Contact definitions on face sets: a pair (each surface's nodes against the
other's faces), general contact on a surface (self contact) or on every
exterior face of the solids; Coulomb friction with static and kinetic
coefficients and an exponential decay, a penalty scale and the choice to
adjust or refuse initially penetrating nodes, validated with caps. Rigid
walls of explicit steps gain friction. The check wants solid faces on
contact surfaces; sets used by a contact cannot be deleted.
fem::contact: the closest point of a triangle (Ericson) or a bilinear
quadrilateral (projected Newton) to a point, and a bucket grid of face
boxes for the global search, deterministic and bounded.
Commands interaction.contact.pair, .general, .edit, .delete and .find
(the touching parts of the mesh, with face sets and pairs made of them);
contacts in the model browser. Abaqus writes *SURFACE INTERACTION with
*FRICTION (exponential decay), *CONTACT PAIR and general *CONTACT; LS-DYNA
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE and _SINGLE_SURFACE with FS, FD, DC;
Nastran refuses contact by name. The implicit solver refuses contact by
name (M7); the explicit solver refuses it until the next commit.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
nl-hybrid-incompressible: 27 hybrid hex8 at ν = 0.4999 pulled to λ = 2 by a prescribed
displacement, against the compressible neo-Hookean closed form (1.4·10⁻⁹); its first
iterations need the LDLᵀ factorization. ROADMAP and parity updated, docs/oracle.md regenerated.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A tangent that is not positive definite is factored by LDLᵀ (Cholesky first, one symbolic
factorization) instead of cutting the increment back; each iteration records the tangent's
negative eigenvalues and the summary warns when a converged state keeps some. A model that can
move freely is still an error at its first tangent, located by the undeformed Cholesky factor;
a singular tangent later is a cutback. Nearly incompressible hybrid hex8, hex20 and tet10 at
ν = 0.4999 under prescribed displacement now converge to the closed form.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A supernodal LDLᵀ on the Cholesky factor's symbolic factorization (ordering, supernodes, fill,
cap and parallel subtree tasks shared through a Symbolic and a generic driver; the Cholesky
numerics are unchanged bit for bit). Pivots by the bounded Bunch–Kaufman (rook) search of
Ashcraft, Grimes & Lewis among each supernode's columns (Schenk & Gärtner's supernode
Bunch–Kaufman), 1 × 1 and 2 × 2, right-looking panels whose trailing update reuses the tiled
Cholesky kernels with L D formed on the fly. Reports the inertia; a column without a pivot above
round-off is a located Singular error; NaN is refused. Symmetric::new picks Cholesky, LDLᵀ, or
Cholesky with an LDLᵀ fallback; solve_refined adds iterative refinement. 1.1–1.2× the Cholesky
time on a 140 000-equation hex8 pattern.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The owner's final choice: the hex-meshed cube with its banded fringe,
drawn a little larger, with no streamlines at any size. The Icon Composer
document is back to the floor and cube groups.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The streamlines are now banded in the fringe colours along their length, as
flow speed would be, with a thin dark halo and arrowheads. Two sweep in from
the lower left, climb over the cube and run straight into the top face's
mesh lines, narrowing to them, so the flow becomes the structure; one wraps
around the cube. The small variant keeps the cube with one bold coloured
streamline and one arrow.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The mark is now a hex-meshed cube in isometric view whose faces carry a
banded colour fringe (blue to red, hottest at the front corner), with three
tapered streamlines wrapping around it. The small variant keeps the fringe
cube and one bolder streamline. Icon Composer gets four groups (floor, flow
behind, cube with mesh, flow in front), its maximum. The title-bar mark is
redrawn as the fringe cube.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
pages.yml builds apps/fieldcraft-web (trunk build --release, wasm-opt)
on pushes to main and develop and on manual runs, and deploys the site
to https://jyviko.github.io/fieldcraft/ only from main (releases).
Develop merges that touch code build without deploying, so a broken
web build shows up before a release. Trunk is a pinned release binary
checked against its SHA-256; actions are pinned by commit.
Part of #119
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The desktop's ModelApp in a web page through eframe's web runner
(WebGPU, WebGL2 fallback, `?webgl` to force it), built with Trunk.
The browser services sit behind the UI's host-service seams: File >
Open uses the browser's file picker, dropped files are read
asynchronously, and both go into the page's files; Save, Save As and
Write Deck hand the written file out as a download. Autosaves go to the
recovery folder as on the desktop and are mirrored to local storage, so
File > Recover offers what a closed or crashed page left. `?sample=NAME`
opens a sample. A loading screen with the app icon, and a message
instead of the app when neither WebGPU nor WebGL2 is available.
