The PDF is read with lopdf (MIT, native builds only) on the worker that reads the file: the first
300 pages and 200,000 characters, with a cap on what a page's content may inflate to and a catch
for a reader panic. Text-less (scanned) or unreadable PDFs are pointed to by path.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
* claim: #9332
* Expressions: hyperbolic functions sinh, cosh, tanh and their inverses (#9332)
`sinh`, `cosh`, `tanh` and `asinh` take a unit-less number and give one,
as `exp` does; `acosh` refuses x < 1 and `atanh` refuses |x| >= 1, worded
as `asin` and `ln` refuse what lies outside their domain. They are in the
expression help (`FUNCTIONS`, the parameter autocomplete), the
hostile-expression fuzz tokens, the parameters and equation-curve guides,
and the checking skill's list of functions. Horn flares (Salmon's hypex
`cosh h + T sinh h`), catenaries and the tractrix can now be written as
they are stated instead of expanded into `exp`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6896
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013WUFq6QFBqmZjxwaKy2J1M
* Sketch: find profiles island by island and say when the search is cut short (#6896)
Every live drag frame ran the profile search over the whole sketch: every
spline piece of a large imported logo was sampled, checked for crossings
and traced again, about 100 ms per frame at 6,000 degrees of freedom.
Profile curves are now split into islands whose enlarged piece boxes do
not touch (the margin exceeds every tolerance the search uses), each
island's loops are found on their own and cached by a fingerprint of its
geometry, and the loops are nested into profiles over the whole sketch in
the order a single search finds them, so profiles, ids, holes and
whole-boundary certificates stay the same. The end-near-crossing lookup
covers its whole reach, independent of the grid's cell size.
The 50,000-piece limit now holds per island, and a sketch that still
exceeds it gets a warning instead of silently missing profiles.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013WUFq6QFBqmZjxwaKy2J1M
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
`schema::issues(id, params)` looks up the command's typed contract and
lists what is wrong with the parameters (none without a contract or for
an unknown id). The batch preflight, the MCP bridge's pre-send check and
the app's check of a design-switching command each did that lookup
themselves; they now call it. No behaviour change.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
sketch_consumed knew Extrude, Revolve and the sheet metal flanges but not
a Hole drilled at a sketch's points, so that sketch stayed drawn inside the
holes. A Multiple hole now counts as using its sketch; Edit Feature shows
its points while the dialog is open. Single holes keep their sketch shown,
since a hidden sketch offers no points to snap to.
* claim: #9299
* Recover unsaved design?: Discard all, with a confirmation (#9299)
A Discard all button at the left of the window's footer, away from Later.
It asks first ("Discard 7 autosaves? This cannot be undone.", Discard all /
Cancel), then discards every listed autosave the way its own Discard button
does (doc.recovery_discard on each id). Cancel goes back to the list
unchanged. The button publishes the handle recovery:discard_all.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9265
Co-Authored-By: Claude Code <noreply@anthropic.com>
* Preferences: a setting in a row with no label is found by the words it shows (#9265)
A search hid a row by its label before its controls were drawn. Seven checkbox rows and two segmented rows on the Interface page have no label, so "status bar", "search results" and "reveal on hover" found nothing. A hidden row with no label now keeps the words of its controls, as a dropdown keeps its value, and is drawn when they match. Each such row has a key of its own, so its neighbours are not drawn with it.
Co-Authored-By: Claude Code <noreply@anthropic.com>
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
* claim: #9195
* Viewport: an opened design is framed at the home view, in the middle of the view (#9195)
A fit the viewport cannot frame yet is kept and asked again on the next frame instead
of being consumed and lost, so a design opened from the Home screen or by launching on
a file no longer comes up at the default camera, wherever its own coordinates put it.
Framing then centres the model on the view rather than on the free rectangle between
the floating controls: `Layout` carries `middle`, the view without the panels, computed
before the compass, the navigation bar and the OK/Cancel corner join the obstacles. The
model is still sized to `drawable`, so it stays clear of every control, but only a panel
moves it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Viewport: keep the centred fit clear of the controls, and frame --sample (#9195)
The fit framed the model into a rectangle the size of `drawable` centred on `middle`, which is
not inside `drawable`, so the model could sit under the compass, the navigation bar or the
render product frame. The model is now sized to `drawable` as before, and only the room left
over in it moves the model towards the middle of the view.
`s27_move_triad` turns the Z ring on its far side: the model now sits further right, and the
Move dialog that docks on the right afterwards covered the ring's old grab point.
