]>
Commit | Line | Data |
---|---|---|
d60e5448 | 1 | @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001 Free Software Foundation, Inc. |
c1f7febf RK |
2 | @c This is part of the GCC manual. |
3 | @c For copying conditions, see the file gcc.texi. | |
4 | ||
5 | @node C Extensions | |
6 | @chapter Extensions to the C Language Family | |
7 | @cindex extensions, C language | |
8 | @cindex C language extensions | |
9 | ||
84330467 | 10 | @opindex pedantic |
5490d604 | 11 | GNU C provides several language features not found in ISO standard C. |
f0523f02 | 12 | (The @option{-pedantic} option directs GCC to print a warning message if |
c1f7febf RK |
13 | any of these features is used.) To test for the availability of these |
14 | features in conditional compilation, check for a predefined macro | |
f0523f02 | 15 | @code{__GNUC__}, which is always defined under GCC. |
c1f7febf | 16 | |
2147b154 | 17 | These extensions are available in C and Objective-C. Most of them are |
c1f7febf RK |
18 | also available in C++. @xref{C++ Extensions,,Extensions to the |
19 | C++ Language}, for extensions that apply @emph{only} to C++. | |
20 | ||
4b404517 JM |
21 | Some features that are in ISO C99 but not C89 or C++ are also, as |
22 | extensions, accepted by GCC in C89 mode and in C++. | |
5490d604 | 23 | |
c1f7febf RK |
24 | @c The only difference between the two versions of this menu is that the |
25 | @c version for clear INTERNALS has an extra node, "Constraints" (which | |
26 | @c appears in a separate chapter in the other version of the manual). | |
27 | @ifset INTERNALS | |
28 | @menu | |
29 | * Statement Exprs:: Putting statements and declarations inside expressions. | |
30 | * Local Labels:: Labels local to a statement-expression. | |
31 | * Labels as Values:: Getting pointers to labels, and computed gotos. | |
32 | * Nested Functions:: As in Algol and Pascal, lexical scoping of functions. | |
33 | * Constructing Calls:: Dispatching a call to another function. | |
34 | * Naming Types:: Giving a name to the type of some expression. | |
35 | * Typeof:: @code{typeof}: referring to the type of an expression. | |
36 | * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues. | |
37 | * Conditionals:: Omitting the middle operand of a @samp{?:} expression. | |
38 | * Long Long:: Double-word integers---@code{long long int}. | |
39 | * Complex:: Data types for complex numbers. | |
6f4d7222 | 40 | * Hex Floats:: Hexadecimal floating-point constants. |
c1f7febf RK |
41 | * Zero Length:: Zero-length arrays. |
42 | * Variable Length:: Arrays whose length is computed at run time. | |
ccd96f0a NB |
43 | * Variadic Macros:: Macros with a variable number of arguments. |
44 | * Escaped Newlines:: Slightly looser rules for escaped newlines. | |
45 | * Multi-line Strings:: String literals with embedded newlines. | |
c1f7febf RK |
46 | * Subscripting:: Any array can be subscripted, even if not an lvalue. |
47 | * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers. | |
48 | * Initializers:: Non-constant initializers. | |
4b404517 | 49 | * Compound Literals:: Compound literals give structures, unions |
c1f7febf | 50 | or arrays as values. |
4b404517 | 51 | * Designated Inits:: Labeling elements of initializers. |
c1f7febf RK |
52 | * Cast to Union:: Casting to union type from any member of the union. |
53 | * Case Ranges:: `case 1 ... 9' and such. | |
4b404517 | 54 | * Mixed Declarations:: Mixing declarations and code. |
c1f7febf RK |
55 | * Function Attributes:: Declaring that functions have no side effects, |
56 | or that they can never return. | |
2c5e91d2 | 57 | * Attribute Syntax:: Formal syntax for attributes. |
c1f7febf RK |
58 | * Function Prototypes:: Prototype declarations and old-style definitions. |
59 | * C++ Comments:: C++ comments are recognized. | |
60 | * Dollar Signs:: Dollar sign is allowed in identifiers. | |
61 | * Character Escapes:: @samp{\e} stands for the character @key{ESC}. | |
62 | * Variable Attributes:: Specifying attributes of variables. | |
63 | * Type Attributes:: Specifying attributes of types. | |
64 | * Alignment:: Inquiring about the alignment of a type or variable. | |
65 | * Inline:: Defining inline functions (as fast as macros). | |
66 | * Extended Asm:: Assembler instructions with C expressions as operands. | |
67 | (With them you can define ``built-in'' functions.) | |
68 | * Asm Labels:: Specifying the assembler name to use for a C symbol. | |
69 | * Explicit Reg Vars:: Defining variables residing in specified registers. | |
70 | * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files. | |
71 | * Incomplete Enums:: @code{enum foo;}, with details to follow. | |
72 | * Function Names:: Printable strings which are the name of the current | |
73 | function. | |
74 | * Return Address:: Getting the return or frame address of a function. | |
185ebd6c | 75 | * Other Builtins:: Other built-in functions. |
c1f7febf RK |
76 | @end menu |
77 | @end ifset | |
78 | @ifclear INTERNALS | |
79 | @menu | |
80 | * Statement Exprs:: Putting statements and declarations inside expressions. | |
81 | * Local Labels:: Labels local to a statement-expression. | |
82 | * Labels as Values:: Getting pointers to labels, and computed gotos. | |
83 | * Nested Functions:: As in Algol and Pascal, lexical scoping of functions. | |
84 | * Constructing Calls:: Dispatching a call to another function. | |
85 | * Naming Types:: Giving a name to the type of some expression. | |
86 | * Typeof:: @code{typeof}: referring to the type of an expression. | |
87 | * Lvalues:: Using @samp{?:}, @samp{,} and casts in lvalues. | |
88 | * Conditionals:: Omitting the middle operand of a @samp{?:} expression. | |
89 | * Long Long:: Double-word integers---@code{long long int}. | |
90 | * Complex:: Data types for complex numbers. | |
6f4d7222 | 91 | * Hex Floats:: Hexadecimal floating-point constants. |
c1f7febf RK |
92 | * Zero Length:: Zero-length arrays. |
93 | * Variable Length:: Arrays whose length is computed at run time. | |
ccd96f0a NB |
94 | * Variadic Macros:: Macros with a variable number of arguments. |
95 | * Escaped Newlines:: Slightly looser rules for escaped newlines. | |
96 | * Multi-line Strings:: String literals with embedded newlines. | |
c1f7febf RK |
97 | * Subscripting:: Any array can be subscripted, even if not an lvalue. |
98 | * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers. | |
99 | * Initializers:: Non-constant initializers. | |
4b404517 | 100 | * Compound Literals:: Compound literals give structures, unions |
c1f7febf | 101 | or arrays as values. |
4b404517 | 102 | * Designated Inits:: Labeling elements of initializers. |
c1f7febf RK |
103 | * Cast to Union:: Casting to union type from any member of the union. |
104 | * Case Ranges:: `case 1 ... 9' and such. | |
4b404517 | 105 | * Mixed Declarations:: Mixing declarations and code. |
c1f7febf RK |
106 | * Function Attributes:: Declaring that functions have no side effects, |
107 | or that they can never return. | |
2c5e91d2 | 108 | * Attribute Syntax:: Formal syntax for attributes. |
c1f7febf RK |
109 | * Function Prototypes:: Prototype declarations and old-style definitions. |
110 | * C++ Comments:: C++ comments are recognized. | |
111 | * Dollar Signs:: Dollar sign is allowed in identifiers. | |
112 | * Character Escapes:: @samp{\e} stands for the character @key{ESC}. | |
113 | * Variable Attributes:: Specifying attributes of variables. | |
114 | * Type Attributes:: Specifying attributes of types. | |
115 | * Alignment:: Inquiring about the alignment of a type or variable. | |
116 | * Inline:: Defining inline functions (as fast as macros). | |
117 | * Extended Asm:: Assembler instructions with C expressions as operands. | |
118 | (With them you can define ``built-in'' functions.) | |
119 | * Constraints:: Constraints for asm operands | |
120 | * Asm Labels:: Specifying the assembler name to use for a C symbol. | |
121 | * Explicit Reg Vars:: Defining variables residing in specified registers. | |
122 | * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files. | |
123 | * Incomplete Enums:: @code{enum foo;}, with details to follow. | |
124 | * Function Names:: Printable strings which are the name of the current | |
125 | function. | |
126 | * Return Address:: Getting the return or frame address of a function. | |
c5c76735 | 127 | * Other Builtins:: Other built-in functions. |
c1f7febf RK |
128 | @end menu |
129 | @end ifclear | |
130 | ||
131 | @node Statement Exprs | |
132 | @section Statements and Declarations in Expressions | |
133 | @cindex statements inside expressions | |
134 | @cindex declarations inside expressions | |
135 | @cindex expressions containing statements | |
136 | @cindex macros, statements in expressions | |
137 | ||
138 | @c the above section title wrapped and causes an underfull hbox.. i | |
139 | @c changed it from "within" to "in". --mew 4feb93 | |
140 | ||
141 | A compound statement enclosed in parentheses may appear as an expression | |
142 | in GNU C. This allows you to use loops, switches, and local variables | |
143 | within an expression. | |
144 | ||
145 | Recall that a compound statement is a sequence of statements surrounded | |
146 | by braces; in this construct, parentheses go around the braces. For | |
147 | example: | |
148 | ||
149 | @example | |
150 | (@{ int y = foo (); int z; | |
151 | if (y > 0) z = y; | |
152 | else z = - y; | |
153 | z; @}) | |
154 | @end example | |
155 | ||
156 | @noindent | |
157 | is a valid (though slightly more complex than necessary) expression | |
158 | for the absolute value of @code{foo ()}. | |
159 | ||
160 | The last thing in the compound statement should be an expression | |
161 | followed by a semicolon; the value of this subexpression serves as the | |
162 | value of the entire construct. (If you use some other kind of statement | |
163 | last within the braces, the construct has type @code{void}, and thus | |
164 | effectively no value.) | |
165 | ||
166 | This feature is especially useful in making macro definitions ``safe'' (so | |
167 | that they evaluate each operand exactly once). For example, the | |
168 | ``maximum'' function is commonly defined as a macro in standard C as | |
169 | follows: | |
170 | ||
171 | @example | |
172 | #define max(a,b) ((a) > (b) ? (a) : (b)) | |
173 | @end example | |
174 | ||
175 | @noindent | |
176 | @cindex side effects, macro argument | |
177 | But this definition computes either @var{a} or @var{b} twice, with bad | |
178 | results if the operand has side effects. In GNU C, if you know the | |
179 | type of the operands (here let's assume @code{int}), you can define | |
180 | the macro safely as follows: | |
181 | ||
182 | @example | |
183 | #define maxint(a,b) \ | |
184 | (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @}) | |
185 | @end example | |
186 | ||
187 | Embedded statements are not allowed in constant expressions, such as | |
c771326b | 188 | the value of an enumeration constant, the width of a bit-field, or |
c1f7febf RK |
189 | the initial value of a static variable. |
190 | ||
191 | If you don't know the type of the operand, you can still do this, but you | |
192 | must use @code{typeof} (@pxref{Typeof}) or type naming (@pxref{Naming | |
193 | Types}). | |
194 | ||
b98e139b MM |
195 | Statement expressions are not supported fully in G++, and their fate |
196 | there is unclear. (It is possible that they will become fully supported | |
197 | at some point, or that they will be deprecated, or that the bugs that | |
198 | are present will continue to exist indefinitely.) Presently, statement | |
02f52e19 | 199 | expressions do not work well as default arguments. |
b98e139b MM |
200 | |
201 | In addition, there are semantic issues with statement-expressions in | |
202 | C++. If you try to use statement-expressions instead of inline | |
203 | functions in C++, you may be surprised at the way object destruction is | |
204 | handled. For example: | |
205 | ||
206 | @example | |
207 | #define foo(a) (@{int b = (a); b + 3; @}) | |
208 | @end example | |
209 | ||
210 | @noindent | |
211 | does not work the same way as: | |
212 | ||
213 | @example | |
54e1d3a6 | 214 | inline int foo(int a) @{ int b = a; return b + 3; @} |
b98e139b MM |
215 | @end example |
216 | ||
217 | @noindent | |
218 | In particular, if the expression passed into @code{foo} involves the | |
219 | creation of temporaries, the destructors for those temporaries will be | |
220 | run earlier in the case of the macro than in the case of the function. | |
221 | ||
222 | These considerations mean that it is probably a bad idea to use | |
223 | statement-expressions of this form in header files that are designed to | |
54e1d3a6 MM |
224 | work with C++. (Note that some versions of the GNU C Library contained |
225 | header files using statement-expression that lead to precisely this | |
226 | bug.) | |
b98e139b | 227 | |
c1f7febf RK |
228 | @node Local Labels |
229 | @section Locally Declared Labels | |
230 | @cindex local labels | |
231 | @cindex macros, local labels | |
232 | ||
233 | Each statement expression is a scope in which @dfn{local labels} can be | |
234 | declared. A local label is simply an identifier; you can jump to it | |
235 | with an ordinary @code{goto} statement, but only from within the | |
236 | statement expression it belongs to. | |
237 | ||
238 | A local label declaration looks like this: | |
239 | ||
240 | @example | |
241 | __label__ @var{label}; | |
242 | @end example | |
243 | ||
244 | @noindent | |
245 | or | |
246 | ||
247 | @example | |
248 | __label__ @var{label1}, @var{label2}, @dots{}; | |
249 | @end example | |
250 | ||
251 | Local label declarations must come at the beginning of the statement | |
252 | expression, right after the @samp{(@{}, before any ordinary | |
253 | declarations. | |
254 | ||
255 | The label declaration defines the label @emph{name}, but does not define | |
256 | the label itself. You must do this in the usual way, with | |
257 | @code{@var{label}:}, within the statements of the statement expression. | |
258 | ||
259 | The local label feature is useful because statement expressions are | |
260 | often used in macros. If the macro contains nested loops, a @code{goto} | |
261 | can be useful for breaking out of them. However, an ordinary label | |
262 | whose scope is the whole function cannot be used: if the macro can be | |
263 | expanded several times in one function, the label will be multiply | |
264 | defined in that function. A local label avoids this problem. For | |
265 | example: | |
266 | ||
267 | @example | |
268 | #define SEARCH(array, target) \ | |
310668e8 | 269 | (@{ \ |
c1f7febf RK |
270 | __label__ found; \ |
271 | typeof (target) _SEARCH_target = (target); \ | |
272 | typeof (*(array)) *_SEARCH_array = (array); \ | |
273 | int i, j; \ | |
274 | int value; \ | |
275 | for (i = 0; i < max; i++) \ | |
276 | for (j = 0; j < max; j++) \ | |
277 | if (_SEARCH_array[i][j] == _SEARCH_target) \ | |
310668e8 | 278 | @{ value = i; goto found; @} \ |
c1f7febf RK |
279 | value = -1; \ |
280 | found: \ | |
281 | value; \ | |
282 | @}) | |
283 | @end example | |
284 | ||
285 | @node Labels as Values | |
286 | @section Labels as Values | |
287 | @cindex labels as values | |
288 | @cindex computed gotos | |
289 | @cindex goto with computed label | |
290 | @cindex address of a label | |
291 | ||
292 | You can get the address of a label defined in the current function | |
293 | (or a containing function) with the unary operator @samp{&&}. The | |
294 | value has type @code{void *}. This value is a constant and can be used | |
295 | wherever a constant of that type is valid. For example: | |
296 | ||
297 | @example | |
298 | void *ptr; | |
299 | @dots{} | |
300 | ptr = &&foo; | |
301 | @end example | |
302 | ||
303 | To use these values, you need to be able to jump to one. This is done | |
304 | with the computed goto statement@footnote{The analogous feature in | |
305 | Fortran is called an assigned goto, but that name seems inappropriate in | |
306 | C, where one can do more than simply store label addresses in label | |
307 | variables.}, @code{goto *@var{exp};}. For example, | |
308 | ||
309 | @example | |
310 | goto *ptr; | |
311 | @end example | |
312 | ||
313 | @noindent | |
314 | Any expression of type @code{void *} is allowed. | |
315 | ||
316 | One way of using these constants is in initializing a static array that | |
317 | will serve as a jump table: | |
318 | ||
319 | @example | |
320 | static void *array[] = @{ &&foo, &&bar, &&hack @}; | |
321 | @end example | |
322 | ||
323 | Then you can select a label with indexing, like this: | |
324 | ||
325 | @example | |
326 | goto *array[i]; | |
327 | @end example | |
328 | ||
329 | @noindent | |
330 | Note that this does not check whether the subscript is in bounds---array | |
331 | indexing in C never does that. | |
332 | ||
333 | Such an array of label values serves a purpose much like that of the | |
334 | @code{switch} statement. The @code{switch} statement is cleaner, so | |
335 | use that rather than an array unless the problem does not fit a | |
336 | @code{switch} statement very well. | |
337 | ||
338 | Another use of label values is in an interpreter for threaded code. | |
339 | The labels within the interpreter function can be stored in the | |
340 | threaded code for super-fast dispatching. | |
341 | ||
02f52e19 | 342 | You may not use this mechanism to jump to code in a different function. |
47620e09 | 343 | If you do that, totally unpredictable things will happen. The best way to |
c1f7febf RK |
344 | avoid this is to store the label address only in automatic variables and |
345 | never pass it as an argument. | |
346 | ||
47620e09 RH |
347 | An alternate way to write the above example is |
348 | ||
349 | @example | |
310668e8 JM |
350 | static const int array[] = @{ &&foo - &&foo, &&bar - &&foo, |
351 | &&hack - &&foo @}; | |
47620e09 RH |
352 | goto *(&&foo + array[i]); |
353 | @end example | |
354 | ||
355 | @noindent | |
356 | This is more friendly to code living in shared libraries, as it reduces | |
357 | the number of dynamic relocations that are needed, and by consequence, | |
358 | allows the data to be read-only. | |
359 | ||
c1f7febf RK |
360 | @node Nested Functions |
361 | @section Nested Functions | |
362 | @cindex nested functions | |
363 | @cindex downward funargs | |
364 | @cindex thunks | |
365 | ||
366 | A @dfn{nested function} is a function defined inside another function. | |
367 | (Nested functions are not supported for GNU C++.) The nested function's | |
368 | name is local to the block where it is defined. For example, here we | |
369 | define a nested function named @code{square}, and call it twice: | |
370 | ||
371 | @example | |
372 | @group | |
373 | foo (double a, double b) | |
374 | @{ | |
375 | double square (double z) @{ return z * z; @} | |
376 | ||
377 | return square (a) + square (b); | |
378 | @} | |
379 | @end group | |
380 | @end example | |
381 | ||
382 | The nested function can access all the variables of the containing | |
383 | function that are visible at the point of its definition. This is | |
384 | called @dfn{lexical scoping}. For example, here we show a nested | |
385 | function which uses an inherited variable named @code{offset}: | |
386 | ||
387 | @example | |
388 | bar (int *array, int offset, int size) | |
389 | @{ | |
390 | int access (int *array, int index) | |
391 | @{ return array[index + offset]; @} | |
392 | int i; | |
393 | @dots{} | |
394 | for (i = 0; i < size; i++) | |
395 | @dots{} access (array, i) @dots{} | |
396 | @} | |
397 | @end example | |
398 | ||
399 | Nested function definitions are permitted within functions in the places | |
400 | where variable definitions are allowed; that is, in any block, before | |
401 | the first statement in the block. | |
402 | ||
403 | It is possible to call the nested function from outside the scope of its | |
404 | name by storing its address or passing the address to another function: | |
405 | ||
406 | @example | |
407 | hack (int *array, int size) | |
408 | @{ | |
409 | void store (int index, int value) | |
410 | @{ array[index] = value; @} | |
411 | ||
412 | intermediate (store, size); | |
413 | @} | |
414 | @end example | |
415 | ||
416 | Here, the function @code{intermediate} receives the address of | |
417 | @code{store} as an argument. If @code{intermediate} calls @code{store}, | |
418 | the arguments given to @code{store} are used to store into @code{array}. | |
419 | But this technique works only so long as the containing function | |
420 | (@code{hack}, in this example) does not exit. | |
421 | ||
422 | If you try to call the nested function through its address after the | |
423 | containing function has exited, all hell will break loose. If you try | |
424 | to call it after a containing scope level has exited, and if it refers | |
425 | to some of the variables that are no longer in scope, you may be lucky, | |
426 | but it's not wise to take the risk. If, however, the nested function | |
427 | does not refer to anything that has gone out of scope, you should be | |
428 | safe. | |
429 | ||
f0523f02 | 430 | GCC implements taking the address of a nested function using a |
674032e2 | 431 | technique called @dfn{trampolines}. A paper describing them is |
9734e80c | 432 | available as @uref{http://people.debian.org/~karlheg/Usenix88-lexic.pdf}. |
c1f7febf RK |
433 | |
434 | A nested function can jump to a label inherited from a containing | |
435 | function, provided the label was explicitly declared in the containing | |
436 | function (@pxref{Local Labels}). Such a jump returns instantly to the | |
437 | containing function, exiting the nested function which did the | |
438 | @code{goto} and any intermediate functions as well. Here is an example: | |
439 | ||
440 | @example | |
441 | @group | |
442 | bar (int *array, int offset, int size) | |
443 | @{ | |
444 | __label__ failure; | |
445 | int access (int *array, int index) | |
446 | @{ | |
447 | if (index > size) | |
448 | goto failure; | |
449 | return array[index + offset]; | |
450 | @} | |
451 | int i; | |
452 | @dots{} | |
453 | for (i = 0; i < size; i++) | |
454 | @dots{} access (array, i) @dots{} | |
455 | @dots{} | |
456 | return 0; | |
457 | ||
458 | /* @r{Control comes here from @code{access} | |
459 | if it detects an error.} */ | |
460 | failure: | |
461 | return -1; | |
462 | @} | |
463 | @end group | |
464 | @end example | |
465 | ||
466 | A nested function always has internal linkage. Declaring one with | |
467 | @code{extern} is erroneous. If you need to declare the nested function | |
468 | before its definition, use @code{auto} (which is otherwise meaningless | |
469 | for function declarations). | |
470 | ||
471 | @example | |
472 | bar (int *array, int offset, int size) | |
473 | @{ | |
474 | __label__ failure; | |
475 | auto int access (int *, int); | |
476 | @dots{} | |
477 | int access (int *array, int index) | |
478 | @{ | |
479 | if (index > size) | |
480 | goto failure; | |
481 | return array[index + offset]; | |
482 | @} | |
483 | @dots{} | |
484 | @} | |
485 | @end example | |
486 | ||
487 | @node Constructing Calls | |
488 | @section Constructing Function Calls | |
489 | @cindex constructing calls | |
490 | @cindex forwarding calls | |
491 | ||
492 | Using the built-in functions described below, you can record | |
493 | the arguments a function received, and call another function | |
494 | with the same arguments, without knowing the number or types | |
495 | of the arguments. | |
496 | ||
497 | You can also record the return value of that function call, | |
498 | and later return that value, without knowing what data type | |
499 | the function tried to return (as long as your caller expects | |
500 | that data type). | |
501 | ||
84330467 JM |
502 | @deftypefn {Built-in Function} {void *} __builtin_apply_args () |
503 | This built-in function returns a pointer to data | |
c1f7febf RK |
504 | describing how to perform a call with the same arguments as were passed |
505 | to the current function. | |
506 | ||
507 | The function saves the arg pointer register, structure value address, | |
508 | and all registers that might be used to pass arguments to a function | |
509 | into a block of memory allocated on the stack. Then it returns the | |
510 | address of that block. | |
84330467 | 511 | @end deftypefn |
c1f7febf | 512 | |
84330467 JM |
513 | @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size}) |
514 | This built-in function invokes @var{function} | |
515 | with a copy of the parameters described by @var{arguments} | |
516 | and @var{size}. | |
c1f7febf RK |
517 | |
518 | The value of @var{arguments} should be the value returned by | |
519 | @code{__builtin_apply_args}. The argument @var{size} specifies the size | |
520 | of the stack argument data, in bytes. | |
521 | ||
84330467 | 522 | This function returns a pointer to data describing |
c1f7febf RK |
523 | how to return whatever value was returned by @var{function}. The data |
524 | is saved in a block of memory allocated on the stack. | |
525 | ||
526 | It is not always simple to compute the proper value for @var{size}. The | |
527 | value is used by @code{__builtin_apply} to compute the amount of data | |
528 | that should be pushed on the stack and copied from the incoming argument | |
529 | area. | |
84330467 | 530 | @end deftypefn |
c1f7febf | 531 | |
84330467 | 532 | @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result}) |
c1f7febf RK |
533 | This built-in function returns the value described by @var{result} from |
534 | the containing function. You should specify, for @var{result}, a value | |
535 | returned by @code{__builtin_apply}. | |
84330467 | 536 | @end deftypefn |
c1f7febf RK |
537 | |
538 | @node Naming Types | |
539 | @section Naming an Expression's Type | |
540 | @cindex naming types | |
541 | ||
542 | You can give a name to the type of an expression using a @code{typedef} | |
543 | declaration with an initializer. Here is how to define @var{name} as a | |
544 | type name for the type of @var{exp}: | |
545 | ||
546 | @example | |
547 | typedef @var{name} = @var{exp}; | |
548 | @end example | |
549 | ||
550 | This is useful in conjunction with the statements-within-expressions | |
551 | feature. Here is how the two together can be used to define a safe | |
552 | ``maximum'' macro that operates on any arithmetic type: | |
553 | ||
554 | @example | |
555 | #define max(a,b) \ | |
556 | (@{typedef _ta = (a), _tb = (b); \ | |
557 | _ta _a = (a); _tb _b = (b); \ | |
558 | _a > _b ? _a : _b; @}) | |
559 | @end example | |
560 | ||
561 | @cindex underscores in variables in macros | |
562 | @cindex @samp{_} in variables in macros | |
563 | @cindex local variables in macros | |
564 | @cindex variables, local, in macros | |
565 | @cindex macros, local variables in | |
566 | ||
567 | The reason for using names that start with underscores for the local | |
568 | variables is to avoid conflicts with variable names that occur within the | |
569 | expressions that are substituted for @code{a} and @code{b}. Eventually we | |
570 | hope to design a new form of declaration syntax that allows you to declare | |
571 | variables whose scopes start only after their initializers; this will be a | |
572 | more reliable way to prevent such conflicts. | |
573 | ||
574 | @node Typeof | |
575 | @section Referring to a Type with @code{typeof} | |
576 | @findex typeof | |
577 | @findex sizeof | |
578 | @cindex macros, types of arguments | |
579 | ||
580 | Another way to refer to the type of an expression is with @code{typeof}. | |
581 | The syntax of using of this keyword looks like @code{sizeof}, but the | |
582 | construct acts semantically like a type name defined with @code{typedef}. | |
583 | ||
584 | There are two ways of writing the argument to @code{typeof}: with an | |
585 | expression or with a type. Here is an example with an expression: | |
586 | ||
587 | @example | |
588 | typeof (x[0](1)) | |
589 | @end example | |
590 | ||
591 | @noindent | |
89aed483 JM |
592 | This assumes that @code{x} is an array of pointers to functions; |
593 | the type described is that of the values of the functions. | |
c1f7febf RK |
594 | |
595 | Here is an example with a typename as the argument: | |
596 | ||
597 | @example | |
598 | typeof (int *) | |
599 | @end example | |
600 | ||
601 | @noindent | |
602 | Here the type described is that of pointers to @code{int}. | |
603 | ||
5490d604 | 604 | If you are writing a header file that must work when included in ISO C |
c1f7febf RK |
605 | programs, write @code{__typeof__} instead of @code{typeof}. |
606 | @xref{Alternate Keywords}. | |
607 | ||
608 | A @code{typeof}-construct can be used anywhere a typedef name could be | |
609 | used. For example, you can use it in a declaration, in a cast, or inside | |
610 | of @code{sizeof} or @code{typeof}. | |
611 | ||
612 | @itemize @bullet | |
613 | @item | |
614 | This declares @code{y} with the type of what @code{x} points to. | |
615 | ||
616 | @example | |
617 | typeof (*x) y; | |
618 | @end example | |
619 | ||
620 | @item | |
621 | This declares @code{y} as an array of such values. | |
622 | ||
623 | @example | |
624 | typeof (*x) y[4]; | |
625 | @end example | |
626 | ||
627 | @item | |
628 | This declares @code{y} as an array of pointers to characters: | |
629 | ||
630 | @example | |
631 | typeof (typeof (char *)[4]) y; | |
632 | @end example | |
633 | ||
634 | @noindent | |
635 | It is equivalent to the following traditional C declaration: | |
636 | ||
637 | @example | |
638 | char *y[4]; | |
639 | @end example | |
640 | ||
641 | To see the meaning of the declaration using @code{typeof}, and why it | |
642 | might be a useful way to write, let's rewrite it with these macros: | |
643 | ||
644 | @example | |
645 | #define pointer(T) typeof(T *) | |
646 | #define array(T, N) typeof(T [N]) | |
647 | @end example | |
648 | ||
649 | @noindent | |
650 | Now the declaration can be rewritten this way: | |
651 | ||
652 | @example | |
653 | array (pointer (char), 4) y; | |
654 | @end example | |
655 | ||
656 | @noindent | |
657 | Thus, @code{array (pointer (char), 4)} is the type of arrays of 4 | |
658 | pointers to @code{char}. | |
659 | @end itemize | |
660 | ||
661 | @node Lvalues | |
662 | @section Generalized Lvalues | |
663 | @cindex compound expressions as lvalues | |
664 | @cindex expressions, compound, as lvalues | |
665 | @cindex conditional expressions as lvalues | |
