Found while building an out-of-tree scripting plugin from scratch against
this doc and the actual ocs_plugin_api/host source. Two categories:
Fixed outright (docs/examples didn't match the real code):
- The worked CadModule example returned Vec<RibbonGroup> from
ribbon_groups(); the real trait requires &[RibbonGroup] - the example
wouldn't compile. Fixed with the OnceLock pattern the host's own
benchmark code already uses.
- Reconciled a contradiction between "process isolation removes the need
for a shared Rust ABI" and the exact-rustc_version requirement described
a few paragraphs earlier in the same doc.
- The roadmap's "Next" list still had external automation as unstarted;
--mcp/--serve are real and operational (src/cli.rs, src/mcp.rs).
- Added the missing plugin.toml to docs/plugin-template-v2 (its own
DESIGN.md already claimed one could be installed).
- Documented that plugin-template's CI floats rust-toolchain@stable, which
doesn't guarantee matching the host's exact rustc - added acadrust_source
to both templates' plugin.toml plus the release.yml step that fills it
from Cargo.lock's resolved acadrust source, alongside the existing
rustc_version fill.
Flagging for your call, not fixed here (found the behavior, not sure it's
the intended one, prefer to ask than guess at a fix):
- acadrust_compatible()'s ABI gate only runs at all if plugin.toml declares
acadrust_source - an undeclared field is silently treated as compatible,
unlike rustc_version (undeclared = incompatible). The same shortcut
appears in two places: ExternalPlugin::acadrust_compatible
(src/plugin/external.rs) and the inline check inside fetch_release_info
(src/plugin/marketplace.rs). If this is deliberate (e.g. back-compat with
manifests from before the field existed), no action needed beyond a
comment; if not, both call sites need the same fix.
- The command line still uppercases typed/pasted input even with the `>`
literal-spaces prefix (Message::CommandInput, src/app/update/mod.rs) -
`>` only stops a space from auto-submitting the line, it doesn't skip
`s.to_uppercase()`. A plugin command needing a case-sensitive typed
argument has no escape today short of automation.
- fetch_release_info (src/plugin/marketplace.rs) returns every release
under a linked repo with no filter on the manifest's own id field (which
it already parses to build ReleaseInfo) - a repo hosting more than one
plugin under tag-namespaced releases surfaces all of them mixed
together in one plugin's version list.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Remove the dead catalog entry for the deleted XREF dead-end text and add the 389 source strings missing from the catalog (data extraction, drafting settings, panels, commands). Every locale gets the English text as placeholder under common/insert following existing practice. scripts/test_locales.py passes (5077 lookups).
Each menu row is now a single mouse_area that publishes on PRESS instead of a button nested in a mouse_area. iced_aw ContextMenu keeps its overlay element in the widget struct, so every view rebuild wipes the overlay subtree state - including a button's is_pressed between press and release - and release-time clicks published nothing. Headless simulator tests through the real overlay guard row clicks, flyout clicks, the Change Path Type hover trigger, and pin the subtree-reset mechanism. Requires iced_test as a dev-dependency; the Cargo.lock gpu-allocator/windows edge is pinned back to 0.62.2 (see docs/review-context-menu-click-delivery.md).
Row menu Select New Path and Find and Replace now act on the right-clicked row via dedicated XrefRowPathPick / XrefRowFindReplacePrompt messages (prefill + command-line focus, XOPEN output included). Rows use a RightClickArea wrapper so right-click selects the row even though ContextMenu captures the press, and opening or acting on the menu resets the Change Path Type flyout state.
The Change Path Type row shows a right-triangle affordance and reveals the Make Absolute / Make Relative / Remove Path submenu beside it. The flyout opens on hover only and a shared submenu-row style removes dead code from the earlier iterations.
Row right-click menu is exactly Open, Attach..., Unload, Reload, Detach, Change Path Type group with Make Absolute, Make Relative, Remove Path, Select New Path, Find and Replace. Change Path Type renders as indented group. Open uses OCS for DWG and OS default for images and PDFs. Preview decodes on selection change, not only at refresh. Details shows placed display width and height for image instances.
Detach and Bind on nested rows report the CLI-worded per-entry error instead of skipping silently. Adds GUI-path regression tests for every row operation on direct and nested rows, plus Reload All, Overlay, Change Path, and Find and Replace coverage.
