1<html lang="en"> 2<head> 3<title>C++ Dialect Options - Using the GNU Compiler Collection (GCC)</title> 4<meta http-equiv="Content-Type" content="text/html"> 5<meta name="description" content="Using the GNU Compiler Collection (GCC)"> 6<meta name="generator" content="makeinfo 4.13"> 7<link title="Top" rel="start" href="index.html#Top"> 8<link rel="up" href="Invoking-GCC.html#Invoking-GCC" title="Invoking GCC"> 9<link rel="prev" href="C-Dialect-Options.html#C-Dialect-Options" title="C Dialect Options"> 10<link rel="next" href="Objective_002dC-and-Objective_002dC_002b_002b-Dialect-Options.html#Objective_002dC-and-Objective_002dC_002b_002b-Dialect-Options" title="Objective-C and Objective-C++ Dialect Options"> 11<link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage"> 12<!-- 13Copyright (C) 1988-2013 Free Software Foundation, Inc. 14 15Permission is granted to copy, distribute and/or modify this document 16under the terms of the GNU Free Documentation License, Version 1.3 or 17any later version published by the Free Software Foundation; with the 18Invariant Sections being ``Funding Free Software'', the Front-Cover 19Texts being (a) (see below), and with the Back-Cover Texts being (b) 20(see below). A copy of the license is included in the section entitled 21``GNU Free Documentation License''. 22 23(a) The FSF's Front-Cover Text is: 24 25 A GNU Manual 26 27(b) The FSF's Back-Cover Text is: 28 29 You have freedom to copy and modify this GNU Manual, like GNU 30 software. Copies published by the Free Software Foundation raise 31 funds for GNU development.--> 32<meta http-equiv="Content-Style-Type" content="text/css"> 33<style type="text/css"><!-- 34 pre.display { font-family:inherit } 35 pre.format { font-family:inherit } 36 pre.smalldisplay { font-family:inherit; font-size:smaller } 37 pre.smallformat { font-family:inherit; font-size:smaller } 38 pre.smallexample { font-size:smaller } 39 pre.smalllisp { font-size:smaller } 40 span.sc { font-variant:small-caps } 41 span.roman { font-family:serif; font-weight:normal; } 42 span.sansserif { font-family:sans-serif; font-weight:normal; } 43--></style> 44<link rel="stylesheet" type="text/css" href="../cs.css"> 45</head> 46<body> 47<div class="node"> 48<a name="C++-Dialect-Options"></a> 49<a name="C_002b_002b-Dialect-Options"></a> 50<p> 51Next: <a rel="next" accesskey="n" href="Objective_002dC-and-Objective_002dC_002b_002b-Dialect-Options.html#Objective_002dC-and-Objective_002dC_002b_002b-Dialect-Options">Objective-C and Objective-C++ Dialect Options</a>, 52Previous: <a rel="previous" accesskey="p" href="C-Dialect-Options.html#C-Dialect-Options">C Dialect Options</a>, 53Up: <a rel="up" accesskey="u" href="Invoking-GCC.html#Invoking-GCC">Invoking GCC</a> 54<hr> 55</div> 56 57<h3 class="section">3.5 Options Controlling C++ Dialect</h3> 58 59<p><a name="index-compiler-options_002c-C_002b_002b-130"></a><a name="index-C_002b_002b-options_002c-command_002dline-131"></a><a name="index-options_002c-C_002b_002b-132"></a>This section describes the command-line options that are only meaningful 60for C++ programs. You can also use most of the GNU compiler options 61regardless of what language your program is in. For example, you 62might compile a file <code>firstClass.C</code> like this: 63 64<pre class="smallexample"> g++ -g -frepo -O -c firstClass.C 65</pre> 66 <p class="noindent">In this example, only <samp><span class="option">-frepo</span></samp> is an option meant 67only for C++ programs; you can use the other options with any 68language supported by GCC. 69 70 <p>Here is a list of options that are <em>only</em> for compiling C++ programs: 71 72 <dl> 73<dt><code>-fabi-version=</code><var>n</var><dd><a name="index-fabi_002dversion-133"></a>Use version <var>n</var> of the C++ ABI. The default is version 2. 74 75 <p>Version 0 refers to the version conforming most closely to 76the C++ ABI specification. Therefore, the ABI obtained using version 0 77will change in different versions of G++ as ABI bugs are fixed. 78 79 <p>Version 1 is the version of the C++ ABI that first appeared in G++ 3.2. 80 81 <p>Version 2 is the version of the C++ ABI that first appeared in G++ 3.4. 82 83 <p>Version 3 corrects an error in mangling a constant address as a 84template argument. 85 86 <p>Version 4, which first appeared in G++ 4.5, implements a standard 87mangling for vector types. 88 89 <p>Version 5, which first appeared in G++ 4.6, corrects the mangling of 90attribute const/volatile on function pointer types, decltype of a 91plain decl, and use of a function parameter in the declaration of 92another parameter. 93 94 <p>Version 6, which first appeared in G++ 4.7, corrects the promotion 95behavior of C++11 scoped enums and the mangling of template argument 96packs, const/static_cast, prefix ++ and –, and a class scope function 97used as a template argument. 