3 commits, squashed; the originals stay reachable from tag archive/merge-main-develop.
- a57ce13e7 jyviko: thread: a shaft thicker than a modelled thread is refused, never turned down
- 6b9f5e0e4 jyviko: thread: main's screw direction (#4414) out; the thread's hand is the one direction
- 5fc160732 jyviko: Main's new callers of the placing calls that now return their error
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5M41G8Nua7HvuFPZkmJ
2 commits, squashed; the originals stay reachable from tag archive/merge-main-develop.
- d63aa814f jyviko: threads: main's tapped-hole sizes and shaft turn-down on the reworked thread
- cfc393d68 jyviko: Main's new callers of the placing calls that now return their error
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5M41G8Nua7HvuFPZkmJ
* Viewport: palette-indexed scene and incremental appearance updates
- Move triangle, line and instance colours and opacity into a per-scene
palette (an R32Uint texture, WebGL2-safe); appearance edits update
palette entries instead of vertex bytes.
- Build CPU body triangles as one canonical vertex stream per block with
opaque/glass/faded index lists; an opacity class change rewrites only
the index list.
- Cache resolved body appearances by a per-body key, so an edit resolves
only the bodies whose look changed.
- Split the model scene key from the session revision: selection and
view-only commands no longer rebuild or upload the model scene.
- Replace the O(n^2) section-index lookup and per-placement instance scan
with hash lookups.
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Viewport: skip unchanged palette work when rebuilding
- Reuse each body's appearance key while the document is unchanged.
- Leave a palette slot alone when its value did not change, and allocate
every body's slots once before the build loop.
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Viewport: measure scene rebuild cost by edit kind
An ignored scale test times the scene build after selection, appearance,
theme and one-body geometry edits at 200, 1,000 and 5,000 bodies.
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Viewport: read faded grid alphas through the scene palette
The grid-fade tests now resolve line palette indices instead of reading the
line record's old colour byte.
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Viewport: measure rebuild cost on few large meshes
Adds the 20/100-torus case where a block rebuild used to re-emit a large
part of the model.
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Viewport: measure rebuild cost on an assembly
Unique detailed parts plus one component placed many times, timed for
selection, part and component colour, opacity, theme, hide and move.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: draw block classes as vertex runs, not index lists
A body's vertices are three per triangle in mesh order, so each draw class
is a few runs of the block's vertex buffer: all-opaque blocks need no index
buffer, and an opacity class change uploads nothing. A fragmented class
(more than 64 runs) falls back to one index buffer.
Each triangle is written as one record into a block reserved up front,
which takes the cold scene build of a 7.1M-triangle assembly from 97 ms
to 55 ms (main: 83 ms).
Also fix reindexing a block whose bodies are drawn elsewhere: it walked
the mesh's triangles from that body's start, past its own vertices.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: draw the highlight's own block, and light faded bodies without copies
The highlight scene keeps its feedback (a hovered face, a body copied on
the CPU) in a block of its own; since blocks became vertex runs, only the
model's blocks were drawn, so that feedback was uploaded and never shown.
Draw the highlight's blocks with its triangles.
Whole faded and see-through bodies (another component active, or an
appearance below full opacity) were lit by copying every triangle on the
CPU at each hover change: 85 ms for a design row of 812 bodies. They are
now drawn again from the model's own block vertices, in the highlight
palette with their slots set to the highlight colour, or as placements of
the model's shared triangles. The picture is the same as the copy's.
Adds a GPU frame benchmark for document-tree hover (ignored test).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: count scene uploads, and keep the sorted glass stream across colour edits
Renderer::uploads reports the bytes the model scene's last change uploaded
(geometry, placements, palette rows, planned colours, re-evaluated vertices).
