![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gcc.info-8 CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [GCC 2.x] / gcc|
Return to gcc.info-8 CVS log | Up to [GCC 2.x] / gcc |
1.1.1.9 ! root 1: This is Info file gcc.info, produced by Makeinfo version 1.67 from the ! 2: input file gcc.texi. 1.1 root 3: 4: This file documents the use and the internals of the GNU compiler. 5: 1.1.1.8 root 6: Published by the Free Software Foundation 59 Temple Place - Suite 330 7: Boston, MA 02111-1307 USA 1.1.1.5 root 8: 1.1.1.8 root 9: Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995 Free Software 10: Foundation, Inc. 1.1 root 11: 1.1.1.3 root 12: Permission is granted to make and distribute verbatim copies of this 13: manual provided the copyright notice and this permission notice are 14: preserved on all copies. 1.1 root 15: 16: Permission is granted to copy and distribute modified versions of 17: this manual under the conditions for verbatim copying, provided also 1.1.1.7 root 18: that the sections entitled "GNU General Public License," "Funding for 19: Free Software," and "Protect Your Freedom--Fight `Look And Feel'" are 20: included exactly as in the original, and provided that the entire 21: resulting derived work is distributed under the terms of a permission 22: notice identical to this one. 1.1 root 23: 24: Permission is granted to copy and distribute translations of this 25: manual into another language, under the above conditions for modified 1.1.1.3 root 26: versions, except that the sections entitled "GNU General Public 1.1.1.7 root 27: License," "Funding for Free Software," and "Protect Your Freedom--Fight 28: `Look And Feel'", and this permission notice, may be included in 29: translations approved by the Free Software Foundation instead of in the 30: original English. 31: 32: 1.1.1.8 root 33: File: gcc.info, Node: Build Cross, Prev: Cross Runtime, Up: Cross-Compiler 1.1.1.7 root 34: 1.1.1.8 root 35: Actually Building the Cross-Compiler 36: ------------------------------------ 1.1.1.7 root 37: 1.1.1.8 root 38: Now you can proceed just as for compiling a single-machine compiler 39: through the step of building stage 1. If you have not provided some 40: sort of `libgcc1.a', then compilation will give up at the point where 41: it needs that file, printing a suitable error message. If you do 42: provide `libgcc1.a', then building the compiler will automatically 43: compile and link a test program called `libgcc1-test'; if you get 44: errors in the linking, it means that not all of the necessary routines 45: in `libgcc1.a' are available. 46: 47: You must provide the header file `float.h'. One way to do this is 48: to compile `enquire' and run it on your target machine. The job of 49: `enquire' is to run on the target machine and figure out by experiment 50: the nature of its floating point representation. `enquire' records its 51: findings in the header file `float.h'. If you can't produce this file 52: by running `enquire' on the target machine, then you will need to come 53: up with a suitable `float.h' in some other way (or else, avoid using it 54: in your programs). 55: 56: Do not try to build stage 2 for a cross-compiler. It doesn't work to 57: rebuild GNU CC as a cross-compiler using the cross-compiler, because 58: that would produce a program that runs on the target machine, not on the 59: host. For example, if you compile a 386-to-68030 cross-compiler with 60: itself, the result will not be right either for the 386 (because it was 61: compiled into 68030 code) or for the 68030 (because it was configured 62: for a 386 as the host). If you want to compile GNU CC into 68030 code, 63: whether you compile it on a 68030 or with a cross-compiler on a 386, you 64: must specify a 68030 as the host when you configure it. 1.1.1.7 root 65: 1.1.1.8 root 66: To install the cross-compiler, use `make install', as usual. 1.1.1.7 root 67: 1.1.1.8 root 68: 69: File: gcc.info, Node: Sun Install, Next: VMS Install, Prev: Cross-Compiler, Up: Installation 70: 71: Installing GNU CC on the Sun 72: ============================ 73: 74: On Solaris (version 2.1), do not use the linker or other tools in 75: `/usr/ucb' to build GNU CC. Use `/usr/ccs/bin'. 1.1.1.7 root 76: 1.1.1.8 root 77: Make sure the environment variable `FLOAT_OPTION' is not set when 78: you compile `libgcc.a'. If this option were set to `f68881' when 79: `libgcc.a' is compiled, the resulting code would demand to be linked 80: with a special startup file and would not link properly without special 81: pains. 82: 83: There is a bug in `alloca' in certain versions of the Sun library. 84: To avoid this bug, install the binaries of GNU CC that were compiled by 85: GNU CC. They use `alloca' as a built-in function and never the one in 86: the library. 87: 88: Some versions of the Sun compiler crash when compiling GNU CC. The 89: problem is a segmentation fault in cpp. This problem seems to be due to 90: the bulk of data in the environment variables. You may be able to avoid 91: it by using the following command to compile GNU CC with Sun CC: 1.1.1.7 root 92: 1.1.1.8 root 93: make CC="TERMCAP=x OBJS=x LIBFUNCS=x STAGESTUFF=x cc" 1.1.1.7 root 94: 1.1.1.8 root 95: 96: File: gcc.info, Node: VMS Install, Next: Collect2, Prev: Sun Install, Up: Installation 97: 98: Installing GNU CC on VMS 99: ======================== 1.1.1.7 root 100: 1.1.1.8 root 101: The VMS version of GNU CC is distributed in a backup saveset 102: containing both source code and precompiled binaries. 1.1.1.7 root 103: 1.1.1.8 root 104: To install the `gcc' command so you can use the compiler easily, in 105: the same manner as you use the VMS C compiler, you must install the VMS 106: CLD file for GNU CC as follows: 107: 108: 1. Define the VMS logical names `GNU_CC' and `GNU_CC_INCLUDE' to 109: point to the directories where the GNU CC executables 110: (`gcc-cpp.exe', `gcc-cc1.exe', etc.) and the C include files are 111: kept respectively. This should be done with the commands: 112: 113: $ assign /system /translation=concealed - 114: disk:[gcc.] gnu_cc 115: $ assign /system /translation=concealed - 116: disk:[gcc.include.] gnu_cc_include 117: 118: with the appropriate disk and directory names. These commands can 119: be placed in your system startup file so they will be executed 120: whenever the machine is rebooted. You may, if you choose, do this 121: via the `GCC_INSTALL.COM' script in the `[GCC]' directory. 