1 @c Copyright (C) 1988,89,92,93,94,96,98,99,2000,01 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Extensions to the C Language Family
7 @cindex extensions, C language
8 @cindex C language extensions
10 GNU C provides several language features not found in ANSI standard C.
11 (The @samp{-pedantic} option directs GNU CC to print a warning message if
12 any of these features is used.) To test for the availability of these
13 features in conditional compilation, check for a predefined macro
14 @code{__GNUC__}, which is always defined under GNU CC.
16 These extensions are available in C and Objective C. Most of them are
17 also available in C++. @xref{C++ Extensions,,Extensions to the
18 C++ Language}, for extensions that apply @emph{only} to C++.
20 @c The only difference between the two versions of this menu is that the
21 @c version for clear INTERNALS has an extra node, "Constraints" (which
22 @c appears in a separate chapter in the other version of the manual).
25 * Statement Exprs:: Putting statements and declarations inside expressions.
26 * Local Labels:: Labels local to a statement-expression.
27 * Labels as Values:: Getting pointers to labels, and computed gotos.
28 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
29 * Constructing Calls:: Dispatching a call to another function.
30 * Naming Types:: Giving a name to the type of some expression.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Hex Floats:: Hexadecimal floating-point constants.
37 * Zero Length:: Zero-length arrays.
38 * Variable Length:: Arrays whose length is computed at run time.
39 * Macro Varargs:: Macros with variable number of arguments.
40 * Subscripting:: Any array can be subscripted, even if not an lvalue.
41 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
42 * Initializers:: Non-constant initializers.
43 * Constructors:: Constructor expressions give structures, unions
45 * Labeled Elements:: Labeling elements of initializers.
46 * Cast to Union:: Casting to union type from any member of the union.
47 * Case Ranges:: `case 1 ... 9' and such.
48 * Function Attributes:: Declaring that functions have no side effects,
49 or that they can never return.
50 * Function Prototypes:: Prototype declarations and old-style definitions.
51 * C++ Comments:: C++ comments are recognized.
52 * Dollar Signs:: Dollar sign is allowed in identifiers.
53 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
54 * Variable Attributes:: Specifying attributes of variables.
55 * Type Attributes:: Specifying attributes of types.
56 * Alignment:: Inquiring about the alignment of a type or variable.
57 * Inline:: Defining inline functions (as fast as macros).
58 * Extended Asm:: Assembler instructions with C expressions as operands.
59 (With them you can define ``built-in'' functions.)
60 * Asm Labels:: Specifying the assembler name to use for a C symbol.
61 * Explicit Reg Vars:: Defining variables residing in specified registers.
62 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
63 * Incomplete Enums:: @code{enum foo;}, with details to follow.
64 * Function Names:: Printable strings which are the name of the current
66 * Return Address:: Getting the return or frame address of a function.
67 * Other Builtins:: Other built-in functions.
68 * Deprecated Features:: Things might disappear from g++.
69 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
74 * Statement Exprs:: Putting statements and declarations inside expressions.
75 * Local Labels:: Labels local to a statement-expression.
76 * Labels as Values:: Getting pointers to labels, and computed gotos.
77 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
78 * Constructing Calls:: Dispatching a call to another function.
79 * Naming Types:: Giving a name to the type of some expression.
80 * Typeof:: @code{typeof}: referring to the type of an expression.
81 * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues.
82 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
83 * Long Long:: Double-word integers---@code{long long int}.
84 * Complex:: Data types for complex numbers.
85 * Hex Floats:: Hexadecimal floating-point constants.
86 * Zero Length:: Zero-length arrays.
87 * Variable Length:: Arrays whose length is computed at run time.
88 * Macro Varargs:: Macros with variable number of arguments.
89 * Subscripting:: Any array can be subscripted, even if not an lvalue.
90 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
91 * Initializers:: Non-constant initializers.
92 * Constructors:: Constructor expressions give structures, unions
94 * Labeled Elements:: Labeling elements of initializers.
95 * Cast to Union:: Casting to union type from any member of the union.
96 * Case Ranges:: `case 1 ... 9' and such.
97 * Function Attributes:: Declaring that functions have no side effects,
98 or that they can never return.
99 * Function Prototypes:: Prototype declarations and old-style definitions.
100 * C++ Comments:: C++ comments are recognized.
101 * Dollar Signs:: Dollar sign is allowed in identifiers.
102 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
103 * Variable Attributes:: Specifying attributes of variables.
104 * Type Attributes:: Specifying attributes of types.
105 * Alignment:: Inquiring about the alignment of a type or variable.
106 * Inline:: Defining inline functions (as fast as macros).
107 * Extended Asm:: Assembler instructions with C expressions as operands.
108 (With them you can define ``built-in'' functions.)
109 * Constraints:: Constraints for asm operands
110 * Asm Labels:: Specifying the assembler name to use for a C symbol.
111 * Explicit Reg Vars:: Defining variables residing in specified registers.
112 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
113 * Incomplete Enums:: @code{enum foo;}, with details to follow.
114 * Function Names:: Printable strings which are the name of the current
116 * Return Address:: Getting the return or frame address of a function.
117 * Deprecated Features:: Things might disappear from g++.
118 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
119 * Other Builtins:: Other built-in functions.
123 @node Statement Exprs
124 @section Statements and Declarations in Expressions
125 @cindex statements inside expressions
126 @cindex declarations inside expressions
127 @cindex expressions containing statements
128 @cindex macros, statements in expressions
130 @c the above section title wrapped and causes an underfull hbox.. i
131 @c changed it from "within" to "in". --mew 4feb93
133 A compound statement enclosed in parentheses may appear as an expression
134 in GNU C. This allows you to use loops, switches, and local variables
135 within an expression.
137 Recall that a compound statement is a sequence of statements surrounded
138 by braces; in this construct, parentheses go around the braces. For
142 (@{ int y = foo (); int z;
149 is a valid (though slightly more complex than necessary) expression
150 for the absolute value of @code{foo ()}.
152 The last thing in the compound statement should be an expression
153 followed by a semicolon; the value of this subexpression serves as the
154 value of the entire construct. (If you use some other kind of statement
155 last within the braces, the construct has type @code{void}, and thus
156 effectively no value.)
158 This feature is especially useful in making macro definitions ``safe'' (so
159 that they evaluate each operand exactly once). For example, the
160 ``maximum'' function is commonly defined as a macro in standard C as
164 #define max(a,b) ((a) > (b) ? (a) : (b))
168 @cindex side effects, macro argument
169 But this definition computes either @var{a} or @var{b} twice, with bad
170 results if the operand has side effects. In GNU C, if you know the
171 type of the operands (here let's assume @code{int}), you can define
172 the macro safely as follows:
175 #define maxint(a,b) \
176 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
179 Embedded statements are not allowed in constant expressions, such as
180 the value of an enumeration constant, the width of a bit field, or
181 the initial value of a static variable.
183 If you don't know the type of the operand, you can still do this, but you
184 must use @code{typeof} (@pxref{Typeof}) or type naming (@pxref{Naming
187 Statement expressions are not supported fully in G++, and their fate
188 there is unclear. (It is possible that they will become fully supported
189 at some point, or that they will be deprecated, or that the bugs that
190 are present will continue to exist indefinitely.) Presently, statement
191 expressions do not work well as default arguments.
193 In addition, there are semantic issues with statement-expressions in
194 C++. If you try to use statement-expressions instead of inline
195 functions in C++, you may be surprised at the way object destruction is
196 handled. For example:
199 #define foo(a) (@{int b = (a); b + 3; @})
203 does not work the same way as:
206 inline int foo(int a) @{ int b = a; return b + 3; @}
210 In particular, if the expression passed into @code{foo} involves the
211 creation of temporaries, the destructors for those temporaries will be
212 run earlier in the case of the macro than in the case of the function.
214 These considerations mean that it is probably a bad idea to use
215 statement-expressions of this form in header files that are designed to
216 work with C++. (Note that some versions of the GNU C Library contained
217 header files using statement-expression that lead to precisely this
221 @section Locally Declared Labels
223 @cindex macros, local labels
225 Each statement expression is a scope in which @dfn{local labels} can be
226 declared. A local label is simply an identifier; you can jump to it
227 with an ordinary @code{goto} statement, but only from within the
228 statement expression it belongs to.
230 A local label declaration looks like this:
233 __label__ @var{label};
240 __label__ @var{label1}, @var{label2}, @dots{};
243 Local label declarations must come at the beginning of the statement
244 expression, right after the @samp{(@{}, before any ordinary
247 The label declaration defines the label @emph{name}, but does not define
248 the label itself. You must do this in the usual way, with
249 @code{@var{label}:}, within the statements of the statement expression.
251 The local label feature is useful because statement expressions are
252 often used in macros. If the macro contains nested loops, a @code{goto}
253 can be useful for breaking out of them. However, an ordinary label
254 whose scope is the whole function cannot be used: if the macro can be
255 expanded several times in one function, the label will be multiply
256 defined in that function. A local label avoids this problem. For
260 #define SEARCH(array, target) \
263 typeof (target) _SEARCH_target = (target); \
264 typeof (*(array)) *_SEARCH_array = (array); \
267 for (i = 0; i < max; i++) \
268 for (j = 0; j < max; j++) \
269 if (_SEARCH_array[i][j] == _SEARCH_target) \
270 @{ value = i; goto found; @} \
277 @node Labels as Values
278 @section Labels as Values
279 @cindex labels as values
280 @cindex computed gotos
281 @cindex goto with computed label
282 @cindex address of a label
284 You can get the address of a label defined in the current function
285 (or a containing function) with the unary operator @samp{&&}. The
286 value has type @code{void *}. This value is a constant and can be used
287 wherever a constant of that type is valid. For example:
295 To use these values, you need to be able to jump to one. This is done
296 with the computed goto statement@footnote{The analogous feature in
297 Fortran is called an assigned goto, but that name seems inappropriate in
298 C, where one can do more than simply store label addresses in label
299 variables.}, @code{goto *@var{exp};}. For example,
306 Any expression of type @code{void *} is allowed.
308 One way of using these constants is in initializing a static array that
309 will serve as a jump table:
312 static void *array[] = @{ &&foo, &&bar, &&hack @};
315 Then you can select a label with indexing, like this:
322 Note that this does not check whether the subscript is in bounds---array
323 indexing in C never does that.
325 Such an array of label values serves a purpose much like that of the
326 @code{switch} statement. The @code{switch} statement is cleaner, so
327 use that rather than an array unless the problem does not fit a
328 @code{switch} statement very well.
330 Another use of label values is in an interpreter for threaded code.
331 The labels within the interpreter function can be stored in the
332 threaded code for super-fast dispatching.
334 You may not use this mechanism to jump to code in a different function.
335 If you do that, totally unpredictable things will happen. The best way to
336 avoid this is to store the label address only in automatic variables and
337 never pass it as an argument.
339 An alternate way to write the above example is
342 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo, &&hack - &&foo @};
343 goto *(&&foo + array[i]);
347 This is more friendly to code living in shared libraries, as it reduces
348 the number of dynamic relocations that are needed, and by consequence,
349 allows the data to be read-only.
351 @node Nested Functions
352 @section Nested Functions
353 @cindex nested functions
354 @cindex downward funargs
357 A @dfn{nested function} is a function defined inside another function.
358 (Nested functions are not supported for GNU C++.) The nested function's
359 name is local to the block where it is defined. For example, here we
360 define a nested function named @code{square}, and call it twice:
364 foo (double a, double b)
366 double square (double z) @{ return z * z; @}
368 return square (a) + square (b);
373 The nested function can access all the variables of the containing
374 function that are visible at the point of its definition. This is
375 called @dfn{lexical scoping}. For example, here we show a nested
376 function which uses an inherited variable named @code{offset}:
379 bar (int *array, int offset, int size)
381 int access (int *array, int index)
382 @{ return array[index + offset]; @}
385 for (i = 0; i < size; i++)
386 @dots{} access (array, i) @dots{}
390 Nested function definitions are permitted within functions in the places
391 where variable definitions are allowed; that is, in any block, before
392 the first statement in the block.
394 It is possible to call the nested function from outside the scope of its
395 name by storing its address or passing the address to another function:
398 hack (int *array, int size)
400 void store (int index, int value)
401 @{ array[index] = value; @}
403 intermediate (store, size);
407 Here, the function @code{intermediate} receives the address of
408 @code{store} as an argument. If @code{intermediate} calls @code{store},
409 the arguments given to @code{store} are used to store into @code{array}.
410 But this technique works only so long as the containing function
411 (@code{hack}, in this example) does not exit.
413 If you try to call the nested function through its address after the
414 containing function has exited, all hell will break loose. If you try
415 to call it after a containing scope level has exited, and if it refers
416 to some of the variables that are no longer in scope, you may be lucky,
417 but it's not wise to take the risk. If, however, the nested function
418 does not refer to anything that has gone out of scope, you should be
421 GNU CC implements taking the address of a nested function using a
422 technique called @dfn{trampolines}. A paper describing them is
423 available as @uref{http://people.debian.org/~karlheg/Usenix88-lexic.pdf}.
425 A nested function can jump to a label inherited from a containing
426 function, provided the label was explicitly declared in the containing
427 function (@pxref{Local Labels}). Such a jump returns instantly to the
428 containing function, exiting the nested function which did the
429 @code{goto} and any intermediate functions as well. Here is an example:
433 bar (int *array, int offset, int size)
436 int access (int *array, int index)
440 return array[index + offset];
444 for (i = 0; i < size; i++)
445 @dots{} access (array, i) @dots{}
449 /* @r{Control comes here from @code{access}
450 if it detects an error.} */
457 A nested function always has internal linkage. Declaring one with
458 @code{extern} is erroneous. If you need to declare the nested function
459 before its definition, use @code{auto} (which is otherwise meaningless
460 for function declarations).
463 bar (int *array, int offset, int size)
466 auto int access (int *, int);
468 int access (int *array, int index)
472 return array[index + offset];
478 @node Constructing Calls
479 @section Constructing Function Calls
480 @cindex constructing calls
481 @cindex forwarding calls
483 Using the built-in functions described below, you can record
484 the arguments a function received, and call another function
485 with the same arguments, without knowing the number or types
488 You can also record the return value of that function call,
489 and later return that value, without knowing what data type
490 the function tried to return (as long as your caller expects
494 @findex __builtin_apply_args
495 @item __builtin_apply_args ()
496 This built-in function returns a pointer of type @code{void *} to data
497 describing how to perform a call with the same arguments as were passed
498 to the current function.
500 The function saves the arg pointer register, structure value address,
501 and all registers that might be used to pass arguments to a function
502 into a block of memory allocated on the stack. Then it returns the
503 address of that block.
505 @findex __builtin_apply
506 @item __builtin_apply (@var{function}, @var{arguments}, @var{size})
507 This built-in function invokes @var{function} (type @code{void (*)()})
508 with a copy of the parameters described by @var{arguments} (type
509 @code{void *}) and @var{size} (type @code{int}).
511 The value of @var{arguments} should be the value returned by
512 @code{__builtin_apply_args}. The argument @var{size} specifies the size
513 of the stack argument data, in bytes.
515 This function returns a pointer of type @code{void *} to data describing
516 how to return whatever value was returned by @var{function}. The data
517 is saved in a block of memory allocated on the stack.
519 It is not always simple to compute the proper value for @var{size}. The
520 value is used by @code{__builtin_apply} to compute the amount of data
521 that should be pushed on the stack and copied from the incoming argument
524 @findex __builtin_return
525 @item __builtin_return (@var{result})
526 This built-in function returns the value described by @var{result} from
527 the containing function. You should specify, for @var{result}, a value
528 returned by @code{__builtin_apply}.
532 @section Naming an Expression's Type
535 You can give a name to the type of an expression using a @code{typedef}
536 declaration with an initializer. Here is how to define @var{name} as a
537 type name for the type of @var{exp}:
540 typedef @var{name} = @var{exp};
543 This is useful in conjunction with the statements-within-expressions
544 feature. Here is how the two together can be used to define a safe
545 ``maximum'' macro that operates on any arithmetic type:
549 (@{typedef _ta = (a), _tb = (b); \
550 _ta _a = (a); _tb _b = (b); \
551 _a > _b ? _a : _b; @})
554 @cindex underscores in variables in macros
555 @cindex @samp{_} in variables in macros
556 @cindex local variables in macros
557 @cindex variables, local, in macros
558 @cindex macros, local variables in
560 The reason for using names that start with underscores for the local
561 variables is to avoid conflicts with variable names that occur within the
562 expressions that are substituted for @code{a} and @code{b}. Eventually we
563 hope to design a new form of declaration syntax that allows you to declare
564 variables whose scopes start only after their initializers; this will be a
565 more reliable way to prevent such conflicts.
568 @section Referring to a Type with @code{typeof}
571 @cindex macros, types of arguments
573 Another way to refer to the type of an expression is with @code{typeof}.
574 The syntax of using of this keyword looks like @code{sizeof}, but the
575 construct acts semantically like a type name defined with @code{typedef}.
