Pulls the acadrust translate fix: MOVE now carries a block's attributes
and the whole MultiLeader (text/base/block-content anchors + break
points), not just the insert point / leader line. No OCS code change —
the transform path already calls translate and re-tessellates.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The selection-filter popup now has the same Select All / Clear All header
as the OSNAP popup: Select All clears every exclusion (all types
selectable), Clear All excludes every present type. Each button disables
when it would be a no-op.
Also remove the "Object Snap Settings…" footer from the OSNAP popup.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The bottom status bar was hard to read: size-10 text labels in a 26px
strip. Make it legible and consistent with the drawing tab bar:
- Height 26 → 30 to match the document tab bar; the command-line input
pill matches too, so the three strips line up.
- Replace the drafting-aid toggle labels (ORTHO/POLAR/OSNAP/OTRACK/DYN/
LWT and the OCS pills ISO/QP/FILTER/SC/CLEAN/TPY) with 17px SVG icons;
the name stays in each pill's tooltip. Twelve new icons under
assets/icons/status, drawn in the existing 24×24 stroke style.
- Vertically centre every element via center_y, and bump all pill text to
size 12 with uniform padding plus larger chrome icons (menu, scroll
arrows, osnap dropdown) so nothing reads smaller than the toggle icons.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Grow the attribute editor into a three-tab enhanced editor:
- Attribute: the tag/prompt/value list; click a row to select it, edit
its value below.
- Text Options: the selected attribute's Text Style, Justification,
Height, Rotation, Width Factor, Oblique Angle, Backwards, Upside down.
- Properties: the selected attribute's Layer, Linetype, Color,
Lineweight (the colour list also surfaces a non-standard current
colour so it displays and round-trips).
Applying writes value, text-formatting and common-property edits back to
each attribute positionally (guarded on block name + count), with undo
and a repaint. OK/Cancel are replaced by a single Apply button in a top
toolbar that commits but keeps the dialog open, matching the other modal
windows — the frame ✕ closes and discards un-applied edits. The frame
adopts the shared style-window palette, tab styling and sizing.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Each row in the attribute editor now shows the attribute's prompt — the
text defined on the block's ATTDEF — falling back to the tag when the
block defines no prompt. The prompt is read from the block definition,
since attribute instances carry only tag + value.
Bumps the acadrust pin to pull the matching reader fix: its DWG object
reader was discarding the ATTDEF prompt, so every prompt came back empty.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Double-clicking a block reference that has attributes, or running ATTEDIT
on one, now opens an editor dialog listing every attribute (tag + editable
value) with OK / Cancel. OK writes the values back to the block, with undo
and a repaint; Cancel discards. A block with no attributes still enters
in-place block edit (REFEDIT) on double-click as before. (#192)
ATTEDIT opens the dialog directly when a suitable block is already
selected; otherwise it runs the pick command and the dialog opens once a
block is chosen. This replaces the earlier command-line, per-attribute
prompt flow, which is removed along with its __ATTEDIT__ sentinel path and
the now-unused attedit_set_attrs trait hook; the ATTEDIT command is reduced
to a plain block picker.
The dialog is a tab-scoped in-canvas modal (Plan B): its working copy holds
a document-local handle, so closing that tab or switching away dismisses the
editor rather than risk applying edits to another tab's document.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Selecting an INSERT with attributes now adds an editable "Attributes"
section to the Properties panel — each tag is a row whose value can be
changed inline (Enter commits, undo works, the attribute text repaints).
Previously attribute values could only be reached through ATTEDIT, which
didn't let them be changed (#192).
Attribute tags are runtime strings, so they can't ride the geometry
edit path whose field key is `&'static str`; a dedicated PropValue::
AttrText variant plus PropAttrInput/PropAttrCommit messages carry the
tag. Values are stored verbatim (no expression evaluation, since
attribute text is free-form).
Also guard the panel's in-progress edit buffer: it now only carries
across a rebuild when the selection is unchanged (a commit-triggered
refresh), so a typed-but-uncommitted value can't display or commit onto
a different entity — e.g. two title blocks sharing a REV1 tag. This
closes a latent stale-buffer leak that affected geometry fields too.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The ribbon layer dropdown grew unbounded so a long layer list ran off the
bottom of the screen with no way to scroll. Cap the panel height and make
the list scrollable; short lists still shrink to fit.