666 | @cindex expressions, conditional, as lvalues | |
667 | @cindex casts as lvalues | |
668 | @cindex generalized lvalues | |
669 | @cindex lvalues, generalized | |
670 | @cindex extensions, @code{?:} | |
671 | @cindex @code{?:} extensions | |
672 | Compound expressions, conditional expressions and casts are allowed as | |
673 | lvalues provided their operands are lvalues. This means that you can take | |
674 | their addresses or store values into them. | |
675 | ||
676 | Standard C++ allows compound expressions and conditional expressions as | |
677 | lvalues, and permits casts to reference type, so use of this extension | |
678 | is deprecated for C++ code. | |
679 | ||
680 | For example, a compound expression can be assigned, provided the last | |
681 | expression in the sequence is an lvalue. These two expressions are | |
682 | equivalent: | |
683 | ||
684 | @example | |
685 | (a, b) += 5 | |
686 | a, (b += 5) | |
687 | @end example | |
688 | ||
689 | Similarly, the address of the compound expression can be taken. These two | |
690 | expressions are equivalent: | |
691 | ||
692 | @example | |
693 | &(a, b) | |
694 | a, &b | |
695 | @end example | |
696 | ||
697 | A conditional expression is a valid lvalue if its type is not void and the | |
698 | true and false branches are both valid lvalues. For example, these two | |
699 | expressions are equivalent: | |
700 | ||
701 | @example | |
702 | (a ? b : c) = 5 | |
703 | (a ? b = 5 : (c = 5)) | |
704 | @end example | |
705 | ||
706 | A cast is a valid lvalue if its operand is an lvalue. A simple | |
707 | assignment whose left-hand side is a cast works by converting the | |
708 | right-hand side first to the specified type, then to the type of the | |
709 | inner left-hand side expression. After this is stored, the value is | |
710 | converted back to the specified type to become the value of the | |
711 | assignment. Thus, if @code{a} has type @code{char *}, the following two | |
712 | expressions are equivalent: | |
713 | ||
714 | @example | |
715 | (int)a = 5 | |
716 | (int)(a = (char *)(int)5) | |
717 | @end example | |
718 | ||
719 | An assignment-with-arithmetic operation such as @samp{+=} applied to a cast | |
720 | performs the arithmetic using the type resulting from the cast, and then | |
721 | continues as in the previous case. Therefore, these two expressions are | |
722 | equivalent: | |
723 | ||
724 | @example | |
725 | (int)a += 5 | |
726 | (int)(a = (char *)(int) ((int)a + 5)) | |
727 | @end example | |
728 | ||
729 | You cannot take the address of an lvalue cast, because the use of its | |
730 | address would not work out coherently. Suppose that @code{&(int)f} were | |
731 | permitted, where @code{f} has type @code{float}. Then the following | |
732 | statement would try to store an integer bit-pattern where a floating | |
733 | point number belongs: | |
734 | ||
735 | @example | |
736 | *&(int)f = 1; | |
737 | @end example | |
738 | ||
739 | This is quite different from what @code{(int)f = 1} would do---that | |
740 | would convert 1 to floating point and store it. Rather than cause this | |
741 | inconsistency, we think it is better to prohibit use of @samp{&} on a cast. | |
742 | ||
743 | If you really do want an @code{int *} pointer with the address of | |
744 | @code{f}, you can simply write @code{(int *)&f}. | |
745 | ||
746 | @node Conditionals | |
747 | @section Conditionals with Omitted Operands | |
748 | @cindex conditional expressions, extensions | |
749 | @cindex omitted middle-operands | |
750 | @cindex middle-operands, omitted | |
751 | @cindex extensions, @code{?:} | |
752 | @cindex @code{?:} extensions | |
753 | ||
754 | The middle operand in a conditional expression may be omitted. Then | |
755 | if the first operand is nonzero, its value is the value of the conditional | |
756 | expression. | |
757 | ||
758 | Therefore, the expression | |
759 | ||
760 | @example | |
761 | x ? : y | |
762 | @end example | |
763 | ||
764 | @noindent | |
765 | has the value of @code{x} if that is nonzero; otherwise, the value of | |
766 | @code{y}. | |
767 | ||
768 | This example is perfectly equivalent to | |
769 | ||
770 | @example | |
771 | x ? x : y | |
772 | @end example | |
773 | ||
774 | @cindex side effect in ?: | |
775 | @cindex ?: side effect | |
776 | @noindent | |
777 | In this simple case, the ability to omit the middle operand is not | |
778 | especially useful. When it becomes useful is when the first operand does, | |
779 | or may (if it is a macro argument), contain a side effect. Then repeating | |
780 | the operand in the middle would perform the side effect twice. Omitting | |
781 | the middle operand uses the value already computed without the undesirable | |
782 | effects of recomputing it. | |
783 | ||
784 | @node Long Long | |
785 | @section Double-Word Integers | |
786 | @cindex @code{long long} data types | |
787 | @cindex double-word arithmetic | |
788 | @cindex multiprecision arithmetic | |
4b404517 JM |
789 | @cindex @code{LL} integer suffix |
790 | @cindex @code{ULL} integer suffix | |
c1f7febf | 791 | |
4b404517 JM |
792 | ISO C99 supports data types for integers that are at least 64 bits wide, |
793 | and as an extension GCC supports them in C89 mode and in C++. | |
794 | Simply write @code{long long int} for a signed integer, or | |
c1f7febf | 795 | @code{unsigned long long int} for an unsigned integer. To make an |
84330467 | 796 | integer constant of type @code{long long int}, add the suffix @samp{LL} |
c1f7febf | 797 | to the integer. To make an integer constant of type @code{unsigned long |
84330467 | 798 | long int}, add the suffix @samp{ULL} to the integer. |
c1f7febf RK |
799 | |
800 | You can use these types in arithmetic like any other integer types. | |
801 | Addition, subtraction, and bitwise boolean operations on these types | |
802 | are open-coded on all types of machines. Multiplication is open-coded | |
803 | if the machine supports fullword-to-doubleword a widening multiply | |
804 | instruction. Division and shifts are open-coded only on machines that | |
805 | provide special support. The operations that are not open-coded use | |
f0523f02 | 806 | special library routines that come with GCC. |
c1f7febf RK |
807 | |
808 | There may be pitfalls when you use @code{long long} types for function | |
809 | arguments, unless you declare function prototypes. If a function | |
810 | expects type @code{int} for its argument, and you pass a value of type | |
811 | @code{long long int}, confusion will result because the caller and the | |
812 | subroutine will disagree about the number of bytes for the argument. | |
813 | Likewise, if the function expects @code{long long int} and you pass | |
814 | @code{int}. The best way to avoid such problems is to use prototypes. | |
815 | ||
816 | @node Complex | |
817 | @section Complex Numbers | |
818 | @cindex complex numbers | |
4b404517 JM |
819 | @cindex @code{_Complex} keyword |
820 | @cindex @code{__complex__} keyword | |
c1f7febf | 821 | |
4b404517 JM |
822 | ISO C99 supports complex floating data types, and as an extension GCC |
823 | supports them in C89 mode and in C++, and supports complex integer data | |
824 | types which are not part of ISO C99. You can declare complex types | |
825 | using the keyword @code{_Complex}. As an extension, the older GNU | |
826 | keyword @code{__complex__} is also supported. | |
c1f7febf | 827 | |
4b404517 | 828 | For example, @samp{_Complex double x;} declares @code{x} as a |
c1f7febf | 829 | variable whose real part and imaginary part are both of type |
4b404517 | 830 | @code{double}. @samp{_Complex short int y;} declares @code{y} to |
c1f7febf RK |
831 | have real and imaginary parts of type @code{short int}; this is not |
832 | likely to be useful, but it shows that the set of complex types is | |
833 | complete. | |
834 | ||
835 | To write a constant with a complex data type, use the suffix @samp{i} or | |
836 | @samp{j} (either one; they are equivalent). For example, @code{2.5fi} | |
4b404517 JM |
837 | has type @code{_Complex float} and @code{3i} has type |
838 | @code{_Complex int}. Such a constant always has a pure imaginary | |
c1f7febf | 839 | value, but you can form any complex value you like by adding one to a |
4b404517 JM |
840 | real constant. This is a GNU extension; if you have an ISO C99 |
841 | conforming C library (such as GNU libc), and want to construct complex | |
842 | constants of floating type, you should include @code{<complex.h>} and | |
843 | use the macros @code{I} or @code{_Complex_I} instead. | |
c1f7febf | 844 | |
4b404517 JM |
845 | @cindex @code{__real__} keyword |
846 | @cindex @code{__imag__} keyword | |
c1f7febf RK |
847 | To extract the real part of a complex-valued expression @var{exp}, write |
848 | @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to | |
4b404517 JM |
849 | extract the imaginary part. This is a GNU extension; for values of |
850 | floating type, you should use the ISO C99 functions @code{crealf}, | |
851 | @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and | |
852 | @code{cimagl}, declared in @code{<complex.h>} and also provided as | |
84330467 | 853 | built-in functions by GCC. |
c1f7febf | 854 | |
4b404517 | 855 | @cindex complex conjugation |
c1f7febf | 856 | The operator @samp{~} performs complex conjugation when used on a value |
4b404517 JM |
857 | with a complex type. This is a GNU extension; for values of |
858 | floating type, you should use the ISO C99 functions @code{conjf}, | |
859 | @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also | |
84330467 | 860 | provided as built-in functions by GCC. |
c1f7febf | 861 | |
f0523f02 | 862 | GCC can allocate complex automatic variables in a noncontiguous |
c1f7febf RK |
863 | fashion; it's even possible for the real part to be in a register while |
864 | the imaginary part is on the stack (or vice-versa). None of the | |
865 | supported debugging info formats has a way to represent noncontiguous | |
f0523f02 | 866 | allocation like this, so GCC describes a noncontiguous complex |
c1f7febf RK |
867 | variable as if it were two separate variables of noncomplex type. |
868 | If the variable's actual name is @code{foo}, the two fictitious | |
869 | variables are named @code{foo$real} and @code{foo$imag}. You can | |
870 | examine and set these two fictitious variables with your debugger. | |
871 | ||
872 | A future version of GDB will know how to recognize such pairs and treat | |
873 | them as a single variable with a complex type. | |
874 | ||
6f4d7222 | 875 | @node Hex Floats |
6b42b9ea UD |
876 | @section Hex Floats |
877 | @cindex hex floats | |
c5c76735 | 878 | |
4b404517 | 879 | ISO C99 supports floating-point numbers written not only in the usual |
6f4d7222 | 880 | decimal notation, such as @code{1.55e1}, but also numbers such as |
4b404517 JM |
881 | @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC |
882 | supports this in C89 mode (except in some cases when strictly | |
883 | conforming) and in C++. In that format the | |
84330467 | 884 | @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are |
6f4d7222 | 885 | mandatory. The exponent is a decimal number that indicates the power of |
84330467 JM |
886 | 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is |
887 | 1 15/16, @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3} | |
6f4d7222 UD |
888 | is the same as @code{1.55e1}. |
889 | ||
890 | Unlike for floating-point numbers in the decimal notation the exponent | |
891 | is always required in the hexadecimal notation. Otherwise the compiler | |
892 | would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This | |
84330467 | 893 | could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the |
6f4d7222 UD |
894 | extension for floating-point constants of type @code{float}. |
895 | ||
c1f7febf RK |
896 | @node Zero Length |
897 | @section Arrays of Length Zero | |
898 | @cindex arrays of length zero | |
899 | @cindex zero-length arrays | |
900 | @cindex length-zero arrays | |
ffc5c6a9 | 901 | @cindex flexible array members |
c1f7febf | 902 | |
584ef5fe RH |
903 | Zero-length arrays are allowed in GNU C. They are very useful as the |
904 | last element of a structure which is really a header for a variable-length | |
c1f7febf RK |
905 | object: |
906 | ||
907 | @example | |
908 | struct line @{ | |
909 | int length; | |
910 | char contents[0]; | |
911 | @}; | |
912 | ||
584ef5fe RH |
913 | struct line *thisline = (struct line *) |
914 | malloc (sizeof (struct line) + this_length); | |
915 | thisline->length = this_length; | |
c1f7febf RK |
916 | @end example |
917 | ||
a25f1211 | 918 | In ISO C89, you would have to give @code{contents} a length of 1, which |
c1f7febf RK |
919 | means either you waste space or complicate the argument to @code{malloc}. |
920 | ||
02f52e19 | 921 | In ISO C99, you would use a @dfn{flexible array member}, which is |
584ef5fe RH |
922 | slightly different in syntax and semantics: |
923 | ||
924 | @itemize @bullet | |
925 | @item | |
926 | Flexible array members are written as @code{contents[]} without | |
927 | the @code{0}. | |
928 | ||
929 | @item | |
930 | Flexible array members have incomplete type, and so the @code{sizeof} | |
931 | operator may not be applied. As a quirk of the original implementation | |
932 | of zero-length arrays, @code{sizeof} evaluates to zero. | |
933 | ||
934 | @item | |
935 | Flexible array members may only appear as the last member of a | |
02f52e19 | 936 | @code{struct} that is otherwise non-empty. GCC currently allows |
584ef5fe RH |
937 | zero-length arrays anywhere. You may encounter problems, however, |
938 | defining structures containing only a zero-length array. Such usage | |
939 | is deprecated, and we recommend using zero-length arrays only in | |
940 | places in which flexible array members would be allowed. | |
ffc5c6a9 | 941 | @end itemize |
a25f1211 | 942 | |
ffc5c6a9 RH |
943 | GCC versions before 3.0 allowed zero-length arrays to be statically |
944 | initialized. In addition to those cases that were useful, it also | |
945 | allowed initializations in situations that would corrupt later data. | |
946 | Non-empty initialization of zero-length arrays is now deprecated. | |
947 | ||
948 | Instead GCC allows static initialization of flexible array members. | |
949 | This is equivalent to defining a new structure containing the original | |
950 | structure followed by an array of sufficient size to contain the data. | |
951 | I.e. in the following, @code{f1} is constructed as if it were declared | |
952 | like @code{f2}. | |
a25f1211 RH |
953 | |
954 | @example | |
ffc5c6a9 RH |
955 | struct f1 @{ |
956 | int x; int y[]; | |
957 | @} f1 = @{ 1, @{ 2, 3, 4 @} @}; | |
958 | ||
959 | struct f2 @{ | |
960 | struct f1 f1; int data[3]; | |
961 | @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @}; | |
962 | @end example | |
584ef5fe | 963 | |
ffc5c6a9 RH |
964 | @noindent |
965 | The convenience of this extension is that @code{f1} has the desired | |
966 | type, eliminating the need to consistently refer to @code{f2.f1}. | |
967 | ||
968 | This has symmetry with normal static arrays, in that an array of | |
969 | unknown size is also written with @code{[]}. | |
a25f1211 | 970 | |
ffc5c6a9 RH |
971 | Of course, this extension only makes sense if the extra data comes at |
972 | the end of a top-level object, as otherwise we would be overwriting | |
973 | data at subsequent offsets. To avoid undue complication and confusion | |
974 | with initialization of deeply nested arrays, we simply disallow any | |
975 | non-empty initialization except when the structure is the top-level | |
976 | object. For example: | |
584ef5fe | 977 | |
ffc5c6a9 RH |
978 | @example |
979 | struct foo @{ int x; int y[]; @}; | |
980 | struct bar @{ struct foo z; @}; | |
981 | ||
982 | struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // Legal. | |
983 | struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // Illegal. | |
984 | struct bar c = @{ @{ 1, @{ @} @} @}; // Legal. | |
985 | struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // Illegal. | |
a25f1211 | 986 | @end example |
4b606faf | 987 | |
c1f7febf RK |
988 | @node Variable Length |
989 | @section Arrays of Variable Length | |
990 | @cindex variable-length arrays | |
991 | @cindex arrays of variable length | |
4b404517 | 992 | @cindex VLAs |
c1f7febf | 993 | |
4b404517 JM |
994 | Variable-length automatic arrays are allowed in ISO C99, and as an |
995 | extension GCC accepts them in C89 mode and in C++. (However, GCC's | |
996 | implementation of variable-length arrays does not yet conform in detail | |
997 | to the ISO C99 standard.) These arrays are | |
c1f7febf RK |
998 | declared like any other automatic arrays, but with a length that is not |
999 | a constant expression. The storage is allocated at the point of | |
1000 | declaration and deallocated when the brace-level is exited. For | |
1001 | example: | |
1002 | ||
1003 | @example | |
1004 | FILE * | |
1005 | concat_fopen (char *s1, char *s2, char *mode) | |
1006 | @{ | |
1007 | char str[strlen (s1) + strlen (s2) + 1]; | |
1008 | strcpy (str, s1); | |
1009 | strcat (str, s2); | |
1010 | return fopen (str, mode); | |
1011 | @} | |
1012 | @end example | |
1013 | ||
1014 | @cindex scope of a variable length array | |
1015 | @cindex variable-length array scope | |
1016 | @cindex deallocating variable length arrays | |
1017 | Jumping or breaking out of the scope of the array name deallocates the | |
1018 | storage. Jumping into the scope is not allowed; you get an error | |
1019 | message for it. | |
1020 | ||
1021 | @cindex @code{alloca} vs variable-length arrays | |
1022 | You can use the function @code{alloca} to get an effect much like | |
1023 | variable-length arrays. The function @code{alloca} is available in | |
1024 | many other C implementations (but not in all). On the other hand, | |
1025 | variable-length arrays are more elegant. | |
1026 | ||
1027 | There are other differences between these two methods. Space allocated | |
1028 | with @code{alloca} exists until the containing @emph{function} returns. | |
1029 | The space for a variable-length array is deallocated as soon as the array | |
1030 | name's scope ends. (If you use both variable-length arrays and | |
1031 | @code{alloca} in the same function, deallocation of a variable-length array | |
1032 | will also deallocate anything more recently allocated with @code{alloca}.) | |
1033 | ||
1034 | You can also use variable-length arrays as arguments to functions: | |
1035 | ||
1036 | @example | |
1037 | struct entry | |
1038 | tester (int len, char data[len][len]) | |
1039 | @{ | |
1040 | @dots{} | |
1041 | @} | |
1042 | @end example | |
1043 | ||
1044 | The length of an array is computed once when the storage is allocated | |
1045 | and is remembered for the scope of the array in case you access it with | |
1046 | @code{sizeof}. | |
1047 | ||
1048 | If you want to pass the array first and the length afterward, you can | |
1049 | use a forward declaration in the parameter list---another GNU extension. | |
1050 | ||
1051 | @example | |
1052 | struct entry | |
1053 | tester (int len; char data[len][len], int len) | |
1054 | @{ | |
1055 | @dots{} | |
1056 | @} | |
1057 | @end example | |
1058 | ||
1059 | @cindex parameter forward declaration | |
1060 | The @samp{int len} before the semicolon is a @dfn{parameter forward | |
1061 | declaration}, and it serves the purpose of making the name @code{len} | |
1062 | known when the declaration of @code{data} is parsed. | |
1063 | ||
1064 | You can write any number of such parameter forward declarations in the | |
1065 | parameter list. They can be separated by commas or semicolons, but the | |
1066 | last one must end with a semicolon, which is followed by the ``real'' | |
1067 | parameter declarations. Each forward declaration must match a ``real'' | |
4b404517 JM |
1068 | declaration in parameter name and data type. ISO C99 does not support |
1069 | parameter forward declarations. | |
c1f7febf | 1070 | |
ccd96f0a NB |
1071 | @node Variadic Macros |
1072 | @section Macros with a Variable Number of Arguments. | |
c1f7febf RK |
1073 | @cindex variable number of arguments |
1074 | @cindex macro with variable arguments | |
1075 | @cindex rest argument (in macro) | |
ccd96f0a | 1076 | @cindex variadic macros |
c1f7febf | 1077 | |
ccd96f0a NB |
1078 | In the ISO C standard of 1999, a macro can be declared to accept a |
1079 | variable number of arguments much as a function can. The syntax for | |
1080 | defining the macro is similar to that of a function. Here is an | |
1081 | example: | |
c1f7febf RK |
1082 | |
1083 | @example | |
ccd96f0a | 1084 | #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__) |
c1f7febf RK |
1085 | @end example |
1086 | ||
ccd96f0a NB |
1087 | Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of |
1088 | such a macro, it represents the zero or more tokens until the closing | |
1089 | parenthesis that ends the invocation, including any commas. This set of | |
1090 | tokens replaces the identifier @code{__VA_ARGS__} in the macro body | |
1091 | wherever it appears. See the CPP manual for more information. | |
1092 | ||
1093 | GCC has long supported variadic macros, and used a different syntax that | |
1094 | allowed you to give a name to the variable arguments just like any other | |
1095 | argument. Here is an example: | |
c1f7febf RK |
1096 | |
1097 | @example | |
ccd96f0a | 1098 | #define debug(format, args...) fprintf (stderr, format, args) |
c1f7febf RK |
1099 | @end example |
1100 | ||
ccd96f0a NB |
1101 | This is in all ways equivalent to the ISO C example above, but arguably |
1102 | more readable and descriptive. | |
c1f7febf | 1103 | |
ccd96f0a NB |
1104 | GNU CPP has two further variadic macro extensions, and permits them to |
1105 | be used with either of the above forms of macro definition. | |
1106 | ||
1107 | In standard C, you are not allowed to leave the variable argument out | |
1108 | entirely; but you are allowed to pass an empty argument. For example, | |
1109 | this invocation is invalid in ISO C, because there is no comma after | |
1110 | the string: | |
c1f7febf RK |
1111 | |
1112 | @example | |
ccd96f0a | 1113 | debug ("A message") |
c1f7febf RK |
1114 | @end example |
1115 | ||
ccd96f0a NB |
1116 | GNU CPP permits you to completely omit the variable arguments in this |
1117 | way. In the above examples, the compiler would complain, though since | |
1118 | the expansion of the macro still has the extra comma after the format | |
1119 | string. | |
1120 | ||
1121 | To help solve this problem, CPP behaves specially for variable arguments | |
1122 | used with the token paste operator, @samp{##}. If instead you write | |
c1f7febf RK |
1123 | |
1124 | @example | |
ccd96f0a | 1125 | #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__) |
c1f7febf RK |
1126 | @end example |
1127 | ||
ccd96f0a NB |
1128 | and if the variable arguments are omitted or empty, the @samp{##} |
1129 | operator causes the preprocessor to remove the comma before it. If you | |
1130 | do provide some variable arguments in your macro invocation, GNU CPP | |
1131 | does not complain about the paste operation and instead places the | |
1132 | variable arguments after the comma. Just like any other pasted macro | |
1133 | argument, these arguments are not macro expanded. | |
1134 | ||
1135 | @node Escaped Newlines | |
1136 | @section Slightly Looser Rules for Escaped Newlines | |
1137 | @cindex escaped newlines | |
1138 | @cindex newlines (escaped) | |
1139 | ||
1140 | Recently, the non-traditional preprocessor has relaxed its treatment of | |
1141 | escaped newlines. Previously, the newline had to immediately follow a | |
1142 | backslash. The current implementation allows whitespace in the form of | |
1143 | spaces, horizontal and vertical tabs, and form feeds between the | |
1144 | backslash and the subsequent newline. The preprocessor issues a | |
1145 | warning, but treats it as a valid escaped newline and combines the two | |
1146 | lines to form a single logical line. This works within comments and | |
1147 | tokens, including multi-line strings, as well as between tokens. | |
1148 | Comments are @emph{not} treated as whitespace for the purposes of this | |
1149 | relaxation, since they have not yet been replaced with spaces. | |
1150 | ||
1151 | @node Multi-line Strings | |
1152 | @section String Literals with Embedded Newlines | |
1153 | @cindex multi-line string literals | |
1154 | ||
1155 | As an extension, GNU CPP permits string literals to cross multiple lines | |
1156 | without escaping the embedded newlines. Each embedded newline is | |
1157 | replaced with a single @samp{\n} character in the resulting string | |
1158 | literal, regardless of what form the newline took originally. | |
1159 | ||
1160 | CPP currently allows such strings in directives as well (other than the | |
1161 | @samp{#include} family). This is deprecated and will eventually be | |
1162 | removed. | |
c1f7febf RK |
1163 | |
1164 | @node Subscripting | |
1165 | @section Non-Lvalue Arrays May Have Subscripts | |
1166 | @cindex subscripting | |
1167 | @cindex arrays, non-lvalue | |
1168 | ||
1169 | @cindex subscripting and function values | |
1170 | Subscripting is allowed on arrays that are not lvalues, even though the | |
4b404517 JM |
1171 | unary @samp{&} operator is not. (In ISO C99, both are allowed (though |
1172 | the array may not be used after the next sequence point), but this ISO | |
1173 | C99 feature is not yet fully supported in GCC.) For example, | |
1174 | this is valid in GNU C though not valid in C89: | |
c1f7febf RK |
1175 | |
1176 | @example | |
1177 | @group | |
1178 | struct foo @{int a[4];@}; | |
1179 | ||
1180 | struct foo f(); | |
1181 | ||
1182 | bar (int index) | |
1183 | @{ | |
1184 | return f().a[index]; | |
1185 | @} | |
1186 | @end group | |
1187 | @end example | |
1188 | ||
1189 | @node Pointer Arith | |
1190 | @section Arithmetic on @code{void}- and Function-Pointers | |
1191 | @cindex void pointers, arithmetic | |
1192 | @cindex void, size of pointer to | |
1193 | @cindex function pointers, arithmetic | |
1194 | @cindex function, size of pointer to | |
1195 | ||
1196 | In GNU C, addition and subtraction operations are supported on pointers to | |
1197 | @code{void} and on pointers to functions. This is done by treating the | |
1198 | size of a @code{void} or of a function as 1. | |
1199 | ||
1200 | A consequence of this is that @code{sizeof} is also allowed on @code{void} | |
1201 | and on function types, and returns 1. | |
1202 | ||
84330467 JM |
1203 | @opindex Wpointer-arith |
1204 | The option @option{-Wpointer-arith} requests a warning if these extensions | |
c1f7febf RK |
1205 | are used. |
1206 | ||
1207 | @node Initializers | |
1208 | @section Non-Constant Initializers | |
1209 | @cindex initializers, non-constant | |
1210 | @cindex non-constant initializers | |
1211 | ||
4b404517 | 1212 | As in standard C++ and ISO C99, the elements of an aggregate initializer for an |
c1f7febf RK |
1213 | automatic variable are not required to be constant expressions in GNU C. |
1214 | Here is an example of an initializer with run-time varying elements: | |
1215 | ||
1216 | @example | |
1217 | foo (float f, float g) | |
1218 | @{ | |
1219 | float beat_freqs[2] = @{ f-g, f+g @}; | |
1220 | @dots{} | |
1221 | @} | |
1222 | @end example | |
1223 | ||
4b404517 JM |
1224 | @node Compound Literals |
1225 | @section Compound Literals | |
c1f7febf RK |
1226 | @cindex constructor expressions |
1227 | @cindex initializations in expressions | |
1228 | @cindex structures, constructor expression | |
1229 | @cindex expressions, constructor | |
4b404517 JM |
1230 | @cindex compound literals |
1231 | @c The GNU C name for what C99 calls compound literals was "constructor expressions". | |
c1f7febf | 1232 | |
4b404517 | 1233 | ISO C99 supports compound literals. A compound literal looks like |
c1f7febf RK |
1234 | a cast containing an initializer. Its value is an object of the |
1235 | type specified in the cast, containing the elements specified in | |
4b404517 JM |
1236 | the initializer. (GCC does not yet implement the full ISO C99 semantics |
1237 | for compound literals.) As an extension, GCC supports compound literals | |
1238 | in C89 mode and in C++. | |
c1f7febf RK |
1239 | |
1240 | Usually, the specified type is a structure. Assume that | |
1241 | @code{struct foo} and @code{structure} are declared as shown: | |
1242 | ||
1243 | @example | |
1244 | struct foo @{int a; char b[2];@} structure; | |
1245 | @end example | |
1246 | ||
1247 | @noindent | |
4b404517 | 1248 | Here is an example of constructing a @code{struct foo} with a compound literal: |
c1f7febf RK |
1249 | |
1250 | @example | |
1251 | structure = ((struct foo) @{x + y, 'a', 0@}); | |
1252 | @end example | |
1253 | ||
1254 | @noindent | |
1255 | This is equivalent to writing the following: | |
1256 | ||
1257 | @example | |
1258 | @{ | |
1259 | struct foo temp = @{x + y, 'a', 0@}; | |
1260 | structure = temp; | |
1261 | @} | |
1262 | @end example | |
1263 | ||
4b404517 | 1264 | You can also construct an array. If all the elements of the compound literal |
c1f7febf | 1265 | are (made up of) simple constant expressions, suitable for use in |
4b404517 | 1266 | initializers, then the compound literal is an lvalue and can be coerced to a |
c1f7febf RK |
1267 | pointer to its first element, as shown here: |
1268 | ||
1269 | @example | |
1270 | char **foo = (char *[]) @{ "x", "y", "z" @}; | |
1271 | @end example | |
1272 | ||
4b404517 JM |
1273 | Array compound literals whose elements are not simple constants are |
1274 | not very useful, because the compound literal is not an lvalue; ISO C99 | |
1275 | specifies that it is, being a temporary object with automatic storage | |
1276 | duration associated with the enclosing block, but GCC does not yet | |
1277 | implement this. There are currently only two valid ways to use it with | |
1278 | GCC: to subscript it, or initialize | |
c1f7febf RK |
1279 | an array variable with it. The former is probably slower than a |
1280 | @code{switch} statement, while the latter does the same thing an | |
1281 | ordinary C initializer would do. Here is an example of | |
4b404517 | 1282 | subscripting an array compound literal: |
c1f7febf RK |
1283 | |
1284 | @example | |
1285 | output = ((int[]) @{ 2, x, 28 @}) [input]; | |
1286 | @end example | |
