binutils - ar, nm, objdump

libjpeg.a 라는 녀석을 발견하고는 어떻게 생겼나 궁금증이 생겼다.

$ ll libjpeg.a
-rw-r--r-- 1 root root 166012 2007-10-01 23:36 libjpeg.a

$ file libjpeg.a
libjpeg.a: current ar archive

$ nm -A libjpeg.a
libjpeg.a:jcapimin.o:00000330 T jpeg_CreateCompress
libjpeg.a:jcapimin.o:         U jpeg_abort
libjpeg.a:jcapistd.o:         U jinit_compress_master
libjpeg.a:jcapistd.o:000001b0 T jpeg_start_compress
libjpeg.a:jctrans.o:000000c0 t compress_output
libjpeg.a:jctrans.o:         U jinit_c_master_control
libjpeg.a:jcparam.o:000008a0 t add_huff_table
libjpeg.a:jcparam.o:000000e2 r bits_ac_chrominance.3846
... 파일별로 반복
'
U,T,r 등은 symbol type 이라고 하며,
위의 U는 Undefined 을, T는 Text section(code) 을 의미한다.

$ nm --help
Usage: nm [option(s)] [file(s)]
 List symbols in [file(s)] (a.out by default).
 The options are:
  -a, --debug-syms       Display debugger-only symbols
  -A, --print-file-name  Print name of the input file before every symbol
  -B                     Same as --format=bsd
  -C, --demangle[=STYLE] Decode low-level symbol names into user-level names
                          The STYLE, if specified, can be `auto' (the default),
                          `gnu', `lucid', `arm', `hp', `edg', `gnu-v3', `java'
                          or `gnat'
      --no-demangle      Do not demangle low-level symbol names
  -D, --dynamic          Display dynamic symbols instead of normal symbols
      --defined-only     Display only defined symbols
  -e                     (ignored)
  -f, --format=FORMAT    Use the output format FORMAT.  FORMAT can be `bsd',
                           `sysv' or `posix'.  The default is `bsd'
  -g, --extern-only      Display only external symbols
  -l, --line-numbers     Use debugging information to find a filename and
                           line number for each symbol
  -n, --numeric-sort     Sort symbols numerically by address
  -o                     Same as -A
  -p, --no-sort          Do not sort the symbols
  -P, --portability      Same as --format=posix
  -r, --reverse-sort     Reverse the sense of the sort
  -S, --print-size       Print size of defined symbols
  -s, --print-armap      Include index for symbols from archive members
      --size-sort        Sort symbols by size
      --special-syms     Include special symbols in the output
      --synthetic        Display synthetic symbols as well
  -t, --radix=RADIX      Use RADIX for printing symbol values
      --target=BFDNAME   Specify the target object format as BFDNAME
  -u, --undefined-only   Display only undefined symbols
  -X 32_64               (ignored)
  @FILE                  Read options from FILE
  -h, --help             Display this information
  -V, --version          Display this program's version number

nm: supported targets: elf32-i386 a.out-i386-linux efi-app-ia32 efi-bsdrv-ia32 efi-rtdrv-ia32 elf32-little elf32-big elf64-x86-64 efi-app-x86_64 efi-bsdrv-x86_64 efi-rtdrv-x86_64 elf64-little elf64-big srec symbolsrec tekhex binary ihex trad-core
Report bugs to <http://www.sourceware.org/bugzilla/>.

$ man nm
NM(1)                                     GNU Development Tools                                    NM(1)

NAME
       nm - list symbols from object files

SYNOPSIS
       nm [-a|--debug-syms] [-g|--extern-only]
          [-B] [-C|--demangle[=style]] [-D|--dynamic]
          [-S|--print-size] [-s|--print-armap]
          [-A|-o|--print-file-name][--special-syms]
          [-n|-v|--numeric-sort] [-p|--no-sort]
          [-r|--reverse-sort] [--size-sort] [-u|--undefined-only]
          [-t radix|--radix=radix] [-P|--portability]
          [--target=bfdname] [-fformat|--format=format]
          [--defined-only] [-l|--line-numbers] [--no-demangle]
          [-V|--version] [-X 32_64] [--help]  [objfile...]

DESCRIPTION
       GNU  nm  lists  the symbols from object files objfile....  If no object files are listed as argu-
       ments, nm assumes the file a.out.

       For each symbol, nm shows:

       o   The symbol value, in the radix selected by options (see below), or hexadecimal by default.

       o   The symbol type.  At least the following types are used; others are, as  well,  depending  on
           the  object  file  format.   If  lowercase,  the symbol is local; if uppercase, the symbol is
           global (external).

           "A" The symbol's value is absolute, and will not be changed by further linking.

           "B"
           "b" The symbol is in the uninitialized data section (known as BSS).