While a dropdown was open, the 3D viewport beneath it kept tracking the
cursor over the panel's empty areas, and the OS cursor vanished there. In
iced 0.14 mouse_area/opaque only capture button presses, never CursorMoved,
so a higher stack layer can't swallow motion — gate it state-side instead:
- ignore viewport motion in on_viewport_move while ribbon.open_dropdown is set
- suppress the crosshair overlay's Hidden cursor (and crosshair draw) so the
normal OS cursor shows over an open dropdown
- route all ribbon dropdown backdrops through a shared dropdown_backdrop()
Co-Authored-By: Claude Opus 4.8 <[email protected]>
PURGE removed unused layers/styles/linetypes from the document but never
refreshed the layer panel + ribbon caches, so the removed definitions
stayed visible in the UI (issue 228). It also never purged blocks.
- call refresh_layer_panel() after a purge so panel/dropdowns update
- purge unused block definitions (no INSERT/Dimension reference), skipping
anonymous, layout, and xref records; drop their member entities too
- add PURGE BLOCKS sub-arg; bare PURGE / PURGE ALL now include blocks
- report a per-type breakdown in the result message
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Layer color/linetype/lineweight and MATCHPROP edits set the
document dirty flag but never bumped geometry_epoch. The wire
caches key on that epoch and bake a ByLayer entity's resolved
color at tessellation time, so they returned stale wires until
a new entity forced a rebuild. Bump geometry on those four
handlers so the edit repaints immediately.
Also remove REGEN/REGENALL/REDRAW/REDRWALL: the GPU raster
pipeline keeps the display continuously in sync, so on-demand
regen/redraw is a no-op concept in OCS.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Opening / navigating a heavy 3D DWG (e.g. a BIM model with ~10k ACIS solids)
was effectively frozen. Several independent costs, all addressed here:
- Face3D edge/fill buffers were re-uploaded on every camera move (keyed on
camera_generation). They are world-space and selection-independent, so gate
them on (geometry_epoch, fill_mode) instead — a pan no longer re-walks every
wire.
- Solid meshes were drawn one buffer + one draw call per solid (~10k draws/
frame), which strangled the GPU front end. Concatenate every solid's LOD0
into a few large buffers (split only to stay under the 256 MB buffer cap) and
draw the whole set in a handful of calls. Built once per geometry epoch, so
selection / hover never re-pack it.
- Selection / hover highlight is now a small tinted overlay of just the picked
solids, drawn last with an Always-depth pipeline so it shows on top even when
the solid is occluded — instead of re-tinting and re-uploading all meshes.
- Hover / pick hit-testing scanned every triangle of every solid each move
(seconds on a heavy model). Add a 3D-AABB screen-rect broad phase
(MeshLodSet now carries z_aabb) and reuse the renderer's cached expanded mesh
set, so picking is O(solids) cheap projections plus the few solids actually
under the cursor.
- Block-internal hatch hit-testing exploded every INSERT on every hover.
Cache the result per geometry epoch and skip blocks that contain no hatch.
- Curved-surface (cone/sphere/torus) tessellation tolerance is now radius-
relative (CURVE_REL_TOL) so a cylinder's facet count is size-independent and
matches the circle/arc wire tessellation, instead of exploding on large radii.
Also adds docs/tessellation.md mapping every EntityType to its tessellation
path (truck B-rep vs truck curve topology vs direct).
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Pulls the quadratic→linear AcDs blob-extraction fix in acadrust. 3D-heavy
DWGs (and master files that xref them) that took 28s+ to parse — minutes
for a multi-discipline set — now open in well under a second each.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Bare BACKGROUND now opens an interactive colour prompt (via
ValuePromptCommand, same idiom as PDMODE/LTSCALE) instead of just
printing usage, so the command works both step-by-step and as a
one-shot `BACKGROUND <colour>`. Options reordered to
Default/Black/DarkGray/Gray/LightGray/White; DEFAULT restores the app
default, now unified to rgb(33,40,48) across startup, load and reset.
