149 Clang aims to produce beautiful diagnostics by default, particularly for
150 new users that first come to Clang. However, different people have
151 different preferences, and sometimes Clang is driven by another program
152 that wants to parse simple and consistent output, not a person. For
153 these cases, Clang provides a wide range of options to control the exact
154 output format of the diagnostics that it generates.
156 .. _opt_fshow-column:
158 **-f[no-]show-column**
159 Print column number in diagnostic.
161 This option, which defaults to on, controls whether or not Clang
162 prints the column number of a diagnostic. For example, when this is
163 enabled, Clang will print something like:
167 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
172 When this is disabled, Clang will print "test.c:28: warning..." with
175 The printed column numbers count bytes from the beginning of the
176 line; take care if your source contains multibyte characters.
178 .. _opt_fshow-source-location:
180 **-f[no-]show-source-location**
181 Print source file/line/column information in diagnostic.
183 This option, which defaults to on, controls whether or not Clang
184 prints the filename, line number and column number of a diagnostic.
185 For example, when this is enabled, Clang will print something like:
189 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
194 When this is disabled, Clang will not print the "test.c:28:8: "
197 .. _opt_fcaret-diagnostics:
199 **-f[no-]caret-diagnostics**
200 Print source line and ranges from source code in diagnostic.
201 This option, which defaults to on, controls whether or not Clang
202 prints the source line, source ranges, and caret when emitting a
203 diagnostic. For example, when this is enabled, Clang will print
208 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
213 **-f[no-]color-diagnostics**
214 This option, which defaults to on when a color-capable terminal is
215 detected, controls whether or not Clang prints diagnostics in color.
217 When this option is enabled, Clang will use colors to highlight
218 specific parts of the diagnostic, e.g.,
220 .. nasty hack to not lose our dignity
225 <b><span style="color:black">test.c:28:8: <span style="color:magenta">warning</span>: extra tokens at end of #endif directive [-Wextra-tokens]</span></b>
227 <span style="color:green">^</span>
228 <span style="color:green">//</span>
231 When this is disabled, Clang will just print:
235 test.c:2:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
240 **-fansi-escape-codes**
241 Controls whether ANSI escape codes are used instead of the Windows Console
242 API to output colored diagnostics. This option is only used on Windows and
245 .. option:: -fdiagnostics-format=clang/msvc/vi
247 Changes diagnostic output format to better match IDEs and command line tools.
249 This option controls the output format of the filename, line number,
250 and column printed in diagnostic messages. The options, and their
251 affect on formatting a simple conversion diagnostic, follow:
256 t.c:3:11: warning: conversion specifies type 'char *' but the argument has type 'int'
261 t.c(3,11) : warning: conversion specifies type 'char *' but the argument has type 'int'
266 t.c +3:11: warning: conversion specifies type 'char *' but the argument has type 'int'
268 .. _opt_fdiagnostics-show-option:
270 **-f[no-]diagnostics-show-option**
271 Enable ``[-Woption]`` information in diagnostic line.
273 This option, which defaults to on, controls whether or not Clang
274 prints the associated :ref:`warning group <cl_diag_warning_groups>`
275 option name when outputting a warning diagnostic. For example, in
280 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
285 Passing **-fno-diagnostics-show-option** will prevent Clang from
286 printing the [:ref:`-Wextra-tokens <opt_Wextra-tokens>`] information in
287 the diagnostic. This information tells you the flag needed to enable
288 or disable the diagnostic, either from the command line or through
289 :ref:`#pragma GCC diagnostic <pragma_GCC_diagnostic>`.
291 .. _opt_fdiagnostics-show-category:
293 .. option:: -fdiagnostics-show-category=none/id/name
295 Enable printing category information in diagnostic line.
297 This option, which defaults to "none", controls whether or not Clang
298 prints the category associated with a diagnostic when emitting it.
299 Each diagnostic may or many not have an associated category, if it
300 has one, it is listed in the diagnostic categorization field of the
301 diagnostic line (in the []'s).
303 For example, a format string warning will produce these three
304 renditions based on the setting of this option:
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,1]
310 t.c:3:11: warning: conversion specifies type 'char *' but the argument has type 'int' [-Wformat,Format String]
312 This category can be used by clients that want to group diagnostics
313 by category, so it should be a high level category. We want dozens
314 of these, not hundreds or thousands of them.
316 .. _opt_fdiagnostics-fixit-info:
318 **-f[no-]diagnostics-fixit-info**
319 Enable "FixIt" information in the diagnostics output.
321 This option, which defaults to on, controls whether or not Clang
322 prints the information on how to fix a specific diagnostic
323 underneath it when it knows. For example, in this output:
327 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
332 Passing **-fno-diagnostics-fixit-info** will prevent Clang from
333 printing the "//" line at the end of the message. This information
334 is useful for users who may not understand what is wrong, but can be
335 confusing for machine parsing.
337 .. _opt_fdiagnostics-print-source-range-info:
339 **-fdiagnostics-print-source-range-info**
340 Print machine parsable information about source ranges.
341 This option makes Clang print information about source ranges in a machine
342 parsable format after the file/line/column number information. The
343 information is a simple sequence of brace enclosed ranges, where each range
344 lists the start and end line/column locations. For example, in this output:
348 exprs.c:47:15:{47:8-47:14}{47:17-47:24}: error: invalid operands to binary expression ('int *' and '_Complex float')
349 P = (P-42) + Gamma*4;
352 The {}'s are generated by -fdiagnostics-print-source-range-info.
354 The printed column numbers count bytes from the beginning of the
355 line; take care if your source contains multibyte characters.
357 .. option:: -fdiagnostics-parseable-fixits
359 Print Fix-Its in a machine parseable form.
361 This option makes Clang print available Fix-Its in a machine
362 parseable format at the end of diagnostics. The following example
363 illustrates the format:
367 fix-it:"t.cpp":{7:25-7:29}:"Gamma"
369 The range printed is a half-open range, so in this example the
370 characters at column 25 up to but not including column 29 on line 7
371 in t.cpp should be replaced with the string "Gamma". Either the
372 range or the replacement string may be empty (representing strict
373 insertions and strict erasures, respectively). Both the file name
374 and the insertion string escape backslash (as "\\\\"), tabs (as
375 "\\t"), newlines (as "\\n"), double quotes(as "\\"") and
376 non-printable characters (as octal "\\xxx").
378 The printed column numbers count bytes from the beginning of the
379 line; take care if your source contains multibyte characters.
381 .. option:: -fno-elide-type
383 Turns off elision in template type printing.
385 The default for template type printing is to elide as many template
386 arguments as possible, removing those which are the same in both
387 template types, leaving only the differences. Adding this flag will
388 print all the template arguments. If supported by the terminal,
389 highlighting will still appear on differing arguments.
395 t.cc:4:5: note: candidate function not viable: no known conversion from 'vector<map<[...], map<float, [...]>>>' to 'vector<map<[...], map<double, [...]>>>' for 1st argument;
401 t.cc:4:5: note: candidate function not viable: no known conversion from 'vector<map<int, map<float, int>>>' to 'vector<map<int, map<double, int>>>' for 1st argument;
403 .. option:: -fdiagnostics-show-template-tree
405 Template type diffing prints a text tree.
407 For diffing large templated types, this option will cause Clang to
408 display the templates as an indented text tree, one argument per
409 line, with differences marked inline. This is compatible with
416 t.cc:4:5: note: candidate function not viable: no known conversion from 'vector<map<[...], map<float, [...]>>>' to 'vector<map<[...], map<double, [...]>>>' for 1st argument;
418 With :option:`-fdiagnostics-show-template-tree`:
422 t.cc:4:5: note: candidate function not viable: no known conversion for 1st argument;
430 .. _cl_diag_warning_groups:
432 Individual Warning Groups
433 ^^^^^^^^^^^^^^^^^^^^^^^^^
435 TODO: Generate this from tblgen. Define one anchor per warning group.
437 .. _opt_wextra-tokens:
439 .. option:: -Wextra-tokens
441 Warn about excess tokens at the end of a preprocessor directive.
443 This option, which defaults to on, enables warnings about extra
444 tokens at the end of preprocessor directives. For example:
448 test.c:28:8: warning: extra tokens at end of #endif directive [-Wextra-tokens]
452 These extra tokens are not strictly conforming, and are usually best
453 handled by commenting them out.
