Repository navigation
GPU Source Qualified Face Scene
Issue #1072. Native COMPLETE source-only route; later adoption/publication stages remain separate.
Authentic saved #1057 warped input contains six separate original Source0 files, each a Model0/mm with five geometry objects (one quadratic Surface/four original boundary Curves), plus a validated forward FaceBundle. "Model5" shorthand means five objects, NOT schema/model version5: actual files declare version0. Repeated face_local/loop_local/c0..c3 IDs remain local to explicit sourceId A..F. No merged thirty-object source or Boolean mesh retag.
Actual current accepted compiled CAD provider (root42884d31, reused by GPU1059) independently compiled all six source files, bit-compared each complete model to asset.face.geometry, validated original FaceBundle and tessellated original Surface17x17. Existing public GPU packet/prepareMesh accepted all six independently:289vertices/512triangles/10008bytes each;1734vertices/3072triangles/60048bytes aggregate. New1057 closed-shell constructor/validator was NOT invoked. NativeFace is forward only; parent uses source-native reverseSurfaceAxis plus a new forward Face, not an arbitrary reverse flag. First exploratory harness incorrectly treated validateFaceBundle's returned report as a FaceBundle; corrected before this contract. Exact first failure/probe results retained.
Existing GPU PrimitiveScene0/1/2 admits only1..4 entries. Existing #941 Boolean Shell profile requires its own actual Boolean extrusion result and cannot accept1057 incidence. Four-Face subset/two canvases lose the required complete six-source view; changing packet.surfaceId to faceId loses original Surface identity. The minimum GPU gap is a new bounded source-owned six-Face preview with qualified Face/Surface/loop IDs and shared native depth, reusing existing native Face/tessellation/math and mesh shader. Loader1067 owns separate Three incidence inspection and remains independent. No duplicate GPU source-qualified Face scene card was found in all-state search.
GPU selects this card immediately after native1059 completion/handoff. Immutable parent249c0a6c9ec2b665dddd56f80ed0dcad1ff8415f. New isolated clean path/branch claim recorded before code. Root owns new GPU qualified-face-scene.ts and additive index exports; renderer helper owns only new qualified-face-scene-renderer.ts; flow helper owns only new examples/gpu-scene/qualified-face-scene-flow.mjs. Helpers are not alone and preserve edits. No existing renderer/native/compiler/camera/root/lock/manifest/Viewer/editor/Loader/shared changes.
The source-only route uses actual already available compileCadModel/validateFaceBundle/tessellateSurface. New1057 global constructor adoption is required only if a future caller invokes that constructor or claims its result; it is not a universal dependency for these authentic independent sources. Global closed incidence is not rendered authority in this frame. No constructor/target availability is fabricated.
Add prepareQualifiedFaceScene(input:unknown):QualifiedFaceSceneV0, QualifiedFaceSceneError with code INPUT|VERSION|BUDGET|SOURCE|SOURCE_BINDING|NATIVE|GPU|DISPOSED|CONTEXT, types QualifiedFaceSceneV0/QualifiedFaceAssetV0/QualifiedFaceSceneDiagnosticCode. Add mountQualifiedFaceScene(canvas,scene):QualifiedFaceSceneRenderer and class updateScene(scene):UpdateResult, resize(width,height):UpdateResult, getScene():independent frozen full source scene, readPixels():Uint8Array, dispose():idempotent. No caller GLSL/GL handles or new native kernel. Private resolveQualifiedFaceScene returns admitted scene/derived meshes once per candidate.
Exact own frame fields: {format:'barocss-gpu-qualified-face-scene',version:0,profile:'source-qualified-forward-face-v0',sourceVersion:0,basis:'authored-xyz-matches-rhyup',assets,countU,countV,camera,mmToSceneUnit,translate,background}. Exactly2..6 assets; each exact {sourceId,source,face,color}. sourceId ASCII1..48 and unique; preserve original user namespace, not inferred model name. Source max65536 UTF8 bytes/code units each, aggregate<=393216 UTF8 bytes. Original Model0/mm exactly5 geometry objects, one selected original Surface and four original boundary Curves, must Object.is bit-match face.geometry including -0 and ignoring object-key ordering. Public Source0 compile validates actual model; public validateFaceBundle validates original topology/forward orientation/loop/pcurves/native geometry. No fabricated FaceBundle return from its report.