Also drop the unreachable OPTIONS arm in dispatch_display.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Entities inside a block definition that sit on layer "0" must take on the
properties of the layer the block reference is inserted on; every other
layer is "sticky" and keeps its own identity. OpenCADStudio resolved a
block child's ByLayer color/linetype/lineweight against the child's own
layer in all cases, so layer-0 content stayed layer-0 (white) instead of
inheriting the instance layer.
Resolution now follows the standard rule for block children — explicit
wins, then ByBlock → the insert's own style, then layer-0 + ByLayer →
the insert's *layer* style, else the child's own layer. The insert's
layer style threads through the whole block-expansion chain (and chains
correctly through nested inserts), with the child's own transparency
preserved.
Covered across every block-render path:
- wires (line/arc/circle/polyline/text) via the block cache + the
explode fallback + insert attributes
- hatch / solid fills (also fixes the PDF/print export path)
- 3D solid block children (per-instance mesh recolor)
- Dimension / Table baked-block sub-entities
New shared helpers in scene::view::render: InheritStyle, layer_render_style,
lineweight_to_px; render_style_for_block_sub gained the layer-0 branch.
Adds unit tests covering the layer-0, sticky-layer, ByBlock, explicit and
transparency cases.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Modify commands invoked with nothing pre-selected (ERASE, MOVE, COPY,
ROTATE, SCALE, MIRROR, ARRAY*, STRETCH, EXPLODE, GROUP, BLOCK, the
clipboard cuts, LAYOFF/FRZ/LCK/ULK, ...) ran the command on the FIRST
picked object and terminated. You could not build up a selection.
SelectObjectsCommand now keeps gathering: each pick accumulates into the
set (Shift removes), the prompt shows the running count, and the command
fires only when the user presses Enter or right-clicks. Reuses the
existing JOIN gather pattern, so no viewport/host changes are needed.
LAYMCUR acts on a single object's layer, so it keeps the immediate
first-pick behaviour via the new SelectObjectsCommand::instant().
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The preview box rendered a hardcoded `text("AaBbCc 0123")` that ignored
every edited property. Replace it with a canvas that tessellates the
sample string through the same stroke path the text entities use
(`lff::tessellate_text_ex`), so font, width factor, oblique angle and the
backward/upside-down flags are reflected live as fields change. Effective
font follows the entity rule (TrueType name wins, else stroke font file);
the result is fit to the preview box with Y flipped.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The v0.7.0 web (GitHub Pages) release failed to compile under
--no-default-features. Two symbols leaked into the wasm build:
* `solid_section` (SECTION command) used truck-meshalgo directly
(`triangulation`/`extract_interference`), but that crate is gated
behind the `solid3d` feature which is off on wasm. Gate the function
with `#[cfg(feature = "solid3d")]` and add a stub for builds without
it, mirroring `solid_model::boolean`.
* `register_as_handler` was `#[cfg(not(wasm32))]` while its caller
`set_default_app_blocking` was not, so wasm couldn't resolve it. Drop
the wasm gate; the existing non-win/lin/mac arm already returns `Ok`
on wasm, matching `unregister_handler`.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Two gaps in the autosave recovery files:
1. A brand-new, never-saved drawing was never autosaved — on_autosave only
wrote tabs that already had a path. A shared autosave_target() now writes
saved drawings to <file>.sv$ and unsaved ones to the system temp dir
(OpenCADStudio_<name>.sv$).
2. Closing a file (tab) left its .sv$ behind, which also defeated the
program-close cleanup (it only scans still-open tabs). on_tab_close and the
unsaved-discard path now delete the closing tab's .sv$, and Save/Save-As
drop the temp copy a drawing used while it was still unsaved.