122: 123: 2. Install the `GCC' command with the command line: 124: 125: $ set command /table=sys$common:[syslib]dcltables - 126: /output=sys$common:[syslib]dcltables gnu_cc:[000000]gcc 127: $ install replace sys$common:[syslib]dcltables 128: 129: 3. To install the help file, do the following: 130: 131: $ library/help sys$library:helplib.hlb gcc.hlp 132: 133: Now you can invoke the compiler with a command like `gcc /verbose 134: file.c', which is equivalent to the command `gcc -v -c file.c' in 135: Unix. 136: 137: If you wish to use GNU C++ you must first install GNU CC, and then 138: perform the following steps: 139: 140: 1. Define the VMS logical name `GNU_GXX_INCLUDE' to point to the 141: directory where the preprocessor will search for the C++ header 142: files. This can be done with the command: 143: 144: $ assign /system /translation=concealed - 145: disk:[gcc.gxx_include.] gnu_gxx_include 146: 147: with the appropriate disk and directory name. If you are going to 148: be using libg++, this is where the libg++ install procedure will 149: install the libg++ header files. 150: 151: 2. Obtain the file `gcc-cc1plus.exe', and place this in the same 152: directory that `gcc-cc1.exe' is kept. 153: 154: The GNU C++ compiler can be invoked with a command like `gcc /plus 155: /verbose file.cc', which is equivalent to the command `g++ -v -c 156: file.cc' in Unix. 157: 158: We try to put corresponding binaries and sources on the VMS 159: distribution tape. But sometimes the binaries will be from an older 160: version than the sources, because we don't always have time to update 161: them. (Use the `/version' option to determine the version number of 162: the binaries and compare it with the source file `version.c' to tell 163: whether this is so.) In this case, you should use the binaries you get 164: to recompile the sources. If you must recompile, here is how: 165: 166: 1. Execute the command procedure `vmsconfig.com' to set up the files 167: `tm.h', `config.h', `aux-output.c', and `md.', and to create files 168: `tconfig.h' and `hconfig.h'. This procedure also creates several 169: linker option files used by `make-cc1.com' and a data file used by 170: `make-l2.com'. 171: 172: $ @vmsconfig.com 173: 174: 2. Setup the logical names and command tables as defined above. In 175: addition, define the VMS logical name `GNU_BISON' to point at the 176: to the directories where the Bison executable is kept. This 177: should be done with the command: 178: 179: $ assign /system /translation=concealed - 180: disk:[bison.] gnu_bison 181: 182: You may, if you choose, use the `INSTALL_BISON.COM' script in the 183: `[BISON]' directory. 184: 185: 3. Install the `BISON' command with the command line: 186: 187: $ set command /table=sys$common:[syslib]dcltables - 188: /output=sys$common:[syslib]dcltables - 189: gnu_bison:[000000]bison 190: $ install replace sys$common:[syslib]dcltables 191: 192: 4. Type `@make-gcc' to recompile everything (alternatively, submit 193: the file `make-gcc.com' to a batch queue). If you wish to build 194: the GNU C++ compiler as well as the GNU CC compiler, you must 195: first edit `make-gcc.com' and follow the instructions that appear 196: in the comments. 197: 198: 5. In order to use GCC, you need a library of functions which GCC 199: compiled code will call to perform certain tasks, and these 200: functions are defined in the file `libgcc2.c'. To compile this 201: you should use the command procedure `make-l2.com', which will 202: generate the library `libgcc2.olb'. `libgcc2.olb' should be built 203: using the compiler built from the same distribution that 204: `libgcc2.c' came from, and `make-gcc.com' will automatically do 205: all of this for you. 206: 207: To install the library, use the following commands: 208: 209: $ library gnu_cc:[000000]gcclib/delete=(new,eprintf) 210: $ library gnu_cc:[000000]gcclib/delete=L_* 211: $ library libgcc2/extract=*/output=libgcc2.obj 212: $ library gnu_cc:[000000]gcclib libgcc2.obj 213: 214: The first command simply removes old modules that will be replaced 215: with modules from `libgcc2' under different module names. The 216: modules `new' and `eprintf' may not actually be present in your 217: `gcclib.olb'--if the VMS librarian complains about those modules 218: not being present, simply ignore the message and continue on with 219: the next command. The second command removes the modules that 220: came from the previous version of the library `libgcc2.c'. 221: 222: Whenever you update the compiler on your system, you should also 223: update the library with the above procedure. 224: 225: 6. You may wish to build GCC in such a way that no files are written 226: to the directory where the source files reside. An example would 227: be the when the source files are on a read-only disk. In these 228: cases, execute the following DCL commands (substituting your 229: actual path names): 230: 231: $ assign dua0:[gcc.build_dir.]/translation=concealed, - 232: dua1:[gcc.source_dir.]/translation=concealed gcc_build 233: $ set default gcc_build:[000000] 234: 235: where the directory `dua1:[gcc.source_dir]' contains the source 236: code, and the directory `dua0:[gcc.build_dir]' is meant to contain 237: all of the generated object files and executables. Once you have 238: done this, you can proceed building GCC as described above. (Keep 239: in mind that `gcc_build' is a rooted logical name, and thus the 240: device names in each element of the search list must be an actual 241: physical device name rather than another rooted logical name). 242: 243: 7. *If you are building GNU CC with a previous version of GNU CC, you 244: also should check to see that you have the newest version of the 245: assembler*. In particular, GNU CC version 2 treats global constant 246: variables slightly differently from GNU CC version 1, and GAS 247: version 1.38.1 does not have the patches required to work with GCC 248: version 2. If you use GAS 1.38.1, then `extern const' variables 249: will not have the read-only bit set, and the linker will generate 250: warning messages about mismatched psect attributes for these 251: variables. These warning messages are merely a nuisance, and can 252: safely be ignored. 253: 254: If you are compiling with a version of GNU CC older than 1.33, 255: specify `/DEFINE=("inline=")' as an option in all the 256: compilations. This requires editing all the `gcc' commands in 257: `make-cc1.com'. (The older versions had problems supporting 258: `inline'.) Once you have a working 1.33 or newer GNU CC, you can 259: change this file back. 260: 261: 8. If you want to build GNU CC with the VAX C compiler, you will need 262: to make minor changes in `make-cccp.com' and `make-cc1.com' to 263: choose alternate definitions of `CC', `CFLAGS', and `LIBS'. See 264: comments in those files. However, you must also have a working 265: version of the GNU assembler (GNU as, aka GAS) as it is used as 266: the back-end for GNU CC to produce binary object modules and is 267: not included in the GNU CC sources. GAS is also needed to compile 268: `libgcc2' in order to build `gcclib' (see above); `make-l2.com' 269: expects to be able to find it operational in 270: `gnu_cc:[000000]gnu-as.exe'. 271: 272: To use GNU CC on VMS, you need the VMS driver programs `gcc.exe', 273: `gcc.com', and `gcc.cld'. They are distributed with the VMS 274: binaries (`gcc-vms') rather than the GNU CC sources. GAS is also 275: included in `gcc-vms', as is Bison. 