           "C" The symbol is common.  Common symbols are uninitialized  data.   When  linking,  multiple
               common  symbols  may  appear  with the same name.  If the symbol is defined anywhere, the
               common symbols are treated as undefined references.

           "D"
           "d" The symbol is in the initialized data section.

           "G"
           "g" The symbol is in an initialized data section for small objects.  Some object file formats
               permit  more  efficient  access  to  small data objects, such as a global int variable as
               opposed to a large global array.

           "I" The symbol is an indirect reference to another symbol.  This is a GNU  extension  to  the
               a.out object file format which is rarely used.

           "i" The symbol is in a section specific to the implementation of DLLs.

           "N" The symbol is a debugging symbol.

           "p" The symbols is in a stack unwind section.

           "R"

           "r" The symbol is in a read only data section.

           "S"
           "s" The symbol is in an uninitialized data section for small objects.

           "T"
           "t" The symbol is in the text (code) section.

           "U" The symbol is undefined.

           "V"
           "v" The  symbol is a weak object.  When a weak defined symbol is linked with a normal defined
               symbol, the normal defined symbol is used with no error.  When a weak undefined symbol is
               linked  and  the symbol is not defined, the value of the weak symbol becomes zero with no
               error.  On some systems, uppercase indicates that a default value has been specified.

           "W"
           "w" The symbol is a weak symbol that has not been specifically tagged as a weak  object  sym-
               bol.   When  a  weak  defined  symbol  is linked with a normal defined symbol, the normal
               defined symbol is used with no error.  When a weak undefined symbol  is  linked  and  the
               symbol  is not defined, the value of the symbol is determined in a system-specific manner
               without error.  On some systems, uppercase indicates that a default value has been speci-
               fied.

           "-" The  symbol  is  a  stabs  symbol in an a.out object file.  In this case, the next values
               printed are the stabs other field, the stabs desc field, and the stab type.   Stabs  sym-
               bols are used to hold debugging information.

           "?" The symbol type is unknown, or object file format specific.

$ ar -t libjpeg.a
jcapimin.o
jcapistd.o
jctrans.o
jcparam.o
jdatadst.o
...

$ ar --help
Usage: ar [emulation options] [-]{dmpqrstx}[abcfilNoPsSuvV] [member-name] [count] archive-file file...
       ar -M [<mri-script]
 commands:
  d            - delete file(s) from the archive
  m[ab]        - move file(s) in the archive
  p            - print file(s) found in the archive
  q[f]         - quick append file(s) to the archive
  r[ab][f][u]  - replace existing or insert new file(s) into the archive
  t            - display contents of archive
  x[o]         - extract file(s) from the archive
 command specific modifiers:
  [a]          - put file(s) after [member-name]
  [b]          - put file(s) before [member-name] (same as [i])
  [N]          - use instance [count] of name
  [f]          - truncate inserted file names
  [P]          - use full path names when matching
  [o]          - preserve original dates
  [u]          - only replace files that are newer than current archive contents
 generic modifiers:
  [c]          - do not warn if the library had to be created
  [s]          - create an archive index (cf. ranlib)
  [S]          - do not build a symbol table
  [T]          - make a thin archive
  [v]          - be verbose
  [V]          - display the version number
  @<file>      - read options from <file>
 emulation options:
  No emulation specific options
ar: supported targets: elf32-i386 a.out-i386-linux efi-app-ia32 efi-bsdrv-ia32 efi-rtdrv-ia32 elf32-little elf32-big elf64-x86-64 efi-app-x86_64 efi-bsdrv-x86_64 efi-rtdrv-x86_64 elf64-little elf64-big srec symbolsrec tekhex binary ihex trad-core
Report bugs to <http://www.sourceware.org/bugzilla/>

$ objdump -x libjpeg.a
In archive libjpeg.a:

jcapimin.o:     file format elf32-i386
rw-r--r-- 0/0   2296 Oct  1 23:36 2007 jcapimin.o
architecture: i386, flags 0x00000011:
HAS_RELOC, HAS_SYMS
start address 0x00000000

Sections:
SYMBOL TABLE:
RELOCATION RECORDS FOR [.text]:
... 파일별로 반복

$ objdump --help
Usage: objdump <option(s)> <file(s)>
 Display information from object <file(s)>.
 At least one of the following switches must be given:
  -a, --archive-headers    Display archive header information
  -f, --file-headers       Display the contents of the overall file header
  -p, --private-headers    Display object format specific file header contents
  -h, --[section-]headers  Display the contents of the section headers
  -x, --all-headers        Display the contents of all headers
  -d, --disassemble        Display assembler contents of executable sections
  -D, --disassemble-all    Display assembler contents of all sections
  -S, --source             Intermix source code with disassembly
  -s, --full-contents      Display the full contents of all sections requested
  -g, --debugging          Display debug information in object file
  -e, --debugging-tags     Display debug information using ctags style
  -G, --stabs              Display (in raw form) any STABS info in the file
  -W, --dwarf              Display DWARF info in the file
  -t, --syms               Display the contents of the symbol table(s)
  -T, --dynamic-syms       Display the contents of the dynamic symbol table
  -r, --reloc              Display the relocation entries in the file
  -R, --dynamic-reloc      Display the dynamic relocation entries in the file
  @<file>                  Read options from <file>
  -v, --version            Display this program's version number
  -i, --info               List object formats and architectures supported
  -H, --help               Display this information