Without order-independent transparency the global glass stream is kept while
the blocks' see-through triangles are the same, so a colour edit on a glass
body uploads palette rows only. Tests assert both on a GPU.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: keep palette slots and looks without rehashing every body
A rebuild found every body's palette slots by name and wrote them back,
allocating and hashing a new key per body, and copied every shown name into a
set to check the arena still belongs to the design. Slots are written back only
when a body's faces grew, and the looks cache is pruned only when it holds more
bodies than are shown. Adds the STEP-file benchmark (SOLIDCRAFT_SCALE_STEP) and
the colour command's own time to the assembly benchmark.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: round-pick construction lines through the scene palette
The axis and centre feedback for round picks (#4610) drew with a raw colour;
on the palette scene it is interned once like the other construction geometry.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Viewport: draw faded and see-through bodies from their plans with OIT
A body faded outside the active component, or see-through, left the GPU
tessellation path and was built triangle by triangle on the CPU, so working in
a component of a large assembly rebuilt every other body. Where the renderer
blends see-through surfaces order-independently (GpuTarget::oit, set by the
renderer), a body whose faces share one class is drawn from its plan: chunks
are split by class, opaque passes, AO and the section draw only opaque chunks,
and the OIT pass draws the rest. The sorted fallback keeps them on the CPU.
The model and its highlights decide through one rule (model_plan_class).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
The blue, green, red scale's lightness rose and then fell (L* 44 up to
72 and back to 50), so a flat face and a tight radius looked alike with
hue lost, and its upper part merged for a red-green loss. The stops are
now dark blue, a middle grey-gold and light yellow in one straight
blend, as measured in the issue: L* rises the whole way and every
quarter step stays clear for normal, protan, deutan and tritan vision.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
With a section shown, the marking pass decided where the cap shows (the stencil) and wrote the
depth everything after it tests against. #1768 gave it the faces' FACE_BIAS for the second job,
so edges draw over their own faces (#1434). For the first job its slope part is unfair: next to an
edge where a back face seen through the cut meets a front face, the front face won a band about a
pixel wide although the back face is nearer. The cap skipped that band, and the edge line showed
through the cut.
#6363 marked those samples again from the back faces, pulled toward the eye as far as FACE_BIAS
pushes. That closes the gaps beside the edges, but the pull also lets back faces win wherever they
lie within the push of a front face, along every body's far silhouettes, and the cap hatches that
band. On bherbruck's block, the band away from the cut went from 58 px orthographic and 534 px
perspective before #6363 to 618 and 780 with it. In the app, his filleted hub showed it in every
frame of a 36-step orbit in both projections.
This replaces that pass. While a section is shown the opaque model now draws in four steps (`Cut`):
1. mark: the faces by their true depth (no push): colour, the stencil and depth;
2. a triangle over the viewport sets the depth back to the far plane, colour and stencil kept;
3. push: the faces again with FACE_BIAS, depth only, so the depth is what unsectioned frames have
and edges draw over their faces as before;
4. cap: depth only, as #6728 made it; the mark pass has drawn the cap's colour and hatch, as fs_solid
draws a back face seen through the cut.
The window pass's depth reset and this one are one helper, `depth_reset_pipeline`, for either
depth direction, and its passes are built by #6728's `solid_writes`. Unsectioned frames are unchanged. The CPU rasterizer already chose the cap by
true depth.
Tests (crates/gpu/tests/depth.rs), with the device's float depth and the 24-bit fallback:
- a_cut_hides_the_edges_behind_it (new): bherbruck's block cut at y = -16, home view, orthographic
and perspective. No edge shows through the cap, and away from the cut the picture is a box of
just the kept part's: 0 px orthographic and 1 perspective (main: 618 and 780).
- bherbruck's pocket and hub tests from #6363 are kept and pass (pocket 0 px; hub 1 and 2 px over
its twelve views, as with #6363's pass).
- an_edge_on_a_face_draws_whole_at_any_angle and
an_edge_just_behind_a_face_stays_hidden_on_a_large_design also run with a section shown.
In the app (safe mode, WARP), with 625 patterned cylinders cut by a section, the frame time is
that of main (#6363): 33.5 against 34.3 ms at the home view, 34.1 against 35.2 ms with the model
filling the view.
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #4214
* Viewport: the grid fades with the distance instead of cutting off (#4214)
The ground grid stopped at a fixed distance from the view, so its edge showed
as a cut across the picture. It now fades out with the distance from the
middle of the view, and it is gone before either of the two lines that used to
cut it: its own edge, and the picture's far plane.
Each grid line is drawn in pieces, and each piece is as faint as the fade
makes it at the distance where it lies. The 'under' batch is drawn by the
overlay pipeline with alpha blending, so the colour's alpha reaches the
frame.