1287 | ||
4b404517 JM |
1288 | Compound literals for scalar types and union types are is |
1289 | also allowed, but then the compound literal is equivalent | |
c1f7febf RK |
1290 | to a cast. |
1291 | ||
4b404517 JM |
1292 | @node Designated Inits |
1293 | @section Designated Initializers | |
c1f7febf RK |
1294 | @cindex initializers with labeled elements |
1295 | @cindex labeled elements in initializers | |
1296 | @cindex case labels in initializers | |
4b404517 | 1297 | @cindex designated initializers |
c1f7febf | 1298 | |
26d4fec7 | 1299 | Standard C89 requires the elements of an initializer to appear in a fixed |
c1f7febf RK |
1300 | order, the same as the order of the elements in the array or structure |
1301 | being initialized. | |
1302 | ||
26d4fec7 JM |
1303 | In ISO C99 you can give the elements in any order, specifying the array |
1304 | indices or structure field names they apply to, and GNU C allows this as | |
1305 | an extension in C89 mode as well. This extension is not | |
c1f7febf RK |
1306 | implemented in GNU C++. |
1307 | ||
26d4fec7 | 1308 | To specify an array index, write |
c1f7febf RK |
1309 | @samp{[@var{index}] =} before the element value. For example, |
1310 | ||
1311 | @example | |
26d4fec7 | 1312 | int a[6] = @{ [4] = 29, [2] = 15 @}; |
c1f7febf RK |
1313 | @end example |
1314 | ||
1315 | @noindent | |
1316 | is equivalent to | |
1317 | ||
1318 | @example | |
1319 | int a[6] = @{ 0, 0, 15, 0, 29, 0 @}; | |
1320 | @end example | |
1321 | ||
1322 | @noindent | |
1323 | The index values must be constant expressions, even if the array being | |
1324 | initialized is automatic. | |
1325 | ||
26d4fec7 JM |
1326 | An alternative syntax for this which has been obsolete since GCC 2.5 but |
1327 | GCC still accepts is to write @samp{[@var{index}]} before the element | |
1328 | value, with no @samp{=}. | |
1329 | ||
c1f7febf | 1330 | To initialize a range of elements to the same value, write |
26d4fec7 JM |
1331 | @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU |
1332 | extension. For example, | |
c1f7febf RK |
1333 | |
1334 | @example | |
1335 | int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @}; | |
1336 | @end example | |
1337 | ||
8b6a5902 JJ |
1338 | @noindent |
1339 | If the value in it has side-effects, the side-effects will happen only once, | |
1340 | not for each initialized field by the range initializer. | |
1341 | ||
c1f7febf RK |
1342 | @noindent |
1343 | Note that the length of the array is the highest value specified | |
1344 | plus one. | |
1345 | ||
1346 | In a structure initializer, specify the name of a field to initialize | |
26d4fec7 | 1347 | with @samp{.@var{fieldname} =} before the element value. For example, |
c1f7febf RK |
1348 | given the following structure, |
1349 | ||
1350 | @example | |
1351 | struct point @{ int x, y; @}; | |
1352 | @end example | |
1353 | ||
1354 | @noindent | |
1355 | the following initialization | |
1356 | ||
1357 | @example | |
26d4fec7 | 1358 | struct point p = @{ .y = yvalue, .x = xvalue @}; |
c1f7febf RK |
1359 | @end example |
1360 | ||
1361 | @noindent | |
1362 | is equivalent to | |
1363 | ||
1364 | @example | |
1365 | struct point p = @{ xvalue, yvalue @}; | |
1366 | @end example | |
1367 | ||
26d4fec7 JM |
1368 | Another syntax which has the same meaning, obsolete since GCC 2.5, is |
1369 | @samp{@var{fieldname}:}, as shown here: | |
c1f7febf RK |
1370 | |
1371 | @example | |
26d4fec7 | 1372 | struct point p = @{ y: yvalue, x: xvalue @}; |
c1f7febf RK |
1373 | @end example |
1374 | ||
4b404517 JM |
1375 | @cindex designators |
1376 | The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a | |
1377 | @dfn{designator}. You can also use a designator (or the obsolete colon | |
1378 | syntax) when initializing a union, to specify which element of the union | |
1379 | should be used. For example, | |
c1f7febf RK |
1380 | |
1381 | @example | |
1382 | union foo @{ int i; double d; @}; | |
1383 | ||
26d4fec7 | 1384 | union foo f = @{ .d = 4 @}; |
c1f7febf RK |
1385 | @end example |
1386 | ||
1387 | @noindent | |
1388 | will convert 4 to a @code{double} to store it in the union using | |
1389 | the second element. By contrast, casting 4 to type @code{union foo} | |
1390 | would store it into the union as the integer @code{i}, since it is | |
1391 | an integer. (@xref{Cast to Union}.) | |
1392 | ||
1393 | You can combine this technique of naming elements with ordinary C | |
1394 | initialization of successive elements. Each initializer element that | |
4b404517 | 1395 | does not have a designator applies to the next consecutive element of the |
c1f7febf RK |
1396 | array or structure. For example, |
1397 | ||
1398 | @example | |
1399 | int a[6] = @{ [1] = v1, v2, [4] = v4 @}; | |
1400 | @end example | |
1401 | ||
1402 | @noindent | |
1403 | is equivalent to | |
1404 | ||
1405 | @example | |
1406 | int a[6] = @{ 0, v1, v2, 0, v4, 0 @}; | |
1407 | @end example | |
1408 | ||
1409 | Labeling the elements of an array initializer is especially useful | |
1410 | when the indices are characters or belong to an @code{enum} type. | |
1411 | For example: | |
1412 | ||
1413 | @example | |
1414 | int whitespace[256] | |
1415 | = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1, | |
1416 | ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @}; | |
1417 | @end example | |
1418 | ||
4b404517 | 1419 | @cindex designator lists |
26d4fec7 | 1420 | You can also write a series of @samp{.@var{fieldname}} and |
4b404517 | 1421 | @samp{[@var{index}]} designators before an @samp{=} to specify a |
26d4fec7 JM |
1422 | nested subobject to initialize; the list is taken relative to the |
1423 | subobject corresponding to the closest surrounding brace pair. For | |
1424 | example, with the @samp{struct point} declaration above: | |
1425 | ||
1426 | @example | |
1427 | struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @}; | |
1428 | @end example | |
1429 | ||
8b6a5902 JJ |
1430 | @noindent |
1431 | If the same field is initialized multiple times, it will have value from | |
1432 | the last initialization. If any such overridden initialization has | |
1433 | side-effect, it is unspecified whether the side-effect happens or not. | |
1434 | Currently, gcc will discard them and issue a warning. | |
1435 | ||
c1f7febf RK |
1436 | @node Case Ranges |
1437 | @section Case Ranges | |
1438 | @cindex case ranges | |
1439 | @cindex ranges in case statements | |
1440 | ||
1441 | You can specify a range of consecutive values in a single @code{case} label, | |
1442 | like this: | |
1443 | ||
1444 | @example | |
1445 | case @var{low} ... @var{high}: | |
1446 | @end example | |
1447 | ||
1448 | @noindent | |
1449 | This has the same effect as the proper number of individual @code{case} | |
1450 | labels, one for each integer value from @var{low} to @var{high}, inclusive. | |
1451 | ||
1452 | This feature is especially useful for ranges of ASCII character codes: | |
1453 | ||
1454 | @example | |
1455 | case 'A' ... 'Z': | |
1456 | @end example | |
1457 | ||
1458 | @strong{Be careful:} Write spaces around the @code{...}, for otherwise | |
1459 | it may be parsed wrong when you use it with integer values. For example, | |
1460 | write this: | |
1461 | ||
1462 | @example | |
1463 | case 1 ... 5: | |
1464 | @end example | |
1465 | ||
1466 | @noindent | |
1467 | rather than this: | |
1468 | ||
1469 | @example | |
1470 | case 1...5: | |
1471 | @end example | |
1472 | ||
1473 | @node Cast to Union | |
1474 | @section Cast to a Union Type | |
1475 | @cindex cast to a union | |
1476 | @cindex union, casting to a | |
1477 | ||
1478 | A cast to union type is similar to other casts, except that the type | |
1479 | specified is a union type. You can specify the type either with | |
1480 | @code{union @var{tag}} or with a typedef name. A cast to union is actually | |
1481 | a constructor though, not a cast, and hence does not yield an lvalue like | |
4b404517 | 1482 | normal casts. (@xref{Compound Literals}.) |
c1f7febf RK |
1483 | |
1484 | The types that may be cast to the union type are those of the members | |
1485 | of the union. Thus, given the following union and variables: | |
1486 | ||
1487 | @example | |
1488 | union foo @{ int i; double d; @}; | |
1489 | int x; | |
1490 | double y; | |
1491 | @end example | |
1492 | ||
1493 | @noindent | |
1494 | both @code{x} and @code{y} can be cast to type @code{union} foo. | |
1495 | ||
1496 | Using the cast as the right-hand side of an assignment to a variable of | |
1497 | union type is equivalent to storing in a member of the union: | |
1498 | ||
1499 | @example | |
1500 | union foo u; | |
1501 | @dots{} | |
1502 | u = (union foo) x @equiv{} u.i = x | |
1503 | u = (union foo) y @equiv{} u.d = y | |
1504 | @end example | |
1505 | ||
1506 | You can also use the union cast as a function argument: | |
1507 | ||
1508 | @example | |
1509 | void hack (union foo); | |
1510 | @dots{} | |
1511 | hack ((union foo) x); | |
1512 | @end example | |
1513 | ||
4b404517 JM |
1514 | @node Mixed Declarations |
1515 | @section Mixed Declarations and Code | |
1516 | @cindex mixed declarations and code | |
1517 | @cindex declarations, mixed with code | |
1518 | @cindex code, mixed with declarations | |
1519 | ||
1520 | ISO C99 and ISO C++ allow declarations and code to be freely mixed | |
1521 | within compound statements. As an extension, GCC also allows this in | |
1522 | C89 mode. For example, you could do: | |
1523 | ||
1524 | @example | |
1525 | int i; | |
1526 | @dots{} | |
1527 | i++; | |
1528 | int j = i + 2; | |
1529 | @end example | |
1530 | ||
1531 | Each identifier is visible from where it is declared until the end of | |
1532 | the enclosing block. | |
1533 | ||
c1f7febf RK |
1534 | @node Function Attributes |
1535 | @section Declaring Attributes of Functions | |
1536 | @cindex function attributes | |
1537 | @cindex declaring attributes of functions | |
1538 | @cindex functions that never return | |
1539 | @cindex functions that have no side effects | |
1540 | @cindex functions in arbitrary sections | |
2a59078d | 1541 | @cindex functions that behave like malloc |
c1f7febf RK |
1542 | @cindex @code{volatile} applied to function |
1543 | @cindex @code{const} applied to function | |
26f6672d | 1544 | @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments |
c1f7febf RK |
1545 | @cindex functions that are passed arguments in registers on the 386 |
1546 | @cindex functions that pop the argument stack on the 386 | |
1547 | @cindex functions that do not pop the argument stack on the 386 | |
1548 | ||
1549 | In GNU C, you declare certain things about functions called in your program | |
1550 | which help the compiler optimize function calls and check your code more | |
1551 | carefully. | |
1552 | ||
1553 | The keyword @code{__attribute__} allows you to specify special | |
1554 | attributes when making a declaration. This keyword is followed by an | |
9f1bbeaa JM |
1555 | attribute specification inside double parentheses. Fourteen attributes, |
1556 | @code{noreturn}, @code{pure}, @code{const}, @code{format}, | |
1557 | @code{format_arg}, @code{no_instrument_function}, @code{section}, | |
1558 | @code{constructor}, @code{destructor}, @code{unused}, @code{weak}, | |
1559 | @code{malloc}, @code{alias} and @code{no_check_memory_usage} are | |
1560 | currently defined for functions. Several other attributes are defined | |
1561 | for functions on particular target systems. Other attributes, including | |
c1f7febf RK |
1562 | @code{section} are supported for variables declarations (@pxref{Variable |
1563 | Attributes}) and for types (@pxref{Type Attributes}). | |
1564 | ||
1565 | You may also specify attributes with @samp{__} preceding and following | |
1566 | each keyword. This allows you to use them in header files without | |
1567 | being concerned about a possible macro of the same name. For example, | |
1568 | you may use @code{__noreturn__} instead of @code{noreturn}. | |
1569 | ||
2c5e91d2 JM |
1570 | @xref{Attribute Syntax}, for details of the exact syntax for using |
1571 | attributes. | |
1572 | ||
c1f7febf RK |
1573 | @table @code |
1574 | @cindex @code{noreturn} function attribute | |
1575 | @item noreturn | |
1576 | A few standard library functions, such as @code{abort} and @code{exit}, | |
f0523f02 | 1577 | cannot return. GCC knows this automatically. Some programs define |
c1f7febf RK |
1578 | their own functions that never return. You can declare them |
1579 | @code{noreturn} to tell the compiler this fact. For example, | |
1580 | ||
1581 | @smallexample | |
1582 | void fatal () __attribute__ ((noreturn)); | |
1583 | ||
1584 | void | |
1585 | fatal (@dots{}) | |
1586 | @{ | |
1587 | @dots{} /* @r{Print error message.} */ @dots{} | |
1588 | exit (1); | |
1589 | @} | |
1590 | @end smallexample | |
1591 | ||
1592 | The @code{noreturn} keyword tells the compiler to assume that | |
1593 | @code{fatal} cannot return. It can then optimize without regard to what | |
1594 | would happen if @code{fatal} ever did return. This makes slightly | |
1595 | better code. More importantly, it helps avoid spurious warnings of | |
1596 | uninitialized variables. | |
1597 | ||
1598 | Do not assume that registers saved by the calling function are | |
1599 | restored before calling the @code{noreturn} function. | |
1600 | ||
1601 | It does not make sense for a @code{noreturn} function to have a return | |
1602 | type other than @code{void}. | |
1603 | ||
f0523f02 | 1604 | The attribute @code{noreturn} is not implemented in GCC versions |
c1f7febf RK |
1605 | earlier than 2.5. An alternative way to declare that a function does |
1606 | not return, which works in the current version and in some older | |
1607 | versions, is as follows: | |
1608 | ||
1609 | @smallexample | |
1610 | typedef void voidfn (); | |
1611 | ||
1612 | volatile voidfn fatal; | |
1613 | @end smallexample | |
1614 | ||
2a8f6b90 JH |
1615 | @cindex @code{pure} function attribute |
1616 | @item pure | |
1617 | Many functions have no effects except the return value and their | |
d4047e24 | 1618 | return value depends only on the parameters and/or global variables. |
2a8f6b90 | 1619 | Such a function can be subject |
c1f7febf RK |
1620 | to common subexpression elimination and loop optimization just as an |
1621 | arithmetic operator would be. These functions should be declared | |
2a8f6b90 | 1622 | with the attribute @code{pure}. For example, |
c1f7febf RK |
1623 | |
1624 | @smallexample | |
2a8f6b90 | 1625 | int square (int) __attribute__ ((pure)); |
c1f7febf RK |
1626 | @end smallexample |
1627 | ||
1628 | @noindent | |
1629 | says that the hypothetical function @code{square} is safe to call | |
1630 | fewer times than the program says. | |
1631 | ||
2a8f6b90 JH |
1632 | Some of common examples of pure functions are @code{strlen} or @code{memcmp}. |
1633 | Interesting non-pure functions are functions with infinite loops or those | |
1634 | depending on volatile memory or other system resource, that may change between | |
2a59078d | 1635 | two consecutive calls (such as @code{feof} in a multithreading environment). |
2a8f6b90 | 1636 | |
f0523f02 | 1637 | The attribute @code{pure} is not implemented in GCC versions earlier |
2a8f6b90 JH |
1638 | than 2.96. |
1639 | @cindex @code{const} function attribute | |
1640 | @item const | |
1641 | Many functions do not examine any values except their arguments, and | |
1642 | have no effects except the return value. Basically this is just slightly | |
84330467 | 1643 | more strict class than the @code{pure} attribute above, since function is not |
2a59078d | 1644 | allowed to read global memory. |
2a8f6b90 JH |
1645 | |
1646 | @cindex pointer arguments | |
1647 | Note that a function that has pointer arguments and examines the data | |
1648 | pointed to must @emph{not} be declared @code{const}. Likewise, a | |
1649 | function that calls a non-@code{const} function usually must not be | |
1650 | @code{const}. It does not make sense for a @code{const} function to | |
1651 | return @code{void}. | |
1652 | ||
f0523f02 | 1653 | The attribute @code{const} is not implemented in GCC versions earlier |
c1f7febf RK |
1654 | than 2.5. An alternative way to declare that a function has no side |
1655 | effects, which works in the current version and in some older versions, | |
1656 | is as follows: | |
1657 | ||
1658 | @smallexample | |
1659 | typedef int intfn (); | |
1660 | ||
1661 | extern const intfn square; | |
1662 | @end smallexample | |
1663 | ||
1664 | This approach does not work in GNU C++ from 2.6.0 on, since the language | |
1665 | specifies that the @samp{const} must be attached to the return value. | |
1666 | ||
c1f7febf RK |
1667 | |
1668 | @item format (@var{archetype}, @var{string-index}, @var{first-to-check}) | |
1669 | @cindex @code{format} function attribute | |
84330467 | 1670 | @opindex Wformat |
bb72a084 | 1671 | The @code{format} attribute specifies that a function takes @code{printf}, |
26f6672d JM |
1672 | @code{scanf}, @code{strftime} or @code{strfmon} style arguments which |
1673 | should be type-checked against a format string. For example, the | |
1674 | declaration: | |
c1f7febf RK |
1675 | |
1676 | @smallexample | |
1677 | extern int | |
1678 | my_printf (void *my_object, const char *my_format, ...) | |
1679 | __attribute__ ((format (printf, 2, 3))); | |
1680 | @end smallexample | |
1681 | ||
1682 | @noindent | |
1683 | causes the compiler to check the arguments in calls to @code{my_printf} | |
1684 | for consistency with the @code{printf} style format string argument | |
1685 | @code{my_format}. | |
1686 | ||
1687 | The parameter @var{archetype} determines how the format string is | |
26f6672d JM |
1688 | interpreted, and should be @code{printf}, @code{scanf}, @code{strftime} |
1689 | or @code{strfmon}. (You can also use @code{__printf__}, | |
1690 | @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The | |
c1f7febf RK |
1691 | parameter @var{string-index} specifies which argument is the format |
1692 | string argument (starting from 1), while @var{first-to-check} is the | |
1693 | number of the first argument to check against the format string. For | |
1694 | functions where the arguments are not available to be checked (such as | |
1695 | @code{vprintf}), specify the third parameter as zero. In this case the | |
b722c82c JM |
1696 | compiler only checks the format string for consistency. For |
1697 | @code{strftime} formats, the third parameter is required to be zero. | |
c1f7febf RK |
1698 | |
1699 | In the example above, the format string (@code{my_format}) is the second | |
1700 | argument of the function @code{my_print}, and the arguments to check | |
1701 | start with the third argument, so the correct parameters for the format | |
1702 | attribute are 2 and 3. | |
1703 | ||
84330467 | 1704 | @opindex ffreestanding |
c1f7febf | 1705 | The @code{format} attribute allows you to identify your own functions |
f0523f02 | 1706 | which take format strings as arguments, so that GCC can check the |
b722c82c | 1707 | calls to these functions for errors. The compiler always (unless |
84330467 | 1708 | @option{-ffreestanding} is used) checks formats |
b722c82c | 1709 | for the standard library functions @code{printf}, @code{fprintf}, |
bb72a084 | 1710 | @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime}, |
c1f7febf | 1711 | @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such |
84330467 | 1712 | warnings are requested (using @option{-Wformat}), so there is no need to |
b722c82c JM |
1713 | modify the header file @file{stdio.h}. In C99 mode, the functions |
1714 | @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and | |
26f6672d JM |
1715 | @code{vsscanf} are also checked. Except in strictly conforming C |
1716 | standard modes, the X/Open function @code{strfmon} is also checked. | |
b722c82c | 1717 | @xref{C Dialect Options,,Options Controlling C Dialect}. |
c1f7febf RK |
1718 | |
1719 | @item format_arg (@var{string-index}) | |
1720 | @cindex @code{format_arg} function attribute | |
84330467 | 1721 | @opindex Wformat-nonliteral |
26f6672d JM |
1722 | The @code{format_arg} attribute specifies that a function takes a format |
1723 | string for a @code{printf}, @code{scanf}, @code{strftime} or | |
1724 | @code{strfmon} style function and modifies it (for example, to translate | |
1725 | it into another language), so the result can be passed to a | |
1726 | @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style | |
1727 | function (with the remaining arguments to the format function the same | |
1728 | as they would have been for the unmodified string). For example, the | |
1729 | declaration: | |
c1f7febf RK |
1730 | |
1731 | @smallexample | |
1732 | extern char * | |
1733 | my_dgettext (char *my_domain, const char *my_format) | |
1734 | __attribute__ ((format_arg (2))); | |
1735 | @end smallexample | |
1736 | ||
1737 | @noindent | |
26f6672d JM |
1738 | causes the compiler to check the arguments in calls to a @code{printf}, |
1739 | @code{scanf}, @code{strftime} or @code{strfmon} type function, whose | |
1740 | format string argument is a call to the @code{my_dgettext} function, for | |
1741 | consistency with the format string argument @code{my_format}. If the | |
1742 | @code{format_arg} attribute had not been specified, all the compiler | |
1743 | could tell in such calls to format functions would be that the format | |
1744 | string argument is not constant; this would generate a warning when | |
84330467 | 1745 | @option{-Wformat-nonliteral} is used, but the calls could not be checked |
26f6672d | 1746 | without the attribute. |
c1f7febf RK |
1747 | |
1748 | The parameter @var{string-index} specifies which argument is the format | |
1749 | string argument (starting from 1). | |
1750 | ||
1751 | The @code{format-arg} attribute allows you to identify your own | |
f0523f02 | 1752 | functions which modify format strings, so that GCC can check the |
26f6672d JM |
1753 | calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} |
1754 | type function whose operands are a call to one of your own function. | |
1755 | The compiler always treats @code{gettext}, @code{dgettext}, and | |
1756 | @code{dcgettext} in this manner except when strict ISO C support is | |
84330467 JM |
1757 | requested by @option{-ansi} or an appropriate @option{-std} option, or |
1758 | @option{-ffreestanding} is used. @xref{C Dialect Options,,Options | |
26f6672d | 1759 | Controlling C Dialect}. |
c1f7febf | 1760 | |
07417085 KR |
1761 | @item no_instrument_function |
1762 | @cindex @code{no_instrument_function} function attribute | |
84330467 JM |
1763 | @opindex finstrument-functions |
1764 | If @option{-finstrument-functions} is given, profiling function calls will | |
07417085 KR |
1765 | be generated at entry and exit of most user-compiled functions. |
1766 | Functions with this attribute will not be so instrumented. | |
1767 | ||
84330467 | 1768 | @item section ("@var{section-name}") |
c1f7febf RK |
1769 | @cindex @code{section} function attribute |
1770 | Normally, the compiler places the code it generates in the @code{text} section. | |
1771 | Sometimes, however, you need additional sections, or you need certain | |
1772 | particular functions to appear in special sections. The @code{section} | |
1773 | attribute specifies that a function lives in a particular section. | |
1774 | For example, the declaration: | |
1775 | ||
1776 | @smallexample | |
1777 | extern void foobar (void) __attribute__ ((section ("bar"))); | |
1778 | @end smallexample | |
1779 | ||
1780 | @noindent | |
1781 | puts the function @code{foobar} in the @code{bar} section. | |
1782 | ||
1783 | Some file formats do not support arbitrary sections so the @code{section} | |
1784 | attribute is not available on all platforms. | |
1785 | If you need to map the entire contents of a module to a particular | |
1786 | section, consider using the facilities of the linker instead. | |
1787 | ||
1788 | @item constructor | |
1789 | @itemx destructor | |
1790 | @cindex @code{constructor} function attribute | |
1791 | @cindex @code{destructor} function attribute | |
1792 | The @code{constructor} attribute causes the function to be called | |
1793 | automatically before execution enters @code{main ()}. Similarly, the | |
1794 | @code{destructor} attribute causes the function to be called | |
1795 | automatically after @code{main ()} has completed or @code{exit ()} has | |
1796 | been called. Functions with these attributes are useful for | |
1797 | initializing data that will be used implicitly during the execution of | |
1798 | the program. | |
1799 | ||
2147b154 | 1800 | These attributes are not currently implemented for Objective-C. |
c1f7febf RK |
1801 | |
1802 | @item unused | |
1803 | This attribute, attached to a function, means that the function is meant | |
f0523f02 | 1804 | to be possibly unused. GCC will not produce a warning for this |
c1f7febf RK |
1805 | function. GNU C++ does not currently support this attribute as |
1806 | definitions without parameters are valid in C++. | |
1807 | ||
1808 | @item weak | |
1809 | @cindex @code{weak} attribute | |
1810 | The @code{weak} attribute causes the declaration to be emitted as a weak | |
1811 | symbol rather than a global. This is primarily useful in defining | |
1812 | library functions which can be overridden in user code, though it can | |
1813 | also be used with non-function declarations. Weak symbols are supported | |
1814 | for ELF targets, and also for a.out targets when using the GNU assembler | |
1815 | and linker. | |
1816 | ||
140592a0 AG |
1817 | @item malloc |
1818 | @cindex @code{malloc} attribute | |
1819 | The @code{malloc} attribute is used to tell the compiler that a function | |
1820 | may be treated as if it were the malloc function. The compiler assumes | |
1821 | that calls to malloc result in a pointers that cannot alias anything. | |
1822 | This will often improve optimization. | |
1823 | ||
84330467 | 1824 | @item alias ("@var{target}") |
c1f7febf RK |
1825 | @cindex @code{alias} attribute |
1826 | The @code{alias} attribute causes the declaration to be emitted as an | |
1827 | alias for another symbol, which must be specified. For instance, | |
1828 | ||
1829 | @smallexample | |
1830 | void __f () @{ /* do something */; @} | |
1831 | void f () __attribute__ ((weak, alias ("__f"))); | |
1832 | @end smallexample | |
1833 | ||
1834 | declares @samp{f} to be a weak alias for @samp{__f}. In C++, the | |
1835 | mangled name for the target must be used. | |
1836 | ||
af3e86c2 RK |
1837 | Not all target machines support this attribute. |
1838 | ||
7d384cc0 KR |
1839 | @item no_check_memory_usage |
1840 | @cindex @code{no_check_memory_usage} function attribute | |
84330467 | 1841 | @opindex fcheck-memory-usage |
f0523f02 | 1842 | The @code{no_check_memory_usage} attribute causes GCC to omit checks |
c5c76735 | 1843 | of memory references when it generates code for that function. Normally |
84330467 | 1844 | if you specify @option{-fcheck-memory-usage} (see @pxref{Code Gen |
f0523f02 | 1845 | Options}), GCC generates calls to support routines before most memory |
c5c76735 | 1846 | accesses to permit support code to record usage and detect uses of |
f0523f02 | 1847 | uninitialized or unallocated storage. Since GCC cannot handle |
c5c76735 | 1848 | @code{asm} statements properly they are not allowed in such functions. |
f0523f02 | 1849 | If you declare a function with this attribute, GCC will not generate |
7d384cc0 | 1850 | memory checking code for that function, permitting the use of @code{asm} |
c5c76735 JL |
1851 | statements without having to compile that function with different |
1852 | options. This also allows you to write support routines of your own if | |
1853 | you wish, without getting infinite recursion if they get compiled with | |
630d3d5a | 1854 | @option{-fcheck-memory-usage}. |
7d384cc0 | 1855 | |
c1f7febf RK |
1856 | @item regparm (@var{number}) |
1857 | @cindex functions that are passed arguments in registers on the 386 | |
1858 | On the Intel 386, the @code{regparm} attribute causes the compiler to | |
84330467 JM |
1859 | pass up to @var{number} integer arguments in registers EAX, |
1860 | EDX, and ECX instead of on the stack. Functions that take a | |
c1f7febf RK |
1861 | variable number of arguments will continue to be passed all of their |
1862 | arguments on the stack. | |
1863 | ||
1864 | @item stdcall | |
1865 | @cindex functions that pop the argument stack on the 386 | |
1866 | On the Intel 386, the @code{stdcall} attribute causes the compiler to | |
1867 | assume that the called function will pop off the stack space used to | |
1868 | pass arguments, unless it takes a variable number of arguments. | |
1869 | ||
1870 | The PowerPC compiler for Windows NT currently ignores the @code{stdcall} | |
1871 | attribute. | |
1872 | ||
1873 | @item cdecl | |
1874 | @cindex functions that do pop the argument stack on the 386 | |
84330467 | 1875 | @opindex mrtd |
c1f7febf RK |
1876 | On the Intel 386, the @code{cdecl} attribute causes the compiler to |
1877 | assume that the calling function will pop off the stack space used to | |
1878 | pass arguments. This is | |
84330467 | 1879 | useful to override the effects of the @option{-mrtd} switch. |
c1f7febf RK |
1880 | |
1881 | The PowerPC compiler for Windows NT currently ignores the @code{cdecl} | |
1882 | attribute. | |
1883 | ||
1884 | @item longcall | |
1885 | @cindex functions called via pointer on the RS/6000 and PowerPC | |
1886 | On the RS/6000 and PowerPC, the @code{longcall} attribute causes the | |
1887 | compiler to always call the function via a pointer, so that functions | |
1888 | which reside further than 64 megabytes (67,108,864 bytes) from the | |
1889 | current location can be called. | |
1890 | ||
c27ba912 DM |
1891 | @item long_call/short_call |
1892 | @cindex indirect calls on ARM | |
1893 | This attribute allows to specify how to call a particular function on | |
630d3d5a | 1894 | ARM. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options}) |
c27ba912 DM |
1895 | command line switch and @code{#pragma long_calls} settings. The |
1896 | @code{long_call} attribute causes the compiler to always call the | |
1897 | function by first loading its address into a register and then using the | |
1898 | contents of that register. The @code{short_call} attribute always places | |
1899 | the offset to the function from the call site into the @samp{BL} | |
