GCC使用指南

使用語法：
       gcc [ option | filename ]...
       g++ [ option | filename ]...

       其中 option   為 gcc 使用時的選項(后面會再詳述)，而 filename 為欲以 gcc 處理的文件
說明：
      這 C 與 C++ 的 compiler 已將產生新程序的相關程序整合起來。產生一個新的程序需要經過四個階段：預處理、編譯、匯編,連結，而這兩個編譯器都能將輸入的文件做不同階段的處理。雖然原始程序的擴展名可用來分辨編寫原始程序碼所用的語言，但不同的 compiler ，其預設的處理程序卻各不相同：

       gcc    預設經由預處理過(擴展名為.i)的文件為 C 語言，并於程式連結階段以 C 的連結方式處理。

       g++    預設經由預處理過(擴展名為.i)的文件為 C++ 語言，并於程序連結階段以 C++ 的連結方式處理。

使用語法：
       gcc [ option | filename ]...
       g++ [ option | filename ]...

       其中 option   為 gcc 使用時的選項(后面會再詳述)，而 filename 為欲以 gcc 處理的文件
說明：
      這 C 與 C++ 的 compiler 已將產生新程序的相關程序整合起來。產生一個新的程序需要經過四個階段：預處理、編譯、匯編,連結，而這兩個編譯器都能將輸入的文件做不同階段的處理。雖然原始程序的擴展名可用來分辨編寫原始程序碼所用的語言，但不同的 compiler ，其預設的處理程序卻各不相同：

       gcc    預設經由預處理過(擴展名為.i)的文件為 C 語言，并於程式連結階段以 C 的連結方式處理。

       g++    預設經由預處理過(擴展名為.i)的文件為 C++ 語言，并於程序連結階段以 C++ 的連結方式處理。


       原始程序碼的擴展名指出所用編寫程序所用的語言，以及相對應的處
       理方法：

       .c    C 原始程序                     ；   預處理、編譯、匯編
       .C    C++ 原始程序                   ；   預處理、編譯、匯編
       .cc   C++ 原始程序                   ；   預處理、編譯、匯編
       .cxx  C++ 原始程序                   ；   預處理、編譯、匯編
       .m    Objective-C 原始程序           ；   預處理、編譯、匯編
       .i    已經過預處理之 C 原始程序    ；   編譯、匯編
       .ii   已經過預處理之 C++ 原始程序  ；   編譯、匯編
       .s    組合語言原始程序               ；   匯編
       .S    組合語言原始程序               ；   預處理、匯編
       .h    預處理文件(標頭文件)           ；   (不常出現在指令行)


       其他擴展名的文件是由連結程序來處理，通常有：

       .o    Object file
       .a    Archive file


       除非編譯過程出現錯誤，否則 "連結" 一定是產生一個新程序的最後階段。然而你也可以以 -c、-s 或 -E 等選項，將整個過程自四個階段中的其中一個停止。在連結階段，所有與原始碼相對應的.o 文件、程序庫、和其他無法自文件名辨明屬性的文件(包括不以 .o 為擴展名的 object file 以及擴展名為 .a 的 archive file)都會交由連結程序來處理(在指令行將那些文件當作連結程序的參數傳給連結程序)。


選項：
       不同的選項必須分開來下：例如 `-dr' 這個選項就與 `-d -r' 大不相同。

       絕大部份的 `-f' 及 `-W' 選項都有正反兩種形式：-fname 及-fno-name (或 -Wname 及 -Wno-name)。以下只列出非預設的那個形式。

       以下是所有選項的摘要。以形式來分類。選項的意義將另辟小節說明。

       一般性(概略、常用的)選項
              -c -S -E -o file -pipe -v -x language

       程序語言選項
              -ansi -fall-virtual -fcond-mismatch
              -fdollars-in-identifiers -fenum-int-equiv
              -fexternal-templates -fno-asm -fno-builtin
              -fno-strict-prototype -fsigned-bitfields
              -fsigned-char -fthis-is-variable
              -funsigned-bitfields -funsigned-char
              -fwritable-strings -traditional -traditional-cpp
              -trigraphs