577 There are two ways of writing the argument to @code{typeof}: with an
578 expression or with a type. Here is an example with an expression:
585 This assumes that @code{x} is an array of functions; the type described
586 is that of the values of the functions.
588 Here is an example with a typename as the argument:
595 Here the type described is that of pointers to @code{int}.
597 If you are writing a header file that must work when included in ANSI C
598 programs, write @code{__typeof__} instead of @code{typeof}.
599 @xref{Alternate Keywords}.
601 A @code{typeof}-construct can be used anywhere a typedef name could be
602 used. For example, you can use it in a declaration, in a cast, or inside
603 of @code{sizeof} or @code{typeof}.
607 This declares @code{y} with the type of what @code{x} points to.
614 This declares @code{y} as an array of such values.
621 This declares @code{y} as an array of pointers to characters:
624 typeof (typeof (char *)[4]) y;
628 It is equivalent to the following traditional C declaration:
634 To see the meaning of the declaration using @code{typeof}, and why it
635 might be a useful way to write, let's rewrite it with these macros:
638 #define pointer(T) typeof(T *)
639 #define array(T, N) typeof(T [N])
643 Now the declaration can be rewritten this way:
646 array (pointer (char), 4) y;
650 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
651 pointers to @code{char}.
655 @section Generalized Lvalues
656 @cindex compound expressions as lvalues
657 @cindex expressions, compound, as lvalues
658 @cindex conditional expressions as lvalues
659 @cindex expressions, conditional, as lvalues
660 @cindex casts as lvalues
661 @cindex generalized lvalues
662 @cindex lvalues, generalized
663 @cindex extensions, @code{?:}
664 @cindex @code{?:} extensions
665 Compound expressions, conditional expressions and casts are allowed as
666 lvalues provided their operands are lvalues. This means that you can take
667 their addresses or store values into them.
669 Standard C++ allows compound expressions and conditional expressions as
670 lvalues, and permits casts to reference type, so use of this extension
671 is deprecated for C++ code.
673 For example, a compound expression can be assigned, provided the last
674 expression in the sequence is an lvalue. These two expressions are
682 Similarly, the address of the compound expression can be taken. These two
683 expressions are equivalent:
690 A conditional expression is a valid lvalue if its type is not void and the
691 true and false branches are both valid lvalues. For example, these two
692 expressions are equivalent:
696 (a ? b = 5 : (c = 5))
699 A cast is a valid lvalue if its operand is an lvalue. A simple
700 assignment whose left-hand side is a cast works by converting the
701 right-hand side first to the specified type, then to the type of the
702 inner left-hand side expression. After this is stored, the value is
703 converted back to the specified type to become the value of the
704 assignment. Thus, if @code{a} has type @code{char *}, the following two
705 expressions are equivalent:
709 (int)(a = (char *)(int)5)
712 An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
713 performs the arithmetic using the type resulting from the cast, and then
714 continues as in the previous case. Therefore, these two expressions are
719 (int)(a = (char *)(int) ((int)a + 5))
722 You cannot take the address of an lvalue cast, because the use of its
723 address would not work out coherently. Suppose that @code{&(int)f} were
724 permitted, where @code{f} has type @code{float}. Then the following
725 statement would try to store an integer bit-pattern where a floating
726 point number belongs:
732 This is quite different from what @code{(int)f = 1} would do---that
733 would convert 1 to floating point and store it. Rather than cause this
734 inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
736 If you really do want an @code{int *} pointer with the address of
737 @code{f}, you can simply write @code{(int *)&f}.
740 @section Conditionals with Omitted Operands
741 @cindex conditional expressions, extensions
742 @cindex omitted middle-operands
743 @cindex middle-operands, omitted
744 @cindex extensions, @code{?:}
745 @cindex @code{?:} extensions
747 The middle operand in a conditional expression may be omitted. Then
748 if the first operand is nonzero, its value is the value of the conditional
751 Therefore, the expression
758 has the value of @code{x} if that is nonzero; otherwise, the value of
761 This example is perfectly equivalent to
767 @cindex side effect in ?:
768 @cindex ?: side effect
770 In this simple case, the ability to omit the middle operand is not
771 especially useful. When it becomes useful is when the first operand does,
772 or may (if it is a macro argument), contain a side effect. Then repeating
773 the operand in the middle would perform the side effect twice. Omitting
774 the middle operand uses the value already computed without the undesirable
775 effects of recomputing it.
778 @section Double-Word Integers
779 @cindex @code{long long} data types
780 @cindex double-word arithmetic
781 @cindex multiprecision arithmetic
783 GNU C supports data types for integers that are twice as long as
784 @code{int}. Simply write @code{long long int} for a signed integer, or
785 @code{unsigned long long int} for an unsigned integer. To make an
786 integer constant of type @code{long long int}, add the suffix @code{LL}
787 to the integer. To make an integer constant of type @code{unsigned long
788 long int}, add the suffix @code{ULL} to the integer.
790 You can use these types in arithmetic like any other integer types.
791 Addition, subtraction, and bitwise boolean operations on these types
792 are open-coded on all types of machines. Multiplication is open-coded
793 if the machine supports fullword-to-doubleword a widening multiply
794 instruction. Division and shifts are open-coded only on machines that
795 provide special support. The operations that are not open-coded use
796 special library routines that come with GNU CC.
798 There may be pitfalls when you use @code{long long} types for function
799 arguments, unless you declare function prototypes. If a function
800 expects type @code{int} for its argument, and you pass a value of type
801 @code{long long int}, confusion will result because the caller and the
802 subroutine will disagree about the number of bytes for the argument.
803 Likewise, if the function expects @code{long long int} and you pass
804 @code{int}. The best way to avoid such problems is to use prototypes.
807 @section Complex Numbers
808 @cindex complex numbers
810 GNU C supports complex data types. You can declare both complex integer
811 types and complex floating types, using the keyword @code{__complex__}.
813 For example, @samp{__complex__ double x;} declares @code{x} as a
814 variable whose real part and imaginary part are both of type
815 @code{double}. @samp{__complex__ short int y;} declares @code{y} to
816 have real and imaginary parts of type @code{short int}; this is not
817 likely to be useful, but it shows that the set of complex types is
820 To write a constant with a complex data type, use the suffix @samp{i} or
821 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
822 has type @code{__complex__ float} and @code{3i} has type
823 @code{__complex__ int}. Such a constant always has a pure imaginary
824 value, but you can form any complex value you like by adding one to a
827 To extract the real part of a complex-valued expression @var{exp}, write
828 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
829 extract the imaginary part.
831 The operator @samp{~} performs complex conjugation when used on a value
834 GNU CC can allocate complex automatic variables in a noncontiguous
835 fashion; it's even possible for the real part to be in a register while
836 the imaginary part is on the stack (or vice-versa). None of the
837 supported debugging info formats has a way to represent noncontiguous
838 allocation like this, so GNU CC describes a noncontiguous complex
839 variable as if it were two separate variables of noncomplex type.
840 If the variable's actual name is @code{foo}, the two fictitious
841 variables are named @code{foo$real} and @code{foo$imag}. You can
842 examine and set these two fictitious variables with your debugger.
844 A future version of GDB will know how to recognize such pairs and treat
845 them as a single variable with a complex type.
851 GNU CC recognizes floating-point numbers written not only in the usual
852 decimal notation, such as @code{1.55e1}, but also numbers such as
853 @code{0x1.fp3} written in hexadecimal format. In that format the
854 @code{0x} hex introducer and the @code{p} or @code{P} exponent field are
855 mandatory. The exponent is a decimal number that indicates the power of
856 2 by which the significant part will be multiplied. Thus @code{0x1.f} is
857 1 15/16, @code{p3} multiplies it by 8, and the value of @code{0x1.fp3}
858 is the same as @code{1.55e1}.
860 Unlike for floating-point numbers in the decimal notation the exponent
861 is always required in the hexadecimal notation. Otherwise the compiler
862 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
863 could mean @code{1.0f} or @code{1.9375} since @code{f} is also the
864 extension for floating-point constants of type @code{float}.
867 @section Arrays of Length Zero
868 @cindex arrays of length zero
869 @cindex zero-length arrays
870 @cindex length-zero arrays
871 @cindex flexible array members
873 Zero-length arrays are allowed in GNU C. They are very useful as the
874 last element of a structure which is really a header for a variable-length
883 struct line *thisline = (struct line *)
884 malloc (sizeof (struct line) + this_length);
885 thisline->length = this_length;
888 In ISO C89, you would have to give @code{contents} a length of 1, which
889 means either you waste space or complicate the argument to @code{malloc}.
891 In ISO C99, you would use a @dfn{flexible array member}, which is
892 slightly different in syntax and semantics:
896 Flexible array members are written as @code{contents[]} without
900 Flexible array members have incomplete type, and so the @code{sizeof}
901 operator may not be applied. As a quirk of the original implementation
902 of zero-length arrays, @code{sizeof} evaluates to zero.
905 Flexible array members may only appear as the last member of a
906 @code{struct} that is otherwise non-empty. GCC currently allows
907 zero-length arrays anywhere. You may encounter problems, however,
908 defining structures containing only a zero-length array. Such usage
909 is deprecated, and we recommend using zero-length arrays only in
910 places in which flexible array members would be allowed.
913 GCC versions before 3.0 allowed zero-length arrays to be statically
914 initialized. In addition to those cases that were useful, it also
915 allowed initializations in situations that would corrupt later data.
916 Non-empty initialization of zero-length arrays is now deprecated.
918 Instead GCC allows static initialization of flexible array members.
919 This is equivalent to defining a new structure containing the original
920 structure followed by an array of sufficient size to contain the data.
921 I.e. in the following, @code{f1} is constructed as if it were declared
927 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
930 struct f1 f1; int data[3];
931 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
935 The convenience of this extension is that @code{f1} has the desired
936 type, eliminating the need to consistently refer to @code{f2.f1}.
938 This has symmetry with normal static arrays, in that an array of
939 unknown size is also written with @code{[]}.
941 Of course, this extension only makes sense if the extra data comes at
942 the end of a top-level object, as otherwise we would be overwriting
943 data at subsequent offsets. To avoid undue complication and confusion
944 with initialization of deeply nested arrays, we simply disallow any
945 non-empty initialization except when the structure is the top-level
949 struct foo @{ int x; int y[]; @};
950 struct bar @{ struct foo z; @};
952 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // Legal.
953 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // Illegal.
954 struct bar c = @{ @{ 1, @{ @} @} @}; // Legal.
955 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // Illegal.
958 @node Variable Length
959 @section Arrays of Variable Length
960 @cindex variable-length arrays
961 @cindex arrays of variable length
963 Variable-length automatic arrays are allowed in GNU C. These arrays are
964 declared like any other automatic arrays, but with a length that is not
965 a constant expression. The storage is allocated at the point of
966 declaration and deallocated when the brace-level is exited. For
971 concat_fopen (char *s1, char *s2, char *mode)
973 char str[strlen (s1) + strlen (s2) + 1];
976 return fopen (str, mode);
980 @cindex scope of a variable length array
981 @cindex variable-length array scope
982 @cindex deallocating variable length arrays
983 Jumping or breaking out of the scope of the array name deallocates the
984 storage. Jumping into the scope is not allowed; you get an error
987 @cindex @code{alloca} vs variable-length arrays
988 You can use the function @code{alloca} to get an effect much like
989 variable-length arrays. The function @code{alloca} is available in
990 many other C implementations (but not in all). On the other hand,
991 variable-length arrays are more elegant.
993 There are other differences between these two methods. Space allocated
994 with @code{alloca} exists until the containing @emph{function} returns.
995 The space for a variable-length array is deallocated as soon as the array
996 name's scope ends. (If you use both variable-length arrays and
997 @code{alloca} in the same function, deallocation of a variable-length array
998 will also deallocate anything more recently allocated with @code{alloca}.)
1000 You can also use variable-length arrays as arguments to functions:
1004 tester (int len, char data[len][len])
1010 The length of an array is computed once when the storage is allocated
1011 and is remembered for the scope of the array in case you access it with
1014 If you want to pass the array first and the length afterward, you can
1015 use a forward declaration in the parameter list---another GNU extension.
1019 tester (int len; char data[len][len], int len)
1025 @cindex parameter forward declaration
1026 The @samp{int len} before the semicolon is a @dfn{parameter forward
1027 declaration}, and it serves the purpose of making the name @code{len}
1028 known when the declaration of @code{data} is parsed.
1030 You can write any number of such parameter forward declarations in the
1031 parameter list. They can be separated by commas or semicolons, but the
1032 last one must end with a semicolon, which is followed by the ``real''
1033 parameter declarations. Each forward declaration must match a ``real''
1034 declaration in parameter name and data type.
1037 @section Macros with Variable Numbers of Arguments
1038 @cindex variable number of arguments
1039 @cindex macro with variable arguments
1040 @cindex rest argument (in macro)
1042 In GNU C, a macro can accept a variable number of arguments, much as a
1043 function can. The syntax for defining the macro looks much like that
1044 used for a function. Here is an example:
1047 #define eprintf(format, args...) \
1048 fprintf (stderr, format , ## args)
1051 Here @code{args} is a @dfn{rest argument}: it takes in zero or more
1052 arguments, as many as the call contains. All of them plus the commas
1053 between them form the value of @code{args}, which is substituted into
1054 the macro body where @code{args} is used. Thus, we have this expansion:
1057 eprintf ("%s:%d: ", input_file_name, line_number)
1059 fprintf (stderr, "%s:%d: " , input_file_name, line_number)
1063 Note that the comma after the string constant comes from the definition
1064 of @code{eprintf}, whereas the last comma comes from the value of
1067 The reason for using @samp{##} is to handle the case when @code{args}
1068 matches no arguments at all. In this case, @code{args} has an empty
1069 value. In this case, the second comma in the definition becomes an
1070 embarrassment: if it got through to the expansion of the macro, we would
1071 get something like this:
1074 fprintf (stderr, "success!\n" , )
1078 which is invalid C syntax. @samp{##} gets rid of the comma, so we get
1079 the following instead:
1082 fprintf (stderr, "success!\n")
1085 This is a special feature of the GNU C preprocessor: @samp{##} before a
1086 rest argument that is empty discards the preceding sequence of
1087 non-whitespace characters from the macro definition. (If another macro
1088 argument precedes, none of it is discarded.)
1090 It might be better to discard the last preprocessor token instead of the
1091 last preceding sequence of non-whitespace characters; in fact, we may
1092 someday change this feature to do so. We advise you to write the macro
1093 definition so that the preceding sequence of non-whitespace characters
1094 is just a single token, so that the meaning will not change if we change
1095 the definition of this feature.
1098 @section Non-Lvalue Arrays May Have Subscripts
1099 @cindex subscripting
1100 @cindex arrays, non-lvalue
1102 @cindex subscripting and function values
1103 Subscripting is allowed on arrays that are not lvalues, even though the
1104 unary @samp{&} operator is not. For example, this is valid in GNU C though
1105 not valid in other C dialects:
1109 struct foo @{int a[4];@};
1115 return f().a[index];
1121 @section Arithmetic on @code{void}- and Function-Pointers
1122 @cindex void pointers, arithmetic
1123 @cindex void, size of pointer to
1124 @cindex function pointers, arithmetic
1125 @cindex function, size of pointer to
1127 In GNU C, addition and subtraction operations are supported on pointers to
1128 @code{void} and on pointers to functions. This is done by treating the
1129 size of a @code{void} or of a function as 1.
1131 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1132 and on function types, and returns 1.
1134 The option @samp{-Wpointer-arith} requests a warning if these extensions
1138 @section Non-Constant Initializers
1139 @cindex initializers, non-constant
1140 @cindex non-constant initializers
1142 As in standard C++, the elements of an aggregate initializer for an
1143 automatic variable are not required to be constant expressions in GNU C.
1144 Here is an example of an initializer with run-time varying elements:
1147 foo (float f, float g)
1149 float beat_freqs[2] = @{ f-g, f+g @};
1155 @section Constructor Expressions
1156 @cindex constructor expressions
1157 @cindex initializations in expressions
1158 @cindex structures, constructor expression
1159 @cindex expressions, constructor
1161 GNU C supports constructor expressions. A constructor looks like
1162 a cast containing an initializer. Its value is an object of the
1163 type specified in the cast, containing the elements specified in
1166 Usually, the specified type is a structure. Assume that
1167 @code{struct foo} and @code{structure} are declared as shown:
1170 struct foo @{int a; char b[2];@} structure;
1174 Here is an example of constructing a @code{struct foo} with a constructor:
1177 structure = ((struct foo) @{x + y, 'a', 0@});
1181 This is equivalent to writing the following:
1185 struct foo temp = @{x + y, 'a', 0@};
1190 You can also construct an array. If all the elements of the constructor
1191 are (made up of) simple constant expressions, suitable for use in
1192 initializers, then the constructor is an lvalue and can be coerced to a
1193 pointer to its first element, as shown here:
1196 char **foo = (char *[]) @{ "x", "y", "z" @};
1199 Array constructors whose elements are not simple constants are
1200 not very useful, because the constructor is not an lvalue. There
1201 are only two valid ways to use it: to subscript it, or initialize
1202 an array variable with it. The former is probably slower than a
1203 @code{switch} statement, while the latter does the same thing an
1204 ordinary C initializer would do. Here is an example of
1205 subscripting an array constructor:
1208 output = ((int[]) @{ 2, x, 28 @}) [input];
1211 Constructor expressions for scalar types and union types are is
1212 also allowed, but then the constructor expression is equivalent
1215 @node Labeled Elements
1216 @section Labeled Elements in Initializers
1217 @cindex initializers with labeled elements
1218 @cindex labeled elements in initializers
1219 @cindex case labels in initializers
1221 Standard C89 requires the elements of an initializer to appear in a fixed
1222 order, the same as the order of the elements in the array or structure
1225 In ISO C99 you can give the elements in any order, specifying the array
1226 indices or structure field names they apply to, and GNU C allows this as
1227 an extension in C89 mode as well. This extension is not
1228 implemented in GNU C++.
