Firstly I acknowledge that this is very borderline opinionated, but I'd like to emphasize this is more a thing of possibilities given some shortcomings of the default just writing down the offsets of things.
I do a lot of reverse engineering and then I write C/C++ code to mess with what I've learned. Suppose there is a class like this, representing a monster:
OFFSET SIZE VALUE 0 8 Pointer to methods 1280 4 X coordinate 1284 4 Y coordinate 128C 4 Z coordinate 1400 4 Health
where, for example,
*(float*)((uint8_t*)obj + 0x1280) would give the X coordinate. The rest of the class is unknown, and the only known method is
*(uint32_t*)((uint8_t*)*(uint32_t*)obj + 0xC), which just sets the health.
These kinds of classes can be updated very frequently, making it incredibly annoying to just have recorded the offsets of different things. I have seen it suggested to create a padded class with just
uint8_t[next off - start], but even that just feels really awkward.
Right now the best thing I can think of to do would be to create a bunch of different bytescans (excluding the offsets) for different methods of the object that access its members and automatically determine the offsets, but still.
My problem is that although I don't actually mind the updating itself, my code looks terrible. I effectively treat these objects as just integers floating around and have a bunch of scattered functions that operate on them like objects, but really they aren't.
Is there a "proper" way to define an object with "suggested" offsets, with or without llibraries, and regardless, what would be a decent way to have clean code working with undocumented and mostly unknown objects like these?