1900 | instruction directly. | |
1901 | ||
c1f7febf RK |
1902 | @item dllimport |
1903 | @cindex functions which are imported from a dll on PowerPC Windows NT | |
1904 | On the PowerPC running Windows NT, the @code{dllimport} attribute causes | |
1905 | the compiler to call the function via a global pointer to the function | |
1906 | pointer that is set up by the Windows NT dll library. The pointer name | |
1907 | is formed by combining @code{__imp_} and the function name. | |
1908 | ||
1909 | @item dllexport | |
1910 | @cindex functions which are exported from a dll on PowerPC Windows NT | |
1911 | On the PowerPC running Windows NT, the @code{dllexport} attribute causes | |
1912 | the compiler to provide a global pointer to the function pointer, so | |
1913 | that it can be called with the @code{dllimport} attribute. The pointer | |
1914 | name is formed by combining @code{__imp_} and the function name. | |
1915 | ||
1916 | @item exception (@var{except-func} [, @var{except-arg}]) | |
1917 | @cindex functions which specify exception handling on PowerPC Windows NT | |
1918 | On the PowerPC running Windows NT, the @code{exception} attribute causes | |
1919 | the compiler to modify the structured exception table entry it emits for | |
1920 | the declared function. The string or identifier @var{except-func} is | |
1921 | placed in the third entry of the structured exception table. It | |
1922 | represents a function, which is called by the exception handling | |
1923 | mechanism if an exception occurs. If it was specified, the string or | |
1924 | identifier @var{except-arg} is placed in the fourth entry of the | |
1925 | structured exception table. | |
1926 | ||
1927 | @item function_vector | |
1928 | @cindex calling functions through the function vector on the H8/300 processors | |
1929 | Use this option on the H8/300 and H8/300H to indicate that the specified | |
1930 | function should be called through the function vector. Calling a | |
1931 | function through the function vector will reduce code size, however; | |
1932 | the function vector has a limited size (maximum 128 entries on the H8/300 | |
1933 | and 64 entries on the H8/300H) and shares space with the interrupt vector. | |
1934 | ||
1935 | You must use GAS and GLD from GNU binutils version 2.7 or later for | |
1936 | this option to work correctly. | |
1937 | ||
6d3d9133 NC |
1938 | @item interrupt |
1939 | @cindex interrupt handler functions | |
1940 | Use this option on the ARM, AVR and M32R/D ports to indicate that the | |
1941 | specified function is an interrupt handler. The compiler will generate | |
1942 | function entry and exit sequences suitable for use in an interrupt | |
1943 | handler when this attribute is present. | |
1944 | ||
b93e3893 AO |
1945 | Note, interrupt handlers for the H8/300, H8/300H and SH processors can |
1946 | be specified via the @code{interrupt_handler} attribute. | |
6d3d9133 NC |
1947 | |
1948 | Note, on the AVR interrupts will be enabled inside the function. | |
1949 | ||
1950 | Note, for the ARM you can specify the kind of interrupt to be handled by | |
1951 | adding an optional parameter to the interrupt attribute like this: | |
1952 | ||
1953 | @smallexample | |
1954 | void f () __attribute__ ((interrupt ("IRQ"))); | |
1955 | @end smallexample | |
1956 | ||
a7cf60a2 | 1957 | Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF. |
6d3d9133 | 1958 | |
b93e3893 AO |
1959 | @item interrupt_handler |
1960 | @cindex interrupt handler functions on the H8/300 and SH processors | |
1961 | Use this option on the H8/300, H8/300H and SH to indicate that the | |
1962 | specified function is an interrupt handler. The compiler will generate | |
1963 | function entry and exit sequences suitable for use in an interrupt | |
1964 | handler when this attribute is present. | |
1965 | ||
1966 | @item sp_switch | |
1967 | Use this option on the SH to indicate an @code{interrupt_handler} | |
1968 | function should switch to an alternate stack. It expects a string | |
1969 | argument that names a global variable holding the address of the | |
1970 | alternate stack. | |
1971 | ||
1972 | @smallexample | |
1973 | void *alt_stack; | |
1974 | void f () __attribute__ ((interrupt_handler, sp_switch ("alt_stack"))); | |
1975 | @end smallexample | |
1976 | ||
1977 | @item trap_exit | |
1978 | Use this option on the SH for an @code{interrupt_handle} to return using | |
1979 | @code{trapa} instead of @code{rte}. This attribute expects an integer | |
1980 | argument specifying the trap number to be used. | |
1981 | ||
c1f7febf RK |
1982 | @item eightbit_data |
1983 | @cindex eight bit data on the H8/300 and H8/300H | |
1984 | Use this option on the H8/300 and H8/300H to indicate that the specified | |
1985 | variable should be placed into the eight bit data section. | |
1986 | The compiler will generate more efficient code for certain operations | |
1987 | on data in the eight bit data area. Note the eight bit data area is limited to | |
1988 | 256 bytes of data. | |
1989 | ||
1990 | You must use GAS and GLD from GNU binutils version 2.7 or later for | |
1991 | this option to work correctly. | |
1992 | ||
1993 | @item tiny_data | |
1994 | @cindex tiny data section on the H8/300H | |
1995 | Use this option on the H8/300H to indicate that the specified | |
1996 | variable should be placed into the tiny data section. | |
1997 | The compiler will generate more efficient code for loads and stores | |
1998 | on data in the tiny data section. Note the tiny data area is limited to | |
1999 | slightly under 32kbytes of data. | |
845da534 | 2000 | |
052a4b28 DC |
2001 | @item signal |
2002 | @cindex signal handler functions on the AVR processors | |
2003 | Use this option on the AVR to indicate that the specified | |
2004 | function is an signal handler. The compiler will generate function | |
2005 | entry and exit sequences suitable for use in an signal handler when this | |
767094dd | 2006 | attribute is present. Interrupts will be disabled inside function. |
052a4b28 DC |
2007 | |
2008 | @item naked | |
6d3d9133 NC |
2009 | @cindex function without a prologue/epilogue code |
2010 | Use this option on the ARM or AVR ports to indicate that the specified | |
2011 | function do not need prologue/epilogue sequences generated by the | |
2012 | compiler. It is up to the programmer to provide these sequences. | |
052a4b28 | 2013 | |
845da534 DE |
2014 | @item model (@var{model-name}) |
2015 | @cindex function addressability on the M32R/D | |
2016 | Use this attribute on the M32R/D to set the addressability of an object, | |
2017 | and the code generated for a function. | |
2018 | The identifier @var{model-name} is one of @code{small}, @code{medium}, | |
2019 | or @code{large}, representing each of the code models. | |
2020 | ||
2021 | Small model objects live in the lower 16MB of memory (so that their | |
2022 | addresses can be loaded with the @code{ld24} instruction), and are | |
2023 | callable with the @code{bl} instruction. | |
2024 | ||
02f52e19 | 2025 | Medium model objects may live anywhere in the 32-bit address space (the |
845da534 DE |
2026 | compiler will generate @code{seth/add3} instructions to load their addresses), |
2027 | and are callable with the @code{bl} instruction. | |
2028 | ||
02f52e19 | 2029 | Large model objects may live anywhere in the 32-bit address space (the |
845da534 DE |
2030 | compiler will generate @code{seth/add3} instructions to load their addresses), |
2031 | and may not be reachable with the @code{bl} instruction (the compiler will | |
2032 | generate the much slower @code{seth/add3/jl} instruction sequence). | |
2033 | ||
c1f7febf RK |
2034 | @end table |
2035 | ||
2036 | You can specify multiple attributes in a declaration by separating them | |
2037 | by commas within the double parentheses or by immediately following an | |
2038 | attribute declaration with another attribute declaration. | |
2039 | ||
2040 | @cindex @code{#pragma}, reason for not using | |
2041 | @cindex pragma, reason for not using | |
9f1bbeaa JM |
2042 | Some people object to the @code{__attribute__} feature, suggesting that |
2043 | ISO C's @code{#pragma} should be used instead. At the time | |
2044 | @code{__attribute__} was designed, there were two reasons for not doing | |
2045 | this. | |
c1f7febf RK |
2046 | |
2047 | @enumerate | |
2048 | @item | |
2049 | It is impossible to generate @code{#pragma} commands from a macro. | |
2050 | ||
2051 | @item | |
2052 | There is no telling what the same @code{#pragma} might mean in another | |
2053 | compiler. | |
2054 | @end enumerate | |
2055 | ||
9f1bbeaa JM |
2056 | These two reasons applied to almost any application that might have been |
2057 | proposed for @code{#pragma}. It was basically a mistake to use | |
2058 | @code{#pragma} for @emph{anything}. | |
2059 | ||
2060 | The ISO C99 standard includes @code{_Pragma}, which now allows pragmas | |
2061 | to be generated from macros. In addition, a @code{#pragma GCC} | |
2062 | namespace is now in use for GCC-specific pragmas. However, it has been | |
2063 | found convenient to use @code{__attribute__} to achieve a natural | |
2064 | attachment of attributes to their corresponding declarations, whereas | |
2065 | @code{#pragma GCC} is of use for constructs that do not naturally form | |
2066 | part of the grammar. @xref{Other Directives,,Miscellaneous | |
2067 | Preprocessing Directives, cpp, The C Preprocessor}. | |
c1f7febf | 2068 | |
2c5e91d2 JM |
2069 | @node Attribute Syntax |
2070 | @section Attribute Syntax | |
2071 | @cindex attribute syntax | |
2072 | ||
2073 | This section describes the syntax with which @code{__attribute__} may be | |
2074 | used, and the constructs to which attribute specifiers bind, for the C | |
2147b154 | 2075 | language. Some details may vary for C++ and Objective-C. Because of |
2c5e91d2 JM |
2076 | infelicities in the grammar for attributes, some forms described here |
2077 | may not be successfully parsed in all cases. | |
2078 | ||
2079 | @xref{Function Attributes}, for details of the semantics of attributes | |
2080 | applying to functions. @xref{Variable Attributes}, for details of the | |
2081 | semantics of attributes applying to variables. @xref{Type Attributes}, | |
2082 | for details of the semantics of attributes applying to structure, union | |
2083 | and enumerated types. | |
2084 | ||
2085 | An @dfn{attribute specifier} is of the form | |
2086 | @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list} | |
2087 | is a possibly empty comma-separated sequence of @dfn{attributes}, where | |
2088 | each attribute is one of the following: | |
2089 | ||
2090 | @itemize @bullet | |
2091 | @item | |
2092 | Empty. Empty attributes are ignored. | |
2093 | ||
2094 | @item | |
2095 | A word (which may be an identifier such as @code{unused}, or a reserved | |
2096 | word such as @code{const}). | |
2097 | ||
2098 | @item | |
2099 | A word, followed by, in parentheses, parameters for the attribute. | |
2100 | These parameters take one of the following forms: | |
2101 | ||
2102 | @itemize @bullet | |
2103 | @item | |
2104 | An identifier. For example, @code{mode} attributes use this form. | |
2105 | ||
2106 | @item | |
2107 | An identifier followed by a comma and a non-empty comma-separated list | |
2108 | of expressions. For example, @code{format} attributes use this form. | |
2109 | ||
2110 | @item | |
2111 | A possibly empty comma-separated list of expressions. For example, | |
2112 | @code{format_arg} attributes use this form with the list being a single | |
2113 | integer constant expression, and @code{alias} attributes use this form | |
2114 | with the list being a single string constant. | |
2115 | @end itemize | |
2116 | @end itemize | |
2117 | ||
2118 | An @dfn{attribute specifier list} is a sequence of one or more attribute | |
2119 | specifiers, not separated by any other tokens. | |
2120 | ||
2121 | An attribute specifier list may appear after the colon following a | |
2122 | label, other than a @code{case} or @code{default} label. The only | |
2123 | attribute it makes sense to use after a label is @code{unused}. This | |
2124 | feature is intended for code generated by programs which contains labels | |
2125 | that may be unused but which is compiled with @option{-Wall}. It would | |
2126 | not normally be appropriate to use in it human-written code, though it | |
2127 | could be useful in cases where the code that jumps to the label is | |
2128 | contained within an @code{#ifdef} conditional. | |
2129 | ||
2130 | An attribute specifier list may appear as part of a @code{struct}, | |
2131 | @code{union} or @code{enum} specifier. It may go either immediately | |
2132 | after the @code{struct}, @code{union} or @code{enum} keyword, or after | |
2133 | the closing brace. It is ignored if the content of the structure, union | |
2134 | or enumerated type is not defined in the specifier in which the | |
2135 | attribute specifier list is used---that is, in usages such as | |
2136 | @code{struct __attribute__((foo)) bar} with no following opening brace. | |
2137 | Where attribute specifiers follow the closing brace, they are considered | |
2138 | to relate to the structure, union or enumerated type defined, not to any | |
2139 | enclosing declaration the type specifier appears in, and the type | |
2140 | defined is not complete until after the attribute specifiers. | |
2141 | @c Otherwise, there would be the following problems: a shift/reduce | |
2142 | @c conflict between attributes binding the the struct/union/enum and | |
2143 | @c binding to the list of specifiers/qualifiers; and "aligned" | |
2144 | @c attributes could use sizeof for the structure, but the size could be | |
2145 | @c changed later by "packed" attributes. | |
2146 | ||
2147 | Otherwise, an attribute specifier appears as part of a declaration, | |
2148 | counting declarations of unnamed parameters and type names, and relates | |
2149 | to that declaration (which may be nested in another declaration, for | |
2150 | example in the case of a parameter declaration). In future, attribute | |
2151 | specifiers in some places may however apply to a particular declarator | |
ff867905 JM |
2152 | within a declaration instead; these cases are noted below. Where an |
2153 | attribute specifier is applied to a parameter declared as a function or | |
2154 | an array, it should apply to the function or array rather than the | |
2155 | pointer to which the parameter is implicitly converted, but this is not | |
2156 | yet correctly implemented. | |
2c5e91d2 JM |
2157 | |
2158 | Any list of specifiers and qualifiers at the start of a declaration may | |
2159 | contain attribute specifiers, whether or not such a list may in that | |
2160 | context contain storage class specifiers. (Some attributes, however, | |
2161 | are essentially in the nature of storage class specifiers, and only make | |
2162 | sense where storage class specifiers may be used; for example, | |
2163 | @code{section}.) There is one necessary limitation to this syntax: the | |
2164 | first old-style parameter declaration in a function definition cannot | |
2165 | begin with an attribute specifier, because such an attribute applies to | |
2166 | the function instead by syntax described below (which, however, is not | |
2167 | yet implemented in this case). In some other cases, attribute | |
2168 | specifiers are permitted by this grammar but not yet supported by the | |
2169 | compiler. All attribute specifiers in this place relate to the | |
c771326b | 2170 | declaration as a whole. In the obsolescent usage where a type of |
2c5e91d2 JM |
2171 | @code{int} is implied by the absence of type specifiers, such a list of |
2172 | specifiers and qualifiers may be an attribute specifier list with no | |
2173 | other specifiers or qualifiers. | |
2174 | ||
2175 | An attribute specifier list may appear immediately before a declarator | |
2176 | (other than the first) in a comma-separated list of declarators in a | |
2177 | declaration of more than one identifier using a single list of | |
2178 | specifiers and qualifiers. At present, such attribute specifiers apply | |
2179 | not only to the identifier before whose declarator they appear, but to | |
2180 | all subsequent identifiers declared in that declaration, but in future | |
2181 | they may apply only to that single identifier. For example, in | |
2182 | @code{__attribute__((noreturn)) void d0 (void), | |
2183 | __attribute__((format(printf, 1, 2))) d1 (const char *, ...), d2 | |
2184 | (void)}, the @code{noreturn} attribute applies to all the functions | |
2185 | declared; the @code{format} attribute should only apply to @code{d1}, | |
2186 | but at present applies to @code{d2} as well (and so causes an error). | |
2187 | ||
2188 | An attribute specifier list may appear immediately before the comma, | |
2189 | @code{=} or semicolon terminating the declaration of an identifier other | |
2190 | than a function definition. At present, such attribute specifiers apply | |
2191 | to the declared object or function, but in future they may attach to the | |
2192 | outermost adjacent declarator. In simple cases there is no difference, | |
2193 | but, for example, in @code{void (****f)(void) | |
2194 | __attribute__((noreturn));}, at present the @code{noreturn} attribute | |
2195 | applies to @code{f}, which causes a warning since @code{f} is not a | |
2196 | function, but in future it may apply to the function @code{****f}. The | |
2197 | precise semantics of what attributes in such cases will apply to are not | |
2198 | yet specified. Where an assembler name for an object or function is | |
2199 | specified (@pxref{Asm Labels}), at present the attribute must follow the | |
2200 | @code{asm} specification; in future, attributes before the @code{asm} | |
2201 | specification may apply to the adjacent declarator, and those after it | |
2202 | to the declared object or function. | |
2203 | ||
2204 | An attribute specifier list may, in future, be permitted to appear after | |
2205 | the declarator in a function definition (before any old-style parameter | |
2206 | declarations or the function body). | |
2207 | ||
0e03329a JM |
2208 | Attribute specifiers may be mixed with type qualifiers appearing inside |
2209 | the @code{[]} of a parameter array declarator, in the C99 construct by | |
2210 | which such qualifiers are applied to the pointer to which the array is | |
2211 | implicitly converted. Such attribute specifiers apply to the pointer, | |
2212 | not to the array, but at present this is not implemented and they are | |
2213 | ignored. | |
2214 | ||
2c5e91d2 JM |
2215 | An attribute specifier list may appear at the start of a nested |
2216 | declarator. At present, there are some limitations in this usage: the | |
c771326b | 2217 | attributes apply to the identifier declared, and to all subsequent |
2c5e91d2 JM |
2218 | identifiers declared in that declaration (if it includes a |
2219 | comma-separated list of declarators), rather than to a specific | |
2220 | declarator. When attribute specifiers follow the @code{*} of a pointer | |
2221 | declarator, they must presently follow any type qualifiers present, and | |
2222 | cannot be mixed with them. The following describes intended future | |
2223 | semantics which make this syntax more useful only. It will make the | |
2224 | most sense if you are familiar with the formal specification of | |
2225 | declarators in the ISO C standard. | |
2226 | ||
2227 | Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T | |
2228 | D1}, where @code{T} contains declaration specifiers that specify a type | |
2229 | @var{Type} (such as @code{int}) and @code{D1} is a declarator that | |
2230 | contains an identifier @var{ident}. The type specified for @var{ident} | |
2231 | for derived declarators whose type does not include an attribute | |
2232 | specifier is as in the ISO C standard. | |
2233 | ||
2234 | If @code{D1} has the form @code{( @var{attribute-specifier-list} D )}, | |
2235 | and the declaration @code{T D} specifies the type | |
2236 | ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then | |
2237 | @code{T D1} specifies the type ``@var{derived-declarator-type-list} | |
2238 | @var{attribute-specifier-list} @var{Type}'' for @var{ident}. | |
2239 | ||
2240 | If @code{D1} has the form @code{* | |
2241 | @var{type-qualifier-and-attribute-specifier-list} D}, and the | |
2242 | declaration @code{T D} specifies the type | |
2243 | ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then | |
2244 | @code{T D1} specifies the type ``@var{derived-declarator-type-list} | |
2245 | @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for | |
2246 | @var{ident}. | |
2247 | ||
2248 | For example, @code{void (__attribute__((noreturn)) ****f)();} specifies | |
2249 | the type ``pointer to pointer to pointer to pointer to non-returning | |
2250 | function returning @code{void}''. As another example, @code{char | |
2251 | *__attribute__((aligned(8))) *f;} specifies the type ``pointer to | |
2252 | 8-byte-aligned pointer to @code{char}''. Note again that this describes | |
2253 | intended future semantics, not current implementation. | |
2254 | ||
c1f7febf RK |
2255 | @node Function Prototypes |
2256 | @section Prototypes and Old-Style Function Definitions | |
2257 | @cindex function prototype declarations | |
2258 | @cindex old-style function definitions | |
2259 | @cindex promotion of formal parameters | |
2260 | ||
5490d604 | 2261 | GNU C extends ISO C to allow a function prototype to override a later |
c1f7febf RK |
2262 | old-style non-prototype definition. Consider the following example: |
2263 | ||
2264 | @example | |
2265 | /* @r{Use prototypes unless the compiler is old-fashioned.} */ | |
d863830b | 2266 | #ifdef __STDC__ |
c1f7febf RK |
2267 | #define P(x) x |
2268 | #else | |
2269 | #define P(x) () | |
2270 | #endif | |
2271 | ||
2272 | /* @r{Prototype function declaration.} */ | |
2273 | int isroot P((uid_t)); | |
2274 | ||
2275 | /* @r{Old-style function definition.} */ | |
2276 | int | |
2277 | isroot (x) /* ??? lossage here ??? */ | |
2278 | uid_t x; | |
2279 | @{ | |
2280 | return x == 0; | |
2281 | @} | |
2282 | @end example | |
2283 | ||
5490d604 | 2284 | Suppose the type @code{uid_t} happens to be @code{short}. ISO C does |
c1f7febf RK |
2285 | not allow this example, because subword arguments in old-style |
2286 | non-prototype definitions are promoted. Therefore in this example the | |
2287 | function definition's argument is really an @code{int}, which does not | |
2288 | match the prototype argument type of @code{short}. | |
2289 | ||
5490d604 | 2290 | This restriction of ISO C makes it hard to write code that is portable |
c1f7febf RK |
2291 | to traditional C compilers, because the programmer does not know |
2292 | whether the @code{uid_t} type is @code{short}, @code{int}, or | |
2293 | @code{long}. Therefore, in cases like these GNU C allows a prototype | |
2294 | to override a later old-style definition. More precisely, in GNU C, a | |
2295 | function prototype argument type overrides the argument type specified | |
2296 | by a later old-style definition if the former type is the same as the | |
2297 | latter type before promotion. Thus in GNU C the above example is | |
2298 | equivalent to the following: | |
2299 | ||
2300 | @example | |
2301 | int isroot (uid_t); | |
2302 | ||
2303 | int | |
2304 | isroot (uid_t x) | |
2305 | @{ | |
2306 | return x == 0; | |
2307 | @} | |
2308 | @end example | |
2309 | ||
2310 | GNU C++ does not support old-style function definitions, so this | |
2311 | extension is irrelevant. | |
2312 | ||
2313 | @node C++ Comments | |
2314 | @section C++ Style Comments | |
2315 | @cindex // | |
2316 | @cindex C++ comments | |
2317 | @cindex comments, C++ style | |
2318 | ||
2319 | In GNU C, you may use C++ style comments, which start with @samp{//} and | |
2320 | continue until the end of the line. Many other C implementations allow | |
2321 | such comments, and they are likely to be in a future C standard. | |
2322 | However, C++ style comments are not recognized if you specify | |
84330467 JM |
2323 | @w{@option{-ansi}}, a @option{-std} option specifying a version of ISO C |
2324 | before C99, or @w{@option{-traditional}}, since they are incompatible | |
c1f7febf RK |
2325 | with traditional constructs like @code{dividend//*comment*/divisor}. |
2326 | ||
2327 | @node Dollar Signs | |
2328 | @section Dollar Signs in Identifier Names | |
2329 | @cindex $ | |
2330 | @cindex dollar signs in identifier names | |
2331 | @cindex identifier names, dollar signs in | |
2332 | ||
79188db9 RK |
2333 | In GNU C, you may normally use dollar signs in identifier names. |
2334 | This is because many traditional C implementations allow such identifiers. | |
2335 | However, dollar signs in identifiers are not supported on a few target | |
2336 | machines, typically because the target assembler does not allow them. | |
c1f7febf RK |
2337 | |
2338 | @node Character Escapes | |
2339 | @section The Character @key{ESC} in Constants | |
2340 | ||
2341 | You can use the sequence @samp{\e} in a string or character constant to | |
2342 | stand for the ASCII character @key{ESC}. | |
2343 | ||
2344 | @node Alignment | |
2345 | @section Inquiring on Alignment of Types or Variables | |
2346 | @cindex alignment | |
2347 | @cindex type alignment | |
2348 | @cindex variable alignment | |
2349 | ||
2350 | The keyword @code{__alignof__} allows you to inquire about how an object | |
2351 | is aligned, or the minimum alignment usually required by a type. Its | |
2352 | syntax is just like @code{sizeof}. | |
2353 | ||
2354 | For example, if the target machine requires a @code{double} value to be | |
2355 | aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8. | |
2356 | This is true on many RISC machines. On more traditional machine | |
2357 | designs, @code{__alignof__ (double)} is 4 or even 2. | |
2358 | ||
2359 | Some machines never actually require alignment; they allow reference to any | |
2360 | data type even at an odd addresses. For these machines, @code{__alignof__} | |
2361 | reports the @emph{recommended} alignment of a type. | |
2362 | ||
2363 | When the operand of @code{__alignof__} is an lvalue rather than a type, the | |
2364 | value is the largest alignment that the lvalue is known to have. It may | |
2365 | have this alignment as a result of its data type, or because it is part of | |
2366 | a structure and inherits alignment from that structure. For example, after | |
2367 | this declaration: | |
2368 | ||
2369 | @example | |
2370 | struct foo @{ int x; char y; @} foo1; | |
2371 | @end example | |
2372 | ||
2373 | @noindent | |
2374 | the value of @code{__alignof__ (foo1.y)} is probably 2 or 4, the same as | |
2375 | @code{__alignof__ (int)}, even though the data type of @code{foo1.y} | |
bd819a4a | 2376 | does not itself demand any alignment. |
c1f7febf | 2377 | |
9d27bffe SS |
2378 | It is an error to ask for the alignment of an incomplete type. |
2379 | ||
c1f7febf RK |
2380 | A related feature which lets you specify the alignment of an object is |
2381 | @code{__attribute__ ((aligned (@var{alignment})))}; see the following | |
2382 | section. | |
2383 | ||
2384 | @node Variable Attributes | |
2385 | @section Specifying Attributes of Variables | |
2386 | @cindex attribute of variables | |
2387 | @cindex variable attributes | |
2388 | ||
2389 | The keyword @code{__attribute__} allows you to specify special | |
2390 | attributes of variables or structure fields. This keyword is followed | |
2391 | by an attribute specification inside double parentheses. Eight | |
2392 | attributes are currently defined for variables: @code{aligned}, | |
2393 | @code{mode}, @code{nocommon}, @code{packed}, @code{section}, | |
9f1bbeaa JM |
2394 | @code{transparent_union}, @code{unused}, and @code{weak}. Some other |
2395 | attributes are defined for variables on particular target systems. Other | |
c1f7febf | 2396 | attributes are available for functions (@pxref{Function Attributes}) and |
6c0a4eab | 2397 | for types (@pxref{Type Attributes}). Other front ends might define more |
5c25e11d | 2398 | attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}). |
c1f7febf RK |
2399 | |
2400 | You may also specify attributes with @samp{__} preceding and following | |
2401 | each keyword. This allows you to use them in header files without | |
2402 | being concerned about a possible macro of the same name. For example, | |
2403 | you may use @code{__aligned__} instead of @code{aligned}. | |
2404 | ||
2c5e91d2 JM |
2405 | @xref{Attribute Syntax}, for details of the exact syntax for using |
2406 | attributes. | |
2407 | ||
c1f7febf RK |
2408 | @table @code |
2409 | @cindex @code{aligned} attribute | |
2410 | @item aligned (@var{alignment}) | |
2411 | This attribute specifies a minimum alignment for the variable or | |
2412 | structure field, measured in bytes. For example, the declaration: | |
2413 | ||
2414 | @smallexample | |
2415 | int x __attribute__ ((aligned (16))) = 0; | |
2416 | @end smallexample | |
2417 | ||
2418 | @noindent | |
2419 | causes the compiler to allocate the global variable @code{x} on a | |
2420 | 16-byte boundary. On a 68040, this could be used in conjunction with | |
2421 | an @code{asm} expression to access the @code{move16} instruction which | |
2422 | requires 16-byte aligned operands. | |
2423 | ||
2424 | You can also specify the alignment of structure fields. For example, to | |
2425 | create a double-word aligned @code{int} pair, you could write: | |
2426 | ||
2427 | @smallexample | |
2428 | struct foo @{ int x[2] __attribute__ ((aligned (8))); @}; | |
2429 | @end smallexample | |
2430 | ||
2431 | @noindent | |
2432 | This is an alternative to creating a union with a @code{double} member | |
2433 | that forces the union to be double-word aligned. | |
2434 | ||
2435 | It is not possible to specify the alignment of functions; the alignment | |
2436 | of functions is determined by the machine's requirements and cannot be | |
2437 | changed. You cannot specify alignment for a typedef name because such a | |
2438 | name is just an alias, not a distinct type. | |
2439 | ||
2440 | As in the preceding examples, you can explicitly specify the alignment | |
2441 | (in bytes) that you wish the compiler to use for a given variable or | |
2442 | structure field. Alternatively, you can leave out the alignment factor | |
2443 | and just ask the compiler to align a variable or field to the maximum | |
