I'm learning x86 assembly on OS X 10.6, how do I compile?

I'm about to learn x86 assembly language on Mac OS X. I'm using as instruction to compile assembly file in commend window. But I have several errors and I don't know how I can get through.

Here are the errors and my assembly code, which is quite simple.

**ung-mi-lims-macbook-pro:pa2 ungmi$ as swap.s
swap.s:16:Unknown pseudo-op: .type
swap.s:16:Rest of line ignored. 1st junk character valued 115 (s).
swap.s:19:suffix or operands invalid for `push'
swap.s:46:suffix or operands invalid for `pop'
ung-mi-lims-macbook-pro:pa2 ungmi$** 

and the source is

.text
      开发者_如何学Go  .align 4
.globl  swap
        .type swap,@function

swap:
        pushl   %ebp
        movl    %esp, %ebp

        movl    %ebp, %esp
        popl    %ebp
        ret

and I searched some solution which is I have to put -arch i386 than

**ung-mi-lims-macbook-pro:pa2 ungmi$ as -arch i386 swap.s
swap.s:16:Unknown pseudo-op: .type
swap.s:16:Rest of line ignored. 1st junk character valued 115 (s).
ung-mi-lims-macbook-pro:pa2 ungmi$** 

Could you help me out? Just let me know what I need to compile assembly file. I have XCode already and I'd rather to do this with commend window and vi editor.

Your code is 32-bit assembly. Compilation on Snow Leopard is 64-bit by default, but you can use gcc -m32 for compiling 32-bit code. There are of course also options that you can pass to as and ld but I found that remembering only the option for gcc was enough, since it is a front-end to all these things.

Use gcc to show you an example of assembly code that works: type a minimal C function in file test.c and use gcc -S test.c to produce the assembly in file test.s.

Example:

int x;

void f(void)
{
  int i;
  for (i = 0; i < 5; i++)  x = x + 1;
}

is compiled in assembly on this Leopard Mac:

    .text
.globl _f
_f:
    pushl   %ebp
    movl    %esp, %ebp
    subl    $24, %esp
    call    L6
"L00000000001$pb":
L6:
    popl    %ecx
    movl    $0, -12(%ebp)
    jmp L2
L3:
    leal    L_x$non_lazy_ptr-"L00000000001$pb"(%ecx), %eax
    movl    (%eax), %eax
    movl    (%eax), %eax
    leal    1(%eax), %edx
    leal    L_x$non_lazy_ptr-"L00000000001$pb"(%ecx), %eax
    movl    (%eax), %eax
    movl    %edx, (%eax)
    leal    -12(%ebp), %eax
    incl    (%eax)
L2:
    cmpl    $4, -12(%ebp)
    jle L3
    leave
    ret
.comm _x,4,2
    .section __IMPORT,__pointers,non_lazy_symbol_pointers
L_x$non_lazy_ptr:
    .indirect_symbol _x
    .long   0
    .subsections_via_symbols

You can make it simpler by using gcc option -fno-PIC:

    .text
.globl _f
_f:
    pushl   %ebp
    movl    %esp, %ebp
    subl    $24, %esp
    movl    $0, -12(%ebp)
    jmp L2
L3:
    movl    _x, %eax
    incl    %eax
    movl    %eax, _x
    leal    -12(%ebp), %eax
    incl    (%eax)
L2:
    cmpl    $4, -12(%ebp)
    jle L3
    leave
    ret
.comm _x,4,2
    .subsections_via_symbols

On Snow Leopard you will get 64-bit assembly (amd64). You could get the same thing on Leopard by using the -m64 gcc option, and you can get on Snow Leopard the same thing you'd get running gcc on Leopard with the -m32 option.

Use GCC, it'll compile and link for you: cc swap.s

Apple's as is forked from binutils 1.38, while the current version is 2.20, so there may be many new directives not available. I can't find the manual for 1.38, but the oldest one (2.10) I could find states that .type is associated only with COFF format output. The object format on Mac is Mach-O, so the .type directive is meaningless. Just remove it.

Recently I have ran in to this issue as well trying to compile Intel x86 on Mac OS X:

For nasm:

-o hello.tmp - outfile
-f macho - specify format
Linux - elf or elf64
Mac OSX - macho

For ld:

-arch i386 - specify architecture (32 bit assembly)
-macosx_version_min 10.6 (Mac OSX - complains about default specification)
-no_pie (Mac OSX - removes ld warning)
-e main - specify main symbol name (Mac OSX - default is start)
-o hello.o - outfile

For Shell:

./hello.o - execution

One-liner:

nasm -o hello.tmp -f macho hello.s && ld -arch i386 -macosx_version_min 10.6 -no_pie -e _main -o hello.o hello.tmp && ./hello.o

Let me know if this helps!

I wrote how to do it on my blog here:

http://blog.burrowsapps.com/2013/07/how-to-compile-helloworld-in-intel-x86.html

For a more verbose explanation, I explained on my Github here:

https://github.com/jaredsburrows/Assembly