98 99 <p>See also <samp><span class="option">-Wabi</span></samp>. 100 101 <br><dt><code>-fno-access-control</code><dd><a name="index-fno_002daccess_002dcontrol-134"></a>Turn off all access checking. This switch is mainly useful for working 102around bugs in the access control code. 103 104 <br><dt><code>-fcheck-new</code><dd><a name="index-fcheck_002dnew-135"></a>Check that the pointer returned by <code>operator new</code> is non-null 105before attempting to modify the storage allocated. This check is 106normally unnecessary because the C++ standard specifies that 107<code>operator new</code> only returns <code>0</code> if it is declared 108‘<samp><span class="samp">throw()</span></samp>’, in which case the compiler always checks the 109return value even without this option. In all other cases, when 110<code>operator new</code> has a non-empty exception specification, memory 111exhaustion is signalled by throwing <code>std::bad_alloc</code>. See also 112‘<samp><span class="samp">new (nothrow)</span></samp>’. 113 114 <br><dt><code>-fconstexpr-depth=</code><var>n</var><dd><a name="index-fconstexpr_002ddepth-136"></a>Set the maximum nested evaluation depth for C++11 constexpr functions 115to <var>n</var>. A limit is needed to detect endless recursion during 116constant expression evaluation. The minimum specified by the standard 117is 512. 118 119 <br><dt><code>-fdeduce-init-list</code><dd><a name="index-fdeduce_002dinit_002dlist-137"></a>Enable deduction of a template type parameter as 120<code>std::initializer_list</code> from a brace-enclosed initializer list, i.e. 121 122 <pre class="smallexample"> template <class T> auto forward(T t) -> decltype (realfn (t)) 123 { 124 return realfn (t); 125 } 126 127 void f() 128 { 129 forward({1,2}); // call forward<std::initializer_list<int>> 130 } 131</pre> 132 <p>This deduction was implemented as a possible extension to the 133originally proposed semantics for the C++11 standard, but was not part 134of the final standard, so it is disabled by default. This option is 135deprecated, and may be removed in a future version of G++. 136 137 <br><dt><code>-ffriend-injection</code><dd><a name="index-ffriend_002dinjection-138"></a>Inject friend functions into the enclosing namespace, so that they are 138visible outside the scope of the class in which they are declared. 139Friend functions were documented to work this way in the old Annotated 140C++ Reference Manual, and versions of G++ before 4.1 always worked 141that way. However, in ISO C++ a friend function that is not declared 142in an enclosing scope can only be found using argument dependent 143lookup. This option causes friends to be injected as they were in 144earlier releases. 145 146 <p>This option is for compatibility, and may be removed in a future 147release of G++. 148 149 <br><dt><code>-fno-elide-constructors</code><dd><a name="index-fno_002delide_002dconstructors-139"></a>The C++ standard allows an implementation to omit creating a temporary 150that is only used to initialize another object of the same type. 151Specifying this option disables that optimization, and forces G++ to 152call the copy constructor in all cases. 153 154 <br><dt><code>-fno-enforce-eh-specs</code><dd><a name="index-fno_002denforce_002deh_002dspecs-140"></a>Don't generate code to check for violation of exception specifications 155at run time. This option violates the C++ standard, but may be useful 156for reducing code size in production builds, much like defining 157‘<samp><span class="samp">NDEBUG</span></samp>’. This does not give user code permission to throw 158exceptions in violation of the exception specifications; the compiler 159still optimizes based on the specifications, so throwing an 160unexpected exception results in undefined behavior at run time. 161 162 <br><dt><code>-fextern-tls-init</code><dt><code>-fno-extern-tls-init</code><dd><a name="index-fextern_002dtls_002dinit-141"></a><a name="index-fno_002dextern_002dtls_002dinit-142"></a>The C++11 and OpenMP standards allow ‘<samp><span class="samp">thread_local</span></samp>’ and 163‘<samp><span class="samp">threadprivate</span></samp>’ variables to have dynamic (runtime) 164initialization. To support this, any use of such a variable goes 165through a wrapper function that performs any necessary initialization. 166When the use and definition of the variable are in the same 167translation unit, this overhead can be optimized away, but when the 168use is in a different translation unit there is significant overhead 169even if the variable doesn't actually need dynamic initialization. If 170the programmer can be sure that no use of the variable in a 171non-defining TU needs to trigger dynamic initialization (either 172because the variable is statically initialized, or a use of the 173variable in the defining TU will be executed before any uses in 174another TU), they can avoid this overhead with the 175<samp><span class="option">-fno-extern-tls-init</span></samp> option. 