A launch with `--sample` builds the design straight into the session and never asked for a
fit; it now comes up at the home view, framed, like an opened design.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Viewport tests: the cursor orbit's pivot keeps the pixel it was picked on (#9195)
With the model now centred on the view, one of the test's perspective drag points lands within
the pick tolerance of the box's top edge, and the pick snaps the pivot onto that edge, a few
pixels from the cursor. The orbit keeps that point still, as it should; the test compared it
with the cursor instead of with where it started.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Both commands need a choice, but start_with_variant_memory only special-
cased the parameterless display toggles, so a click on the ribbon (or
command search, a shortcut, ui.start) ran them with {} and showed only
the engine's "style must be ..." / "tool must be ..." error.
They now open a short menu, the way Units does: under the ribbon row
that started it (a second click closes it), or in the middle of the view.
Visual style lists the four styles with the current one checked; the
Navigation tool lists Orbit, Pan and Zoom, and picking the latched one
releases it, as its navigation-bar button does. The four styles come
from one list shared with the navigation bar's menu.
A sketch that failed (a dimension set negative by a parameter) left every
feature built on it with "unknown sketch 1 (it must come earlier in the
timeline)", which points at the timeline order. The evaluation now notes
the features that failed earlier in the pass, and a feature whose input
sketch is one of them fails with "sketch `Sketch1` failed: <its error>".
That result is never cached, since the cache key holds neither the
sketch's name nor its error: a second failing value or a rename shows
up at once.
A page window (Preferences) filled its frame with
dialog_body.gamma_multiply(0.97), which also scales the alpha, so the
ribbon's labels showed through its title bar, search field and heading.
It now uses the opaque dialog_body, as every other window does; the
navigation bar keeps its own translucent resting fill.
* i18n(uk): add pending translations for missing strings
* i18n(uk): add pending translations for new strings from upstream
* i18n(uk): improve pending translations
* i18n(uk): professional translation of all pending strings
* i18n(uk): finish the pending translations for 29 October
Every value in uk-sync-29oct.json is now Ukrainian except the seven symbols
and formulas that must stay unchanged: the modifier keys, HDR and the three
inertia tensors.
Terminology follows uk.json: spot light is "прожектор", directional light is
"спрямоване", body is "тіло", render is "рендеринг". Camera orientation keeps
тангаж, рискання and крен.
Removes history-origin-bodies.json, which was empty after its content moved
into this file.
- worst_status searched the history and the results for each feature of a folded folder, the square of the history's length every frame
- A folder of 16 features or more now finds places and results in tables built once, first of an id winning as feature_index and Model::result do
- Frame on a 4,880-body design: 20.4 ms -> 7.9 ms (M4 Max, release, headless)
- Test a_folders_mark_is_the_worst_of_its_features_own
Co-authored-by: P'tite Tête <asso@ptitetete.com>
A new user guide page, Equation curves, under Sketching: the first curve step by step, Explicit and
Parametric explained with tables of rules to type (a line, a parabola, a wave, a circle, a spiral),
radians, parameters with the involute of a base circle, building a part (a closed flower curve
extruded), editing, and common mistakes. The technical details stay in a short section for scripts
and agents.
Seven pictures, rendered by `cargo xtask book` from new scripts in docs/book/scripts/equation-*.json
(gitignored output, attributed in attribution/pending/docs-equation-curves.md). The Sketching page
keeps a short summary and links to the new page.
* claim: #9184
* Control channel: a request's own param_check level, applied and confirmed by the app (#9184)
`engine.execute`, `engine.batch` and `engine.script` take an optional
`param_check: "off"|"warn"|"strict"` in their params. The app runs the
request at that level through the same `run_checked` path and batch
preflight a headless session uses, puts its own level back, and every
reply says the level it ran at as `param_check` beside `ok`. A command
that switches designs keeps the app's level; any other word is refused.
Readings and exports answered on a copy off the window thread keep the
level too.
`control.capabilities`, answered at once on the control-server thread,
lists what the channel does beyond the methods every version has:
`{"param_check": ["off", "warn", "strict"]}`. An older app answers
"unknown method".
The MCP bridge asks it once per connection. When the app lists the
level, every call and batch goes with it and the app checks against its
own commands; otherwise the bridge checks before sending, as #9073 does.