276: 277: Once you have successfully built GNU CC with VAX C, you should use 278: the resulting compiler to rebuild itself. Before doing this, be 279: sure to restore the `CC', `CFLAGS', and `LIBS' definitions in 280: `make-cccp.com' and `make-cc1.com'. The second generation 281: compiler will be able to take advantage of many optimizations that 282: must be suppressed when building with other compilers. 283: 284: Under previous versions of GNU CC, the generated code would 285: occasionally give strange results when linked with the sharable 286: `VAXCRTL' library. Now this should work. 287: 288: Even with this version, however, GNU CC itself should not be linked 289: with the sharable `VAXCRTL'. The version of `qsort' in `VAXCRTL' has a 290: bug (known to be present in VMS versions V4.6 through V5.5) which 291: causes the compiler to fail. 292: 293: The executables are generated by `make-cc1.com' and `make-cccp.com' 294: use the object library version of `VAXCRTL' in order to make use of the 295: `qsort' routine in `gcclib.olb'. If you wish to link the compiler 296: executables with the shareable image version of `VAXCRTL', you should 297: edit the file `tm.h' (created by `vmsconfig.com') to define the macro 298: `QSORT_WORKAROUND'. 1.1.1.7 root 299: 1.1.1.8 root 300: `QSORT_WORKAROUND' is always defined when GNU CC is compiled with 301: VAX C, to avoid a problem in case `gcclib.olb' is not yet available. 1.1.1.7 root 302: 303: 1.1.1.8 root 304: File: gcc.info, Node: Collect2, Next: Header Dirs, Prev: VMS Install, Up: Installation 1.1.1.7 root 305: 1.1.1.8 root 306: `collect2' 307: ========== 1.1.1.7 root 308: 1.1.1.8 root 309: Many target systems do not have support in the assembler and linker 310: for "constructors"--initialization functions to be called before the 311: official "start" of `main'. On such systems, GNU CC uses a utility 312: called `collect2' to arrange to call these functions at start time. 1.1.1.7 root 313: 1.1.1.8 root 314: The program `collect2' works by linking the program once and looking 315: through the linker output file for symbols with particular names 316: indicating they are constructor functions. If it finds any, it creates 317: a new temporary `.c' file containing a table of them, compiles it, and 318: links the program a second time including that file. 1.1.1.7 root 319: 1.1.1.8 root 320: The actual calls to the constructors are carried out by a subroutine 321: called `__main', which is called (automatically) at the beginning of 322: the body of `main' (provided `main' was compiled with GNU CC). Calling 323: `__main' is necessary, even when compiling C code, to allow linking C 324: and C++ object code together. (If you use `-nostdlib', you get an 325: unresolved reference to `__main', since it's defined in the standard 326: GCC library. Include `-lgcc' at the end of your compiler command line 327: to resolve this reference.) 1.1.1.7 root 328: 1.1.1.8 root 329: The program `collect2' is installed as `ld' in the directory where 330: the passes of the compiler are installed. When `collect2' needs to 331: find the *real* `ld', it tries the following file names: 1.1.1.7 root 332: 1.1.1.8 root 333: * `real-ld' in the directories listed in the compiler's search 334: directories. 1.1.1.7 root 335: 1.1.1.8 root 336: * `real-ld' in the directories listed in the environment variable 337: `PATH'. 1.1.1.7 root 338: 1.1.1.8 root 339: * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 340: if specified. 1.1.1.7 root 341: 1.1.1.8 root 342: * `ld' in the compiler's search directories, except that `collect2' 343: will not execute itself recursively. 344: 345: * `ld' in `PATH'. 346: 347: "The compiler's search directories" means all the directories where 348: `gcc' searches for passes of the compiler. This includes directories 349: that you specify with `-B'. 350: 351: Cross-compilers search a little differently: 352: 353: * `real-ld' in the compiler's search directories. 354: 355: * `TARGET-real-ld' in `PATH'. 356: 357: * The file specified in the `REAL_LD_FILE_NAME' configuration macro, 358: if specified. 359: 360: * `ld' in the compiler's search directories. 361: 362: * `TARGET-ld' in `PATH'. 363: 364: `collect2' explicitly avoids running `ld' using the file name under 365: which `collect2' itself was invoked. In fact, it remembers up a list 366: of such names--in case one copy of `collect2' finds another copy (or 367: version) of `collect2' installed as `ld' in a second place in the 368: search path. 369: 370: `collect2' searches for the utilities `nm' and `strip' using the 371: same algorithm as above for `ld'. 1.1.1.7 root 372: 373: 1.1.1.8 root 374: File: gcc.info, Node: Header Dirs, Prev: Collect2, Up: Installation 1.1.1.7 root 375: 1.1.1.8 root 376: Standard Header File Directories 377: ================================ 1.1.1.7 root 378: 1.1.1.8 root 379: `GCC_INCLUDE_DIR' means the same thing for native and cross. It is 380: where GNU CC stores its private include files, and also where GNU CC 381: stores the fixed include files. A cross compiled GNU CC runs 382: `fixincludes' on the header files in `$(tooldir)/include'. (If the 383: cross compilation header files need to be fixed, they must be installed 384: before GNU CC is built. If the cross compilation header files are 385: already suitable for ANSI C and GNU CC, nothing special need be done). 386: 387: `GPLUS_INCLUDE_DIR' means the same thing for native and cross. It 388: is where `g++' looks first for header files. `libg++' installs only 389: target independent header files in that directory. 390: 391: `LOCAL_INCLUDE_DIR' is used only for a native compiler. It is 392: normally `/usr/local/include'. GNU CC searches this directory so that 393: users can install header files in `/usr/local/include'. 394: 395: `CROSS_INCLUDE_DIR' is used only for a cross compiler. GNU CC 396: doesn't install anything there. 397: 398: `TOOL_INCLUDE_DIR' is used for both native and cross compilers. It 399: is the place for other packages to install header files that GNU CC will 400: use. For a cross-compiler, this is the equivalent of `/usr/include'. 401: When you build a cross-compiler, `fixincludes' processes any header 402: files in this directory. 1.1.1.7 root 403: 404: 1.1.1.8 root 405: File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: Installation, Up: Top 1.1.1.7 root 406: 1.1.1.8 root 407: Extensions to the C Language Family 408: *********************************** 1.1.1.7 root 409: 1.1.1.8 root 410: GNU C provides several language features not found in ANSI standard 411: C. (The `-pedantic' option directs GNU CC to print a warning message if 412: any of these features is used.) To test for the availability of these 413: features in conditional compilation, check for a predefined macro 414: `__GNUC__', which is always defined under GNU CC. 415: 416: These extensions are available in C and Objective C. Most of them 417: are also available in C++. *Note Extensions to the C++ Language: C++ 418: Extensions, for extensions that apply *only* to C++. 