176 177 <p>On targets that support symbol aliases, the default is 178<samp><span class="option">-fextern-tls-init</span></samp>. On targets that do not support symbol 179aliases, the default is <samp><span class="option">-fno-extern-tls-init</span></samp>. 180 181 <br><dt><code>-ffor-scope</code><dt><code>-fno-for-scope</code><dd><a name="index-ffor_002dscope-143"></a><a name="index-fno_002dfor_002dscope-144"></a>If <samp><span class="option">-ffor-scope</span></samp> is specified, the scope of variables declared in 182a <i>for-init-statement</i> is limited to the ‘<samp><span class="samp">for</span></samp>’ loop itself, 183as specified by the C++ standard. 184If <samp><span class="option">-fno-for-scope</span></samp> is specified, the scope of variables declared in 185a <i>for-init-statement</i> extends to the end of the enclosing scope, 186as was the case in old versions of G++, and other (traditional) 187implementations of C++. 188 189 <p>If neither flag is given, the default is to follow the standard, 190but to allow and give a warning for old-style code that would 191otherwise be invalid, or have different behavior. 192 193 <br><dt><code>-fno-gnu-keywords</code><dd><a name="index-fno_002dgnu_002dkeywords-145"></a>Do not recognize <code>typeof</code> as a keyword, so that code can use this 194word as an identifier. You can use the keyword <code>__typeof__</code> instead. 195<samp><span class="option">-ansi</span></samp> implies <samp><span class="option">-fno-gnu-keywords</span></samp>. 196 197 <br><dt><code>-fno-implicit-templates</code><dd><a name="index-fno_002dimplicit_002dtemplates-146"></a>Never emit code for non-inline templates that are instantiated 198implicitly (i.e. by use); only emit code for explicit instantiations. 199See <a href="Template-Instantiation.html#Template-Instantiation">Template Instantiation</a>, for more information. 200 201 <br><dt><code>-fno-implicit-inline-templates</code><dd><a name="index-fno_002dimplicit_002dinline_002dtemplates-147"></a>Don't emit code for implicit instantiations of inline templates, either. 202The default is to handle inlines differently so that compiles with and 203without optimization need the same set of explicit instantiations. 204 205 <br><dt><code>-fno-implement-inlines</code><dd><a name="index-fno_002dimplement_002dinlines-148"></a>To save space, do not emit out-of-line copies of inline functions 206controlled by ‘<samp><span class="samp">#pragma implementation</span></samp>’. This causes linker 207errors if these functions are not inlined everywhere they are called. 208 209 <br><dt><code>-fms-extensions</code><dd><a name="index-fms_002dextensions-149"></a>Disable Wpedantic warnings about constructs used in MFC, such as implicit 210int and getting a pointer to member function via non-standard syntax. 211 212 <br><dt><code>-fno-nonansi-builtins</code><dd><a name="index-fno_002dnonansi_002dbuiltins-150"></a>Disable built-in declarations of functions that are not mandated by 213ANSI/ISO C. These include <code>ffs</code>, <code>alloca</code>, <code>_exit</code>, 214<code>index</code>, <code>bzero</code>, <code>conjf</code>, and other related functions. 215 216 <br><dt><code>-fnothrow-opt</code><dd><a name="index-fnothrow_002dopt-151"></a>Treat a <code>throw()</code> exception specification as if it were a 217<code>noexcept</code> specification to reduce or eliminate the text size 218overhead relative to a function with no exception specification. If 219the function has local variables of types with non-trivial 220destructors, the exception specification actually makes the 221function smaller because the EH cleanups for those variables can be 222optimized away. The semantic effect is that an exception thrown out of 223a function with such an exception specification results in a call 224to <code>terminate</code> rather than <code>unexpected</code>. 225 226 <br><dt><code>-fno-operator-names</code><dd><a name="index-fno_002doperator_002dnames-152"></a>Do not treat the operator name keywords <code>and</code>, <code>bitand</code>, 227<code>bitor</code>, <code>compl</code>, <code>not</code>, <code>or</code> and <code>xor</code> as 228synonyms as keywords. 229 230 <br><dt><code>-fno-optional-diags</code><dd><a name="index-fno_002doptional_002ddiags-153"></a>Disable diagnostics that the standard says a compiler does not need to 231issue. Currently, the only such diagnostic issued by G++ is the one for 232a name having multiple meanings within a class. 233 234 <br><dt><code>-fpermissive</code><dd><a name="index-fpermissive-154"></a>Downgrade some diagnostics about nonconformant code from errors to 235warnings. Thus, using <samp><span class="option">-fpermissive</span></samp> allows some 236nonconforming code to compile. 