Always use repr(C) for vertices.

[PieroV]

I think it'd be useful to always give the guarantee that vertices are repr(C), not only with vulkano.

[simnalamburt]

🤔 I wonder if it's breaking change or not

[PieroV]

FWIW, tests passes, and when using vulkano you already have this representation.

[PieroV]

TBH, I think also bytemuck without Volcano would be great 🙂.

[simnalamburt]

bytemuck without Volcano would be great

I think this should be possible.

always give the guarantee that vertices are repr(C)

Would you mind if I asked where you're using the repr(C) attribute? Does GLSL require repr(C) condition for some reason?

[PieroV]

always give the guarantee that vertices are repr(C)

Would you mind if I asked where you're using the repr(C) attribute? Does GLSL require repr(C) condition for some reason?

No, it doesn't.
However, the Rust representation on my machine for the various vertices is the same as repr(C) (e.g., TexturedVertex is 36 bytes, the same as 9 f32).
However, it seems to me that with repr(C) you're more sure you can pass Vec<V> (with V a vertex type) slices to glBufferData.
From repr(C) in the Rustonomicon:

This [repr(C)] is also necessary to soundly do more elaborate tricks with data layout such as reinterpreting values as a different type.

Hence the request.

Also, you are also sure you can send these types through FFI in this way.
For example, nalgebra also uses #[repr(C)], even though not strictly needed.

For my case, using bytemuck would also work, but for now I had to use reinterpret, instead.