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]>
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]>
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]>
Paper-space commands could not pick model geometry displayed by a layout
viewport: `hit_test_wires` deliberately excludes viewport content in PSPACE,
and that has to stay true for selection. This adds the capability as a separate
query rather than by changing those semantics.
`OpenCADStudio::paper_viewport_snap` walks the current layout's displayed
content viewports top-most first, skips those whose display is off or whose
clip boundary excludes the cursor, and for each one runs the EXISTING snap
engine against that viewport's model wires using the viewport's own camera and
screen rectangle (`viewport_edit_frame_for`). The accepted model point is
mapped back onto the sheet with `ViewportFrame::model_to_paper`, so everything
downstream keeps working in paper coordinates — LINE still draws at the
projected paper point. Grid snap and the paper-space rubber-band origin are
suppressed for that pass; results outside the clip are rejected so no snap
lands on a clip cut. Locked viewports snap like unlocked ones.
No projected geometry is rebuilt per pointer event: the wire set is the
renderer's resident per-viewport Arc (`model_wires_for_viewport_arc`) and the
broad phase reuses the shared interaction index over it.
The result is merged with the ordinary paper-sheet snap by `snap::merge_snap`,
which applies the engine's own priority-then-distance rule, in both the
cursor-move and the click paths.
`SnapResult` gains Copy fields `viewport: Option<Handle>` and
`source: Option<DimensionAssociationSource>`; the engine fills `source` from
the owning wire wherever it knows it. `AcceptedSnap::from_snap` converts a
result plus frame into the shared PR2/PR3 contract type.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Constraint delete: a Properties-panel row delete button, viewport
click-to-select on a constraint glyph pill with Delete-key support, and a
right-click "Delete Constraint" menu item — all routed through one
undoable Scene::selected_constraint-aware path.
Fixes two named-parameter bugs found while testing that feature:
- DIMCONSTRAINT: a parameter named like a mode-switch keyword ("r", "d",
"a", "x", "y") could never be referenced, since the keyword check ran
before the parameter-name check.
- PARAMETERS modal Apply rebuilt Scene::named_parameters but never
refreshed the Properties panel, leaving renamed/deleted rows stale
until an unrelated refresh (e.g. a selection change) happened to occur.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Include decoded IMAGE/OLE content in the shared PDF plotting path,
with clipping, transforms, transparency, draw order, and resource reuse.
Use named page content for export and printing.
Fixes#1266
Entities that carry the standard PE_URL XDATA record (what HYPERLINK writes)
can now be followed from the canvas: Ctrl+click opens the link in the system
browser and consumes the click, so no box/lasso selection starts.
- scene::selection::pe_url_of() reads the record with the same rule as the
Properties panel hyperlink field; unit-tested.
- The ViewCube/compass hit region wins over a hyperlink drawn beneath it.
- Unchanged behaviour when Ctrl is not held or the entity has no link.
A day of pan/zoom that had suddenly become unusable on a 1.2M-entity
drawing turned out to be no regression at all: an unattended NVIDIA
userspace update had been applied without a reboot, the Vulkan ICD could
no longer talk to the still-loaded kernel module, and wgpu silently fell
through to Mesa's llvmpipe. Nothing in the app said so. iced picks the
adapter and names it only under RUST_LOG, so the only evidence was 32
`llvmpipe-N` threads in /proc. The same silence covers WARP on Windows,
SwiftShader in browsers, and the case where wgpu finds no adapter at all
and iced quietly draws the chrome with tiny-skia while the shader widget
is dropped, leaving a blank viewport. This commit makes all of that
visible, on every platform, without new dependencies.
Detection (src/scene/pipeline/mod.rs)
-------------------------------------
`GpuStatus` is derived from two signals:
* `record_gpu_adapter`, run from `MultiPipeline::new` (once per device,
next to the existing uncaptured-error handler), reads
`Device::adapter_info()`, prints one `[gpu] adapter: <name> (<backend>,
<device type>, driver …)` line — stderr natively, `console.warn` on the
web, which has no stderr — and keeps a copy.
* `note_scene_draw`, run from `ViewportPane::draw`, counts how often the
viewport asked for a primitive. Every renderer makes that call; only
wgpu goes on to `prepare` and create a device. Three draws with no
device ever recorded therefore mean the primitive is being dropped.