At strict, a reply that doesn't say the app checked the call is an error,
and the next calls are checked by the bridge. Also: `add_param_warnings`
takes the reply by `&mut`, and `with_param_check` sits beside
`Remote::new`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Control channel: a design-switching command is checked at the request's level
Review fixes for #9184. `component.edit_in_place`, `component.finish_edit`,
`view.next_tab` and `view.previous_tab` ran at the app's own level and
echoed `off`, so a misspelt call went through at strict. They now have
their parameters checked at the request's level before they run (and
still run at the app's own level, as each design keeps its own), and the
reply echoes the requested level. The MCP bridge at strict now requires
the echo to match its level, not just to be there. `param_check` is read
only by `engine.execute`, `engine.batch` and `engine.script`; other
methods ignore it. The off-thread test now shows the design change
computing off the window thread before it is refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch corners: a round that would use up a whole arc is refused
A quarter arc of radius 5 running into a line along it, rounded with
radius 5, put the round's tangent point at the arc's other end and left
an arc of no length. round_corner now skips a round whose tangent point
falls on either curve's other end; with none left it says that radius
does not fit, as for two lines (found by a local trim fuzzer).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch corners: a round where the curves meet smoothly is refused
A line running on from an arc's end along its tangent (off only by rounding)
has no corner, but the round's tangent points landed within 1e-7 mm of it and
the round came out as an arc of no length (found by a local trim fuzzer). The
corner test's line met its arc this way too and was checking that output; it
now meets the arc at a corner.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9179
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019FAWPuFoh9qxn2uS2j53QZ
* sketch.inspect: each profile loop names a curve once, not once per segment (#9179)
A profile records one curve id per piece its loop walks, and a conic or
spline is walked as many short pieces, so sketch.inspect listed a single
Bezier curve 50 times in a profile's `curves` and as often in a hole loop.
The reply now names each curve of a loop once, in the order the loop walks
them (solidcraft_sketch::distinct_curves, as delete and profile references
already do). The solved profiles keep their per-piece lists.
The lists are now a profile pick as they stand: a whole_boundaries pick
refuses a curve named twice, and the regression test extrudes the wall
from the reply's lists.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019FAWPuFoh9qxn2uS2j53QZ
---------
Co-authored-by: Claude <noreply@anthropic.com>
An SVG import leaves a sketch with regions and no body. The marquee (box select) took only the bodies, so the imported geometry was unreachable with the selection tools: a marquee over it selected nothing, and a click on a region did not take it either.
The marquee now takes the regions of the shown sketches when the design has no body. A design that already has a body keeps the old exclusive behaviour: the marquee takes the bodies alone, not the regions of the sketches that are shown. A dialog that asks for edges or faces (the Patch dialog) never takes the regions: they are neither.
* claim: #2829 (construct.* contracts)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Typed contracts for the construction axes, points and planes
The 19 construct.* commands get input contracts and FeatureResult as
their output, with shared types for references, planes, a plane or
face base, an axis line and a construction plane's name, each calling
the reader its handler uses (split out of geo, plane_param, the face
capture and named_axis_line).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9188
* Preference: keep the view when a sketch opens, Look At faces it on request (#9188)
Preferences > Sketch > "Turn to face a sketch when it opens" (on by default,
so nothing changes unless it is turned off). Off: creating or editing a
sketch leaves the view as it is, a sketch started on a face no longer
re-centres on the picked point, and Look At (the palette button, Ctrl+8)
still turns to face the sketch. The control channel sets it as
`ui.prefs {"sketch_open_look_at": false}`.
Changes a default: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9087
* fix(#9087): ui.home refuses a sample index that names no sample
The control channel passed the sample index straight to open_sample, which drops an index that names no sample and says nothing. The request then replied ok, so an agent could not tell a typo from a sample that had opened.
The index is checked before it reaches open_sample: an index that names no sample is refused with the valid range in the error. A sample that exists still opens, as before.
* claim: #9086 (edge lengths)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Measure: an exact circle's or arc's length agrees with its circumference (#9086)
Measure gave a circle's or an arc's length_mm as the length of the
polyline drawn for it, which runs inside the curve, while the same reply
gave its circumference or arc_length from the exact circle. When the edge
is known exactly (the kernel's circle, or a sketch's own circle or arc),
length_mm is now its radius times its angle. A circle fitted to a mesh
body's polyline keeps the polyline's length.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* claim: #9029
Co-authored-by: Cursor <cursoragent@cursor.com>
* Square sketch axes when looking at a turned plane (#9029)
Look At only yaw-aligned sketches facing world up/down, so a sketch on a rotated component stayed tilted on screen. Roll (or yaw at the poles) so the sketch's x runs right and y is upright.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
* claim: #9108
Co-authored-by: Cursor <cursoragent@cursor.com>
* Canvas calibrate: pick Point A and B with viewport clicks (#9108).