Each complete original Source0/model/Face/Surface/Curve/edge/vertex/loop/pcurve/weights/UV identity is retained in independent owned frame snapshots. Qualified display identity is sourceId+face.topology.face.id, with exact original surfaceId and outerLoopId retained separately. It is not native persistent topology naming or a certified solid. No caller display packet is accepted; derive actual tessellateSurface with explicit countU/countV integers2..17 and revision0. Native packet retains original surfaceId/UV; do not retag it as Face or Shell. Six standalone Face validation reports remain incidenceValid/geometric-not-certified/solid:false; no global Shell/Body/Boolean result is created.
Reflection65536values/depth24/array1536/record16/key128/strings65536; exact plain own enumerable data only, no invoked accessors/symbols/foreign/sparse/cyclic/nonfinite values. Existing Face validator's stricter8192/depth10/per-array/native limits can honestly refuse larger data. Source/model mismatch gets SOURCE_BINDING; source diagnostics/native failures retain actual cause. Fresh copies/freeze preserve original F64 vs private F32/U16 buffers. Direct per candidate<=6compile+6Facevalidate+6tessellate calls; public existing native descendant work is not invented telemetry. At maximum1734samples/3072triangles/7680 native deviation probes; reports remain non-certified.
Mandatory public MeshCamera; native authoredXYZ/mm basis explicit. Every original position converts once by nativeMM*mmToSceneUnit+scene translate, common view/projection. No optional pose/material/layers/picking. Background and normalized opaque RGB colors explicit. Native finite±1e6mm/unitnormal/UV/index/extrema and derived±1000scene/F32triangle rules reuse existing packet/prepareMesh. Legacy primitive capacities/profiles/Boolean/uniform/SDFframe0 remain unchanged.
One shared RGBA8/DEPTH24 candidate, native six draws with same camera, LESS/depth writes/culling off/default lighting. One shared native shader program perframe, one VAO/3buffers per Face. Declared accepted+candidate peak2programs/2temporaryshaders/12VAOs/36buffers/2framebuffer-texture-depth sets. Perframe6*10008=60048 geometry-arraybytes; staging120096, excludes source/F64/JS/driver memory. Framebuffer1..1024axis/<=1048576pixels. All budgets before GL allocation; sequential shader cleanup. Declare bounds, not performance/devicecapacity.
Full source/assets/face/color/camera/conversion candidate is copied/validated/lowered before GL. Own every handle before subsequent allocation; render all Faces before present/accept. Failed live source/native/shader/FBO/late allocation releases only candidate and retains accepted source/topology/pixels/handles. Resize stages preserved scene before changing canvas; live failed presentation restores prior dimensions/target, else reports context unavailable with CPU authority retained. getScene remains readable after loss/dispose; mutating/readPixels methods refuse; remove owned listeners; no automatic context recreation/loseContext. Exclusive caller canvas/context; caller owns DOM. No host document or combined editor Save transaction.
Affected GPU build once with real current unchanged compiled CAD source/Face provider. One owned source/API/file/browser flow uses byte-authentic six saved1057 Source0 assets and Face metadata; no public new closed-shell constructor. Actual original Source0/model/geometry IDs/UV/forward Face and source-qualified duplicate local IDs preserved. Real existing four-entry API must refuse6 before new profile admits full6 on one shared depth target. Inspect actual1734 vertices/3072 triangles/60048-byte F32/U16 uploads, colors/shared depth/camera/resize/copies. Authentic interior F pole edit changes its original Surface/frame/pixels while all other five assets/boundary geometry/IDs remain; global closure is not asserted by GPU.
Save/reopen actual complete scene and original sources/native data through real files; cold public compile/Face validation/tessellation/scene reconstruction checks full bits. Reject wrong profile/version, duplicate/foreign source ref, source/model mismatch, reverse Face flag, getter, oversized grids, unsafe F32 and late buffer null while retaining accepted complete scene/pixels/handles. Live console/GL and double disposal0; same authorized52215, coordinate occupancy and never bypass denial. No parent/producer/globalaccuracy/whole-gallery/performance campaign or certified Solid/native inverse/picking/editor claim.
Root records exact clean parent/head, first failures/repairs/current provider maps/hashes/results/support limits, sends immutable delta once to sole Integration and actual consumers, then chooses next actual need. Native completion, Integration, Viewer/Loader/editor adoption and publication remain separate. No new permanent Work/model/cadence/product push/PR/merge/release; paused629 and original hosts/scenes/inherited render/kiosk/history preserved. Planner receives only new Issue number/title.
