I am doing something wrong but I'm not sure what...

Code:

format PE console entry start include 'C:\fasm\include\win32ax.inc' section '.idata' import data readable  library kernel,'kernel32.dll', msvcrt,'msvcrt.dll', User32,'User32.dll'        import kernel, ExitProcess,'ExitProcess'   import msvcrt, printf,'printf'     import User32, wsprintf,'wsprintfA' section '.data' data readable writeable   strInteger db '%d', 13, 10 , 0     strBuffer rb 64      a_msg db 'a: ', 0 section '.code' code readable executable start:  mov eax, -22         mov ebx, 10  xor edx, edx         div ebx      mov esi, eax         ccall [wsprintf], strBuffer, strInteger, esi         ccall [printf], a_msg        ccall [printf], strBuffer    stdcall [ExitProcess],0

Basically, I am doing an integer division: (-22) / 10
I would expect the result to be (-2) but instead, I get 429'496'727
The division works correctly with positive numbers.

Use idiv instruction.
[edit]And cdq instead of "xor edx, edx"

That was fast!

Very interesting. I'll have to look into cdq, it looks like idiv doesn't give the right result without it.

IDIV works with EDX:EAX and you set EDX to zero - EDX must have sign extended over all its bits.
CDQ does just that - it takes the bit sign from EAX (bit #31) and sets all bits in EDX to that bit value.