In the following code-
000000D3 invoke MessageBoxA,0,soffptr,0,0
000000D3 6A00 * push 0
000000D5 6A00 * push 0
000000D7 68A9000000 * push offset soffptr
000000DC 6A00 * push 0
000000DE E800000000 * call MessageBoxA
Why does it say " push offset soffptr " when in fact it is pushing the value of soffptr?
soffptr is probably defined as
soffptr DB "Text of the message",0
so it is not a value (text of the message), which is being pushed, but offset of the string (encoded as immediate number, similary as other three operands). The string is at position 0x000000A9 in data segment during assembly-time, and those four bytes will be relocated to its virtual address at run-time.
The listing looks wrong indeed, but it is impossible to judge without seeing the complete source.
Tested with Asmc Version 2.31.17 :
.386
.model flat,stdcall
option casemap:none
MessageBoxA PROTO :DWORD,:DWORD,:DWORD,:DWORD
ExitProcess PROTO :DWORD
.data
soffptr db 'Hello',0
.code
start:
invoke MessageBoxA,0,soffptr,0,0
invoke ExitProcess,0
END start
Disassembling the object module with Agner Fog's objconv :
.386
option dotname
.model flat
public _start
extern _ExitProcess@4: near
extern _MessageBoxA@16: near
_text SEGMENT DWORD PUBLIC 'CODE' ; section number 1
_start PROC NEAR
push 0 ; 0000 _ 6A, 00
push 0 ; 0002 _ 6A, 00
movzx eax, byte ptr [_soffptr] ; 0004 _ 0F B6. 05, 00000000(d)
push eax ; 000B _ 50
push 0 ; 000C _ 6A, 00
call _MessageBoxA@16 ; 000E _ E8, 00000000(rel)
push 0 ; 0013 _ 6A, 00
; Note: Function does not end with ret or jmp
call _ExitProcess@4 ; 0015 _ E8, 00000000(rel)
_start ENDP
_text ENDS
_data SEGMENT DWORD PUBLIC 'DATA' ; section number 2
_soffptr label byte
db 48H, 65H, 6CH, 6CH, 6FH, 00H ; 0000 _ Hello.
soffptr is defined as just a number
soffptr = $
It happens to be an offset into memory, but there is no such thing as "offset soffptr" in this case, it's just a number.
I can see the assembler internally saying "I know this is a relocatable reference to a memory location", indeed, how would it get the right number after the relocation takes place, but it is not semantically an offset.
It threw me when I saw it in the assembly listing and I briefly thought it was pushing the address of some place holding the actual number, but it wasn't.
In any case, it pushes the correct value, so this was just a curiosity question to see if there was some logic I wasn't seeing.
Hi Jimg!
I think you code is wrong, must be:Invoke MessageBoxA, 0,addr soffptr, 0,0
I can't test in this phone
Sorry I see pretty bad the problem. In this case $ means "offset of this position", not whatever number.
include \masm32\include\masm32rt.inc
.code
TheTitle db "the title", 0
start:
.listall
soffptr=$
invoke MessageBox, 0, soffptr, addr TheTitle, MB_OK
.nolist
exit
end start
UAsm64 listing:
00000000 746865207469746C65 TheTitle db "the title", 0
0000000A start:
.listall
0000000A = A soffptr=$
0000000A invoke MessageBoxA, 0, soffptr, addr TheTitle, MB_OK
0000000A 6A00 * push MB_OK
0000000C 6800000000 * push offset TheTitle
00000011 680A000000 * push offset soffptr
00000016 6A00 * push 0
00000018 E800000000 * call MessageBoxA
As the disassembly shows, the code assembles correctly: the ModuleEntryPoint is being pushed, and the MsgBox shows the character 'j', which is the push 0 (6A translates to a j).
00401000 . 74 68 65 20 74 69 74 6C ascii "the title",0 ; ASCII "the title"
<ModuleEntryPoint> Ú$ 6A 00 push 0 ; ÚType = MB_OK|MB_DEFBUTTON1|MB_APPLMODAL
0040100C ³. 68 00104000 push 00401000 ; ³Caption = "the title"
00401011 ³. 68 0A104000 push <ModuleEntryPoint> ; ³Text = "j"
00401016 ³. 6A 00 push 0 ; ³hOwner = NULL
00401018 ³. E8 07000000 call <jmp.&user32.MessageBoxA> ; ÀUSER32.MessageBoxA
0040101D ³. 6A 00 push 0 ; ÚExitCode = 0
0040101F À. E8 06000000 call <jmp.&kernel32.ExitProcess> ; ÀKERNEL32.ExitProcess
00401024 $ FF25 4C204000 jmp near [<&user32.MessageBoxA>]
0040102A $ FF25 54204000 jmp near [<&kernel32.ExitProcess>]
So the issue here is a slightly incorrect interpretation of invoke MessageBox, 0, soffptr, ... in the listing only. The assembly of this rather exotic piece of code behaves correctly.
MASM (all versions) doesn't like the source (error A2004: symbol type conflict). But there is nothing wrong in trying to push $, so that's another MASM bug.
P.S., just in case somebody wants to defend MASM against the ugly word "bug":
include \masm32\include\masm32rt.inc
.code
TheTitle db "it's a bug, folks!!", 0
start:
invoke MessageBox, 0, $, 0, MB_OK
exit
end start
Have fun with the disassembly :cool:
Pretty silly jj.
;dotst1=1
.data
ifdef dotst1
soffptr = $ ;;
testptr db "test",0
else
soffptr = $
tstptr2 db "test2",0
endif
.code
invoke MessageBox,0,soffptr,0,0