1.1.1.7 root 419: 1.1.1.8 root 420: * Menu: 1.1.1.7 root 421: 1.1.1.8 root 422: * Statement Exprs:: Putting statements and declarations inside expressions. 423: * Local Labels:: Labels local to a statement-expression. 424: * Labels as Values:: Getting pointers to labels, and computed gotos. 425: * Nested Functions:: As in Algol and Pascal, lexical scoping of functions. 426: * Constructing Calls:: Dispatching a call to another function. 427: * Naming Types:: Giving a name to the type of some expression. 428: * Typeof:: `typeof': referring to the type of an expression. 429: * Lvalues:: Using `?:', `,' and casts in lvalues. 430: * Conditionals:: Omitting the middle operand of a `?:' expression. 431: * Long Long:: Double-word integers--`long long int'. 432: * Complex:: Data types for complex numbers. 433: * Zero Length:: Zero-length arrays. 434: * Variable Length:: Arrays whose length is computed at run time. 435: * Macro Varargs:: Macros with variable number of arguments. 436: * Subscripting:: Any array can be subscripted, even if not an lvalue. 437: * Pointer Arith:: Arithmetic on `void'-pointers and function pointers. 438: * Initializers:: Non-constant initializers. 439: * Constructors:: Constructor expressions give structures, unions 440: or arrays as values. 441: * Labeled Elements:: Labeling elements of initializers. 442: * Cast to Union:: Casting to union type from any member of the union. 443: * Case Ranges:: `case 1 ... 9' and such. 444: * Function Attributes:: Declaring that functions have no side effects, 445: or that they can never return. 446: * Function Prototypes:: Prototype declarations and old-style definitions. 447: * C++ Comments:: C++ comments are recognized. 448: * Dollar Signs:: Dollar sign is allowed in identifiers. 1.1.1.9 ! root 449: * Character Escapes:: `\e' stands for the character <ESC>. 1.1.1.8 root 450: * Variable Attributes:: Specifying attributes of variables. 451: * Type Attributes:: Specifying attributes of types. 452: * Alignment:: Inquiring about the alignment of a type or variable. 453: * Inline:: Defining inline functions (as fast as macros). 454: * Extended Asm:: Assembler instructions with C expressions as operands. 455: (With them you can define "built-in" functions.) 456: * Asm Labels:: Specifying the assembler name to use for a C symbol. 457: * Explicit Reg Vars:: Defining variables residing in specified registers. 458: * Alternate Keywords:: `__const__', `__asm__', etc., for header files. 459: * Incomplete Enums:: `enum foo;', with details to follow. 460: * Function Names:: Printable strings which are the name of the current 461: function. 1.1.1.7 root 462: 1.1.1.8 root 463: 464: File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions 1.1.1.7 root 465: 1.1.1.8 root 466: Statements and Declarations in Expressions 467: ========================================== 468: 469: A compound statement enclosed in parentheses may appear as an 470: expression in GNU C. This allows you to use loops, switches, and local 471: variables within an expression. 472: 473: Recall that a compound statement is a sequence of statements 474: surrounded by braces; in this construct, parentheses go around the 475: braces. For example: 476: 477: ({ int y = foo (); int z; 478: if (y > 0) z = y; 479: else z = - y; 480: z; }) 481: 482: is a valid (though slightly more complex than necessary) expression for 483: the absolute value of `foo ()'. 484: 485: The last thing in the compound statement should be an expression 486: followed by a semicolon; the value of this subexpression serves as the 487: value of the entire construct. (If you use some other kind of statement 488: last within the braces, the construct has type `void', and thus 489: effectively no value.) 490: 491: This feature is especially useful in making macro definitions "safe" 492: (so that they evaluate each operand exactly once). For example, the 493: "maximum" function is commonly defined as a macro in standard C as 494: follows: 495: 496: #define max(a,b) ((a) > (b) ? (a) : (b)) 497: 498: But this definition computes either A or B twice, with bad results if 499: the operand has side effects. In GNU C, if you know the type of the 500: operands (here let's assume `int'), you can define the macro safely as 501: follows: 502: 503: #define maxint(a,b) \ 504: ({int _a = (a), _b = (b); _a > _b ? _a : _b; }) 505: 506: Embedded statements are not allowed in constant expressions, such as 507: the value of an enumeration constant, the width of a bit field, or the 508: initial value of a static variable. 509: 510: If you don't know the type of the operand, you can still do this, 511: but you must use `typeof' (*note Typeof::.) or type naming (*note 512: Naming Types::.). 1.1.1.7 root 513: 514: 1.1.1.8 root 515: File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions 1.1.1.7 root 516: 1.1.1.8 root 517: Locally Declared Labels 518: ======================= 519: 520: Each statement expression is a scope in which "local labels" can be 521: declared. A local label is simply an identifier; you can jump to it 522: with an ordinary `goto' statement, but only from within the statement 523: expression it belongs to. 524: 525: A local label declaration looks like this: 526: 527: __label__ LABEL; 528: 529: or 1.1.1.7 root 530: 1.1.1.8 root 531: __label__ LABEL1, LABEL2, ...; 1.1.1.7 root 532: 1.1.1.8 root 533: Local label declarations must come at the beginning of the statement 534: expression, right after the `({', before any ordinary declarations. 1.1.1.7 root 535: 1.1.1.8 root 536: The label declaration defines the label *name*, but does not define 537: the label itself. You must do this in the usual way, with `LABEL:', 538: within the statements of the statement expression. 539: 540: The local label feature is useful because statement expressions are 541: often used in macros. If the macro contains nested loops, a `goto' can 542: be useful for breaking out of them. However, an ordinary label whose 543: scope is the whole function cannot be used: if the macro can be 544: expanded several times in one function, the label will be multiply 545: defined in that function. A local label avoids this problem. For 1.1.1.7 root 546: example: 547: 1.1.1.8 root 548: #define SEARCH(array, target) \ 549: ({ \ 550: __label__ found; \ 551: typeof (target) _SEARCH_target = (target); \ 552: typeof (*(array)) *_SEARCH_array = (array); \ 553: int i, j; \ 554: int value; \ 555: for (i = 0; i < max; i++) \ 556: for (j = 0; j < max; j++) \ 557: if (_SEARCH_array[i][j] == _SEARCH_target) \ 558: { value = i; goto found; } \ 559: value = -1; \ 560: found: \ 561: value; \ 562: }) 1.1.1.7 root 563: 564: 1.1.1.8 root 565: File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions 1.1.1.7 root 566: 1.1.1.8 root 567: Labels as Values 568: ================ 1.1.1.7 root 569: 1.1.1.8 root 570: You can get the address of a label defined in the current function 571: (or a containing function) with the unary operator `&&'. The value has 572: type `void *'. This value is a constant and can be used wherever a 573: constant of that type is valid. For example: 1.1.1.7 root 574: 1.1.1.8 root 575: void *ptr; 576: ... 