Clean249c0a6c9ec2b665dddd56f80ed0dcad1ff8415f→fc0ed9b85683c3c12ac0e6164cff51dc5048d99d; receipt6085106477. Four owned GPU/example files. First build2 rejected nonexistent native.type; source-file-only degreeU/degree guard repaired, build0. Earlier exploratory FaceReport-as-bundle failure also retained. ONE first actual browser/source/API/file flow passed five captures/eleven refusals, oldfour-entry actualrefusal, eighteen actualuploads60048bytes, original1734vertices/3072triangles, authentic F interior edit retaining otherfive/sourceboundary/IDs, commoncamera/resize/copies/latebufferretention, two complete cold scenes/twelve sourcefiles. Getters0, live/runtimeconsoleempty/GL0, seven handle classesdisposed0; observed peaks2programs/2shaders/12VAOs/36buffers/2sets. No runtime/file/flow repair or replay.
Original Source0/Model0/five-object forwardFaceBundle geometry/topology/UV/qualification remain source authority, distinct from derived F32/U16 display. Global1057constructor/validator notinvoked, noBooleanretag/Solid/globalclosure/picking/embedding/devicecapacity/performance/contextrecovery/editor/publicationcertificate. GPUroot53b492638d64400b3ca976a8fe1d2df931b15610a5ef22a9a6b5fcc3790a5847; unchangedCAD42884d31 reused. Scoped clean native receipt/evidence/acceptedPNG retained and immutable handoff sent once to soleIntegration and actual Viewer/RenderEditor consumers. NativeCOMPLETE; later stages not inferred.
Next actual1069 source-only target is its original degree4 worldSource through already available native-uniform-curve-v0. Current CPU admission129samples passes and five paired SurfaceUVdegree3/PlaneUVdegree4 observations reproduce original parentpoints (max local observed residual1.7763568394002505e-15mm). Dual UV remain dimensionless JSON and original parentSource models remain separate; no new1069builder/adoption or adaptiveupgrade presumed. A bounded owned actual display/file example is selected next, with no GPU API delta or duplicate Loader1078 card.
Exact native 249c0a6c→fc0ed9b8 is present in local develop at 711aa374354d0fd98521a19490f67efb9949e5a9; serial merge receipt6085656614. Scope: 2..6 source-qualified forward Faces; Model0 with five original geometry objects, no global Shell/Solid certificate. GPU verified the exact Git merge parent/child/ancestry and easylogic author/committer, without editing Integration or replaying native flows. The immediate Integration notice was received and reconciled by GPU; this records agent receipt, not a human read or new verification.
Code is available for compatible downstream implementation with the preserved native build/actual-flow/first-failure evidence. Current integrated build/browser/Save/Open/recovery checks were NOT run in this merge-first step. Track those required follow-ups under existing #674/Integration queue; do not call them passed, fully adopted by a consumer or published. Real profile/ABI/build failures remain concrete gates. All old native/success/failure/denial receipts remain intact.
Actual CAD native 82bde19d3873de1b35b61e70d7563473e7ab08b0 → e9d59973512b06b7d67ed177487942b8ce6cb8c3 is merged at 5c1d0c9ec84b0a692745c04ab25699092f7a8751; native receipt, merge receipt. GPU inspected the actual public export and complete result assembly. Stage: merged; integration validation pending. Native five-family/27-refusal proof belongs to CAD; no GPU composition run was repeated here.
For a future named Surface authoring consumer, use editSourceSurfaceControlNet({source,sourceVersion:0,surfaceId,edits}) and retain the returned complete result.source and result.model. result.edit.source/model is the child single-Surface proof, not the complete five-object Face asset. Keep sourceId, local topology IDs, original source siblings and native binary64 values separate from newly derived display arrays. The GPU source/Face binding requires the complete candidate model to match candidate face.geometry; replacing only the source or using the child model fails that contract. A matching model alone does not prove unchanged boundary/topology geometry. Revalidate native Face/topology and any required boundary relationship under the actual native provider before adopting the candidate; boundary edits must not silently reuse old boundary Curves or global incidence claims.