`cargo xtask wasm` (and so `cargo xtask ci`) now checks fieldcraft-web
for wasm32 with the library crates. docs/web.md: running it locally,
what works in the browser and what does not.
Part of #119
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The browser build has no threads, so a background job (opening a file,
meshing from a form, solving) ran inline in the click and the page froze
before it said what it was doing. Now the work waits until the window
has drawn the job's progress, then runs on the window thread in the
next frame. Native jobs still start on a worker thread at once.
Part of #119
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
file.open, geometry.open, geometry.import.mesh and reopening a model's
geometry read through the io crate's file edge instead of SolidCraft's
vfs, so in the browser they see the files the page holds (the same
store saves write to). Natively it is the same capped regular-file read.
fieldcraft-geometry drops its path-based readers (read_file,
open_design_file, read_3mf_file): it works on bytes, and reading files
is the io crate's job.
Part of #119
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The browser has no file system a page can call synchronously. On wasm32,
fieldcraft_io::fs now reads and writes files kept in memory (memory.rs,
after SolidCraft's vfs): the web host puts picked and dropped files in
and takes out what commands wrote. The recovery folder works over the
same calls (list, remove, is_file), with in-memory entry locks and a
host-provided clock on wasm, so autosave and File > Recover work in the
browser as on the desktop. Native behaviour is unchanged.
Part of #119
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
Every Cargo update changes the licences file, which Dependabot cannot regenerate, so all its pull
requests failed the licences check before testing anything. CI regenerates it for them and a new
workflow commits it. num-bigint 0.5 alone cannot build: the predicates' BigInt is num-rational's.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
Nothing checked Linux beyond x86_64. A weekly and on-demand workflow
builds fieldcraft and fieldcraft-cli on a native arm64 runner (release
profile, as shipped), runs both binaries and the CLI and engine tests
there, and cross-builds both for riscv64gc-unknown-linux-gnu with
Ubuntu's GNU toolchain, linked, then runs them under QEMU. The GUI opens
Wayland, X11 and GL at run time, so the cross build needs no target
system libraries beyond libc.
The RISC-V job is adapted from SolidCraft's CI.
Closes#122
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The release pipeline's scripts and metadata were exercised only by a
full release run. A workflow of a few seconds, no Rust build, now checks
them on every push to develop and pull request that touches
.github/workflows or packaging, and on demand: actionlint (pinned,
checksum-verified, with shellcheck on every run: script), shellcheck on
the packaging scripts and AppRun, a parse and PSScriptAnalyzer pass
(errors and warnings) over the Windows scripts, xmllint on the macOS
property lists, desktop-file-validate on the desktop entry, and the
entry's StartupWMClass against the app id the window sets.
Adapted from SolidCraft's packaging-lint.yml (itself from PhotoCraft's).
Closes#121
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The scripts carry `# shellcheck source=../env.sh`, but shellcheck reads
that path from the working directory, so from the repository root it
could not open env.sh: SC1091 on every script, and SC2153 in
macos/package.sh because APP_ID, set in env.sh, looked unassigned.
packaging/.shellcheckrc resolves source paths from the script's own
directory (SolidCraft's setting); every packaging script now passes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
Dependabot opens one grouped pull request a week per ecosystem against
develop: compatible Cargo updates together (a breaking release on its
own) and every action together, keeping the SHA pins and their version
comments. The SolidCraft pins, the truck and egui-wgpu patches taken
from that commit and craft-ui's rev move by hand; egui and eframe take
patch releases only, since their line follows SolidCraft's.
CI skips same-repo pull requests, so Dependabot's would merge untested:
the gate, the oracle and the advisories now also run on pull requests
opened by dependabot[bot]. The heavy job stays off pull requests.
Closes#120
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
Quality from the ribbon printed its whole JSON report into the status
bar, cut off after the first metric's histogram, so a user saw braces
and no answer. The report now carries a `summary` line ("30676 elements,
all valid; scaled Jacobian min 0.305, mean 0.963", with the invalid and
undetermined counts when there are any), and the status bar shows a
reply's summary instead of its JSON whenever it has one. The CLI and MCP
get the same line.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
OK on a mesher's form ran the command on the window thread: octree hex
on the bolted assembly froze the window for over a minute, with no
progress and no way to stop it. The tet and hex meshers' forms now run
their command as a background job on a copy of the session, as opening
a file and solving do: the window keeps drawing with the job's name and
Cancel, the form closes when the mesh is in and shows the error when it
is not. ui.dialog {run} stays synchronous for programmatic callers.