* The headless grid fades with the distance too (#4214)
The headless path (solidcraft-cli snapshot, the home-page thumbnail, the MCP
screenshot) drew the grid from the engine's palette, where a line had no
alpha: the grid still ended in a cut, and at a low camera angle the cut was
in the picture.
A SceneLine now carries an alpha, the CPU rasterizer blends with it and the
GPU capture passes it on, so the engine's grid can fade. The fade itself
moves to the engine, next to the grid's other geometry, and both paths use
it: the headless grid now reaches as far on the ground as the view is wide
(the ground stretches as it is seen at a slant) and dissolves into the
distance instead of ending in a line.
* Fix the headless/viewport grid fade divergence (#4214)
The engine stretched the grid's size with the slant of the plane it is seen at, the viewport stretched it only on the ground, so on a sketch's plane the picture's grid reached further than the window's (153.42 mm against 120.00 mm). Both now share one grid_ext, which the picture and the window draw the same grid from.
* The grid fades out before the picture's far plane cuts the ground (#4214)
The window's grid faded against a far plane of its own: the scene's radius
clipped by the projection's camera (260 mm away at 400 mm of distance), while
the picture itself cuts the ground at the far plane of its own projection
(223.7 mm at the same distance, because the projection frames the scene's
bounds and not its radius). The grid was still a third of its strength where
the picture cut the ground, so the cut showed as a straight line across the
view.
The fade now ends at the far plane the picture's projection uses, and the
grid is gone before that plane cuts the ground.
The test that watched the fade now measures the ink of each row of the
picture, and checks that the grid is already gone in the ten rows above the
row where the far plane cuts the ground: the point it compared before lay
beyond that plane, where both renders draw nothing at all, so the comparison
was always true.
The cap pass redraws every back face behind a marked pixel with depth test
Always to put the section plane's depth there, and it also wrote colour.
The last back face drawn painted the cut, so a body behind the cut (one
inside another's hollow, an assembly's inner parts) showed through it in
its own colour and outline, like an X-ray. The pass now writes depth only;
the marking passes have already drawn the nearest back face's colour (the
same split as #4622).
Test: a_body_behind_the_cut_does_not_colour_the_cap in
crates/gpu/tests/capture.rs (a red body in a green box's hollow, 4 mm
behind the cut: the picture is the same with and without it, hatched and
plain, orthographic and perspective). On main 472 pixels differ.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Viewport: lift selected and hovered edge highlights toward the eye
A selected edge's highlight (3 px core, 6 px halo) is a screen-space band that
was depth-tested at the edge's own depth. A face next to the edge seen at a
grazing angle is pushed back by only about a pixel of its slope, so it buried
part of the band, unevenly along the edge: the highlight showed as a thin
wedge instead of a line.
The band is now drawn like a thin cylinder around the edge: its depth is
pulled toward the eye by half its width plus a pixel, measured on screen so
it holds at every zoom. Plain model edges keep their depth (#1434 still
holds); the part of a highlighted edge the model really hides stays dimmed
through hidden_line.
Not yet compiled or tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Viewport: lift an edge highlight by how steeply its faces are seen
The first lift (half the band's width plus a pixel, the same on every edge)
was not tied to the faces beside the edge. Each highlight segment now carries
the normals of the two faces it borders (octahedral, packed with its width in
the line's width slot, read as u32 by vs_line): the band is pulled toward the
eye by its half width times how steeply those faces are seen, so a grazing
face no longer buries it, and an edge seen square to its faces is not lifted
at all.
Test: a_selected_edge_on_a_grazing_face_draws_whole (tilts 60 to 88 degrees,
small and large designs, orthographic and perspective; the band 5 mm behind
a face stays hidden). Capturer::rgba_highlighted draws a highlight scene as
the viewport does, for that test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Since #1768 the section's stencil marking draws every face with FACE_BIAS.
Where a back face seen through the cut meets a front face along an edge (a
pocket's wall and its side wall or floor, a fillet and the face it runs
into), the slope push decided which was nearer beside that edge; a front
face winning left those samples unmarked, the cap skipped them, and the
edge, drawn at its own depth, showed through the hatch. In perspective, and
while orbiting in orthographic too, edges popped in and out from frame to
frame.