455 .. option:: -Wambiguous-member-template
457 Warn about unqualified uses of a member template whose name resolves to
458 another template at the location of the use.
460 This option, which defaults to on, enables a warning in the
465 template<typename T> struct set{};
466 template<typename T> struct trait { typedef const T& type; };
468 template<typename T> void set(typename trait<T>::type value) {}
475 C++ [basic.lookup.classref] requires this to be an error, but,
476 because it's hard to work around, Clang downgrades it to a warning
479 .. option:: -Wbind-to-temporary-copy
481 Warn about an unusable copy constructor when binding a reference to a
484 This option enables warnings about binding a
485 reference to a temporary when the temporary doesn't have a usable
486 copy constructor. For example:
493 NonCopyable(const NonCopyable&);
495 void foo(const NonCopyable&);
497 foo(NonCopyable()); // Disallowed in C++98; allowed in C++11.
502 struct NonCopyable2 {
504 NonCopyable2(NonCopyable2&);
506 void foo(const NonCopyable2&);
508 foo(NonCopyable2()); // Disallowed in C++98; allowed in C++11.
511 Note that if ``NonCopyable2::NonCopyable2()`` has a default argument
512 whose instantiation produces a compile error, that error will still
513 be a hard error in C++98 mode even if this warning is turned off.
515 Options to Control Clang Crash Diagnostics
516 ------------------------------------------
518 As unbelievable as it may sound, Clang does crash from time to time.
519 Generally, this only occurs to those living on the `bleeding
520 edge <http://llvm.org/releases/download.html#svn>`_. Clang goes to great
521 lengths to assist you in filing a bug report. Specifically, Clang
522 generates preprocessed source file(s) and associated run script(s) upon
523 a crash. These files should be attached to a bug report to ease
524 reproducibility of the failure. Below are the command line options to
525 control the crash diagnostics.
527 .. option:: -fno-crash-diagnostics
529 Disable auto-generation of preprocessed source files during a clang crash.
531 The -fno-crash-diagnostics flag can be helpful for speeding the process
532 of generating a delta reduced test case.
534 Options to Emit Optimization Reports
535 ------------------------------------
537 Optimization reports trace, at a high-level, all the major decisions
538 done by compiler transformations. For instance, when the inliner
539 decides to inline function ``foo()`` into ``bar()``, or the loop unroller
540 decides to unroll a loop N times, or the vectorizer decides to
541 vectorize a loop body.
543 Clang offers a family of flags which the optimizers can use to emit
544 a diagnostic in three cases:
546 1. When the pass makes a transformation (:option:`-Rpass`).
548 2. When the pass fails to make a transformation (:option:`-Rpass-missed`).
550 3. When the pass determines whether or not to make a transformation
551 (:option:`-Rpass-analysis`).
553 NOTE: Although the discussion below focuses on :option:`-Rpass`, the exact
554 same options apply to :option:`-Rpass-missed` and :option:`-Rpass-analysis`.
556 Since there are dozens of passes inside the compiler, each of these flags
557 take a regular expression that identifies the name of the pass which should
558 emit the associated diagnostic. For example, to get a report from the inliner,
559 compile the code with:
561 .. code-block:: console
563 $ clang -O2 -Rpass=inline code.cc -o code
564 code.cc:4:25: remark: foo inlined into bar [-Rpass=inline]
565 int bar(int j) { return foo(j, j - 2); }
568 Note that remarks from the inliner are identified with `[-Rpass=inline]`.
569 To request a report from every optimization pass, you should use
570 :option:`-Rpass=.*` (in fact, you can use any valid POSIX regular
571 expression). However, do not expect a report from every transformation
572 made by the compiler. Optimization remarks do not really make sense
573 outside of the major transformations (e.g., inlining, vectorization,
574 loop optimizations) and not every optimization pass supports this
580 1. Optimization remarks that refer to function names will display the
581 mangled name of the function. Since these remarks are emitted by the
582 back end of the compiler, it does not know anything about the input
583 language, nor its mangling rules.
585 2. Some source locations are not displayed correctly. The front end has
586 a more detailed source location tracking than the locations included
587 in the debug info (e.g., the front end can locate code inside macro
588 expansions). However, the locations used by :option:`-Rpass` are
589 translated from debug annotations. That translation can be lossy,
590 which results in some remarks having no location information.
593 Language and Target-Independent Features
594 ========================================
596 Controlling Errors and Warnings
597 -------------------------------
599 Clang provides a number of ways to control which code constructs cause
600 it to emit errors and warning messages, and how they are displayed to
603 Controlling How Clang Displays Diagnostics
604 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
606 When Clang emits a diagnostic, it includes rich information in the
607 output, and gives you fine-grain control over which information is
608 printed. Clang has the ability to print this information, and these are
609 the options that control it:
611 #. A file/line/column indicator that shows exactly where the diagnostic
612 occurs in your code [:ref:`-fshow-column <opt_fshow-column>`,
613 :ref:`-fshow-source-location <opt_fshow-source-location>`].
614 #. A categorization of the diagnostic as a note, warning, error, or
616 #. A text string that describes what the problem is.
617 #. An option that indicates how to control the diagnostic (for
618 diagnostics that support it)
619 [:ref:`-fdiagnostics-show-option <opt_fdiagnostics-show-option>`].
620 #. A :ref:`high-level category <diagnostics_categories>` for the diagnostic
621 for clients that want to group diagnostics by class (for diagnostics
623 [:ref:`-fdiagnostics-show-category <opt_fdiagnostics-show-category>`].
624 #. The line of source code that the issue occurs on, along with a caret
625 and ranges that indicate the important locations
626 [:ref:`-fcaret-diagnostics <opt_fcaret-diagnostics>`].
627 #. "FixIt" information, which is a concise explanation of how to fix the
628 problem (when Clang is certain it knows)
629 [:ref:`-fdiagnostics-fixit-info <opt_fdiagnostics-fixit-info>`].
630 #. A machine-parsable representation of the ranges involved (off by
632 [:ref:`-fdiagnostics-print-source-range-info <opt_fdiagnostics-print-source-range-info>`].
634 For more information please see :ref:`Formatting of
635 Diagnostics <cl_diag_formatting>`.
640 All diagnostics are mapped into one of these 6 classes:
649 .. _diagnostics_categories:
651 Diagnostic Categories
652 ^^^^^^^^^^^^^^^^^^^^^
654 Though not shown by default, diagnostics may each be associated with a
655 high-level category. This category is intended to make it possible to
656 triage builds that produce a large number of errors or warnings in a
659 Categories are not shown by default, but they can be turned on with the
660 :ref:`-fdiagnostics-show-category <opt_fdiagnostics-show-category>` option.
661 When set to "``name``", the category is printed textually in the
662 diagnostic output. When it is set to "``id``", a category number is
663 printed. The mapping of category names to category id's can be obtained
664 by running '``clang --print-diagnostic-categories``'.
666 Controlling Diagnostics via Command Line Flags
667 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
669 TODO: -W flags, -pedantic, etc
671 .. _pragma_gcc_diagnostic:
673 Controlling Diagnostics via Pragmas
674 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
676 Clang can also control what diagnostics are enabled through the use of
677 pragmas in the source code. This is useful for turning off specific
678 warnings in a section of source code. Clang supports GCC's pragma for
679 compatibility with existing source code, as well as several extensions.
681 The pragma may control any warning that can be used from the command
682 line. Warnings may be set to ignored, warning, error, or fatal. The
683 following example code will tell Clang or GCC to ignore the -Wall
688 #pragma GCC diagnostic ignored "-Wall"
690 In addition to all of the functionality provided by GCC's pragma, Clang
691 also allows you to push and pop the current warning state. This is
692 particularly useful when writing a header file that will be compiled by
693 other people, because you don't know what warning flags they build with.
695 In the below example :option:`-Wmultichar` is ignored for only a single line of
696 code, after which the diagnostics return to whatever state had previously
701 #pragma clang diagnostic push
702 #pragma clang diagnostic ignored "-Wmultichar"
704 char b = 'df'; // no warning.