2444 | useful alignment for the target machine you are compiling for. For | |
2445 | example, you could write: | |
2446 | ||
2447 | @smallexample | |
2448 | short array[3] __attribute__ ((aligned)); | |
2449 | @end smallexample | |
2450 | ||
2451 | Whenever you leave out the alignment factor in an @code{aligned} attribute | |
2452 | specification, the compiler automatically sets the alignment for the declared | |
2453 | variable or field to the largest alignment which is ever used for any data | |
2454 | type on the target machine you are compiling for. Doing this can often make | |
2455 | copy operations more efficient, because the compiler can use whatever | |
2456 | instructions copy the biggest chunks of memory when performing copies to | |
2457 | or from the variables or fields that you have aligned this way. | |
2458 | ||
2459 | The @code{aligned} attribute can only increase the alignment; but you | |
2460 | can decrease it by specifying @code{packed} as well. See below. | |
2461 | ||
2462 | Note that the effectiveness of @code{aligned} attributes may be limited | |
2463 | by inherent limitations in your linker. On many systems, the linker is | |
2464 | only able to arrange for variables to be aligned up to a certain maximum | |
2465 | alignment. (For some linkers, the maximum supported alignment may | |
2466 | be very very small.) If your linker is only able to align variables | |
2467 | up to a maximum of 8 byte alignment, then specifying @code{aligned(16)} | |
2468 | in an @code{__attribute__} will still only provide you with 8 byte | |
2469 | alignment. See your linker documentation for further information. | |
2470 | ||
2471 | @item mode (@var{mode}) | |
2472 | @cindex @code{mode} attribute | |
2473 | This attribute specifies the data type for the declaration---whichever | |
2474 | type corresponds to the mode @var{mode}. This in effect lets you | |
2475 | request an integer or floating point type according to its width. | |
2476 | ||
2477 | You may also specify a mode of @samp{byte} or @samp{__byte__} to | |
2478 | indicate the mode corresponding to a one-byte integer, @samp{word} or | |
2479 | @samp{__word__} for the mode of a one-word integer, and @samp{pointer} | |
2480 | or @samp{__pointer__} for the mode used to represent pointers. | |
2481 | ||
2482 | @item nocommon | |
2483 | @cindex @code{nocommon} attribute | |
84330467 | 2484 | @opindex fno-common |
f0523f02 | 2485 | This attribute specifies requests GCC not to place a variable |
c1f7febf | 2486 | ``common'' but instead to allocate space for it directly. If you |
f0523f02 | 2487 | specify the @option{-fno-common} flag, GCC will do this for all |
c1f7febf RK |
2488 | variables. |
2489 | ||
2490 | Specifying the @code{nocommon} attribute for a variable provides an | |
2491 | initialization of zeros. A variable may only be initialized in one | |
2492 | source file. | |
2493 | ||
2494 | @item packed | |
2495 | @cindex @code{packed} attribute | |
2496 | The @code{packed} attribute specifies that a variable or structure field | |
2497 | should have the smallest possible alignment---one byte for a variable, | |
2498 | and one bit for a field, unless you specify a larger value with the | |
2499 | @code{aligned} attribute. | |
2500 | ||
2501 | Here is a structure in which the field @code{x} is packed, so that it | |
2502 | immediately follows @code{a}: | |
2503 | ||
2504 | @example | |
2505 | struct foo | |
2506 | @{ | |
2507 | char a; | |
2508 | int x[2] __attribute__ ((packed)); | |
2509 | @}; | |
2510 | @end example | |
2511 | ||
84330467 | 2512 | @item section ("@var{section-name}") |
c1f7febf RK |
2513 | @cindex @code{section} variable attribute |
2514 | Normally, the compiler places the objects it generates in sections like | |
2515 | @code{data} and @code{bss}. Sometimes, however, you need additional sections, | |
2516 | or you need certain particular variables to appear in special sections, | |
2517 | for example to map to special hardware. The @code{section} | |
2518 | attribute specifies that a variable (or function) lives in a particular | |
2519 | section. For example, this small program uses several specific section names: | |
2520 | ||
2521 | @smallexample | |
2522 | struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @}; | |
2523 | struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @}; | |
2524 | char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @}; | |
2525 | int init_data __attribute__ ((section ("INITDATA"))) = 0; | |
2526 | ||
2527 | main() | |
2528 | @{ | |
2529 | /* Initialize stack pointer */ | |
2530 | init_sp (stack + sizeof (stack)); | |
2531 | ||
2532 | /* Initialize initialized data */ | |
2533 | memcpy (&init_data, &data, &edata - &data); | |
2534 | ||
2535 | /* Turn on the serial ports */ | |
2536 | init_duart (&a); | |
2537 | init_duart (&b); | |
2538 | @} | |
2539 | @end smallexample | |
2540 | ||
2541 | @noindent | |
2542 | Use the @code{section} attribute with an @emph{initialized} definition | |
f0523f02 | 2543 | of a @emph{global} variable, as shown in the example. GCC issues |
c1f7febf RK |
2544 | a warning and otherwise ignores the @code{section} attribute in |
2545 | uninitialized variable declarations. | |
2546 | ||
2547 | You may only use the @code{section} attribute with a fully initialized | |
2548 | global definition because of the way linkers work. The linker requires | |
2549 | each object be defined once, with the exception that uninitialized | |
2550 | variables tentatively go in the @code{common} (or @code{bss}) section | |
84330467 JM |
2551 | and can be multiply ``defined''. You can force a variable to be |
2552 | initialized with the @option{-fno-common} flag or the @code{nocommon} | |
c1f7febf RK |
2553 | attribute. |
2554 | ||
2555 | Some file formats do not support arbitrary sections so the @code{section} | |
2556 | attribute is not available on all platforms. | |
2557 | If you need to map the entire contents of a module to a particular | |
2558 | section, consider using the facilities of the linker instead. | |
2559 | ||
593d3a34 MK |
2560 | @item shared |
2561 | @cindex @code{shared} variable attribute | |
02f52e19 AJ |
2562 | On Windows NT, in addition to putting variable definitions in a named |
2563 | section, the section can also be shared among all running copies of an | |
767094dd | 2564 | executable or DLL. For example, this small program defines shared data |
84330467 | 2565 | by putting it in a named section @code{shared} and marking the section |
593d3a34 MK |
2566 | shareable: |
2567 | ||
2568 | @smallexample | |
2569 | int foo __attribute__((section ("shared"), shared)) = 0; | |
2570 | ||
2571 | int | |
2572 | main() | |
2573 | @{ | |
310668e8 JM |
2574 | /* Read and write foo. All running |
2575 | copies see the same value. */ | |
593d3a34 MK |
2576 | return 0; |
2577 | @} | |
2578 | @end smallexample | |
2579 | ||
2580 | @noindent | |
2581 | You may only use the @code{shared} attribute along with @code{section} | |
02f52e19 | 2582 | attribute with a fully initialized global definition because of the way |
593d3a34 MK |
2583 | linkers work. See @code{section} attribute for more information. |
2584 | ||
2585 | The @code{shared} attribute is only available on Windows NT. | |
2586 | ||
c1f7febf RK |
2587 | @item transparent_union |
2588 | This attribute, attached to a function parameter which is a union, means | |
2589 | that the corresponding argument may have the type of any union member, | |
2590 | but the argument is passed as if its type were that of the first union | |
2591 | member. For more details see @xref{Type Attributes}. You can also use | |
2592 | this attribute on a @code{typedef} for a union data type; then it | |
2593 | applies to all function parameters with that type. | |
2594 | ||
2595 | @item unused | |
2596 | This attribute, attached to a variable, means that the variable is meant | |
f0523f02 | 2597 | to be possibly unused. GCC will not produce a warning for this |
c1f7febf RK |
2598 | variable. |
2599 | ||
2600 | @item weak | |
2601 | The @code{weak} attribute is described in @xref{Function Attributes}. | |
845da534 DE |
2602 | |
2603 | @item model (@var{model-name}) | |
2604 | @cindex variable addressability on the M32R/D | |
2605 | Use this attribute on the M32R/D to set the addressability of an object. | |
2606 | The identifier @var{model-name} is one of @code{small}, @code{medium}, | |
2607 | or @code{large}, representing each of the code models. | |
2608 | ||
2609 | Small model objects live in the lower 16MB of memory (so that their | |
2610 | addresses can be loaded with the @code{ld24} instruction). | |
2611 | ||
02f52e19 | 2612 | Medium and large model objects may live anywhere in the 32-bit address space |
845da534 DE |
2613 | (the compiler will generate @code{seth/add3} instructions to load their |
2614 | addresses). | |
2615 | ||
c1f7febf RK |
2616 | @end table |
2617 | ||
2618 | To specify multiple attributes, separate them by commas within the | |
2619 | double parentheses: for example, @samp{__attribute__ ((aligned (16), | |
2620 | packed))}. | |
2621 | ||
2622 | @node Type Attributes | |
2623 | @section Specifying Attributes of Types | |
2624 | @cindex attribute of types | |
2625 | @cindex type attributes | |
2626 | ||
2627 | The keyword @code{__attribute__} allows you to specify special | |
2628 | attributes of @code{struct} and @code{union} types when you define such | |
2629 | types. This keyword is followed by an attribute specification inside | |
9f1bbeaa JM |
2630 | double parentheses. Four attributes are currently defined for types: |
2631 | @code{aligned}, @code{packed}, @code{transparent_union}, and @code{unused}. | |
2632 | Other attributes are defined for functions (@pxref{Function Attributes}) and | |
c1f7febf RK |
2633 | for variables (@pxref{Variable Attributes}). |
2634 | ||
2635 | You may also specify any one of these attributes with @samp{__} | |
2636 | preceding and following its keyword. This allows you to use these | |
2637 | attributes in header files without being concerned about a possible | |
2638 | macro of the same name. For example, you may use @code{__aligned__} | |
2639 | instead of @code{aligned}. | |
2640 | ||
2641 | You may specify the @code{aligned} and @code{transparent_union} | |
2642 | attributes either in a @code{typedef} declaration or just past the | |
2643 | closing curly brace of a complete enum, struct or union type | |
2644 | @emph{definition} and the @code{packed} attribute only past the closing | |
2645 | brace of a definition. | |
2646 | ||
4051959b JM |
2647 | You may also specify attributes between the enum, struct or union |
2648 | tag and the name of the type rather than after the closing brace. | |
2649 | ||
2c5e91d2 JM |
2650 | @xref{Attribute Syntax}, for details of the exact syntax for using |
2651 | attributes. | |
2652 | ||
c1f7febf RK |
2653 | @table @code |
2654 | @cindex @code{aligned} attribute | |
2655 | @item aligned (@var{alignment}) | |
2656 | This attribute specifies a minimum alignment (in bytes) for variables | |
2657 | of the specified type. For example, the declarations: | |
2658 | ||
2659 | @smallexample | |
f69eecfb JL |
2660 | struct S @{ short f[3]; @} __attribute__ ((aligned (8))); |
2661 | typedef int more_aligned_int __attribute__ ((aligned (8))); | |
c1f7febf RK |
2662 | @end smallexample |
2663 | ||
2664 | @noindent | |
d863830b | 2665 | force the compiler to insure (as far as it can) that each variable whose |
c1f7febf RK |
2666 | type is @code{struct S} or @code{more_aligned_int} will be allocated and |
2667 | aligned @emph{at least} on a 8-byte boundary. On a Sparc, having all | |
2668 | variables of type @code{struct S} aligned to 8-byte boundaries allows | |
2669 | the compiler to use the @code{ldd} and @code{std} (doubleword load and | |
2670 | store) instructions when copying one variable of type @code{struct S} to | |
2671 | another, thus improving run-time efficiency. | |
2672 | ||
2673 | Note that the alignment of any given @code{struct} or @code{union} type | |
5490d604 | 2674 | is required by the ISO C standard to be at least a perfect multiple of |
c1f7febf RK |
2675 | the lowest common multiple of the alignments of all of the members of |
2676 | the @code{struct} or @code{union} in question. This means that you @emph{can} | |
2677 | effectively adjust the alignment of a @code{struct} or @code{union} | |
2678 | type by attaching an @code{aligned} attribute to any one of the members | |
2679 | of such a type, but the notation illustrated in the example above is a | |
2680 | more obvious, intuitive, and readable way to request the compiler to | |
2681 | adjust the alignment of an entire @code{struct} or @code{union} type. | |
2682 | ||
2683 | As in the preceding example, you can explicitly specify the alignment | |
2684 | (in bytes) that you wish the compiler to use for a given @code{struct} | |
2685 | or @code{union} type. Alternatively, you can leave out the alignment factor | |
2686 | and just ask the compiler to align a type to the maximum | |
2687 | useful alignment for the target machine you are compiling for. For | |
2688 | example, you could write: | |
2689 | ||
2690 | @smallexample | |
2691 | struct S @{ short f[3]; @} __attribute__ ((aligned)); | |
2692 | @end smallexample | |
2693 | ||
2694 | Whenever you leave out the alignment factor in an @code{aligned} | |
2695 | attribute specification, the compiler automatically sets the alignment | |
2696 | for the type to the largest alignment which is ever used for any data | |
2697 | type on the target machine you are compiling for. Doing this can often | |
2698 | make copy operations more efficient, because the compiler can use | |
2699 | whatever instructions copy the biggest chunks of memory when performing | |
2700 | copies to or from the variables which have types that you have aligned | |
2701 | this way. | |
2702 | ||
2703 | In the example above, if the size of each @code{short} is 2 bytes, then | |
2704 | the size of the entire @code{struct S} type is 6 bytes. The smallest | |
2705 | power of two which is greater than or equal to that is 8, so the | |
2706 | compiler sets the alignment for the entire @code{struct S} type to 8 | |
2707 | bytes. | |
2708 | ||
2709 | Note that although you can ask the compiler to select a time-efficient | |
2710 | alignment for a given type and then declare only individual stand-alone | |
2711 | objects of that type, the compiler's ability to select a time-efficient | |
2712 | alignment is primarily useful only when you plan to create arrays of | |
2713 | variables having the relevant (efficiently aligned) type. If you | |
2714 | declare or use arrays of variables of an efficiently-aligned type, then | |
2715 | it is likely that your program will also be doing pointer arithmetic (or | |
2716 | subscripting, which amounts to the same thing) on pointers to the | |
2717 | relevant type, and the code that the compiler generates for these | |
2718 | pointer arithmetic operations will often be more efficient for | |
2719 | efficiently-aligned types than for other types. | |
2720 | ||
2721 | The @code{aligned} attribute can only increase the alignment; but you | |
2722 | can decrease it by specifying @code{packed} as well. See below. | |
2723 | ||
2724 | Note that the effectiveness of @code{aligned} attributes may be limited | |
2725 | by inherent limitations in your linker. On many systems, the linker is | |
2726 | only able to arrange for variables to be aligned up to a certain maximum | |
2727 | alignment. (For some linkers, the maximum supported alignment may | |
2728 | be very very small.) If your linker is only able to align variables | |
2729 | up to a maximum of 8 byte alignment, then specifying @code{aligned(16)} | |
2730 | in an @code{__attribute__} will still only provide you with 8 byte | |
2731 | alignment. See your linker documentation for further information. | |
2732 | ||
2733 | @item packed | |
2734 | This attribute, attached to an @code{enum}, @code{struct}, or | |
2735 | @code{union} type definition, specified that the minimum required memory | |
2736 | be used to represent the type. | |
2737 | ||
84330467 | 2738 | @opindex fshort-enums |
c1f7febf RK |
2739 | Specifying this attribute for @code{struct} and @code{union} types is |
2740 | equivalent to specifying the @code{packed} attribute on each of the | |
84330467 | 2741 | structure or union members. Specifying the @option{-fshort-enums} |
c1f7febf RK |
2742 | flag on the line is equivalent to specifying the @code{packed} |
2743 | attribute on all @code{enum} definitions. | |
2744 | ||
2745 | You may only specify this attribute after a closing curly brace on an | |
1cd4bca9 BK |
2746 | @code{enum} definition, not in a @code{typedef} declaration, unless that |
2747 | declaration also contains the definition of the @code{enum}. | |
c1f7febf RK |
2748 | |
2749 | @item transparent_union | |
2750 | This attribute, attached to a @code{union} type definition, indicates | |
2751 | that any function parameter having that union type causes calls to that | |
2752 | function to be treated in a special way. | |
2753 | ||
2754 | First, the argument corresponding to a transparent union type can be of | |
2755 | any type in the union; no cast is required. Also, if the union contains | |
2756 | a pointer type, the corresponding argument can be a null pointer | |
2757 | constant or a void pointer expression; and if the union contains a void | |
2758 | pointer type, the corresponding argument can be any pointer expression. | |
2759 | If the union member type is a pointer, qualifiers like @code{const} on | |
2760 | the referenced type must be respected, just as with normal pointer | |
2761 | conversions. | |
2762 | ||
2763 | Second, the argument is passed to the function using the calling | |
2764 | conventions of first member of the transparent union, not the calling | |
2765 | conventions of the union itself. All members of the union must have the | |
2766 | same machine representation; this is necessary for this argument passing | |
2767 | to work properly. | |
2768 | ||
2769 | Transparent unions are designed for library functions that have multiple | |
2770 | interfaces for compatibility reasons. For example, suppose the | |
2771 | @code{wait} function must accept either a value of type @code{int *} to | |
2772 | comply with Posix, or a value of type @code{union wait *} to comply with | |
2773 | the 4.1BSD interface. If @code{wait}'s parameter were @code{void *}, | |
2774 | @code{wait} would accept both kinds of arguments, but it would also | |
2775 | accept any other pointer type and this would make argument type checking | |
2776 | less useful. Instead, @code{<sys/wait.h>} might define the interface | |
2777 | as follows: | |
2778 | ||
2779 | @smallexample | |
2780 | typedef union | |
2781 | @{ | |
2782 | int *__ip; | |
2783 | union wait *__up; | |
2784 | @} wait_status_ptr_t __attribute__ ((__transparent_union__)); | |
2785 | ||
2786 | pid_t wait (wait_status_ptr_t); | |
2787 | @end smallexample | |
2788 | ||
2789 | This interface allows either @code{int *} or @code{union wait *} | |
2790 | arguments to be passed, using the @code{int *} calling convention. | |
2791 | The program can call @code{wait} with arguments of either type: | |
2792 | ||
2793 | @example | |
2794 | int w1 () @{ int w; return wait (&w); @} | |
2795 | int w2 () @{ union wait w; return wait (&w); @} | |
2796 | @end example | |
2797 | ||
2798 | With this interface, @code{wait}'s implementation might look like this: | |
2799 | ||
2800 | @example | |
2801 | pid_t wait (wait_status_ptr_t p) | |
2802 | @{ | |
2803 | return waitpid (-1, p.__ip, 0); | |
2804 | @} | |
2805 | @end example | |
d863830b JL |
2806 | |
2807 | @item unused | |
2808 | When attached to a type (including a @code{union} or a @code{struct}), | |
2809 | this attribute means that variables of that type are meant to appear | |
f0523f02 | 2810 | possibly unused. GCC will not produce a warning for any variables of |
d863830b JL |
2811 | that type, even if the variable appears to do nothing. This is often |
2812 | the case with lock or thread classes, which are usually defined and then | |
2813 | not referenced, but contain constructors and destructors that have | |
956d6950 | 2814 | nontrivial bookkeeping functions. |
d863830b | 2815 | |
c1f7febf RK |
2816 | @end table |
2817 | ||
2818 | To specify multiple attributes, separate them by commas within the | |
2819 | double parentheses: for example, @samp{__attribute__ ((aligned (16), | |
2820 | packed))}. | |
2821 | ||
2822 | @node Inline | |
2823 | @section An Inline Function is As Fast As a Macro | |
2824 | @cindex inline functions | |
2825 | @cindex integrating function code | |
2826 | @cindex open coding | |
2827 | @cindex macros, inline alternative | |
2828 | ||
f0523f02 | 2829 | By declaring a function @code{inline}, you can direct GCC to |
c1f7febf RK |
2830 | integrate that function's code into the code for its callers. This |
2831 | makes execution faster by eliminating the function-call overhead; in | |
2832 | addition, if any of the actual argument values are constant, their known | |
2833 | values may permit simplifications at compile time so that not all of the | |
2834 | inline function's code needs to be included. The effect on code size is | |
2835 | less predictable; object code may be larger or smaller with function | |
2836 | inlining, depending on the particular case. Inlining of functions is an | |
2837 | optimization and it really ``works'' only in optimizing compilation. If | |
84330467 | 2838 | you don't use @option{-O}, no function is really inline. |
c1f7febf | 2839 | |
4b404517 JM |
2840 | Inline functions are included in the ISO C99 standard, but there are |
2841 | currently substantial differences between what GCC implements and what | |
2842 | the ISO C99 standard requires. | |
2843 | ||
c1f7febf RK |
2844 | To declare a function inline, use the @code{inline} keyword in its |
2845 | declaration, like this: | |
2846 | ||
2847 | @example | |
2848 | inline int | |
2849 | inc (int *a) | |
2850 | @{ | |
2851 | (*a)++; | |
2852 | @} | |
2853 | @end example | |
2854 | ||
5490d604 | 2855 | (If you are writing a header file to be included in ISO C programs, write |
c1f7febf | 2856 | @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.) |
c1f7febf | 2857 | You can also make all ``simple enough'' functions inline with the option |
84330467 | 2858 | @option{-finline-functions}. |
247b14bd | 2859 | |
84330467 | 2860 | @opindex Winline |
247b14bd RH |
2861 | Note that certain usages in a function definition can make it unsuitable |
2862 | for inline substitution. Among these usages are: use of varargs, use of | |
2863 | alloca, use of variable sized data types (@pxref{Variable Length}), | |
2864 | use of computed goto (@pxref{Labels as Values}), use of nonlocal goto, | |
84330467 | 2865 | and nested functions (@pxref{Nested Functions}). Using @option{-Winline} |
247b14bd RH |
2866 | will warn when a function marked @code{inline} could not be substituted, |
2867 | and will give the reason for the failure. | |
c1f7febf | 2868 | |
2147b154 | 2869 | Note that in C and Objective-C, unlike C++, the @code{inline} keyword |
c1f7febf RK |
2870 | does not affect the linkage of the function. |
2871 | ||
2872 | @cindex automatic @code{inline} for C++ member fns | |
2873 | @cindex @code{inline} automatic for C++ member fns | |
2874 | @cindex member fns, automatically @code{inline} | |
2875 | @cindex C++ member fns, automatically @code{inline} | |
84330467 | 2876 | @opindex fno-default-inline |
f0523f02 | 2877 | GCC automatically inlines member functions defined within the class |
c1f7febf | 2878 | body of C++ programs even if they are not explicitly declared |
84330467 | 2879 | @code{inline}. (You can override this with @option{-fno-default-inline}; |
c1f7febf RK |
2880 | @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.) |
2881 | ||
2882 | @cindex inline functions, omission of | |
84330467 | 2883 | @opindex fkeep-inline-functions |
c1f7febf RK |
2884 | When a function is both inline and @code{static}, if all calls to the |
2885 | function are integrated into the caller, and the function's address is | |
2886 | never used, then the function's own assembler code is never referenced. | |
f0523f02 | 2887 | In this case, GCC does not actually output assembler code for the |
84330467 | 2888 | function, unless you specify the option @option{-fkeep-inline-functions}. |
c1f7febf RK |
2889 | Some calls cannot be integrated for various reasons (in particular, |
2890 | calls that precede the function's definition cannot be integrated, and | |
2891 | neither can recursive calls within the definition). If there is a | |
2892 | nonintegrated call, then the function is compiled to assembler code as | |
2893 | usual. The function must also be compiled as usual if the program | |
2894 | refers to its address, because that can't be inlined. | |
2895 | ||
2896 | @cindex non-static inline function | |
2897 | When an inline function is not @code{static}, then the compiler must assume | |
2898 | that there may be calls from other source files; since a global symbol can | |
2899 | be defined only once in any program, the function must not be defined in | |
2900 | the other source files, so the calls therein cannot be integrated. | |
2901 | Therefore, a non-@code{static} inline function is always compiled on its | |
2902 | own in the usual fashion. | |
2903 | ||
2904 | If you specify both @code{inline} and @code{extern} in the function | |
2905 | definition, then the definition is used only for inlining. In no case | |
2906 | is the function compiled on its own, not even if you refer to its | |
2907 | address explicitly. Such an address becomes an external reference, as | |
2908 | if you had only declared the function, and had not defined it. | |
2909 | ||
2910 | This combination of @code{inline} and @code{extern} has almost the | |
2911 | effect of a macro. The way to use it is to put a function definition in | |
2912 | a header file with these keywords, and put another copy of the | |
2913 | definition (lacking @code{inline} and @code{extern}) in a library file. | |
2914 | The definition in the header file will cause most calls to the function | |
2915 | to be inlined. If any uses of the function remain, they will refer to | |
2916 | the single copy in the library. | |
2917 | ||
4b404517 JM |
2918 | For future compatibility with when GCC implements ISO C99 semantics for |
2919 | inline functions, it is best to use @code{static inline} only. (The | |
2920 | existing semantics will remain available when @option{-std=gnu89} is | |
2921 | specified, but eventually the default will be @option{-std=gnu99} and | |
2922 | that will implement the C99 semantics, though it does not do so yet.) | |
2923 | ||
f0523f02 | 2924 | GCC does not inline any functions when not optimizing. It is not |
c1f7febf RK |
2925 | clear whether it is better to inline or not, in this case, but we found |
2926 | that a correct implementation when not optimizing was difficult. So we | |
2927 | did the easy thing, and turned it off. | |
2928 | ||
2929 | @node Extended Asm | |
2930 | @section Assembler Instructions with C Expression Operands | |
2931 | @cindex extended @code{asm} | |
2932 | @cindex @code{asm} expressions | |
2933 | @cindex assembler instructions | |
2934 | @cindex registers | |
2935 | ||
c85f7c16 JL |
2936 | In an assembler instruction using @code{asm}, you can specify the |
2937 | operands of the instruction using C expressions. This means you need not | |
2938 | guess which registers or memory locations will contain the data you want | |
c1f7febf RK |
2939 | to use. |
2940 | ||
c85f7c16 JL |
2941 | You must specify an assembler instruction template much like what |
2942 | appears in a machine description, plus an operand constraint string for | |
2943 | each operand. | |
c1f7febf RK |
2944 | |
2945 | For example, here is how to use the 68881's @code{fsinx} instruction: | |
2946 | ||
2947 | @example | |
2948 | asm ("fsinx %1,%0" : "=f" (result) : "f" (angle)); | |
2949 | @end example | |
2950 | ||
2951 | @noindent | |
2952 | Here @code{angle} is the C expression for the input operand while | |
2953 | @code{result} is that of the output operand. Each has @samp{"f"} as its | |
c85f7c16 JL |
2954 | operand constraint, saying that a floating point register is required. |
2955 | The @samp{=} in @samp{=f} indicates that the operand is an output; all | |
2956 | output operands' constraints must use @samp{=}. The constraints use the | |
2957 | same language used in the machine description (@pxref{Constraints}). | |
2958 | ||
2959 | Each operand is described by an operand-constraint string followed by | |
2960 | the C expression in parentheses. A colon separates the assembler | |
2961 | template from the first output operand and another separates the last | |
2962 | output operand from the first input, if any. Commas separate the | |
2963 | operands within each group. The total number of operands is limited to | |
2964 | ten or to the maximum number of operands in any instruction pattern in | |
2965 | the machine description, whichever is greater. | |
2966 | ||
2967 | If there are no output operands but there are input operands, you must | |
2968 | place two consecutive colons surrounding the place where the output | |
c1f7febf RK |
2969 | operands would go. |
2970 | ||
2971 | Output operand expressions must be lvalues; the compiler can check this. | |
c85f7c16 JL |
2972 | The input operands need not be lvalues. The compiler cannot check |
2973 | whether the operands have data types that are reasonable for the | |
2974 | instruction being executed. It does not parse the assembler instruction | |
2975 | template and does not know what it means or even whether it is valid | |
2976 | assembler input. The extended @code{asm} feature is most often used for | |
2977 | machine instructions the compiler itself does not know exist. If | |
2978 | the output expression cannot be directly addressed (for example, it is a | |
f0523f02 | 2979 | bit-field), your constraint must allow a register. In that case, GCC |
c85f7c16 JL |
2980 | will use the register as the output of the @code{asm}, and then store |
2981 | that register into the output. | |
2982 | ||
f0523f02 | 2983 | The ordinary output operands must be write-only; GCC will assume that |
c85f7c16 JL |
2984 | the values in these operands before the instruction are dead and need |