Program-close cleanup (exit_app -> cleanup_autosaves) already worked.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The .sv$ recovery copies only need to survive a crash, so they are removed on
a clean save (Save and Save-As both delete the file's .sv$ after writing) and
on a clean exit — a new exit_app() runs cleanup_autosaves over every open tab
before quitting, and all iced::exit() paths now funnel through it. A leftover
.sv$ therefore signals an unclean shutdown, i.e. something to recover.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
A periodic subscription (every SAVETIME minutes — default 10, 0 disables)
writes a <name>.sv$ recovery copy of each dirty, already-saved drawing at its
own DWG version. Non-destructive: it never touches the original file or the
dirty flag, and skips unsaved (path-less) tabs. The SAVETIME command
sets/queries the interval and persists it. Recovery files (.sv$ and .bak) open
through the normal Open dialog — the loader detects DWG vs DXF from the leading
magic bytes, and a "Backup / Autosave" filter lists them.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Drop the every-launch (re)registration thread. File association is now set up
once via the first-launch prompt (Yes → register + set default; No → decline,
persisted as file_assoc_enabled) and managed afterwards by the FILEASSOC
command. set_default_app now registers the handler entries (.desktop / mime /
ProgIDs) itself before asking the OS to default to us, since boot no longer
does — so the default never points at a missing handler.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
`FILEASSOC 1` (re)registers the app as a .dwg/.dxf/.bak handler; `FILEASSOC 0`
unregisters it (removes the .desktop + CAD mime package on Linux, the per-user
registry keys on Windows); bare `FILEASSOC` reports the current state. The
choice persists (file_assoc_enabled) and gates the every-launch
re-registration, so turning it off stays off across restarts.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- Save: overwriting an existing file first copies it to a sibling <name>.bak
(default on; toggle with the ISAVEBAK command, persisted in settings).
- Open: .bak is a first-class type — the Open dialog lists it and the loader
detects DWG vs DXF from the file's leading magic bytes (the extension hides
the real format).
- File association registers .bak too, every launch (so users who associated
earlier get it automatically): Windows via the registry; Linux via a
high-priority shared-mime-info magic that resolves DWG/DXF content under the
.bak extension, with the generic *.bak glob claimed only once the app is the
default DWG handler — so it never hijacks other apps' backups.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Two in-canvas confirmations on the Save-As path:
- AEC/Civil drop warning: when the drawing carries unsupported AEC/Civil
(application) objects that survive only as verbatim source-version bytes,
saving to a different version or to DXF would silently drop them. The
dialog reports the count and offers "Save in source version" (keep) vs
"Save anyway" (drop). Same-version DWG saves are lossless and skip it.
- Overwrite confirmation: when the chosen filename already exists in the
target folder, "Replace" / "Cancel" before clobbering it.
The two chain (AEC first, then overwrite) via per-flow acknowledgement flags
reset whenever the Save dialog opens or closes. Detection is a new
io::dropped_on_save_count over the document's Unknown (raw-passthrough)
objects/entities.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
When the command line holds a typed-but-unentered command (or option
keyword / value), a viewport right-click now runs it, mirroring Enter.
Routed through CommandFinalize so the MText preview and grip-popup
guards apply and a non-empty line forwards to the submit path. All
other right-click behaviour is unchanged and applies only when the
command line is empty: right-drag orbits, the first right-click during
an active command acts as Enter, a second opens the context menu, and
an idle right-click opens the menu.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
DWG→DXF conversion now drops references to objects that have no DXF form
(unsupported AEC/application objects dropped on conversion), so the output
opens clean instead of auditing 619 dangling handle references.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- FLATSHOT: project the selected solid's edges to 2D Line entities at Z=0
- CONVTOSURFACE: convert the selected solid(s) to Surface entities
Both reuse the cached B-rep's edge wires. COMMANDS.md: 219/262.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Package the drawing plus its referenced files (xrefs + raster images)
into a sibling "<name>_archive" folder via std::fs, reusing the
existing block_records / entities iterators. COMMANDS.md: 217/262.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Convert selected closed polylines and circles into Region entities,
one wire loop each (reuses acadrust Region/Wire). COMMANDS.md: 215/262.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Six more standard commands by reusing existing infrastructure:
- CLOSE: close the active tab (Message::TabClose)
- HIDE / VISUALSTYLES: per-viewport render mode (Message::SetRenderMode)
- QUICKCALC / QC: alias the CAL arithmetic evaluator
- HYPERLINK <url>: attach a hyperlink via the PE_URL XData record
- XOPEN: open a selected xref's source file (Message::OpenRecent)
COMMANDS.md updated to 214 done / 15 partial / 33 missing of 262.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Integrate a large batch of standard CAD commands and fix several
name/behaviour mismatches, with no product references in code.