706 #pragma clang diagnostic pop
708 The push and pop pragmas will save and restore the full diagnostic state
709 of the compiler, regardless of how it was set. That means that it is
710 possible to use push and pop around GCC compatible diagnostics and Clang
711 will push and pop them appropriately, while GCC will ignore the pushes
712 and pops as unknown pragmas. It should be noted that while Clang
713 supports the GCC pragma, Clang and GCC do not support the exact same set
714 of warnings, so even when using GCC compatible #pragmas there is no
715 guarantee that they will have identical behaviour on both compilers.
717 In addition to controlling warnings and errors generated by the compiler, it is
718 possible to generate custom warning and error messages through the following
723 // The following will produce warning messages
724 #pragma message "some diagnostic message"
725 #pragma GCC warning "TODO: replace deprecated feature"
727 // The following will produce an error message
728 #pragma GCC error "Not supported"
730 These pragmas operate similarly to the ``#warning`` and ``#error`` preprocessor
731 directives, except that they may also be embedded into preprocessor macros via
732 the C99 ``_Pragma`` operator, for example:
737 #define DEFER(M,...) M(__VA_ARGS__)
738 #define CUSTOM_ERROR(X) _Pragma(STR(GCC error(X " at line " DEFER(STR,__LINE__))))
740 CUSTOM_ERROR("Feature not available");
742 Controlling Diagnostics in System Headers
743 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
745 Warnings are suppressed when they occur in system headers. By default,
746 an included file is treated as a system header if it is found in an
747 include path specified by ``-isystem``, but this can be overridden in
750 The ``system_header`` pragma can be used to mark the current file as
751 being a system header. No warnings will be produced from the location of
752 the pragma onwards within the same file.
756 char a = 'xy'; // warning
758 #pragma clang system_header
760 char b = 'ab'; // no warning
762 The :option:`--system-header-prefix=` and :option:`--no-system-header-prefix=`
763 command-line arguments can be used to override whether subsets of an include
764 path are treated as system headers. When the name in a ``#include`` directive
765 is found within a header search path and starts with a system prefix, the
766 header is treated as a system header. The last prefix on the
767 command-line which matches the specified header name takes precedence.
770 .. code-block:: console
772 $ clang -Ifoo -isystem bar --system-header-prefix=x/ \
773 --no-system-header-prefix=x/y/
775 Here, ``#include "x/a.h"`` is treated as including a system header, even
776 if the header is found in ``foo``, and ``#include "x/y/b.h"`` is treated
777 as not including a system header, even if the header is found in
780 A ``#include`` directive which finds a file relative to the current
781 directory is treated as including a system header if the including file
782 is treated as a system header.
784 .. _diagnostics_enable_everything:
786 Enabling All Diagnostics
787 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
789 In addition to the traditional ``-W`` flags, one can enable **all**
790 diagnostics by passing :option:`-Weverything`. This works as expected
792 :option:`-Werror`, and also includes the warnings from :option:`-pedantic`.
794 Note that when combined with :option:`-w` (which disables all warnings), that
797 Controlling Static Analyzer Diagnostics
798 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
800 While not strictly part of the compiler, the diagnostics from Clang's
801 `static analyzer <http://clang-analyzer.llvm.org>`_ can also be
802 influenced by the user via changes to the source code. See the available
803 `annotations <http://clang-analyzer.llvm.org/annotations.html>`_ and the
805 page <http://clang-analyzer.llvm.org/faq.html#exclude_code>`_ for more
808 .. _usersmanual-precompiled-headers:
813 `Precompiled headers <http://en.wikipedia.org/wiki/Precompiled_header>`__
814 are a general approach employed by many compilers to reduce compilation
815 time. The underlying motivation of the approach is that it is common for
816 the same (and often large) header files to be included by multiple
817 source files. Consequently, compile times can often be greatly improved
818 by caching some of the (redundant) work done by a compiler to process
819 headers. Precompiled header files, which represent one of many ways to
820 implement this optimization, are literally files that represent an
821 on-disk cache that contains the vital information necessary to reduce
822 some of the work needed to process a corresponding header file. While
823 details of precompiled headers vary between compilers, precompiled
824 headers have been shown to be highly effective at speeding up program
825 compilation on systems with very large system headers (e.g., Mac OS X).
827 Generating a PCH File
828 ^^^^^^^^^^^^^^^^^^^^^
830 To generate a PCH file using Clang, one invokes Clang with the
831 :option:`-x <language>-header` option. This mirrors the interface in GCC
832 for generating PCH files:
834 .. code-block:: console
836 $ gcc -x c-header test.h -o test.h.gch
837 $ clang -x c-header test.h -o test.h.pch
842 A PCH file can then be used as a prefix header when a :option:`-include`
843 option is passed to ``clang``:
845 .. code-block:: console
847 $ clang -include test.h test.c -o test
849 The ``clang`` driver will first check if a PCH file for ``test.h`` is
850 available; if so, the contents of ``test.h`` (and the files it includes)
851 will be processed from the PCH file. Otherwise, Clang falls back to
852 directly processing the content of ``test.h``. This mirrors the behavior
857 Clang does *not* automatically use PCH files for headers that are directly
858 included within a source file. For example:
860 .. code-block:: console
862 $ clang -x c-header test.h -o test.h.pch
865 $ clang test.c -o test
867 In this example, ``clang`` will not automatically use the PCH file for
868 ``test.h`` since ``test.h`` was included directly in the source file and not
869 specified on the command line using :option:`-include`.
871 Relocatable PCH Files
872 ^^^^^^^^^^^^^^^^^^^^^
874 It is sometimes necessary to build a precompiled header from headers
875 that are not yet in their final, installed locations. For example, one
876 might build a precompiled header within the build tree that is then
877 meant to be installed alongside the headers. Clang permits the creation
878 of "relocatable" precompiled headers, which are built with a given path
879 (into the build directory) and can later be used from an installed
882 To build a relocatable precompiled header, place your headers into a
883 subdirectory whose structure mimics the installed location. For example,
884 if you want to build a precompiled header for the header ``mylib.h``
885 that will be installed into ``/usr/include``, create a subdirectory
886 ``build/usr/include`` and place the header ``mylib.h`` into that
887 subdirectory. If ``mylib.h`` depends on other headers, then they can be
888 stored within ``build/usr/include`` in a way that mimics the installed
891 Building a relocatable precompiled header requires two additional
892 arguments. First, pass the ``--relocatable-pch`` flag to indicate that
893 the resulting PCH file should be relocatable. Second, pass
894 :option:`-isysroot /path/to/build`, which makes all includes for your library
895 relative to the build directory. For example:
897 .. code-block:: console
899 # clang -x c-header --relocatable-pch -isysroot /path/to/build /path/to/build/mylib.h mylib.h.pch
901 When loading the relocatable PCH file, the various headers used in the
902 PCH file are found from the system header root. For example, ``mylib.h``
903 can be found in ``/usr/include/mylib.h``. If the headers are installed
904 in some other system root, the :option:`-isysroot` option can be used provide
905 a different system root from which the headers will be based. For
906 example, :option:`-isysroot /Developer/SDKs/MacOSX10.4u.sdk` will look for
907 ``mylib.h`` in ``/Developer/SDKs/MacOSX10.4u.sdk/usr/include/mylib.h``.
909 Relocatable precompiled headers are intended to be used in a limited
910 number of cases where the compilation environment is tightly controlled
911 and the precompiled header cannot be generated after headers have been
914 Controlling Code Generation
915 ---------------------------
917 Clang provides a number of ways to control code generation. The options
920 **-f[no-]sanitize=check1,check2,...**
921 Turn on runtime checks for various forms of undefined or suspicious
924 This option controls whether Clang adds runtime checks for various
925 forms of undefined or suspicious behavior, and is disabled by
926 default. If a check fails, a diagnostic message is produced at
927 runtime explaining the problem. The main checks are:
929 - .. _opt_fsanitize_address:
931 ``-fsanitize=address``:
932 :doc:`AddressSanitizer`, a memory error
934 - ``-fsanitize=integer``: Enables checks for undefined or
935 suspicious integer behavior.
936 - .. _opt_fsanitize_thread:
938 ``-fsanitize=thread``: :doc:`ThreadSanitizer`, a data race detector.