External References dock default 360 to 460 for the column-rich table. Attach split button reads Attach DWG with Attach Image and Attach PDF options, all menu rows full width. List view shows one distinct glyph and theme color per reference status.
Right-click actions bound to the row under the cursor, split-button toolbar groups with shared borders and hover states, and a notice when an operation needs a selection first.
XREF command with status, reload, unload, detach, path, pathtype, bind, overlay, and attach options, plus guided XATTACH with unique naming and placement prompts.
Path normalization with stable keys, relative/absolute/none pathtypes, wildcard matching, and reference entries with loaded, stale, missing, and unreferenced states.
Problem
The open guard rejects circles and arcs with a radius below 1e-10. A normal
open fails when a drawing contains one. The user gets the recovery prompt.
AutoCAD and nanoCAD open these drawings.
The check came from ace99541. Its comment says the tessellator cannot sample
a degenerate curve. That was true of truck's parameter_division. Commit
8c7ebdc9 removed truck. Commit 96be8214 made the kernel sample circles and
arcs by sweep angle with a floored segment count. The reason for the check
no longer exists.
Change
- Remove the near-zero radius check for circles and arcs.
- Keep the checks for non-finite values and radii above 1e10.
- Update the guard comments. The purge rejects parser desync. It does not
reject degenerate geometry. Be generous in what you accept and strict in
what you produce.
- Add tests: a zero radius passes the guard, an absurd radius does not.
- Change the automation purge test sample from radius 0 to radius 1e11.
Behavior
- Zero-radius circles and arcs load. They render as dots.
- Tessellation, hit testing, snapping, and hatch boundaries accept a zero
radius.
- The DXF writer does not change.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Removes PR3's `viewport_frame` seam in `viewport_ref.rs`; the frame is
PR1's camera-derived one everywhere.
The five dimension commands (DIMLINEAR / DIMALIGNED / DIMRADIUS /
DIMDIAMETER / DIMANGULAR) now record a real association when they
measure through a viewport. The commit path's one association gate
(`dimension_association_allowed`, added when PR2 was stacked) gains an
else-branch: where PR2 committed the dimension unassociated and PR1's
inference guard returned an empty source list, `Scene::attach_viewport_
dimension_association` now stores the whole chain — dimension ->
viewport -> block path -> entity -> feature — from the AcceptedSnaps PR1
captured, in definition-point slot order.
* `Scene::dimension_association_slot_points` names that order, including
the radial dimensions `dimension_reference_points` does not cover.
* `OpenCADStudio::viewport_dimension_snaps` builds the slot array. A snap
whose paper point is no longer the definition point it fills has its
model point recomputed from that definition point through the frame:
the object pick records one click for both extension origins, and a
rotated DIMLINEAR projects its origins onto the measuring axis.
* Radial dimensions take their own path inside
`attach_viewport_dimension_association`. Their single reference is an
osnap type 10 carrying the radial marker (which circle / which polyline
arc segment) and the angle around it; `source_reference` would hand
back the generic curve-parameter marker -2, which
`radial_source_for_marker` cannot find.
One measurement rule, documented in `viewport_ref::measurement_rule`:
`actual_measurement` is always the RAW measurement in the dimension's own
space (paper, for a viewport dimension), and the viewport compensation
lives only in the magnitude of the negative DIMLFAC override, applied
once by the formatter. PR2's persistence already followed it; PR3's
refresh did not — it substituted the model distance times
`model_factor()`, which made a refreshed dimension read `user_lfac` times
too much versus an identical freshly-created one. Refresh now writes the
paper measurement (a debug assert pins resolved paper points to the model
points through the frame), and `axis_distance` goes with it.
Block sub-entity snaps: verified that `block_cache` stamps the top-level
INSERT's handle onto every expanded and clipped sub-wire, so
`wire_source` can only ever report the INSERT — nested inserts included.
