Cheat Engine

[tuxlu]

hi!

So I was battling with shared instructions on my cheat and wanted to know if there was a better way.

So I have my instruction I want to change:
- i change it, of course the game crashes because shared instructions
- I do "find out whate adresses this instruction accesses". I have 530 results (that's a lot)
- I choose the correct adress in group one and 3 or 5 others as group 2, and do a "scan for commonalities"
- Of course they don't have obvious commonalities, so I do a structure compare on a register I know is used in the instruction. (ex: RDI in mov [rdi+04],rax )
- I don't really see much obvious commonalities between the 2 groups, but I find what appears to be static values in my "correct" RDI offsets, that are different than what can be found in my "invalid" group.
- so if I compare like 3 of theses RDI offsets values with these "magic static numbers" that should be enough (like RDI +4 == 42 && RDI + 12 == 1337 && ...)
- problem: when I relaunch the game I notice that these values were in fact not static. Well, some of them are, but not the majority.

So what do?
I'd like to compare the same structures across game relaunches.
Of course adresses will be different, but MaxLevel/StructSize will be the same.

As when you save Structure Compare scans results in .sptr , it also saves the results in .sptr.results.x , I was a bit sad that in the Structure Compare window, when you open a .sptr it only loads the addresses, and doesn't even keep the ones from the window's previous scan.
Maybe I'm missing something, but can't I just load the results of a previous scan from last game boot, and compare them with the live result of the new scan with the game running?

For now my solution has been:
- going to the correct RDI address in the memory viewer
- copying everything from RDI-64 to RDI+512 (or more),
- putting the result in a line of a text file
- rebooting the game, refinding the correct RDI and repeating steps
- create a python script comparing the 2 lines of text and finding common values, printing their offset and value.
- repeat this a 3rd time if necessary

and NO, i did not just take screenshots of the memory view to then try to align them in layers with Paint.NET, that would be so hacky and time wasteful

So is there an easier/smarter way?

Thanks again in advance for your help, I always got deep and insightful answers here ( but no pressure )

[ParkourPenguin]

[tuxlu]

thanks for the fast answer (and sorry for my slow one)

so, if I understand well, you advise me to either:

- inject where the data is read instead of written? Like, if it's my player health, find when the game over instruction is called when health is 0, to disable it, instead of changing my health directly?

- do an injection at an instruction modifying a value in my player instance, but that is an"easier" injection point I know is consistent and not accessed by other shared instructions. with this I find the player instance address and check "if (rdi == playerAddr) ..."
That's interesting, but I really don't know if I could find this kind of "better injection point"

- put my debugger on the instruction I patched , after my checks that rdi is the correct one, and then "step out" 2 times, to find where my call to "subtractHealth" is done, and replace it.
Problem in my case: the method substracting my health seems to do quite a lot more than that, and if I try to nop it 1 level above, I just can't move my character.


I litterally coded the game I try to hack, and I know the instruction I try to alter isn't *supposed* to be shared. but it's in Godot, so it's an interpreted language and i fear it does funky stuff.

I have only 1 result instruction in "find out what writes to this address" of my player's health, but by using "find out what adresses this instruction accesses", I have 530 results.

So it's not a simple "class code shared between the player and the enemy" it's "the engine does funky stuff to set all values of all classes in one method" problem (that's my guess at least)

for now my solution seem to work, because I guess my "rdi" points to the user class , with "[rdi+04]" being the player's health.
So I try to find unique and consistent values around this memory region, and it works well, but some are not persistent across reboots.
But in the end, my solution worked, as I found that even across game reboots,

[ParkourPenguin]

[tuxlu]

oh ok, thanks for your insightful reply.

So in my case what I did seemed to be the best solution,
so just as a feature request if it is revelant for other people: being able to load an old "Structure Compare scans result" and compare it with new results would help in my specific case.