Review windows-numerics: fix π literal, add Neg trait for vectors, cross product, and expand test coverage

[Copilot]

What's this all about?

Bug fixes

• matrix4x4.rs: Replace hardcoded 3.141592654 / 180.0 with degree.to_radians() — idiomatic, self-documenting, and avoids literal drift
• lib.rs: Remove clippy::all from #![expect] — it was fulfilled by the old approximate PI literal triggering clippy::approx_constant; after switching to to_radians() no clippy::all lints fire, so the expectation was unfulfilled and broke the clippy CI job
• Cargo.toml (test crate): Add features = ["std"] to the windows-numerics dependency — length, normalize, and distance are #[cfg(feature = "std")] gated, but the workspace default is default-features = false, causing test compilation failures

New functionality

• Neg (-x) unary operator added to the three vector types (Vector2, Vector3, Vector4), both owned and borrowed forms, matching the existing operator impl pattern
• Vector3::cross(&self, rhs: &Self) -> Self — the standard 3D cross-product was absent

let v1 = Vector3 { X: 1.0, Y: 0.0, Z: 0.0 };
let v2 = Vector3 { X: 0.0, Y: 1.0, Z: 0.0 };
assert_eq!(v1.cross(&v2), Vector3 { X: 0.0, Y: 0.0, Z: 1.0 });
assert_eq!(-v1, Vector3 { X: -1.0, Y: 0.0, Z: 0.0 });

Test coverage
Previously only the four arithmetic operators were tested. Added coverage for:

• dot, length, length_squared, normalize, distance, distance_squared — all vector types
• cross including anti-commutativity (v2×v1 == -(v1×v2))
• Neg — all three vector types (Vector2, Vector3, Vector4)
• Matrix3x2 constructors: identity, translation, scale, scale_around
• Matrix3x2 * identity == matrix identity property
• Matrix4x4 constructors: translation, perspective_projection (including zero/negative depth edge cases)

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[kennykerr]

@robmikh would you mind reviewing? I asked Copilot to review this crate and it came up with some improvements. I'm just not 100% sure they're all applicable.

[kennykerr]

@copilot fix build breaks

[Copilot]

@copilot fix build breaks

Fixed in c6d7ad4. Two issues:

1. Clippy: #![expect(clippy::all)] became unfulfilled after replacing 3.141592654 / 180.0 with degree.to_radians() — the old literal was triggering clippy::approx_constant. Removed clippy::all from the #![expect] attribute since nothing in the crate needs it anymore.

2. Tests: length, normalize, and distance are #[cfg(feature = "std")] gated, but the test crate depended on windows-numerics with default-features = false (the workspace default). Added features = ["std"] to the test crate's dependency.

[robmikh]

I think neg makes sense for vectors, and having the cross product is handy for Vec3. But neg for the matrices doesn't make sense to me, I'm not sure when you'd ever use it.

[kennykerr]

I think neg makes sense for vectors, and having the cross product is handy for Vec3. But neg for the matrices doesn't make sense to me, I'm not sure when you'd ever use it.

@copilot please work on this.

[Copilot]

I think neg makes sense for vectors, and having the cross product is handy for Vec3. But neg for the matrices doesn't make sense to me, I'm not sure when you'd ever use it.

@copilot please work o...

Done in df72c75 — removed Neg (and the associated impl_neg helper) from both Matrix3x2 and Matrix4x4, along with the corresponding tests. Neg is kept for Vector2, Vector3, and Vector4.

[kennykerr]

Thanks Rob!