       編譯時的警告選項
              -fsyntax-only -pedantic -pedantic-errors -w -W
              -Wall -Waggregate-return -Wcast-align -Wcast-qual
              -Wchar-subscript -Wcomment -Wconversion
              -Wenum-clash -Werror -Wformat -Wid-clash-len
              -Wimplicit -Winline -Wmissing-prototypes
              -Wmissing-declarations -Wnested-externs -Wno-import
              -Wparentheses -Wpointer-arith -Wredundant-decls
              -Wreturn-type -Wshadow -Wstrict-prototypes -Wswitch
              -Wtemplate-debugging -Wtraditional -Wtrigraphs
              -Wuninitialized -Wunused -Wwrite-strings

       除錯選項
              -a -dletters -fpretend-float -g -glevel -gcoff
              -gxcoff -gxcoff+ -gdwarf -gdwarf+ -gstabs -gstabs+
              -ggdb -p -pg -save-temps -print-file-name=library
              -print-libgcc-file-name -print-prog-name=program

       最佳化選項
              -fcaller-saves -fcse-follow-jumps -fcse-skip-blocks
              -fdelayed-branch -felide-constructors
              -fexpensive-optimizations -ffast-math -ffloat-store
              -fforce-addr -fforce-mem -finline-functions
              -fkeep-inline-functions -fmemoize-lookups
              -fno-default-inline -fno-defer-pop
              -fno-function-cse -fno-inline -fno-peephole
              -fomit-frame-pointer -frerun-cse-after-loop
              -fschedule-insns -fschedule-insns2
              -fstrength-reduce -fthread-jumps -funroll-all-loops
              -funroll-loops -O -O2

       預處理選項
              -Aassertion -C -dD -dM -dN -Dmacro[=defn] -E -H
              -idirafter dir -include file -imacros file -iprefix
              file -iwithprefix dir -M -MD -MM -MMD -nostdinc -P
              -Umacro -undef

       匯編程序選項
              -Wa,option

       連結程序選項
              -llibrary -nostartfiles -nostdlib -static -shared
              -symbolic -Xlinker option -Wl,option -u symbol

       目錄選項
              -Bprefix -Idir -I- -Ldir

       Target Options
              -b  machine -V version

       與機器(平臺)相關的選項
              M680x0 Options
              -m68000 -m68020 -m68020-40 -m68030 -m68040 -m68881
              -mbitfield -mc68000 -mc68020 -mfpa -mnobitfield
              -mrtd -mshort -msoft-float

              VAX Options
              -mg -mgnu -munix

              SPARC Options
              -mepilogue -mfpu -mhard-float -mno-fpu
              -mno-epilogue -msoft-float -msparclite -mv8
              -msupersparc -mcypress

              Convex Options
              -margcount -mc1 -mc2 -mnoargcount

              AMD29K Options
              -m29000 -m29050 -mbw -mdw -mkernel-registers
              -mlarge -mnbw -mnodw -msmall -mstack-check
              -muser-registers

              M88K Options
              -m88000 -m88100 -m88110 -mbig-pic
              -mcheck-zero-division -mhandle-large-shift
              -midentify-revision -mno-check-zero-division
              -mno-ocs-debug-info -mno-ocs-frame-position
              -mno-optimize-arg-area -mno-serialize-volatile
              -mno-underscores -mocs-debug-info
              -mocs-frame-position -moptimize-arg-area
              -mserialize-volatile -mshort-data-num -msvr3 -msvr4
              -mtrap-large-shift -muse-div-instruction
              -mversion-03.00 -mwarn-passed-structs

              RS6000 Options
              -mfp-in-toc -mno-fop-in-toc

              RT Options
              -mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
              -mfull-fp-blocks -mhc-struct-return -min-line-mul
              -mminimum-fp-blocks -mnohc-struct-return