The new test walks a first-time user's path: Open picks a SolidCraft
design, Tet Mesh opens its form, OK with the defaults meshes the part,
and a mapped hex mesh of a holed plate says why it does not fit.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The Open dialog is a native picker the control channel cannot click, so
nothing could walk a person's open path end to end: ui.start file.open
opens the command's JSON form instead. ui.drop {paths} hands the paths
to the next frame with the window system's drops, which open them
through the same route a picked file takes (open_path: by type, in the
background, the unsaved-changes prompt, the recent list). At most 16
paths of at most 4096 bytes each.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The Tet Mesh form opened with {} and failed with "give either body (with
tolerance) or surface", so a first-time user could not mesh a part
without knowing its body name and a chordal tolerance. The tet mesher
now takes its bodies as the hex meshers do: every body of the open
design by default, one named body, or an imported surface, with the
tessellation tolerance defaulting to 1e-3 of each body's diagonal. With
several bodies an error names the body it is about. The reply keeps its
totals (kind, counts, first labels, volumes) and lists each body's
counts, size and quality report under `bodies`.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The bracket's tet mesh is not bit-identical across platforms (#115): its fillet peak is 213.31 MPa
on macOS and 212.42 on Linux CI, which turned the oracle job red after #110. Those two quantities
are no longer frozen; the page names them mesh-dependent instead of printing one platform's value.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
CI runs on merges to develop (and on fork pull requests); the release workflow builds the draft
from a push to main. AGENTS.md, CONTRIBUTING.md and docs/releasing.md follow.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The end-to-end demo met a raw-CDT sliver that the validity gate accepted (exactly positive) and
the solver refused. Now:
- One validity criterion, `fieldcraft_model::admissible_jacobian` (normalized Jacobian above
1e-12), used by the implicit solver (fem `positive_inverse`), the explicit tet4/hex8
kinematics and the mesh gate, which proves it on the whole element (tet4: the solvers' own
Jacobian; curved kinds: a Bernstein floor), so whatever passes the gate passes the solvers.
- CDT rounding repaired by flips on the rounded mesh (Diazzi et al. 2023 §5.1) instead of the
±2-ulp nudge.
- `fieldcraft-mesh::tet`: a mutable tetrahedral complex whose every change is a checked
replacement (positive and admissible tetrahedra, faces matched with opposite orientation) with
an exact undo log; edge removal (Klincsek), multi-face removal, Delaunay insertion with
boundary faces, edge collapse; a curvature- and thickness-graded size field; constrained
Delaunay refinement with encroachment splits; ODT smoothing; the Klingner–Shewchuk schedule;
boundary vertices moved onto the exact B-rep surfaces at the end; tet10 with mid-edge nodes on
the geometry.
- `mesh.tet.generate` gains size, order, quality, quality_target, curvature_angle, grading,
max_passes, max_points and reports the CDT's and the mesh's quality; `solve.probe` takes a set
and reports its largest von Mises stress.
- Oracle cases filleted-bracket-fillet-stress(-coarse): the fillet's peak von Mises stress on
tet10 meshes of 2k and 13.5k elements, 213.3 MPa (a four-level study stays within 1.8%); the
CDT cases keep the raw tetrahedralization.
- Corpus quality columns: all 20 valid parts mesh and pass the gate.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
ui.select now also drives the viewport's selection: the filter, box
mode, items by name and label as model.set.create takes them, a pick at
a window point or the visible items in a box, changed by a mode
(replace, add, remove, toggle). ui.selection reads it back with its set
members and the hover; ui.dialog takes use_selection (the form's Target
row). Selection stays in the window: no engine command. Documented in
docs/control-protocol.md, with a test that picks the cantilever's free
end by a simulated click and a box, makes sets of it and resolves them,
and hostile-input tests (bad points and boxes, labels, huge lists).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
A click in the viewport picks what is under the pointer through the
selection filter (the selection bar, top centre): geometry faces, edges,
vertices and bodies, or mesh nodes, element faces and elements. Shift
adds, Cmd/Ctrl adds or removes, a click on nothing clears, Escape
clears; with Box on, a left drag selects the visible nodes, element
faces or elements inside the box. What a click would pick is
highlighted under the pointer with its name beside it.
The selection is kept in the model's persistent terms: geometry by name
(GeometryRef, named through the mesh's geometry tags or the drawn
bodies' face names), so it survives a remesh, and the mesh by label
(labels the model loses drop out). It reads as a set's members.