2985 | not be generated. Extended asm supports input-output or read-write | |
2986 | operands. Use the constraint character @samp{+} to indicate such an | |
2987 | operand and list it with the output operands. | |
2988 | ||
2989 | When the constraints for the read-write operand (or the operand in which | |
2990 | only some of the bits are to be changed) allows a register, you may, as | |
2991 | an alternative, logically split its function into two separate operands, | |
2992 | one input operand and one write-only output operand. The connection | |
2993 | between them is expressed by constraints which say they need to be in | |
2994 | the same location when the instruction executes. You can use the same C | |
2995 | expression for both operands, or different expressions. For example, | |
2996 | here we write the (fictitious) @samp{combine} instruction with | |
2997 | @code{bar} as its read-only source operand and @code{foo} as its | |
2998 | read-write destination: | |
c1f7febf RK |
2999 | |
3000 | @example | |
3001 | asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar)); | |
3002 | @end example | |
3003 | ||
3004 | @noindent | |
c85f7c16 JL |
3005 | The constraint @samp{"0"} for operand 1 says that it must occupy the |
3006 | same location as operand 0. A digit in constraint is allowed only in an | |
3007 | input operand and it must refer to an output operand. | |
c1f7febf RK |
3008 | |
3009 | Only a digit in the constraint can guarantee that one operand will be in | |
c85f7c16 JL |
3010 | the same place as another. The mere fact that @code{foo} is the value |
3011 | of both operands is not enough to guarantee that they will be in the | |
3012 | same place in the generated assembler code. The following would not | |
3013 | work reliably: | |
c1f7febf RK |
3014 | |
3015 | @example | |
3016 | asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar)); | |
3017 | @end example | |
3018 | ||
3019 | Various optimizations or reloading could cause operands 0 and 1 to be in | |
f0523f02 | 3020 | different registers; GCC knows no reason not to do so. For example, the |
c1f7febf RK |
3021 | compiler might find a copy of the value of @code{foo} in one register and |
3022 | use it for operand 1, but generate the output operand 0 in a different | |
3023 | register (copying it afterward to @code{foo}'s own address). Of course, | |
3024 | since the register for operand 1 is not even mentioned in the assembler | |
f0523f02 | 3025 | code, the result will not work, but GCC can't tell that. |
c1f7febf | 3026 | |
c85f7c16 JL |
3027 | Some instructions clobber specific hard registers. To describe this, |
3028 | write a third colon after the input operands, followed by the names of | |
3029 | the clobbered hard registers (given as strings). Here is a realistic | |
3030 | example for the VAX: | |
c1f7febf RK |
3031 | |
3032 | @example | |
3033 | asm volatile ("movc3 %0,%1,%2" | |
3034 | : /* no outputs */ | |
3035 | : "g" (from), "g" (to), "g" (count) | |
3036 | : "r0", "r1", "r2", "r3", "r4", "r5"); | |
3037 | @end example | |
3038 | ||
c5c76735 JL |
3039 | You may not write a clobber description in a way that overlaps with an |
3040 | input or output operand. For example, you may not have an operand | |
3041 | describing a register class with one member if you mention that register | |
3042 | in the clobber list. There is no way for you to specify that an input | |
3043 | operand is modified without also specifying it as an output | |
3044 | operand. Note that if all the output operands you specify are for this | |
3045 | purpose (and hence unused), you will then also need to specify | |
3046 | @code{volatile} for the @code{asm} construct, as described below, to | |
f0523f02 | 3047 | prevent GCC from deleting the @code{asm} statement as unused. |
8fe1938e | 3048 | |
c1f7febf | 3049 | If you refer to a particular hardware register from the assembler code, |
c85f7c16 JL |
3050 | you will probably have to list the register after the third colon to |
3051 | tell the compiler the register's value is modified. In some assemblers, | |
3052 | the register names begin with @samp{%}; to produce one @samp{%} in the | |
3053 | assembler code, you must write @samp{%%} in the input. | |
3054 | ||
3055 | If your assembler instruction can alter the condition code register, add | |
f0523f02 | 3056 | @samp{cc} to the list of clobbered registers. GCC on some machines |
c85f7c16 JL |
3057 | represents the condition codes as a specific hardware register; |
3058 | @samp{cc} serves to name this register. On other machines, the | |
3059 | condition code is handled differently, and specifying @samp{cc} has no | |
3060 | effect. But it is valid no matter what the machine. | |
c1f7febf RK |
3061 | |
3062 | If your assembler instruction modifies memory in an unpredictable | |
c85f7c16 | 3063 | fashion, add @samp{memory} to the list of clobbered registers. This |
f0523f02 | 3064 | will cause GCC to not keep memory values cached in registers across |
dd40655a GK |
3065 | the assembler instruction. You will also want to add the |
3066 | @code{volatile} keyword if the memory affected is not listed in the | |
3067 | inputs or outputs of the @code{asm}, as the @samp{memory} clobber does | |
3068 | not count as a side-effect of the @code{asm}. | |
c1f7febf | 3069 | |
c85f7c16 | 3070 | You can put multiple assembler instructions together in a single |
8720914b HPN |
3071 | @code{asm} template, separated by the characters normally used in assembly |
3072 | code for the system. A combination that works in most places is a newline | |
3073 | to break the line, plus a tab character to move to the instruction field | |
3074 | (written as @samp{\n\t}). Sometimes semicolons can be used, if the | |
3075 | assembler allows semicolons as a line-breaking character. Note that some | |
3076 | assembler dialects use semicolons to start a comment. | |
3077 | The input operands are guaranteed not to use any of the clobbered | |
c85f7c16 JL |
3078 | registers, and neither will the output operands' addresses, so you can |
3079 | read and write the clobbered registers as many times as you like. Here | |
3080 | is an example of multiple instructions in a template; it assumes the | |
3081 | subroutine @code{_foo} accepts arguments in registers 9 and 10: | |
c1f7febf RK |
3082 | |
3083 | @example | |
8720914b | 3084 | asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo" |
c1f7febf RK |
3085 | : /* no outputs */ |
3086 | : "g" (from), "g" (to) | |
3087 | : "r9", "r10"); | |
3088 | @end example | |
3089 | ||
f0523f02 | 3090 | Unless an output operand has the @samp{&} constraint modifier, GCC |
c85f7c16 JL |
3091 | may allocate it in the same register as an unrelated input operand, on |
3092 | the assumption the inputs are consumed before the outputs are produced. | |
c1f7febf RK |
3093 | This assumption may be false if the assembler code actually consists of |
3094 | more than one instruction. In such a case, use @samp{&} for each output | |
c85f7c16 | 3095 | operand that may not overlap an input. @xref{Modifiers}. |
c1f7febf | 3096 | |
c85f7c16 JL |
3097 | If you want to test the condition code produced by an assembler |
3098 | instruction, you must include a branch and a label in the @code{asm} | |
3099 | construct, as follows: | |
c1f7febf RK |
3100 | |
3101 | @example | |
8720914b | 3102 | asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:" |
c1f7febf RK |
3103 | : "g" (result) |
3104 | : "g" (input)); | |
3105 | @end example | |
3106 | ||
3107 | @noindent | |
3108 | This assumes your assembler supports local labels, as the GNU assembler | |
3109 | and most Unix assemblers do. | |
3110 | ||
3111 | Speaking of labels, jumps from one @code{asm} to another are not | |
c85f7c16 JL |
3112 | supported. The compiler's optimizers do not know about these jumps, and |
3113 | therefore they cannot take account of them when deciding how to | |
c1f7febf RK |
3114 | optimize. |
3115 | ||
3116 | @cindex macros containing @code{asm} | |
3117 | Usually the most convenient way to use these @code{asm} instructions is to | |
3118 | encapsulate them in macros that look like functions. For example, | |
3119 | ||
3120 | @example | |
3121 | #define sin(x) \ | |
3122 | (@{ double __value, __arg = (x); \ | |
3123 | asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \ | |
3124 | __value; @}) | |
3125 | @end example | |
3126 | ||
3127 | @noindent | |
3128 | Here the variable @code{__arg} is used to make sure that the instruction | |
3129 | operates on a proper @code{double} value, and to accept only those | |
3130 | arguments @code{x} which can convert automatically to a @code{double}. | |
3131 | ||
c85f7c16 JL |
3132 | Another way to make sure the instruction operates on the correct data |
3133 | type is to use a cast in the @code{asm}. This is different from using a | |
c1f7febf RK |
3134 | variable @code{__arg} in that it converts more different types. For |
3135 | example, if the desired type were @code{int}, casting the argument to | |
3136 | @code{int} would accept a pointer with no complaint, while assigning the | |
3137 | argument to an @code{int} variable named @code{__arg} would warn about | |
3138 | using a pointer unless the caller explicitly casts it. | |
3139 | ||
f0523f02 | 3140 | If an @code{asm} has output operands, GCC assumes for optimization |
c85f7c16 JL |
3141 | purposes the instruction has no side effects except to change the output |
3142 | operands. This does not mean instructions with a side effect cannot be | |
3143 | used, but you must be careful, because the compiler may eliminate them | |
3144 | if the output operands aren't used, or move them out of loops, or | |
3145 | replace two with one if they constitute a common subexpression. Also, | |
3146 | if your instruction does have a side effect on a variable that otherwise | |
3147 | appears not to change, the old value of the variable may be reused later | |
3148 | if it happens to be found in a register. | |
c1f7febf RK |
3149 | |
3150 | You can prevent an @code{asm} instruction from being deleted, moved | |
3151 | significantly, or combined, by writing the keyword @code{volatile} after | |
3152 | the @code{asm}. For example: | |
3153 | ||
3154 | @example | |
310668e8 JM |
3155 | #define get_and_set_priority(new) \ |
3156 | (@{ int __old; \ | |
3157 | asm volatile ("get_and_set_priority %0, %1" \ | |
3158 | : "=g" (__old) : "g" (new)); \ | |
c85f7c16 | 3159 | __old; @}) |
24f98470 | 3160 | @end example |
c1f7febf RK |
3161 | |
3162 | @noindent | |
f0523f02 | 3163 | If you write an @code{asm} instruction with no outputs, GCC will know |
c85f7c16 | 3164 | the instruction has side-effects and will not delete the instruction or |
e71b34aa | 3165 | move it outside of loops. |
c85f7c16 | 3166 | |
e71b34aa MM |
3167 | The @code{volatile} keyword indicates that the instruction has |
3168 | important side-effects. GCC will not delete a volatile @code{asm} if | |
3169 | it is reachable. (The instruction can still be deleted if GCC can | |
3170 | prove that control-flow will never reach the location of the | |
3171 | instruction.) In addition, GCC will not reschedule instructions | |
3172 | across a volatile @code{asm} instruction. For example: | |
3173 | ||
3174 | @example | |
bd78000b | 3175 | *(volatile int *)addr = foo; |
e71b34aa MM |
3176 | asm volatile ("eieio" : : ); |
3177 | @end example | |
3178 | ||
ebb48a4d | 3179 | @noindent |
e71b34aa MM |
3180 | Assume @code{addr} contains the address of a memory mapped device |
3181 | register. The PowerPC @code{eieio} instruction (Enforce In-order | |
3182 | Execution of I/O) tells the cpu to make sure that the store to that | |
3183 | device register happens before it issues any other I/O. | |
c1f7febf RK |
3184 | |
3185 | Note that even a volatile @code{asm} instruction can be moved in ways | |
3186 | that appear insignificant to the compiler, such as across jump | |
3187 | instructions. You can't expect a sequence of volatile @code{asm} | |
3188 | instructions to remain perfectly consecutive. If you want consecutive | |
e71b34aa MM |
3189 | output, use a single @code{asm}. Also, GCC will perform some |
3190 | optimizations across a volatile @code{asm} instruction; GCC does not | |
3191 | ``forget everything'' when it encounters a volatile @code{asm} | |
3192 | instruction the way some other compilers do. | |
3193 | ||
3194 | An @code{asm} instruction without any operands or clobbers (an ``old | |
3195 | style'' @code{asm}) will be treated identically to a volatile | |
3196 | @code{asm} instruction. | |
c1f7febf RK |
3197 | |
3198 | It is a natural idea to look for a way to give access to the condition | |
3199 | code left by the assembler instruction. However, when we attempted to | |
3200 | implement this, we found no way to make it work reliably. The problem | |
3201 | is that output operands might need reloading, which would result in | |
3202 | additional following ``store'' instructions. On most machines, these | |
3203 | instructions would alter the condition code before there was time to | |
3204 | test it. This problem doesn't arise for ordinary ``test'' and | |
3205 | ``compare'' instructions because they don't have any output operands. | |
3206 | ||
eda3fbbe GB |
3207 | For reasons similar to those described above, it is not possible to give |
3208 | an assembler instruction access to the condition code left by previous | |
3209 | instructions. | |
3210 | ||
5490d604 | 3211 | If you are writing a header file that should be includable in ISO C |
c1f7febf RK |
3212 | programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate |
3213 | Keywords}. | |
3214 | ||
fe0ce426 JH |
3215 | @subsection i386 floating point asm operands |
3216 | ||
3217 | There are several rules on the usage of stack-like regs in | |
3218 | asm_operands insns. These rules apply only to the operands that are | |
3219 | stack-like regs: | |
3220 | ||
3221 | @enumerate | |
3222 | @item | |
3223 | Given a set of input regs that die in an asm_operands, it is | |
3224 | necessary to know which are implicitly popped by the asm, and | |
3225 | which must be explicitly popped by gcc. | |
3226 | ||
3227 | An input reg that is implicitly popped by the asm must be | |
3228 | explicitly clobbered, unless it is constrained to match an | |
3229 | output operand. | |
3230 | ||
3231 | @item | |
3232 | For any input reg that is implicitly popped by an asm, it is | |
3233 | necessary to know how to adjust the stack to compensate for the pop. | |
3234 | If any non-popped input is closer to the top of the reg-stack than | |
3235 | the implicitly popped reg, it would not be possible to know what the | |
84330467 | 3236 | stack looked like---it's not clear how the rest of the stack ``slides |
fe0ce426 JH |
3237 | up''. |
3238 | ||
3239 | All implicitly popped input regs must be closer to the top of | |
3240 | the reg-stack than any input that is not implicitly popped. | |
3241 | ||
3242 | It is possible that if an input dies in an insn, reload might | |
3243 | use the input reg for an output reload. Consider this example: | |
3244 | ||
3245 | @example | |
3246 | asm ("foo" : "=t" (a) : "f" (b)); | |
3247 | @end example | |
3248 | ||
3249 | This asm says that input B is not popped by the asm, and that | |
c771326b | 3250 | the asm pushes a result onto the reg-stack, i.e., the stack is one |
fe0ce426 JH |
3251 | deeper after the asm than it was before. But, it is possible that |
3252 | reload will think that it can use the same reg for both the input and | |
3253 | the output, if input B dies in this insn. | |
3254 | ||
3255 | If any input operand uses the @code{f} constraint, all output reg | |
3256 | constraints must use the @code{&} earlyclobber. | |
3257 | ||
3258 | The asm above would be written as | |
3259 | ||
3260 | @example | |
3261 | asm ("foo" : "=&t" (a) : "f" (b)); | |
3262 | @end example | |
3263 | ||
3264 | @item | |
3265 | Some operands need to be in particular places on the stack. All | |
84330467 | 3266 | output operands fall in this category---there is no other way to |
fe0ce426 JH |
3267 | know which regs the outputs appear in unless the user indicates |
3268 | this in the constraints. | |
3269 | ||
3270 | Output operands must specifically indicate which reg an output | |
3271 | appears in after an asm. @code{=f} is not allowed: the operand | |
3272 | constraints must select a class with a single reg. | |
3273 | ||
3274 | @item | |
3275 | Output operands may not be ``inserted'' between existing stack regs. | |
3276 | Since no 387 opcode uses a read/write operand, all output operands | |
3277 | are dead before the asm_operands, and are pushed by the asm_operands. | |
3278 | It makes no sense to push anywhere but the top of the reg-stack. | |
3279 | ||
3280 | Output operands must start at the top of the reg-stack: output | |
3281 | operands may not ``skip'' a reg. | |
3282 | ||
3283 | @item | |
3284 | Some asm statements may need extra stack space for internal | |
3285 | calculations. This can be guaranteed by clobbering stack registers | |
3286 | unrelated to the inputs and outputs. | |
3287 | ||
3288 | @end enumerate | |
3289 | ||
3290 | Here are a couple of reasonable asms to want to write. This asm | |
3291 | takes one input, which is internally popped, and produces two outputs. | |
3292 | ||
3293 | @example | |
3294 | asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp)); | |
3295 | @end example | |
3296 | ||
3297 | This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode, | |
3298 | and replaces them with one output. The user must code the @code{st(1)} | |
3299 | clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs. | |
3300 | ||
3301 | @example | |
3302 | asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)"); | |
3303 | @end example | |
3304 | ||
c1f7febf RK |
3305 | @ifclear INTERNALS |
3306 | @c Show the details on constraints if they do not appear elsewhere in | |
3307 | @c the manual | |
3308 | @include md.texi | |
3309 | @end ifclear | |
3310 | ||
3311 | @node Asm Labels | |
3312 | @section Controlling Names Used in Assembler Code | |
3313 | @cindex assembler names for identifiers | |
3314 | @cindex names used in assembler code | |
3315 | @cindex identifiers, names in assembler code | |
3316 | ||
3317 | You can specify the name to be used in the assembler code for a C | |
3318 | function or variable by writing the @code{asm} (or @code{__asm__}) | |
3319 | keyword after the declarator as follows: | |
3320 | ||
3321 | @example | |
3322 | int foo asm ("myfoo") = 2; | |
3323 | @end example | |
3324 | ||
3325 | @noindent | |
3326 | This specifies that the name to be used for the variable @code{foo} in | |
3327 | the assembler code should be @samp{myfoo} rather than the usual | |
3328 | @samp{_foo}. | |
3329 | ||
3330 | On systems where an underscore is normally prepended to the name of a C | |
3331 | function or variable, this feature allows you to define names for the | |
3332 | linker that do not start with an underscore. | |
3333 | ||
0adc3c19 MM |
3334 | It does not make sense to use this feature with a non-static local |
3335 | variable since such variables do not have assembler names. If you are | |
3336 | trying to put the variable in a particular register, see @ref{Explicit | |
3337 | Reg Vars}. GCC presently accepts such code with a warning, but will | |
3338 | probably be changed to issue an error, rather than a warning, in the | |
3339 | future. | |
3340 | ||
c1f7febf RK |
3341 | You cannot use @code{asm} in this way in a function @emph{definition}; but |
3342 | you can get the same effect by writing a declaration for the function | |
3343 | before its definition and putting @code{asm} there, like this: | |
3344 | ||
3345 | @example | |
3346 | extern func () asm ("FUNC"); | |
3347 | ||
3348 | func (x, y) | |
3349 | int x, y; | |
3350 | @dots{} | |
3351 | @end example | |
3352 | ||
3353 | It is up to you to make sure that the assembler names you choose do not | |
3354 | conflict with any other assembler symbols. Also, you must not use a | |
f0523f02 JM |
3355 | register name; that would produce completely invalid assembler code. GCC |
3356 | does not as yet have the ability to store static variables in registers. | |
c1f7febf RK |
3357 | Perhaps that will be added. |
3358 | ||
3359 | @node Explicit Reg Vars | |
3360 | @section Variables in Specified Registers | |
3361 | @cindex explicit register variables | |
3362 | @cindex variables in specified registers | |
3363 | @cindex specified registers | |
3364 | @cindex registers, global allocation | |
3365 | ||
3366 | GNU C allows you to put a few global variables into specified hardware | |
3367 | registers. You can also specify the register in which an ordinary | |
3368 | register variable should be allocated. | |
3369 | ||
3370 | @itemize @bullet | |
3371 | @item | |
3372 | Global register variables reserve registers throughout the program. | |
3373 | This may be useful in programs such as programming language | |
3374 | interpreters which have a couple of global variables that are accessed | |
3375 | very often. | |
3376 | ||
3377 | @item | |
3378 | Local register variables in specific registers do not reserve the | |
3379 | registers. The compiler's data flow analysis is capable of determining | |
3380 | where the specified registers contain live values, and where they are | |
8d344fbc | 3381 | available for other uses. Stores into local register variables may be deleted |
0deaf590 JL |
3382 | when they appear to be dead according to dataflow analysis. References |
3383 | to local register variables may be deleted or moved or simplified. | |
c1f7febf RK |
3384 | |
3385 | These local variables are sometimes convenient for use with the extended | |
3386 | @code{asm} feature (@pxref{Extended Asm}), if you want to write one | |
3387 | output of the assembler instruction directly into a particular register. | |
3388 | (This will work provided the register you specify fits the constraints | |
3389 | specified for that operand in the @code{asm}.) | |
3390 | @end itemize | |
3391 | ||
3392 | @menu | |
3393 | * Global Reg Vars:: | |
3394 | * Local Reg Vars:: | |
3395 | @end menu | |
3396 | ||
3397 | @node Global Reg Vars | |
3398 | @subsection Defining Global Register Variables | |
3399 | @cindex global register variables | |
3400 | @cindex registers, global variables in | |
3401 | ||
3402 | You can define a global register variable in GNU C like this: | |
3403 | ||
3404 | @example | |
3405 | register int *foo asm ("a5"); | |
3406 | @end example | |
3407 | ||
3408 | @noindent | |
3409 | Here @code{a5} is the name of the register which should be used. Choose a | |
3410 | register which is normally saved and restored by function calls on your | |
3411 | machine, so that library routines will not clobber it. | |
3412 | ||
3413 | Naturally the register name is cpu-dependent, so you would need to | |
3414 | conditionalize your program according to cpu type. The register | |
3415 | @code{a5} would be a good choice on a 68000 for a variable of pointer | |
3416 | type. On machines with register windows, be sure to choose a ``global'' | |
3417 | register that is not affected magically by the function call mechanism. | |
3418 | ||
3419 | In addition, operating systems on one type of cpu may differ in how they | |
3420 | name the registers; then you would need additional conditionals. For | |
3421 | example, some 68000 operating systems call this register @code{%a5}. | |
3422 | ||
3423 | Eventually there may be a way of asking the compiler to choose a register | |
3424 | automatically, but first we need to figure out how it should choose and | |
3425 | how to enable you to guide the choice. No solution is evident. | |
3426 | ||
3427 | Defining a global register variable in a certain register reserves that | |
3428 | register entirely for this use, at least within the current compilation. | |
3429 | The register will not be allocated for any other purpose in the functions | |
3430 | in the current compilation. The register will not be saved and restored by | |
3431 | these functions. Stores into this register are never deleted even if they | |
3432 | would appear to be dead, but references may be deleted or moved or | |
3433 | simplified. | |
3434 | ||
3435 | It is not safe to access the global register variables from signal | |
3436 | handlers, or from more than one thread of control, because the system | |
3437 | library routines may temporarily use the register for other things (unless | |
3438 | you recompile them specially for the task at hand). | |
3439 | ||
3440 | @cindex @code{qsort}, and global register variables | |
3441 | It is not safe for one function that uses a global register variable to | |
3442 | call another such function @code{foo} by way of a third function | |
3443 | @code{lose} that was compiled without knowledge of this variable (i.e. in a | |
3444 | different source file in which the variable wasn't declared). This is | |
3445 | because @code{lose} might save the register and put some other value there. | |
3446 | For example, you can't expect a global register variable to be available in | |
3447 | the comparison-function that you pass to @code{qsort}, since @code{qsort} | |
3448 | might have put something else in that register. (If you are prepared to | |
3449 | recompile @code{qsort} with the same global register variable, you can | |
3450 | solve this problem.) | |
3451 | ||
3452 | If you want to recompile @code{qsort} or other source files which do not | |
3453 | actually use your global register variable, so that they will not use that | |
3454 | register for any other purpose, then it suffices to specify the compiler | |
84330467 | 3455 | option @option{-ffixed-@var{reg}}. You need not actually add a global |
c1f7febf RK |
3456 | register declaration to their source code. |
3457 | ||
3458 | A function which can alter the value of a global register variable cannot | |
3459 | safely be called from a function compiled without this variable, because it | |
3460 | could clobber the value the caller expects to find there on return. | |
3461 | Therefore, the function which is the entry point into the part of the | |
3462 | program that uses the global register variable must explicitly save and | |
3463 | restore the value which belongs to its caller. | |
3464 | ||
3465 | @cindex register variable after @code{longjmp} | |
3466 | @cindex global register after @code{longjmp} | |
3467 | @cindex value after @code{longjmp} | |
3468 | @findex longjmp | |
3469 | @findex setjmp | |
3470 | On most machines, @code{longjmp} will restore to each global register | |
3471 | variable the value it had at the time of the @code{setjmp}. On some | |
3472 | machines, however, @code{longjmp} will not change the value of global | |
3473 | register variables. To be portable, the function that called @code{setjmp} | |
3474 | should make other arrangements to save the values of the global register | |
3475 | variables, and to restore them in a @code{longjmp}. This way, the same | |
3476 | thing will happen regardless of what @code{longjmp} does. | |
3477 | ||
3478 | All global register variable declarations must precede all function | |
3479 | definitions. If such a declaration could appear after function | |
3480 | definitions, the declaration would be too late to prevent the register from | |
3481 | being used for other purposes in the preceding functions. | |
3482 | ||
3483 | Global register variables may not have initial values, because an | |
3484 | executable file has no means to supply initial contents for a register. | |
3485 | ||
3486 | On the Sparc, there are reports that g3 @dots{} g7 are suitable | |
3487 | registers, but certain library functions, such as @code{getwd}, as well | |
3488 | as the subroutines for division and remainder, modify g3 and g4. g1 and | |
3489 | g2 are local temporaries. | |
3490 | ||
3491 | On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7. | |
3492 | Of course, it will not do to use more than a few of those. | |
3493 | ||
3494 | @node Local Reg Vars | |
3495 | @subsection Specifying Registers for Local Variables | |
3496 | @cindex local variables, specifying registers | |
3497 | @cindex specifying registers for local variables | |
3498 | @cindex registers for local variables | |
3499 | ||
3500 | You can define a local register variable with a specified register | |
3501 | like this: | |
3502 | ||
3503 | @example | |
3504 | register int *foo asm ("a5"); | |
3505 | @end example | |
3506 | ||
3507 | @noindent | |
3508 | Here @code{a5} is the name of the register which should be used. Note | |
3509 | that this is the same syntax used for defining global register | |
3510 | variables, but for a local variable it would appear within a function. | |
3511 | ||
3512 | Naturally the register name is cpu-dependent, but this is not a | |
3513 | problem, since specific registers are most often useful with explicit | |
3514 | assembler instructions (@pxref{Extended Asm}). Both of these things | |
3515 | generally require that you conditionalize your program according to | |
3516 | cpu type. | |
3517 | ||
3518 | In addition, operating systems on one type of cpu may differ in how they | |
3519 | name the registers; then you would need additional conditionals. For | |
3520 | example, some 68000 operating systems call this register @code{%a5}. | |
3521 | ||
c1f7febf RK |
3522 | Defining such a register variable does not reserve the register; it |
3523 | remains available for other uses in places where flow control determines | |
3524 | the variable's value is not live. However, these registers are made | |
e5e809f4 JL |
3525 | unavailable for use in the reload pass; excessive use of this feature |
3526 | leaves the compiler too few available registers to compile certain | |
3527 | functions. | |
3528 | ||
f0523f02 | 3529 | This option does not guarantee that GCC will generate code that has |
e5e809f4 JL |
3530 | this variable in the register you specify at all times. You may not |
3531 | code an explicit reference to this register in an @code{asm} statement | |
3532 | and assume it will always refer to this variable. | |
c1f7febf | 3533 | |
8d344fbc | 3534 | Stores into local register variables may be deleted when they appear to be dead |
0deaf590 JL |
3535 | according to dataflow analysis. References to local register variables may |
3536 | be deleted or moved or simplified. | |
3537 | ||
c1f7febf RK |
3538 | @node Alternate Keywords |
