bastacyclop / rust_box2d Goto Github PK

Hello, found this issue when I were playing with shapes and getting it position.
For start I were using b2::CircleShape::new without passing Vec2 and radius, which I thought was a problem of crash, but once I used new_with() with Vec2 and radius, nothing changes. I have feeling that problem with getting position of CircleShape itself.

Error: process didn't exit successfully: 'target\debug\examples\samples.exe' (exit code: 0xc0000005, STATUS_ACCESS_VIOLATION)

Example:

let test_circle_shape = b2::CircleShape::new_with(b2::Vec2{x: -3., y: 0.}, 3.0);
test_circle_shape.position();

You might run it and it will crash immediately.

Seems to me like this is feasible and I'm happy to help

Hi, this code seems to have taken quite some effort, which is why I'm surprised that it's almost completely undocumented. By that I mean only source code is provided, even the official documentation is just source code with a search bar, no doc comments.

As a new user, this makes it extremely hard to use: things that could otherwise take a minute can take upwards of an hour (not to mention the possibility of something not even being supported).

I know this is supposed to be a thin wrapper around Box2D, however the difference is still large enough to cause confusion and necessitate documenting the edge cases.

Of course, for example, the step function or the idea of fixtures don't have to be documented. But how UserDataTypes are used and hundreds of other such things induced by the language difference totally require explanation.

Structs defined in c++, such as b2Vec2 seem to cause many problems when they are passed over the FFI by-value. I think the reason is different calling conventions for C vs C++.
It seems like functions returning structs as-value or taking structs as input argument as-value cause problems. Take Body::world_point() as example, it does not work and return undefined data or some times crash. (All other demos from the testbed work because they never use any such function).

It is implemented as follows:

pub fn world_point(&self, local: &Vec2) -> Vec2 {
        unsafe { ffi::Body_get_world_point(self.ptr(), local) }
    }

b2Vec2 Body_get_world_point(const b2Body* self, const b2Vec2* local) {
    return self->GetWorldPoint(*local);
}

Here the argument "local" will point to garbage no matter how you call the funcition and the program crashes. This is because the return value is a b2Vec2 as-value and b2Vec2 is compiled as struct/class by c++ code while the function Body_get_world_point is treated as C-code.

Hey, I'd like to use this to make playdate games, but that would require no_std support. It would be great to be able to use this on platforms and devices that don't have access to the standard library.

I see there is a dependency on nalgebra, I belive that weould need default features disabling, not sure about the other dependecies.

As someone with no C++ experience this has been rather confusing... what exactly is expected of the Box2D dependency regards its location and its contents? I've tried pointing to it locally w/ BOX2D_INCLUDE_PATH as well as putting what I think are the correct Box2D contents in /usr/local/include/ to no avail. Help appreciated :-)

macOS, Rust 1.21.0

Spotted in #4: https://travis-ci.org/Bastacyclop/rust_box2d/builds/196365785.
I'll wait a bit to see if things fix themselves.

It looks like the Travis CI is not running on new commits anymore .. this needs to be investigated.

Hi,

thanks for this awsome bindings.

Is there any documentation on how to user UserData?

Thanks