237 238 <br><dt><code>-fno-pretty-templates</code><dd><a name="index-fno_002dpretty_002dtemplates-155"></a>When an error message refers to a specialization of a function 239template, the compiler normally prints the signature of the 240template followed by the template arguments and any typedefs or 241typenames in the signature (e.g. <code>void f(T) [with T = int]</code> 242rather than <code>void f(int)</code>) so that it's clear which template is 243involved. When an error message refers to a specialization of a class 244template, the compiler omits any template arguments that match 245the default template arguments for that template. If either of these 246behaviors make it harder to understand the error message rather than 247easier, you can use <samp><span class="option">-fno-pretty-templates</span></samp> to disable them. 248 249 <br><dt><code>-frepo</code><dd><a name="index-frepo-156"></a>Enable automatic template instantiation at link time. This option also 250implies <samp><span class="option">-fno-implicit-templates</span></samp>. See <a href="Template-Instantiation.html#Template-Instantiation">Template Instantiation</a>, for more information. 251 252 <br><dt><code>-fno-rtti</code><dd><a name="index-fno_002drtti-157"></a>Disable generation of information about every class with virtual 253functions for use by the C++ run-time type identification features 254(‘<samp><span class="samp">dynamic_cast</span></samp>’ and ‘<samp><span class="samp">typeid</span></samp>’). If you don't use those parts 255of the language, you can save some space by using this flag. Note that 256exception handling uses the same information, but G++ generates it as 257needed. The ‘<samp><span class="samp">dynamic_cast</span></samp>’ operator can still be used for casts that 258do not require run-time type information, i.e. casts to <code>void *</code> or to 259unambiguous base classes. 260 261 <br><dt><code>-fstats</code><dd><a name="index-fstats-158"></a>Emit statistics about front-end processing at the end of the compilation. 262This information is generally only useful to the G++ development team. 263 264 <br><dt><code>-fstrict-enums</code><dd><a name="index-fstrict_002denums-159"></a>Allow the compiler to optimize using the assumption that a value of 265enumerated type can only be one of the values of the enumeration (as 266defined in the C++ standard; basically, a value that can be 267represented in the minimum number of bits needed to represent all the 268enumerators). This assumption may not be valid if the program uses a 269cast to convert an arbitrary integer value to the enumerated type. 270 271 <br><dt><code>-ftemplate-backtrace-limit=</code><var>n</var><dd><a name="index-ftemplate_002dbacktrace_002dlimit-160"></a>Set the maximum number of template instantiation notes for a single 272warning or error to <var>n</var>. The default value is 10. 273 274 <br><dt><code>-ftemplate-depth=</code><var>n</var><dd><a name="index-ftemplate_002ddepth-161"></a>Set the maximum instantiation depth for template classes to <var>n</var>. 275A limit on the template instantiation depth is needed to detect 276endless recursions during template class instantiation. ANSI/ISO C++ 277conforming programs must not rely on a maximum depth greater than 17 278(changed to 1024 in C++11). The default value is 900, as the compiler 279can run out of stack space before hitting 1024 in some situations. 280 281 <br><dt><code>-fno-threadsafe-statics</code><dd><a name="index-fno_002dthreadsafe_002dstatics-162"></a>Do not emit the extra code to use the routines specified in the C++ 282ABI for thread-safe initialization of local statics. You can use this 283option to reduce code size slightly in code that doesn't need to be 284thread-safe. 285 286 <br><dt><code>-fuse-cxa-atexit</code><dd><a name="index-fuse_002dcxa_002datexit-163"></a>Register destructors for objects with static storage duration with the 287<code>__cxa_atexit</code> function rather than the <code>atexit</code> function. 288This option is required for fully standards-compliant handling of static 289destructors, but only works if your C library supports 290<code>__cxa_atexit</code>. 291 292 <br><dt><code>-fno-use-cxa-get-exception-ptr</code><dd><a name="index-fno_002duse_002dcxa_002dget_002dexception_002dptr-164"></a>Don't use the <code>__cxa_get_exception_ptr</code> runtime routine. This 293causes <code>std::uncaught_exception</code> to be incorrect, but is necessary 294if the runtime routine is not available. 295 296 <br><dt><code>-fvisibility-inlines-hidden</code><dd><a name="index-fvisibility_002dinlines_002dhidden-165"></a>This switch declares that the user does not attempt to compare 297pointers to inline functions or methods where the addresses of the two functions 298are taken in different shared objects. 