577: ptr = &&foo; 578: 579: To use these values, you need to be able to jump to one. This is 580: done with the computed goto statement(1), `goto *EXP;'. For example, 581: 582: goto *ptr; 1.1.1.7 root 583: 1.1.1.8 root 584: Any expression of type `void *' is allowed. 1.1.1.7 root 585: 1.1.1.8 root 586: One way of using these constants is in initializing a static array 587: that will serve as a jump table: 1.1.1.7 root 588: 1.1.1.8 root 589: static void *array[] = { &&foo, &&bar, &&hack }; 1.1.1.7 root 590: 1.1.1.8 root 591: Then you can select a label with indexing, like this: 1.1.1.7 root 592: 1.1.1.8 root 593: goto *array[i]; 1.1.1.7 root 594: 1.1.1.8 root 595: Note that this does not check whether the subscript is in bounds--array 596: indexing in C never does that. 1.1.1.7 root 597: 1.1.1.8 root 598: Such an array of label values serves a purpose much like that of the 599: `switch' statement. The `switch' statement is cleaner, so use that 600: rather than an array unless the problem does not fit a `switch' 601: statement very well. 1.1.1.7 root 602: 1.1.1.8 root 603: Another use of label values is in an interpreter for threaded code. 604: The labels within the interpreter function can be stored in the 605: threaded code for super-fast dispatching. 606: 607: You can use this mechanism to jump to code in a different function. 608: If you do that, totally unpredictable things will happen. The best way 609: to avoid this is to store the label address only in automatic variables 610: and never pass it as an argument. 611: 612: ---------- Footnotes ---------- 613: 614: (1) The analogous feature in Fortran is called an assigned goto, 615: but that name seems inappropriate in C, where one can do more than 616: simply store label addresses in label variables. 1.1.1.7 root 617: 618: 1.1.1.8 root 619: File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions 1.1.1.7 root 620: 1.1.1.8 root 621: Nested Functions 622: ================ 1.1.1.7 root 623: 1.1.1.8 root 624: A "nested function" is a function defined inside another function. 625: (Nested functions are not supported for GNU C++.) The nested function's 626: name is local to the block where it is defined. For example, here we 627: define a nested function named `square', and call it twice: 1.1.1.7 root 628: 1.1.1.8 root 629: foo (double a, double b) 630: { 631: double square (double z) { return z * z; } 632: 633: return square (a) + square (b); 634: } 635: 636: The nested function can access all the variables of the containing 637: function that are visible at the point of its definition. This is 638: called "lexical scoping". For example, here we show a nested function 639: which uses an inherited variable named `offset': 1.1.1.7 root 640: 1.1.1.8 root 641: bar (int *array, int offset, int size) 1.1.1.7 root 642: { 1.1.1.8 root 643: int access (int *array, int index) 644: { return array[index + offset]; } 645: int i; 646: ... 647: for (i = 0; i < size; i++) 648: ... access (array, i) ... 1.1.1.7 root 649: } 650: 1.1.1.8 root 651: Nested function definitions are permitted within functions in the 652: places where variable definitions are allowed; that is, in any block, 653: before the first statement in the block. 654: 655: It is possible to call the nested function from outside the scope of 656: its name by storing its address or passing the address to another 657: function: 1.1.1.7 root 658: 1.1.1.8 root 659: hack (int *array, int size) 660: { 661: void store (int index, int value) 662: { array[index] = value; } 663: 664: intermediate (store, size); 665: } 666: 667: Here, the function `intermediate' receives the address of `store' as 668: an argument. If `intermediate' calls `store', the arguments given to 669: `store' are used to store into `array'. But this technique works only 670: so long as the containing function (`hack', in this example) does not 671: exit. 672: 673: If you try to call the nested function through its address after the 674: containing function has exited, all hell will break loose. If you try 675: to call it after a containing scope level has exited, and if it refers 676: to some of the variables that are no longer in scope, you may be lucky, 677: but it's not wise to take the risk. If, however, the nested function 678: does not refer to anything that has gone out of scope, you should be 679: safe. 680: 681: GNU CC implements taking the address of a nested function using a 682: technique called "trampolines". A paper describing them is available 683: from `maya.idiap.ch' in directory `pub/tmb', file `usenix88-lexic.ps.Z'. 684: 685: A nested function can jump to a label inherited from a containing 686: function, provided the label was explicitly declared in the containing 687: function (*note Local Labels::.). Such a jump returns instantly to the 688: containing function, exiting the nested function which did the `goto' 689: and any intermediate functions as well. Here is an example: 1.1.1.7 root 690: 1.1.1.8 root 691: bar (int *array, int offset, int size) 692: { 693: __label__ failure; 694: int access (int *array, int index) 695: { 696: if (index > size) 697: goto failure; 698: return array[index + offset]; 699: } 700: int i; 701: ... 702: for (i = 0; i < size; i++) 703: ... access (array, i) ... 704: ... 705: return 0; 1.1.1.7 root 706: 1.1.1.8 root 707: /* Control comes here from `access' 708: if it detects an error. */ 709: failure: 710: return -1; 711: } 1.1.1.7 root 712: 1.1.1.8 root 713: A nested function always has internal linkage. Declaring one with 714: `extern' is erroneous. If you need to declare the nested function 715: before its definition, use `auto' (which is otherwise meaningless for 716: function declarations). 717: 718: bar (int *array, int offset, int size) 719: { 720: __label__ failure; 721: auto int access (int *, int); 722: ... 723: int access (int *array, int index) 724: { 725: if (index > size) 726: goto failure; 727: return array[index + offset]; 728: } 729: ... 730: } 1.1.1.7 root 731: 732: 1.1.1.8 root 733: File: gcc.info, Node: Constructing Calls, Next: Naming Types, Prev: Nested Functions, Up: C Extensions 1.1.1.7 root 734: 1.1.1.8 root 735: Constructing Function Calls 736: =========================== 1.1.1.7 root 737: 1.1.1.8 root 738: Using the built-in functions described below, you can record the 739: arguments a function received, and call another function with the same 740: arguments, without knowing the number or types of the arguments. 