Provider limits are not display limits: CAD1071 admits complete Source0 up to262144 UTF8 bytes/code units and1..16 objects, selected degree1..3/grid<=16²/at most16 pole edits/64 scalar patches. GPU1072 still admits exactly one Surface/four Curves per asset, <=65536 UTF8 bytes/code units per source,2..6 assets and countU/countV2..17. These intersections must be checked before a host prepares/commits a display candidate. No GPU cap/profile/schema widening is required by the current receipt.
PLANNED future acceptance for an actual named authoring caller: edit an authentic interior pole, reconstruct the full original candidate Source0/Face geometry, prepare one complete current GPU scene, update transactionally and save only accepted full sources. Reopen in that same consumer and rederive native/display data; invalid source binding, boundary policy, source size or late allocation must retain prior source/frame/draft. This is preparation, not an executed editor flow or a new public GPU edit API. Existing GPU1072 interior-edit proof remains its original source-only proof.
CAD native receipt6097864946 supplies clean immutable 3ab8b83cab24a875381888b4ef8100769539bcdf -> 51821d21c7880edbdb54ad7f2688b848ae56bde2. Actual serial adoption is PENDING at this receipt. Root read the handoff ledger and new result types; three transient helpers checked source/geometry compatibility and actual callers. No producer/build/flow replay or GPU implementation follows from receiving this provider.
The native restricted Body is now genuinely qualified within its declared family: six entire Face images, an injective source-defined boundary and outward material proof, 8 Vertices/12 shared Edges/6 Faces/24 opposing Uses, one Shell and one restricted certified Body. Preserve this new native authority separately from unchanged parent flags, earlier unqualified standalone Faces and derived display meshes. Generic Boolean, arbitrary curved Solid, WorldXYZ membership, manufacturing, global C1/binary64 and graphics/publication remain outside this native receipt. Its FIRST build and ONE native source/API/file/cold flow both exited0; seven families,168 endpoint D0 observations with gap0,45 refusals and getters0 belong to CAD's run, not a GPU run.
Complete native geometry is18 objects across TWO real compiled Source0 Models [16,2], retaining full sources and qualified references. Frame is authored XYZ with no axis conversion, lengths mm, volume mm3, UV dimensionless. Original support coefficients, source IDs, parent-local origins, global topology, stored effective Use order/senses and exact material certificate remain native authority. Do not flip Uses again using Face.orientation. Native2MiB request/result and262144-character source admission are not GPU capacities; largest producer result2044025bytes is a recorded native observation only.
Concrete saved native bodymain mapping was inspected without execution: models[0] uses sourceId bodymain_source0, model.name bodymain_m0,16 objects and4068 source bytes; models[1] uses bodymain_source1/bodymain_m1,2 objects and274 source bytes. All six supports and first ten World Curves belong to the first Model; closure_base_left/right and corresponding global Edges e10/e11 belong to the second. supportRef.sourceModelId and curveRef.sourceModelId resolve Model names, not the separate sourceId strings. Keep that distinction when designing future display identity mapping; two sources sharing a local label do not establish identity. This is one static saved-result observation, not compiler/renderer acceptance.
Existing GPU #1072 source-qualified-forward-face-v0 is a different contract:2..6 independently named forward FaceBundle assets, each exact complete Model0/mm with one Surface/four Curves, five objects and65536-byte source. It validates that complete source/model binding and samples the whole original Surface parameter rectangle. #1301's global trimmed/reverse Faces, cross-model references and models[16,2] cannot be cast into this profile, split by discarding source siblings, renamed to fake namespaces or flattened into an over-cap Model. Whole untrimmed Support preview does not display the restricted Face image or complete material boundary. Existing native-mesh/Boolean profiles also require their own actual native producer/mesh certificate; generic triangle data can carry a declared derived display mesh only after genuine geometry production and explicit native identity mapping, not by relabelling Body data.
No named GPU Body caller was found. Current Loader #1299/#1313/#1314 consume existing #1261 lower-strip certificates and explicitly show whole untrimmed Support; they do not require #1301 or prove clipped Body fill. Native #1319 WorldXYZ classification remains CAD-owned and is not screen/pixel picking. Therefore this is provider/compatibility preparation, not READY display implementation and not a new renderer Issue.
