Hi guys,

I hope that this thread is still alive!

Following "Shirk" excellent advices and tips, I successfully built FASM 1.70.02 on my OSX machines (32-bits and 64-bits).

I want to share the binary with everyone interested on programming with FASM under OSX (see USAGE file).

QUESTION for "Tomasz Grysztar": "Shrink" provided a patch (fasm-fix-out-of-memory-lion.diff.txt) which allow to compile FASM on OSX. Any chance to get it commited to FASM source code?

Thanks
Zab

good one

Uh, this got sticky - nice

Apparently I'm back to my pet os project, back to fasm and we still need the patched version.. (checked no path in 1.70.3 or 1.71 previews).

Not sure if Tomasz reads this forum / has interest in the patch but big thanks to zab for providing the binaries!

Maybe someone could mail him?

I cannot approve such kind of fix. If that one relocation is processed incorrectly then the other relocations also cannot be trusted and it is inherently unsafe to try fixing it this way. One should rather search for the real source of the problem, that is the underlying bug related to relocations - and report it.

Hi Tomasz,
please don't get me wrong - my post was in no way meant to insult anyone.
It was just written in a challenging tone to lure an answer out of you.

I agree that the real fix would be to determine why this one relocation is messed up. If you ask me - it's a flaw in objconv.. I looked through most of the code by now and everything else seems fine. (could be related to different reloc handling in mach-o).

But to be honest - osx or mach-o is nowhere near your target / planned features list and I neither have the time nor the fasm-guts to dig through the whole code to add a mach-o backend. So it looks like mac users are still dependent on this (possibly flawed, however working-fine-so-far) patched binary.
I'm fine with that and most of us can resort back to a VM if hard comes to hard..

so keep the thread sticky and most of us should be happy

Cheers,
Shirk

P.S. - my vote for a fasm mach-o backend

Working on 10.8.2 with 1.71.06

Weird, can compile and run fasm.asm with fasm, but try to compile for efi and back to "out of memory"

ok changed buffer rb 1000h in fasm.asm to buffer rb 4000h
Now compiles efi

Just compiled fasm.efi and works
I should note that fast.asm for efi already had buffer set to 4000h
so that appears to be the "out of memory" bug

OS X appears to need more buffer space as 1000h leads to "Out Of Memory" errors and also leads to some examples that have buffer set to 1000h, getting errors too.

Is there a reason why OS X needs bigger buffer space??
Of course, it won't work otherwise.

So set the buffer to 4000h and you can do wonders.

tools/libc/listing.asm is one example that now works.
As well as compiling fasm and fasm.efi from newest sources.

All this on Mountain Lion.

Anybody interested???

Yep, as soon as my monitor works again.. And if i can get mountain lion up on a custom

I've got ML running now.. it might be interesting to get fasm running

I found a few OSX Nasm examples too.. there was a quartz and an OPENGL one.. and a boilerplated hello world

http://michaux.ca/articles/assembly-hello-world-for-os-x

http://forum.nasm.us/index.php?topic=1075.0 (OpenGL)

http://osxbook.com/blog/2009/03/15/crafting-a-tiny-mach-o-executable/

https://gist.github.com/1084476

http://www.feiri.de/macho/

http://seriot.ch/hello_macho.php

looks like others have, and continue to play with macho-o. I think the above examples are good for understanding the bare minimum requirements for a macho32 and macho64. I just nasm assembled the second example off git and it works:

Code:

$nasm -O0 -f bin -o tiny_hello gistfile2.asm $chmod +x tiny_hello $wc -c tiny_hello      251 tiny_hello $./tiny_hello  Hello, World!

251 bytes on ML! hehe small enough to get to know macho bit by bit and byte by byte


im trying to see if i can get some function out of tinycc on osx with objconv [edit: tinycc can be used same way fasm works with objconvert and ld.. ld adds a lot of weight, probably a lot unnecessary.]

here are some gui programs used as tools to examine macho:

http://sourceforge.net/projects/machoview/

http://www.affinic.com/?page_id=109

http://opensource.apple.com/release/developer-tools-45/

the source for cctools including ld, otool, as etc

Nice analysis on mach-o so far -- I had a look into a fasm mach-o backend but tbh. couldn't find my way around mach-o 64..