741: 742: You can also record the return value of that function call, and 743: later return that value, without knowing what data type the function 744: tried to return (as long as your caller expects that data type). 745: 746: `__builtin_apply_args ()' 747: This built-in function returns a pointer of type `void *' to data 748: describing how to perform a call with the same arguments as were 749: passed to the current function. 750: 751: The function saves the arg pointer register, structure value 752: address, and all registers that might be used to pass arguments to 753: a function into a block of memory allocated on the stack. Then it 754: returns the address of that block. 755: 756: `__builtin_apply (FUNCTION, ARGUMENTS, SIZE)' 757: This built-in function invokes FUNCTION (type `void (*)()') with a 758: copy of the parameters described by ARGUMENTS (type `void *') and 759: SIZE (type `int'). 760: 761: The value of ARGUMENTS should be the value returned by 762: `__builtin_apply_args'. The argument SIZE specifies the size of 763: the stack argument data, in bytes. 764: 765: This function returns a pointer of type `void *' to data describing 766: how to return whatever value was returned by FUNCTION. The data 767: is saved in a block of memory allocated on the stack. 768: 769: It is not always simple to compute the proper value for SIZE. The 770: value is used by `__builtin_apply' to compute the amount of data 771: that should be pushed on the stack and copied from the incoming 772: argument area. 773: 774: `__builtin_return (RESULT)' 775: This built-in function returns the value described by RESULT from 776: the containing function. You should specify, for RESULT, a value 777: returned by `__builtin_apply'. 1.1.1.7 root 778: 1.1.1.8 root 779: 780: File: gcc.info, Node: Naming Types, Next: Typeof, Prev: Constructing Calls, Up: C Extensions 1.1.1.7 root 781: 1.1.1.8 root 782: Naming an Expression's Type 783: =========================== 784: 785: You can give a name to the type of an expression using a `typedef' 786: declaration with an initializer. Here is how to define NAME as a type 787: name for the type of EXP: 788: 789: typedef NAME = EXP; 790: 791: This is useful in conjunction with the statements-within-expressions 792: feature. Here is how the two together can be used to define a safe 793: "maximum" macro that operates on any arithmetic type: 794: 795: #define max(a,b) \ 796: ({typedef _ta = (a), _tb = (b); \ 797: _ta _a = (a); _tb _b = (b); \ 798: _a > _b ? _a : _b; }) 799: 800: The reason for using names that start with underscores for the local 801: variables is to avoid conflicts with variable names that occur within 802: the expressions that are substituted for `a' and `b'. Eventually we 803: hope to design a new form of declaration syntax that allows you to 804: declare variables whose scopes start only after their initializers; 805: this will be a more reliable way to prevent such conflicts. 1.1.1.7 root 806: 807: 1.1.1.8 root 808: File: gcc.info, Node: Typeof, Next: Lvalues, Prev: Naming Types, Up: C Extensions 809: 810: Referring to a Type with `typeof' 811: ================================= 812: 813: Another way to refer to the type of an expression is with `typeof'. 814: The syntax of using of this keyword looks like `sizeof', but the 815: construct acts semantically like a type name defined with `typedef'. 816: 817: There are two ways of writing the argument to `typeof': with an 818: expression or with a type. Here is an example with an expression: 819: 820: typeof (x[0](1)) 821: 822: This assumes that `x' is an array of functions; the type described is 823: that of the values of the functions. 824: 825: Here is an example with a typename as the argument: 826: 827: typeof (int *) 828: 829: Here the type described is that of pointers to `int'. 1.1.1.7 root 830: 1.1.1.8 root 831: If you are writing a header file that must work when included in 832: ANSI C programs, write `__typeof__' instead of `typeof'. *Note 833: Alternate Keywords::. 1.1.1.7 root 834: 1.1.1.8 root 835: A `typeof'-construct can be used anywhere a typedef name could be 836: used. For example, you can use it in a declaration, in a cast, or 837: inside of `sizeof' or `typeof'. 838: 839: * This declares `y' with the type of what `x' points to. 840: 841: typeof (*x) y; 842: 843: * This declares `y' as an array of such values. 844: 845: typeof (*x) y[4]; 846: 847: * This declares `y' as an array of pointers to characters: 848: 849: typeof (typeof (char *)[4]) y; 850: 851: It is equivalent to the following traditional C declaration: 852: 853: char *y[4]; 854: 855: To see the meaning of the declaration using `typeof', and why it 856: might be a useful way to write, let's rewrite it with these macros: 857: 858: #define pointer(T) typeof(T *) 859: #define array(T, N) typeof(T [N]) 860: 861: Now the declaration can be rewritten this way: 862: 863: array (pointer (char), 4) y; 864: 865: Thus, `array (pointer (char), 4)' is the type of arrays of 4 866: pointers to `char'. 1.1.1.2 root 867: 868: 1.1.1.8 root 869: File: gcc.info, Node: Lvalues, Next: Conditionals, Prev: Typeof, Up: C Extensions 1.1.1.6 root 870: 1.1.1.8 root 871: Generalized Lvalues 872: =================== 1.1.1.6 root 873: 1.1.1.8 root 874: Compound expressions, conditional expressions and casts are allowed 875: as lvalues provided their operands are lvalues. This means that you 876: can take their addresses or store values into them. 877: 878: Standard C++ allows compound expressions and conditional expressions 879: as lvalues, and permits casts to reference type, so use of this 880: extension is deprecated for C++ code. 881: 882: For example, a compound expression can be assigned, provided the last 883: expression in the sequence is an lvalue. These two expressions are 884: equivalent: 885: 886: (a, b) += 5 887: a, (b += 5) 888: 889: Similarly, the address of the compound expression can be taken. 890: These two expressions are equivalent: 891: 892: &(a, b) 893: a, &b 894: 895: A conditional expression is a valid lvalue if its type is not void 896: and the true and false branches are both valid lvalues. For example, 897: these two expressions are equivalent: 898: 899: (a ? b : c) = 5 900: (a ? b = 5 : (c = 5)) 1.1.1.6 root 901: 1.1.1.8 root 902: A cast is a valid lvalue if its operand is an lvalue. A simple 903: assignment whose left-hand side is a cast works by converting the 904: right-hand side first to the specified type, then to the type of the 905: inner left-hand side expression. After this is stored, the value is 906: converted back to the specified type to become the value of the 907: assignment. Thus, if `a' has type `char *', the following two 908: expressions are equivalent: 1.1.1.6 root 909: 1.1.1.8 root 910: (int)a = 5 911: (int)(a = (char *)(int)5) 912: 913: An assignment-with-arithmetic operation such as `+=' applied to a 914: cast performs the arithmetic using the type resulting from the cast, 915: and then continues as in the previous case. Therefore, these two 916: expressions are equivalent: 917: 918: (int)a += 5 919: (int)(a = (char *)(int) ((int)a + 5)) 920: 921: You cannot take the address of an lvalue cast, because the use of its 922: address would not work out coherently. Suppose that `&(int)f' were 923: permitted, where `f' has type `float'. Then the following statement 924: would try to store an integer bit-pattern where a floating point number 925: belongs: 926: 927: *&(int)f = 1; 928: 929: This is quite different from what `(int)f = 1' would do--that would 930: convert 1 to floating point and store it. Rather than cause this 931: inconsistency, we think it is better to prohibit use of `&' on a cast. 932: 933: If you really do want an `int *' pointer with the address of `f', 934: you can simply write `(int *)&f'. 