1230 To specify an array index, write
1231 @samp{[@var{index}] =} before the element value. For example,
1234 int a[6] = @{ [4] = 29, [2] = 15 @};
1241 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1245 The index values must be constant expressions, even if the array being
1246 initialized is automatic.
1248 An alternative syntax for this which has been obsolete since GCC 2.5 but
1249 GCC still accepts is to write @samp{[@var{index}]} before the element
1250 value, with no @samp{=}.
1252 To initialize a range of elements to the same value, write
1253 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1254 extension. For example,
1257 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1261 Note that the length of the array is the highest value specified
1264 In a structure initializer, specify the name of a field to initialize
1265 with @samp{.@var{fieldname} =} before the element value. For example,
1266 given the following structure,
1269 struct point @{ int x, y; @};
1273 the following initialization
1276 struct point p = @{ .y = yvalue, .x = xvalue @};
1283 struct point p = @{ xvalue, yvalue @};
1286 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1287 @samp{@var{fieldname}:}, as shown here:
1290 struct point p = @{ y: yvalue, x: xvalue @};
1293 You can also use an element label (with either the colon syntax or the
1294 period-equal syntax) when initializing a union, to specify which element
1295 of the union should be used. For example,
1298 union foo @{ int i; double d; @};
1300 union foo f = @{ .d = 4 @};
1304 will convert 4 to a @code{double} to store it in the union using
1305 the second element. By contrast, casting 4 to type @code{union foo}
1306 would store it into the union as the integer @code{i}, since it is
1307 an integer. (@xref{Cast to Union}.)
1309 You can combine this technique of naming elements with ordinary C
1310 initialization of successive elements. Each initializer element that
1311 does not have a label applies to the next consecutive element of the
1312 array or structure. For example,
1315 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1322 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1325 Labeling the elements of an array initializer is especially useful
1326 when the indices are characters or belong to an @code{enum} type.
1331 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1332 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1335 You can also write a series of @samp{.@var{fieldname}} and
1336 @samp{[@var{index}]} element labels before an @samp{=} to specify a
1337 nested subobject to initialize; the list is taken relative to the
1338 subobject corresponding to the closest surrounding brace pair. For
1339 example, with the @samp{struct point} declaration above:
1342 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1346 @section Case Ranges
1348 @cindex ranges in case statements
1350 You can specify a range of consecutive values in a single @code{case} label,
1354 case @var{low} ... @var{high}:
1358 This has the same effect as the proper number of individual @code{case}
1359 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1361 This feature is especially useful for ranges of ASCII character codes:
1367 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1368 it may be parsed wrong when you use it with integer values. For example,
1383 @section Cast to a Union Type
1384 @cindex cast to a union
1385 @cindex union, casting to a
1387 A cast to union type is similar to other casts, except that the type
1388 specified is a union type. You can specify the type either with
1389 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1390 a constructor though, not a cast, and hence does not yield an lvalue like
1391 normal casts. (@xref{Constructors}.)
1393 The types that may be cast to the union type are those of the members
1394 of the union. Thus, given the following union and variables:
1397 union foo @{ int i; double d; @};
1403 both @code{x} and @code{y} can be cast to type @code{union} foo.
1405 Using the cast as the right-hand side of an assignment to a variable of
1406 union type is equivalent to storing in a member of the union:
1411 u = (union foo) x @equiv{} u.i = x
1412 u = (union foo) y @equiv{} u.d = y
1415 You can also use the union cast as a function argument:
1418 void hack (union foo);
1420 hack ((union foo) x);
1423 @node Function Attributes
1424 @section Declaring Attributes of Functions
1425 @cindex function attributes
1426 @cindex declaring attributes of functions
1427 @cindex functions that never return
1428 @cindex functions that have no side effects
1429 @cindex functions in arbitrary sections
1430 @cindex functions that behave like malloc
1431 @cindex @code{volatile} applied to function
1432 @cindex @code{const} applied to function
1433 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1434 @cindex functions that are passed arguments in registers on the 386
1435 @cindex functions that pop the argument stack on the 386
1436 @cindex functions that do not pop the argument stack on the 386
1438 In GNU C, you declare certain things about functions called in your program
1439 which help the compiler optimize function calls and check your code more
1442 The keyword @code{__attribute__} allows you to specify special
1443 attributes when making a declaration. This keyword is followed by an
1444 attribute specification inside double parentheses. Fourteen attributes,
1445 @code{noreturn}, @code{pure}, @code{const}, @code{format},
1446 @code{format_arg}, @code{no_instrument_function}, @code{section},
1447 @code{constructor}, @code{destructor}, @code{unused}, @code{weak},
1448 @code{malloc}, @code{alias} and @code{no_check_memory_usage} are
1449 currently defined for functions. Several other attributes are defined
1450 for functions on particular target systems. Other attributes, including
1451 @code{section} are supported for variables declarations (@pxref{Variable
1452 Attributes}) and for types (@pxref{Type Attributes}).
1454 You may also specify attributes with @samp{__} preceding and following
1455 each keyword. This allows you to use them in header files without
1456 being concerned about a possible macro of the same name. For example,
1457 you may use @code{__noreturn__} instead of @code{noreturn}.
1460 @cindex @code{noreturn} function attribute
1462 A few standard library functions, such as @code{abort} and @code{exit},
1463 cannot return. GNU CC knows this automatically. Some programs define
1464 their own functions that never return. You can declare them
1465 @code{noreturn} to tell the compiler this fact. For example,
1468 void fatal () __attribute__ ((noreturn));
1473 @dots{} /* @r{Print error message.} */ @dots{}
1478 The @code{noreturn} keyword tells the compiler to assume that
1479 @code{fatal} cannot return. It can then optimize without regard to what
1480 would happen if @code{fatal} ever did return. This makes slightly
1481 better code. More importantly, it helps avoid spurious warnings of
1482 uninitialized variables.
1484 Do not assume that registers saved by the calling function are
1485 restored before calling the @code{noreturn} function.
1487 It does not make sense for a @code{noreturn} function to have a return
1488 type other than @code{void}.
1490 The attribute @code{noreturn} is not implemented in GNU C versions
1491 earlier than 2.5. An alternative way to declare that a function does
1492 not return, which works in the current version and in some older
1493 versions, is as follows:
1496 typedef void voidfn ();
1498 volatile voidfn fatal;
1501 @cindex @code{pure} function attribute
1503 Many functions have no effects except the return value and their
1504 return value depends only on the parameters and/or global variables.
1505 Such a function can be subject
1506 to common subexpression elimination and loop optimization just as an
1507 arithmetic operator would be. These functions should be declared
1508 with the attribute @code{pure}. For example,
1511 int square (int) __attribute__ ((pure));
1515 says that the hypothetical function @code{square} is safe to call
1516 fewer times than the program says.
1518 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
1519 Interesting non-pure functions are functions with infinite loops or those
1520 depending on volatile memory or other system resource, that may change between
1521 two consecutive calls (such as @code{feof} in a multithreading environment).
1523 The attribute @code{pure} is not implemented in GNU C versions earlier
1525 @cindex @code{const} function attribute
1527 Many functions do not examine any values except their arguments, and
1528 have no effects except the return value. Basically this is just slightly
1529 more strict class than the "pure" attribute above, since function is not
1530 allowed to read global memory.
1532 @cindex pointer arguments
1533 Note that a function that has pointer arguments and examines the data
1534 pointed to must @emph{not} be declared @code{const}. Likewise, a
1535 function that calls a non-@code{const} function usually must not be
1536 @code{const}. It does not make sense for a @code{const} function to
1539 The attribute @code{const} is not implemented in GNU C versions earlier
1540 than 2.5. An alternative way to declare that a function has no side
1541 effects, which works in the current version and in some older versions,
1545 typedef int intfn ();
1547 extern const intfn square;
1550 This approach does not work in GNU C++ from 2.6.0 on, since the language
1551 specifies that the @samp{const} must be attached to the return value.
1554 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1555 @cindex @code{format} function attribute
1556 The @code{format} attribute specifies that a function takes @code{printf},
1557 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1558 should be type-checked against a format string. For example, the
1563 my_printf (void *my_object, const char *my_format, ...)
1564 __attribute__ ((format (printf, 2, 3)));
1568 causes the compiler to check the arguments in calls to @code{my_printf}
1569 for consistency with the @code{printf} style format string argument
1572 The parameter @var{archetype} determines how the format string is
1573 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1574 or @code{strfmon}. (You can also use @code{__printf__},
1575 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1576 parameter @var{string-index} specifies which argument is the format
1577 string argument (starting from 1), while @var{first-to-check} is the
1578 number of the first argument to check against the format string. For
1579 functions where the arguments are not available to be checked (such as
1580 @code{vprintf}), specify the third parameter as zero. In this case the
1581 compiler only checks the format string for consistency. For
1582 @code{strftime} formats, the third parameter is required to be zero.
1584 In the example above, the format string (@code{my_format}) is the second
1585 argument of the function @code{my_print}, and the arguments to check
1586 start with the third argument, so the correct parameters for the format
1587 attribute are 2 and 3.
1589 The @code{format} attribute allows you to identify your own functions
1590 which take format strings as arguments, so that GNU CC can check the
1591 calls to these functions for errors. The compiler always (unless
1592 @samp{-ffreestanding} is used) checks formats
1593 for the standard library functions @code{printf}, @code{fprintf},
1594 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1595 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1596 warnings are requested (using @samp{-Wformat}), so there is no need to
1597 modify the header file @file{stdio.h}. In C99 mode, the functions
1598 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1599 @code{vsscanf} are also checked. Except in strictly conforming C
1600 standard modes, the X/Open function @code{strfmon} is also checked.
1601 @xref{C Dialect Options,,Options Controlling C Dialect}.
1603 @item format_arg (@var{string-index})
1604 @cindex @code{format_arg} function attribute
1605 The @code{format_arg} attribute specifies that a function takes a format
1606 string for a @code{printf}, @code{scanf}, @code{strftime} or
1607 @code{strfmon} style function and modifies it (for example, to translate
1608 it into another language), so the result can be passed to a
1609 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1610 function (with the remaining arguments to the format function the same
1611 as they would have been for the unmodified string). For example, the
1616 my_dgettext (char *my_domain, const char *my_format)
1617 __attribute__ ((format_arg (2)));
1621 causes the compiler to check the arguments in calls to a @code{printf},
1622 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1623 format string argument is a call to the @code{my_dgettext} function, for
1624 consistency with the format string argument @code{my_format}. If the
1625 @code{format_arg} attribute had not been specified, all the compiler
1626 could tell in such calls to format functions would be that the format
1627 string argument is not constant; this would generate a warning when
1628 @code{-Wformat-nonliteral} is used, but the calls could not be checked
1629 without the attribute.
1631 The parameter @var{string-index} specifies which argument is the format
1632 string argument (starting from 1).
1634 The @code{format-arg} attribute allows you to identify your own
1635 functions which modify format strings, so that GNU CC can check the
1636 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1637 type function whose operands are a call to one of your own function.
1638 The compiler always treats @code{gettext}, @code{dgettext}, and
1639 @code{dcgettext} in this manner except when strict ISO C support is
1640 requested by @samp{-ansi} or an appropriate @samp{-std} option, or
1641 @samp{-ffreestanding} is used. @xref{C Dialect Options,,Options
1642 Controlling C Dialect}.
1644 @item no_instrument_function
1645 @cindex @code{no_instrument_function} function attribute
1646 If @samp{-finstrument-functions} is given, profiling function calls will
1647 be generated at entry and exit of most user-compiled functions.
1648 Functions with this attribute will not be so instrumented.
1650 @item section ("section-name")
1651 @cindex @code{section} function attribute
1652 Normally, the compiler places the code it generates in the @code{text} section.
1653 Sometimes, however, you need additional sections, or you need certain
1654 particular functions to appear in special sections. The @code{section}
1655 attribute specifies that a function lives in a particular section.
1656 For example, the declaration:
1659 extern void foobar (void) __attribute__ ((section ("bar")));
1663 puts the function @code{foobar} in the @code{bar} section.
1665 Some file formats do not support arbitrary sections so the @code{section}
1666 attribute is not available on all platforms.
1667 If you need to map the entire contents of a module to a particular
1668 section, consider using the facilities of the linker instead.
1672 @cindex @code{constructor} function attribute
1673 @cindex @code{destructor} function attribute
1674 The @code{constructor} attribute causes the function to be called
1675 automatically before execution enters @code{main ()}. Similarly, the
1676 @code{destructor} attribute causes the function to be called
1677 automatically after @code{main ()} has completed or @code{exit ()} has
1678 been called. Functions with these attributes are useful for
1679 initializing data that will be used implicitly during the execution of
1682 These attributes are not currently implemented for Objective C.
1685 This attribute, attached to a function, means that the function is meant
1686 to be possibly unused. GNU CC will not produce a warning for this
1687 function. GNU C++ does not currently support this attribute as
1688 definitions without parameters are valid in C++.
1691 @cindex @code{weak} attribute
1692 The @code{weak} attribute causes the declaration to be emitted as a weak
1693 symbol rather than a global. This is primarily useful in defining
1694 library functions which can be overridden in user code, though it can
1695 also be used with non-function declarations. Weak symbols are supported
1696 for ELF targets, and also for a.out targets when using the GNU assembler
1700 @cindex @code{malloc} attribute
1701 The @code{malloc} attribute is used to tell the compiler that a function
1702 may be treated as if it were the malloc function. The compiler assumes
1703 that calls to malloc result in a pointers that cannot alias anything.
1704 This will often improve optimization.
1706 @item alias ("target")
1707 @cindex @code{alias} attribute
1708 The @code{alias} attribute causes the declaration to be emitted as an
1709 alias for another symbol, which must be specified. For instance,
1712 void __f () @{ /* do something */; @}
1713 void f () __attribute__ ((weak, alias ("__f")));
1716 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
1717 mangled name for the target must be used.
1719 Not all target machines support this attribute.
1721 @item no_check_memory_usage
1722 @cindex @code{no_check_memory_usage} function attribute
1723 The @code{no_check_memory_usage} attribute causes GNU CC to omit checks
1724 of memory references when it generates code for that function. Normally
1725 if you specify @samp{-fcheck-memory-usage} (see @pxref{Code Gen
1726 Options}), GNU CC generates calls to support routines before most memory
1727 accesses to permit support code to record usage and detect uses of
1728 uninitialized or unallocated storage. Since GNU CC cannot handle
1729 @code{asm} statements properly they are not allowed in such functions.
1730 If you declare a function with this attribute, GNU CC will not generate
1731 memory checking code for that function, permitting the use of @code{asm}
1732 statements without having to compile that function with different
1733 options. This also allows you to write support routines of your own if
1734 you wish, without getting infinite recursion if they get compiled with
1735 @code{-fcheck-memory-usage}.
1737 @item regparm (@var{number})
1738 @cindex functions that are passed arguments in registers on the 386
1739 On the Intel 386, the @code{regparm} attribute causes the compiler to
1740 pass up to @var{number} integer arguments in registers @var{EAX},
1741 @var{EDX}, and @var{ECX} instead of on the stack. Functions that take a
1742 variable number of arguments will continue to be passed all of their
1743 arguments on the stack.
1746 @cindex functions that pop the argument stack on the 386
1747 On the Intel 386, the @code{stdcall} attribute causes the compiler to
1748 assume that the called function will pop off the stack space used to
1749 pass arguments, unless it takes a variable number of arguments.
1751 The PowerPC compiler for Windows NT currently ignores the @code{stdcall}
1755 @cindex functions that do pop the argument stack on the 386
1756 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1757 assume that the calling function will pop off the stack space used to
1758 pass arguments. This is
1759 useful to override the effects of the @samp{-mrtd} switch.
1761 The PowerPC compiler for Windows NT currently ignores the @code{cdecl}
1765 @cindex functions called via pointer on the RS/6000 and PowerPC
1766 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
1767 compiler to always call the function via a pointer, so that functions
1768 which reside further than 64 megabytes (67,108,864 bytes) from the
1769 current location can be called.
