This commit also adds means to manually disable warnings
in `code` tags where it's a false positive with the new
`skip-lint` attribute.
Warnings are now enabled on CI to prevent future errors.
We don't use that info for anything, and it generates unnecessary diffs
every time we bump the minor version (and CI failures if we forget to
sync some files from opt-in modules (mono, text_server_fb).
* Typo fixes.
* Navigation agent's `velocity_computed` signal is not emitted by
`set_velocity`, but at the end of that frame if `set_velocity` is
called.
* Simplify link in navigation agent's `avoidance_enabled` description.
* Unify `navigation_layers` description across `NavigationAgent{2,3}D`.
* Unify `is_normalized()` description across `Vector2`, `Vector3`, and
`Vector4`
The main change is to caculate tangent directly from bezier curve, without going
through discretized polyline, avoiding pitfalls of discretization. A similar refacor
had been applied to Curve3D.
The test cases for Curve2D is updated, comparing floating point with is_equal_approxmiate()
instead of `==`, in order to avoid float precision problem.
The main change is to caculate tangent directly from bezier curve, without going
through discretized polyline, avoiding pitfalls of discretization.
Other changes are:
1. Add an bezier_derivative() method for Vector3, Vector2, and Math;
2. Add an tesselate_even_length() method to Curve3D, which tesselate bezier curve to even length segments adaptively;
3. Cache the tangent vectors in baked_tangent_vector_cache;
This makes it easier to spot syntax errors when editing the
class reference. The schema is referenced locally so validation
can still work offline.
Each class XML's schema conformance is also checked on GitHub Actions.
Notably:
* `Packed*Array.size()` and `Array.size()`.
* Shared methods of `Transform2D` and `Transform3D`.
* Shared methods of `Vector2`, `Vector3`, and `Vector2i`.
This reduces the Deja Vu when translating the class reference :)
The description was probably copied from Vector3.reflect(), and
unfortunately did not match the 2D behaviour (where n is apparently the
direction vector of the symmetry line, not the normal).
For the time being we don't support writing a description for those, preferring
having all details in the method's description.
Using self-closing tags saves half the lines, and prevents contributors from
thinking that they should write the argument or return documentation there.