For a future actual caller, settle producer-side restricted Face tessellation/trim coverage and native-to-display mapping first, then full-source/model budgets, fixed topology/resource counts, explicit authored-XYZ-to-scene basis/mm conversion exactly once, normals/winding versus parent Use convention, identity/copies and refused-candidate retention. A future ONE acceptance must use an authentic saved complete Body, all six actual Face images/shared edges and TWO original Models; verify clipped coverage/absence outside the roof, shared depth and original identities, cold full-source reconstruction, malformed/binding and late-resource retention, and normal disposal. This proposed acceptance is UNRUN, with no available new tessellator inferred from source/native qualification. Reuse a matching actual consumer card if one arrives; no forced Body renderer, source cap widening, speculative backlog or policy bypass.
Additional existing-path limits: selected uniform Curve preview requires a complete1..4-object Model, so the16-object first Body Model exceeds that profile. The2-object second Model is only a possible separately named Curve preview; no admission or Body display was run. prepareKernelFacePreview requires genuine native mesh chunks, which #1301 does not output; its part-xyz adapter transforms positions/normals as [x,z,-y]. That route is not automatically compatible with #1301's explicit authored-XYZ/no-axis-conversion frame. Preserve the native frame and freeze an explicit compatible display basis before any future adapter; do not inherit a conversion by choosing an old profile. These are static contract findings, not new runtime refusals or consumer support.
Canonical scoped closure6097913350 now records actual adoption ded1ef68b9877de2b487721bc7228e37daed2865; #1301 is CLOSED. GPU independently read the five native handoff blob IDs and compared all five adopted final blobs exactly, with actual adoption base 957e5b7042d0aa29b58c96889fc132b36f8d4890 and current Root ancestry valid. No inspection failure, native replay, build or GPU flow. The earlier PENDING statements above preserve the state when the native notice arrived; provider code is now merged with integration validation pending. This removes adoption uncertainty but does not make the incompatible GPU1072 forward/Model5 profile or untrimmed Support renderer accept the restricted Body. Named caller, real trim-aware derived mesh and explicit frame/identity contract remain unselected. Current Root #1254/#674, GPU graphics/consumer and publication stay separate.
New CAD #1319 native receipt supplies classifyTranslatedSurfaceStripWorldPoints and its validator at bed16bc6cd124700756a3aee7621218c50aff1e0, parent 4c7b0db85ebe43e745b49b6dc5791191d4194225. Adoption is PENDING at this notice. Its first build and one native flow passed, including 141 actual represented XYZ/mm queries, 64 refusals and getters0. These are CAD results, not a GPU run. The provider retains complete qualified #1287 region authority and false Body flags; it does not borrow #1301 Body qualification. Maximum16 supplied points, complete family files <=1MiB in UTF16 and UTF8, original sources/Models and exact rational or surd witnesses remain provider-owned.
This is membership of the supplied represented World XYZ point, not Surface projection, ray intersection, trim tessellation or a raster hit. Classifying a derived Float32 point does not recover the original native point or certify the corresponding screen pixel. A quadratic inverse witness need not provide a represented native U suitable for an evaluator. Therefore no GPU picker or clipped Face mesh can be inferred from this provider. Source bounds #1320 cover Source1/2 implicit primitives only and do not admit CAD Body or NURBS bounds.
Loader #1321 is PLANNED complete Body ingestion with a distinct profile; filled trimmed display is explicitly excluded. Preserve #1301's complete 2,044,025-byte native result and two Models [16,2], proof/reference identities and signed zero. Serialization clarification records that native demo JSON emits numeric -0 explicitly and JSON.parse retains it; ordinary JSON.stringify loses that sign. CAD has no public Body codec. Loader must freeze its accepted bytes/text/value and serialization policy; GPU must retain full original source/native values separately from any future derived buffers, without assuming a codec or widening an old packet cap.
Future display remains PLANNED: an actual named caller and genuine trim-aware mesh producer must establish six clipped Face images, shared edges, original-to-display identity, explicit authored XYZ/mm-to-scene basis conversion exactly once, normals and retained-candidate policy. Existing #1072 forward/exact-five-object/whole-Support contracts remain incompatible. Do not cast models, discard siblings, retag profiles or double-flip effective Uses. If region membership is later part of the caller's acceptance, compare independently represented original points with #1319 and preserve its exact witness/limits; that complements, and cannot replace, clipped coverage/depth/pixel/lifetime evidence. The future ONE complete Body/source/cold/refusal/display flow remains UNRUN. No new GPU Issue/API, producer replay, graphics retry or handoff follows from these availability notices.