3539 | @section Alternate Keywords | |
3540 | @cindex alternate keywords | |
3541 | @cindex keywords, alternate | |
3542 | ||
5490d604 JM |
3543 | The option @option{-traditional} disables certain keywords; |
3544 | @option{-ansi} and the various @option{-std} options disable certain | |
3545 | others. This causes trouble when you want to use GNU C extensions, or | |
3546 | ISO C features, in a general-purpose header file that should be usable | |
3547 | by all programs, including ISO C programs and traditional ones. The | |
3548 | keywords @code{asm}, @code{typeof} and @code{inline} cannot be used | |
3549 | since they won't work in a program compiled with @option{-ansi} | |
3550 | (although @code{inline} can be used in a program compiled with | |
3551 | @option{-std=c99}), while the keywords @code{const}, @code{volatile}, | |
3552 | @code{signed}, @code{typeof} and @code{inline} won't work in a program | |
3553 | compiled with @option{-traditional}. The ISO C99 keyword | |
3554 | @code{restrict} is only available when @option{-std=gnu99} (which will | |
3555 | eventually be the default) or @option{-std=c99} (or the equivalent | |
bd819a4a | 3556 | @option{-std=iso9899:1999}) is used. |
c1f7febf RK |
3557 | |
3558 | The way to solve these problems is to put @samp{__} at the beginning and | |
3559 | end of each problematical keyword. For example, use @code{__asm__} | |
3560 | instead of @code{asm}, @code{__const__} instead of @code{const}, and | |
3561 | @code{__inline__} instead of @code{inline}. | |
3562 | ||
3563 | Other C compilers won't accept these alternative keywords; if you want to | |
3564 | compile with another compiler, you can define the alternate keywords as | |
3565 | macros to replace them with the customary keywords. It looks like this: | |
3566 | ||
3567 | @example | |
3568 | #ifndef __GNUC__ | |
3569 | #define __asm__ asm | |
3570 | #endif | |
3571 | @end example | |
3572 | ||
6e6b0525 | 3573 | @findex __extension__ |
84330467 JM |
3574 | @opindex pedantic |
3575 | @option{-pedantic} and other options cause warnings for many GNU C extensions. | |
dbe519e0 | 3576 | You can |
c1f7febf RK |
3577 | prevent such warnings within one expression by writing |
3578 | @code{__extension__} before the expression. @code{__extension__} has no | |
3579 | effect aside from this. | |
3580 | ||
3581 | @node Incomplete Enums | |
3582 | @section Incomplete @code{enum} Types | |
3583 | ||
3584 | You can define an @code{enum} tag without specifying its possible values. | |
3585 | This results in an incomplete type, much like what you get if you write | |
3586 | @code{struct foo} without describing the elements. A later declaration | |
3587 | which does specify the possible values completes the type. | |
3588 | ||
3589 | You can't allocate variables or storage using the type while it is | |
3590 | incomplete. However, you can work with pointers to that type. | |
3591 | ||
3592 | This extension may not be very useful, but it makes the handling of | |
3593 | @code{enum} more consistent with the way @code{struct} and @code{union} | |
3594 | are handled. | |
3595 | ||
3596 | This extension is not supported by GNU C++. | |
3597 | ||
3598 | @node Function Names | |
3599 | @section Function Names as Strings | |
4b404517 JM |
3600 | @cindex @code{__FUNCTION__} identifier |
3601 | @cindex @code{__PRETTY_FUNCTION__} identifier | |
3602 | @cindex @code{__func__} identifier | |
c1f7febf | 3603 | |
f0523f02 | 3604 | GCC predefines two magic identifiers to hold the name of the current |
767094dd JM |
3605 | function. The identifier @code{__FUNCTION__} holds the name of the function |
3606 | as it appears in the source. The identifier @code{__PRETTY_FUNCTION__} | |
22acfb79 NM |
3607 | holds the name of the function pretty printed in a language specific |
3608 | fashion. | |
c1f7febf RK |
3609 | |
3610 | These names are always the same in a C function, but in a C++ function | |
3611 | they may be different. For example, this program: | |
3612 | ||
3613 | @smallexample | |
3614 | extern "C" @{ | |
3615 | extern int printf (char *, ...); | |
3616 | @} | |
3617 | ||
3618 | class a @{ | |
3619 | public: | |
3620 | sub (int i) | |
3621 | @{ | |
3622 | printf ("__FUNCTION__ = %s\n", __FUNCTION__); | |
3623 | printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__); | |
3624 | @} | |
3625 | @}; | |
3626 | ||
3627 | int | |
3628 | main (void) | |
3629 | @{ | |
3630 | a ax; | |
3631 | ax.sub (0); | |
3632 | return 0; | |
3633 | @} | |
3634 | @end smallexample | |
3635 | ||
3636 | @noindent | |
3637 | gives this output: | |
3638 | ||
3639 | @smallexample | |
3640 | __FUNCTION__ = sub | |
3641 | __PRETTY_FUNCTION__ = int a::sub (int) | |
3642 | @end smallexample | |
3643 | ||
22acfb79 | 3644 | The compiler automagically replaces the identifiers with a string |
767094dd | 3645 | literal containing the appropriate name. Thus, they are neither |
22acfb79 | 3646 | preprocessor macros, like @code{__FILE__} and @code{__LINE__}, nor |
767094dd JM |
3647 | variables. This means that they catenate with other string literals, and |
3648 | that they can be used to initialize char arrays. For example | |
22acfb79 NM |
3649 | |
3650 | @smallexample | |
3651 | char here[] = "Function " __FUNCTION__ " in " __FILE__; | |
3652 | @end smallexample | |
3653 | ||
3654 | On the other hand, @samp{#ifdef __FUNCTION__} does not have any special | |
c1f7febf RK |
3655 | meaning inside a function, since the preprocessor does not do anything |
3656 | special with the identifier @code{__FUNCTION__}. | |
3657 | ||
f0523f02 | 3658 | GCC also supports the magic word @code{__func__}, defined by the |
4b404517 | 3659 | ISO standard C99: |
22acfb79 NM |
3660 | |
3661 | @display | |
3662 | The identifier @code{__func__} is implicitly declared by the translator | |
3663 | as if, immediately following the opening brace of each function | |
3664 | definition, the declaration | |
3665 | ||
3666 | @smallexample | |
3667 | static const char __func__[] = "function-name"; | |
3668 | @end smallexample | |
3669 | ||
3670 | appeared, where function-name is the name of the lexically-enclosing | |
767094dd | 3671 | function. This name is the unadorned name of the function. |
22acfb79 NM |
3672 | @end display |
3673 | ||
3674 | By this definition, @code{__func__} is a variable, not a string literal. | |
3675 | In particular, @code{__func__} does not catenate with other string | |
3676 | literals. | |
3677 | ||
3678 | In @code{C++}, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} are | |
3679 | variables, declared in the same way as @code{__func__}. | |
3680 | ||
c1f7febf RK |
3681 | @node Return Address |
3682 | @section Getting the Return or Frame Address of a Function | |
3683 | ||
3684 | These functions may be used to get information about the callers of a | |
3685 | function. | |
3686 | ||
84330467 | 3687 | @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level}) |
c1f7febf RK |
3688 | This function returns the return address of the current function, or of |
3689 | one of its callers. The @var{level} argument is number of frames to | |
3690 | scan up the call stack. A value of @code{0} yields the return address | |
3691 | of the current function, a value of @code{1} yields the return address | |
3692 | of the caller of the current function, and so forth. | |
3693 | ||
3694 | The @var{level} argument must be a constant integer. | |
3695 | ||
3696 | On some machines it may be impossible to determine the return address of | |
3697 | any function other than the current one; in such cases, or when the top | |
3698 | of the stack has been reached, this function will return @code{0}. | |
3699 | ||
3700 | This function should only be used with a non-zero argument for debugging | |
3701 | purposes. | |
84330467 | 3702 | @end deftypefn |
c1f7febf | 3703 | |
84330467 | 3704 | @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level}) |
c1f7febf RK |
3705 | This function is similar to @code{__builtin_return_address}, but it |
3706 | returns the address of the function frame rather than the return address | |
3707 | of the function. Calling @code{__builtin_frame_address} with a value of | |
3708 | @code{0} yields the frame address of the current function, a value of | |
3709 | @code{1} yields the frame address of the caller of the current function, | |
3710 | and so forth. | |
3711 | ||
3712 | The frame is the area on the stack which holds local variables and saved | |
3713 | registers. The frame address is normally the address of the first word | |
3714 | pushed on to the stack by the function. However, the exact definition | |
3715 | depends upon the processor and the calling convention. If the processor | |
3716 | has a dedicated frame pointer register, and the function has a frame, | |
3717 | then @code{__builtin_frame_address} will return the value of the frame | |
3718 | pointer register. | |
3719 | ||
3720 | The caveats that apply to @code{__builtin_return_address} apply to this | |
3721 | function as well. | |
84330467 | 3722 | @end deftypefn |
c1f7febf | 3723 | |
185ebd6c | 3724 | @node Other Builtins |
f0523f02 | 3725 | @section Other built-in functions provided by GCC |
c771326b | 3726 | @cindex built-in functions |
01702459 JM |
3727 | @findex __builtin_isgreater |
3728 | @findex __builtin_isgreaterequal | |
3729 | @findex __builtin_isless | |
3730 | @findex __builtin_islessequal | |
3731 | @findex __builtin_islessgreater | |
3732 | @findex __builtin_isunordered | |
3733 | @findex abort | |
3734 | @findex abs | |
3735 | @findex alloca | |
3736 | @findex bcmp | |
3737 | @findex bzero | |
341e3d11 JM |
3738 | @findex cimag |
3739 | @findex cimagf | |
3740 | @findex cimagl | |
3741 | @findex conj | |
3742 | @findex conjf | |
3743 | @findex conjl | |
01702459 JM |
3744 | @findex cos |
3745 | @findex cosf | |
3746 | @findex cosl | |
341e3d11 JM |
3747 | @findex creal |
3748 | @findex crealf | |
3749 | @findex creall | |
01702459 JM |
3750 | @findex exit |
3751 | @findex _exit | |
796cdb65 | 3752 | @findex _Exit |
01702459 JM |
3753 | @findex fabs |
3754 | @findex fabsf | |
3755 | @findex fabsl | |
3756 | @findex ffs | |
18f988a0 | 3757 | @findex fprintf |
01702459 | 3758 | @findex fputs |
e78f4a97 | 3759 | @findex imaxabs |
c7b6c6cd | 3760 | @findex index |
01702459 JM |
3761 | @findex labs |
3762 | @findex llabs | |
3763 | @findex memcmp | |
3764 | @findex memcpy | |
3765 | @findex memset | |
3766 | @findex printf | |
c7b6c6cd | 3767 | @findex rindex |
01702459 JM |
3768 | @findex sin |
3769 | @findex sinf | |
3770 | @findex sinl | |
3771 | @findex sqrt | |
3772 | @findex sqrtf | |
3773 | @findex sqrtl | |
d118937d | 3774 | @findex strcat |
01702459 JM |
3775 | @findex strchr |
3776 | @findex strcmp | |
3777 | @findex strcpy | |
d118937d | 3778 | @findex strcspn |
01702459 | 3779 | @findex strlen |
d118937d | 3780 | @findex strncat |
da9e9f08 KG |
3781 | @findex strncmp |
3782 | @findex strncpy | |
01702459 JM |
3783 | @findex strpbrk |
3784 | @findex strrchr | |
d118937d | 3785 | @findex strspn |
01702459 | 3786 | @findex strstr |
185ebd6c | 3787 | |
f0523f02 | 3788 | GCC provides a large number of built-in functions other than the ones |
185ebd6c RH |
3789 | mentioned above. Some of these are for internal use in the processing |
3790 | of exceptions or variable-length argument lists and will not be | |
3791 | documented here because they may change from time to time; we do not | |
3792 | recommend general use of these functions. | |
3793 | ||
3794 | The remaining functions are provided for optimization purposes. | |
3795 | ||
84330467 | 3796 | @opindex fno-builtin |
f0523f02 | 3797 | GCC includes built-in versions of many of the functions in the |
01702459 JM |
3798 | standard C library. The versions prefixed with @code{__builtin_} will |
3799 | always be treated as having the same meaning as the C library function | |
84330467 | 3800 | even if you specify the @option{-fno-builtin} (@pxref{C Dialect Options}) |
01702459 JM |
3801 | option. Many of these functions are only optimized in certain cases; if |
3802 | not optimized in a particular case, a call to the library function will | |
3803 | be emitted. | |
3804 | ||
84330467 JM |
3805 | @opindex ansi |
3806 | @opindex std | |
796cdb65 JM |
3807 | The functions @code{abort}, @code{exit}, @code{_Exit} and @code{_exit} |
3808 | are recognized and presumed not to return, but otherwise are not built | |
84330467 JM |
3809 | in. @code{_exit} is not recognized in strict ISO C mode (@option{-ansi}, |
3810 | @option{-std=c89} or @option{-std=c99}). @code{_Exit} is not recognized in | |
3811 | strict C89 mode (@option{-ansi} or @option{-std=c89}). | |
01702459 JM |
3812 | |
3813 | Outside strict ISO C mode, the functions @code{alloca}, @code{bcmp}, | |
c7b6c6cd | 3814 | @code{bzero}, @code{index}, @code{rindex} and @code{ffs} may be handled |
84330467 | 3815 | as built-in functions. Corresponding versions @code{__builtin_alloca}, |
c7b6c6cd KG |
3816 | @code{__builtin_bcmp}, @code{__builtin_bzero}, @code{__builtin_index}, |
3817 | @code{__builtin_rindex} and @code{__builtin_ffs} are also recognized in | |
01702459 JM |
3818 | strict ISO C mode. |
3819 | ||
341e3d11 JM |
3820 | The ISO C99 functions @code{conj}, @code{conjf}, @code{conjl}, |
3821 | @code{creal}, @code{crealf}, @code{creall}, @code{cimag}, @code{cimagf}, | |
84330467 JM |
3822 | @code{cimagl}, @code{llabs} and @code{imaxabs} are handled as built-in functions |
3823 | except in strict ISO C89 mode. There are also built-in versions of the ISO C99 | |
01702459 JM |
3824 | functions @code{cosf}, @code{cosl}, @code{fabsf}, @code{fabsl}, |
3825 | @code{sinf}, @code{sinl}, @code{sqrtf}, and @code{sqrtl}, that are | |
3826 | recognized in any mode since ISO C89 reserves these names for the | |
3827 | purpose to which ISO C99 puts them. All these functions have | |
3828 | corresponding versions prefixed with @code{__builtin_}. | |
3829 | ||
84330467 JM |
3830 | The following ISO C89 functions are recognized as built-in functions unless |
3831 | @option{-fno-builtin} is specified: @code{abs}, @code{cos}, @code{fabs}, | |
18f988a0 KG |
3832 | @code{fprintf}, @code{fputs}, @code{labs}, @code{memcmp}, @code{memcpy}, |
3833 | @code{memset}, @code{printf}, @code{sin}, @code{sqrt}, @code{strcat}, | |
3834 | @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn}, | |
3835 | @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy}, | |
3836 | @code{strpbrk}, @code{strrchr}, @code{strspn}, and @code{strstr}. All | |
3837 | of these functions have corresponding versions prefixed with | |
3838 | @code{__builtin_}, except that the version for @code{sqrt} is called | |
da9e9f08 | 3839 | @code{__builtin_fsqrt}. |
01702459 | 3840 | |
f0523f02 | 3841 | GCC provides built-in versions of the ISO C99 floating point |
01702459 JM |
3842 | comparison macros (that avoid raising exceptions for unordered |
3843 | operands): @code{__builtin_isgreater}, @code{__builtin_isgreaterequal}, | |
3844 | @code{__builtin_isless}, @code{__builtin_islessequal}, | |
3845 | @code{__builtin_islessgreater}, and @code{__builtin_isunordered}. | |
3846 | ||
185ebd6c | 3847 | |
84330467 JM |
3848 | @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp}) |
3849 | You can use the built-in function @code{__builtin_constant_p} to | |
185ebd6c | 3850 | determine if a value is known to be constant at compile-time and hence |
f0523f02 | 3851 | that GCC can perform constant-folding on expressions involving that |
185ebd6c RH |
3852 | value. The argument of the function is the value to test. The function |
3853 | returns the integer 1 if the argument is known to be a compile-time | |
3854 | constant and 0 if it is not known to be a compile-time constant. A | |
3855 | return of 0 does not indicate that the value is @emph{not} a constant, | |
f0523f02 | 3856 | but merely that GCC cannot prove it is a constant with the specified |
84330467 | 3857 | value of the @option{-O} option. |
185ebd6c RH |
3858 | |
3859 | You would typically use this function in an embedded application where | |
3860 | memory was a critical resource. If you have some complex calculation, | |
3861 | you may want it to be folded if it involves constants, but need to call | |
3862 | a function if it does not. For example: | |
3863 | ||
4d390518 | 3864 | @smallexample |
310668e8 JM |
3865 | #define Scale_Value(X) \ |
3866 | (__builtin_constant_p (X) \ | |
3867 | ? ((X) * SCALE + OFFSET) : Scale (X)) | |
185ebd6c RH |
3868 | @end smallexample |
3869 | ||
84330467 | 3870 | You may use this built-in function in either a macro or an inline |
185ebd6c | 3871 | function. However, if you use it in an inlined function and pass an |
f0523f02 | 3872 | argument of the function as the argument to the built-in, GCC will |
185ebd6c | 3873 | never return 1 when you call the inline function with a string constant |
4b404517 | 3874 | or compound literal (@pxref{Compound Literals}) and will not return 1 |
185ebd6c | 3875 | when you pass a constant numeric value to the inline function unless you |
84330467 JM |
3876 | specify the @option{-O} option. |
3877 | @end deftypefn | |
185ebd6c | 3878 | |
84330467 JM |
3879 | @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c}) |
3880 | @opindex fprofile-arcs | |
02f52e19 | 3881 | You may use @code{__builtin_expect} to provide the compiler with |
994a57cd | 3882 | branch prediction information. In general, you should prefer to |
84330467 | 3883 | use actual profile feedback for this (@option{-fprofile-arcs}), as |
994a57cd | 3884 | programmers are notoriously bad at predicting how their programs |
60b6e1f5 | 3885 | actually perform. However, there are applications in which this |
994a57cd RH |
3886 | data is hard to collect. |
3887 | ||
3888 | The return value is the value of @var{exp}, which should be an | |
3889 | integral expression. The value of @var{c} must be a compile-time | |
84330467 | 3890 | constant. The semantics of the built-in are that it is expected |
994a57cd RH |
3891 | that @var{exp} == @var{c}. For example: |
3892 | ||
3893 | @smallexample | |
3894 | if (__builtin_expect (x, 0)) | |
3895 | foo (); | |
3896 | @end smallexample | |
3897 | ||
3898 | @noindent | |
3899 | would indicate that we do not expect to call @code{foo}, since | |
3900 | we expect @code{x} to be zero. Since you are limited to integral | |
3901 | expressions for @var{exp}, you should use constructions such as | |
3902 | ||
3903 | @smallexample | |
3904 | if (__builtin_expect (ptr != NULL, 1)) | |
3905 | error (); | |
3906 | @end smallexample | |
3907 | ||
3908 | @noindent | |
3909 | when testing pointer or floating-point values. | |
84330467 | 3910 | @end deftypefn |
994a57cd | 3911 | |
c1f7febf RK |
3912 | @node C++ Extensions |
3913 | @chapter Extensions to the C++ Language | |
3914 | @cindex extensions, C++ language | |
3915 | @cindex C++ language extensions | |
3916 | ||
3917 | The GNU compiler provides these extensions to the C++ language (and you | |
3918 | can also use most of the C language extensions in your C++ programs). If you | |
3919 | want to write code that checks whether these features are available, you can | |
3920 | test for the GNU compiler the same way as for C programs: check for a | |
3921 | predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to | |
3922 | test specifically for GNU C++ (@pxref{Standard Predefined,,Standard | |
3923 | Predefined Macros,cpp.info,The C Preprocessor}). | |
3924 | ||
3925 | @menu | |
c1f7febf | 3926 | * Min and Max:: C++ Minimum and maximum operators. |
02cac427 | 3927 | * Volatiles:: What constitutes an access to a volatile object. |
49419c8f | 3928 | * Restricted Pointers:: C99 restricted pointers and references. |
7a81cf7f | 3929 | * Vague Linkage:: Where G++ puts inlines, vtables and such. |
c1f7febf | 3930 | * C++ Interface:: You can use a single C++ header file for both |
e6f3b89d | 3931 | declarations and definitions. |
c1f7febf | 3932 | * Template Instantiation:: Methods for ensuring that exactly one copy of |
e6f3b89d | 3933 | each needed template instantiation is emitted. |
0ded1f18 JM |
3934 | * Bound member functions:: You can extract a function pointer to the |
3935 | method denoted by a @samp{->*} or @samp{.*} expression. | |
e6f3b89d | 3936 | * C++ Attributes:: Variable, function, and type attributes for C++ only. |
1f730ff7 | 3937 | * Java Exceptions:: Tweaking exception handling to work with Java. |
e6f3b89d PE |
3938 | * Deprecated Features:: Things might disappear from g++. |
3939 | * Backwards Compatibility:: Compatibilities with earlier definitions of C++. | |
c1f7febf RK |
3940 | @end menu |
3941 | ||
c1f7febf RK |
3942 | @node Min and Max |
3943 | @section Minimum and Maximum Operators in C++ | |
3944 | ||
3945 | It is very convenient to have operators which return the ``minimum'' or the | |
3946 | ``maximum'' of two arguments. In GNU C++ (but not in GNU C), | |
3947 | ||
3948 | @table @code | |
3949 | @item @var{a} <? @var{b} | |
3950 | @findex <? | |
3951 | @cindex minimum operator | |
3952 | is the @dfn{minimum}, returning the smaller of the numeric values | |
3953 | @var{a} and @var{b}; | |
3954 | ||
3955 | @item @var{a} >? @var{b} | |
3956 | @findex >? | |
3957 | @cindex maximum operator | |
3958 | is the @dfn{maximum}, returning the larger of the numeric values @var{a} | |
3959 | and @var{b}. | |
3960 | @end table | |
3961 | ||
3962 | These operations are not primitive in ordinary C++, since you can | |
3963 | use a macro to return the minimum of two things in C++, as in the | |
3964 | following example. | |
3965 | ||
3966 | @example | |
3967 | #define MIN(X,Y) ((X) < (Y) ? : (X) : (Y)) | |
3968 | @end example | |
3969 | ||
3970 | @noindent | |
3971 | You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to | |
3972 | the minimum value of variables @var{i} and @var{j}. | |
3973 | ||
3974 | However, side effects in @code{X} or @code{Y} may cause unintended | |
3975 | behavior. For example, @code{MIN (i++, j++)} will fail, incrementing | |
3976 | the smaller counter twice. A GNU C extension allows you to write safe | |
3977 | macros that avoid this kind of problem (@pxref{Naming Types,,Naming an | |
3978 | Expression's Type}). However, writing @code{MIN} and @code{MAX} as | |
3979 | macros also forces you to use function-call notation for a | |
3980 | fundamental arithmetic operation. Using GNU C++ extensions, you can | |
3981 | write @w{@samp{int min = i <? j;}} instead. | |
3982 | ||
3983 | Since @code{<?} and @code{>?} are built into the compiler, they properly | |
3984 | handle expressions with side-effects; @w{@samp{int min = i++ <? j++;}} | |
3985 | works correctly. | |
3986 | ||
02cac427 NS |
3987 | @node Volatiles |
3988 | @section When is a Volatile Object Accessed? | |
3989 | @cindex accessing volatiles | |
3990 | @cindex volatile read | |
3991 | @cindex volatile write | |
3992 | @cindex volatile access | |
3993 | ||
767094dd JM |
3994 | Both the C and C++ standard have the concept of volatile objects. These |
3995 | are normally accessed by pointers and used for accessing hardware. The | |
8117da65 | 3996 | standards encourage compilers to refrain from optimizations |
02cac427 | 3997 | concerning accesses to volatile objects that it might perform on |
767094dd JM |
3998 | non-volatile objects. The C standard leaves it implementation defined |
3999 | as to what constitutes a volatile access. The C++ standard omits to | |
02cac427 | 4000 | specify this, except to say that C++ should behave in a similar manner |
767094dd | 4001 | to C with respect to volatiles, where possible. The minimum either |
8117da65 | 4002 | standard specifies is that at a sequence point all previous accesses to |
02cac427 | 4003 | volatile objects have stabilized and no subsequent accesses have |
767094dd | 4004 | occurred. Thus an implementation is free to reorder and combine |
02cac427 | 4005 | volatile accesses which occur between sequence points, but cannot do so |
767094dd | 4006 | for accesses across a sequence point. The use of volatiles does not |
02cac427 NS |
4007 | allow you to violate the restriction on updating objects multiple times |
4008 | within a sequence point. | |
4009 | ||
4010 | In most expressions, it is intuitively obvious what is a read and what is | |
767094dd | 4011 | a write. For instance |
02cac427 NS |
4012 | |
4013 | @example | |
c771326b JM |
4014 | volatile int *dst = @var{somevalue}; |
4015 | volatile int *src = @var{someothervalue}; | |
02cac427 NS |
4016 | *dst = *src; |
4017 | @end example | |
4018 | ||
4019 | @noindent | |
4020 | will cause a read of the volatile object pointed to by @var{src} and stores the | |
767094dd | 4021 | value into the volatile object pointed to by @var{dst}. There is no |
02cac427 NS |
4022 | guarantee that these reads and writes are atomic, especially for objects |
4023 | larger than @code{int}. | |
4024 | ||
4025 | Less obvious expressions are where something which looks like an access | |
767094dd | 4026 | is used in a void context. An example would be, |
02cac427 NS |
4027 | |
4028 | @example | |
c771326b | 4029 | volatile int *src = @var{somevalue}; |
02cac427 NS |
4030 | *src; |
4031 | @end example | |
4032 | ||
4033 | With C, such expressions are rvalues, and as rvalues cause a read of | |
f0523f02 | 4034 | the object, GCC interprets this as a read of the volatile being pointed |
767094dd | 4035 | to. The C++ standard specifies that such expressions do not undergo |
02cac427 | 4036 | lvalue to rvalue conversion, and that the type of the dereferenced |
767094dd | 4037 | object may be incomplete. The C++ standard does not specify explicitly |
02cac427 | 4038 | that it is this lvalue to rvalue conversion which is responsible for |
767094dd JM |
4039 | causing an access. However, there is reason to believe that it is, |
4040 | because otherwise certain simple expressions become undefined. However, | |
f0523f02 | 4041 | because it would surprise most programmers, G++ treats dereferencing a |
02cac427 | 4042 | pointer to volatile object of complete type in a void context as a read |
767094dd | 4043 | of the object. When the object has incomplete type, G++ issues a |
02cac427 NS |
4044 | warning. |
4045 | ||
4046 | @example | |
4047 | struct S; | |
4048 | struct T @{int m;@}; | |
c771326b JM |
4049 | volatile S *ptr1 = @var{somevalue}; |
4050 | volatile T *ptr2 = @var{somevalue}; | |
02cac427 NS |
4051 | *ptr1; |
4052 | *ptr2; | |
4053 | @end example | |
4054 | ||
4055 | In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2} | |
767094dd | 4056 | causes a read of the object pointed to. If you wish to force an error on |
02cac427 NS |
4057 | the first case, you must force a conversion to rvalue with, for instance |
4058 | a static cast, @code{static_cast<S>(*ptr1)}. | |
4059 | ||
f0523f02 | 4060 | When using a reference to volatile, G++ does not treat equivalent |
02cac427 | 4061 | expressions as accesses to volatiles, but instead issues a warning that |
767094dd | 4062 | no volatile is accessed. The rationale for this is that otherwise it |
02cac427 NS |
4063 | becomes difficult to determine where volatile access occur, and not |
4064 | possible to ignore the return value from functions returning volatile | |
767094dd | 4065 | references. Again, if you wish to force a read, cast the reference to |
02cac427 NS |
4066 | an rvalue. |
4067 | ||
535233a8 NS |
4068 | @node Restricted Pointers |
4069 | @section Restricting Pointer Aliasing | |
4070 | @cindex restricted pointers | |
4071 | @cindex restricted references | |
4072 | @cindex restricted this pointer | |
4073 | ||
49419c8f | 4074 | As with gcc, g++ understands the C99 feature of restricted pointers, |
535233a8 | 4075 | specified with the @code{__restrict__}, or @code{__restrict} type |
767094dd | 4076 | qualifier. Because you cannot compile C++ by specifying the @option{-std=c99} |
535233a8 NS |
4077 | language flag, @code{restrict} is not a keyword in C++. |
4078 | ||
4079 | In addition to allowing restricted pointers, you can specify restricted | |
4080 | references, which indicate that the reference is not aliased in the local | |
4081 | context. | |
4082 | ||
4083 | @example | |
4084 | void fn (int *__restrict__ rptr, int &__restrict__ rref) | |
4085 | @{ | |
4086 | @dots{} | |
4087 | @} | |
4088 | @end example | |
4089 | ||
4090 | @noindent | |
4091 | In the body of @code{fn}, @var{rptr} points to an unaliased integer and | |
4092 | @var{rref} refers to a (different) unaliased integer. | |
4093 | ||
4094 | You may also specify whether a member function's @var{this} pointer is | |
4095 | unaliased by using @code{__restrict__} as a member function qualifier. | |
4096 | ||
4097 | @example | |
4098 | void T::fn () __restrict__ | |
4099 | @{ | |
4100 | @dots{} | |
4101 | @} | |
4102 | @end example | |
4103 | ||
4104 | @noindent | |
4105 | Within the body of @code{T::fn}, @var{this} will have the effective | |
767094dd | 4106 | definition @code{T *__restrict__ const this}. Notice that the |
535233a8 NS |
4107 | interpretation of a @code{__restrict__} member function qualifier is |
4108 | different to that of @code{const} or @code{volatile} qualifier, in that it | |
767094dd | 4109 | is applied to the pointer rather than the object. This is consistent with |
535233a8 NS |
4110 | other compilers which implement restricted pointers. |
4111 | ||
4112 | As with all outermost parameter qualifiers, @code{__restrict__} is | |
767094dd | 4113 | ignored in function definition matching. This means you only need to |
535233a8 NS |
4114 | specify @code{__restrict__} in a function definition, rather than |
4115 | in a function prototype as well. | |
4116 | ||
7a81cf7f JM |
4117 | @node Vague Linkage |
4118 | @section Vague Linkage | |
4119 | @cindex vague linkage | |
4120 | ||
4121 | There are several constructs in C++ which require space in the object | |
4122 | file but are not clearly tied to a single translation unit. We say that | |
4123 | these constructs have ``vague linkage''. Typically such constructs are | |
4124 | emitted wherever they are needed, though sometimes we can be more | |
4125 | clever. | |
4126 | ||
4127 | @table @asis | |
4128 | @item Inline Functions | |
4129 | Inline functions are typically defined in a header file which can be | |
4130 | included in many different compilations. Hopefully they can usually be | |
4131 | inlined, but sometimes an out-of-line copy is necessary, if the address | |
4132 | of the function is taken or if inlining fails. In general, we emit an | |
