# Interface recovery validation ledger

2026-09-09. Automated implementation evidence; **human AC10 pending**.
Use `cargo +stable`, `--locked`, approved opt1 dev/test environment from
`development-validation.md` (assertions and overflow checks enabled). Worker
commands occasionally omitted explicit +stable/--locked but used the same
inherited stable opt1 target; parent commands below explicitly used both.
Source changed between impact runs. This is complete target coverage plus
focused correction evidence, **not one frozen all-green invocation**.

Metadata auditor generated **14** target-covering partitions in
`target/interface-partitions.json` after adding terminal_color_reference.

| Partition | Evidence/status |
|---|---|
| workspace excluding CLI, all targets | PASS parent; later renderer-only optimization separately passed full render112 tests and972 exact reference pairs |
| CLI bin hexwork | PASS,0 tests |
| CLI routes | PASS,1 test |
| CLI lib | Parent full run258pass/4fail/2ignored; one compiled stale ribbon expectation subsequently fixed; three self-executable spawn ENOENT failures under concurrent relinking. All4 serial focused reruns PASS. New semantic marker test included in separate5/5 production semantic run. No remaining known lib failure; no claim of a fresh monolithic265-pass run |
| contracts_open | PASS10 worker,85.99s; original bounded capture failure repaired through correct physical route/map exit |
| hud_essentials | PASS9 worker,25.10s; optional legacy0 collision/map close and explicit focus CSI normalization |
| phase11_gate | Initial44pass5fail; four repaired eye/feedback cases pass focused,267.15s five-case run had only script parser failure. Added presentation-only slice verbs; last presentation case PASS parent11.21s. Existing hash/history/v4/death durability assertions unchanged |
| player_interface_pty | Parent entire target PASS19,220.29s, includes10 physical scenarios plus decoder/support cases |
| tactical_orientation | PASS11 worker,66.18s final; strict unique player assertion retained, caught and verified real Entry center-rounding fix |
| terminal_color_reference | Parent PASS2, independent VT oracle on actual full/diff encoders |
| tinker_live | PASS11 worker,379.16s; actual modal title synchronization fixes premature Play-hint match; exact25cu/15cu checks intact |
| transfers_lifecycle | PASS10 worker,122.57s |
| transfers_live | PASS11 worker,174.24s; canonical-route physical key conversion explicit |
| wield_live | PASS10 worker,300.61s; unchanged target |

Additional evidence: Python VT7 pass; core light7 parent pass; term62 pass;
App camera/movement9 pass; production semantic5 pass; renderer112 pass.
No tests disabled to bypass failures. Two ignored CLI tests are the pre-existing
transfer timing harness and new explicitly opt-in render benchmark.

Final `cargo +stable fmt --all -- --check` and `git diff --check`: PASS.
Vendor bincode/bincode_derive lifetime warnings remain, no new compile warning in
parent CLI check. The Rust language server failed initialization; compiler/tests
and targeted reads supplied verification.

## Performance

Machine: Linux x86_64, Intel Processor U300,6 logical CPUs; rustc1.98.1
(48a229cea,2026-09-01), LLVM22.1.8.

`cargo +stable test --locked -p hexwork-cli --lib
interface_benchmark::render_trajectory_benchmark -- --exact --ignored --nocapture`
after conservative projected billboard bounds optimization:

| Frame cells | Samples after10 warmups | median | p95 |
|---|---:|---:|---:|
|80×24|100|3.909ms|4.541ms|
|120×36|100|8.640ms|11.161ms|

Includes chase collision, active catalog collection, derived light, descriptors,
render, gamma and full truecolor encode. Excludes App snapshot, terminal flush,
and canonical simulation. Allocations unmeasured for this broader path; known
per-frame BTreeMap/Vec work remains. Meets16.7ms for this trajectory, not a
universal60fps or tighter per-stage budget claim. Concurrent heavy earlier runs
missed the target (p9522/38ms); neither contention attribution nor optimization
speedup is rigorously isolated. Renderer-only allocator benchmark has separate
scope in `../crates/hexwork-render/BENCHMARK-recovery.md`.

## Release and manual boundary

`cargo +stable build --locked --release -p hexwork-cli`: PASS.
Binary: `target/dev-loop-ab-57-opt1/release/hexwork`
SHA-256: `1697b68a8309e4573c44c5e630d4211054fa37503db976a845993efdb674981b`.

The identified release also passed an actual seed140 Story truecolor physical
PTY route: look/right/up/reset/exit, chase, inventory/return, tactical slice and
rejected off-floor movement, Home, too-small/restore/grow/shrink, map exit/save.
18 decoded stages; opt-in JSON/SVG artifacts:
`target/player-interface-release-evidence/`. This is terminal-protocol evidence,
not human font/monitor readability.

**AC10 `manual_normal_brightness_orientation_route` remains pending.** Follow the
README route on this binary and record screenshots and explicit human pass/fail.
No full HEXWORK1.0 completion, accessibility certification, or human usability
acceptance is claimed.

## Explicit remaining matrix limits

The implementation and tests are materially broader than the baseline, but not
an exhaustive Cartesian matrix of every seed/mode/palette/gamma/remap/stance.
Actual catalog service/exit projection tests use real anchors in synthetic clear
geometry, not a walked service/exit route. Color oracle exercises public renderer
palettes and actual ANSI; CLI gamma-specific decoded fixture matrix is incomplete.
Allocation counts are measured only in the narrow renderer bench. Fully blocked
invalid player geometry cannot provide a safe own-column eye; fallback policy is
finite/deterministic and documented. These limits must accompany any acceptance
claim; automated checks cannot replace AC10.