1771 @item long_call/short_call
1772 @cindex indirect calls on ARM
1773 This attribute allows to specify how to call a particular function on
1774 ARM. Both attributes override the @code{-mlong-calls} (@pxref{ARM Options})
1775 command line switch and @code{#pragma long_calls} settings. The
1776 @code{long_call} attribute causes the compiler to always call the
1777 function by first loading its address into a register and then using the
1778 contents of that register. The @code{short_call} attribute always places
1779 the offset to the function from the call site into the @samp{BL}
1780 instruction directly.
1783 @cindex functions which are imported from a dll on PowerPC Windows NT
1784 On the PowerPC running Windows NT, the @code{dllimport} attribute causes
1785 the compiler to call the function via a global pointer to the function
1786 pointer that is set up by the Windows NT dll library. The pointer name
1787 is formed by combining @code{__imp_} and the function name.
1790 @cindex functions which are exported from a dll on PowerPC Windows NT
1791 On the PowerPC running Windows NT, the @code{dllexport} attribute causes
1792 the compiler to provide a global pointer to the function pointer, so
1793 that it can be called with the @code{dllimport} attribute. The pointer
1794 name is formed by combining @code{__imp_} and the function name.
1796 @item exception (@var{except-func} [, @var{except-arg}])
1797 @cindex functions which specify exception handling on PowerPC Windows NT
1798 On the PowerPC running Windows NT, the @code{exception} attribute causes
1799 the compiler to modify the structured exception table entry it emits for
1800 the declared function. The string or identifier @var{except-func} is
1801 placed in the third entry of the structured exception table. It
1802 represents a function, which is called by the exception handling
1803 mechanism if an exception occurs. If it was specified, the string or
1804 identifier @var{except-arg} is placed in the fourth entry of the
1805 structured exception table.
1807 @item function_vector
1808 @cindex calling functions through the function vector on the H8/300 processors
1809 Use this option on the H8/300 and H8/300H to indicate that the specified
1810 function should be called through the function vector. Calling a
1811 function through the function vector will reduce code size, however;
1812 the function vector has a limited size (maximum 128 entries on the H8/300
1813 and 64 entries on the H8/300H) and shares space with the interrupt vector.
1815 You must use GAS and GLD from GNU binutils version 2.7 or later for
1816 this option to work correctly.
1818 @item interrupt_handler
1819 @cindex interrupt handler functions on the H8/300 processors
1820 Use this option on the H8/300 and H8/300H to indicate that the specified
1821 function is an interrupt handler. The compiler will generate function
1822 entry and exit sequences suitable for use in an interrupt handler when this
1823 attribute is present.
1826 @cindex interrupt handler functions
1827 Use this option on the ARM, AVR and M32R/D ports to indicate that the
1828 specified function is an interrupt handler. The compiler will generate
1829 function entry and exit sequences suitable for use in an interrupt
1830 handler when this attribute is present.
1832 Note, interrupt handlers for the H8/300 and H8/300H processors can be
1833 specified via the @code{interrupt_handler} attribute.
1835 Note, on the AVR interrupts will be enabled inside the function.
1837 Note, for the ARM you can specify the kind of interrupt to be handled by
1838 adding an optional parameter to the interrupt attribute like this:
1841 void f () __attribute__ ((interrupt ("IRQ")));
1844 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF.
1847 @cindex eight bit data on the H8/300 and H8/300H
1848 Use this option on the H8/300 and H8/300H to indicate that the specified
1849 variable should be placed into the eight bit data section.
1850 The compiler will generate more efficient code for certain operations
1851 on data in the eight bit data area. Note the eight bit data area is limited to
1854 You must use GAS and GLD from GNU binutils version 2.7 or later for
1855 this option to work correctly.
1858 @cindex tiny data section on the H8/300H
1859 Use this option on the H8/300H to indicate that the specified
1860 variable should be placed into the tiny data section.
1861 The compiler will generate more efficient code for loads and stores
1862 on data in the tiny data section. Note the tiny data area is limited to
1863 slightly under 32kbytes of data.
1866 @cindex signal handler functions on the AVR processors
1867 Use this option on the AVR to indicate that the specified
1868 function is an signal handler. The compiler will generate function
1869 entry and exit sequences suitable for use in an signal handler when this
1870 attribute is present. Interrupts will be disabled inside function.
1873 @cindex function without a prologue/epilogue code
1874 Use this option on the ARM or AVR ports to indicate that the specified
1875 function do not need prologue/epilogue sequences generated by the
1876 compiler. It is up to the programmer to provide these sequences.
1878 @item model (@var{model-name})
1879 @cindex function addressability on the M32R/D
1880 Use this attribute on the M32R/D to set the addressability of an object,
1881 and the code generated for a function.
1882 The identifier @var{model-name} is one of @code{small}, @code{medium},
1883 or @code{large}, representing each of the code models.
1885 Small model objects live in the lower 16MB of memory (so that their
1886 addresses can be loaded with the @code{ld24} instruction), and are
1887 callable with the @code{bl} instruction.
1889 Medium model objects may live anywhere in the 32 bit address space (the
1890 compiler will generate @code{seth/add3} instructions to load their addresses),
1891 and are callable with the @code{bl} instruction.
1893 Large model objects may live anywhere in the 32 bit address space (the
1894 compiler will generate @code{seth/add3} instructions to load their addresses),
1895 and may not be reachable with the @code{bl} instruction (the compiler will
1896 generate the much slower @code{seth/add3/jl} instruction sequence).
1900 You can specify multiple attributes in a declaration by separating them
1901 by commas within the double parentheses or by immediately following an
1902 attribute declaration with another attribute declaration.
1904 @cindex @code{#pragma}, reason for not using
1905 @cindex pragma, reason for not using
1906 Some people object to the @code{__attribute__} feature, suggesting that
1907 ISO C's @code{#pragma} should be used instead. At the time
1908 @code{__attribute__} was designed, there were two reasons for not doing
1913 It is impossible to generate @code{#pragma} commands from a macro.
1916 There is no telling what the same @code{#pragma} might mean in another
1920 These two reasons applied to almost any application that might have been
1921 proposed for @code{#pragma}. It was basically a mistake to use
1922 @code{#pragma} for @emph{anything}.
1924 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
1925 to be generated from macros. In addition, a @code{#pragma GCC}
1926 namespace is now in use for GCC-specific pragmas. However, it has been
1927 found convenient to use @code{__attribute__} to achieve a natural
1928 attachment of attributes to their corresponding declarations, whereas
1929 @code{#pragma GCC} is of use for constructs that do not naturally form
1930 part of the grammar. @xref{Other Directives,,Miscellaneous
1931 Preprocessing Directives, cpp, The C Preprocessor}.
1933 @node Function Prototypes
1934 @section Prototypes and Old-Style Function Definitions
1935 @cindex function prototype declarations
1936 @cindex old-style function definitions
1937 @cindex promotion of formal parameters
1939 GNU C extends ANSI C to allow a function prototype to override a later
1940 old-style non-prototype definition. Consider the following example:
1943 /* @r{Use prototypes unless the compiler is old-fashioned.} */
1950 /* @r{Prototype function declaration.} */
1951 int isroot P((uid_t));
1953 /* @r{Old-style function definition.} */
1955 isroot (x) /* ??? lossage here ??? */
1962 Suppose the type @code{uid_t} happens to be @code{short}. ANSI C does
1963 not allow this example, because subword arguments in old-style
1964 non-prototype definitions are promoted. Therefore in this example the
1965 function definition's argument is really an @code{int}, which does not
1966 match the prototype argument type of @code{short}.
1968 This restriction of ANSI C makes it hard to write code that is portable
1969 to traditional C compilers, because the programmer does not know
1970 whether the @code{uid_t} type is @code{short}, @code{int}, or
1971 @code{long}. Therefore, in cases like these GNU C allows a prototype
1972 to override a later old-style definition. More precisely, in GNU C, a
1973 function prototype argument type overrides the argument type specified
1974 by a later old-style definition if the former type is the same as the
1975 latter type before promotion. Thus in GNU C the above example is
1976 equivalent to the following:
1988 GNU C++ does not support old-style function definitions, so this
1989 extension is irrelevant.
1992 @section C++ Style Comments
1994 @cindex C++ comments
1995 @cindex comments, C++ style
1997 In GNU C, you may use C++ style comments, which start with @samp{//} and
1998 continue until the end of the line. Many other C implementations allow
1999 such comments, and they are likely to be in a future C standard.
2000 However, C++ style comments are not recognized if you specify
2001 @w{@samp{-ansi}} or @w{@samp{-traditional}}, since they are incompatible
2002 with traditional constructs like @code{dividend//*comment*/divisor}.
2005 @section Dollar Signs in Identifier Names
2007 @cindex dollar signs in identifier names
2008 @cindex identifier names, dollar signs in
2010 In GNU C, you may normally use dollar signs in identifier names.
2011 This is because many traditional C implementations allow such identifiers.
2012 However, dollar signs in identifiers are not supported on a few target
2013 machines, typically because the target assembler does not allow them.
2015 @node Character Escapes
2016 @section The Character @key{ESC} in Constants
2018 You can use the sequence @samp{\e} in a string or character constant to
2019 stand for the ASCII character @key{ESC}.
2022 @section Inquiring on Alignment of Types or Variables
2024 @cindex type alignment
2025 @cindex variable alignment
2027 The keyword @code{__alignof__} allows you to inquire about how an object
2028 is aligned, or the minimum alignment usually required by a type. Its
2029 syntax is just like @code{sizeof}.
2031 For example, if the target machine requires a @code{double} value to be
2032 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2033 This is true on many RISC machines. On more traditional machine
2034 designs, @code{__alignof__ (double)} is 4 or even 2.
2036 Some machines never actually require alignment; they allow reference to any
2037 data type even at an odd addresses. For these machines, @code{__alignof__}
2038 reports the @emph{recommended} alignment of a type.
2040 When the operand of @code{__alignof__} is an lvalue rather than a type, the
2041 value is the largest alignment that the lvalue is known to have. It may
2042 have this alignment as a result of its data type, or because it is part of
2043 a structure and inherits alignment from that structure. For example, after
2047 struct foo @{ int x; char y; @} foo1;
2051 the value of @code{__alignof__ (foo1.y)} is probably 2 or 4, the same as
2052 @code{__alignof__ (int)}, even though the data type of @code{foo1.y}
2053 does not itself demand any alignment.@refill
2055 It is an error to ask for the alignment of an incomplete type.
2057 A related feature which lets you specify the alignment of an object is
2058 @code{__attribute__ ((aligned (@var{alignment})))}; see the following
2061 @node Variable Attributes
2062 @section Specifying Attributes of Variables
2063 @cindex attribute of variables
2064 @cindex variable attributes
2066 The keyword @code{__attribute__} allows you to specify special
2067 attributes of variables or structure fields. This keyword is followed
2068 by an attribute specification inside double parentheses. Eight
2069 attributes are currently defined for variables: @code{aligned},
2070 @code{mode}, @code{nocommon}, @code{packed}, @code{section},
2071 @code{transparent_union}, @code{unused}, and @code{weak}. Some other
2072 attributes are defined for variables on particular target systems. Other
2073 attributes are available for functions (@pxref{Function Attributes}) and
2074 for types (@pxref{Type Attributes}). Other front-ends might define more
2075 attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2077 You may also specify attributes with @samp{__} preceding and following
2078 each keyword. This allows you to use them in header files without
2079 being concerned about a possible macro of the same name. For example,
2080 you may use @code{__aligned__} instead of @code{aligned}.
2083 @cindex @code{aligned} attribute
2084 @item aligned (@var{alignment})
2085 This attribute specifies a minimum alignment for the variable or
2086 structure field, measured in bytes. For example, the declaration:
2089 int x __attribute__ ((aligned (16))) = 0;
2093 causes the compiler to allocate the global variable @code{x} on a
2094 16-byte boundary. On a 68040, this could be used in conjunction with
2095 an @code{asm} expression to access the @code{move16} instruction which
2096 requires 16-byte aligned operands.
2098 You can also specify the alignment of structure fields. For example, to
2099 create a double-word aligned @code{int} pair, you could write:
2102 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2106 This is an alternative to creating a union with a @code{double} member
2107 that forces the union to be double-word aligned.
2109 It is not possible to specify the alignment of functions; the alignment
2110 of functions is determined by the machine's requirements and cannot be
2111 changed. You cannot specify alignment for a typedef name because such a
2112 name is just an alias, not a distinct type.
2114 As in the preceding examples, you can explicitly specify the alignment
2115 (in bytes) that you wish the compiler to use for a given variable or
2116 structure field. Alternatively, you can leave out the alignment factor
2117 and just ask the compiler to align a variable or field to the maximum
2118 useful alignment for the target machine you are compiling for. For
2119 example, you could write:
2122 short array[3] __attribute__ ((aligned));
2125 Whenever you leave out the alignment factor in an @code{aligned} attribute
2126 specification, the compiler automatically sets the alignment for the declared
2127 variable or field to the largest alignment which is ever used for any data
2128 type on the target machine you are compiling for. Doing this can often make
2129 copy operations more efficient, because the compiler can use whatever
2130 instructions copy the biggest chunks of memory when performing copies to
2131 or from the variables or fields that you have aligned this way.
2133 The @code{aligned} attribute can only increase the alignment; but you
2134 can decrease it by specifying @code{packed} as well. See below.
2136 Note that the effectiveness of @code{aligned} attributes may be limited
2137 by inherent limitations in your linker. On many systems, the linker is
2138 only able to arrange for variables to be aligned up to a certain maximum
2139 alignment. (For some linkers, the maximum supported alignment may
2140 be very very small.) If your linker is only able to align variables
2141 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
2142 in an @code{__attribute__} will still only provide you with 8 byte
2143 alignment. See your linker documentation for further information.
2145 @item mode (@var{mode})
2146 @cindex @code{mode} attribute
2147 This attribute specifies the data type for the declaration---whichever
2148 type corresponds to the mode @var{mode}. This in effect lets you
2149 request an integer or floating point type according to its width.
2151 You may also specify a mode of @samp{byte} or @samp{__byte__} to
2152 indicate the mode corresponding to a one-byte integer, @samp{word} or
2153 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
2154 or @samp{__pointer__} for the mode used to represent pointers.
2157 @cindex @code{nocommon} attribute
2158 This attribute specifies requests GNU CC not to place a variable
2159 ``common'' but instead to allocate space for it directly. If you
2160 specify the @samp{-fno-common} flag, GNU CC will do this for all
2163 Specifying the @code{nocommon} attribute for a variable provides an
2164 initialization of zeros. A variable may only be initialized in one
2168 @cindex @code{packed} attribute
2169 The @code{packed} attribute specifies that a variable or structure field
2170 should have the smallest possible alignment---one byte for a variable,
2171 and one bit for a field, unless you specify a larger value with the
2172 @code{aligned} attribute.
2174 Here is a structure in which the field @code{x} is packed, so that it
2175 immediately follows @code{a}:
2181 int x[2] __attribute__ ((packed));
2185 @item section ("section-name")
2186 @cindex @code{section} variable attribute
2187 Normally, the compiler places the objects it generates in sections like
2188 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
2189 or you need certain particular variables to appear in special sections,
2190 for example to map to special hardware. The @code{section}
2191 attribute specifies that a variable (or function) lives in a particular
2192 section. For example, this small program uses several specific section names:
2195 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
2196 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
2197 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
2198 int init_data __attribute__ ((section ("INITDATA"))) = 0;
2202 /* Initialize stack pointer */
2203 init_sp (stack + sizeof (stack));
2205 /* Initialize initialized data */
2206 memcpy (&init_data, &data, &edata - &data);
2208 /* Turn on the serial ports */
2215 Use the @code{section} attribute with an @emph{initialized} definition
2216 of a @emph{global} variable, as shown in the example. GNU CC issues
2217 a warning and otherwise ignores the @code{section} attribute in
2218 uninitialized variable declarations.
2220 You may only use the @code{section} attribute with a fully initialized
2221 global definition because of the way linkers work. The linker requires
2222 each object be defined once, with the exception that uninitialized
2223 variables tentatively go in the @code{common} (or @code{bss}) section
2224 and can be multiply "defined". You can force a variable to be
2225 initialized with the @samp{-fno-common} flag or the @code{nocommon}
2228 Some file formats do not support arbitrary sections so the @code{section}
2229 attribute is not available on all platforms.
2230 If you need to map the entire contents of a module to a particular
2231 section, consider using the facilities of the linker instead.
2234 @cindex @code{shared} variable attribute
2235 On Windows NT, in addition to putting variable definitions in a named
2236 section, the section can also be shared among all running copies of an
2237 executable or DLL. For example, this small program defines shared data
2238 by putting it in a named section "shared" and marking the section
2242 int foo __attribute__((section ("shared"), shared)) = 0;
2247 /* Read and write foo. All running copies see the same value. */
2253 You may only use the @code{shared} attribute along with @code{section}
2254 attribute with a fully initialized global definition because of the way
2255 linkers work. See @code{section} attribute for more information.
