engine::view::build (Style, Show, Palette) is the viewport's builder,
moved from ui-egui; ui-egui maps theme::Tokens onto a Palette and calls
it. Palette::LIGHT now equals the light theme's tokens. Test: the app's
scene and engine::view::frame's are identical for the same session.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The app keeps a capturer on its own device; ui.render draws with the
viewport's pipelines there, on the CPU with the reason otherwise, and
the reply names the renderer.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The window's paint callback (slots, prepare, paint) draws the sample,
its symbols and an overlay into the capture's offscreen pass; the
pixels equal Capturer::rgba's.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Capturer::rgba becomes offscreen() (sizes, error scopes, MSAA, depth,
read-back) plus finish() (background and labels), so another painter,
such as the window's paint callback, can draw into the same pass.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
With a user's partition on a symmetric body, applying the whole body's plan cut it again by its
symmetry plane and mirrored one cell, leaving the other half unmeshed ("the mirror images do not
reach cell Base/b/a"). Such a body is its cells: the symmetry is not used and the plan is refused
with the cells named, as the replay of the saved Pi 5 model shows.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
After geometry.partition cuts a body, the body is its cells: a projection plan of the whole body
could not open it and failed with "no body `Base` (open a design first)". It now names the cells
and says to plan and apply a cell (mesh.plan {"body": "Base/a"}) or delete the partitions. Tests
cover a design opened directly and a saved model opened again, partitioned after opening.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The drawing tests skipped silently without an adapter. Under
FIELDCRAFT_REQUIRE_GPU=1 they fail instead, and the snapshot and MCP
screenshot tests require a GPU renderer. CI runs them with
FIELDCRAFT_GPU=software on lavapipe.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
mesh.plan.apply finds the planes a projection plan's body is its own mirror image across
(through the middle of its extent, square to the sweep's in-plane axes, through no ring: every
triangle middle of its tessellation, a little inside and outside, has its mirror image inside and
outside alike). It cuts the body by them, plans and meshes the cell on their negative side, and
mirrors that mesh across each plane in turn (mesh.mirror, joined node to node on the planes), so
mirror-image regions mesh exactly alike. `symmetry: false` meshes the whole body.
A cell of a partitioned body is planned from the features that built its body, with the rings it
holds; a partitioned body is planned whole as before.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The compass search after the sweeps moves the nodes of the worst elements one at a time, each to
its own best place. Since it lifts every element below 0.7 (#462) it moves many surface nodes,
and a column of them on a face ends at scattered places: the enclosure base's inner walls showed
zigzag element lines in a column next to each corner, full height. A last pass now moves each
surface node back toward the mean of its neighbours on the surface, as far as no element round
it falls below the lower of its own value and the mesh's smallest after the search, so the lines
straighten and the mesh's worst element is no worse.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
XY Plot opens a section in the results panel: a quantity picker, the
series from results.xy plotted against the step's own axis with the
frame or mode shown marked (a click shows the nearest), nodal fields
read at the probed node, and Export CSV through results.xy. The energy
plot and the XY plot share one line-plot function.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Projecting a point onto one B-rep face searched one box tree over every triangle of the
boundary, skipping the other faces' triangles: when other faces lie nearer the point than the
target face, it walked most of the tree before reaching a triangle it could use. Smoothing a
projected mesh projects every boundary node many times, onto every face of its target group,
so the Pi 5 enclosure base spent 95 % of its 11 minutes there. Each face (and each face pair's
curve) now has a tree of its own: the same nearest triangle, found among its own boxes only.
Ties between triangles at the same distance went to whichever the tree's walk met first; they
now go to the lowest triangle, so the result no longer depends on how a tree is built. With
that rule the per-face trees and the whole-boundary tree give bitwise the same mesh.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A nodal field at a node per frame (nonlinear static displacements from
load factor 0, explicit frames, heat temperatures) or per mode, the
explicit energies and histories, history nodes, and the frequencies or
buckling load factors; capped at 100000 points with `every`; stale
results refused; `path` writes a two-column CSV. XY Plot is ready.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The crank arm's back pocket ends on the planes x = 48 and x = 130, its frame sides. The plan
put its frame cuts there; the kernel cannot cut along a face (#427), so both were skipped and
the cell between the bosses swept the pocket's rounded ends into fans. A frame side on a face
of the body now moves one element out of the opening (away from the rectangle's middle),
leaving a strip one element wide; both cuts are made.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
#462 visited nodes by a mirror-invariant key (distance from the centroid). On the cup with a
handle that order sweeps ring by ring round the axis and converges worse in its 100 sweeps: the
refined cup fell to a scaled Jacobian of 0.481 (needs 0.5, #440). In the lattice's own order
it holds, and the order gained the enclosure base little (its corners differ by 0.02 either
way), so it goes; the compass search's lift to 0.7 stays.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The compass search after the sweeps only lifted elements below 0.5 (the quality floor), so a
mesh whose worst elements sat just above it stayed there: every corner of an enclosure base
stopped at a scaled Jacobian of 0.50-0.52 where its walls meet the corner bosses. It now lifts
every element below 0.7, never lowering another below the lower of its own value and 0.5.
