The Andromeda Decompiler is not available to the public at present, so the only example of its use is usually limited to the very impressive demo program. I asked the author to compile the switch test program; here is the result:

<verbatim>
void sub_401080(DWORD arg_4)
{
	DWORD	var_c;
	sub_401440();
	__main();
	switch (arg_4)
	{
	case 7:
		var_c = "Seven!";
		goto loc_4010D0;
	case 6:
		var_c = "Six!";
		goto loc_4010D0;
	case 5:
		var_c = "Five!";
		goto loc_4010D0;
	case 4:
		var_c = "Four!";
		goto loc_4010D0;
	case 3:
		var_c = "Three!";
		goto loc_4010D0;
	case 2:
		var_c = "Two!";
loc_4010D0:
		puts((void *)var_c);
		return;
	}
	var_c = "Other!";
	goto loc_4010D0;
}//sub_401080(:[1])
</verbatim>

This will compile with little effort, and looks like it will run correctly. Obviously, the goto statements are unfortunate. However, this is possibly because the author has not tested before on gcc-compiled input programs.

I'm not sure where the binary for this program came from; it is in a file called =switch_gcc.dc=. To give some idea of the original binary, this is the assembler view:

<pre>
sub_401080	proc near
  mov	 [eax], ebp
  mov	 eax, 0
  mov	 ebp, esp
  sub	 esp, 8
  and	 esp, 0FFFFFFF0h
  mov	 [eax-4], ebx
  mov	 ebx, [eax+8]
  call	sub_401440
  call	__main
  cmp	 ebx, 7
  ja	  loc_4010E0
  <i>T32 = ebx * 4, T32 = &off_4010A8 + T32, (*T32)</i>
loc_40110D:
  mov	 dword ptr [eax], offset aSeven
  jmp	 loc_4010D0

loc_40110D4:
  mov	 dword ptr [eax], offset aSix
  jmp	 loc_4010D0
...
loc_4010C8:
  mov	 fdword ptr [eax], offset aTwo
loc_4010D0:
  call	puts
  mov	 ebx, [eax-4]
  mov	 eax, 0
  mov	 esp, ebp
  mov	 ebp, [eax]
  ret

loc_4010E0:
  mov	 dword ptr [eax], offset aOther
  jmp	 loc_4010D0
sub_401080	endp
</pre>

This assembler code is like nothing I've seen; I suspect that esp is being displayed as eax and possibly some instructions are displayed out of sequence. This was from version 0.62; unfortunately the =.dc= file won't read in later versions of the GUI.

It's a real shame that this decompiler doesn't seem to have progressed since May 2005.

-- Main.MikeVanEmmerik - 24 Mar 2007