@STLVNUB - the buffer size problems could be a result of things like the additional alignment requirements and other features like these that Xnu enforces.

I'm getting close to declaring official support of Mac OS X in my C compiler and just for the fun of it I made FASM executables out of fasm.o for Linux, Mac OS X and DOS/DPMI.

I used my compiler's C library and two pages of simple asm & C code to translate names (leading underscores are used by default in my compiler), preserve regs (my compiler preserves only the obvious: EBP, ESP) and intercept fopen() (it should really be called with "rb" and "wb" and not "r" and "w") and gettimeofday() (not implemented in my library, not part of the C standard).

The result seems to be working so far. I compiled a few C apps with my compiler instructed to use FASM in place of NASM.

Btw, the DOS/DPMI version is somewhat "cleaner" in that it does not try to use unreal mode.

If anyone is interested I could share the code & instructions and the binary(-ies).

So, it looks like my Smaller C compiler is working on Mac OS X 10.6 (Snow Leopard) through Mac OS X 10.12 (Sierra).

Here's how I turn fasm.o into various executables.

Code:

; file: compata.asm bits 32 section .text %macro wrapper 2 extern %2 global %1 %1:     push    ebx     push    ecx     push    edx     push    esi     push    edi     pushfd     push    dword [esp+(6+4)*4]     push    dword [esp+(6+4)*4]     push    dword [esp+(6+4)*4]     push    dword [esp+(6+4)*4]     call    %2     add     esp, 4*4     popfd     pop     edi     pop     esi     pop     edx     pop     ecx     pop     ebx     ret %endmacro wrapper malloc, _malloc wrapper free,   _free wrapper getenv, _getenv wrapper fopen,  _fopen_wrapper wrapper fclose, _fclose wrapper fread,  _fread wrapper fwrite, _fwrite wrapper fseek,  _fseek wrapper ftell,  _ftell wrapper time,   _time wrapper exit,   _exit wrapper write,  _write wrapper gettimeofday, _gettimeofday_wrapper extern main global _main _main:     cld     jmp     main

Code:

// file: compatc.c #include <string.h> #include <stdio.h> #include <time.h> FILE* fopen_wrapper(const char* name, const char* mode) {   if (!strcmp(mode, "r"))     mode = "rb";   else if (!strcmp(mode, "w"))     mode = "wb";   else     return NULL;   return fopen(name, mode); } struct my_timeval {   unsigned tv_sec, tv_usec; }; int gettimeofday_wrapper(struct my_timeval* tp, void* tzp) {   (void)tzp;   tp->tv_sec = time(NULL);   tp->tv_usec = 0;   return 0; }

Compiling for MacOS:

Code:

smlrcc compata.asm compatc.c fasm.o -o fasmm -macos

Compiling for DOS/DPMI:

Code:

smlrcc compata.asm compatc.c fasm.o -o fasmdp.exe -dosp -minheap 17000000

Compiling for Windows:

Code:

smlrcc compata.asm compatc.c fasm.o -o fasmw.exe -win

My compiler depends on NASM and the above asm code is for NASM, so you should have NASM 2.03 or newer if you want to recompile FASM using Smaller C and the above pieces of code.

Smaller C should work out of the box on Windows and DOS, just include the path to v0100\binw or v0100\bind in %PATH%.

On Linux and Mac OS X you can simply recompile the compiler, using the usual (you will need NASM already installed):

Code:

./configure make sudo make install

On Mac you may see an error from readlink, but it should still work.

I'm attaching the binaries.

What else?.. Feel free to poke around the compiler's library source for Mac OS X system calls. Ditto for the linker if you want to make 32-bit Mach-O executables by hand. Or you could just use the linker as-is.