2257 The @code{shared} attribute is only available on Windows NT.
2259 @item transparent_union
2260 This attribute, attached to a function parameter which is a union, means
2261 that the corresponding argument may have the type of any union member,
2262 but the argument is passed as if its type were that of the first union
2263 member. For more details see @xref{Type Attributes}. You can also use
2264 this attribute on a @code{typedef} for a union data type; then it
2265 applies to all function parameters with that type.
2268 This attribute, attached to a variable, means that the variable is meant
2269 to be possibly unused. GNU CC will not produce a warning for this
2273 The @code{weak} attribute is described in @xref{Function Attributes}.
2275 @item model (@var{model-name})
2276 @cindex variable addressability on the M32R/D
2277 Use this attribute on the M32R/D to set the addressability of an object.
2278 The identifier @var{model-name} is one of @code{small}, @code{medium},
2279 or @code{large}, representing each of the code models.
2281 Small model objects live in the lower 16MB of memory (so that their
2282 addresses can be loaded with the @code{ld24} instruction).
2284 Medium and large model objects may live anywhere in the 32 bit address space
2285 (the compiler will generate @code{seth/add3} instructions to load their
2290 To specify multiple attributes, separate them by commas within the
2291 double parentheses: for example, @samp{__attribute__ ((aligned (16),
2294 @node Type Attributes
2295 @section Specifying Attributes of Types
2296 @cindex attribute of types
2297 @cindex type attributes
2299 The keyword @code{__attribute__} allows you to specify special
2300 attributes of @code{struct} and @code{union} types when you define such
2301 types. This keyword is followed by an attribute specification inside
2302 double parentheses. Four attributes are currently defined for types:
2303 @code{aligned}, @code{packed}, @code{transparent_union}, and @code{unused}.
2304 Other attributes are defined for functions (@pxref{Function Attributes}) and
2305 for variables (@pxref{Variable Attributes}).
2307 You may also specify any one of these attributes with @samp{__}
2308 preceding and following its keyword. This allows you to use these
2309 attributes in header files without being concerned about a possible
2310 macro of the same name. For example, you may use @code{__aligned__}
2311 instead of @code{aligned}.
2313 You may specify the @code{aligned} and @code{transparent_union}
2314 attributes either in a @code{typedef} declaration or just past the
2315 closing curly brace of a complete enum, struct or union type
2316 @emph{definition} and the @code{packed} attribute only past the closing
2317 brace of a definition.
2319 You may also specify attributes between the enum, struct or union
2320 tag and the name of the type rather than after the closing brace.
2323 @cindex @code{aligned} attribute
2324 @item aligned (@var{alignment})
2325 This attribute specifies a minimum alignment (in bytes) for variables
2326 of the specified type. For example, the declarations:
2329 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
2330 typedef int more_aligned_int __attribute__ ((aligned (8)));
2334 force the compiler to insure (as far as it can) that each variable whose
2335 type is @code{struct S} or @code{more_aligned_int} will be allocated and
2336 aligned @emph{at least} on a 8-byte boundary. On a Sparc, having all
2337 variables of type @code{struct S} aligned to 8-byte boundaries allows
2338 the compiler to use the @code{ldd} and @code{std} (doubleword load and
2339 store) instructions when copying one variable of type @code{struct S} to
2340 another, thus improving run-time efficiency.
2342 Note that the alignment of any given @code{struct} or @code{union} type
2343 is required by the ANSI C standard to be at least a perfect multiple of
2344 the lowest common multiple of the alignments of all of the members of
2345 the @code{struct} or @code{union} in question. This means that you @emph{can}
2346 effectively adjust the alignment of a @code{struct} or @code{union}
2347 type by attaching an @code{aligned} attribute to any one of the members
2348 of such a type, but the notation illustrated in the example above is a
2349 more obvious, intuitive, and readable way to request the compiler to
2350 adjust the alignment of an entire @code{struct} or @code{union} type.
2352 As in the preceding example, you can explicitly specify the alignment
2353 (in bytes) that you wish the compiler to use for a given @code{struct}
2354 or @code{union} type. Alternatively, you can leave out the alignment factor
2355 and just ask the compiler to align a type to the maximum
2356 useful alignment for the target machine you are compiling for. For
2357 example, you could write:
2360 struct S @{ short f[3]; @} __attribute__ ((aligned));
2363 Whenever you leave out the alignment factor in an @code{aligned}
2364 attribute specification, the compiler automatically sets the alignment
2365 for the type to the largest alignment which is ever used for any data
2366 type on the target machine you are compiling for. Doing this can often
2367 make copy operations more efficient, because the compiler can use
2368 whatever instructions copy the biggest chunks of memory when performing
2369 copies to or from the variables which have types that you have aligned
2372 In the example above, if the size of each @code{short} is 2 bytes, then
2373 the size of the entire @code{struct S} type is 6 bytes. The smallest
2374 power of two which is greater than or equal to that is 8, so the
2375 compiler sets the alignment for the entire @code{struct S} type to 8
2378 Note that although you can ask the compiler to select a time-efficient
2379 alignment for a given type and then declare only individual stand-alone
2380 objects of that type, the compiler's ability to select a time-efficient
2381 alignment is primarily useful only when you plan to create arrays of
2382 variables having the relevant (efficiently aligned) type. If you
2383 declare or use arrays of variables of an efficiently-aligned type, then
2384 it is likely that your program will also be doing pointer arithmetic (or
2385 subscripting, which amounts to the same thing) on pointers to the
2386 relevant type, and the code that the compiler generates for these
2387 pointer arithmetic operations will often be more efficient for
2388 efficiently-aligned types than for other types.
2390 The @code{aligned} attribute can only increase the alignment; but you
2391 can decrease it by specifying @code{packed} as well. See below.
2393 Note that the effectiveness of @code{aligned} attributes may be limited
2394 by inherent limitations in your linker. On many systems, the linker is
2395 only able to arrange for variables to be aligned up to a certain maximum
2396 alignment. (For some linkers, the maximum supported alignment may
2397 be very very small.) If your linker is only able to align variables
2398 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
2399 in an @code{__attribute__} will still only provide you with 8 byte
2400 alignment. See your linker documentation for further information.
2403 This attribute, attached to an @code{enum}, @code{struct}, or
2404 @code{union} type definition, specified that the minimum required memory
2405 be used to represent the type.
2407 Specifying this attribute for @code{struct} and @code{union} types is
2408 equivalent to specifying the @code{packed} attribute on each of the
2409 structure or union members. Specifying the @samp{-fshort-enums}
2410 flag on the line is equivalent to specifying the @code{packed}
2411 attribute on all @code{enum} definitions.
2413 You may only specify this attribute after a closing curly brace on an
2414 @code{enum} definition, not in a @code{typedef} declaration, unless that
2415 declaration also contains the definition of the @code{enum}.
2417 @item transparent_union
2418 This attribute, attached to a @code{union} type definition, indicates
2419 that any function parameter having that union type causes calls to that
2420 function to be treated in a special way.
2422 First, the argument corresponding to a transparent union type can be of
2423 any type in the union; no cast is required. Also, if the union contains
2424 a pointer type, the corresponding argument can be a null pointer
2425 constant or a void pointer expression; and if the union contains a void
2426 pointer type, the corresponding argument can be any pointer expression.
2427 If the union member type is a pointer, qualifiers like @code{const} on
2428 the referenced type must be respected, just as with normal pointer
2431 Second, the argument is passed to the function using the calling
2432 conventions of first member of the transparent union, not the calling
2433 conventions of the union itself. All members of the union must have the
2434 same machine representation; this is necessary for this argument passing
2437 Transparent unions are designed for library functions that have multiple
2438 interfaces for compatibility reasons. For example, suppose the
2439 @code{wait} function must accept either a value of type @code{int *} to
2440 comply with Posix, or a value of type @code{union wait *} to comply with
2441 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
2442 @code{wait} would accept both kinds of arguments, but it would also
2443 accept any other pointer type and this would make argument type checking
2444 less useful. Instead, @code{<sys/wait.h>} might define the interface
2452 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
2454 pid_t wait (wait_status_ptr_t);
2457 This interface allows either @code{int *} or @code{union wait *}
2458 arguments to be passed, using the @code{int *} calling convention.
2459 The program can call @code{wait} with arguments of either type:
2462 int w1 () @{ int w; return wait (&w); @}
2463 int w2 () @{ union wait w; return wait (&w); @}
2466 With this interface, @code{wait}'s implementation might look like this:
2469 pid_t wait (wait_status_ptr_t p)
2471 return waitpid (-1, p.__ip, 0);
2476 When attached to a type (including a @code{union} or a @code{struct}),
2477 this attribute means that variables of that type are meant to appear
2478 possibly unused. GNU CC will not produce a warning for any variables of
2479 that type, even if the variable appears to do nothing. This is often
2480 the case with lock or thread classes, which are usually defined and then
2481 not referenced, but contain constructors and destructors that have
2482 nontrivial bookkeeping functions.
2486 To specify multiple attributes, separate them by commas within the
2487 double parentheses: for example, @samp{__attribute__ ((aligned (16),
2491 @section An Inline Function is As Fast As a Macro
2492 @cindex inline functions
2493 @cindex integrating function code
2495 @cindex macros, inline alternative
2497 By declaring a function @code{inline}, you can direct GNU CC to
2498 integrate that function's code into the code for its callers. This
2499 makes execution faster by eliminating the function-call overhead; in
2500 addition, if any of the actual argument values are constant, their known
2501 values may permit simplifications at compile time so that not all of the
2502 inline function's code needs to be included. The effect on code size is
2503 less predictable; object code may be larger or smaller with function
2504 inlining, depending on the particular case. Inlining of functions is an
2505 optimization and it really ``works'' only in optimizing compilation. If
2506 you don't use @samp{-O}, no function is really inline.
2508 To declare a function inline, use the @code{inline} keyword in its
2509 declaration, like this:
2519 (If you are writing a header file to be included in ANSI C programs, write
2520 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
2521 You can also make all ``simple enough'' functions inline with the option
2522 @samp{-finline-functions}.
2524 Note that certain usages in a function definition can make it unsuitable
2525 for inline substitution. Among these usages are: use of varargs, use of
2526 alloca, use of variable sized data types (@pxref{Variable Length}),
2527 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
2528 and nested functions (@pxref{Nested Functions}). Using @samp{-Winline}
2529 will warn when a function marked @code{inline} could not be substituted,
2530 and will give the reason for the failure.
2532 Note that in C and Objective C, unlike C++, the @code{inline} keyword
2533 does not affect the linkage of the function.
2535 @cindex automatic @code{inline} for C++ member fns
2536 @cindex @code{inline} automatic for C++ member fns
2537 @cindex member fns, automatically @code{inline}
2538 @cindex C++ member fns, automatically @code{inline}
2539 GNU CC automatically inlines member functions defined within the class
2540 body of C++ programs even if they are not explicitly declared
2541 @code{inline}. (You can override this with @samp{-fno-default-inline};
2542 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
2544 @cindex inline functions, omission of
2545 When a function is both inline and @code{static}, if all calls to the
2546 function are integrated into the caller, and the function's address is
2547 never used, then the function's own assembler code is never referenced.
2548 In this case, GNU CC does not actually output assembler code for the
2549 function, unless you specify the option @samp{-fkeep-inline-functions}.
2550 Some calls cannot be integrated for various reasons (in particular,
2551 calls that precede the function's definition cannot be integrated, and
2552 neither can recursive calls within the definition). If there is a
2553 nonintegrated call, then the function is compiled to assembler code as
2554 usual. The function must also be compiled as usual if the program
2555 refers to its address, because that can't be inlined.
2557 @cindex non-static inline function
2558 When an inline function is not @code{static}, then the compiler must assume
2559 that there may be calls from other source files; since a global symbol can
2560 be defined only once in any program, the function must not be defined in
2561 the other source files, so the calls therein cannot be integrated.
2562 Therefore, a non-@code{static} inline function is always compiled on its
2563 own in the usual fashion.
2565 If you specify both @code{inline} and @code{extern} in the function
2566 definition, then the definition is used only for inlining. In no case
2567 is the function compiled on its own, not even if you refer to its
2568 address explicitly. Such an address becomes an external reference, as
2569 if you had only declared the function, and had not defined it.
2571 This combination of @code{inline} and @code{extern} has almost the
2572 effect of a macro. The way to use it is to put a function definition in
2573 a header file with these keywords, and put another copy of the
2574 definition (lacking @code{inline} and @code{extern}) in a library file.
2575 The definition in the header file will cause most calls to the function
2576 to be inlined. If any uses of the function remain, they will refer to
2577 the single copy in the library.
2579 GNU C does not inline any functions when not optimizing. It is not
2580 clear whether it is better to inline or not, in this case, but we found
2581 that a correct implementation when not optimizing was difficult. So we
2582 did the easy thing, and turned it off.
2585 @section Assembler Instructions with C Expression Operands
2586 @cindex extended @code{asm}
2587 @cindex @code{asm} expressions
2588 @cindex assembler instructions
2591 In an assembler instruction using @code{asm}, you can specify the
2592 operands of the instruction using C expressions. This means you need not
2593 guess which registers or memory locations will contain the data you want
2596 You must specify an assembler instruction template much like what
2597 appears in a machine description, plus an operand constraint string for
2600 For example, here is how to use the 68881's @code{fsinx} instruction:
2603 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
2607 Here @code{angle} is the C expression for the input operand while
2608 @code{result} is that of the output operand. Each has @samp{"f"} as its
2609 operand constraint, saying that a floating point register is required.
2610 The @samp{=} in @samp{=f} indicates that the operand is an output; all
2611 output operands' constraints must use @samp{=}. The constraints use the
2612 same language used in the machine description (@pxref{Constraints}).
2614 Each operand is described by an operand-constraint string followed by
2615 the C expression in parentheses. A colon separates the assembler
2616 template from the first output operand and another separates the last
2617 output operand from the first input, if any. Commas separate the
2618 operands within each group. The total number of operands is limited to
2619 ten or to the maximum number of operands in any instruction pattern in
2620 the machine description, whichever is greater.
2622 If there are no output operands but there are input operands, you must
2623 place two consecutive colons surrounding the place where the output
2626 Output operand expressions must be lvalues; the compiler can check this.
2627 The input operands need not be lvalues. The compiler cannot check
2628 whether the operands have data types that are reasonable for the
2629 instruction being executed. It does not parse the assembler instruction
2630 template and does not know what it means or even whether it is valid
2631 assembler input. The extended @code{asm} feature is most often used for
2632 machine instructions the compiler itself does not know exist. If
2633 the output expression cannot be directly addressed (for example, it is a
2634 bit field), your constraint must allow a register. In that case, GNU CC
2635 will use the register as the output of the @code{asm}, and then store
2636 that register into the output.
2638 The ordinary output operands must be write-only; GNU CC will assume that
2639 the values in these operands before the instruction are dead and need
2640 not be generated. Extended asm supports input-output or read-write
2641 operands. Use the constraint character @samp{+} to indicate such an
2642 operand and list it with the output operands.
2644 When the constraints for the read-write operand (or the operand in which
2645 only some of the bits are to be changed) allows a register, you may, as
2646 an alternative, logically split its function into two separate operands,
2647 one input operand and one write-only output operand. The connection
2648 between them is expressed by constraints which say they need to be in
2649 the same location when the instruction executes. You can use the same C
2650 expression for both operands, or different expressions. For example,
2651 here we write the (fictitious) @samp{combine} instruction with
2652 @code{bar} as its read-only source operand and @code{foo} as its
2653 read-write destination:
2656 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
2660 The constraint @samp{"0"} for operand 1 says that it must occupy the
2661 same location as operand 0. A digit in constraint is allowed only in an
2662 input operand and it must refer to an output operand.
2664 Only a digit in the constraint can guarantee that one operand will be in
2665 the same place as another. The mere fact that @code{foo} is the value
2666 of both operands is not enough to guarantee that they will be in the
2667 same place in the generated assembler code. The following would not
2671 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
2674 Various optimizations or reloading could cause operands 0 and 1 to be in
2675 different registers; GNU CC knows no reason not to do so. For example, the
2676 compiler might find a copy of the value of @code{foo} in one register and
2677 use it for operand 1, but generate the output operand 0 in a different
2678 register (copying it afterward to @code{foo}'s own address). Of course,
2679 since the register for operand 1 is not even mentioned in the assembler
2680 code, the result will not work, but GNU CC can't tell that.
2682 Some instructions clobber specific hard registers. To describe this,
2683 write a third colon after the input operands, followed by the names of
2684 the clobbered hard registers (given as strings). Here is a realistic
2685 example for the VAX:
2688 asm volatile ("movc3 %0,%1,%2"
2690 : "g" (from), "g" (to), "g" (count)
2691 : "r0", "r1", "r2", "r3", "r4", "r5");
2694 You may not write a clobber description in a way that overlaps with an
2695 input or output operand. For example, you may not have an operand
2696 describing a register class with one member if you mention that register
2697 in the clobber list. There is no way for you to specify that an input
2698 operand is modified without also specifying it as an output
2699 operand. Note that if all the output operands you specify are for this
2700 purpose (and hence unused), you will then also need to specify
2701 @code{volatile} for the @code{asm} construct, as described below, to
2702 prevent GNU CC from deleting the @code{asm} statement as unused.