4133 | out-of-line copy in all translation units where one is needed. As an | |
4134 | exception, we only emit inline virtual functions with the vtable, since | |
4135 | it will always require a copy. | |
4136 | ||
4137 | Local static variables and string constants used in an inline function | |
4138 | are also considered to have vague linkage, since they must be shared | |
4139 | between all inlined and out-of-line instances of the function. | |
4140 | ||
4141 | @item VTables | |
4142 | @cindex vtable | |
4143 | C++ virtual functions are implemented in most compilers using a lookup | |
4144 | table, known as a vtable. The vtable contains pointers to the virtual | |
4145 | functions provided by a class, and each object of the class contains a | |
4146 | pointer to its vtable (or vtables, in some multiple-inheritance | |
4147 | situations). If the class declares any non-inline, non-pure virtual | |
4148 | functions, the first one is chosen as the ``key method'' for the class, | |
4149 | and the vtable is only emitted in the translation unit where the key | |
4150 | method is defined. | |
4151 | ||
4152 | @emph{Note:} If the chosen key method is later defined as inline, the | |
4153 | vtable will still be emitted in every translation unit which defines it. | |
4154 | Make sure that any inline virtuals are declared inline in the class | |
4155 | body, even if they are not defined there. | |
4156 | ||
4157 | @item type_info objects | |
4158 | @cindex type_info | |
4159 | @cindex RTTI | |
4160 | C++ requires information about types to be written out in order to | |
4161 | implement @samp{dynamic_cast}, @samp{typeid} and exception handling. | |
4162 | For polymorphic classes (classes with virtual functions), the type_info | |
4163 | object is written out along with the vtable so that @samp{dynamic_cast} | |
4164 | can determine the dynamic type of a class object at runtime. For all | |
4165 | other types, we write out the type_info object when it is used: when | |
4166 | applying @samp{typeid} to an expression, throwing an object, or | |
4167 | referring to a type in a catch clause or exception specification. | |
4168 | ||
4169 | @item Template Instantiations | |
4170 | Most everything in this section also applies to template instantiations, | |
4171 | but there are other options as well. | |
4172 | @xref{Template Instantiation,,Where's the Template?}. | |
4173 | ||
4174 | @end table | |
4175 | ||
4176 | When used with GNU ld version 2.8 or later on an ELF system such as | |
4177 | Linux/GNU or Solaris 2, or on Microsoft Windows, duplicate copies of | |
4178 | these constructs will be discarded at link time. This is known as | |
4179 | COMDAT support. | |
4180 | ||
4181 | On targets that don't support COMDAT, but do support weak symbols, GCC | |
4182 | will use them. This way one copy will override all the others, but | |
4183 | the unused copies will still take up space in the executable. | |
4184 | ||
4185 | For targets which do not support either COMDAT or weak symbols, | |
4186 | most entities with vague linkage will be emitted as local symbols to | |
4187 | avoid duplicate definition errors from the linker. This will not happen | |
4188 | for local statics in inlines, however, as having multiple copies will | |
4189 | almost certainly break things. | |
4190 | ||
4191 | @xref{C++ Interface,,Declarations and Definitions in One Header}, for | |
4192 | another way to control placement of these constructs. | |
4193 | ||
c1f7febf RK |
4194 | @node C++ Interface |
4195 | @section Declarations and Definitions in One Header | |
4196 | ||
4197 | @cindex interface and implementation headers, C++ | |
4198 | @cindex C++ interface and implementation headers | |
4199 | C++ object definitions can be quite complex. In principle, your source | |
4200 | code will need two kinds of things for each object that you use across | |
4201 | more than one source file. First, you need an @dfn{interface} | |
4202 | specification, describing its structure with type declarations and | |
4203 | function prototypes. Second, you need the @dfn{implementation} itself. | |
4204 | It can be tedious to maintain a separate interface description in a | |
4205 | header file, in parallel to the actual implementation. It is also | |
4206 | dangerous, since separate interface and implementation definitions may | |
4207 | not remain parallel. | |
4208 | ||
4209 | @cindex pragmas, interface and implementation | |
4210 | With GNU C++, you can use a single header file for both purposes. | |
4211 | ||
4212 | @quotation | |
4213 | @emph{Warning:} The mechanism to specify this is in transition. For the | |
4214 | nonce, you must use one of two @code{#pragma} commands; in a future | |
4215 | release of GNU C++, an alternative mechanism will make these | |
4216 | @code{#pragma} commands unnecessary. | |
4217 | @end quotation | |
4218 | ||
4219 | The header file contains the full definitions, but is marked with | |
4220 | @samp{#pragma interface} in the source code. This allows the compiler | |
4221 | to use the header file only as an interface specification when ordinary | |
4222 | source files incorporate it with @code{#include}. In the single source | |
4223 | file where the full implementation belongs, you can use either a naming | |
4224 | convention or @samp{#pragma implementation} to indicate this alternate | |
4225 | use of the header file. | |
4226 | ||
4227 | @table @code | |
4228 | @item #pragma interface | |
4229 | @itemx #pragma interface "@var{subdir}/@var{objects}.h" | |
4230 | @kindex #pragma interface | |
4231 | Use this directive in @emph{header files} that define object classes, to save | |
4232 | space in most of the object files that use those classes. Normally, | |
4233 | local copies of certain information (backup copies of inline member | |
4234 | functions, debugging information, and the internal tables that implement | |
4235 | virtual functions) must be kept in each object file that includes class | |
4236 | definitions. You can use this pragma to avoid such duplication. When a | |
4237 | header file containing @samp{#pragma interface} is included in a | |
4238 | compilation, this auxiliary information will not be generated (unless | |
4239 | the main input source file itself uses @samp{#pragma implementation}). | |
4240 | Instead, the object files will contain references to be resolved at link | |
4241 | time. | |
4242 | ||
4243 | The second form of this directive is useful for the case where you have | |
4244 | multiple headers with the same name in different directories. If you | |
4245 | use this form, you must specify the same string to @samp{#pragma | |
4246 | implementation}. | |
4247 | ||
4248 | @item #pragma implementation | |
4249 | @itemx #pragma implementation "@var{objects}.h" | |
4250 | @kindex #pragma implementation | |
4251 | Use this pragma in a @emph{main input file}, when you want full output from | |
4252 | included header files to be generated (and made globally visible). The | |
4253 | included header file, in turn, should use @samp{#pragma interface}. | |
4254 | Backup copies of inline member functions, debugging information, and the | |
4255 | internal tables used to implement virtual functions are all generated in | |
4256 | implementation files. | |
4257 | ||
4258 | @cindex implied @code{#pragma implementation} | |
4259 | @cindex @code{#pragma implementation}, implied | |
4260 | @cindex naming convention, implementation headers | |
4261 | If you use @samp{#pragma implementation} with no argument, it applies to | |
4262 | an include file with the same basename@footnote{A file's @dfn{basename} | |
4263 | was the name stripped of all leading path information and of trailing | |
4264 | suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source | |
4265 | file. For example, in @file{allclass.cc}, giving just | |
4266 | @samp{#pragma implementation} | |
4267 | by itself is equivalent to @samp{#pragma implementation "allclass.h"}. | |
4268 | ||
4269 | In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as | |
4270 | an implementation file whenever you would include it from | |
4271 | @file{allclass.cc} even if you never specified @samp{#pragma | |
4272 | implementation}. This was deemed to be more trouble than it was worth, | |
4273 | however, and disabled. | |
4274 | ||
4275 | If you use an explicit @samp{#pragma implementation}, it must appear in | |
4276 | your source file @emph{before} you include the affected header files. | |
4277 | ||
4278 | Use the string argument if you want a single implementation file to | |
4279 | include code from multiple header files. (You must also use | |
4280 | @samp{#include} to include the header file; @samp{#pragma | |
4281 | implementation} only specifies how to use the file---it doesn't actually | |
4282 | include it.) | |
4283 | ||
4284 | There is no way to split up the contents of a single header file into | |
4285 | multiple implementation files. | |
4286 | @end table | |
4287 | ||
4288 | @cindex inlining and C++ pragmas | |
4289 | @cindex C++ pragmas, effect on inlining | |
4290 | @cindex pragmas in C++, effect on inlining | |
4291 | @samp{#pragma implementation} and @samp{#pragma interface} also have an | |
4292 | effect on function inlining. | |
4293 | ||
4294 | If you define a class in a header file marked with @samp{#pragma | |
4295 | interface}, the effect on a function defined in that class is similar to | |
4296 | an explicit @code{extern} declaration---the compiler emits no code at | |
4297 | all to define an independent version of the function. Its definition | |
4298 | is used only for inlining with its callers. | |
4299 | ||
84330467 | 4300 | @opindex fno-implement-inlines |
c1f7febf RK |
4301 | Conversely, when you include the same header file in a main source file |
4302 | that declares it as @samp{#pragma implementation}, the compiler emits | |
4303 | code for the function itself; this defines a version of the function | |
4304 | that can be found via pointers (or by callers compiled without | |
4305 | inlining). If all calls to the function can be inlined, you can avoid | |
84330467 | 4306 | emitting the function by compiling with @option{-fno-implement-inlines}. |
c1f7febf RK |
4307 | If any calls were not inlined, you will get linker errors. |
4308 | ||
4309 | @node Template Instantiation | |
4310 | @section Where's the Template? | |
4311 | ||
4312 | @cindex template instantiation | |
4313 | ||
4314 | C++ templates are the first language feature to require more | |
4315 | intelligence from the environment than one usually finds on a UNIX | |
4316 | system. Somehow the compiler and linker have to make sure that each | |
4317 | template instance occurs exactly once in the executable if it is needed, | |
4318 | and not at all otherwise. There are two basic approaches to this | |
4319 | problem, which I will refer to as the Borland model and the Cfront model. | |
4320 | ||
4321 | @table @asis | |
4322 | @item Borland model | |
4323 | Borland C++ solved the template instantiation problem by adding the code | |
469b759e JM |
4324 | equivalent of common blocks to their linker; the compiler emits template |
4325 | instances in each translation unit that uses them, and the linker | |
4326 | collapses them together. The advantage of this model is that the linker | |
4327 | only has to consider the object files themselves; there is no external | |
4328 | complexity to worry about. This disadvantage is that compilation time | |
4329 | is increased because the template code is being compiled repeatedly. | |
4330 | Code written for this model tends to include definitions of all | |
4331 | templates in the header file, since they must be seen to be | |
4332 | instantiated. | |
c1f7febf RK |
4333 | |
4334 | @item Cfront model | |
4335 | The AT&T C++ translator, Cfront, solved the template instantiation | |
4336 | problem by creating the notion of a template repository, an | |
469b759e JM |
4337 | automatically maintained place where template instances are stored. A |
4338 | more modern version of the repository works as follows: As individual | |
4339 | object files are built, the compiler places any template definitions and | |
4340 | instantiations encountered in the repository. At link time, the link | |
4341 | wrapper adds in the objects in the repository and compiles any needed | |
4342 | instances that were not previously emitted. The advantages of this | |
4343 | model are more optimal compilation speed and the ability to use the | |
4344 | system linker; to implement the Borland model a compiler vendor also | |
c1f7febf | 4345 | needs to replace the linker. The disadvantages are vastly increased |
469b759e JM |
4346 | complexity, and thus potential for error; for some code this can be |
4347 | just as transparent, but in practice it can been very difficult to build | |
c1f7febf | 4348 | multiple programs in one directory and one program in multiple |
469b759e JM |
4349 | directories. Code written for this model tends to separate definitions |
4350 | of non-inline member templates into a separate file, which should be | |
4351 | compiled separately. | |
c1f7febf RK |
4352 | @end table |
4353 | ||
469b759e | 4354 | When used with GNU ld version 2.8 or later on an ELF system such as |
a4b3b54a JM |
4355 | Linux/GNU or Solaris 2, or on Microsoft Windows, g++ supports the |
4356 | Borland model. On other systems, g++ implements neither automatic | |
4357 | model. | |
469b759e JM |
4358 | |
4359 | A future version of g++ will support a hybrid model whereby the compiler | |
4360 | will emit any instantiations for which the template definition is | |
4361 | included in the compile, and store template definitions and | |
4362 | instantiation context information into the object file for the rest. | |
4363 | The link wrapper will extract that information as necessary and invoke | |
4364 | the compiler to produce the remaining instantiations. The linker will | |
4365 | then combine duplicate instantiations. | |
4366 | ||
4367 | In the mean time, you have the following options for dealing with | |
4368 | template instantiations: | |
c1f7febf RK |
4369 | |
4370 | @enumerate | |
d863830b | 4371 | @item |
84330467 JM |
4372 | @opindex frepo |
4373 | Compile your template-using code with @option{-frepo}. The compiler will | |
d863830b JL |
4374 | generate files with the extension @samp{.rpo} listing all of the |
4375 | template instantiations used in the corresponding object files which | |
4376 | could be instantiated there; the link wrapper, @samp{collect2}, will | |
4377 | then update the @samp{.rpo} files to tell the compiler where to place | |
4378 | those instantiations and rebuild any affected object files. The | |
4379 | link-time overhead is negligible after the first pass, as the compiler | |
4380 | will continue to place the instantiations in the same files. | |
4381 | ||
4382 | This is your best option for application code written for the Borland | |
4383 | model, as it will just work. Code written for the Cfront model will | |
4384 | need to be modified so that the template definitions are available at | |
4385 | one or more points of instantiation; usually this is as simple as adding | |
4386 | @code{#include <tmethods.cc>} to the end of each template header. | |
4387 | ||
4388 | For library code, if you want the library to provide all of the template | |
4389 | instantiations it needs, just try to link all of its object files | |
4390 | together; the link will fail, but cause the instantiations to be | |
4391 | generated as a side effect. Be warned, however, that this may cause | |
4392 | conflicts if multiple libraries try to provide the same instantiations. | |
4393 | For greater control, use explicit instantiation as described in the next | |
4394 | option. | |
4395 | ||
c1f7febf | 4396 | @item |
84330467 JM |
4397 | @opindex fno-implicit-templates |
4398 | Compile your code with @option{-fno-implicit-templates} to disable the | |
c1f7febf RK |
4399 | implicit generation of template instances, and explicitly instantiate |
4400 | all the ones you use. This approach requires more knowledge of exactly | |
4401 | which instances you need than do the others, but it's less | |
4402 | mysterious and allows greater control. You can scatter the explicit | |
4403 | instantiations throughout your program, perhaps putting them in the | |
4404 | translation units where the instances are used or the translation units | |
4405 | that define the templates themselves; you can put all of the explicit | |
4406 | instantiations you need into one big file; or you can create small files | |
4407 | like | |
4408 | ||
4409 | @example | |
4410 | #include "Foo.h" | |
4411 | #include "Foo.cc" | |
4412 | ||
4413 | template class Foo<int>; | |
4414 | template ostream& operator << | |
4415 | (ostream&, const Foo<int>&); | |
4416 | @end example | |
4417 | ||
4418 | for each of the instances you need, and create a template instantiation | |
4419 | library from those. | |
4420 | ||
4421 | If you are using Cfront-model code, you can probably get away with not | |
84330467 | 4422 | using @option{-fno-implicit-templates} when compiling files that don't |
c1f7febf RK |
4423 | @samp{#include} the member template definitions. |
4424 | ||
4425 | If you use one big file to do the instantiations, you may want to | |
84330467 | 4426 | compile it without @option{-fno-implicit-templates} so you get all of the |
c1f7febf RK |
4427 | instances required by your explicit instantiations (but not by any |
4428 | other files) without having to specify them as well. | |
4429 | ||
4430 | g++ has extended the template instantiation syntax outlined in the | |
03d0f4af | 4431 | Working Paper to allow forward declaration of explicit instantiations |
4003d7f9 JM |
4432 | (with @code{extern}), instantiation of the compiler support data for a |
4433 | template class (i.e. the vtable) without instantiating any of its | |
4434 | members (with @code{inline}), and instantiation of only the static data | |
4435 | members of a template class, without the support data or member | |
4436 | functions (with (@code{static}): | |
c1f7febf RK |
4437 | |
4438 | @example | |
4439 | extern template int max (int, int); | |
c1f7febf | 4440 | inline template class Foo<int>; |
4003d7f9 | 4441 | static template class Foo<int>; |
c1f7febf RK |
4442 | @end example |
4443 | ||
4444 | @item | |
4445 | Do nothing. Pretend g++ does implement automatic instantiation | |
4446 | management. Code written for the Borland model will work fine, but | |
4447 | each translation unit will contain instances of each of the templates it | |
4448 | uses. In a large program, this can lead to an unacceptable amount of code | |
4449 | duplication. | |
4450 | ||
4451 | @item | |
84330467 | 4452 | @opindex fexternal-templates |
c1f7febf RK |
4453 | Add @samp{#pragma interface} to all files containing template |
4454 | definitions. For each of these files, add @samp{#pragma implementation | |
4455 | "@var{filename}"} to the top of some @samp{.C} file which | |
4456 | @samp{#include}s it. Then compile everything with | |
84330467 | 4457 | @option{-fexternal-templates}. The templates will then only be expanded |
c1f7febf RK |
4458 | in the translation unit which implements them (i.e. has a @samp{#pragma |
4459 | implementation} line for the file where they live); all other files will | |
4460 | use external references. If you're lucky, everything should work | |
4461 | properly. If you get undefined symbol errors, you need to make sure | |
4462 | that each template instance which is used in the program is used in the | |
4463 | file which implements that template. If you don't have any use for a | |
4464 | particular instance in that file, you can just instantiate it | |
4465 | explicitly, using the syntax from the latest C++ working paper: | |
4466 | ||
4467 | @example | |
4468 | template class A<int>; | |
4469 | template ostream& operator << (ostream&, const A<int>&); | |
4470 | @end example | |
4471 | ||
4472 | This strategy will work with code written for either model. If you are | |
4473 | using code written for the Cfront model, the file containing a class | |
4474 | template and the file containing its member templates should be | |
4475 | implemented in the same translation unit. | |
4476 | ||
84330467 | 4477 | @opindex falt-external-templates |
c1f7febf | 4478 | A slight variation on this approach is to instead use the flag |
84330467 | 4479 | @option{-falt-external-templates}; this flag causes template |
c1f7febf RK |
4480 | instances to be emitted in the translation unit that implements the |
4481 | header where they are first instantiated, rather than the one which | |
4482 | implements the file where the templates are defined. This header must | |
4483 | be the same in all translation units, or things are likely to break. | |
4484 | ||
4485 | @xref{C++ Interface,,Declarations and Definitions in One Header}, for | |
4486 | more discussion of these pragmas. | |
4487 | @end enumerate | |
4488 | ||
0ded1f18 JM |
4489 | @node Bound member functions |
4490 | @section Extracting the function pointer from a bound pointer to member function | |
4491 | ||
4492 | @cindex pmf | |
4493 | @cindex pointer to member function | |
4494 | @cindex bound pointer to member function | |
4495 | ||
4496 | In C++, pointer to member functions (PMFs) are implemented using a wide | |
4497 | pointer of sorts to handle all the possible call mechanisms; the PMF | |
4498 | needs to store information about how to adjust the @samp{this} pointer, | |
4499 | and if the function pointed to is virtual, where to find the vtable, and | |
4500 | where in the vtable to look for the member function. If you are using | |
4501 | PMFs in an inner loop, you should really reconsider that decision. If | |
4502 | that is not an option, you can extract the pointer to the function that | |
4503 | would be called for a given object/PMF pair and call it directly inside | |
4504 | the inner loop, to save a bit of time. | |
4505 | ||
4506 | Note that you will still be paying the penalty for the call through a | |
4507 | function pointer; on most modern architectures, such a call defeats the | |
4508 | branch prediction features of the CPU. This is also true of normal | |
4509 | virtual function calls. | |
4510 | ||
4511 | The syntax for this extension is | |
4512 | ||
4513 | @example | |
4514 | extern A a; | |
4515 | extern int (A::*fp)(); | |
4516 | typedef int (*fptr)(A *); | |
4517 | ||
4518 | fptr p = (fptr)(a.*fp); | |
4519 | @end example | |
4520 | ||
0fb6bbf5 | 4521 | For PMF constants (i.e. expressions of the form @samp{&Klasse::Member}), |
767094dd | 4522 | no object is needed to obtain the address of the function. They can be |
0fb6bbf5 ML |
4523 | converted to function pointers directly: |
4524 | ||
4525 | @example | |
4526 | fptr p1 = (fptr)(&A::foo); | |
4527 | @end example | |
4528 | ||
84330467 JM |
4529 | @opindex Wno-pmf-conversions |
4530 | You must specify @option{-Wno-pmf-conversions} to use this extension. | |
0ded1f18 | 4531 | |
5c25e11d PE |
4532 | @node C++ Attributes |
4533 | @section C++-Specific Variable, Function, and Type Attributes | |
4534 | ||
4535 | Some attributes only make sense for C++ programs. | |
4536 | ||
4537 | @table @code | |
4538 | @item init_priority (@var{priority}) | |
4539 | @cindex init_priority attribute | |
4540 | ||
4541 | ||
4542 | In Standard C++, objects defined at namespace scope are guaranteed to be | |
4543 | initialized in an order in strict accordance with that of their definitions | |
4544 | @emph{in a given translation unit}. No guarantee is made for initializations | |
4545 | across translation units. However, GNU C++ allows users to control the | |
3844cd2e | 4546 | order of initialization of objects defined at namespace scope with the |
5c25e11d PE |
4547 | @code{init_priority} attribute by specifying a relative @var{priority}, |
4548 | a constant integral expression currently bounded between 101 and 65535 | |
4549 | inclusive. Lower numbers indicate a higher priority. | |
4550 | ||
4551 | In the following example, @code{A} would normally be created before | |
4552 | @code{B}, but the @code{init_priority} attribute has reversed that order: | |
4553 | ||
4554 | @example | |
4555 | Some_Class A __attribute__ ((init_priority (2000))); | |
4556 | Some_Class B __attribute__ ((init_priority (543))); | |
4557 | @end example | |
4558 | ||
4559 | @noindent | |
4560 | Note that the particular values of @var{priority} do not matter; only their | |
4561 | relative ordering. | |
4562 | ||
60c87482 BM |
4563 | @item java_interface |
4564 | @cindex java_interface attribute | |
4565 | ||
02f52e19 | 4566 | This type attribute informs C++ that the class is a Java interface. It may |
60c87482 | 4567 | only be applied to classes declared within an @code{extern "Java"} block. |
02f52e19 AJ |
4568 | Calls to methods declared in this interface will be dispatched using GCJ's |
4569 | interface table mechanism, instead of regular virtual table dispatch. | |
60c87482 | 4570 | |
5c25e11d PE |
4571 | @end table |
4572 | ||
1f730ff7 ZW |
4573 | @node Java Exceptions |
4574 | @section Java Exceptions | |
4575 | ||
4576 | The Java language uses a slightly different exception handling model | |
4577 | from C++. Normally, GNU C++ will automatically detect when you are | |
4578 | writing C++ code that uses Java exceptions, and handle them | |
4579 | appropriately. However, if C++ code only needs to execute destructors | |
4580 | when Java exceptions are thrown through it, GCC will guess incorrectly. | |
4581 | Sample problematic code: | |
4582 | ||
4583 | @example | |
4584 | struct S @{ ~S(); @}; | |
4585 | extern void bar(); // is implemented in Java and may throw exceptions | |
4586 | void foo() | |
4587 | @{ | |
4588 | S s; | |
4589 | bar(); | |
4590 | @} | |
4591 | @end example | |
4592 | ||
4593 | @noindent | |
4594 | The usual effect of an incorrect guess is a link failure, complaining of | |
4595 | a missing routine called @samp{__gxx_personality_v0}. | |
4596 | ||
4597 | You can inform the compiler that Java exceptions are to be used in a | |
4598 | translation unit, irrespective of what it might think, by writing | |
4599 | @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This | |
4600 | @samp{#pragma} must appear before any functions that throw or catch | |
4601 | exceptions, or run destructors when exceptions are thrown through them. | |
4602 | ||
4603 | You cannot mix Java and C++ exceptions in the same translation unit. It | |
4604 | is believed to be safe to throw a C++ exception from one file through | |
4605 | another file compiled for the for the Java exception model, or vice | |
4606 | versa, but there may be bugs in this area. | |
4607 | ||
e6f3b89d PE |
4608 | @node Deprecated Features |
4609 | @section Deprecated Features | |
4610 | ||
4611 | In the past, the GNU C++ compiler was extended to experiment with new | |
767094dd | 4612 | features, at a time when the C++ language was still evolving. Now that |
e6f3b89d | 4613 | the C++ standard is complete, some of those features are superseded by |
767094dd JM |
4614 | superior alternatives. Using the old features might cause a warning in |
4615 | some cases that the feature will be dropped in the future. In other | |
e6f3b89d PE |
4616 | cases, the feature might be gone already. |
4617 | ||
4618 | While the list below is not exhaustive, it documents some of the options | |
4619 | that are now deprecated: | |
4620 | ||
4621 | @table @code | |
4622 | @item -fexternal-templates | |
4623 | @itemx -falt-external-templates | |
4624 | These are two of the many ways for g++ to implement template | |
767094dd | 4625 | instantiation. @xref{Template Instantiation}. The C++ standard clearly |
e6f3b89d | 4626 | defines how template definitions have to be organized across |
767094dd | 4627 | implementation units. g++ has an implicit instantiation mechanism that |
e6f3b89d PE |
4628 | should work just fine for standard-conforming code. |
4629 | ||
4630 | @item -fstrict-prototype | |
4631 | @itemx -fno-strict-prototype | |
4632 | Previously it was possible to use an empty prototype parameter list to | |
4633 | indicate an unspecified number of parameters (like C), rather than no | |
767094dd | 4634 | parameters, as C++ demands. This feature has been removed, except where |
e6f3b89d PE |
4635 | it is required for backwards compatibility @xref{Backwards Compatibility}. |
4636 | @end table | |
4637 | ||
4638 | The named return value extension has been deprecated, and will be | |
4639 | removed from g++ at some point. | |
4640 | ||
82c18d5c NS |
4641 | The use of initializer lists with new expressions has been deprecated, |
4642 | and will be removed from g++ at some point. | |
4643 | ||
e6f3b89d PE |
4644 | @node Backwards Compatibility |
4645 | @section Backwards Compatibility | |
4646 | @cindex Backwards Compatibility | |
4647 | @cindex ARM [Annotated C++ Reference Manual] | |
4648 | ||
4649 | Now that there is a definitive ISO standard C++, g++ has a specification | |
767094dd | 4650 | to adhere to. The C++ language evolved over time, and features that |
e6f3b89d | 4651 | used to be acceptable in previous drafts of the standard, such as the ARM |
767094dd | 4652 | [Annotated C++ Reference Manual], are no longer accepted. In order to allow |
e6f3b89d | 4653 | compilation of C++ written to such drafts, g++ contains some backwards |
767094dd | 4654 | compatibilities. @emph{All such backwards compatibility features are |
e6f3b89d PE |
4655 | liable to disappear in future versions of g++.} They should be considered |
4656 | deprecated @xref{Deprecated Features}. | |
4657 | ||
4658 | @table @code | |
4659 | @item For scope | |
4660 | If a variable is declared at for scope, it used to remain in scope until | |
4661 | the end of the scope which contained the for statement (rather than just | |
767094dd | 4662 | within the for scope). g++ retains this, but issues a warning, if such a |
e6f3b89d PE |
4663 | variable is accessed outside the for scope. |
4664 | ||
4665 | @item implicit C language | |
630d3d5a | 4666 | Old C system header files did not contain an @code{extern "C" @{@dots{}@}} |
767094dd JM |
4667 | scope to set the language. On such systems, all header files are |
4668 | implicitly scoped inside a C language scope. Also, an empty prototype | |
e6f3b89d PE |
4669 | @code{()} will be treated as an unspecified number of arguments, rather |
4670 | than no arguments, as C++ demands. | |
4671 | @end table |