mirror of
https://github.com/bherbruck/solidcraft.git
synced 2026-11-22 16:41:06 +00:00
b19001830f
* Drawings: one place says how a view's bodies share geometry and which edges it draws mesh_lines::prepare grouped the placed bodies (which placements share one copy of a definition's geometry, and which are their own group, such as a mirror image) and chose the edges a view draws inline. Both now live in their own functions in meshes.rs, `placed` and `drawn_edges`, so the exact hidden lines can use the same rules. Nothing a view draws changes. A new test checks the grouping: a copy moved by a rigid transform shares its original's geometry, and a mirror image is a group of its own. Part of #5866. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drawings: a view's lines come from the exact model, and from the mesh only for bodies the exact lines can't draw A drawing view now draws its visible and hidden lines with hlr, from each body's exact faces and edges: straight edges and circles come out exact, other curves as chords within the asked error. A body falls back to the old mesh projection when the exact lines can't draw it (mesh bodies, split round faces, faces or trims hlr can't read, a mesh body that may stand in front of it), and the view's status names how many bodies fell back and why. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drawing views: a test that a cylinder's side lines meet its rims, known to fail in hlr (#2455) A cylinder seen from each upper corner at 1:2: each side line must end where a piece of its bottom rim ends, visible and hidden, to 1e-6 mm. Two corners miss by 0.75 mm on paper: hlr gives the rim no split where the side line touches it, and the visibility change falls in the first tenth of the rim's one stretch, which classify_range doesn't search. The other sides are 7.7e-6 mm off, at the grazing ray's own-face margin. The test is listed in tests/KNOWN_FAILING_TESTS until hlr splits the rim there; the fix drops the line. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drawing views: reference parts are drawn from their exact curves (#2455) A reference part's lines now come from a second hidden_lines call, not from the mesh. That call holds every reference part first, then each model body that may hide one. So reference lines are tested against everything, the model still sees through reference parts unless the view says they hide it, and a reference line on a model line keeps all of it. Reference lines still stop at the view's boundary: a piece inside keeps its exact curve, a line is cut exactly, and an arc is cut on its chords and put back on its circle, as the mesh projection does. Views with reference parts get faster (83 to 63 ms with one, 103 to 79 ms with three, on a plate with four holes and a boss) and certify any max_error_mm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drawing views: a split round face draws only its split from the mesh (#2455) A body with a face split on one cylinder, cone, sphere or torus stays drawn from its exact curves. Only the split edge, which hidden lines don't draw, comes from the mesh projection, picked by body and edge. The provenance and the max_error_mm refusal name those edges. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drawing views: prepare each body for exact lines once per sheet (#13218) A sheet's views each prepared every body for hidden-line removal, though preparing a body does not depend on the view. The sheet now hands its views one PreparedBodies set, keyed by each body's place in the model state, so a body is prepared once and later views reuse it. A view of another model state prepares its own. A cancelled prepare keeps nothing. Each view still checks the edges it draws against the body it uses. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