1.1.1.6 root 935: 936: 1.1.1.8 root 937: File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Lvalues, Up: C Extensions 1.1.1.6 root 938: 1.1.1.8 root 939: Conditionals with Omitted Operands 940: ================================== 941: 942: The middle operand in a conditional expression may be omitted. Then 943: if the first operand is nonzero, its value is the value of the 944: conditional expression. 945: 946: Therefore, the expression 1.1.1.6 root 947: 1.1.1.8 root 948: x ? : y 1.1.1.6 root 949: 1.1.1.8 root 950: has the value of `x' if that is nonzero; otherwise, the value of `y'. 1.1.1.6 root 951: 1.1.1.8 root 952: This example is perfectly equivalent to 1.1.1.6 root 953: 1.1.1.8 root 954: x ? x : y 955: 956: In this simple case, the ability to omit the middle operand is not 957: especially useful. When it becomes useful is when the first operand 958: does, or may (if it is a macro argument), contain a side effect. Then 959: repeating the operand in the middle would perform the side effect 960: twice. Omitting the middle operand uses the value already computed 961: without the undesirable effects of recomputing it. 1.1.1.7 root 962: 1.1.1.6 root 963: 1.1.1.8 root 964: File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions 1.1.1.6 root 965: 1.1.1.8 root 966: Double-Word Integers 967: ==================== 1.1.1.6 root 968: 1.1.1.8 root 969: GNU C supports data types for integers that are twice as long as 970: `long int'. Simply write `long long int' for a signed integer, or 971: `unsigned long long int' for an unsigned integer. To make an integer 972: constant of type `long long int', add the suffix `LL' to the integer. 973: To make an integer constant of type `unsigned long long int', add the 974: suffix `ULL' to the integer. 975: 976: You can use these types in arithmetic like any other integer types. 977: Addition, subtraction, and bitwise boolean operations on these types 978: are open-coded on all types of machines. Multiplication is open-coded 979: if the machine supports fullword-to-doubleword a widening multiply 980: instruction. Division and shifts are open-coded only on machines that 981: provide special support. The operations that are not open-coded use 982: special library routines that come with GNU CC. 983: 984: There may be pitfalls when you use `long long' types for function 985: arguments, unless you declare function prototypes. If a function 986: expects type `int' for its argument, and you pass a value of type `long 987: long int', confusion will result because the caller and the subroutine 988: will disagree about the number of bytes for the argument. Likewise, if 989: the function expects `long long int' and you pass `int'. The best way 990: to avoid such problems is to use prototypes. 1.1.1.6 root 991: 1.1.1.8 root 992: 993: File: gcc.info, Node: Complex, Next: Zero Length, Prev: Long Long, Up: C Extensions 1.1.1.6 root 994: 1.1.1.8 root 995: Complex Numbers 996: =============== 997: 998: GNU C supports complex data types. You can declare both complex 999: integer types and complex floating types, using the keyword 1000: `__complex__'. 1001: 1002: For example, `__complex__ double x;' declares `x' as a variable 1003: whose real part and imaginary part are both of type `double'. 1004: `__complex__ short int y;' declares `y' to have real and imaginary 1005: parts of type `short int'; this is not likely to be useful, but it 1006: shows that the set of complex types is complete. 1007: 1008: To write a constant with a complex data type, use the suffix `i' or 1009: `j' (either one; they are equivalent). For example, `2.5fi' has type 1010: `__complex__ float' and `3i' has type `__complex__ int'. Such a 1011: constant always has a pure imaginary value, but you can form any 1012: complex value you like by adding one to a real constant. 1013: 1014: To extract the real part of a complex-valued expression EXP, write 1015: `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary part. 1016: 1017: The operator `~' performs complex conjugation when used on a value 1018: with a complex type. 1019: 1020: GNU CC can allocate complex automatic variables in a noncontiguous 1021: fashion; it's even possible for the real part to be in a register while 1022: the imaginary part is on the stack (or vice-versa). None of the 1023: supported debugging info formats has a way to represent noncontiguous 1024: allocation like this, so GNU CC describes a noncontiguous complex 1025: variable as if it were two separate variables of noncomplex type. If 1026: the variable's actual name is `foo', the two fictitious variables are 1027: named `foo$real' and `foo$imag'. You can examine and set these two 1028: fictitious variables with your debugger. 1029: 1030: A future version of GDB will know how to recognize such pairs and 1031: treat them as a single variable with a complex type. 1032: 1033: 1034: File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Complex, Up: C Extensions 1035: 1036: Arrays of Length Zero 1037: ===================== 1038: 1039: Zero-length arrays are allowed in GNU C. They are very useful as 1040: the last element of a structure which is really a header for a 1041: variable-length object: 1042: 1043: struct line { 1044: int length; 1045: char contents[0]; 1.1.1.6 root 1046: }; 1047: 1048: { 1.1.1.8 root 1049: struct line *thisline = (struct line *) 1050: malloc (sizeof (struct line) + this_length); 1051: thisline->length = this_length; 1.1.1.6 root 1052: } 1053: 1.1.1.8 root 1054: In standard C, you would have to give `contents' a length of 1, which 1055: means either you waste space or complicate the argument to `malloc'. 