939 - .. _opt_fsanitize_memory:
941 ``-fsanitize=memory``: :doc:`MemorySanitizer`,
942 an *experimental* detector of uninitialized reads. Not ready for
944 - .. _opt_fsanitize_undefined:
946 ``-fsanitize=undefined``: Fast and compatible undefined behavior
947 checker. Enables the undefined behavior checks that have small
948 runtime cost and no impact on address space layout or ABI. This
949 includes all of the checks listed below other than
950 ``unsigned-integer-overflow``.
952 - ``-fsanitize=undefined-trap``: This includes all sanitizers
953 included by ``-fsanitize=undefined``, except those that require
954 runtime support. This group of sanitizers is intended to be
955 used in conjunction with the ``-fsanitize-undefined-trap-on-error``
956 flag. This includes all of the checks listed below other than
957 ``unsigned-integer-overflow`` and ``vptr``.
958 - ``-fsanitize=dataflow``: :doc:`DataFlowSanitizer`, a general data
960 - ``-fsanitize=cfi``: :doc:`control flow integrity <ControlFlowIntegrity>`
961 checks. Implies ``-flto``.
963 The following more fine-grained checks are also available:
965 - ``-fsanitize=alignment``: Use of a misaligned pointer or creation
966 of a misaligned reference.
967 - ``-fsanitize=bool``: Load of a ``bool`` value which is neither
968 ``true`` nor ``false``.
969 - ``-fsanitize=bounds``: Out of bounds array indexing, in cases
970 where the array bound can be statically determined.
971 - ``-fsanitize=cfi-cast-strict``: Enables :ref:`strict cast checks
973 - ``-fsanitize=cfi-derived-cast``: Base-to-derived cast to the wrong
974 dynamic type. Implies ``-flto``.
975 - ``-fsanitize=cfi-unrelated-cast``: Cast from ``void*`` or another
976 unrelated type to the wrong dynamic type. Implies ``-flto``.
977 - ``-fsanitize=cfi-vptr``: Use of an object whose vptr is of the
978 wrong dynamic type. Implies ``-flto``.
979 - ``-fsanitize=enum``: Load of a value of an enumerated type which
980 is not in the range of representable values for that enumerated
982 - ``-fsanitize=float-cast-overflow``: Conversion to, from, or
983 between floating-point types which would overflow the
985 - ``-fsanitize=float-divide-by-zero``: Floating point division by
987 - ``-fsanitize=function``: Indirect call of a function through a
988 function pointer of the wrong type (Linux, C++ and x86/x86_64 only).
989 - ``-fsanitize=integer-divide-by-zero``: Integer division by zero.
990 - ``-fsanitize=nonnull-attribute``: Passing null pointer as a function
991 parameter which is declared to never be null.
992 - ``-fsanitize=null``: Use of a null pointer or creation of a null
994 - ``-fsanitize=object-size``: An attempt to use bytes which the
995 optimizer can determine are not part of the object being
996 accessed. The sizes of objects are determined using
997 ``__builtin_object_size``, and consequently may be able to detect
998 more problems at higher optimization levels.
999 - ``-fsanitize=return``: In C++, reaching the end of a
1000 value-returning function without returning a value.
1001 - ``-fsanitize=returns-nonnull-attribute``: Returning null pointer
1002 from a function which is declared to never return null.
1003 - ``-fsanitize=shift``: Shift operators where the amount shifted is
1004 greater or equal to the promoted bit-width of the left hand side
1005 or less than zero, or where the left hand side is negative. For a
1006 signed left shift, also checks for signed overflow in C, and for
1007 unsigned overflow in C++. You can use ``-fsanitize=shift-base`` or
1008 ``-fsanitize=shift-exponent`` to check only left-hand side or
1009 right-hand side of shift operation, respectively.
1010 - ``-fsanitize=signed-integer-overflow``: Signed integer overflow,
1011 including all the checks added by ``-ftrapv``, and checking for
1012 overflow in signed division (``INT_MIN / -1``).
1013 - ``-fsanitize=unreachable``: If control flow reaches
1014 ``__builtin_unreachable``.
1015 - ``-fsanitize=unsigned-integer-overflow``: Unsigned integer
1017 - ``-fsanitize=vla-bound``: A variable-length array whose bound
1018 does not evaluate to a positive value.
1019 - ``-fsanitize=vptr``: Use of an object whose vptr indicates that
1020 it is of the wrong dynamic type, or that its lifetime has not
1021 begun or has ended. Incompatible with ``-fno-rtti``.
1023 You can turn off or modify checks for certain source files, functions
1024 or even variables by providing a special file:
1026 - ``-fsanitize-blacklist=/path/to/blacklist/file``: disable or modify
1027 sanitizer checks for objects listed in the file. See
1028 :doc:`SanitizerSpecialCaseList` for file format description.
1029 - ``-fno-sanitize-blacklist``: don't use blacklist file, if it was
1030 specified earlier in the command line.
1032 Extra features of MemorySanitizer (require explicit
1033 ``-fsanitize=memory``):
1035 - ``-fsanitize-memory-track-origins[=level]``: Enables origin tracking in
1036 MemorySanitizer. Adds a second section to MemorySanitizer
1037 reports pointing to the heap or stack allocation the
1038 uninitialized bits came from. Slows down execution by additional
1041 Possible values for level are 0 (off), 1, 2 (default). Level 2
1042 adds more sections to MemorySanitizer reports describing the
1043 order of memory stores the uninitialized value went
1044 through. This mode may use extra memory in programs that copy
1045 uninitialized memory a lot.
1047 Extra features of UndefinedBehaviorSanitizer:
1049 - ``-fsanitize-undefined-trap-on-error``: Causes traps to be emitted
1050 rather than calls to runtime libraries when a problem is detected.
1051 This option is intended for use in cases where the sanitizer runtime
1052 cannot be used (for instance, when building libc or a kernel module).
1053 This is only compatible with the sanitizers in the ``undefined-trap``
1056 The ``-fsanitize=`` argument must also be provided when linking, in
1057 order to link to the appropriate runtime library. When using
1058 ``-fsanitize=vptr`` (or a group that includes it, such as
1059 ``-fsanitize=undefined``) with a C++ program, the link must be
1060 performed by ``clang++``, not ``clang``, in order to link against the
1061 C++-specific parts of the runtime library.
1063 It is not possible to combine more than one of the ``-fsanitize=address``,
1064 ``-fsanitize=thread``, and ``-fsanitize=memory`` checkers in the same
1065 program. The ``-fsanitize=undefined`` checks can only be combined with
1066 ``-fsanitize=address``.
1068 **-f[no-]sanitize-recover=check1,check2,...**
1070 Controls which checks enabled by ``-fsanitize=`` flag are non-fatal.
1071 If the check is fatal, program will halt after the first error
1072 of this kind is detected and error report is printed.
1074 By default, non-fatal checks are those enabled by UndefinedBehaviorSanitizer,
1075 except for ``-fsanitize=return`` and ``-fsanitize=unreachable``. Some
1076 sanitizers (e.g. :doc:`AddressSanitizer`) may not support recovery,
1077 and always crash the program after the issue is detected.
1079 .. option:: -fno-assume-sane-operator-new
1081 Don't assume that the C++'s new operator is sane.
1083 This option tells the compiler to do not assume that C++'s global
1084 new operator will always return a pointer that does not alias any
1085 other pointer when the function returns.
1087 .. option:: -ftrap-function=[name]
1089 Instruct code generator to emit a function call to the specified
1090 function name for ``__builtin_trap()``.
1092 LLVM code generator translates ``__builtin_trap()`` to a trap
1093 instruction if it is supported by the target ISA. Otherwise, the
1094 builtin is translated into a call to ``abort``. If this option is
1095 set, then the code generator will always lower the builtin to a call
1096 to the specified function regardless of whether the target ISA has a
1097 trap instruction. This option is useful for environments (e.g.
1098 deeply embedded) where a trap cannot be properly handled, or when
1099 some custom behavior is desired.
1101 .. option:: -ftls-model=[model]
1103 Select which TLS model to use.
1105 Valid values are: ``global-dynamic``, ``local-dynamic``,
1106 ``initial-exec`` and ``local-exec``. The default value is
1107 ``global-dynamic``. The compiler may use a different model if the
1108 selected model is not supported by the target, or if a more
1109 efficient model can be used. The TLS model can be overridden per
1110 variable using the ``tls_model`` attribute.