28 lines
4.5 KiB
Plaintext
28 lines
4.5 KiB
Plaintext
# The cargo tests that fail today on the kernel, solidcore, one a line: `<test> <why it fails
|
|
# today>`, with the issue that tracks it when there is one. Sorted. `<test>` is nextest's
|
|
# `<binary id>::<test name>`, as the junit report names it. CI's test jobs (`cargo xtask known-failing`)
|
|
# fail on a test that fails and is not listed here, and on a listed test that passes: a fix removes its
|
|
# line in the same pull request. The tests stay un-ignored, so every run shows whether they still fail.
|
|
# Refresh: download the junit artifacts of develop's latest CI run, one directory each
|
|
# (`gh run download <run> -R tangency-labs/solidcraft -n <artifact> -D <dir>/<artifact>` for junit
|
|
# and junit-gpu-1..4), run `cargo xtask known-failing --junit <dir>`, then drop each FIXED line and
|
|
# add each REGRESSION line with its cause, from the test's failure message.
|
|
solidcraft-cli::action_matrix::action_matrix_sample 2 baseline actions fail: 2 shells (Bad Loop Orient on issue-616-lid-ribs; status 1001 on filleted-shelled-box)
|
|
solidcraft-cli::naming_corpus::corpus_designs_name_alike_every_build_and_fail_where_a_reference_is_gone oracle/83-boundary-loop-thicken does not load: step 8, the offset of face 0 by -1 is not proved regular, so thicken is refused with no size named
|
|
solidcraft-engine::cmd::drawing_dimension::tests::a_new_hole_never_takes_the_dimension_of_an_undone_one a hole made after an undo takes the undone hole's feature id, and with it the undone hole's dimension (gap 2 on #4892, #5308)
|
|
solidcraft-engine::cmd::drawing_dimension::tests::renaming_the_plate_keeps_its_five_dimensions a dimension's reference names its body, so renaming the body loses every reference to it (gap 1 on #4892, #5308)
|
|
solidcraft-engine::cmd::drawing_view::tests::a_cylinders_sides_meet_its_rims_from_every_corner hlr finds no split where a cylinder's side line touches its rim and doesn't search a stretch's first and last tenth: in two corner views the rim's visibility changes at its seam, 0.75 mm on paper from the side line; the others are 7.7e-6 off, at the grazing ray's own-face margin (#2455)
|
|
solidcraft-engine::cmd::edit::tests::inserting_variable_fillet_control_preserves_parameters_at_existing_positions solidcore refuses a variable fillet whose law has a corner (solidcore#430)
|
|
solidcraft-engine::extent_tests::cut_to_the_far_side_of_a_curved_wall main's mesh-based version (#8634) not taken; the exact replacement is to come
|
|
solidcraft-engine::hole_face_tests::a_hole_wider_than_the_part_says_it_removes_all_of_it a 40 mm hole at (10, 10) touches the plate's edge y = 30 at one point; the kernel parts the plate there into 2 bodies, the test wants truck's 1, as on SMLib
|
|
solidcraft-engine::tests::chord_and_variable_fillets its 2 → 4 → 2 waist: solidcore refuses a variable fillet whose law has a corner (solidcore#430)
|
|
solidcraft-engine::tests::fillet_fit_taper_closes_to_a_point fillet status 1001 on the taper fit (radius 16 closing to 0), as on C++
|
|
solidcraft-engine::tests::mirror_placed::dragging_a_mirrored_part_meshes_its_copies_once a copy of a copy drawn from its first copy's triangles (#10101) is triangulated differently from a fresh mesh of it, where truck's meshes are the same under a rigid move, as on SMLib
|
|
solidcraft-kernel::blend::corner_tests::a_round_ending_in_an_earlier_round_matches_both_together fillet status 1001 on the round that ends in the earlier one, as on C++
|
|
solidcraft-kernel::blend_corner::tests::issue2422_vertical_after_separate_bottom_rounds the vertical edge after the separate bottom rounds is refused as smooth ("its faces meet tangentially, so there is nothing to round"): the rounds leave it tangent at solidcore 8ddce9016
|
|
solidcraft-kernel::blend_variable::tests::variable_radius_with_a_corner_through_a_middle_point solidcore refuses a variable fillet whose law has a corner (solidcore#430)
|
|
solidcraft-kernel::corner_g2::tests::a_long_setback_keeps_the_g2_patch_inside_the_box at set backs 11 to 15 the G2 corner patch bulges past the box, as on C++
|
|
solidcraft-kernel::corner_g2::tests::curvature_rounds_meet_at_a_g2_corner the corner between curvature rounds jumps 2.07 in curvature over 538 stations on solidcore; the test wants it curvature continuous, as SMLib f916a8d's setback corner (smlib#104), which solidcore ports in solidcore#2310 builds it
|
|
solidcraft-kernel::corner_g2::tests::the_box_corner_g2_patch_is_a_closed_solid_curvature_continuous_to_its_neighbours the fitted G2 corner patch bulges 1.7e-5 past the box, as on C++
|
|
solidcraft-smlib::ops::tests::a_variable_round_with_a_corner_in_its_law_is_its_balls_union solidcore refuses a variable fillet whose law has a corner (solidcore#430)
|