circular, convex, concave) and performs efficient collision detection agains these build
volumes. As of now, collision detection is performed against a convex
hull of a concave build volume for efficency.
GCodeProcessor::Result renamed out of GCodeProcessor to GCodeProcessorResult,
so it could be forward declared.
Plater newly exports BuildVolume, not Bed3D. Bed3D is a rendering class,
while BuildVolume is a purely geometric class.
Reduced usage of global wxGetApp, the Bed3D is passed as a parameter
to View3D/Preview/GLCanvas.
Convex hull code was extracted from Geometry.cpp/hpp to Geometry/ConvexHulll.cpp,hpp.
New test inside_convex_polygon().
New efficent point inside polygon test: Decompose convex hull
to bottom / top parts and use the decomposition to detect point inside
a convex polygon in O(log n). decompose_convex_polygon_top_bottom(),
inside_convex_polygon().
New Circle constructing functions: circle_ransac() and circle_taubin_newton().
New polygon_is_convex() test with unit tests.
Fixed some old support and infill issues.
Fixes support problem #4295
Fixes Parts of interface layer extends beyond supports and cannot be printed
Fixes support missing under horizontal overhang #6058
Fixes Slicer double-traces small sections of Rectilinear Supports, causes
Fixes plastic buildup and nozzle crashes #4951
Fixes Add "Angle Interface layers" #2969
limited anchors, while before a full perimeter segment was always
taken if possible.
2) Adapted the line infills (grid, stars, triangles, cubic) to 1).
This also solves a long standing issue of these infills producing
anchors for each sweep direction independently, thus possibly
overlapping and overextruding, which was quite detrimental
in narrow areas.
3) Refactored cubic adaptive infill anchroing algorithm
for performance and clarity.
1) Octree is built directly from the triangle mesh by checking
overlap of a triangle with an octree cell. This shall produce
a tighter octree with less dense cells.
2) The same method is used for both the adaptive / support cubic infill,
where for the support cubic infill the non-overhang triangles are
ignored.
The AABB tree is no more used.
3) Optimized extraction of continuous infill lines in O(1) instead of O(n^2)
The offset curve extractor is already quite usable,
though singular cases are still not covered yet
when the offset curve intersects or nearly intersects
a Voronoi vertex.
Removal of the PRINTF_ZU "%zu" Visual Studio printf compatibility macro.
Fixes of a contours self intersection test for collinear segments.
SVG exporter now exports white background, so that the GNOME Eye viewer is usable.
BoundingBox constructor will no more throw for empty vector of points.
GMP allowed for Vojtech's fork of boost::polygon Voronoi implementation.
Added libslic3r tests for boost::polygon Voronoi. All Voronoi issues
ever reported on the Internet are captured by the tests. Two issues
reported (the two test cases) are real issues which may influence
PrusaSlicer negatively, namely
https://github.com/boostorg/polygon/issues/43
into the 3MF / AMF.
Improved accuracy of ModelVolume's mesh transform back from Object's
coordinate space to its own coordinate space after reloading
from 3MF / AMF.
1) Reworked logic for pasting volumes
2) Fixed paste of volumes into different objects
3) Do not apply offset when pasting into the copied object
4) Keep source transformation matrix and relative positions when copy/pasting volumes into another object
Ulocking the "anisotropic" scaling checkbox will bake the transformation
into meshes to allow for scaling in world axes.
Optimized and templated the stl_transform functions, now also available for 3x3 matrices.
The Canvas3D::reload_scene() now maintains selection even if all volumes of an instance changed their IDs.
is_left_handed() method on transformations and volumes
rendering of GLVolumes in left handed coordinate systems by glFrontFace(GL_CW);
SLA slicing on left hand oriented instances by flipping the mesh for SLAPrintObject in X.
rendering of the SLA cutting plane in left handed systems
resetting the SLA clipping planes on 3D preview invalidation
Sphere and Cylinder are scaled to the same volume as Box
Newly entered modifier meshes are rotated parallell to the world coordinates.
If the instance coordinate system is rotated and scaled, it is not possible
to create an unskewed modifier to world transformation. In that case
the best possible transformation is found to minimize least squares error
of the 8 corners of the new modifier mesh bounding box using
Levenberg-Marquardt algorithm.
FIXME:
1) The Levenberg-Marquardt non-linear least squares does not converge
nicely, it may require some tuning.
2) Above all, if 1) is called, then often the skew of the modifier mesh
is so high, that it is likely more useful to display the modifier
with zero rotation and inverse scaling, so that the modifier will be
of correct size, but not parallel to the world coordinates.
