Modifying stack args

[jj2007]

Simple question: Could modifying arguments passed on stack as shown below cause any (OS-related) problems?

include \masm32\include\masm32rt.inc

MyTest   PROTO: DWORD, :DWORD

.code
AppName   db "Masm32 is great!", 0
Hello   db "A message:", 0

start:
   invoke MyTest, offset AppName, addr Hello
   exit

MyTest proc arg1:DWORD, arg2:DWORD
  mov arg1, chr$("Masm32 is wonderful!")
  MsgBox 0, arg1, arg2, MB_OK
  ret
MyTest endp

end start

[Vortex]

You can think that arg1 is acting as a local variable. Modifying the argument does not cause a problem.

[sinsi]

I've used args as local storage before with no problems. I can maybe see a problem if the caller does something odd but not the OS.
Do you have a problem jj or is it idle curiosity?

[jj2007]

I've used args as local storage before with no problems. I can maybe see a problem if the caller does something odd but not the OS.
Do you have a problem jj or is it idle curiosity?

Idle curiosity. Some HLL languages choke when you try that...
Thanks for reassuring me :bg

[dedndave]

that's because you can never tell if they are passing the argument or a pointer to it - lol
C tends to be over-bloated because they lean too much toward pointers, when the argument would do
good to see you on Jochen
Z sends a kiss

[sinsi]

Hmm...I was talking about using a dword arg as temp storage, but I think you dave are talking about using the target.
Reassure me here jj.

[dedndave]

what i was thinking was that a C program might push an arg on the stack, then use it again, assuming it to be unmodified
an example is, it passes a pointer to a variable for one call, then,
because it may use the same variable for the next call, assumes the pointer is still on the stack
for instance: y = 2x^2 + x + 2
it will pass a pointer to x to square it, then again to add it
same var - leave it on the stack and don't pop it off

in assembler, the arguments are rarely treated this way, although they could be if desired
i would think, as long as you are popping all the arguments off at the end of the routine, it is safe to modify them

[sinsi]

Well, that was one of my odd occurences...cdecl is different (to me) in that the caller 'owns' the args, since it has to clean up the stack after the call.

[jj2007]

Hmm...I was talking about using a dword arg as temp storage, but I think you dave are talking about using the target.
Reassure me here jj.

See my example on top of this thread. I modify what I find on the stack, and under normal circumstances that should indeed pose no problems - even if it is a pointer. However, on one occasion I also reuse these args in a different proc; that is definitely bad practice but the only alternative would be a global variable...

[dedndave]

well - as it is written, they are discarded when you are done with them
i see no problems
(don't go by what i say, though - lol)