fix bindgen

[khanghugo]

Bindgen 0.71 and latest rust nightly version.

Bindgen does not right away so there are some changes to bindgen such as c_uint to c_int or *mut T to *const T. With that, bxt-rs code can remain unchanged. Bindgen also makes function type to be Option<unsafe extern ...> instead of the usual unsafe extern .... So, for some parts, I add .unwrap() wherever applicable. There are a lot of changes needed doing after generating bindings so I write all instructions along with the bindgen command inside each binding files.

Some files are deleted because the new bindgen conveniently generates bunch of related structs. There are some re-declared structs such as hull_s where its members are defined in some files and not defined in some files. Doesn't seem like a problem.

I am able to load a TAS and interact with the editor just fine. If anything to worry about, it is the tas_server_time_fix because there is a small change to the code that does pointer comparison. Though the changed code is from rust_analyzer suggestion so I guess it is all okay.

[YaLTeR]

Oh right, I forgot I did a bunch of manual cleanup on the generated bindings. Thanks for documenting the steps.

Bindgen also makes function type to be Option<unsafe extern ...> instead of the usual unsafe extern .... So, for some parts, I add .unwrap() wherever applicable.

I got rid of all Option<> manually in the generated bindings, I think that's a better move here rather than adding unwrap(), since we know that those pointers are always present. (Only leave Option<> when the pointer can actually be NULL.)

I also replaced all ::std::os::raw:: and such with use manually, it's easy enough, just do a search and replace for ::std::os::raw:: to nothing, and same for ::std::mem:: (add the new offset_of import). ::std::option:: can also be removed.

For types like vec3_t I replaced them with [f32; 3] manually to avoid these spammy type aliases. Also doable with a search and replace. Same with edict_s -> edict_t etc.

Remove Copy and Clone too since the types are not semantically copyable.

Also if you don't mind, move structs like usercmd_s back to their separate files.

With these cleanups it should be easy to see if anything changed in the bindings unexpectedly. Actually I don't think much should change in the actual types, it's the layout tests that changed.

I don't think we need to regen this often, so it's fine that more manual work is involved. Actually I doubt we'll have to regen this for a while if at all because the major problem with bindgen layout tests had been fixed now with offset_of! and I don't think there are any other big problems.

[khanghugo]

Take a look again and see. Doesn't seem to have any problem loading and interacting with a TAS

[YaLTeR]

Thanks, looks much better. Left some suggestions and comments

[YaLTeR]

Thanks, I hope I didn't break anything with those last changes