The Calibrate card only had typed coordinates, so users could not mark the image. Clicks on the canvas now fill A then B, with on-image markers and a short hint.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Calibrate canvas: ja/zh/ar/he/vi strings for the click hints
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Persistent naming: a proposal from a second implementation
A design proposal for the persistent-naming failures still open (#4341,
#5198, #5308, #5857, #5858, #5860, #5861), from building persistent naming
in Form Studio: fall back by lineage, never by geometry alone; a suppressed
maker is not a lost reference; names saved as structured one-level values;
ordinals from the feature, not the kernel's face order; split picks resolve
to all pieces; body identity apart from the display name. Documentation
only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Persistent naming proposal: body identity follows the #5308 plan
Change 6 now agrees with bluecmd's plan on #5308: an id minted from
next_id and stored per feature slot, better than the derived identity
the first version proposed. It adds that a stored id is only as stable
as the slot assignment, which today follows each lump's least face
name, and gives the lineage signature from the #5308 discussion.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #4515
Co-authored-by: Cursor <cursoragent@cursor.com>
* Inspect: keep analysis legends clear of a left History (#4515).
Draft, Accessibility and Curvature legends were painted at the viewport's left edge under a left History sidebar. Place them in the free workspace content instead.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
offscreen_device() asked wgpu for its default adapter and ignored the
SOLIDCRAFT_GPU policy that headless pictures follow, so with
SOLIDCRAFT_GPU=software the ui-egui offscreen tests still opened the
hardware GPU. Under hardware or software the adapter now comes from
solidcraft_gpu::select (new pick_among, which pick uses too); under cpu
there is none; unset (auto) keeps wgpu's default adapter as before.
Part of #9040.
Claude-Session: https://claude.ai/code/session_013WUFq6QFBqmZjxwaKy2J1M
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9147
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: the Sketch Grid box in the Sketch panel turns the sketch's grid off (#9147)
The box sets the open sketch's hide_grid, but the viewport drew the grid
whenever a sketch was open, so hide_grid only reached the ground grid and
the box did nothing. An open sketch now shows its grid whatever the Grid
switch says, unless its own Sketch Grid option is off; the window guides
that renders and the control channel use follow the same rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* claim: #7521
* fix: headless pictures return their draw counts with the picture
`the_renderer_counts_what_it_draws` drew with `Capturer::rgba` and then
read `Capturer::last_stats()`. The capturer is shared with the binary's
other tests, which draw on it from parallel threads, so another picture
could land between the two calls and the test read its counts (e.g. 10224
triangles for the empty design).
`Capturer::rgba_with_stats` returns the picture and its counts, read while
the call still holds the device. `last_stats()` is removed; this test was
its only caller (the performance overlay reads the viewport renderer's own
stats handle).
New test `pictures_drawn_at_once_each_count_their_own` draws empty and box
designs from eight threads on the shared capturer. Against the old pair it
failed 25 of 40 full runs of the `capture` binary; it passes with the new
call.
* fix: a headless picture waits for its read-back's answer
A picture maps its read-back buffer, waits for its submission with
`device.poll(PollType::Wait)`, and then checked once with `try_recv`
whether the map callback had run. wgpu calls a finished mapping's
callback from whichever thread next maintains the device, in a poll or
in any `queue.submit`, after it lets go of the device. When another
thread got there first, the wait returned with the mapping done but its
callback not yet called, and the picture failed with "not supported yet:
reading a picture back without waiting for the browser", the error meant
for a browser.
On the shared capturer, the other thread is another picture's device
check (`probe_lost` submits nothing before it takes the capturer's lock),
so the `capture` tests failed now and then: `bad_picture_sizes_are_errors`
(its last 8 x 8 picture), `pictures_show_a_plain_cut_without_the_hatch`,
`pictures_blend_a_line_that_is_not_opaque`. In the app, the window's
frames are submitted on the device its thumbnails and `ui.render`
pictures are drawn on.
A native picture now waits for the answer, up to the picture timeout; a
browser still answers later, on its own.
New test `pictures_read_back_while_another_thread_uses_the_device`: 2,000
small pictures while another thread polls their device. Without the fix
it failed 19 of 20 runs (1 to 7 pictures each); with it, 150 of 150 runs
pass. `bad_picture_sizes_are_errors` now prints the error it gets.
bherbruck