20 lines
3.5 KiB
Markdown
20 lines
3.5 KiB
Markdown
# Saved verification checkpoint — 2026-09-12
|
|
|
|
This document preserves verification and saved source identities, not live task status. The user explicitly directed continued work and removed misleading stop notices from Gitea. No usage-limit stop instruction applies. Live coordination remains Gitea issues #3 and #15. The separate Linux/WSL networking deferral remains in effect.
|
|
|
|
## Saved allocation work
|
|
|
|
Root branch `codex/behavior-authoring-20260910`, published as `codex/release1-956atf`. Source commit `3840ca29652e3128ffbcdebd9f39f364e56079ea`; subsequent `f1bd5149db0f637fa09d4dcf93ebd2f3cf221593` changes only decision documentation. The fixed toolbar destination identity, frozen canvas target outline and gesture-local full-name inspection are implemented, with actual mounted minimum-window and tooltip-hover tests. `PROJECT_GOAL.md` user input was preserved.
|
|
|
|
Full Windows command: `cargo test --locked -p oa-desktop --bin arcadia-studio -- --test-threads=1`, using `.local/toolchain/env.ps1` and `.local/run-allocation-gate.py`. Run started 11:46:24 UTC and finished 11:52:31 UTC, exit 0: **796 passed, 0 failed, 0 ignored**. All 526 identified inputs matched source 3840ca2 and remained unchanged. Input manifest SHA-256: `60892c37f9a288bbd58d93c691c90b68b58bac2a91baaa9153cb2e5a85cbd538`. Strict Clippy, packaged GUI and native acceptance have not run for this source.
|
|
|
|
Evidence: [archive](experiments/allocation-3840ca2-pause-20260912/archive-manifest.json) contains exact full-run log/input/result receipts, two earlier failed focused-run receipts, source review and worker handoff, and six losslessly encoded actual renderer PNGs with metadata. Optional render capture used `ARCADIA_RENDER_EVIDENCE_DIR=.local/allocation-render-3840ca2`; original RGBA remains local. Earlier failures were test-fixture snap-tie and unsupported tooltip geometry-inspection assumptions, corrected before the full passing run. They are retained to avoid repeating those approaches.
|
|
|
|
Visual inspection of OA minimum normal/top, OA full-name hover, and SA minimum top frames found the toolbar identity readable and the moving work name unobscured by opaque target labels. A remaining feedback defect was observed: during held-pointer hover outside the canvas, the status/footer says “Work drag canceled outside the canvas” although the gesture remains live and can return to the canvas. On resume, distinguish preview-outside wording from actual release cancellation. Do not claim full native usability acceptance from these headless frames.
|
|
|
|
## Saved separate Delete candidate
|
|
|
|
Branch `codex/diagram-delete-956atf-20260912`, isolated worktree `.local/diagram-delete-20260912`, commit `7dcfb3cb107a8f2fc28664fe5b7ad9cda16f4dda`, based on 5a70181. Thirteen code files plus its durable `harness/diagram-delete-candidate-20260912.md` report are committed. Architecture/state canvas focus, Delete placement removal, captured model confirmation and ten mounted test functions are authored. Formatting checks passed; compilation and tests **have not run**. This is an unintegrated WIP, not a deliverable or completed issue #15.
|
|
|
|
The candidate report identifies review boundaries. Preserve the newer root allocation toolbar when merging the older-base candidate's `views.rs`; do not copy that file wholesale. Review context freshness, focus, confirmation and allocation interaction; process/scenario/explorer removal gaps remain explicit. Serialize compiler runs. Finish appropriate source-specific checks and coordinated integration before claiming acceptance or publishing a release.
|