Configurable pointer-sizes

[albertz]

One other issue I've had that I'm not sure worth writing whole issue about, seems pointer-sizes are always handled as host-size with this, 64-bit python will read any pointers parsed from the data as 64-bit pointers, would it be possible to have some option to change the pointer size regardless of host size? (eg. for structs meant for PSX all pointers would be 32-bit)

Seems pointers get lowered to ctypes ctypes.POINTER(t)/ctypes.c_void_p which always use host size, maybe there could be some option to use c_uint32/c_uint64 if it's set?

Originally posted by @emoose in #19

[albertz]

Couldn't you just not use a pointer type then but write uint32_t in your struct instead?

Or you think it's nicer to have it as a flag?

Note, using such a flag would basically break the current C interpretation. At least how the C interpretation is handled right now. Because right now, all the pointers in C interpreted code are actual real pointers pointing to real memory. But I guess you would anyway not make use of that?

[emoose]

Well ideally I'd like to feed it whole headers/source straight from my games C project (or cpp-preprocessed first), PyCParser seems to handle that as input pretty well so far other than these pointers (and a separate issue with bitfields, but I think that's a problem with ctypes itself I can maybe workaround)
Having to replace each pointer during that could be difficult when they could have dozens of nested structs (and in these game decomps the structs change often too as new things get figured out, which is why I'd want to pass latest full headers to it)

For my code I'm only really using interpreter's getCType, doesn't try to dereference pointers or run the code at all, so for my case here I guess it wouldn't really need to be actual pointers.
Maybe by default it could keep using the current setup with ctypes.POINTER, but if some setting is non-zero and set to 32/64, switch to using uint32/uint64 (or could allow passing a ctypes class to use for them)?

E: after some more testing it looks like func pointers are also affected, and long which gets treated as ctypes.c_long is also host-size dependent.

Found that I can add monkeypatches for these though, to expose a --bitness param in my script, if you think a flag for it here isn't ideal I'm happy to just use those instead.

[albertz]

I added that as an option to the Interpreter. See test_pointer_size_override as an example. You can now set interpreter.pointer_size = 32 and then getCType should give it to you in the right format.