299 300 <p>The effect of this is that GCC may, effectively, mark inline methods with 301<code>__attribute__ ((visibility ("hidden")))</code> so that they do not 302appear in the export table of a DSO and do not require a PLT indirection 303when used within the DSO. Enabling this option can have a dramatic effect 304on load and link times of a DSO as it massively reduces the size of the 305dynamic export table when the library makes heavy use of templates. 306 307 <p>The behavior of this switch is not quite the same as marking the 308methods as hidden directly, because it does not affect static variables 309local to the function or cause the compiler to deduce that 310the function is defined in only one shared object. 311 312 <p>You may mark a method as having a visibility explicitly to negate the 313effect of the switch for that method. For example, if you do want to 314compare pointers to a particular inline method, you might mark it as 315having default visibility. Marking the enclosing class with explicit 316visibility has no effect. 317 318 <p>Explicitly instantiated inline methods are unaffected by this option 319as their linkage might otherwise cross a shared library boundary. 320See <a href="Template-Instantiation.html#Template-Instantiation">Template Instantiation</a>. 321 322 <br><dt><code>-fvisibility-ms-compat</code><dd><a name="index-fvisibility_002dms_002dcompat-166"></a>This flag attempts to use visibility settings to make GCC's C++ 323linkage model compatible with that of Microsoft Visual Studio. 324 325 <p>The flag makes these changes to GCC's linkage model: 326 327 <ol type=1 start=1> 328<li>It sets the default visibility to <code>hidden</code>, like 329<samp><span class="option">-fvisibility=hidden</span></samp>. 330 331 <li>Types, but not their members, are not hidden by default. 332 333 <li>The One Definition Rule is relaxed for types without explicit 334visibility specifications that are defined in more than one 335shared object: those declarations are permitted if they are 336permitted when this option is not used. 337 </ol> 338 339 <p>In new code it is better to use <samp><span class="option">-fvisibility=hidden</span></samp> and 340export those classes that are intended to be externally visible. 341Unfortunately it is possible for code to rely, perhaps accidentally, 342on the Visual Studio behavior. 343 344 <p>Among the consequences of these changes are that static data members 345of the same type with the same name but defined in different shared 346objects are different, so changing one does not change the other; 347and that pointers to function members defined in different shared 348objects may not compare equal. When this flag is given, it is a 349violation of the ODR to define types with the same name differently. 350 351 <br><dt><code>-fno-weak</code><dd><a name="index-fno_002dweak-167"></a>Do not use weak symbol support, even if it is provided by the linker. 352By default, G++ uses weak symbols if they are available. This 353option exists only for testing, and should not be used by end-users; 354it results in inferior code and has no benefits. This option may 355be removed in a future release of G++. 356 357 <br><dt><code>-nostdinc++</code><dd><a name="index-nostdinc_002b_002b-168"></a>Do not search for header files in the standard directories specific to 358C++, but do still search the other standard directories. (This option 359is used when building the C++ library.) 360</dl> 361 362 <p>In addition, these optimization, warning, and code generation options 363have meanings only for C++ programs: 364 365 <dl> 366<dt><code>-fno-default-inline</code><dd><a name="index-fno_002ddefault_002dinline-169"></a>Do not assume ‘<samp><span class="samp">inline</span></samp>’ for functions defined inside a class scope. 367See <a href="Optimize-Options.html#Optimize-Options">Options That Control Optimization</a>. Note that these 368functions have linkage like inline functions; they just aren't 369inlined by default. 370 371 <br><dt><code>-Wabi </code><span class="roman">(C, Objective-C, C++ and Objective-C++ only)</span><dd><a name="index-Wabi-170"></a><a name="index-Wno_002dabi-171"></a>Warn when G++ generates code that is probably not compatible with the 372vendor-neutral C++ ABI. Although an effort has been made to warn about 373all such cases, there are probably some cases that are not warned about, 374even though G++ is generating incompatible code. There may also be 375cases where warnings are emitted even though the code that is generated 376is compatible. 377 378 <p>You should rewrite your code to avoid these warnings if you are 379concerned about the fact that code generated by G++ may not be binary 380compatible with code generated by other compilers. 