1112 .. option:: -mhwdiv=[values]
1114 Select the ARM modes (arm or thumb) that support hardware division
1117 Valid values are: ``arm``, ``thumb`` and ``arm,thumb``.
1118 This option is used to indicate which mode (arm or thumb) supports
1119 hardware division instructions. This only applies to the ARM
1122 .. option:: -m[no-]crc
1124 Enable or disable CRC instructions.
1126 This option is used to indicate whether CRC instructions are to
1127 be generated. This only applies to the ARM architecture.
1129 CRC instructions are enabled by default on ARMv8.
1131 .. option:: -mgeneral-regs-only
1133 Generate code which only uses the general purpose registers.
1135 This option restricts the generated code to use general registers
1136 only. This only applies to the AArch64 architecture.
1138 **-f[no-]max-unknown-pointer-align=[number]**
1139 Instruct the code generator to not enforce a higher alignment than the given
1140 number (of bytes) when accessing memory via an opaque pointer or reference.
1141 This cap is ignored when directly accessing a variable or when the pointee
1142 type has an explicit “aligned” attribute.
1144 The value should usually be determined by the properties of the system allocator.
1145 Some builtin types, especially vector types, have very high natural alignments;
1146 when working with values of those types, Clang usually wants to use instructions
1147 that take advantage of that alignment. However, many system allocators do
1148 not promise to return memory that is more than 8-byte or 16-byte-aligned. Use
1149 this option to limit the alignment that the compiler can assume for an arbitrary
1150 pointer, which may point onto the heap.
1152 This option does not affect the ABI alignment of types; the layout of structs and
1153 unions and the value returned by the alignof operator remain the same.
1155 This option can be overridden on a case-by-case basis by putting an explicit
1156 “aligned” alignment on a struct, union, or typedef. For example:
1158 .. code-block:: console
1160 #include <immintrin.h>
1161 // Make an aligned typedef of the AVX-512 16-int vector type.
1162 typedef __v16si __aligned_v16si __attribute__((aligned(64)));
1164 void initialize_vector(__aligned_v16si *v) {
1165 // The compiler may assume that ‘v’ is 64-byte aligned, regardless of the
1166 // value of -fmax-unknown-pointer-align.
1170 Profile Guided Optimization
1171 ---------------------------
1173 Profile information enables better optimization. For example, knowing that a
1174 branch is taken very frequently helps the compiler make better decisions when
1175 ordering basic blocks. Knowing that a function ``foo`` is called more
1176 frequently than another function ``bar`` helps the inliner.
1178 Clang supports profile guided optimization with two different kinds of
1179 profiling. A sampling profiler can generate a profile with very low runtime
1180 overhead, or you can build an instrumented version of the code that collects
1181 more detailed profile information. Both kinds of profiles can provide execution
1182 counts for instructions in the code and information on branches taken and
1183 function invocation.
1185 Regardless of which kind of profiling you use, be careful to collect profiles
1186 by running your code with inputs that are representative of the typical
1187 behavior. Code that is not exercised in the profile will be optimized as if it
1188 is unimportant, and the compiler may make poor optimization choices for code
1189 that is disproportionately used while profiling.
1191 Using Sampling Profilers
1192 ^^^^^^^^^^^^^^^^^^^^^^^^
1194 Sampling profilers are used to collect runtime information, such as
1195 hardware counters, while your application executes. They are typically
1196 very efficient and do not incur a large runtime overhead. The
1197 sample data collected by the profiler can be used during compilation
1198 to determine what the most executed areas of the code are.
1200 Using the data from a sample profiler requires some changes in the way
1201 a program is built. Before the compiler can use profiling information,
1202 the code needs to execute under the profiler. The following is the
1203 usual build cycle when using sample profilers for optimization:
1205 1. Build the code with source line table information. You can use all the
1206 usual build flags that you always build your application with. The only
1207 requirement is that you add ``-gline-tables-only`` or ``-g`` to the
1208 command line. This is important for the profiler to be able to map
1209 instructions back to source line locations.
1211 .. code-block:: console
1213 $ clang++ -O2 -gline-tables-only code.cc -o code
1215 2. Run the executable under a sampling profiler. The specific profiler
1216 you use does not really matter, as long as its output can be converted
1217 into the format that the LLVM optimizer understands. Currently, there
1218 exists a conversion tool for the Linux Perf profiler
1219 (https://perf.wiki.kernel.org/), so these examples assume that you
1220 are using Linux Perf to profile your code.
1222 .. code-block:: console
1224 $ perf record -b ./code
1226 Note the use of the ``-b`` flag. This tells Perf to use the Last Branch
1227 Record (LBR) to record call chains. While this is not strictly required,
1228 it provides better call information, which improves the accuracy of
1231 3. Convert the collected profile data to LLVM's sample profile format.
1232 This is currently supported via the AutoFDO converter ``create_llvm_prof``.
1233 It is available at http://github.com/google/autofdo. Once built and
1234 installed, you can convert the ``perf.data`` file to LLVM using
1237 .. code-block:: console
1239 $ create_llvm_prof --binary=./code --out=code.prof
1241 This will read ``perf.data`` and the binary file ``./code`` and emit
1242 the profile data in ``code.prof``. Note that if you ran ``perf``
1243 without the ``-b`` flag, you need to use ``--use_lbr=false`` when
1244 calling ``create_llvm_prof``.
1246 4. Build the code again using the collected profile. This step feeds
1247 the profile back to the optimizers. This should result in a binary
1248 that executes faster than the original one. Note that you are not
1249 required to build the code with the exact same arguments that you
1250 used in the first step. The only requirement is that you build the code
1251 with ``-gline-tables-only`` and ``-fprofile-sample-use``.
1253 .. code-block:: console
1255 $ clang++ -O2 -gline-tables-only -fprofile-sample-use=code.prof code.cc -o code
1258 Sample Profile Format
1259 """""""""""""""""""""
1261 If you are not using Linux Perf to collect profiles, you will need to
1262 write a conversion tool from your profiler to LLVM's format. This section
1263 explains the file format expected by the backend.
1265 Sample profiles are written as ASCII text. The file is divided into sections,
1266 which correspond to each of the functions executed at runtime. Each
1267 section has the following format (taken from
1268 https://github.com/google/autofdo/blob/master/profile_writer.h):
1270 .. code-block:: console
1272 function1:total_samples:total_head_samples
1273 offset1[.discriminator]: number_of_samples [fn1:num fn2:num ... ]
1274 offset2[.discriminator]: number_of_samples [fn3:num fn4:num ... ]
1276 offsetN[.discriminator]: number_of_samples [fn5:num fn6:num ... ]
1278 The file may contain blank lines between sections and within a
1279 section. However, the spacing within a single line is fixed. Additional
1280 spaces will result in an error while reading the file.
1282 Function names must be mangled in order for the profile loader to
1283 match them in the current translation unit. The two numbers in the
1284 function header specify how many total samples were accumulated in the
1285 function (first number), and the total number of samples accumulated
1286 in the prologue of the function (second number). This head sample
1287 count provides an indicator of how frequently the function is invoked.
1289 Each sampled line may contain several items. Some are optional (marked
1292 a. Source line offset. This number represents the line number
1293 in the function where the sample was collected. The line number is
1294 always relative to the line where symbol of the function is
1295 defined. So, if the function has its header at line 280, the offset
1296 13 is at line 293 in the file.
1298 Note that this offset should never be a negative number. This could
1299 happen in cases like macros. The debug machinery will register the
1300 line number at the point of macro expansion. So, if the macro was
1301 expanded in a line before the start of the function, the profile
1302 converter should emit a 0 as the offset (this means that the optimizers
1303 will not be able to associate a meaningful weight to the instructions
1306 b. [OPTIONAL] Discriminator. This is used if the sampled program
1307 was compiled with DWARF discriminator support
1308 (http://wiki.dwarfstd.org/index.php?title=Path_Discriminators).
1309 DWARF discriminators are unsigned integer values that allow the
1310 compiler to distinguish between multiple execution paths on the
1311 same source line location.