A second marking pass draws the back faces only (front faces culled, depth
test <=, no depth write), pulled toward the eye as far as FACE_BIAS pushes
faces back, before the cap. The depth buffer keeps FACE_BIAS, so edges on
visible faces still draw whole (#1434).
Tests in crates/gpu/tests/depth.rs: a pocket 2 mm behind the cut, and a
filleted hub with a bore and a cross hole seen from twelve directions, both
depths, orthographic and perspective. On main the pocket shows 45 px of
edges through the cap in perspective and the hub 25 px over the views; with
this change 0 and at most 2.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Performance overlay: frame-time graph and histogram, triangles, draw calls (#2548)
Net change of the perf-overlay branch, rebased onto main as one commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Document frame_rate_details in ui.prefs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* claim: #4563
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Section Analysis: a Hatch option, off for a plain cut in the body's colour (#4563)
The Section Analysis dialog gets a Hatch check box beside Flip, and
inspect.section a hatch?: bool (default true), so scripts, the CLI and the
control channel reach it too. Hatched, a cut is its body's colour tinted
toward the cap colour with 45 degree stripes, as before; off, it is the
body's own solid colour. The viewport (cap alpha in the uniform), the CPU
rasterizer and ui.render follow it, and the section keeps it when hidden
and shown again or edited.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
NVIDIA's EGL answers wgpu's request for a GL 3.3 context with exactly 3.3
(GLSL 3.30) yet lists GL_ARB_compute_shader. wgpu-hal then reports compute
shaders, wgpu-core builds its indirect-call validation pipeline, naga cannot
emit compute GLSL below 4.30, and request_device fails as Device(Lost).
The test draws nothing indirect, so it opts out of that validation.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Snapshot: draw a see-through appearance see-through, and give the CLI --body, --tool and --section
A headless picture dropped a body's opacity: the scene carried only a colour, so a design with
a translucent appearance was drawn opaque by `solidcraft-cli snapshot`, by MCP screenshots and
in a design's thumbnail, although the window has always drawn it as glass. There was then no
way to look at the body a cut would remove inside the material it meets: only a volume to
measure and reason about, which hides exactly the errors a picture makes obvious (a tool that
stops short, or lands in the wrong place).
- `SceneMesh` carries the body's opacity; `view::build_scene` takes it from the design
(`model_body_opacity`), and a selected body stays solid so the selection colour reads.
- The GPU capture routes a see-through body into the scene's glass at its own alpha, as the
viewport does; the CPU rasterizer draws the solid bodies first (they own the depth buffer),
then the see-through ones blended over them, far sides before near sides and without writing
depth, so what lies inside shows and a line between the two is not hidden.
- `snapshot --body NAME:opacity=0.3,color=#rrggbb`, `--tool NAME` (that body solid in the tool
colour, the rest see-through) and `--section z=27` run the commands the app itself runs
(`appearance.assign`, `inspect.section`): the options are shortcuts, not a second renderer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Tests: measure the tool colour as it is drawn through a see-through target
The tool-colour filter was derived on paper and demanded a red more than twice either other
channel, which the target in front of it tints away (measured: (200, 108, 85) through opacity
0.3); and the colour check painted the tool, which a solid target hides.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* fmt
* Snapshot options: strip the colour's # rather than slicing it; README
Slicing `value[1..]` on an input-derived string is what the house rules ask us not to write,
even where a length check has already made it safe. The error a failed assignment reports now
names the body rather than claiming it came from --body, since --tool reaches it too.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* --tool: assign the bodies around the tool in groups, not one command at a time
Each appearance.assign is an undo step with a document snapshot behind it, so one per body
made --tool cost grow with the size of the design for no reason: the bodies around the tool
all get the same opacity, and the ones that share a colour can be named in one call. A design
whose bodies all wear the default grey now costs two commands whatever its size.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Rasterizer: say where its glass and the GPU's differ
Splitting each see-through body by facing orders its own two layers; it does not sort across
bodies, which the GPU does. Worth stating rather than leaving a reader to find it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* claim: #2823
* fix(view): lines thinner than a pixel no longer vanish without multisampling (#2823)
A line quad narrower than a pixel covers no pixel centre when it falls between
two rows, so the grid's 0.7 px lines came and went as the view zoomed where the
viewport draws without 4x MSAA (some browsers and devices). Such lines are now
drawn a pixel wide, faded by the width they lack.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
jyviko