381 382 <p>The known incompatibilities in <samp><span class="option">-fabi-version=2</span></samp> (the default) include: 383 384 <ul> 385<li>A template with a non-type template parameter of reference type is 386mangled incorrectly: 387 <pre class="smallexample"> extern int N; 388 template <int &> struct S {}; 389 void n (S<N>) {2} 390</pre> 391 <p>This is fixed in <samp><span class="option">-fabi-version=3</span></samp>. 392 393 <li>SIMD vector types declared using <code>__attribute ((vector_size))</code> are 394mangled in a non-standard way that does not allow for overloading of 395functions taking vectors of different sizes. 396 397 <p>The mangling is changed in <samp><span class="option">-fabi-version=4</span></samp>. 398</ul> 399 400 <p>The known incompatibilities in <samp><span class="option">-fabi-version=1</span></samp> include: 401 402 <ul> 403<li>Incorrect handling of tail-padding for bit-fields. G++ may attempt to 404pack data into the same byte as a base class. For example: 405 406 <pre class="smallexample"> struct A { virtual void f(); int f1 : 1; }; 407 struct B : public A { int f2 : 1; }; 408</pre> 409 <p class="noindent">In this case, G++ places <code>B::f2</code> into the same byte 410as <code>A::f1</code>; other compilers do not. You can avoid this problem 411by explicitly padding <code>A</code> so that its size is a multiple of the 412byte size on your platform; that causes G++ and other compilers to 413lay out <code>B</code> identically. 414 415 <li>Incorrect handling of tail-padding for virtual bases. G++ does not use 416tail padding when laying out virtual bases. For example: 417 418 <pre class="smallexample"> struct A { virtual void f(); char c1; }; 419 struct B { B(); char c2; }; 420 struct C : public A, public virtual B {}; 421</pre> 422 <p class="noindent">In this case, G++ does not place <code>B</code> into the tail-padding for 423<code>A</code>; other compilers do. You can avoid this problem by 424explicitly padding <code>A</code> so that its size is a multiple of its 425alignment (ignoring virtual base classes); that causes G++ and other 426compilers to lay out <code>C</code> identically. 427 428 <li>Incorrect handling of bit-fields with declared widths greater than that 429of their underlying types, when the bit-fields appear in a union. For 430example: 431 432 <pre class="smallexample"> union U { int i : 4096; }; 433</pre> 434 <p class="noindent">Assuming that an <code>int</code> does not have 4096 bits, G++ makes the 435union too small by the number of bits in an <code>int</code>. 436 437 <li>Empty classes can be placed at incorrect offsets. For example: 438 439 <pre class="smallexample"> struct A {}; 440 441 struct B { 442 A a; 443 virtual void f (); 444 }; 445 446 struct C : public B, public A {}; 447</pre> 448 <p class="noindent">G++ places the <code>A</code> base class of <code>C</code> at a nonzero offset; 449it should be placed at offset zero. G++ mistakenly believes that the 450<code>A</code> data member of <code>B</code> is already at offset zero. 451 452 <li>Names of template functions whose types involve <code>typename</code> or 453template template parameters can be mangled incorrectly. 454 455 <pre class="smallexample"> template <typename Q> 456 void f(typename Q::X) {} 457 458 template <template <typename> class Q> 459 void f(typename Q<int>::X) {} 460</pre> 461 <p class="noindent">Instantiations of these templates may be mangled incorrectly. 462 463 </ul> 464 465 <p>It also warns about psABI-related changes. The known psABI changes at this 466point include: 467 468 <ul> 469<li>For SysV/x86-64, unions with <code>long double</code> members are 470passed in memory as specified in psABI. For example: 471 472 <pre class="smallexample"> union U { 473 long double ld; 474 int i; 475 }; 476</pre> 477 <p class="noindent"><code>union U</code> is always passed in memory. 478 479 </ul> 480 481 <br><dt><code>-Wctor-dtor-privacy </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wctor_002ddtor_002dprivacy-172"></a><a name="index-Wno_002dctor_002ddtor_002dprivacy-173"></a>Warn when a class seems unusable because all the constructors or 482destructors in that class are private, and it has neither friends nor 483public static member functions. Also warn if there are no non-private 484methods, and there's at least one private member function that isn't 485a constructor or destructor. 486 487 <br><dt><code>-Wdelete-non-virtual-dtor </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wdelete_002dnon_002dvirtual_002ddtor-174"></a><a name="index-Wno_002ddelete_002dnon_002dvirtual_002ddtor-175"></a>Warn when ‘<samp><span class="samp">delete</span></samp>’ is used to destroy an instance of a class that 488has virtual functions and non-virtual destructor. It is unsafe to delete 489an instance of a derived class through a pointer to a base class if the 490base class does not have a virtual destructor. This warning is enabled 491by <samp><span class="option">-Wall</span></samp>. 