1313 For example, consider the line of code ``if (cond) foo(); else bar();``.
1314 If the predicate ``cond`` is true 80% of the time, then the edge
1315 into function ``foo`` should be considered to be taken most of the
1316 time. But both calls to ``foo`` and ``bar`` are at the same source
1317 line, so a sample count at that line is not sufficient. The
1318 compiler needs to know which part of that line is taken more
1321 This is what discriminators provide. In this case, the calls to
1322 ``foo`` and ``bar`` will be at the same line, but will have
1323 different discriminator values. This allows the compiler to correctly
1324 set edge weights into ``foo`` and ``bar``.
1326 c. Number of samples. This is an integer quantity representing the
1327 number of samples collected by the profiler at this source
1330 d. [OPTIONAL] Potential call targets and samples. If present, this
1331 line contains a call instruction. This models both direct and
1332 number of samples. For example,
1334 .. code-block:: console
1336 130: 7 foo:3 bar:2 baz:7
1338 The above means that at relative line offset 130 there is a call
1339 instruction that calls one of ``foo()``, ``bar()`` and ``baz()``,
1340 with ``baz()`` being the relatively more frequently called target.
1343 Profiling with Instrumentation
1344 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1346 Clang also supports profiling via instrumentation. This requires building a
1347 special instrumented version of the code and has some runtime
1348 overhead during the profiling, but it provides more detailed results than a
1349 sampling profiler. It also provides reproducible results, at least to the
1350 extent that the code behaves consistently across runs.
1352 Here are the steps for using profile guided optimization with
1355 1. Build an instrumented version of the code by compiling and linking with the
1356 ``-fprofile-instr-generate`` option.
1358 .. code-block:: console
1360 $ clang++ -O2 -fprofile-instr-generate code.cc -o code
1362 2. Run the instrumented executable with inputs that reflect the typical usage.
1363 By default, the profile data will be written to a ``default.profraw`` file
1364 in the current directory. You can override that default by setting the
1365 ``LLVM_PROFILE_FILE`` environment variable to specify an alternate file.
1366 Any instance of ``%p`` in that file name will be replaced by the process
1367 ID, so that you can easily distinguish the profile output from multiple
1370 .. code-block:: console
1372 $ LLVM_PROFILE_FILE="code-%p.profraw" ./code
1374 3. Combine profiles from multiple runs and convert the "raw" profile format to
1375 the input expected by clang. Use the ``merge`` command of the llvm-profdata
1378 .. code-block:: console
1380 $ llvm-profdata merge -output=code.profdata code-*.profraw
1382 Note that this step is necessary even when there is only one "raw" profile,
1383 since the merge operation also changes the file format.
1385 4. Build the code again using the ``-fprofile-instr-use`` option to specify the
1386 collected profile data.
1388 .. code-block:: console
1390 $ clang++ -O2 -fprofile-instr-use=code.profdata code.cc -o code
1392 You can repeat step 4 as often as you like without regenerating the
1393 profile. As you make changes to your code, clang may no longer be able to
1394 use the profile data. It will warn you when this happens.
1397 Controlling Size of Debug Information
1398 -------------------------------------
1400 Debug info kind generated by Clang can be set by one of the flags listed
1401 below. If multiple flags are present, the last one is used.
1405 Don't generate any debug info (default).
1407 .. option:: -gline-tables-only
1409 Generate line number tables only.
1411 This kind of debug info allows to obtain stack traces with function names,
1412 file names and line numbers (by such tools as ``gdb`` or ``addr2line``). It
1413 doesn't contain any other data (e.g. description of local variables or
1414 function parameters).
1416 .. option:: -fstandalone-debug
1418 Clang supports a number of optimizations to reduce the size of debug
1419 information in the binary. They work based on the assumption that
1420 the debug type information can be spread out over multiple
1421 compilation units. For instance, Clang will not emit type
1422 definitions for types that are not needed by a module and could be
1423 replaced with a forward declaration. Further, Clang will only emit
1424 type info for a dynamic C++ class in the module that contains the
1425 vtable for the class.
1427 The **-fstandalone-debug** option turns off these optimizations.
1428 This is useful when working with 3rd-party libraries that don't come
1429 with debug information. Note that Clang will never emit type
1430 information for types that are not referenced at all by the program.
1432 .. option:: -fno-standalone-debug
1434 On Darwin **-fstandalone-debug** is enabled by default. The
1435 **-fno-standalone-debug** option can be used to get to turn on the
1436 vtable-based optimization described above.
1440 Generate complete debug info.
1442 Comment Parsing Options
1443 -----------------------
1445 Clang parses Doxygen and non-Doxygen style documentation comments and attaches
1446 them to the appropriate declaration nodes. By default, it only parses
1447 Doxygen-style comments and ignores ordinary comments starting with ``//`` and
1450 .. option:: -Wdocumentation
1452 Emit warnings about use of documentation comments. This warning group is off
1455 This includes checking that ``\param`` commands name parameters that actually
1456 present in the function signature, checking that ``\returns`` is used only on
1457 functions that actually return a value etc.
1459 .. option:: -Wno-documentation-unknown-command
1461 Don't warn when encountering an unknown Doxygen command.
1463 .. option:: -fparse-all-comments
1465 Parse all comments as documentation comments (including ordinary comments
1466 starting with ``//`` and ``/*``).
1468 .. option:: -fcomment-block-commands=[commands]
1470 Define custom documentation commands as block commands. This allows Clang to
1471 construct the correct AST for these custom commands, and silences warnings
1472 about unknown commands. Several commands must be separated by a comma
1473 *without trailing space*; e.g. ``-fcomment-block-commands=foo,bar`` defines
1474 custom commands ``\foo`` and ``\bar``.
1476 It is also possible to use ``-fcomment-block-commands`` several times; e.g.
1477 ``-fcomment-block-commands=foo -fcomment-block-commands=bar`` does the same
1485 The support for standard C in clang is feature-complete except for the
1486 C99 floating-point pragmas.
1488 Extensions supported by clang
1489 -----------------------------
1491 See :doc:`LanguageExtensions`.
1493 Differences between various standard modes
1494 ------------------------------------------
1496 clang supports the -std option, which changes what language mode clang
1497 uses. The supported modes for C are c89, gnu89, c94, c99, gnu99, c11,
1498 gnu11, and various aliases for those modes. If no -std option is
1499 specified, clang defaults to gnu11 mode. Many C99 and C11 features are
1500 supported in earlier modes as a conforming extension, with a warning. Use
1501 ``-pedantic-errors`` to request an error if a feature from a later standard
1502 revision is used in an earlier mode.
1504 Differences between all ``c*`` and ``gnu*`` modes:
1506 - ``c*`` modes define "``__STRICT_ANSI__``".
1507 - Target-specific defines not prefixed by underscores, like "linux",
1508 are defined in ``gnu*`` modes.
1509 - Trigraphs default to being off in ``gnu*`` modes; they can be enabled by
1510 the -trigraphs option.
1511 - The parser recognizes "asm" and "typeof" as keywords in ``gnu*`` modes;
1512 the variants "``__asm__``" and "``__typeof__``" are recognized in all
1514 - The Apple "blocks" extension is recognized by default in ``gnu*`` modes
1515 on some platforms; it can be enabled in any mode with the "-fblocks"
1517 - Arrays that are VLA's according to the standard, but which can be
1518 constant folded by the frontend are treated as fixed size arrays.
1519 This occurs for things like "int X[(1, 2)];", which is technically a
1520 VLA. ``c*`` modes are strictly compliant and treat these as VLAs.
1522 Differences between ``*89`` and ``*99`` modes:
1524 - The ``*99`` modes default to implementing "inline" as specified in C99,
1525 while the ``*89`` modes implement the GNU version. This can be
1526 overridden for individual functions with the ``__gnu_inline__``
1528 - Digraphs are not recognized in c89 mode.
1529 - The scope of names defined inside a "for", "if", "switch", "while",
1530 or "do" statement is different. (example: "``if ((struct x {int
1532 - ``__STDC_VERSION__`` is not defined in ``*89`` modes.
1533 - "inline" is not recognized as a keyword in c89 mode.
1534 - "restrict" is not recognized as a keyword in ``*89`` modes.
1535 - Commas are allowed in integer constant expressions in ``*99`` modes.
1536 - Arrays which are not lvalues are not implicitly promoted to pointers
1538 - Some warnings are different.
1540 Differences between ``*99`` and ``*11`` modes:
1542 - Warnings for use of C11 features are disabled.