The verdict is one of `Unknown`, `Hardware`, `Software(adapter)` or
`NoRenderer`. `Software` is `DeviceType::Cpu` — what Vulkan reports for
llvmpipe, DX12 for WARP and GL for a renderer string that gives it away —
plus a name match on the fallbacks browsers expose (SwiftShader, llvmpipe,
lavapipe, "Microsoft Basic Render"), because the WebGPU backend reports
every adapter as `Other` (gfx-rs/wgpu#8819). Metal has no software device,
so on macOS only `NoRenderer` can occur (a VM without a GPU); it is caught
by the draw count like everywhere else.
`gpu_status_if_changed` costs one atomic load per app message: a
generation counter moves only when a device is recorded or the draw count
crosses its threshold, so the app re-derives the verdict a handful of
times per session.
Notification (src/app/startup.rs, view/modal.rs, ui/statusbar)
-------------------------------------------------------------
`refresh_gpu_status` runs after every message. On a degraded verdict it
speaks through three channels, because many users never read the command
line:
* A popup, `ModalKind::GpuWarning`, queued once per verdict through the
startup-modal queue so it waits its turn behind the file open. It names
the adapter, states the consequence (slow navigation, unstable or
missing rendering) and gives the remedy that usually applies on the
running platform: a driver update not yet followed by a reboot on
Linux, a missing driver or a Remote Desktop / VM session on Windows, a
VM without a Metal device on macOS, hardware acceleration switched off
in the browser. "Don't show again for this device" stores
`GpuStatus::identity()` in `settings.gpu_warning_silenced`; keying it by
verdict means silencing "software rendering on llvmpipe" does not
silence a later, different failure.
* A status-bar pill in the theme's warning colours, `⚠ Software rendering`
or `⚠ No GPU renderer`. It is not a `StatusPill`, so it cannot be hidden,
and it stays for the session; clicking it reopens the popup, silenced or
not.
* A command-line entry through the new `CommandLine::push_warning`, which
is styled like an error but deliberately does not touch `last_error` /
`error_revision`: automation reads those to decide whether the command it
just ran failed, and a graphics warning is not a failed command.
`subscription()` requests frames only while the verdict is `Unknown` on a
document tab (the Start page draws no viewport), so the probe costs one
frame on a working GPU, three on none, then stops.
Verification
------------
* Live on the machine that had the incident, with the Vulkan loader
restricted to Mesa (`VK_DRIVER_FILES=…/lvp_icd.json WGPU_BACKEND=vulkan`):
popup → OK → pill → click reopens; "Don't show again" → restart → pill
only, no popup.
* `gpu_tests::the_adapter_is_reported_once_per_device` (ignored, needs an
adapter) asserts the classification on a real device; it was run on both
llvmpipe (`Cpu`) and an NVIDIA GT 1030 (`DiscreteGpu`).
* Unit tests cover the name heuristic, the verdict function, the identity
rule, the warning text with an adapter name that itself contains
parentheses and commas, the settings round trip and the command-line
semantics. New strings are in all 21 locales as the `gpu` Fluent group.
Co-Authored-By: Claude Opus 5 <[email protected]>
Builds from main reported only the release version (2026.36) even when
HEAD was hundreds of commits past the tag. build.rs now also emits
OCS_TAG_DISTANCE, OCS_COMMIT_DATE, OCS_BUILD_METADATA and
OCS_FULL_VERSION: a clean checkout of the release tag stays "2026.36",
anything else becomes "2026.36+194.gef189d77", with ".dirty" appended
for uncommitted changes and only the hash when no tag can be counted
from (shallow clone, tarball). OCS_APP_VERSION is unchanged so the
update check, donation gate, User-Agents and window title keep working.
The About dialog gains Build / Commit date / Profile cards, Copy Info,
the REPORT issue template and recovery reports carry the full version,
and `-V`/`--version` print it instead of Cargo's 2026.36.0. Release
workflows check out full history so release binaries see the tag.
Also fixes the build being marked dirty when there is no repository:
diff-index exits 128 there, not 1.
Co-Authored-By: Claude Opus 5 <[email protected]>
The MTP/M2P midpoint-between-two-points modifier existed as a typed transparent command only. Add it to the Shift+RMB snap override menu as an icon-only cell after Parallel (hover tooltip, no text) with a new osnap/mtp.svg marker, plus a context-menu entry while a point prompt is active.
Christopher Overbeck