Home » Newsletter » May 2011

By V. Subhash

Recently, I saw a question on StackOverflow.com that asked for the source code of a "hello world" program that could create a fully self-contained binary in 20 bytes or less. The accepted answer there is a Macro Assembler script.

In the mid-90s, I had read a Kris Jamsa book titled Jamsa's 1001 DOS and PC Tips that had introduced me to DEBUG Scripts. So, I was sure this task could be done using a DEBUG script.

In Windows versions up to XP, DEBUG is available as a DOS command. It can act as an interpreter and allow you to interactively enter assembly instructions and execute them. In other words, you literally tell what the CPU needs to do next in realtime. (Save raw opcode, this is as close you can get to the CPU.) Alternatively, you can store all your keyboard input in a text file and send it as input to the command. In both cases, DEBUG will allow you to generate a binary executable from your assembly code.

So, I went searching for the correct DOS interrupt and relevant instructions. I found that the INT21 instruction could output a string to the console but I soon got stuck trying to find a way to store the string in memory. Then, I found this link where I found a script that pretty much matched mine except for the last part. With a 20-byte limitation, there are not many ways of writing this script. The script stores the string after a RET instruction.

The following is my "hello world" DEBUG script. I named the file as HELLO.TXT.

A
MOV AH,9
MOV DX,108
INT 21
RET
DB 'HELLO WORLD$'

R CX
14
N MYHELLO.COM
W
Q
 

The A command allows you to enter the assembly instructions at code segment position 100 (CS:100). (DEBUG will also display current code position in memory after each instruction.) To output a string to the console using the DOS interrupt 21, set the AH register with the value of 09h and the DX register with the location where the string is stored. (You need to do test run first to determine the location of the string after all instructions have been entered.) These assembly instructions will take up 8 bytes. So, the location of the string will be 108. After setting the DX register, call the venerable Interrupt 21. Then, enter the RET instruction for exiting the program. At this point, use the DB instruction to enter the "Hello world" string. This string must end with the '$' character. Press the Enter key to signify the end of the assembly code. Now, you will see that the code position is at 14h. (In hexadecimal, this means the binary size is 20 bytes.) Next, set this value to the CX register. Use the N command to provide the name of the file to which the binary instructions needs to be saved. (I gave the name as MYHELLO.COM.) Use the W command to write to the file. Press the Enter key again to exit DEBUG.

With all the instructions in a text file (named HELLO.TXT), redirect it as the input for the DEBUG command at the DOS prompt.

DEBUG < HELLO.TXT

DEBUG will then create a binary executable. In my case, it was MYHELLO.COM. If you double-click this file in Windows Explorer, you will see a DOS window flash by and disappear. So, remain at the DOS prompt and execute the binary using its file name. I just typed MYHELLO and there it was - the smallest self-contained "Hello World" executable at 20 bytes! It does not rely on any library functions. It directly uses a DOS interrupt and finishes the job.

---oO0Oo---

* - Apparently, Windows 7 does not have DEBUG. (I am still holding on to XP.) It is really the end of an era!. Here is an interesting account from someone in Microsoft.

I once went into CompUSA in Chicago in the 80s, and just for fun I hacked up the Command.com on a few of their DOS-based laptops that were on display. I searched for the string "invalid command or filename" and replaced it with "Shop at Elek-Tek and SAVE!"

The cool thing about this trick was that the bogus error message didn't appear until after the computer had been rebooted, so even if a friendly little salesman snuck up on me, I could play dumb, say I got lost in something, and the computer seemed to work fine. But the next day, when the salesman made a mistake showing off some feature to a customer, up came that replacement message I had entered for him.