Scoped closure records sole Integration adoption ed8c545b76480aff2c7b37a594856cea576f64a9, actual base dd6f3a1263d1dff4ef247d4316bbb94f3d7094ac. GPU independently compared the exact leaf/demo blobs and complete base plus exact native index/README suffix: four paths match, current Root ancestry valid. Earlier PENDING text preserves the native-notice stage; the provider is now code merged with integration validation pending. No producer build/flow or GPU runtime ran, and no display/picking/Body caller is inferred.
Actual signature is classifyTranslatedSurfaceStripWorldPoints(volume, queries, {evaluationId, maximumQueries:16}); queries contain queryId and original pointMM, with1..16 supplied points. The existing whole-result validator reconstructs its actual volume/queries/options/proof/report. Preserve the full #1287 volume, exact represented XYZ/mm and false Body flags. Explicit native-query authority does not remove the existing trim-mesh, ray-to-native-point, display-basis or real caller gates. Loader #1321 remains its separate complete Body ingestion preparation, without filled trimmed display.
Read-only adoption checks had no failure. A separate documentation lookup first addressed the Wiki filename from the integration checkout and failed; reading the actual Wiki checkout resolved it. No product/source/verification change followed. GPU's static evidence is provider-1319-adoption-static.json in its own affine-renderer evidence directory; donor records remain untouched. No repeated availability message, new Issue or handoff.
Loader preparation6098119009 selects isolated parent 2cb2f890d19cfb628c899be10b5b77eec0e25bd4 with actual #1301 adoption ancestry. This is private preparation only: no native profile implementation, product gates or GPU consumer availability. Proposed ./qualified-body text and owned-data entries return independently mutable native data plus validation report, profile and exact sourceText (or null for owned data). The actual native validator returns a report, so the complete copied native candidate must be retained separately. No byte/URL/async/cancellation/serializer/dispose capability is implied.
The selected proposal fixes2MiB UTF8/UTF16 text caps and independent sign-preserving canonical size accounting, with copied occurrences1048576/depth64/record fields64/dense arrays8192/string262144/key128; aliases are copied and charged independently. Strict decoded duplicate-key, Unicode, finite JSON Number and original -0 policies are Loader-owned. Standard JSON conversion is binary64, not exact decimal arithmetic. These are proposed ingestion limits, not GPU resource budgets or an accepted native packet profile. Whole Models [16,2], original source/UV/IDs/authored XYZ/mm/mm3 and complete restricted Body proof remain unchanged. GPU waits only for a real compatible display need/trim producer before adopting any Body display contract; unrelated grammar or source work stays independent.
Loader #1325 now identifies a concrete neutral packet consumer for all six trimmed boundary Faces of #1301. This supersedes the earlier absence of a named packet consumer; it does not establish a GPU renderer caller or an available packet. Loader owns neutral projection and explicit unitsPerMillimeter conversion. Native/display provider ownership and public trimmed-six-Face DTO/schema/version are still unresolved; #1321 ingestion and that producer are real start gates. No CAD kernel ownership transfers to GPU or Loader.
The producer must supply actual bounded vertices/indices/normals/UV, full source-qualified Face/Edge/Use correspondence, shared-edge seams, orientation and approximation/work limits, with retained original output and exact source/native parent/file identity. Whole untrimmed Support, #1316 open Loft and #1319 point membership do not supply that clipped Body boundary. Existing GPU generic geometry/mesh APIs can admit a suitable real derived packet later; a native restricted Body certificate cannot be cast into their data or into #1072's forward five-object profile.
Reconcile future GPU budgets only against the actual accepted neutral schema and complete native identities. Preserve two Models[16,2]/all18 geometry objects, original XYZ/mm/mm3/dimensionless UV/-0 and effective Use order, scale/basis exactly once and native versus derived buffers. The one future GPU consumer flow must observe six clipped Face coverage, excluded Support area, declared seams/orientation/common depth, original-source cold reconstruction, malformed/stale/late-resource retention and disposal. That acceptance remains UNRUN. Reuse existing #1325 for the neutral dependency; no duplicate adapter, fake producer, speculative GPU Issue or qualification inferred from planning.
Loader native receipt6098359048 now completes 2cb2f890d19cfb628c899be10b5b77eec0e25bd4 -> 57dec18694252b418765a4a8de308c0864591205, nine owned interop paths and additive ./qualified-body entry. Exact local adoption is PENDING at this receipt. This supersedes the earlier preparation-only state; it is not current integrated runtime or GPU Body availability.