492 493 <br><dt><code>-Wliteral-suffix </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wliteral_002dsuffix-176"></a><a name="index-Wno_002dliteral_002dsuffix-177"></a>Warn when a string or character literal is followed by a ud-suffix which does 494not begin with an underscore. As a conforming extension, GCC treats such 495suffixes as separate preprocessing tokens in order to maintain backwards 496compatibility with code that uses formatting macros from <code><inttypes.h></code>. 497For example: 498 499 <pre class="smallexample"> #define __STDC_FORMAT_MACROS 500 #include <inttypes.h> 501 #include <stdio.h> 502 503 int main() { 504 int64_t i64 = 123; 505 printf("My int64: %"PRId64"\n", i64); 506 } 507</pre> 508 <p>In this case, <code>PRId64</code> is treated as a separate preprocessing token. 509 510 <p>This warning is enabled by default. 511 512 <br><dt><code>-Wnarrowing </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wnarrowing-178"></a><a name="index-Wno_002dnarrowing-179"></a>Warn when a narrowing conversion prohibited by C++11 occurs within 513‘<samp><span class="samp">{ }</span></samp>’, e.g. 514 515 <pre class="smallexample"> int i = { 2.2 }; // error: narrowing from double to int 516</pre> 517 <p>This flag is included in <samp><span class="option">-Wall</span></samp> and <samp><span class="option">-Wc++11-compat</span></samp>. 518 519 <p>With <samp><span class="option">-std=c++11</span></samp>, <samp><span class="option">-Wno-narrowing</span></samp> suppresses the diagnostic 520required by the standard. Note that this does not affect the meaning 521of well-formed code; narrowing conversions are still considered 522ill-formed in SFINAE context. 523 524 <br><dt><code>-Wnoexcept </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wnoexcept-180"></a><a name="index-Wno_002dnoexcept-181"></a>Warn when a noexcept-expression evaluates to false because of a call 525to a function that does not have a non-throwing exception 526specification (i.e. ‘<samp><span class="samp">throw()</span></samp>’ or ‘<samp><span class="samp">noexcept</span></samp>’) but is known by 527the compiler to never throw an exception. 528 529 <br><dt><code>-Wnon-virtual-dtor </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wnon_002dvirtual_002ddtor-182"></a><a name="index-Wno_002dnon_002dvirtual_002ddtor-183"></a>Warn when a class has virtual functions and an accessible non-virtual 530destructor, in which case it is possible but unsafe to delete 531an instance of a derived class through a pointer to the base class. 532This warning is also enabled if <samp><span class="option">-Weffc++</span></samp> is specified. 533 534 <br><dt><code>-Wreorder </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wreorder-184"></a><a name="index-Wno_002dreorder-185"></a><a name="index-reordering_002c-warning-186"></a><a name="index-warning-for-reordering-of-member-initializers-187"></a>Warn when the order of member initializers given in the code does not 535match the order in which they must be executed. For instance: 536 537 <pre class="smallexample"> struct A { 538 int i; 539 int j; 540 A(): j (0), i (1) { } 541 }; 542</pre> 543 <p class="noindent">The compiler rearranges the member initializers for ‘<samp><span class="samp">i</span></samp>’ 544and ‘<samp><span class="samp">j</span></samp>’ to match the declaration order of the members, emitting 545a warning to that effect. This warning is enabled by <samp><span class="option">-Wall</span></samp>. 546 547 <br><dt><code>-fext-numeric-literals </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-fext_002dnumeric_002dliterals-188"></a><a name="index-fno_002dext_002dnumeric_002dliterals-189"></a>Accept imaginary, fixed-point, or machine-defined 548literal number suffixes as GNU extensions. 549When this option is turned off these suffixes are treated 550as C++11 user-defined literal numeric suffixes. 551This is on by default for all pre-C++11 dialects and all GNU dialects: 552<samp><span class="option">-std=c++98</span></samp>, <samp><span class="option">-std=gnu++98</span></samp>, <samp><span class="option">-std=gnu++11</span></samp>, 553<samp><span class="option">-std=gnu++1y</span></samp>. 554This option is off by default 555for ISO C++11 onwards (<samp><span class="option">-std=c++11</span></samp>, ...). 556</dl> 557 558 <p>The following <samp><span class="option">-W...</span></samp> options are not affected by <samp><span class="option">-Wall</span></samp>. 559 560 <dl> 561<dt><code>-Weffc++ </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Weffc_002b_002b-190"></a><a name="index-Wno_002deffc_002b_002b-191"></a>Warn about violations of the following style guidelines from Scott Meyers' 562<cite>Effective C++, Second Edition</cite> book: 563 564 <ul> 565<li>Item 11: Define a copy constructor and an assignment operator for classes 566with dynamically-allocated memory. 