2704 If you refer to a particular hardware register from the assembler code,
2705 you will probably have to list the register after the third colon to
2706 tell the compiler the register's value is modified. In some assemblers,
2707 the register names begin with @samp{%}; to produce one @samp{%} in the
2708 assembler code, you must write @samp{%%} in the input.
2710 If your assembler instruction can alter the condition code register, add
2711 @samp{cc} to the list of clobbered registers. GNU CC on some machines
2712 represents the condition codes as a specific hardware register;
2713 @samp{cc} serves to name this register. On other machines, the
2714 condition code is handled differently, and specifying @samp{cc} has no
2715 effect. But it is valid no matter what the machine.
2717 If your assembler instruction modifies memory in an unpredictable
2718 fashion, add @samp{memory} to the list of clobbered registers. This
2719 will cause GNU CC to not keep memory values cached in registers across
2720 the assembler instruction. You will also want to add the
2721 @code{volatile} keyword if the memory affected is not listed in the
2722 inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
2723 not count as a side-effect of the @code{asm}.
2725 You can put multiple assembler instructions together in a single
2726 @code{asm} template, separated either with newlines (written as
2727 @samp{\n}) or with semicolons if the assembler allows such semicolons.
2728 The GNU assembler allows semicolons and most Unix assemblers seem to do
2729 so. The input operands are guaranteed not to use any of the clobbered
2730 registers, and neither will the output operands' addresses, so you can
2731 read and write the clobbered registers as many times as you like. Here
2732 is an example of multiple instructions in a template; it assumes the
2733 subroutine @code{_foo} accepts arguments in registers 9 and 10:
2736 asm ("movl %0,r9;movl %1,r10;call _foo"
2738 : "g" (from), "g" (to)
2742 Unless an output operand has the @samp{&} constraint modifier, GNU CC
2743 may allocate it in the same register as an unrelated input operand, on
2744 the assumption the inputs are consumed before the outputs are produced.
2745 This assumption may be false if the assembler code actually consists of
2746 more than one instruction. In such a case, use @samp{&} for each output
2747 operand that may not overlap an input. @xref{Modifiers}.
2749 If you want to test the condition code produced by an assembler
2750 instruction, you must include a branch and a label in the @code{asm}
2751 construct, as follows:
2754 asm ("clr %0;frob %1;beq 0f;mov #1,%0;0:"
2760 This assumes your assembler supports local labels, as the GNU assembler
2761 and most Unix assemblers do.
2763 Speaking of labels, jumps from one @code{asm} to another are not
2764 supported. The compiler's optimizers do not know about these jumps, and
2765 therefore they cannot take account of them when deciding how to
2768 @cindex macros containing @code{asm}
2769 Usually the most convenient way to use these @code{asm} instructions is to
2770 encapsulate them in macros that look like functions. For example,
2774 (@{ double __value, __arg = (x); \
2775 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
2780 Here the variable @code{__arg} is used to make sure that the instruction
2781 operates on a proper @code{double} value, and to accept only those
2782 arguments @code{x} which can convert automatically to a @code{double}.
2784 Another way to make sure the instruction operates on the correct data
2785 type is to use a cast in the @code{asm}. This is different from using a
2786 variable @code{__arg} in that it converts more different types. For
2787 example, if the desired type were @code{int}, casting the argument to
2788 @code{int} would accept a pointer with no complaint, while assigning the
2789 argument to an @code{int} variable named @code{__arg} would warn about
2790 using a pointer unless the caller explicitly casts it.
2792 If an @code{asm} has output operands, GNU CC assumes for optimization
2793 purposes the instruction has no side effects except to change the output
2794 operands. This does not mean instructions with a side effect cannot be
2795 used, but you must be careful, because the compiler may eliminate them
2796 if the output operands aren't used, or move them out of loops, or
2797 replace two with one if they constitute a common subexpression. Also,
2798 if your instruction does have a side effect on a variable that otherwise
2799 appears not to change, the old value of the variable may be reused later
2800 if it happens to be found in a register.
2802 You can prevent an @code{asm} instruction from being deleted, moved
2803 significantly, or combined, by writing the keyword @code{volatile} after
2804 the @code{asm}. For example:
2807 #define get_and_set_priority(new) \
2809 asm volatile ("get_and_set_priority %0, %1": "=g" (__old) : "g" (new)); \
2814 If you write an @code{asm} instruction with no outputs, GNU CC will know
2815 the instruction has side-effects and will not delete the instruction or
2816 move it outside of loops. If the side-effects of your instruction are
2817 not purely external, but will affect variables in your program in ways
2818 other than reading the inputs and clobbering the specified registers or
2819 memory, you should write the @code{volatile} keyword to prevent future
2820 versions of GNU CC from moving the instruction around within a core
2823 An @code{asm} instruction without any operands or clobbers (and ``old
2824 style'' @code{asm}) will not be deleted or moved significantly,
2825 regardless, unless it is unreachable, the same way as if you had
2826 written a @code{volatile} keyword.
2828 Note that even a volatile @code{asm} instruction can be moved in ways
2829 that appear insignificant to the compiler, such as across jump
2830 instructions. You can't expect a sequence of volatile @code{asm}
2831 instructions to remain perfectly consecutive. If you want consecutive
2832 output, use a single @code{asm}.
2834 It is a natural idea to look for a way to give access to the condition
2835 code left by the assembler instruction. However, when we attempted to
2836 implement this, we found no way to make it work reliably. The problem
2837 is that output operands might need reloading, which would result in
2838 additional following ``store'' instructions. On most machines, these
2839 instructions would alter the condition code before there was time to
2840 test it. This problem doesn't arise for ordinary ``test'' and
2841 ``compare'' instructions because they don't have any output operands.
2843 For reasons similar to those described above, it is not possible to give
2844 an assembler instruction access to the condition code left by previous
2847 If you are writing a header file that should be includable in ANSI C
2848 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
2851 @subsection i386 floating point asm operands
2853 There are several rules on the usage of stack-like regs in
2854 asm_operands insns. These rules apply only to the operands that are
2859 Given a set of input regs that die in an asm_operands, it is
2860 necessary to know which are implicitly popped by the asm, and
2861 which must be explicitly popped by gcc.
2863 An input reg that is implicitly popped by the asm must be
2864 explicitly clobbered, unless it is constrained to match an
2868 For any input reg that is implicitly popped by an asm, it is
2869 necessary to know how to adjust the stack to compensate for the pop.
2870 If any non-popped input is closer to the top of the reg-stack than
2871 the implicitly popped reg, it would not be possible to know what the
2872 stack looked like --- it's not clear how the rest of the stack ``slides
2875 All implicitly popped input regs must be closer to the top of
2876 the reg-stack than any input that is not implicitly popped.
2878 It is possible that if an input dies in an insn, reload might
2879 use the input reg for an output reload. Consider this example:
2882 asm ("foo" : "=t" (a) : "f" (b));
2885 This asm says that input B is not popped by the asm, and that
2886 the asm pushes a result onto the reg-stack, ie, the stack is one
2887 deeper after the asm than it was before. But, it is possible that
2888 reload will think that it can use the same reg for both the input and
2889 the output, if input B dies in this insn.
2891 If any input operand uses the @code{f} constraint, all output reg
2892 constraints must use the @code{&} earlyclobber.
2894 The asm above would be written as
2897 asm ("foo" : "=&t" (a) : "f" (b));
2901 Some operands need to be in particular places on the stack. All
2902 output operands fall in this category --- there is no other way to
2903 know which regs the outputs appear in unless the user indicates
2904 this in the constraints.
2906 Output operands must specifically indicate which reg an output
2907 appears in after an asm. @code{=f} is not allowed: the operand
2908 constraints must select a class with a single reg.
2911 Output operands may not be ``inserted'' between existing stack regs.
2912 Since no 387 opcode uses a read/write operand, all output operands
2913 are dead before the asm_operands, and are pushed by the asm_operands.
2914 It makes no sense to push anywhere but the top of the reg-stack.
2916 Output operands must start at the top of the reg-stack: output
2917 operands may not ``skip'' a reg.
2920 Some asm statements may need extra stack space for internal
2921 calculations. This can be guaranteed by clobbering stack registers
2922 unrelated to the inputs and outputs.
2926 Here are a couple of reasonable asms to want to write. This asm
2927 takes one input, which is internally popped, and produces two outputs.
2930 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
2933 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
2934 and replaces them with one output. The user must code the @code{st(1)}
2935 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
2938 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
2942 @c Show the details on constraints if they do not appear elsewhere in
2948 @section Controlling Names Used in Assembler Code
2949 @cindex assembler names for identifiers
2950 @cindex names used in assembler code
2951 @cindex identifiers, names in assembler code
2953 You can specify the name to be used in the assembler code for a C
2954 function or variable by writing the @code{asm} (or @code{__asm__})
2955 keyword after the declarator as follows:
2958 int foo asm ("myfoo") = 2;
2962 This specifies that the name to be used for the variable @code{foo} in
2963 the assembler code should be @samp{myfoo} rather than the usual
2966 On systems where an underscore is normally prepended to the name of a C
2967 function or variable, this feature allows you to define names for the
2968 linker that do not start with an underscore.
2970 You cannot use @code{asm} in this way in a function @emph{definition}; but
2971 you can get the same effect by writing a declaration for the function
2972 before its definition and putting @code{asm} there, like this:
2975 extern func () asm ("FUNC");
2982 It is up to you to make sure that the assembler names you choose do not
2983 conflict with any other assembler symbols. Also, you must not use a
2984 register name; that would produce completely invalid assembler code. GNU
2985 CC does not as yet have the ability to store static variables in registers.
2986 Perhaps that will be added.
2988 @node Explicit Reg Vars
2989 @section Variables in Specified Registers
2990 @cindex explicit register variables
2991 @cindex variables in specified registers
2992 @cindex specified registers
2993 @cindex registers, global allocation
2995 GNU C allows you to put a few global variables into specified hardware
2996 registers. You can also specify the register in which an ordinary
2997 register variable should be allocated.
3001 Global register variables reserve registers throughout the program.
3002 This may be useful in programs such as programming language
3003 interpreters which have a couple of global variables that are accessed
3007 Local register variables in specific registers do not reserve the
3008 registers. The compiler's data flow analysis is capable of determining
3009 where the specified registers contain live values, and where they are
3010 available for other uses. Stores into local register variables may be deleted
3011 when they appear to be dead according to dataflow analysis. References
3012 to local register variables may be deleted or moved or simplified.
3014 These local variables are sometimes convenient for use with the extended
3015 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
3016 output of the assembler instruction directly into a particular register.
3017 (This will work provided the register you specify fits the constraints
3018 specified for that operand in the @code{asm}.)
3026 @node Global Reg Vars
3027 @subsection Defining Global Register Variables
3028 @cindex global register variables
3029 @cindex registers, global variables in
3031 You can define a global register variable in GNU C like this:
3034 register int *foo asm ("a5");
3038 Here @code{a5} is the name of the register which should be used. Choose a
3039 register which is normally saved and restored by function calls on your
3040 machine, so that library routines will not clobber it.
3042 Naturally the register name is cpu-dependent, so you would need to
3043 conditionalize your program according to cpu type. The register
3044 @code{a5} would be a good choice on a 68000 for a variable of pointer
3045 type. On machines with register windows, be sure to choose a ``global''
3046 register that is not affected magically by the function call mechanism.
3048 In addition, operating systems on one type of cpu may differ in how they
3049 name the registers; then you would need additional conditionals. For
3050 example, some 68000 operating systems call this register @code{%a5}.
3052 Eventually there may be a way of asking the compiler to choose a register
3053 automatically, but first we need to figure out how it should choose and
3054 how to enable you to guide the choice. No solution is evident.
3056 Defining a global register variable in a certain register reserves that
3057 register entirely for this use, at least within the current compilation.
3058 The register will not be allocated for any other purpose in the functions
3059 in the current compilation. The register will not be saved and restored by
3060 these functions. Stores into this register are never deleted even if they
3061 would appear to be dead, but references may be deleted or moved or
3064 It is not safe to access the global register variables from signal
3065 handlers, or from more than one thread of control, because the system
3066 library routines may temporarily use the register for other things (unless
3067 you recompile them specially for the task at hand).
3069 @cindex @code{qsort}, and global register variables
3070 It is not safe for one function that uses a global register variable to
3071 call another such function @code{foo} by way of a third function
3072 @code{lose} that was compiled without knowledge of this variable (i.e. in a
3073 different source file in which the variable wasn't declared). This is
3074 because @code{lose} might save the register and put some other value there.
3075 For example, you can't expect a global register variable to be available in
3076 the comparison-function that you pass to @code{qsort}, since @code{qsort}
3077 might have put something else in that register. (If you are prepared to
3078 recompile @code{qsort} with the same global register variable, you can
3079 solve this problem.)
3081 If you want to recompile @code{qsort} or other source files which do not
3082 actually use your global register variable, so that they will not use that
3083 register for any other purpose, then it suffices to specify the compiler
3084 option @samp{-ffixed-@var{reg}}. You need not actually add a global
3085 register declaration to their source code.
3087 A function which can alter the value of a global register variable cannot
3088 safely be called from a function compiled without this variable, because it
3089 could clobber the value the caller expects to find there on return.
3090 Therefore, the function which is the entry point into the part of the
3091 program that uses the global register variable must explicitly save and
3092 restore the value which belongs to its caller.
3094 @cindex register variable after @code{longjmp}
3095 @cindex global register after @code{longjmp}
3096 @cindex value after @code{longjmp}
3099 On most machines, @code{longjmp} will restore to each global register
3100 variable the value it had at the time of the @code{setjmp}. On some
3101 machines, however, @code{longjmp} will not change the value of global
3102 register variables. To be portable, the function that called @code{setjmp}
3103 should make other arrangements to save the values of the global register
3104 variables, and to restore them in a @code{longjmp}. This way, the same
3105 thing will happen regardless of what @code{longjmp} does.
3107 All global register variable declarations must precede all function
3108 definitions. If such a declaration could appear after function
3109 definitions, the declaration would be too late to prevent the register from
3110 being used for other purposes in the preceding functions.
3112 Global register variables may not have initial values, because an
3113 executable file has no means to supply initial contents for a register.
3115 On the Sparc, there are reports that g3 @dots{} g7 are suitable
3116 registers, but certain library functions, such as @code{getwd}, as well
3117 as the subroutines for division and remainder, modify g3 and g4. g1 and
3118 g2 are local temporaries.
3120 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
3121 Of course, it will not do to use more than a few of those.
3123 @node Local Reg Vars
3124 @subsection Specifying Registers for Local Variables
3125 @cindex local variables, specifying registers
3126 @cindex specifying registers for local variables
3127 @cindex registers for local variables
3129 You can define a local register variable with a specified register
3133 register int *foo asm ("a5");
3137 Here @code{a5} is the name of the register which should be used. Note
3138 that this is the same syntax used for defining global register
3139 variables, but for a local variable it would appear within a function.
3141 Naturally the register name is cpu-dependent, but this is not a
3142 problem, since specific registers are most often useful with explicit
3143 assembler instructions (@pxref{Extended Asm}). Both of these things
3144 generally require that you conditionalize your program according to
3147 In addition, operating systems on one type of cpu may differ in how they
3148 name the registers; then you would need additional conditionals. For
3149 example, some 68000 operating systems call this register @code{%a5}.
3151 Defining such a register variable does not reserve the register; it
3152 remains available for other uses in places where flow control determines
3153 the variable's value is not live. However, these registers are made
3154 unavailable for use in the reload pass; excessive use of this feature
3155 leaves the compiler too few available registers to compile certain
3158 This option does not guarantee that GNU CC will generate code that has
3159 this variable in the register you specify at all times. You may not
3160 code an explicit reference to this register in an @code{asm} statement
3161 and assume it will always refer to this variable.
3163 Stores into local register variables may be deleted when they appear to be dead
3164 according to dataflow analysis. References to local register variables may
3165 be deleted or moved or simplified.
3167 @node Alternate Keywords
3168 @section Alternate Keywords
3169 @cindex alternate keywords
3170 @cindex keywords, alternate
3172 The option @samp{-traditional} disables certain keywords; @samp{-ansi}
3173 disables certain others. This causes trouble when you want to use GNU C
3174 extensions, or ANSI C features, in a general-purpose header file that
3175 should be usable by all programs, including ANSI C programs and traditional
3176 ones. The keywords @code{asm}, @code{typeof} and @code{inline} cannot be
3177 used since they won't work in a program compiled with @samp{-ansi}, while
3178 the keywords @code{const}, @code{volatile}, @code{signed}, @code{typeof}
3179 and @code{inline} won't work in a program compiled with
3180 @samp{-traditional}.@refill
3182 The way to solve these problems is to put @samp{__} at the beginning and
3183 end of each problematical keyword. For example, use @code{__asm__}
3184 instead of @code{asm}, @code{__const__} instead of @code{const}, and
3185 @code{__inline__} instead of @code{inline}.