`OpenCADStudio::resolve_source_block_path` now descends from that INSERT
at accept time (using the object-pick path's existing
`resolve_measurable_entity`, so arbitrary depth, not just one level) and
fills `SnapSourceRef.block_path` with the real inner entity. Where the
descent cannot identify one — a non-planar-pickable type such as text,
hatch, spline or solid, or nesting past eight levels — the source stays
the INSERT with an empty path, which is coarser but never wrong. Both the
limitation and where it bites are documented on
`Scene::resolve_block_snap_source` and `SnapResult::source`.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
A hand-written DXF small enough to read: one layout viewport at a 1:2
paper-per-model scale, one line, circle, arc, bulged lightweight
polyline, spline and block instance in model space, and one associative
aligned dimension per feature drawn on the sheet through the viewport.
Each association exercises a different part of the chain and feature
resolution: endpoints on a line, quadrants on a circle, endpoints on an
arc, an endpoint and a bulge segment on a polyline, a tangent-to-spline
reference (the 115448 shape), a viewport -> insert -> line chain through
a block instance, and an intersection with a second object chain. Every
association is marked trans_space, so all of them fail on the pre-PR3
resolver and all of them are expected to resolve now.
The expected values are checkable by hand: the viewport maps model
(x, y) to paper (150 + (x - 150)/2, 100 + (y - 150)/2), so a 100-unit
model length is 50 paper units and must still display as 100.
Verification tests come in the next pass; the fixture is committed now
so they have something to read.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Panning, zooming, twisting or rescaling a layout viewport moves model
geometry across the sheet without touching that geometry, so nothing in
the ordinary entity-change path noticed and the dimensions drawn through
the viewport stayed where they were — detached from what they measure.
`Scene::notify_viewport_changed` announces the viewport itself as a
modified entity. That is all it takes: every trans-space reference
carries the viewport handle as the first link of its chain, so the
dependency lookup in `refresh_associative_dimensions` already finds the
associations from there, and the refresh re-places and re-measures them.
Called from each mutator that changes the paper<->model mapping: pan,
zoom, fit-to-bounds, view normalisation on entering MSPACE, the
view-direction/twist snaps, the viewcube camera path, and both
annotation-scale setters.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
COPY, MIRROR and ARRAY duplicated a dimension's baked block but left the
copy with no association at all. Now a copy keeps one, on one condition:
every handle in every chain — the viewport, each INSERT on the block
path, the source entity, and an intersection's second chain — must have
been copied too. Then the whole chain is remapped onto the copies, so
the copy measures the copied geometry.
The condition is all-or-nothing on purpose. A partially-remapped chain
would leave a copy quietly driven by the original's geometry, which is
worse than no association; and copying a dimension on its own still
produces an unassociated dimension, which is exactly what the model
space path did before. It is the same rule `copy_complete_groups` uses
for groups, applied to a longer chain.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
This wires the chain walker into the association system proper, and adds
the entry point PR2 calls.
Resolution. `resolve_reference_chain` walks the chain, evaluates the
osnap feature in the entity's own coordinates, lifts it to model space
through the block path, and — when the chain goes through a viewport —
projects it onto the sheet with that viewport's `ViewportFrame`. Both
points are kept: the dimension holds paper points, the value it displays
is a model length. `ChainMap` is the one place that conversion lives,
including a `map_radial` that carries a radial source's plane into the
dimension's space so the radius/diameter/arc paths need no viewport
special-casing of their own. Which space a stored osnap point was
written in is not recoverable from the record — our own writer used
model/WCS, AutoCAD's trans-space records use the dimension's space, and
both turn up in the same drawing — so both readings are mapped down to
the entity and the one that lands on the geometry wins. It is only ever
a hint that disambiguates candidates, never the answer.
Creation. `Scene::attach_viewport_dimension_association` takes one
`AcceptedSnap` per definition point and writes the whole chain: viewport
handle, block path, entity, osnap type, sub-entity marker and parameter,
plus `trans_space` on the association. Every link becomes a reactor
target, not just the innermost entity, so a change anywhere along a
chain reaches the refresh through the ordinary dependency index.
Refresh. `refresh_associative_dimensions` now matches a changed handle
anywhere in a chain (including the second chain of an intersection), so
a viewport change reaches the dimensions drawn through it. Perpendicular
and tangent references resolve in a second pass, measured from the other
definition point. A dimension carries its sheet annotation with it: the
dimension-line standoff and a user-dragged text position move by the
centroid shift of the points that actually moved, so a one-ended edit
keeps the dimension line centred instead of collapsing it. For a
trans-space linear or aligned dimension the displayed value is
recomputed from the model points and multiplied only by the user's
DIMLFAC (`MeasurementScale::model_factor`) — the viewport compensation
is inherent in having measured in model space, and applying it again is
the double-scaling this three-PR effort exists to stop.