3D solids (reusing the truck B-rep pipeline):
- SLICE, INTERFERE, PRESSPULL, POLYSOLID, PYRAMID, SECTION
- 3DROTATE, 3DMIRROR, 3DALIGN (cached-solid transforms)
Annotation / text:
- ARCTEXT, SPLINEFIT (Catmull-Rom to Bezier), DIMJOGGED, OBJECTSCALE
(XData markers honoured by the existing geometry-gen / tessellator)
Blocks / misc:
- NCOPY (block-content extraction via explode_entity)
- OPTIONS, SYNCPVIEWPORTS, THICKEN, plus earlier toggles, aliases,
system variables, and layer/style/view commands
POINTCLOUD/RECAP and UNDERLAYLAYERS/UOSNAP are intentionally left out:
they require a proprietary binary / PDF-layer parser, a GPU point
renderer, and real test files.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
PR #213 introduced rustfmt drift in tip_style / swatch indentation and an
inline if-expression in two previously fmt-clean files. Format pass only;
no behaviour change.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
printpdf's default = ["html"] pulled in azul-layout, rust-fontconfig,
kuchiki, xmlparser and their transitive trees (~40+ crates including
two with cdylib targets) — none of which are used by the project's
PDF export (only basic primitives like Mm, Line, Polygon, Rgb).
Also fixes a MinGW linker error (export ordinal too large: 100035)
triggered when ld.bfd tries to build a DLL for printpdf with >65535
exported symbols.
Consolidates 21 occurrences of identical Color literals scattered across widgets.rs and mod.rs into 9 named constants (COMBO_BG, COMBO_HOVER_BG, COMBO_OPEN_BG, COMBO_BORDER, COMBO_ACTIVE_BORDER, COMBO_ARROW, SWATCH_BORDER, TIP_BG, HIST_INACTIVE_BG). -85 net lines, zero behaviour change.
Section labels (Draw, Modify, Annotation, …) were left-aligned with a small padding. Set align_x(Center) on the parent column so the label container — still sized to its text width — is centred within the column, which itself remains sized by the tool buttons above. Sections keep their natural content-based width, no wasted space.
The recursion-depth guard fixes the parser stack overflow, but user input
actually flows through eval_to_string (command line, viewport, dynamic
input — synchronous on the UI thread). Its embedded-expression fallback
re-tokenises every (start, end) window, so cost is ≈O(n³): a few-KB paste
of non-evaluable text or deeply nested parens froze the app for seconds to
minutes even though eval_number fast-returns None in microseconds
(measured: "x"×3200 → 16 s, 4 KB of parens → 51 s).
Cap the fallback scan at MAX_EVAL_SCAN_LEN (256) — whole-string numbers
and expressions of any length still evaluate via the fast paths; only the
mixed free-text scan is bounded, returning long input unchanged. Clarify
that MAX_RECURSION counts parser frames (256 ≈ 41 nested parens, not 256
levels). Add regression tests: the 41/42 nesting boundary, unary-chain and
function-arg-nesting rejection, and an eval_to_string over-cap promptness
test (50k junk + 100k nested parens return instantly).
Co-Authored-By: Claude Opus 4.8 <[email protected]>
R2018 cross-version saves now open in AutoCAD with NO recovery: the
MLEADER entity tail (block labels + dir/bottom/top attach order) is
correct, so all 3 MLEADERs are accepted — the Civil3D test file opens
clean (only the unconvertible AEC objects are dropped, as expected).
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- expand_defn: dedup the <1px baseline LOD branch the same way the <5px
greek branch does, so colour-split MTEXT no longer stacks N overlapping
baselines (one per \C segment) over the same OBB.
- sysfont: memoise canonical_family_name process-wide. resolve_font calls
it once per word on the MTEXT measure hot path and Face::resolve re-runs
it right after; the fontdb query + linear family scans were uncached.