1543 - ``__STDC_VERSION__`` is defined to ``201112L`` rather than ``199901L``.
1545 c94 mode is identical to c89 mode except that digraphs are enabled in
1546 c94 mode (FIXME: And ``__STDC_VERSION__`` should be defined!).
1548 GCC extensions not implemented yet
1549 ----------------------------------
1551 clang tries to be compatible with gcc as much as possible, but some gcc
1552 extensions are not implemented yet:
1554 - clang does not support #pragma weak (`bug
1555 3679 <http://llvm.org/bugs/show_bug.cgi?id=3679>`_). Due to the uses
1556 described in the bug, this is likely to be implemented at some point,
1558 - clang does not support decimal floating point types (``_Decimal32`` and
1559 friends) or fixed-point types (``_Fract`` and friends); nobody has
1560 expressed interest in these features yet, so it's hard to say when
1561 they will be implemented.
1562 - clang does not support nested functions; this is a complex feature
1563 which is infrequently used, so it is unlikely to be implemented
1564 anytime soon. In C++11 it can be emulated by assigning lambda
1565 functions to local variables, e.g:
1569 auto const local_function = [&](int parameter) {
1575 - clang does not support global register variables; this is unlikely to
1576 be implemented soon because it requires additional LLVM backend
1578 - clang does not support static initialization of flexible array
1579 members. This appears to be a rarely used extension, but could be
1580 implemented pending user demand.
1581 - clang does not support
1582 ``__builtin_va_arg_pack``/``__builtin_va_arg_pack_len``. This is
1583 used rarely, but in some potentially interesting places, like the
1584 glibc headers, so it may be implemented pending user demand. Note
1585 that because clang pretends to be like GCC 4.2, and this extension
1586 was introduced in 4.3, the glibc headers will not try to use this
1587 extension with clang at the moment.
1588 - clang does not support the gcc extension for forward-declaring
1589 function parameters; this has not shown up in any real-world code
1590 yet, though, so it might never be implemented.
1592 This is not a complete list; if you find an unsupported extension
1593 missing from this list, please send an e-mail to cfe-dev. This list
1594 currently excludes C++; see :ref:`C++ Language Features <cxx>`. Also, this
1595 list does not include bugs in mostly-implemented features; please see
1597 tracker <http://llvm.org/bugs/buglist.cgi?quicksearch=product%3Aclang+component%3A-New%2BBugs%2CAST%2CBasic%2CDriver%2CHeaders%2CLLVM%2BCodeGen%2Cparser%2Cpreprocessor%2CSemantic%2BAnalyzer>`_
1598 for known existing bugs (FIXME: Is there a section for bug-reporting
1599 guidelines somewhere?).
1601 Intentionally unsupported GCC extensions
1602 ----------------------------------------
1604 - clang does not support the gcc extension that allows variable-length
1605 arrays in structures. This is for a few reasons: one, it is tricky to
1606 implement, two, the extension is completely undocumented, and three,
1607 the extension appears to be rarely used. Note that clang *does*
1608 support flexible array members (arrays with a zero or unspecified
1609 size at the end of a structure).
1610 - clang does not have an equivalent to gcc's "fold"; this means that
1611 clang doesn't accept some constructs gcc might accept in contexts
1612 where a constant expression is required, like "x-x" where x is a
1614 - clang does not support ``__builtin_apply`` and friends; this extension
1615 is extremely obscure and difficult to implement reliably.
1619 Microsoft extensions
1620 --------------------
1622 clang has some experimental support for extensions from Microsoft Visual
1623 C++; to enable it, use the ``-fms-extensions`` command-line option. This is
1624 the default for Windows targets. Note that the support is incomplete.
1625 Some constructs such as ``dllexport`` on classes are ignored with a warning,
1626 and others such as `Microsoft IDL annotations
1627 <http://msdn.microsoft.com/en-us/library/8tesw2eh.aspx>`_ are silently
1630 clang has a ``-fms-compatibility`` flag that makes clang accept enough
1631 invalid C++ to be able to parse most Microsoft headers. For example, it
1632 allows `unqualified lookup of dependent base class members
1633 <http://clang.llvm.org/compatibility.html#dep_lookup_bases>`_, which is
1634 a common compatibility issue with clang. This flag is enabled by default
1635 for Windows targets.
1637 ``-fdelayed-template-parsing`` lets clang delay parsing of function template
1638 definitions until the end of a translation unit. This flag is enabled by
1639 default for Windows targets.
1641 - clang allows setting ``_MSC_VER`` with ``-fmsc-version=``. It defaults to
1642 1700 which is the same as Visual C/C++ 2012. Any number is supported
1643 and can greatly affect what Windows SDK and c++stdlib headers clang
1645 - clang does not support the Microsoft extension where anonymous record
1646 members can be declared using user defined typedefs.
1647 - clang supports the Microsoft ``#pragma pack`` feature for controlling
1648 record layout. GCC also contains support for this feature, however
1649 where MSVC and GCC are incompatible clang follows the MSVC
1651 - clang supports the Microsoft ``#pragma comment(lib, "foo.lib")`` feature for
1652 automatically linking against the specified library. Currently this feature
1653 only works with the Visual C++ linker.
1654 - clang supports the Microsoft ``#pragma comment(linker, "/flag:foo")`` feature
1655 for adding linker flags to COFF object files. The user is responsible for
1656 ensuring that the linker understands the flags.
1657 - clang defaults to C++11 for Windows targets.
1661 C++ Language Features
1662 =====================
1664 clang fully implements all of standard C++98 except for exported
1665 templates (which were removed in C++11), and all of standard C++11
1666 and the current draft standard for C++1y.
1668 Controlling implementation limits
1669 ---------------------------------
1671 .. option:: -fbracket-depth=N
1673 Sets the limit for nested parentheses, brackets, and braces to N. The
1676 .. option:: -fconstexpr-depth=N
1678 Sets the limit for recursive constexpr function invocations to N. The
1681 .. option:: -ftemplate-depth=N
1683 Sets the limit for recursively nested template instantiations to N. The
1686 .. option:: -foperator-arrow-depth=N
1688 Sets the limit for iterative calls to 'operator->' functions to N. The
1693 Objective-C Language Features
1694 =============================
1698 Objective-C++ Language Features
1699 ===============================
1702 .. _target_features:
1704 Target-Specific Features and Limitations
1705 ========================================
1707 CPU Architectures Features and Limitations
1708 ------------------------------------------
1713 The support for X86 (both 32-bit and 64-bit) is considered stable on
1714 Darwin (Mac OS X), Linux, FreeBSD, and Dragonfly BSD: it has been tested
1715 to correctly compile many large C, C++, Objective-C, and Objective-C++
1718 On ``x86_64-mingw32``, passing i128(by value) is incompatible with the
1719 Microsoft x64 calling convention. You might need to tweak
1720 ``WinX86_64ABIInfo::classify()`` in lib/CodeGen/TargetInfo.cpp.
1722 For the X86 target, clang supports the :option:`-m16` command line
1723 argument which enables 16-bit code output. This is broadly similar to
1724 using ``asm(".code16gcc")`` with the GNU toolchain. The generated code
1725 and the ABI remains 32-bit but the assembler emits instructions
1726 appropriate for a CPU running in 16-bit mode, with address-size and
1727 operand-size prefixes to enable 32-bit addressing and operations.
1732 The support for ARM (specifically ARMv6 and ARMv7) is considered stable
1733 on Darwin (iOS): it has been tested to correctly compile many large C,
1734 C++, Objective-C, and Objective-C++ codebases. Clang only supports a
1735 limited number of ARM architectures. It does not yet fully support
1741 The support for PowerPC (especially PowerPC64) is considered stable
1742 on Linux and FreeBSD: it has been tested to correctly compile many
1743 large C and C++ codebases. PowerPC (32bit) is still missing certain
1744 features (e.g. PIC code on ELF platforms).
1749 clang currently contains some support for other architectures (e.g. Sparc);
1750 however, significant pieces of code generation are still missing, and they
1751 haven't undergone significant testing.