Broken references. Brokenness is derived, never written: a reference
whose source was erased simply stops resolving, the dimension keeps its
last valid appearance, and the record is left byte-for-byte as it was.
An undo that restores the source restores the link with no extra
bookkeeping. `Scene::dimension_association_status` reports it per slot.
`dimension_is_associative` now walks the chain too, instead of accepting
any surviving handle in `xrefs` — which was true even for a trans-space
reference whose actual geometry had gone, because the viewport was still
there.
Undo. Creating an association journals the new object and every reactor
list it touches; refreshing a dimension journals the dimension's
before-image. Both were previously invisible to the delta recorder.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
An `AssocDimensionReference` addresses a feature through a chain of
handles — dimension -> [viewport] -> [insert ...] -> entity -> feature —
stored in `xrefs` (AutoCAD's `mainObjectIds`). The existing resolver read
only `xrefs[0]`. That is right for a model-space dimension on top-level
geometry and wrong for everything else: a paper-space dimension measuring
through a layout viewport has the VIEWPORT first, so resolution stopped
on a viewport entity, found no curve and gave up. Nine of the ten
association objects in the 115448 regression drawing are that shape.
`dimension_assoc_chain` walks the chain positionally, which is all the
file gives us: a leading VIEWPORT is the trans-space hop, INSERTs after
it are the block-instance path (with their composed transform), and the
first handle that is neither is the geometry. A chain of nothing but
INSERTs ends on the innermost one, which is how a block reference's own
insertion point is dimensioned. A handle that is no longer in the
document comes back as `ChainError::Missing` rather than as a silent
failure, so the caller can tell a broken link from an unsupported one.
The module also evaluates the osnap feature itself, in the entity's own
coordinates: centre, midpoint, quadrant, node/insertion, nearest,
perpendicular, tangent (closed form on a circle or arc; a Newton
refinement of (C(u) - P) . C'(u) = 0 on a spline, seeded from the stored
parameter — the shape of the 115448 tangent-to-spline reference), and
intersection / apparent intersection against the second object chain.
Types we do not evaluate return None and keep their stored data, so an
unsupported reference is left unresolved instead of being replaced by a
guess.
`Transform` has no inverse of its own and a block path is always affine,
so the 3x3 cofactor inverse lives here too. All arithmetic is f64.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
PR3 has to map a resolved model point onto the sheet, which needs the
layout viewport's paper<->model mapping. PR1 (feat/viewport-snap) owns
`Scene::viewport_frame` and is not visible from this branch, so this adds
the minimal stand-in the shared contract in `viewport_ref.rs` describes:
paper_center from the viewport rectangle's centre, model_target from the
saved view target, scale from `vp_effective_scale`, twist from the saved
view twist.
It is fenced off in one clearly marked block with merge instructions.
PR1's version is camera-derived, because it must agree with what is
actually drawn — including the auto-fit rescue `camera_for_viewport`
applies to stale saved views. Nothing else in PR3 reads viewport fields
directly, so replacing this one function is the whole merge.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Removes PR2's two seams now that PR1 (viewport snapping) is underneath.
* `Scene::viewport_frame` is PR1's camera-derived implementation. The
viewport-record version is gone; PR1's returns `None` for oblique /
perspective viewports, which is what we want — every entry point in
`viewport_dim_seam` then declines rather than guessing a planar
mapping. `viewport_dim_seam` is now just the dimension object-pick
path, and its module doc says so.
* There is one accepted-snap store: PR1's `OpenCADStudio::accepted_snaps`.
PR2's parallel `dim_accepted_snaps` (and its command-name ownership
tracking, which PR1's clear-on-command-start/end already does) is
deleted, along with the `record_dimension_accepted_snap` seam.
`accepted_snap_for_point` reads PR1's list.