567 568 <li>Item 12: Prefer initialization to assignment in constructors. 569 570 <li>Item 14: Make destructors virtual in base classes. 571 572 <li>Item 15: Have <code>operator=</code> return a reference to <code>*this</code>. 573 574 <li>Item 23: Don't try to return a reference when you must return an object. 575 576 </ul> 577 578 <p>Also warn about violations of the following style guidelines from 579Scott Meyers' <cite>More Effective C++</cite> book: 580 581 <ul> 582<li>Item 6: Distinguish between prefix and postfix forms of increment and 583decrement operators. 584 585 <li>Item 7: Never overload <code>&&</code>, <code>||</code>, or <code>,</code>. 586 587 </ul> 588 589 <p>When selecting this option, be aware that the standard library 590headers do not obey all of these guidelines; use ‘<samp><span class="samp">grep -v</span></samp>’ 591to filter out those warnings. 592 593 <br><dt><code>-Wstrict-null-sentinel </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wstrict_002dnull_002dsentinel-192"></a><a name="index-Wno_002dstrict_002dnull_002dsentinel-193"></a>Warn about the use of an uncasted <code>NULL</code> as sentinel. When 594compiling only with GCC this is a valid sentinel, as <code>NULL</code> is defined 595to <code>__null</code>. Although it is a null pointer constant rather than a 596null pointer, it is guaranteed to be of the same size as a pointer. 597But this use is not portable across different compilers. 598 599 <br><dt><code>-Wno-non-template-friend </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wno_002dnon_002dtemplate_002dfriend-194"></a><a name="index-Wnon_002dtemplate_002dfriend-195"></a>Disable warnings when non-templatized friend functions are declared 600within a template. Since the advent of explicit template specification 601support in G++, if the name of the friend is an unqualified-id (i.e., 602‘<samp><span class="samp">friend foo(int)</span></samp>’), the C++ language specification demands that the 603friend declare or define an ordinary, nontemplate function. (Section 60414.5.3). Before G++ implemented explicit specification, unqualified-ids 605could be interpreted as a particular specialization of a templatized 606function. Because this non-conforming behavior is no longer the default 607behavior for G++, <samp><span class="option">-Wnon-template-friend</span></samp> allows the compiler to 608check existing code for potential trouble spots and is on by default. 609This new compiler behavior can be turned off with 610<samp><span class="option">-Wno-non-template-friend</span></samp>, which keeps the conformant compiler code 611but disables the helpful warning. 612 613 <br><dt><code>-Wold-style-cast </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wold_002dstyle_002dcast-196"></a><a name="index-Wno_002dold_002dstyle_002dcast-197"></a>Warn if an old-style (C-style) cast to a non-void type is used within 614a C++ program. The new-style casts (‘<samp><span class="samp">dynamic_cast</span></samp>’, 615‘<samp><span class="samp">static_cast</span></samp>’, ‘<samp><span class="samp">reinterpret_cast</span></samp>’, and ‘<samp><span class="samp">const_cast</span></samp>’) are 616less vulnerable to unintended effects and much easier to search for. 617 618 <br><dt><code>-Woverloaded-virtual </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Woverloaded_002dvirtual-198"></a><a name="index-Wno_002doverloaded_002dvirtual-199"></a><a name="index-overloaded-virtual-function_002c-warning-200"></a><a name="index-warning-for-overloaded-virtual-function-201"></a>Warn when a function declaration hides virtual functions from a 619base class. For example, in: 620 621 <pre class="smallexample"> struct A { 622 virtual void f(); 623 }; 624 625 struct B: public A { 626 void f(int); 627 }; 628</pre> 629 <p>the <code>A</code> class version of <code>f</code> is hidden in <code>B</code>, and code 630like: 631 632 <pre class="smallexample"> B* b; 633 b->f(); 634</pre> 635 <p class="noindent">fails to compile. 636 637 <br><dt><code>-Wno-pmf-conversions </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wno_002dpmf_002dconversions-202"></a><a name="index-Wpmf_002dconversions-203"></a>Disable the diagnostic for converting a bound pointer to member function 638to a plain pointer. 639 640 <br><dt><code>-Wsign-promo </code><span class="roman">(C++ and Objective-C++ only)</span><dd><a name="index-Wsign_002dpromo-204"></a><a name="index-Wno_002dsign_002dpromo-205"></a>Warn when overload resolution chooses a promotion from unsigned or 641enumerated type to a signed type, over a conversion to an unsigned type of 642the same size. Previous versions of G++ tried to preserve 643unsignedness, but the standard mandates the current behavior. 644</dl> 645 646 </body></html> 647 648