Decompilation techniques for DOS .COM files

Question

I've found a binary copy of a DOS freeware game I was fond of back in the day (Firefighter, downloadable as FIRE.COM) but the source code is not available. I'd like to re-implement it in C or Python using Unix curses and while I'm fully competent to do that part, I know almost nothing about tools for reverse-engineering and decompiling DOS .COM file binaries.

Going by documentation that says .COM is headerless raw machine code, I tried running objdump -D -b binary -m i386 on it. The result didn’t look valid. objdump -D -b binary -m i8086 –adjust-vma=0×100 appeared to do the trick as did using NASM with ndisasm -b 16 -a -o 0×100 and both of those do seem to produce plausible assembly listings.

Ideally, though, I'm looking for something that decompiles and renders to C in as-if-it-were-assembler style with generated variable names and useful annotations on the DOS traps. I understand that such a beast may not exist, but on the chance one does I'm seeking recommendations.

Answer 1

My answer is a little late; newcomer to this site. The Decompiler project was initiated in order to decompile MS-DOS EXE and COM binaries. The project has both a command-line and a GUI tool:

https://sourceforge.net/projects/decompiler/

Use the following command with the command-line tool to decompile COM programs:

decompile --default-to ms-dos-com myprog.com

In the GUI, use the menu command File > Open as... to open the COM file and specify a start address like 0800:0100.

Answer 2

IDA Pro and a number of other disassemblers will disassemble your file, assuming that it isn't compressed or otherwise obfuscated. But this is only the first step in producing a working assembly language program, which you will need if you are to translate it into a higher language.

Essentially you are looking at two projects. The first is to produce a working assembly language version of your game. The second is to translate the game mechanics and logic into C++ or whatever. The first is easier than the second.

Answer 3

I recently found a DOS program which turns .COM and .EXE into a simple "pseudo-code". The program is --- exe2c.zip

http://www.codeforge.com/dlpre/201558/cb2e19437871.zip__.html

Not an "excellent" tool, but i found it useful. Start by executing ENVMNU.EXE and then entering the .COM or .EXE filename. ---(hint?) If better results are necessary, use the individual modules ( such as the DOS to ASM module ) and hand-edit ASM discrepancies then feed them to the ASM to C module and observe the results.

-- William Gallant

Answer 4

no beast exist that will decompile 16 bit com program
if you prefer to debug it look for grdb from ladsoft
also have in hand a copy of ralf browns interrupt list for all the
int 21 / int 10 SetCursor / GetCursor calls

also make sure you rename the downloaded exe to be 8dot3 compatible
(name less than 8 charecters)

GRDB.EXE firefi.com    

Get Real Debugger Version 9.6  Copyright (c) 1997-2009 David Lindauer (LADSoft)
GRDB comes with ABSOLUTELY NO WARRANTY, for details type `?g'
This is free software, and you are welcome to redistribute it
under certain conditions; type `?gr' for details

DPMI Start code hooked
History enabled
eax:00000000 ebx:00000000 ecx:00007500 edx:00000000 esi:00000000 edi:00000000
ebp:00000000 esp:0000FFEE eip:00000100 flag:00000202 NV UP EI PL NZ NA PO NC
ds:1F25 es:1F25 fs:1F25 gs:1F25 ss:1F25 cs:1F25
1F25:0100 E9 25 27       jmp          2828
Size: 00007500
->  

Answer 5

As I review your application (Fire.COM), I think you should use some DOS debugger like Turbo Debugger 5.0 to debug and reverse this application because I have some encrypted stage in it. If you want to "decompile" these application to C code, it's very hard, I had tried some tools in the past (2008) but there was not a good tool to do it. I also got a C source but there was many inline assembly in it.

Answer 6

For .com (MS DOS) programs I use ndisasm (Netwide Disassembler) as mentioned in this answer. It is very simple, just work without any magic.