3187 Other C compilers won't accept these alternative keywords; if you want to
3188 compile with another compiler, you can define the alternate keywords as
3189 macros to replace them with the customary keywords. It looks like this:
3197 @findex __extension__
3198 @samp{-pedantic} and other options cause warnings for many GNU C extensions.
3200 prevent such warnings within one expression by writing
3201 @code{__extension__} before the expression. @code{__extension__} has no
3202 effect aside from this.
3204 @node Incomplete Enums
3205 @section Incomplete @code{enum} Types
3207 You can define an @code{enum} tag without specifying its possible values.
3208 This results in an incomplete type, much like what you get if you write
3209 @code{struct foo} without describing the elements. A later declaration
3210 which does specify the possible values completes the type.
3212 You can't allocate variables or storage using the type while it is
3213 incomplete. However, you can work with pointers to that type.
3215 This extension may not be very useful, but it makes the handling of
3216 @code{enum} more consistent with the way @code{struct} and @code{union}
3219 This extension is not supported by GNU C++.
3221 @node Function Names
3222 @section Function Names as Strings
3224 GNU CC predefines two magic identifiers to hold the name of the current
3225 function. The identifier @code{__FUNCTION__} holds the name of the function
3226 as it appears in the source. The identifier @code{__PRETTY_FUNCTION__}
3227 holds the name of the function pretty printed in a language specific
3230 These names are always the same in a C function, but in a C++ function
3231 they may be different. For example, this program:
3235 extern int printf (char *, ...);
3242 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
3243 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
3261 __PRETTY_FUNCTION__ = int a::sub (int)
3264 The compiler automagically replaces the identifiers with a string
3265 literal containing the appropriate name. Thus, they are neither
3266 preprocessor macros, like @code{__FILE__} and @code{__LINE__}, nor
3267 variables. This means that they catenate with other string literals, and
3268 that they can be used to initialize char arrays. For example
3271 char here[] = "Function " __FUNCTION__ " in " __FILE__;
3274 On the other hand, @samp{#ifdef __FUNCTION__} does not have any special
3275 meaning inside a function, since the preprocessor does not do anything
3276 special with the identifier @code{__FUNCTION__}.
3278 GNU CC also supports the magic word @code{__func__}, defined by the
3282 The identifier @code{__func__} is implicitly declared by the translator
3283 as if, immediately following the opening brace of each function
3284 definition, the declaration
3287 static const char __func__[] = "function-name";
3290 appeared, where function-name is the name of the lexically-enclosing
3291 function. This name is the unadorned name of the function.
3294 By this definition, @code{__func__} is a variable, not a string literal.
3295 In particular, @code{__func__} does not catenate with other string
3298 In @code{C++}, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} are
3299 variables, declared in the same way as @code{__func__}.
3301 @node Return Address
3302 @section Getting the Return or Frame Address of a Function
3304 These functions may be used to get information about the callers of a
3308 @findex __builtin_return_address
3309 @item __builtin_return_address (@var{level})
3310 This function returns the return address of the current function, or of
3311 one of its callers. The @var{level} argument is number of frames to
3312 scan up the call stack. A value of @code{0} yields the return address
3313 of the current function, a value of @code{1} yields the return address
3314 of the caller of the current function, and so forth.
3316 The @var{level} argument must be a constant integer.
3318 On some machines it may be impossible to determine the return address of
3319 any function other than the current one; in such cases, or when the top
3320 of the stack has been reached, this function will return @code{0}.
3322 This function should only be used with a non-zero argument for debugging
3325 @findex __builtin_frame_address
3326 @item __builtin_frame_address (@var{level})
3327 This function is similar to @code{__builtin_return_address}, but it
3328 returns the address of the function frame rather than the return address
3329 of the function. Calling @code{__builtin_frame_address} with a value of
3330 @code{0} yields the frame address of the current function, a value of
3331 @code{1} yields the frame address of the caller of the current function,
3334 The frame is the area on the stack which holds local variables and saved
3335 registers. The frame address is normally the address of the first word
3336 pushed on to the stack by the function. However, the exact definition
3337 depends upon the processor and the calling convention. If the processor
3338 has a dedicated frame pointer register, and the function has a frame,
3339 then @code{__builtin_frame_address} will return the value of the frame
3342 The caveats that apply to @code{__builtin_return_address} apply to this
3346 @node Other Builtins
3347 @section Other built-in functions provided by GNU CC
3348 @cindex builtin functions
3349 @findex __builtin_isgreater
3350 @findex __builtin_isgreaterequal
3351 @findex __builtin_isless
3352 @findex __builtin_islessequal
3353 @findex __builtin_islessgreater
3354 @findex __builtin_isunordered
3408 GNU CC provides a large number of built-in functions other than the ones
3409 mentioned above. Some of these are for internal use in the processing
3410 of exceptions or variable-length argument lists and will not be
3411 documented here because they may change from time to time; we do not
3412 recommend general use of these functions.
3414 The remaining functions are provided for optimization purposes.
3416 GNU CC includes builtin versions of many of the functions in the
3417 standard C library. The versions prefixed with @code{__builtin_} will
3418 always be treated as having the same meaning as the C library function
3419 even if you specify the @samp{-fno-builtin} (@pxref{C Dialect Options})
3420 option. Many of these functions are only optimized in certain cases; if
3421 not optimized in a particular case, a call to the library function will
3424 The functions @code{abort}, @code{exit}, and @code{_exit} are recognized
3425 and presumed not to return, but otherwise are not built in.
3426 @code{_exit} is not recognized in strict ISO C mode (@samp{-ansi},
3427 @samp{-std=c89} or @samp{-std=c99}).
3429 Outside strict ISO C mode, the functions @code{alloca}, @code{bcmp},
3430 @code{bzero}, @code{index}, @code{rindex} and @code{ffs} may be handled
3431 as builtins. Corresponding versions @code{__builtin_alloca},
3432 @code{__builtin_bcmp}, @code{__builtin_bzero}, @code{__builtin_index},
3433 @code{__builtin_rindex} and @code{__builtin_ffs} are also recognized in
3436 The ISO C99 functions @code{conj}, @code{conjf}, @code{conjl},
3437 @code{creal}, @code{crealf}, @code{creall}, @code{cimag}, @code{cimagf},
3438 @code{cimagl}, @code{llabs} and @code{imaxabs} are handled as builtins
3439 except in strict ISO C89 mode. There are also builtin versions of the ISO C99
3440 functions @code{cosf}, @code{cosl}, @code{fabsf}, @code{fabsl},
3441 @code{sinf}, @code{sinl}, @code{sqrtf}, and @code{sqrtl}, that are
3442 recognized in any mode since ISO C89 reserves these names for the
3443 purpose to which ISO C99 puts them. All these functions have
3444 corresponding versions prefixed with @code{__builtin_}.
3446 The following ISO C89 functions are recognized as builtins unless
3447 @samp{-fno-builtin} is specified: @code{abs}, @code{cos}, @code{fabs},
3448 @code{fputs}, @code{labs}, @code{memcmp}, @code{memcpy}, @code{memset},
3449 @code{printf}, @code{sin}, @code{sqrt}, @code{strcat}, @code{strchr},
3450 @code{strcmp}, @code{strcpy}, @code{strcspn}, @code{strlen},
3451 @code{strncat}, @code{strncmp}, @code{strncpy}, @code{strpbrk},
3452 @code{strrchr}, @code{strspn}, and @code{strstr}. All of these
3453 functions have corresponding versions prefixed with @code{__builtin_},
3454 except that the version for @code{sqrt} is called
3455 @code{__builtin_fsqrt}.
3457 GNU CC provides builtin versions of the ISO C99 floating point
3458 comparison macros (that avoid raising exceptions for unordered
3459 operands): @code{__builtin_isgreater}, @code{__builtin_isgreaterequal},
3460 @code{__builtin_isless}, @code{__builtin_islessequal},
3461 @code{__builtin_islessgreater}, and @code{__builtin_isunordered}.
3465 @findex __builtin_constant_p
3466 @item __builtin_constant_p (@var{exp})
3467 You can use the builtin function @code{__builtin_constant_p} to
3468 determine if a value is known to be constant at compile-time and hence
3469 that GNU CC can perform constant-folding on expressions involving that
3470 value. The argument of the function is the value to test. The function
3471 returns the integer 1 if the argument is known to be a compile-time
3472 constant and 0 if it is not known to be a compile-time constant. A
3473 return of 0 does not indicate that the value is @emph{not} a constant,
3474 but merely that GNU CC cannot prove it is a constant with the specified
3475 value of the @samp{-O} option.
3477 You would typically use this function in an embedded application where
3478 memory was a critical resource. If you have some complex calculation,
3479 you may want it to be folded if it involves constants, but need to call
3480 a function if it does not. For example:
3483 #define Scale_Value(X) \
3484 (__builtin_constant_p (X) ? ((X) * SCALE + OFFSET) : Scale (X))
3487 You may use this builtin function in either a macro or an inline
3488 function. However, if you use it in an inlined function and pass an
3489 argument of the function as the argument to the builtin, GNU CC will
3490 never return 1 when you call the inline function with a string constant
3491 or constructor expression (@pxref{Constructors}) and will not return 1
3492 when you pass a constant numeric value to the inline function unless you
3493 specify the @samp{-O} option.
3495 @findex __builtin_expect
3496 @item __builtin_expect(@var{exp}, @var{c})
3497 You may use @code{__builtin_expect} to provide the compiler with
3498 branch prediction information. In general, you should prefer to
3499 use actual profile feedback for this (@samp{-fprofile-arcs}), as
3500 programmers are notoriously bad at predicting how their programs
3501 actually perform. However, there are applications in which this
3502 data is hard to collect.
3504 The return value is the value of @var{exp}, which should be an
3505 integral expression. The value of @var{c} must be a compile-time
3506 constant. The semantics of the builtin are that it is expected
3507 that @var{exp} == @var{c}. For example:
3510 if (__builtin_expect (x, 0))
3515 would indicate that we do not expect to call @code{foo}, since
3516 we expect @code{x} to be zero. Since you are limited to integral
3517 expressions for @var{exp}, you should use constructions such as
3520 if (__builtin_expect (ptr != NULL, 1))
3525 when testing pointer or floating-point values.
3528 @node Deprecated Features
3529 @section Deprecated Features
3531 In the past, the GNU C++ compiler was extended to experiment with new
3532 features, at a time when the C++ language was still evolving. Now that
3533 the C++ standard is complete, some of those features are superseded by
3534 superior alternatives. Using the old features might cause a warning in
3535 some cases that the feature will be dropped in the future. In other
3536 cases, the feature might be gone already.
3538 While the list below is not exhaustive, it documents some of the options
3539 that are now deprecated:
3542 @item -fexternal-templates
3543 @itemx -falt-external-templates
3544 These are two of the many ways for g++ to implement template
3545 instantiation. @xref{Template Instantiation}. The C++ standard clearly
3546 defines how template definitions have to be organized across
3547 implementation units. g++ has an implicit instantiation mechanism that
3548 should work just fine for standard-conforming code.
3550 @item -fstrict-prototype
3551 @itemx -fno-strict-prototype
3552 Previously it was possible to use an empty prototype parameter list to
3553 indicate an unspecified number of parameters (like C), rather than no
3554 parameters, as C++ demands. This feature has been removed, except where
3555 it is required for backwards compatibility @xref{Backwards Compatibility}.
3558 The named return value extension has been deprecated, and will be
3559 removed from g++ at some point.
3561 @node Backwards Compatibility
3562 @section Backwards Compatibility
3563 @cindex Backwards Compatibility
3566 Now that there is a definitive ISO standard C++, g++ has a specification
3567 to adhere to. The C++ language evolved over time, and features that used
3568 to be acceptable in previous drafts of the standard, such as the ARM, are
3569 no longer accepted. In order to allow compilation of C++ written to such
3570 drafts, g++ contains some backwards compatibilities. @emph{All such
3571 backwards compatibility features are liable to disappear in future
3572 versions of g++.} They should be considered deprecated @xref{Deprecated
3577 If a variable is declared at for scope, it used to remain in scope until
3578 the end of the scope which contained the for statement (rather than just
3579 within the for scope). g++ retains this, but issues a warning, if such a
3580 variable is accessed outside the for scope.
3582 @item implicit C language
3583 Old C system header files did not contain an @code{extern "C" @{...@}}
3584 scope to set the language. On such systems, all header files are
3585 implicitly scoped inside a C language scope. Also, an empty prototype
3586 @code{()} will be treated as an unspecified number of arguments, rather
3587 than no arguments, as C++ demands.
3590 @node C++ Extensions
3591 @chapter Extensions to the C++ Language
3592 @cindex extensions, C++ language
3593 @cindex C++ language extensions
3595 The GNU compiler provides these extensions to the C++ language (and you
3596 can also use most of the C language extensions in your C++ programs). If you
3597 want to write code that checks whether these features are available, you can
3598 test for the GNU compiler the same way as for C programs: check for a
3599 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
3600 test specifically for GNU C++ (@pxref{Standard Predefined,,Standard
3601 Predefined Macros,cpp.info,The C Preprocessor}).
3604 * Min and Max:: C++ Minimum and maximum operators.
3605 * Volatiles:: What constitutes an access to a volatile object.
3606 * Restricted Pointers:: C99 restricted pointers and references.
3607 * C++ Interface:: You can use a single C++ header file for both
3608 declarations and definitions.
3609 * Template Instantiation:: Methods for ensuring that exactly one copy of
3610 each needed template instantiation is emitted.
3611 * Bound member functions:: You can extract a function pointer to the
3612 method denoted by a @samp{->*} or @samp{.*} expression.
3613 * C++ Attributes:: Variable, function, and type attributes for C++ only.
3617 @section Minimum and Maximum Operators in C++
3619 It is very convenient to have operators which return the ``minimum'' or the
3620 ``maximum'' of two arguments. In GNU C++ (but not in GNU C),
3623 @item @var{a} <? @var{b}
3625 @cindex minimum operator
3626 is the @dfn{minimum}, returning the smaller of the numeric values
3627 @var{a} and @var{b};
3629 @item @var{a} >? @var{b}
3631 @cindex maximum operator
3632 is the @dfn{maximum}, returning the larger of the numeric values @var{a}
3636 These operations are not primitive in ordinary C++, since you can
3637 use a macro to return the minimum of two things in C++, as in the
3641 #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
3645 You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
3646 the minimum value of variables @var{i} and @var{j}.
3648 However, side effects in @code{X} or @code{Y} may cause unintended
3649 behavior. For example, @code{MIN (i++, j++)} will fail, incrementing
3650 the smaller counter twice. A GNU C extension allows you to write safe
3651 macros that avoid this kind of problem (@pxref{Naming Types,,Naming an
3652 Expression's Type}). However, writing @code{MIN} and @code{MAX} as
3653 macros also forces you to use function-call notation for a
3654 fundamental arithmetic operation. Using GNU C++ extensions, you can
3655 write @w{@samp{int min = i <? j;}} instead.
3657 Since @code{<?} and @code{>?} are built into the compiler, they properly
3658 handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}}
3662 @section When is a Volatile Object Accessed?
3663 @cindex accessing volatiles
3664 @cindex volatile read
3665 @cindex volatile write
3666 @cindex volatile access
3668 Both the C and C++ standard have the concept of volatile objects. These
3669 are normally accessed by pointers and used for accessing hardware. The
3670 standards encourage compilers to refrain from optimizations
3671 concerning accesses to volatile objects that it might perform on
3672 non-volatile objects. The C standard leaves it implementation defined
3673 as to what constitutes a volatile access. The C++ standard omits to
3674 specify this, except to say that C++ should behave in a similar manner
3675 to C with respect to volatiles, where possible. The minimum either
3676 standard specifies is that at a sequence point all previous accesses to
3677 volatile objects have stabilized and no subsequent accesses have
3678 occurred. Thus an implementation is free to reorder and combine
3679 volatile accesses which occur between sequence points, but cannot do so
3680 for accesses across a sequence point. The use of volatiles does not
3681 allow you to violate the restriction on updating objects multiple times
3682 within a sequence point.
3684 In most expressions, it is intuitively obvious what is a read and what is
3685 a write. For instance
3688 volatile int *dst = <somevalue>;
3689 volatile int *src = <someothervalue>;
3694 will cause a read of the volatile object pointed to by @var{src} and stores the
3695 value into the volatile object pointed to by @var{dst}. There is no
3696 guarantee that these reads and writes are atomic, especially for objects
3697 larger than @code{int}.