Picking runs on the CPU against the triangles both renderers draw. The
selection and the hover are drawn as an overlay in the theme's new
pick_selected and pick_hover colours (light and dark), outlined; the GPU
draws it as a second small layer, so a hover change uploads only that.
Command forms that apply to part of the model (sets, boundary
conditions, loads, sections, couplings) get a Target row whose Use
selection fills the members, or makes a set of the selection and names
it. A right click offers Create set from selection.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The pick module tests rays against triangles with the watertight
ray-triangle intersection of Woop, Benthin and Wald (JCGT 2013): a ray
through an edge or vertex two triangles share hits one of them, the
nearest hit wins and ties go to the lower index, so picks are
deterministic. A Projector holds the camera's projection for
screen-space picks (Camera::to_screen and the CPU renderer use it too),
and a DepthMap of the drawn faces, rasterised by the renderer's own
triangle routine, tells which points are in front (box selection).
Display triangles now carry their element and face number (numbered as
ElementKind::faces numbers them; seen_face gives a shell's bottom side
from behind), line elements their element, and geometry surface
triangles their face tag, so a hit names what a set refers to.
DisplayMesh::pick (results probing) uses the new test.
render_layers draws a second scene over the first: its faces over the
scene faces they repeat (a small depth allowance), then its lines and
points. The viewport draws selection and hover highlights with it
without rebuilding the model's scene.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
A push to the release branch (or workflow_dispatch) builds the macOS universal dmg and CLI zip,
the Linux tar.gz and AppImage and the Windows zip, with the version from Cargo.toml. The last
job checks the set, writes SHA256SUMS.txt and notes saying how the macOS app was signed, and
creates or updates a draft release; it never publishes. Off the release branch it only uploads
the files as a run artifact (dry run). macOS signing and notarization use the APPLE_* secrets
when they exist and fall back to ad-hoc signing (#72).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
packaging/macos/package.sh builds FieldCraft.app (io.github.jyviko.fieldcraft, with icon.icns
and the layered Assets.car), signs it with a Developer ID and notarizes it when the
credentials are there and ad-hoc otherwise, and puts it on a drag-to-Applications dmg that
is mounted and checked; fieldcraft-cli goes in its own zip. packaging/linux/package.sh makes
an FHS tar.gz (binaries, .desktop entry, hicolor icons, licences) and an AppImage made with
appimagetool and the type 2 runtime at pinned, hash-checked releases.
packaging/windows/package.ps1 zips fieldcraft.exe and fieldcraft-cli.exe (static CRT,
checked PE headers and icon). Every package carries README.md, LICENSE-MIT, LICENSE-APACHE,
NOTICE and THIRD-PARTY-LICENSES.txt (#72).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The window's Wayland app_id / X11 class is io.github.jyviko.fieldcraft, the id the release
packages use for the .desktop entry and the macOS bundle (#72). Windows builds embed
assets/app-icon/icon.ico in fieldcraft.exe through embed-resource (MIT, a build dependency only).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
The corpus runs mesh.hex.octree's mesher on every part at 1/25 of its diagonal (the size of
the structured hex column) and reports hexahedra, the smallest and mean scaled Jacobian, the
mesh volume against SolidCraft's and the time; a part it cannot mesh gets the reason. The
column caps the elements at 500 000 so a part whose grid explodes is refused in seconds.
Loaded parts carry their bounding-box diagonal for the stages. docs/meshing/corpus.md
regenerated: 15 of 21 parts mesh, including filleted-bracket, rounded-box, sphere-on-plate
and the analytic sphere.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
A slow theory case: the corpus bracket meshed by mesh.hex.octree at 7, 5 and 4 mm, clamped
at the base and loaded by 2 kN. Checks equilibrium, that every element is certified valid,
the mesh volume against the CAD volume, the beam bending stress of the upright leg
(sigma = M c / I = 145.3 MPa, within 5%) and the convergence of the fillet von Mises stress
against the 4 mm mesh.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
Replaces the mesh with the octree all-hex mesh of a body, an imported surface or every body
of the open design, and tags boundary element faces with their B-rep faces so sets by
geometry resolve. Size, tolerance, refinement depth, element cap, attempts and optimiser
sweeps are validated and capped; the element cap defaults to the session limit
(`Limits::max_hex_elements`), which the hostile-params test keeps small. Body input and face
tagging are shared with mesh.hex.generate. The catalogue's Automatic Hex entry is ready.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm
