* claim: #6720
* Face pick by point: linear work budget, so bodies with many faces take a sketch (#6720)
A face picked by its point alone charged the whole mesh once per face, and
face_contains scanned every triangle for each candidate. A block with a
modelled thread (about 5 000 faces) went past the 2 000 000 limit. The
triangles are grouped by face once, each face's test reads only its own,
and the budget charges triangles + faces.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #3812
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SBN24y2TcqRcm2zn8Gkj27
* Sketch Dimension: value box in the design's units (#3812)
A dimension added without a value restated what it measured as
"<value> mm", whatever the design's units. In an inch design the
Dimension tool's value box therefore opened at "25.4 mm" next to a
dimension whose text read 1.000 (inches).
Document::design_length writes a measured length in the design's units
("1 in"), with enough decimals to keep it to the micrometre, so a
dimension that restates the geometry still moves nothing. The Dimension
tool, AutoConstrain's dimensions and Toggle Driven (making a driven
dimension driving) use it. Millimetre designs are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SBN24y2TcqRcm2zn8Gkj27
* Move the design-units dimension tests into sketch_ux_tests (review)
The #3812 tests go in the existing sketch dimension test file rather than
a new file of their own (AGENTS.md, #3226).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SBN24y2TcqRcm2zn8Gkj27
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The evaluation cache's 64 MB budget charged every cached feature result the whole model, though a
state shares every body it did not change with the state before it. A design whose model is a few
MB kept only its last few dozen results, so setting a parameter back recomputed what had dropped
out. Each result is now charged for the bodies and sketches it does not share with the state it
started from.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6550
* Crash reports: keep journaling through file.new and doc.open (#6550)
`Session::replace_design` swapped in the fresh session without its
`crash_journal` flag, and the app turns it back on only at its next frame.
So `file.new`/`doc.open` never recorded how they ended, the commands sent
after them in the same frame (a control-channel or MCP batch) were not
journaled, and a crash report then named `file.new` as the command in
flight. `replace_design` now keeps the flag, like the other session-wide
fields it keeps; only the design on screen has it set.
Test: `commands_right_after_a_new_design_are_journaled` (crash_tests.rs).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6917
* Tests: the six timing assertions do not depend on the wall clock (#6917)
Six tests asserted on the wall clock. Their result depended on how loaded
the machine was and not only on the code under test: they passed on an idle
machine and failed when the test binaries ran beside each other, which is how
the suite runs (`cargo xtask test` starts every test binary at once).
CONTRIBUTING.md:196 says it: "Tests run beside each other on a busy machine,
so a test must not depend on wall-clock timing."
The change is confined to the six test files the issue names:
- geom: the tangent chain of a helix is a ratio (750 pieces against 3 000)
instead of a limit of 200 ms.
- io: the placement budget is spent inside the reader, so the clock is gone.
- ui-egui: the classification of many faces is a ratio (900 against 3 600)
instead of a limit of 10 s.
- engine: the clearance measurement drops its 30 s limit; the boolean that
gives up reports itself.
- engine: the fillet beside a modelled thread is a ratio (a 1 mm thread
against an 8 mm one) instead of a limit of 30 s.
- ui-egui: the window frame budget goes from 250 ms to 20 s, the limit #6072
gave its own probe.
The helper is the one #649 wrote, copied into the crates that need it, with
the noise of a run as a parameter.
* claim: #6558
* Shortcuts: Ctrl+Shift+S opens Save As in the workspace (#6558)
The Home page already opened Save As with Ctrl+Shift+S, but the design workspace matched Ctrl+S, so the chord saved over the current file instead of offering another one. The workspace now routes the chord to Save As, and the File menu shows that shortcut next to Save As, taken from the command registry.
* claim: #6600
Co-authored-by: Cursor <cursoragent@cursor.com>
* Save an imported mesh that only fits after compression (#6600)
The design file limit was applied to the mesh before Deflate, so a mesh
that had already loaded could not be saved or autosaved.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix: validate and version large mesh saves (#6600)
---------
Co-authored-by: Colin Faulkingham <colinfaulkingham@Colins-Mac-mini.local>
Co-authored-by: Cursor <cursoragent@cursor.com>
* claim: #5679
* Top-level cut and join reach bodies in sub-components (#5679)
An extrude, revolve or press-pull from a sketch on a face of a body in a
component inside the active one now cuts or joins that body, moved or
turned, in its own component, so every placement shows the change. The
automatic operation probes that body where it is shown. A linked part, or
a face on a further placement of a part placed more than once, is refused
with the reason instead of a silent no-op (#5678).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6423
* test: a copied body must keep the look of the body it came from
solid.move with copy: true leaves the copy without the appearance, the
material or the face looks of the source body, so it renders with the
default look. The tests fail for that reason: the copy has no entry in
appearances.bodies, in materials or in appearances.faces.
* fix: a copied body keeps the look of the body it came from (#6423)
A move with `copy: true` pushed a new body but left it with the default look, so the copy came out a different colour than the part it came from. The copy now carries the body's appearance and its material, as the component copy already did.
* test: keep the copy regression to the body look and the material (#6423)
* test: each copied body keeps the look and the material of its own source (#6423)
Two bodies with different appearances and different materials, copied in one solid.move: each copy must keep the look and the material of the body it came from.
Scope of the fix: the look of the BODY and its material are carried over to the copy. The per-face looks are NOT carried over: a face look is keyed by the point on the face, so the copy's faces would have to be matched and their points translated, which is a separate feature.
A drag that isn't live walks to the cursor in steps, each a held solve
on its own budget seeded on the last solved state. Past a failed step
the walk keeps trying every step (a stretch the geometry can't follow
can be followed by one it can), and each failing solve runs the
numerical solver to its limit: one drag in a 65-command sketch took
7.7-9 s, most of it in failing solves.
A failed step leaves the sketch as it was, so the steps after it all
start from the same state and none changes what the next one sees.
They are now solved eight at once on all cores; those that fail in a
row are taken as tried and the walk goes on at the first that doesn't,
as it would one by one. The drag above takes 3.7-4.3 s and every
command's reply is unchanged. Live frames, whose steps share one
frame's budget, are left as they are.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6569
* Sketch: an offset slot can still be moved by the dimension that places it (#6569)
An offset arc is drawn round its source's own centre, and its ends are
held while the sketch solves (the offset is rebuilt from its source
afterwards). Its arc condition, both ends as far from the centre, then
only tied that shared centre, and the source arc with it, to the middle
of the held ends: a slot offset by 1 mm lost a degree of freedom, and
the length that places it was refused as conflicting.
The solver and the constructive reading now leave out the arc condition
of an offset arc round its source's centre (Sketch::offset_arc_follows_source),
as #4466 did for the Parallel between an offset line and its source. The
source's own arc keeps the centre; nothing changes without an offset.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: a test for a rounded rectangle with an offset round it (#6569)
kaievns's case from the review: the offset's corner arcs tied each corner
centre to a 45 degree line, so the height followed the width and every
width change was refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6841
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PsiqyXWBFGTCX9J51bhysc
* Sheet metal: a pattern of a cut made in the flat is kept like the cut (#6841)
A rectangular, circular or path pattern of an extrude cut through sheet metal was refused
("this sheet machining cannot be preserved in its flat material"), though the cut itself is
kept as a cut-out of the flat pattern. The pattern replayed its copies through the generic
boolean, whose sheet check refuses any change the sheet model can't keep; the cut goes through
the sheet's flat machining first.
Each copy of an untapered extrude cut is now machined into the sheets the same way, in its
moved sketch plane (a moved copy keeps the extrude's prism certificate), and then cut from the
other targets. A copy that meets a folded bend is refused as the cut is there. Mirrored copies
(the sketch plane turns over) and patterns of an intersect (which cut with the union of their
copies) keep the generic path.
Fixes#6841.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PsiqyXWBFGTCX9J51bhysc
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #5098
* Sketch: deleting part of a link breaks off the rest instead of rebuilding it (#5098)
Deleting some of a link's curves left the rest linked. The next refresh saw a
different layout and rebuilt the whole link as normal geometry: the deleted
curve came back and the others lost their construction flag (a projected arc
returning as a solid arc, a projected face's lines turning solid).
sketch.delete now breaks the link for what remains, as Break Link does: those
curves stay where they are and keep their flags. Deleting a whole link, and
text links, are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FN5eEMrqWWXqi65FcK8GXq
* Move the #5098 tests up in sketch_project_tests.rs
They sat at the end of the file, where main has since added other
tests; moving them next to the break-link test avoids the merge conflict.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #5853
* Batch: refuse a step that would fail at its turn before the first step runs (#5853)
engine.batch and the MCP batch tool checked only a batch's shape up
front. A misspelled command id in a late step was found at its turn,
after every earlier step had run, and then all of it was rolled back: a
160-step batch with a typo as its last step ran for a minute and
changed nothing.
solidcraft_engine::batch_preflight::check finds, before anything runs,
the first step the batch would certainly refuse at its turn: a step
without a command, an unknown command id (the error names the closest
ids, as at the step's turn) or, with strict parameter checking, a call
that breaks its command's typed contract. Anything that depends on the
design is left to the step's turn.
- engine.batch (control channel) checks against the app's registry and
its session's level; view.pick, which the app answers itself, passes.
- The MCP batch tool checks when the commands run in its own process
(headless, with --param-check from the backend). A running app may be
another version with other commands and checks the batch itself; its
reply's not_run is passed on.
- The refusal is the failing step's reply with completed 0, undone 0,
not_run listing every step and a note.
Tests that used an unknown command only to make a late step fail now
use a step that fails at its turn (a bad extrude distance), so they
still test rollback, timings and file.new as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FN5eEMrqWWXqi65FcK8GXq
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6211
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Joints: a loop whose axes are a hair out of parallel still moves (#6211)
Holes picked on parts are rarely exactly in line. A closed loop through
axes a few hundredths of a microradian apart (a lever, link and a rod on
a cylindrical axle) can't close exactly: it closes to nanometres, and the
gap grows as it turns. The loop solver wanted 1e-7 mm, so it threw such a
closure away (the loop stayed open, its parts at 0 dof) or stopped a drive
or drag after a fraction of a degree.
The conflict check is now 1e-4 mm / 1e-6 rad (it was below what its
message can show), and joint.drive / motion and joint.reach close the
loop at each step as at the start: to the tight tolerance where they get
there, otherwise within the conflict check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #2175
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Body rename: an imported assembly's other bodies keep their names and components (#2175)
An Import keeps its body names by the file's body index, and builds no body for the
file's later instances of a repeated part (`skip`). Renaming a body stored the names by
the body's place in the model instead, so every name after the first skipped instance
landed one slot early (and the new name could land on a skipped slot). The rename looked
right until the design was rebuilt; then those bodies took other names, dropped out of
their components and showed up loose in the import's top component.
`Feature::body_name_slots` maps the bodies a feature builds to their stored slots, and
`Feature::pin_body_names` stores current names through it. The rename, the pinning of
later features' names (rename and suppress) and `Document::body_rename_from` use them.
Fixes#2175
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6348
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: constrain to an edge of any body, a mirrored copy too (#6348)
A constraint tool ignored a click on a body edge, so only edges already in
the sketch (the face it sits on) could be constrained to; a mirrored copy's
edges never were. Constraint commands now take a model edge or vertex
({edge|vertex: [x,y,z], body?}) for an operand and project it first, in the
same undo step, and the pick tools pass a clicked edge or vertex that way.
The link keeps the edge's persistent name, so the constraint follows rebuilds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6646
* Control channel: document.revision, a design revision and a document id (#6646)
A program that follows the open design asks document.revision and exports
only when the answer moves: document_id names the open design (kept through
edits, saves and renames, another for each design started or opened, never
shared between two tabs or two runs) and revision moves whenever the design,
what is hidden or an uncaptured component position changes. It stays put on
reads, selections and view changes, unlike the session's view revision.
The revision is read lazily: the session remembers the design it last
reported (held weakly, so Arc::make_mut can never change it in place and its
address can't be reused) and takes a new number from one counter for the
whole run when the design is another one. The control method answers at once,
also while a slow command computes; engine.execute reaches the same command
for scripts and MCP.
* document.revision: hash the run's id prefix with the process's random keys
A clock set before 1970 gave every run the prefix 0, so a restarted app could
answer a new design with the id and revision of the old one. The docs no
longer promise an answer at once: a screenshot waiting for an opened design to
be meshed still holds the requests behind it.
* document.revision: its label in the complete catalogs (ja, zh)
* claim: #6985
* Control channel: control.subscribe and design.changed events (#6985)
A connection that sends control.subscribe gets the active design's id and
revision in the reply (as document.revision answers), then a
{"event":"design.changed","params":{document_id, revision}} line whenever
either changes, so a live link need not poll.
The window publishes the design at the end of each frame's logic and
drawing while a connection is subscribed (one atomic load otherwise). Each
subscribed connection waits on that from a thread of its own and writes
the latest after a 100 ms gap, so a burst of changes is one line. Replies
and events share one lock per connection, each written whole (#6390).
* control.subscribe: the reply before any event, a second one refused
The events thread started before the subscribe reply was written, so a
change right after the window's answer could reach the client before the
reply that named the older revision. The subscription now starts holding
the connection's line lock and writes the reply under it.
A second control.subscribe on a subscribed connection is refused on the
connection itself: its answer could arrive behind a newer event. When no
thread or socket is left for the events, the reply says it did not
subscribe instead of ok.
* control.subscribe: a second one does not lose a finished late reply
The already-subscribed answer came before the busy check, so a request that timed out and had finished since lost its reply: control.pending then said the app ended it without one. It now comes after the busy check and drops a finished late reply as every other request does.
* control_server: declare the subscribe guard at the top of the loop
Clear of the lines #6643 changes (its token check moves to where the declaration was).
* control_server: the subscription's items beside serve
Away from the lines next to send, where #6643 adds token_of.
* control test: New on a blank design reuses its tab, so redo the box first
---------
Co-authored-by: MateuszSychXpl <151755914+MateuszSychXpl@users.noreply.github.com>
* claim: #617
* fix(sketch): manual Horizontal/Vertical between a circle centre and an edge (#617)
The manual H/V tool silently did nothing when a point and a straight line
were picked or selected together (only two points or a lone line worked),
while auto-constrain inference applied the alignment. A point picked or
selected with a straight line now aligns the point with the line's
midpoint, in the mode the line's dominant axis suggests, through the
existing midpoint alignment in sketch.constraint.horizontal_vertical.
- tools: constraint_preselection consumes [point, line] selections; the
interactive tool applies the alignment on the second pick (a line).
- context menu: Horizontal/Vertical Constraint items for point + line.
- engine + ui regression tests pin centre-edge alignment and the
edge-to-centre dimension from #617.
Written by Qwen.
* style: formatting cleanup for #617 files
---------
Co-authored-by: Dahak Builds <dahakbuilds@users.noreply.github.com>
* Extend: join a stop that is already another curve's end
sketch.extend refused a target that coincided with another curve's end
("the extension target coincides with existing geometry") instead of
extending onto that point. The stop is now joined: a new line ends on the
existing point, and a free end grows onto it held by a Coincident. The
preview shows that stop too, so the line it draws is the line the click adds.
A stop at the end of a spline or conic joins that end the same way, and
extending an infinite line now refuses instead of moving one of its two
points.
🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>
* Extend: close an arc, take an equal stop, preview with the command's search
An arc with nothing on its path now grows all the way to its own far end and
closes into a whole circle instead of reporting that there is nothing to
extend to.
Two curves stopping the extension equally far are both on the way, so the first
stop is taken instead of refusing to choose between them.
The line preview and the command now share one stop search (`line_aim`), so the
preview can no longer draw a line the click refuses.
---------
Co-authored-by: n0tfr3shm1lk <n0tfr3shm1lk@ukr.net>
Co-authored-by: Codebuff <noreply@codebuff.com>
* claim: #1994
* plastic.snap_fit: base_height and new_body options (#1994)
Base Height extends the snap fit's base below the picked face; New Body
makes the snap fit its own body so it can be patterned or moved. Both
optional, serde-defaulted, old designs replay identically.
* ui: snap_fit length-field defaults list gains the optional base height
---------
Co-authored-by: Dahak Builds <dahakbuilds@users.noreply.github.com>
* Browser: a feature row edits its profile sketch, suppresses and rolls back
A sketch button on feature rows with a profile sketch, and the feature menu gains
Edit Profile Sketch, Suppress/Unsuppress and Roll Back Here. Features behind the
history marker or suppressed are drawn faint, as in the timeline.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: the sketch button sits right after the feature's name
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6545
* cli_link: the Unix-only link code and its test build on Unix only, so the UI crate builds on Windows (#6545)
slot() is used only by link_in's Unix version, and links_only_where_it_may
makes its own link with std::os::unix::fs::symlink; gate both like link_in.
On Windows the test failed to compile (E0433) and clippy -D warnings flagged
slot as dead code.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6755
* Frame rate readout stays beside a right-side pane, not under it (#6755)
The readout and its details sat in the viewport's bottom-right corner even under the
floating Assistant. They now move just left of any pane over that corner and come back
when it closes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Every view attached to an API-agent message stayed in the history as a
base64 data URL, and the whole history goes with every request (up to 24
per message). Requests grew by one screenshot per message until the
endpoint refused them with HTTP 413, after which every message failed.
Keep the newest two views as pictures and swap older ones for a short note;
the text sent with them and every other message stay. HTTP 413 now says the
conversation is too large for the endpoint and to start a new chat.
Fixes#2298
Main now speaks the Anthropic protocol too, where a view is an `image`
block rather than an OpenAI `image_url` part: drop_old_views trims both,
before trim_history. ui.ai_http_too_large is in every catalog that has the
other ai_http_* messages.
Co-authored-by: Framewrk CI <ci@framewrklabs.ai>
* claim: #6396
* Sketch: a regular line on a construction line is picked before it (#6396)
Coincident sketch curves tied on pick distance and the one drawn first
won, so a construction line drawn first hid the regular line on part of
it. Construction and centre lines now get half a screen point added to
their pick distance: on a tie the regular curve is hovered and picked,
and a construction line clearly nearer the cursor still wins. Clicking
again and Select Other still reach the construction line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FN5eEMrqWWXqi65FcK8GXq
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
wgpu's default uncaptured-error handler panics ("wgpu error: Out of Memory").
The headless picture devices already report instead; the window's device had
no handler. Errors the code expects stay under error scopes.
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
A line tangent to an arc at the arc's end had no sense: the tangency held
with the line on either side of its touch point, so pinning a line's far
end across the corner left the line cutting back through its fillet, and
when the solver found the mirrored corner the arc kept its stored
direction and became its three-quarter complement, inside out.
The solver now keeps each fillet an arc of its corner. A line tangent at an
arc's end keeps its sense there (it leaves the arc as it did) and the arc
its side of its chord, compared with the positions before an interactive
move or the stored ones. A corner that mirrored gets its arc on the
mirrored side: its ends swapped where the solve ran the arc round the
other way with the lines following, or the arc reflected across the line
that stayed and solved again where a line cut back through. A solve that
would loop the arc the other way round, or fold a line onto it, is refused
as a fold is. A drag's cursor-following steps keep the fillets too, so a
line dragged through its own fillet stops at the arc.
The kit slider adds a real `egui::Slider`, lets egui paint it, then paints an
opaque rectangle over it and draws the kit's track and thumb on top. That mask
was exactly `slider_width` wide, and egui places its handle so the circle's
right edge lands on `rect.right()` at the maximum, so the fill stopped flush
with the handle and its last column survived as a light sliver beside the
value box. The kit's own thumb cannot hide it either: its radius is
`t.space_8` while egui's handle radius is `input_radius`, about two pixels
larger.
Extend the mask one pixel past its right edge. The gap before the editor is
`item_spacing.x`, so nothing else is covered, and only the maximum was
affected: at the minimum the handle's left edge coincides with the mask's left
edge, which a fill does cover.
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
* claim: #6524
* Sketch display: each sketch samples within its own budget (#6524)
After a large DXF was inserted, lines drawn in a later sketch vanished
once finished: only their end points showed and could be picked, and
hiding the DXF sketch brought them back.
The display cache had one sampling budget for every shown sketch (4096
curves, 131 072 vertices per model state and view). Sketches are drawn
and picked in timeline order, so a large imported drawing used it all
and every later curve came back empty (BudgetLimited): not drawn, not
pickable, while the points, drawn separately, stayed.
Each sketch now has its own budget with the same limits (curves,
profiles, vertices and nodes). The cache as a whole is capped at 65 536
curves and 16 384 profiles, so memory stays bounded however many
sketches are shown.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FN5eEMrqWWXqi65FcK8GXq
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #6657
* Context menu: the Appearance flyout closes when the pointer moves to another row (#6657)
Hovering Appearance opened its quick-choice flyout, but resting on a plain
row of the same menu never closed it: only another submenu row replaced it.
Now a flyout closes once the pointer moves onto another row and rests there
for 0.3 s, so a diagonal path to the flyout across a neighbour keeps it open.
A flyout opened with the keys stays open while the pointer is still.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>