The sweeps and the search are Gauss-Seidel, so the node order shaped the result and mirror
image corners smoothed differently. Nodes are now visited by a key that a mirror or a quarter
turn about the nodes' centroid keeps (distance from it, then the sorted offsets along the
axes). What spread is left between mirrored corners (0.686 vs 0.702 on the test base) is most likely the projection onto the body's
tessellation, which the kernel does not make mirror symmetric (not yet proven).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
geometry.check keeps its findings on the session (not an edit: not journalled,
saved or undone) while the checked design stays open; the picture and the app's
viewport draw the faces of small, sliver and near faces in the marker colour,
short edges as thick lines and a dot at each finding over everything. `mark:
false` or a clean check clears them.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A Sides field kind: element faces or edges typed as element:side with
ranges (1-20:3, 25:1), no row cap, and Use selection for picked faces.
Set and mesh-tag forms use it instead of a JSON box (#449).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Two field kinds: NameOr (a name as a bare string, or an object of fields)
and RefText (a text per item of a collection). The mirror's plane is a
datum plane or a point and normal; its images a body name per body
(#449).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A PerAxis field kind: one value for every direction, or Per axis ticked
for one per axis. Thermal conductivity and expansion and the point mass
use it instead of a JSON box (#449).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
egui-wgpu's startup hook and the app's own steps (window, app setup, viewport
pipelines, first frame, running) go into the journal. On Windows,
native_crash.rs (adapted from SolidCraft, the only module allowed unsafe code;
the app crate now denies instead of forbids it) writes the exception code,
faulting module and offset and a minidump. A nightly/on-demand CI job runs the
app's tests on Windows, including a child that faults on purpose.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The journal keeps the startup step the app is in (phase) and the renderer
(set_environment) across rebases; an unclean exit before the first frame is
reported "during startup (<step>, N s after launch)". native_paths() is where
a native exception filter writes what it caught and a minidump; the next
start puts the exception code, module+offset and thread into the report and
moves the minidump into its folder. Adapted from SolidCraft's crash.rs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
#445 also closed the window on SIGHUP, which overrides the SIG_IGN that
`nohup fieldcraft &` sets so the app outlives its terminal. A safe API cannot
ask whether a signal was ignored at launch, so SIGHUP keeps its default.
SIGTERM and SIGINT still close the window the normal way.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The crank ends at 8, 16, 20 and 22 mm, so no one-to-one sweep takes it.
The sweep met one of those caps as a linking face and reported "linking
face has no top side". A plane parallel to the source that only part of a
ring reaches is now named as what it is, with extrude pointed to. The arm's
profile alone sweeps as well as extrude does.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A boss the body's walls run into had its circle crossed by the walls' planes, so the projection
layout refused it and the plan fell back to swept cells, fanning round the boss and its hole.
Such a boss is now framed by the square round its whole circle, subdivided by the planes that
cross it; every square inside (the hole's, the core) is divided alike, so the ring's blocks meet
the walls node to node. On the walls' side the frame stops at their inner plane and its corner
projects onto the middle of the inner fillet, which the ring takes in (a boss just touching the
fillet included: no sliver). Cylinders a fillet's history names but that are a ring's circle are
the ring's faces. Rings are built per patch of grid cells, the one-cell O-grid being the
simplest patch.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
SIGTERM, SIGINT and SIGHUP ended the process on the spot: no last autosave,
and the crash journal left behind became an unclean_exit report at the next
start. The report on #248's machine ("ended without a clean exit between
commands (last: control ui.capture)") has that exact signature and is
reproduced by kill -TERM after a capture; 500 captures of the solved
crank-mirror (50 app runs), 30 CLI snapshots and 250 MCP headless
screenshots exited cleanly. The first signal now asks the window to close;
a second ends the process with the signal's default action.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The kernel's last-resort boolean retry moves the tool 1.3 µm and turns it
2 µrad. For a partition that puts the cut beside the face it was meant to
run along and leaves a cell bounded by slivers (the open boundary at x = 130
on the crank half). halves() now refuses a result-altering retry and names it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A smooth loop (straight sides running tangent into arcs, as a slot or a
rounded pocket) is meshed as a disk. Its middle block was a square halfway
to the loop's quarter points, which on the crank's 6:1 pocket floor
squashed elements to a scaled Jacobian of 0.16. The block is now a
rectangle on the region's principal axes, inset by the same distance all
round (the old square on a circle), its corners paired with the loop nodes
nearest their diagonals, its sides spaced as the nodes they pair with, and
the O-grid is smoothed (untangled first) before the validity check.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Static, frequency (per mode), nonlinear and explicit (per frame, with a
.pvd collection) and heat results; fluid steps are refused. Wedges keep
the model's node order: VTK 9's wedge has the same orientation.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