 The following switches are optional:
  -b, --target=BFDNAME           Specify the target object format as BFDNAME
  -m, --architecture=MACHINE     Specify the target architecture as MACHINE
  -j, --section=NAME             Only display information for section NAME
  -M, --disassembler-options=OPT Pass text OPT on to the disassembler
  -EB --endian=big               Assume big endian format when disassembling
  -EL --endian=little            Assume little endian format when disassembling
      --file-start-context       Include context from start of file (with -S)
  -I, --include=DIR              Add DIR to search list for source files
  -l, --line-numbers             Include line numbers and filenames in output
  -F, --file-offsets             Include file offsets when displaying information
  -C, --demangle[=STYLE]         Decode mangled/processed symbol names
                                  The STYLE, if specified, can be `auto', `gnu',
                                  `lucid', `arm', `hp', `edg', `gnu-v3', `java'
                                  or `gnat'
  -w, --wide                     Format output for more than 80 columns
  -z, --disassemble-zeroes       Do not skip blocks of zeroes when disassembling
      --start-address=ADDR       Only process data whose address is >= ADDR
      --stop-address=ADDR        Only process data whose address is <= ADDR
      --prefix-addresses         Print complete address alongside disassembly
      --[no-]show-raw-insn       Display hex alongside symbolic disassembly
      --adjust-vma=OFFSET        Add OFFSET to all displayed section addresses
      --special-syms             Include special symbols in symbol dumps

objdump: supported targets: elf32-i386 a.out-i386-linux efi-app-ia32 efi-bsdrv-ia32 efi-rtdrv-ia32 elf32-little elf32-big elf64-x86-64 efi-app-x86_64 efi-bsdrv-x86_64 efi-rtdrv-x86_64 elf64-little elf64-big srec symbolsrec tekhex binary ihex trad-core
objdump: supported architectures: i386 i386:x86-64 i8086 i386:intel i386:x86-64:intel

The following i386/x86-64 specific disassembler options are supported for use
with the -M switch (multiple options should be separated by commas):
  x86-64      Disassemble in 64bit mode
  i386        Disassemble in 32bit mode
  i8086       Disassemble in 16bit mode
  att         Display instruction in AT&T syntax
  intel       Display instruction in Intel syntax
  att-mnemonic
              Display instruction in AT&T mnemonic
  intel-mnemonic
              Display instruction in Intel mnemonic
  addr64      Assume 64bit address size
  addr32      Assume 32bit address size
  addr16      Assume 16bit address size
  data32      Assume 32bit data size
  data16      Assume 16bit data size
  suffix      Always display instruction suffix in AT&T syntax
Report bugs to <http://www.sourceware.org/bugzilla/>.

mangling은 컴파일러에서 이름이 중복되지 않도록 독특한 이름을 지어주는 것으로
C++에서 overloading을 지원하기 하는데 사용되기도 한다고 한다. [링크 : http://rubyeye.kr/]

int _cdecl    f (int x) { return 0; }	//	_f
int _stdcall  g (int y) { return 0; }	//	_g@4
int _fastcall h (int z) { return 0; }	//	@h@4

위키피디아에서 참조한 녀석으로 MS 컴파일러가 C언어 맹글링하는 방식의 예제이다.
[링크 : http://en.wikipedia.org/wiki/Name_mangling]

그리고 이러한 녀석을 원래이름으로 돌려주는 유틸리티로 c++filt가 존재한다.
[참고 : http://kldp.org/node/68410]

$ c++filt --help
Usage: c++filt [options] [mangled names]
Options are:
  [-_|--strip-underscore]     Ignore first leading underscore
  [-n|--no-strip-underscore]  Do not ignore a leading underscore (default)
  [-p|--no-params]            Do not display function arguments
  [-i|--no-verbose]           Do not show implementation details (if any)
  [-t|--types]                Also attempt to demangle type encodings
  [-s|--format {none,auto,gnu,lucid,arm,hp,edg,gnu-v3,java,gnat}]
  [@<file>]                   Read extra options from <file>
  [-h|--help]                 Display this information
  [-v|--version]              Show the version information
Demangled names are displayed to stdout.
If a name cannot be demangled it is just echoed to stdout.
If no names are provided on the command line, stdin is read.
Report bugs to <http://www.sourceware.org/bugzilla/>.