Two things layer on top:
- the explicit object pick is not a point step, so it still records its
own snaps, now through `push_accepted_snap` into PR1's list, and now
carrying the picked entity handle + block path as a `SnapSourceRef`
so PR3 can associate it;
- a typed coordinate carries no snap, so `measure_snap` re-attaches the
content viewport under a snap that is free *and* source-less. That
keeps `DIMLINEAR` with typed coordinates inside a 1:10 viewport
measuring the model, without ever reinterpreting a real paper-sheet
snap that happens to sit over a viewport.
* One association gate, not two. `dimension_association_allowed` is the
single predicate, derived from the same `dimension_measure_space`
classification that decides whether to compensate the measurement, so
the two can never disagree. `infer_dimension_sources_guarded` delegates
to it, and the commit path's "commit unassociated when measured through
a viewport" condition is now that predicate rather than a second,
independently-computed bool. It also covers the explicit-pick and
explicit-source-list branches, which the old inference-only guard did
not. Because the classification ignores the trailing dimension-line
placement click, an ordinary paper dimension whose placement click lands
inside a viewport rectangle keeps its association.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The negative DIMLFAC convention is resolved from the dimension's owner block,
but the entity only gets one when `add_entity_to_layout` commits it. With
DIMASSOC = 0 the dimension is exploded to its pieces first, and the baked
text would then have read as a model-space dimension and dropped the
compensation.
Stamp the layout block - the same handle the add assigns - as soon as the
override is written, so the entity is self-describing whichever path consumes
it.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Colocated unit tests pin the persisted representation: a 1:10 viewport with
no user factor writes DIMLFAC -10, a 10-unit paper span then formats as 100,
and the same dimension read back in model space does not scale. A positive
user factor folds into the magnitude, and a drawing-wide negative DIMLFAC
that already carries the compensation writes no override at all. Another
test pins the viewport model->paper affine against ViewportFrame's own
mapping, so object-pick projection and snap projection cannot drift.
Also clears the recorded snaps when a multi-dimension command commits.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Paper-space hit testing only sees sheet entities, so DIMLINEAR / DIMALIGNED /
DIMRADIUS / DIMDIAMETER / DIMANGULAR in "select object" mode found nothing
when the user clicked the model geometry a viewport displays.
A pick that lands inside a content viewport is now mapped into model space
through the viewport frame and tested against that viewport's resident model
wires - the same tessellation it draws, so the pick matches what the user
sees. Ordinary block instances are descended into, with the instance
transform (recursively, up to eight levels) baked into the extracted
geometry, so a line inside a block measures its placed length.
The picked entity is handed to the command already projected onto the sheet,
so all five commands keep working entirely in paper coordinates and unchanged;
the pick is recorded as a viewport snap, so the committed dimension is
compensated exactly like a point-input one. A miss falls through to the
ordinary paper-space pick.
Also narrows mixed-space classification to the measuring points: the final
click of each of these commands is the dimension-line placement, which lands
on bare sheet for viewport measurements and inside a viewport rectangle for
paper ones, and says nothing about what is being measured.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
DIMLINEAR / DIMALIGNED / DIMRADIUS / DIMDIAMETER / DIMANGULAR created on a
sheet now report the model measurement of the geometry seen through a
viewport, at any viewport scale, while definition points, the dimension line
and the text stay in paper coordinates and the annotation keeps its sheet
size (DIMSCALE is untouched).
A viewport frame is a similarity - uniform scale plus in-plane twist - so
projecting the model span onto the paper axis and compensating gives exactly
the raw paper measurement times 1/scale. No geometry is rewritten: the
compensation is persisted once as a negative DIMLFAC override on the
dimension and applied once by the linear formatter, so preview text,
dimension text, the properties measurement row and DXF/DWG export all agree.
User measurement multipliers fold into the same magnitude, and an explicit
text override still bypasses formatting entirely, so it is never re-scaled.
Angular dimensions are never compensated - a similarity preserves angles.
Mixed input (one point through viewport A, another through viewport B or on
the sheet) is refused rather than silently mis-measured: the command line
explains why and the dimension falls back to plain paper measurement.
A dimension measured through a viewport is left unassociated: its definition
points are projected paper coordinates, so inferring a paper-space
association would bind it to geometry that is not there. PR3 records the real
dimension -> viewport -> entity chain.
The result is one ordinary dimension entity carrying one style override, so
undo/redo works through the existing history with no new out-of-document
state.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Hakan Seven