1753 clang contains limited support for the MSP430 embedded processor, but
1754 both the clang support and the LLVM backend support are highly
1757 Other platforms are completely unsupported at the moment. Adding the
1758 minimal support needed for parsing and semantic analysis on a new
1759 platform is quite easy; see ``lib/Basic/Targets.cpp`` in the clang source
1760 tree. This level of support is also sufficient for conversion to LLVM IR
1761 for simple programs. Proper support for conversion to LLVM IR requires
1762 adding code to ``lib/CodeGen/CGCall.cpp`` at the moment; this is likely to
1763 change soon, though. Generating assembly requires a suitable LLVM
1766 Operating System Features and Limitations
1767 -----------------------------------------
1772 Thread Sanitizer is not supported.
1777 Clang has experimental support for targeting "Cygming" (Cygwin / MinGW)
1780 See also :ref:`Microsoft Extensions <c_ms>`.
1785 Clang works on Cygwin-1.7.
1790 Clang works on some mingw32 distributions. Clang assumes directories as
1793 - ``C:/mingw/include``
1795 - ``C:/mingw/lib/gcc/mingw32/4.[3-5].0/include/c++``
1797 On MSYS, a few tests might fail.
1802 For 32-bit (i686-w64-mingw32), and 64-bit (x86\_64-w64-mingw32), Clang
1805 - ``GCC versions 4.5.0 to 4.5.3, 4.6.0 to 4.6.2, or 4.7.0 (for the C++ header search path)``
1806 - ``some_directory/bin/gcc.exe``
1807 - ``some_directory/bin/clang.exe``
1808 - ``some_directory/bin/clang++.exe``
1809 - ``some_directory/bin/../include/c++/GCC_version``
1810 - ``some_directory/bin/../include/c++/GCC_version/x86_64-w64-mingw32``
1811 - ``some_directory/bin/../include/c++/GCC_version/i686-w64-mingw32``
1812 - ``some_directory/bin/../include/c++/GCC_version/backward``
1813 - ``some_directory/bin/../x86_64-w64-mingw32/include``
1814 - ``some_directory/bin/../i686-w64-mingw32/include``
1815 - ``some_directory/bin/../include``
1817 This directory layout is standard for any toolchain you will find on the
1818 official `MinGW-w64 website <http://mingw-w64.sourceforge.net>`_.
1820 Clang expects the GCC executable "gcc.exe" compiled for
1821 ``i686-w64-mingw32`` (or ``x86_64-w64-mingw32``) to be present on PATH.
1823 `Some tests might fail <http://llvm.org/bugs/show_bug.cgi?id=9072>`_ on
1824 ``x86_64-w64-mingw32``.
1831 clang-cl is an alternative command-line interface to Clang driver, designed for
1832 compatibility with the Visual C++ compiler, cl.exe.
1834 To enable clang-cl to find system headers, libraries, and the linker when run
1835 from the command-line, it should be executed inside a Visual Studio Native Tools
1836 Command Prompt or a regular Command Prompt where the environment has been set
1837 up using e.g. `vcvars32.bat <http://msdn.microsoft.com/en-us/library/f2ccy3wt.aspx>`_.
1839 clang-cl can also be used from inside Visual Studio by using an LLVM Platform
1842 Command-Line Options
1843 --------------------
1845 To be compatible with cl.exe, clang-cl supports most of the same command-line
1846 options. Those options can start with either ``/`` or ``-``. It also supports
1847 some of Clang's core options, such as the ``-W`` options.
1849 Options that are known to clang-cl, but not currently supported, are ignored
1850 with a warning. For example:
1854 clang-cl.exe: warning: argument unused during compilation: '/Zi'
1856 To suppress warnings about unused arguments, use the ``-Qunused-arguments`` option.
1858 Options that are not known to clang-cl will cause errors. If they are spelled with a
1859 leading ``/``, they will be mistaken for a filename:
1863 clang-cl.exe: error: no such file or directory: '/foobar'
1865 Please `file a bug <http://llvm.org/bugs/enter_bug.cgi?product=clang&component=Driver>`_
1866 for any valid cl.exe flags that clang-cl does not understand.
1868 Execute ``clang-cl /?`` to see a list of supported options:
1872 CL.EXE COMPATIBILITY OPTIONS:
1873 /? Display available options
1874 /arch:<value> Set architecture for code generation
1875 /C Don't discard comments when preprocessing
1877 /D <macro[=value]> Define macro
1878 /EH<value> Exception handling model
1879 /EP Disable linemarker output and preprocess to stdout
1880 /E Preprocess to stdout
1881 /fallback Fall back to cl.exe if clang-cl fails to compile
1882 /FA Output assembly code file during compilation
1883 /Fa<file or directory> Output assembly code to this file during compilation
1884 /Fe<file or directory> Set output executable file or directory (ends in / or \)
1885 /FI <value> Include file before parsing
1886 /Fi<file> Set preprocess output file name
1887 /Fo<file or directory> Set output object file, or directory (ends in / or \)
1888 /GF- Disable string pooling
1889 /GR- Disable emission of RTTI data
1890 /GR Enable emission of RTTI data
1891 /Gw- Don't put each data item in its own section
1892 /Gw Put each data item in its own section
1893 /Gy- Don't put each function in its own section
1894 /Gy Put each function in its own section
1895 /help Display available options
1896 /I <dir> Add directory to include search path
1897 /J Make char type unsigned
1898 /LDd Create debug DLL
1900 /link <options> Forward options to the linker
1901 /MDd Use DLL debug run-time
1902 /MD Use DLL run-time
1903 /MTd Use static debug run-time
1904 /MT Use static run-time
1905 /Ob0 Disable inlining
1906 /Od Disable optimization
1907 /Oi- Disable use of builtin functions
1908 /Oi Enable use of builtin functions
1909 /Os Optimize for size
1910 /Ot Optimize for speed
1911 /Ox Maximum optimization
1912 /Oy- Disable frame pointer omission
1913 /Oy Enable frame pointer omission
1914 /O<n> Optimization level
1915 /P Preprocess to file
1916 /showIncludes Print info about included files to stderr
1917 /TC Treat all source files as C
1918 /Tc <filename> Specify a C source file
1919 /TP Treat all source files as C++
1920 /Tp <filename> Specify a C++ source file
1921 /U <macro> Undefine macro
1922 /vd<value> Control vtordisp placement
1923 /vmb Use a best-case representation method for member pointers
1924 /vmg Use a most-general representation for member pointers
1925 /vmm Set the default most-general representation to multiple inheritance
1926 /vms Set the default most-general representation to single inheritance
1927 /vmv Set the default most-general representation to virtual inheritance
1928 /W0 Disable all warnings
1934 /WX- Do not treat warnings as errors
1935 /WX Treat warnings as errors
1936 /w Disable all warnings
1937 /Zi Enable debug information
1938 /Zp Set the default maximum struct packing alignment to 1
1939 /Zp<value> Specify the default maximum struct packing alignment
1940 /Zs Syntax-check only
1943 -### Print (but do not run) the commands to run for this compilation
1944 -fms-compatibility-version=<value>
1945 Dot-separated value representing the Microsoft compiler version
1946 number to report in _MSC_VER (0 = don't define it (default))
1947 -fmsc-version=<value> Microsoft compiler version number to report in _MSC_VER (0 = don't
1948 define it (default))
1949 -fsanitize-blacklist=<value>
1950 Path to blacklist file for sanitizers
1951 -fsanitize=<check> Enable runtime instrumentation for bug detection: address (memory
1952 errors) | thread (race detection) | undefined (miscellaneous
1954 -mllvm <value> Additional arguments to forward to LLVM's option processing
1955 -Qunused-arguments Don't emit warning for unused driver arguments
1956 --target=<value> Generate code for the given target
1957 -v Show commands to run and use verbose output
1958 -W<warning> Enable the specified warning
1959 -Xclang <arg> Pass <arg> to the clang compiler
1961 The /fallback Option
1962 ^^^^^^^^^^^^^^^^^^^^
1964 When clang-cl is run with the ``/fallback`` option, it will first try to
1965 compile files itself. For any file that it fails to compile, it will fall back
1966 and try to compile the file by invoking cl.exe.
1968 This option is intended to be used as a temporary means to build projects where
1969 clang-cl cannot successfully compile all the files. clang-cl may fail to compile
1970 a file either because it cannot generate code for some C++ feature, or because
1971 it cannot parse some Microsoft language extension.