              MIPS Options
              -mcpu=cpu type -mips2 -mips3 -mint64 -mlong64
              -mlonglong128 -mmips-as -mgas -mrnames -mno-rnames
              -mgpopt -mno-gpopt -mstats -mno-stats -mmemcpy
              -mno-memcpy -mno-mips-tfile -mmips-tfile
              -msoft-float -mhard-float -mabicalls -mno-abicalls
              -mhalf-pic -mno-half-pic -G num -nocpp

              i386 Options
              -m486 -mno-486 -msoft-float -mno-fp-ret-in-387

              HPPA Options
              -mpa-risc-1-0 -mpa-risc-1-1 -mkernel -mshared-libs
              -mno-shared-libs -mlong-calls -mdisable-fpregs
              -mdisable-indexing -mtrailing-colon

              i960 Options
              -mcpu-type -mnumerics -msoft-float
              -mleaf-procedures -mno-leaf-procedures -mtail-call
              -mno-tail-call -mcomplex-addr -mno-complex-addr
              -mcode-align -mno-code-align -mic-compat
              -mic2.0-compat -mic3.0-compat -masm-compat
              -mintel-asm -mstrict-align -mno-strict-align
              -mold-align -mno-old-align

              DEC Alpha Options
              -mfp-regs -mno-fp-regs -mno-soft-float -msoft-float

              System V Options
              -G -Qy -Qn -YP,paths -Ym,dir

       Code Generation Options
              -fcall-saved-reg -fcall-used-reg -ffixed-reg
              -finhibit-size-directive -fnonnull-objects
              -fno-common -fno-ident -fno-gnu-linker
              -fpcc-struct-return -fpic -fPIC
              -freg-struct-returno -fshared-data -fshort-enums
              -fshort-double -fvolatile -fvolatile-global
              -fverbose-asm

PRAGMAS
       Two  `#pragma'  directives  are  supported for GNU C++, to
       permit using the same header file for two purposes:  as  a
       definition  of  interfaces to a given object class, and as
       the full definition of the contents of that object  class.

       #pragma interface
              (C++  only.)   Use  this  directive in header files
              that define object classes, to save space  in  most
              of  the  object files that use those classes.  Nor-
              mally, local copies of certain information  (backup
              copies of inline member functions, debugging infor-
              mation, and the internal tables that implement vir-
              tual  functions)  must  be kept in each object file
              that includes class definitions.  You can use  this
              pragma  to  avoid  such duplication.  When a header
              file containing `#pragma interface' is included  in
              a  compilation, this auxiliary information will not
              be generated (unless the main input source file it-
              self  uses `#pragma implementation').  Instead, the
              object files will contain references to be resolved
              at link time.

       #pragma implementation

       #pragma implementation "objects.h"
              (C++  only.)  Use this pragma in a main input file,
              when you want  full  output  from  included  header
              files  to be generated (and made globally visible).
              The included  header  file,  in  turn,  should  use
              `#pragma  interface'.  Backup copies of inline mem-
              ber functions, debugging information, and  the  in-
              ternal  tables  used to implement virtual functions
              are all generated in implementation files.

              If you use `#pragma implementation' with  no  argu-
              ment,  it  applies to an include file with the same
              basename as  your  source  file;  for  example,  in
              `allclass.cc',  `#pragma  implementation' by itself
              is   equivalent    to    `#pragma    implementation
              "allclass.h"'.  Use the string argument if you want
              a single implementation file to include  code  from
              multiple header files.

              There  is no way to split up the contents of a sin-
              gle header file into multiple implementation files.

文件說明
       file.c             C source file
       file.h             C header (preprocessor) file
       file.i             經預處理過的 C source file
       file.C             C++ source file
       file.cc            C++ source file
       file.cxx           C++ source file
       file.m             Objective-C source file
       file.s             assembly language file
       file.o             object file
       a.out              link edited output
       TMPDIR/cc*         temporary files
       LIBDIR/cpp         preprocessor
       LIBDIR/cc1         compiler for C
       LIBDIR/cc1plus     compiler for C++
       LIBDIR/collect     linker front end needed on some machines
       LIBDIR/libgcc.a    GCC subroutine library
       /lib/crt[01n].o    start-up routine
       LIBDIR/ccrt0       additional start-up routine for C++
       /lib/libc.a        standard C library, 參閱 man page intro(3)
       /usr/include       standard directory for #include files
       LIBDIR/include     standard gcc directory for #include files
       LIBDIR/g++-include additional g++ directory for #include

       LIBDIR is usually /usr/local/lib/machine/version.
       TMPDIR comes from the environment variable TMPDIR (default
       /usr/tmp if available, else /tmp).

原文轉自：http://www.kjueaiud.com