Actual synchronous text/owned-data ingestion retains the whole13-field native Body, two Models[16,2], full material/topology/source refs, original authored XYZ/mm/mm3/dimensionless UV and exact sourceText. Native data and validation report are independently mutable copies; profile is deeply frozen. Source and emitted first flows each passed10 successful admissions/16 genuine refusals/getter0,60 numeric -0 per original and exact original saved bytes. CAD/interop builds/type/lint and selected emitted declarations/maps passed. Source/emitted summary comparison retained180 numeric -0; no product gate failure. Private prerequisite launcher failure remains a separate retained record. These are Loader's own results, not GPU execution.
Raw2MiB UTF8/UTF16 profile/work limits remain distinct from legacy262144 Source0 transport and future GPU resources. No bytes/URL/async/public serializer/renderer/generic Body or Boolean claim is introduced. #1325 neutral trimmed-boundary need still lacks its actual source-qualified six-Face mesh provider, schema/coverage/seam/approximation contract. Native Body ingestion cannot produce that packet by itself. GPU keeps the current source/native versus derived display and explicit conversion boundaries; no new API, donor replay or graphics flow follows from this notice.
Loader #1321 is now CLOSED and code available at 4ccbb8101024a485158b13a3430db37b07d75aab, actual base 1553dc8bd415be7c158743a454701b685c9199b4; closure receipt6098434326. GPU independently compared nine changed paths: seven new native blobs match, README preserves the full current base plus the exact native suffix, and both source and publishConfig export maps add only the actual qualified-body entry. Current Root ancestry is verified. The first inspection failed because it accounted for only the source export map; checking both actual maps corrected the inspector without changing product bytes or rerunning a producer gate. Evidence: provider-1321-adoption-static.json in GPU's affine-renderer evidence directory. Native ingestion availability removes that dependency from #1325; it does not supply a trimmed display packet.
CAD #1327 native 7b4341ea6cc9d7959121dcd42bff99102141f9f1 -> 876f38fbb21729453fdb5c440322b16564c88307 is code available at 7e96b959746a3fee957644fd1b5ccaf79c992c6c; native receipt6098735141. A scoped helper compared the native/adopted construction leaf without executing it. The complete individual two-cut Band has original parents, Model5, cut World Curves/pcurves and effective Use orientation. Model5 count alone does not make it a GPU1072 forward FaceBundle: that profile requires natural rectangular Surface boundaries. Original reverse Uses and cut domains must not be cast or relabelled. Existing tessellateSurface/prepareMesh can derive the whole Support only; primitive scenes cap at four entries, so Support plus four boundary Curves is not one admitted primitive frame.
Loader #1342 is the actual individual Band consumer. Its existing Three/nativeCapacityV1 route retains complete native proof and labels the whole Support and four boundaries explicitly. It excludes filled Band and needs no GPU API. Reuse that ownership and avoid a duplicate GPU Support viewer.
CAD #1328 native ed8c545b76480aff2c7b37a594856cea576f64a9 -> d0f84deb63fc087838abe20d1c6e6d15456c6cc8 is code available at ed5084bcb443e4844524b21bb3ddd315366b2c97; native receipt6098802571. The public construction and private adapter native/adopted leaves match in the scoped helper inspection. It returns 0..2 complete restricted common-map Bodies, not meshes. Preserve whole parent/native proofs, authored XYZ/mm/mm3/dimensionless UV and source-local/global identities. New Boolean wrapper admission is 6MiB request/12MiB result; individual Body2MiB is unchanged. The old qualified-body reader does not admit the wrapper. Existing extrusion Shell/integer-axis-box buffers validate different native families and cannot consume it by retagging.
Next GPU display preparation remains PLANNED under existing #1325: require a real public trimmed-six-Face provider with exact vertices/indices/normals/UV, effective orientation, seam policy, approximation/work budgets and complete native source/Face/Edge/Use correspondence. Freeze the actual packet before GPU admission or a new bounded owned consumer Issue. Future ONE flow uses that genuine output, verifies clipped coverage/excluded Support/shared depth and full original-source cold reconstruction, then malformed/late-resource retention and live disposal. No producer/build/browser replay, GPU execution, renderer readiness or publication follows from these static adoption checks.