Math: redo Matrix4::normalMatrix() as an inverse transpose.
The article this API was originally based on assumes a scenario which
just *isn't* matching usual practices here, giving wrong results. Too
bad I didn't spend more time questioning the proof there and just
blindly assumed it's correct because everyone said so.
Won't be typing all that reasoning again in the commit message, see the
changelog entry and the comment in the test for details.
- Fixed @relativeref{Corrade,Utility::Tweakable} parsers to compile with
negative @ref Math::Deg / @ref Math::Rad literals such as
@cpp _(-15.0_degf) @ce
- Fixed an assertion when using @ref MeshTools::removeDuplicates() on an
interleaved @ref Trade::MeshData that included padding at the beginning or
end of each vertex
- Fixed a silly mistake where there was no @cpp extern template @ce for a
@ref Frustumd debug printer but two declarations for the float version
instead (see [mosra/magnum#545](https://github.com/mosra/magnum/issues/545))
- Fixed the @ref QuadraticBezier2Dd, @ref QuadraticBezier3Dd,
@ref CubicBezier2Dd and @ref CubicBezier3Dd typedefs to be actually
double-precision
- The @ref Math::Matrix4::normalMatrix() utility was introduced in version
2019.10 based on an article claiming that it's "[a faster calculation of the normal matrix that behaves correctly also in presence of reflections](https://github.com/graphitemaster/normals_revisited)".
Such claim was true only in case the normals were calculated on-the-fly
from vertices in a winding order wrongly flipped due to the reflection,
effectively undoing the original calculation error. As practical uses in
Magnum always involve normals precalculated on an untransformed mesh and
passed to the shader via a vertex attribute, this method gave an incorrect
result while the classic inverse transpose did the right thing. The
function is kept since there's a value in having a convenience API instead
of typing out (and having to read / teach / explain / decipher)
@cpp transformation.rotationScaling().inverted().transposed() @ce every
time, it's just corrected to return an inverse transpose instead of a
@ref Math::Matrix::cofactor() of the upper 3x3 matrix.
- Fixed an assertion when using @ref MeshTools::removeDuplicates() on an
interleaved @ref Trade::MeshData that included padding at the beginning or
end of each vertex
- Fixed @ref MeshTools::compile() to properly take into account index buffer
offsets
- Fixed @ref MeshTools::removeDuplicates() to not take into account random
Vladimír Vondruš
3 years ago