1.1.1.6 root 1056: 1057: 1.1.1.8 root 1058: File: gcc.info, Node: Variable Length, Next: Macro Varargs, Prev: Zero Length, Up: C Extensions 1.1.1.6 root 1059: 1.1.1.8 root 1060: Arrays of Variable Length 1061: ========================= 1.1.1.6 root 1062: 1.1.1.8 root 1063: Variable-length automatic arrays are allowed in GNU C. These arrays 1064: are declared like any other automatic arrays, but with a length that is 1065: not a constant expression. The storage is allocated at the point of 1066: declaration and deallocated when the brace-level is exited. For 1067: example: 1.1.1.6 root 1068: 1.1.1.8 root 1069: FILE * 1070: concat_fopen (char *s1, char *s2, char *mode) 1071: { 1072: char str[strlen (s1) + strlen (s2) + 1]; 1073: strcpy (str, s1); 1074: strcat (str, s2); 1075: return fopen (str, mode); 1076: } 1.1.1.6 root 1077: 1.1.1.8 root 1078: Jumping or breaking out of the scope of the array name deallocates 1079: the storage. Jumping into the scope is not allowed; you get an error 1080: message for it. 1081: 1082: You can use the function `alloca' to get an effect much like 1083: variable-length arrays. The function `alloca' is available in many 1084: other C implementations (but not in all). On the other hand, 1085: variable-length arrays are more elegant. 1086: 1087: There are other differences between these two methods. Space 1088: allocated with `alloca' exists until the containing *function* returns. 1089: The space for a variable-length array is deallocated as soon as the 1090: array name's scope ends. (If you use both variable-length arrays and 1091: `alloca' in the same function, deallocation of a variable-length array 1092: will also deallocate anything more recently allocated with `alloca'.) 1.1.1.6 root 1093: 1.1.1.8 root 1094: You can also use variable-length arrays as arguments to functions: 1095: 1096: struct entry 1097: tester (int len, char data[len][len]) 1.1.1.6 root 1098: { 1099: ... 1100: } 1101: 1.1.1.8 root 1102: The length of an array is computed once when the storage is allocated 1103: and is remembered for the scope of the array in case you access it with 1104: `sizeof'. 1105: 1106: If you want to pass the array first and the length afterward, you can 1107: use a forward declaration in the parameter list--another GNU extension. 1.1.1.6 root 1108: 1.1.1.8 root 1109: struct entry 1110: tester (int len; char data[len][len], int len) 1.1.1.6 root 1111: { 1.1.1.8 root 1112: ... 1.1.1.6 root 1113: } 1114: 1.1.1.8 root 1115: The `int len' before the semicolon is a "parameter forward 1116: declaration", and it serves the purpose of making the name `len' known 1117: when the declaration of `data' is parsed. 1118: 1119: You can write any number of such parameter forward declarations in 1120: the parameter list. They can be separated by commas or semicolons, but 1121: the last one must end with a semicolon, which is followed by the "real" 1122: parameter declarations. Each forward declaration must match a "real" 1123: declaration in parameter name and data type. 1.1.1.6 root 1124: 1.1.1.8 root 1125: 1126: File: gcc.info, Node: Macro Varargs, Next: Subscripting, Prev: Variable Length, Up: C Extensions 1127: 1128: Macros with Variable Numbers of Arguments 1129: ========================================= 1130: 1131: In GNU C, a macro can accept a variable number of arguments, much as 1132: a function can. The syntax for defining the macro looks much like that 1133: used for a function. Here is an example: 1134: 1135: #define eprintf(format, args...) \ 1136: fprintf (stderr, format , ## args) 1137: 1138: Here `args' is a "rest argument": it takes in zero or more 1139: arguments, as many as the call contains. All of them plus the commas 1140: between them form the value of `args', which is substituted into the 1141: macro body where `args' is used. Thus, we have this expansion: 1142: 1143: eprintf ("%s:%d: ", input_file_name, line_number) 1144: ==> 1145: fprintf (stderr, "%s:%d: " , input_file_name, line_number) 1146: 1147: Note that the comma after the string constant comes from the definition 1148: of `eprintf', whereas the last comma comes from the value of `args'. 1149: 1150: The reason for using `##' is to handle the case when `args' matches 1151: no arguments at all. In this case, `args' has an empty value. In this 1152: case, the second comma in the definition becomes an embarrassment: if 1153: it got through to the expansion of the macro, we would get something 1154: like this: 1.1.1.6 root 1155: 1.1.1.8 root 1156: fprintf (stderr, "success!\n" , ) 1.1.1.6 root 1157: 1.1.1.8 root 1158: which is invalid C syntax. `##' gets rid of the comma, so we get the 1159: following instead: 1.1.1.6 root 1160: 1.1.1.8 root 1161: fprintf (stderr, "success!\n") 1162: 1163: This is a special feature of the GNU C preprocessor: `##' before a 1164: rest argument that is empty discards the preceding sequence of 1165: non-whitespace characters from the macro definition. (If another macro 1166: argument precedes, none of it is discarded.) 1167: 1168: It might be better to discard the last preprocessor token instead of 1169: the last preceding sequence of non-whitespace characters; in fact, we 1170: may someday change this feature to do so. We advise you to write the 1171: macro definition so that the preceding sequence of non-whitespace 1172: characters is just a single token, so that the meaning will not change 1173: if we change the definition of this feature. 1174: 1175: 1176: File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Macro Varargs, Up: C Extensions 1177: 1178: Non-Lvalue Arrays May Have Subscripts 1179: ===================================== 1180: 1181: Subscripting is allowed on arrays that are not lvalues, even though 1182: the unary `&' operator is not. For example, this is valid in GNU C 1183: though not valid in other C dialects: 1184: 1185: struct foo {int a[4];}; 1186: 1187: struct foo f(); 1188: 1189: bar (int index) 1.1.1.6 root 1190: { 1.1.1.8 root 1191: return f().a[index]; 1.1.1.6 root 1192: } 1193: 1.1.1.8 root 1194: 1195: File: gcc.info, Node: Pointer Arith, Next: Initializers, Prev: Subscripting, Up: C Extensions 1196: 1197: Arithmetic on `void'- and Function-Pointers 1198: =========================================== 1.1.1.6 root 1199: 1.1.1.8 root 1200: In GNU C, addition and subtraction operations are supported on 1201: pointers to `void' and on pointers to functions. This is done by 1202: treating the size of a `void' or of a function as 1. 1203: 1204: A consequence of this is that `sizeof' is also allowed on `void' and 1205: on function types, and returns 1. 1206: 1207: The option `-Wpointer-arith' requests a warning if these extensions 1208: are used. 1209: 1210: 1211: File: gcc.info, Node: Initializers, Next: Constructors, Prev: Pointer Arith, Up: C Extensions 1212: 1213: Non-Constant Initializers 1214: ========================= 1215: 1216: As in standard C++, the elements of an aggregate initializer for an 1217: automatic variable are not required to be constant expressions in GNU C. 1218: Here is an example of an initializer with run-time varying elements: 1219: 1220: foo (float f, float g) 1.1.1.6 root 1221: { 1.1.1.8 root 1222: float beat_freqs[2] = { f-g, f+g }; 1223: ... 1.1.1.6 root 1224: } 1225:
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.