Eve library

Phase 156 local completion evidence — 2026-07-15

Tracker

Revision 1

On this page

Historical task source. The Eve task board holds current task state.

This audit re-evaluated every unchecked Phase 156 item against the current worktree and host instead of accepting the older “external seam” annotations. Four items were locally actionable and are now implemented. The three remaining unchecked items require physical tablet or headset hardware and cannot be truthfully certified by a headless substitute.

Locally actionable items completed#

Item Authoritative implementation evidence Automated proof
156.8.1 OpenXR controller strokes gpen-vr/src/runtime.rs registers and activates pose/float actions, binds standard hand paths, locates the aim action space, forwards timestamped pressure samples, and releases on tracking loss runtime::tests::registers_standard_openxr_actions_and_captures_a_stroke; runtime::tests::tracking_loss_releases_an_in_progress_stroke
156.8.2 Phase-166 AR placement placement::anchor_from_runtime consumes resolved, actively tracked XR_EXT_spatial_anchor entities and converts runtime Pose6Dof state to GP placement matrices placement::tests::phase_166_runtime_anchor_places_and_updates_a_stroke; unresolved-anchor negative test
156.10.3 Blender GP import gpen-testing/src/blend_import.rs uses embedded DNA/SDNA and imports legacy bGPdata plus Blender 4.3+ GreasePencil/CurvesGeometry, including layers, frames, stroke ranges, material/cyclic state, and point attributes Native imports of committed Blender 4.0.2 and 4.5.4 LTS .blend fixtures; invalid-file negative test
156.10.5 Blender line-art visual diff gpen-testing/src/visual_diff.rs rasterizes Neith contours independently and compares them with the Blender-rendered golden using strict IoU and bidirectional tolerance coverage Blender-reference positive gate plus displaced-mask negative control

The reproducible fixture generator is libs/neith/gpen-testing/tools/generate_blender_fixtures.py. The committed fixtures were generated with Blender 4.0.2 and Blender 4.5.4 LTS:

  • blender-4.0-cube-lineart.pgm — SHA-256 11ad18574d5e7c312640a3cd5825c0f023f32dda2f799f1ad5cbf652f5b7c691
  • blender-4.0-grease-pencil.blend — SHA-256 3015476b866465974f621a76dbcb704d428163cd473e909c1ef9ba3688dea486
  • blender-4.5-grease-pencil.blend — SHA-256 03a339a6d8217b24810abec002c23fcb001a9a3b8208de007178ca1d3e173a82

Verification#

All applicable targeted gates passed on 2026-07-15:

  • cargo fmt --manifest-path libs/neith/gpen-vr/Cargo.toml -- --check
  • cargo clippy --manifest-path libs/neith/gpen-vr/Cargo.toml --all-targets -- -D warnings
  • cargo test --manifest-path libs/neith/gpen-vr/Cargo.toml — 78 passed
  • cargo build --release --manifest-path libs/neith/gpen-vr/Cargo.toml
  • cargo fmt --manifest-path libs/neith/gpen-testing/Cargo.toml -- --check
  • cargo clippy --manifest-path libs/neith/gpen-testing/Cargo.toml --all-targets -- -D warnings
  • cargo test --manifest-path libs/neith/gpen-testing/Cargo.toml — 19 passed
  • cargo build --release --manifest-path libs/neith/gpen-testing/Cargo.toml
  • git diff --check -- TODOS/phase-156.md libs/neith/gpen-testing libs/neith/gpen-vr

Remaining hardware-only acceptance#

  • 156.10.2 is specifically an end-to-end pressure-latency measurement on a physical tablet and windowing input loop. Synthetic timing would not prove the tablet/driver/display path meets 16 ms.
  • 156.10.7 explicitly requires a device-diverse hand-drawn corpus from low-end tablets through Cintiq Pro hardware.
  • 156.10.8 specifically requires an on-headset OpenXR compositor/render frame-time measurement. The locally actionable tube-build compute path is already covered by perf::tests::volumetric_tube_build_is_fast, but that is not substituted for the 11 ms headset acceptance gate.

No other unchecked Phase 156 item remains locally actionable on this host.