- dimension/multileader/table: reword the 6 fill_tris_low FIXMEs. They
cited a stale line and the wrong consumer — these fills render on the
top-level path (panic-safe .get().unwrap_or), not the block cache, so
they cannot trip emit_wire's debug_assert. State the real status:
latent f32-precision debt, not a crash. Also fix the mis-indented
text-fill WireModel in dimension.rs.
- tests: add a host-independent colour-split test (two \C segments in a
block keep ≥2 distinct wire colours — guards the per-wire colour fix)
and a TTF-gated UTM fill test (paired, non-zero fill_tris_low). Note the
arial-style test's LFF-fallback limitation.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
emit_wire reconstructed fill verts via a raw `fill_tris_low[idx]` index,
guarded only by a debug_assert. Slice bounds checks are not gated by
debug-assertions, so a Leader with a filled arrowhead nested in a block
(populated fill_tris, empty fill_tris_low via the low-level tessellate
early-return) panicked in release as well as debug.
Use the panic-safe `.get(idx).copied().unwrap_or([0.0; 3])` — the same
"empty = all-zero low half" contract the other fill consumers
(face3d_gpu, xclip) already honour. The debug_assert stays as a tripwire.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Colour-split MTEXT produces one outline wire per inline \C/\c override. At <5px zoom each outline independently emitted the full OBB greek rectangle — N overlapping boxes with last-colour-wins. Track last_lod_obb and skip duplicate substitutes.
Also corrects the fill-only LOD bypass: suppress fill-only wires at <5px instead of sending them unconditionally through emit_wire. The outline sibling's baseline or greek box suffices at sub-pixel zoom.
dimension.rs (x2), multileader.rs (x3), table.rs (x1) emit fill_tris
without a corresponding fill_tris_low. These will trip the debug_assert_eq!
added in the preceding commit when the entity appears inside a block
definition. Each site is marked with a FIXME referencing block_cache.rs:1397
and the correct fix pattern (split_ds on each vertex).
Three related bugs in tessellate_sub_local and expand_defn:
1. Wire folding lost per-segment colour. Multi-wire tessellator output
was concatenated (NaN-separated) into a single LocalWire using the
first wire's colour. Inline \C/\c MTEXT colour overrides inside block
definitions rendered in the entity colour instead of the override colour.
Fix: tessellate_sub_local returns Vec<LocalWire>; each wire from the
tessellator becomes its own LocalWire with its own colour.
2. fill_tris_low was never forwarded. The double-single precision path
was silently degraded on cached wires. Fix: fill_tris_low forwarded
from WireModel through to LocalWire and BatchEntry. debug_assert_eq!
added at emit site to catch future invariant violations.
3. Fill-only wires hit the text LOD ladder. MTEXT outline/fill pairs
produce separate LocalWires; the fill-only wire was routed through
emit_text_baseline/emit_greeked_text which ignore fill_tris (invisible
double-draw). Fix: is_fill_only field on LocalWire bypasses the LOD
ladder. Also fixes StyleKey::fill_only being hardcoded false at both
emit_wire call sites (latent batch misclassification).
resolve_font previously returned &str, passing raw stems like "arial" verbatim to Face::resolve. System font lookup requires the canonical family name ("Arial"), so inline \f/\F overrides from block-nested MTEXT were silently unresolvable and fell through to no geometry.
Change resolve_font to return Cow<'a, str>:
- no override -> Borrowed(base), zero allocation
- LFF built-in -> Borrowed(font), zero allocation
- system font resolved -> Owned(canonical), one allocation
- unresolvable -> Borrowed(base), falls back to style font
All 4 call sites updated to &font_name (&Cow<str> derefs to &str).
Adds two unit tests covering the fallback and TTF-stem-to-family-name paths. Adds integration test verifying block-nested MTEXT produces geometry and correctly separates outline/fill wires.
- Save As now defaults the version dropdown to DWG/DXF 2018 (latest)
instead of the loaded file's version; the selected version is written
as-is (removed the source-version fallback for version-locked data).
- Pin acadrust 55a2caa: allocate_handle respects header.handle_seed as a
floor, fixing the corruption where opening a file and saving it
silently overwrote existing XRECORD/DICTIONARY objects (handle reuse)
and AutoCAD refused the result.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
