GCC 編譯器使用指南

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

    其中 option   為 gcc 使用時的選項(後面會再詳述)，
      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、彙編,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，並於程
        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，並於程

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、彙編
    .C   C++ 原始程序             ；   預處理、編譯、彙編
    .cc   C++ 原始程序             ；   預處理、編譯、彙編
    .cxx C++ 原始程序             ；   預處理、編譯、彙編
    .m   Objective-C 原始程序       ；   預處理、編譯、彙編
    .i   已經過預處理之 C 原始程序   ；   編譯、彙編
    .ii   已經過預處理之 C++ 原始程序 ；   編譯、彙編
    .s   組合語言原始程序           ；   彙編
    .S   組合語言原始程序           ；   預處理、彙編
    .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。選項的意義將另闢小節說
    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

    彙編程序選項
        -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(平台)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).