3699 Less obvious expressions are where something which looks like an access
3700 is used in a void context. An example would be,
3703 volatile int *src = <somevalue>;
3707 With C, such expressions are rvalues, and as rvalues cause a read of
3708 the object, gcc interprets this as a read of the volatile being pointed
3709 to. The C++ standard specifies that such expressions do not undergo
3710 lvalue to rvalue conversion, and that the type of the dereferenced
3711 object may be incomplete. The C++ standard does not specify explicitly
3712 that it is this lvalue to rvalue conversion which is responsible for
3713 causing an access. However, there is reason to believe that it is,
3714 because otherwise certain simple expressions become undefined. However,
3715 because it would surprise most programmers, g++ treats dereferencing a
3716 pointer to volatile object of complete type in a void context as a read
3717 of the object. When the object has incomplete type, g++ issues a
3722 struct T @{int m;@};
3723 volatile S *ptr1 = <somevalue>;
3724 volatile T *ptr2 = <somevalue>;
3729 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
3730 causes a read of the object pointed to. If you wish to force an error on
3731 the first case, you must force a conversion to rvalue with, for instance
3732 a static cast, @code{static_cast<S>(*ptr1)}.
3734 When using a reference to volatile, g++ does not treat equivalent
3735 expressions as accesses to volatiles, but instead issues a warning that
3736 no volatile is accessed. The rationale for this is that otherwise it
3737 becomes difficult to determine where volatile access occur, and not
3738 possible to ignore the return value from functions returning volatile
3739 references. Again, if you wish to force a read, cast the reference to
3742 @node Restricted Pointers
3743 @section Restricting Pointer Aliasing
3744 @cindex restricted pointers
3745 @cindex restricted references
3746 @cindex restricted this pointer
3748 As with gcc, g++ understands the C99 feature of restricted pointers,
3749 specified with the @code{__restrict__}, or @code{__restrict} type
3750 qualifier. Because you cannot compile C++ by specifying the -std=c99
3751 language flag, @code{restrict} is not a keyword in C++.
3753 In addition to allowing restricted pointers, you can specify restricted
3754 references, which indicate that the reference is not aliased in the local
3758 void fn (int *__restrict__ rptr, int &__restrict__ rref)
3765 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
3766 @var{rref} refers to a (different) unaliased integer.
3768 You may also specify whether a member function's @var{this} pointer is
3769 unaliased by using @code{__restrict__} as a member function qualifier.
3772 void T::fn () __restrict__
3779 Within the body of @code{T::fn}, @var{this} will have the effective
3780 definition @code{T *__restrict__ const this}. Notice that the
3781 interpretation of a @code{__restrict__} member function qualifier is
3782 different to that of @code{const} or @code{volatile} qualifier, in that it
3783 is applied to the pointer rather than the object. This is consistent with
3784 other compilers which implement restricted pointers.
3786 As with all outermost parameter qualifiers, @code{__restrict__} is
3787 ignored in function definition matching. This means you only need to
3788 specify @code{__restrict__} in a function definition, rather than
3789 in a function prototype as well.
3792 @section Declarations and Definitions in One Header
3794 @cindex interface and implementation headers, C++
3795 @cindex C++ interface and implementation headers
3796 C++ object definitions can be quite complex. In principle, your source
3797 code will need two kinds of things for each object that you use across
3798 more than one source file. First, you need an @dfn{interface}
3799 specification, describing its structure with type declarations and
3800 function prototypes. Second, you need the @dfn{implementation} itself.
3801 It can be tedious to maintain a separate interface description in a
3802 header file, in parallel to the actual implementation. It is also
3803 dangerous, since separate interface and implementation definitions may
3804 not remain parallel.
3806 @cindex pragmas, interface and implementation
3807 With GNU C++, you can use a single header file for both purposes.
3810 @emph{Warning:} The mechanism to specify this is in transition. For the
3811 nonce, you must use one of two @code{#pragma} commands; in a future
3812 release of GNU C++, an alternative mechanism will make these
3813 @code{#pragma} commands unnecessary.
3816 The header file contains the full definitions, but is marked with
3817 @samp{#pragma interface} in the source code. This allows the compiler
3818 to use the header file only as an interface specification when ordinary
3819 source files incorporate it with @code{#include}. In the single source
3820 file where the full implementation belongs, you can use either a naming
3821 convention or @samp{#pragma implementation} to indicate this alternate
3822 use of the header file.
3825 @item #pragma interface
3826 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
3827 @kindex #pragma interface
3828 Use this directive in @emph{header files} that define object classes, to save
3829 space in most of the object files that use those classes. Normally,
3830 local copies of certain information (backup copies of inline member
3831 functions, debugging information, and the internal tables that implement
3832 virtual functions) must be kept in each object file that includes class
3833 definitions. You can use this pragma to avoid such duplication. When a
3834 header file containing @samp{#pragma interface} is included in a
3835 compilation, this auxiliary information will not be generated (unless
3836 the main input source file itself uses @samp{#pragma implementation}).
3837 Instead, the object files will contain references to be resolved at link
3840 The second form of this directive is useful for the case where you have
3841 multiple headers with the same name in different directories. If you
3842 use this form, you must specify the same string to @samp{#pragma
3845 @item #pragma implementation
3846 @itemx #pragma implementation "@var{objects}.h"
3847 @kindex #pragma implementation
3848 Use this pragma in a @emph{main input file}, when you want full output from
3849 included header files to be generated (and made globally visible). The
3850 included header file, in turn, should use @samp{#pragma interface}.
3851 Backup copies of inline member functions, debugging information, and the
3852 internal tables used to implement virtual functions are all generated in
3853 implementation files.
3855 @cindex implied @code{#pragma implementation}
3856 @cindex @code{#pragma implementation}, implied
3857 @cindex naming convention, implementation headers
3858 If you use @samp{#pragma implementation} with no argument, it applies to
3859 an include file with the same basename@footnote{A file's @dfn{basename}
3860 was the name stripped of all leading path information and of trailing
3861 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
3862 file. For example, in @file{allclass.cc}, giving just
3863 @samp{#pragma implementation}
3864 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
3866 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
3867 an implementation file whenever you would include it from
3868 @file{allclass.cc} even if you never specified @samp{#pragma
3869 implementation}. This was deemed to be more trouble than it was worth,
3870 however, and disabled.
3872 If you use an explicit @samp{#pragma implementation}, it must appear in
3873 your source file @emph{before} you include the affected header files.
3875 Use the string argument if you want a single implementation file to
3876 include code from multiple header files. (You must also use
3877 @samp{#include} to include the header file; @samp{#pragma
3878 implementation} only specifies how to use the file---it doesn't actually
3881 There is no way to split up the contents of a single header file into
3882 multiple implementation files.
3885 @cindex inlining and C++ pragmas
3886 @cindex C++ pragmas, effect on inlining
3887 @cindex pragmas in C++, effect on inlining
3888 @samp{#pragma implementation} and @samp{#pragma interface} also have an
3889 effect on function inlining.
3891 If you define a class in a header file marked with @samp{#pragma
3892 interface}, the effect on a function defined in that class is similar to
3893 an explicit @code{extern} declaration---the compiler emits no code at
3894 all to define an independent version of the function. Its definition
3895 is used only for inlining with its callers.
3897 Conversely, when you include the same header file in a main source file
3898 that declares it as @samp{#pragma implementation}, the compiler emits
3899 code for the function itself; this defines a version of the function
3900 that can be found via pointers (or by callers compiled without
3901 inlining). If all calls to the function can be inlined, you can avoid
3902 emitting the function by compiling with @samp{-fno-implement-inlines}.
3903 If any calls were not inlined, you will get linker errors.
3905 @node Template Instantiation
3906 @section Where's the Template?
3908 @cindex template instantiation
3910 C++ templates are the first language feature to require more
3911 intelligence from the environment than one usually finds on a UNIX
3912 system. Somehow the compiler and linker have to make sure that each
3913 template instance occurs exactly once in the executable if it is needed,
3914 and not at all otherwise. There are two basic approaches to this
3915 problem, which I will refer to as the Borland model and the Cfront model.
3919 Borland C++ solved the template instantiation problem by adding the code
3920 equivalent of common blocks to their linker; the compiler emits template
3921 instances in each translation unit that uses them, and the linker
3922 collapses them together. The advantage of this model is that the linker
3923 only has to consider the object files themselves; there is no external
3924 complexity to worry about. This disadvantage is that compilation time
3925 is increased because the template code is being compiled repeatedly.
3926 Code written for this model tends to include definitions of all
3927 templates in the header file, since they must be seen to be
3931 The AT&T C++ translator, Cfront, solved the template instantiation
3932 problem by creating the notion of a template repository, an
3933 automatically maintained place where template instances are stored. A
3934 more modern version of the repository works as follows: As individual
3935 object files are built, the compiler places any template definitions and
3936 instantiations encountered in the repository. At link time, the link
3937 wrapper adds in the objects in the repository and compiles any needed
3938 instances that were not previously emitted. The advantages of this
3939 model are more optimal compilation speed and the ability to use the
3940 system linker; to implement the Borland model a compiler vendor also
3941 needs to replace the linker. The disadvantages are vastly increased
3942 complexity, and thus potential for error; for some code this can be
3943 just as transparent, but in practice it can been very difficult to build
3944 multiple programs in one directory and one program in multiple
3945 directories. Code written for this model tends to separate definitions
3946 of non-inline member templates into a separate file, which should be
3947 compiled separately.
3950 When used with GNU ld version 2.8 or later on an ELF system such as
3951 Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the
3952 Borland model. On other systems, g++ implements neither automatic
3955 A future version of g++ will support a hybrid model whereby the compiler
3956 will emit any instantiations for which the template definition is
3957 included in the compile, and store template definitions and
3958 instantiation context information into the object file for the rest.
3959 The link wrapper will extract that information as necessary and invoke
3960 the compiler to produce the remaining instantiations. The linker will
3961 then combine duplicate instantiations.
3963 In the mean time, you have the following options for dealing with
3964 template instantiations:
3968 Compile your template-using code with @samp{-frepo}. The compiler will
3969 generate files with the extension @samp{.rpo} listing all of the
3970 template instantiations used in the corresponding object files which
3971 could be instantiated there; the link wrapper, @samp{collect2}, will
3972 then update the @samp{.rpo} files to tell the compiler where to place
3973 those instantiations and rebuild any affected object files. The
3974 link-time overhead is negligible after the first pass, as the compiler
3975 will continue to place the instantiations in the same files.
3977 This is your best option for application code written for the Borland
3978 model, as it will just work. Code written for the Cfront model will
3979 need to be modified so that the template definitions are available at
3980 one or more points of instantiation; usually this is as simple as adding
3981 @code{#include <tmethods.cc>} to the end of each template header.
3983 For library code, if you want the library to provide all of the template
3984 instantiations it needs, just try to link all of its object files
3985 together; the link will fail, but cause the instantiations to be
3986 generated as a side effect. Be warned, however, that this may cause
3987 conflicts if multiple libraries try to provide the same instantiations.
3988 For greater control, use explicit instantiation as described in the next
3992 Compile your code with @samp{-fno-implicit-templates} to disable the
3993 implicit generation of template instances, and explicitly instantiate
3994 all the ones you use. This approach requires more knowledge of exactly
3995 which instances you need than do the others, but it's less
3996 mysterious and allows greater control. You can scatter the explicit
3997 instantiations throughout your program, perhaps putting them in the
3998 translation units where the instances are used or the translation units
3999 that define the templates themselves; you can put all of the explicit
4000 instantiations you need into one big file; or you can create small files
4007 template class Foo<int>;
4008 template ostream& operator <<
4009 (ostream&, const Foo<int>&);
4012 for each of the instances you need, and create a template instantiation
4015 If you are using Cfront-model code, you can probably get away with not
4016 using @samp{-fno-implicit-templates} when compiling files that don't
4017 @samp{#include} the member template definitions.
4019 If you use one big file to do the instantiations, you may want to
4020 compile it without @samp{-fno-implicit-templates} so you get all of the
4021 instances required by your explicit instantiations (but not by any
4022 other files) without having to specify them as well.
4024 g++ has extended the template instantiation syntax outlined in the
4025 Working Paper to allow forward declaration of explicit instantiations
4026 (with @code{extern}), instantiation of the compiler support data for a
4027 template class (i.e. the vtable) without instantiating any of its
4028 members (with @code{inline}), and instantiation of only the static data
4029 members of a template class, without the support data or member
4030 functions (with (@code{static}):
4033 extern template int max (int, int);
4034 inline template class Foo<int>;
4035 static template class Foo<int>;
4039 Do nothing. Pretend g++ does implement automatic instantiation
4040 management. Code written for the Borland model will work fine, but
4041 each translation unit will contain instances of each of the templates it
4042 uses. In a large program, this can lead to an unacceptable amount of code
4046 Add @samp{#pragma interface} to all files containing template
4047 definitions. For each of these files, add @samp{#pragma implementation
4048 "@var{filename}"} to the top of some @samp{.C} file which
4049 @samp{#include}s it. Then compile everything with
4050 @samp{-fexternal-templates}. The templates will then only be expanded
4051 in the translation unit which implements them (i.e. has a @samp{#pragma
4052 implementation} line for the file where they live); all other files will
4053 use external references. If you're lucky, everything should work
4054 properly. If you get undefined symbol errors, you need to make sure
4055 that each template instance which is used in the program is used in the
4056 file which implements that template. If you don't have any use for a
4057 particular instance in that file, you can just instantiate it
4058 explicitly, using the syntax from the latest C++ working paper:
4061 template class A<int>;
4062 template ostream& operator << (ostream&, const A<int>&);
4065 This strategy will work with code written for either model. If you are
4066 using code written for the Cfront model, the file containing a class
4067 template and the file containing its member templates should be
4068 implemented in the same translation unit.
4070 A slight variation on this approach is to instead use the flag
4071 @samp{-falt-external-templates}; this flag causes template
4072 instances to be emitted in the translation unit that implements the
4073 header where they are first instantiated, rather than the one which
4074 implements the file where the templates are defined. This header must
4075 be the same in all translation units, or things are likely to break.
4077 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
4078 more discussion of these pragmas.
4081 @node Bound member functions
4082 @section Extracting the function pointer from a bound pointer to member function
4085 @cindex pointer to member function
4086 @cindex bound pointer to member function
4088 In C++, pointer to member functions (PMFs) are implemented using a wide
4089 pointer of sorts to handle all the possible call mechanisms; the PMF
4090 needs to store information about how to adjust the @samp{this} pointer,
4091 and if the function pointed to is virtual, where to find the vtable, and
4092 where in the vtable to look for the member function. If you are using
4093 PMFs in an inner loop, you should really reconsider that decision. If
4094 that is not an option, you can extract the pointer to the function that
4095 would be called for a given object/PMF pair and call it directly inside
4096 the inner loop, to save a bit of time.
4098 Note that you will still be paying the penalty for the call through a
4099 function pointer; on most modern architectures, such a call defeats the
4100 branch prediction features of the CPU. This is also true of normal
4101 virtual function calls.
4103 The syntax for this extension is
4107 extern int (A::*fp)();
4108 typedef int (*fptr)(A *);
4110 fptr p = (fptr)(a.*fp);
4113 For PMF constants (i.e. expressions of the form @samp{&Klasse::Member}),
4114 no object is needed to obtain the address of the function. They can be
4115 converted to function pointers directly:
4118 fptr p1 = (fptr)(&A::foo);
4121 You must specify @samp{-Wno-pmf-conversions} to use this extension.
4123 @node C++ Attributes
4124 @section C++-Specific Variable, Function, and Type Attributes
4126 Some attributes only make sense for C++ programs.
4129 @item init_priority (@var{priority})
4130 @cindex init_priority attribute
4133 In Standard C++, objects defined at namespace scope are guaranteed to be
4134 initialized in an order in strict accordance with that of their definitions
4135 @emph{in a given translation unit}. No guarantee is made for initializations
4136 across translation units. However, GNU C++ allows users to control the
4137 order of initialization of objects defined at namespace socpe with the
4138 @code{init_priority} attribute by specifying a relative @var{priority},
4139 a constant integral expression currently bounded between 101 and 65535
4140 inclusive. Lower numbers indicate a higher priority.
4142 In the following example, @code{A} would normally be created before
4143 @code{B}, but the @code{init_priority} attribute has reversed that order:
4146 Some_Class A __attribute__ ((init_priority (2000)));
4147 Some_Class B __attribute__ ((init_priority (543)));
4151 Note that the particular values of @var{priority} do not matter; only their
4156 @cindex com_interface attribute
4158 @c This is based on: 1) grepping the code,
4159 @c 2) http://gcc.gnu.org/ml/gcc-bugs/1999-08n/msg01212.html
4160 @c 3) http://gcc.gnu.org/ml/gcc-bugs/1999-08n/msg01215.html
4161 @c and 4) a lot of guesswork. You can tell I don't use COM. -pme 21Dec00
4163 This type attribute takes no parameters, and marks a class or struct as an
4164 interface for communication via COM; the class will support the COM ABI
4165 rather than the full C++ ABI. Currently this means that RTTI is not possible
4166 with the resulting class heirarchy. The virtual pointer table will be
4167 changed to be COM-compliant. Also, all classes and structs derived from one
4168 marked with this attribute are implicitly marked with the same attribute;
4169 thus, only the base class in a COM hierarchy needs @code{com_interface}.