skara-brae-common/docs/openup/test/test-report-wi-2-architecture-cleanup.md

# WI-2 Test Report — Architecture Cleanup + OpenUP Work Set (skara-brae-server#2, PSP1)

Date: 2026-10-03 · Phase: CMP + TST · Executor: agent (psp-tester session, Kilo/glm-5.3)

## Scope under verification

- **Code** — `skara-brae-server`, branch `wi-2-architecture-cleanup`, 7 commits over base `899fde8`: `337e9bc` fmt baseline, `093012a` constants module, `423997f` DomainError + signature conversions, `6d9f02b` adventure service, `5d5a594` responses module + handler rewiring, `f065942` label helpers + policy constants, `c613142` 2B findings rework. Working tree clean at verification start.
- **Docs** — `skara-brae-common`, branch `wi-2-openup-docs`, 4 commits: `4798f55` WI-2 OpenUP work set move (verbatim), `2343d5e` WI-1 artifact migration, `7a9a5a3` game-canon migration from Joplin, `41f1ec2` EA model preparation. Working tree clean at verification start.

## Verification gates (CMP)

| # | Command | Result | Evidence (summary line) |
|---|---|---|---|
| 1 | `cargo fmt --check` (server) | **pass**, exit 0 | no output — zero formatting diffs |
| 2 | `cargo clippy --all-targets -- -D warnings` (server) | **pass**, exit 0 | `Finished \`dev\` profile [unoptimized + debuginfo] target(s) in 0.11s` — zero warnings emitted |
| 3 | `cargo build --workspace` (server) | **pass**, exit 0 | `Finished \`dev\` profile [unoptimized + debuginfo] target(s) in 2.13s` |
| 4 | Docs-side sanity (common): `git status --porcelain` clean; `python3 scripts/openup/markdown-to-json.py` regenerates the export | **pass**, script exit 0 | `wrote docs/openup/model/exports/uc-realizations.json: 12 realizations, 36 elements`; sha256 before == after: `23c9617f34ae8e92882b49bfc48c990d9c79d975b9756a93b250391dc3aa6f41` — **byte-identical, no drift**; working tree still clean after regeneration |

## Test matrix (TST)

`cargo test --workspace`: **exit 0 — 71 passed, 0 failed, 0 ignored**.

| Target | Count | Summary line |
|---|---|---|
| sb-api (bin unittests) | 6 | `test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out` |
| sb-core (lib unittests) | 63 | `test result: ok. 63 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out` |
| sb-auth / sb-db | 0 | `test result: ok. 0 passed` (both) |
| sb-shared (lib unittests) | 2 | `test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out` |
| Doc-tests (4 targets) | 0 | all `test result: ok. 0 passed` |

Count reconciliation: 45 baseline (WI-1) + 26 new `#[test]` fns (diff vs `899fde8`: 26 added, 0 removed) = 71.

Coverage classes within the 71:

- **Adventure service vectors A1–A8** (8 new, sb-core `adventure.rs`; mapping per `test-cases.md`): A1 `run_adventure_win_applies_rewards_and_loot` (seeded win → xp/gold/nerve/loot/inventory), A2 `run_adventure_defeat_sets_hp_to_zero`, A3 `run_adventure_rejects_incapacitated`, A4 `run_adventure_rejects_insufficient_nerve`, A5 `run_adventure_applies_regen_before_guards`, A6 `run_adventure_charges_spell_upkeep_on_win_and_loss`, A7 `run_adventure_filters_loot_to_catalog_items`, A8 `run_adventure_is_deterministic_under_seeded_rng`. Targeted run: `cargo test -p sb-core run_adventure_ -- --nocapture` → exit 0, `8 passed; 0 failed … 55 filtered out`.
- **Wire-identity goldens W1–W6** (6 new, sb-api `routes/responses.rs`): all six pass under `cargo test -p sb-api wire_identity_ -- --nocapture` → exit 0, `6 passed; 0 failed`. Representative assertions (responses.rs:156/167/237):
  - `assert_eq!(serde_json::to_string(&dto).unwrap(), GOLDEN_W1)` — W1 keys in lexicographic order (`combat_log`, `enemy`, `loot`, `zone`) with nested sorted round/crit keys, pinning the §4.3 `Value` bridge (defect #14).
  - `assert_eq!(serde_json::to_string(&dto).unwrap(), GOLDEN_W2)` — `{"gold":123,"healed":true,"hp":45,"max_hp":80}`.
  - `assert_eq!(serde_json::to_string(&dto).unwrap(), GOLDEN_W6)` — `{"gold":19700,"learned_count":4,"max_spell_points":40,"spell_points":40}` (incl. the Q1 fenced `learned_count` wart, byte-preserved).
- **Baseline 45 + mechanical edits** — the only permitted test edits were the **9 mechanical assertion updates** from design §4.1 / `test-cases.md` §3. Verified in `git diff 899fde8..HEAD`: 3 × `level_up` (`Err("…")` → `Err(DomainError::AlreadyAtMaxLevel | NotEnoughXp | NotEnoughGold)`), 3 × `train_stat` (`NotEnoughEnergy | NotEnoughGold | InvalidStat`), 1 × `equip_item` (`ItemNotInInventory`), 1 × `unequip_item` (`NothingEquippedInSlot`), 1 × `pick_enemy_respects_weights` gains Ok-path `.unwrap()` (service now `Result`-returning). No test logic changes.
- **New table-coverage / data / spot-check tests** (12 new: 10 sb-core, 2 sb-shared): `gym_display_order_covers_all_training_grounds`, `race_table_covers_all_variants`, `class_table_covers_all_variants`, `crit_tables_parse`, `enemies_json_parses`, `zones_json_parses`, `all_items_json_parses`, `caster_spell_level_spot_values`, `spell_learn_cost_guard_spot_checks`, `spells_json_parses`, `item_type_as_str_matches_serde_wire`, `quality_as_str_matches_serde_wire`.

## Wire-invisible deltas — recorded, verified as documented

Both deliberate semantic deltas from design §4.2/§6 are present in the wired code exactly as recorded (1B review finding 5 dispositions), with no reachable wire-visible change:

1. **Lock consolidation** — the refactored adventure handler holds one write-lock span around the service call (`routes/mod.rs:488`); the dead "character vanished between acquisitions" branch is structurally impossible (no character-deletion path exists). Recorded, not claimed away.
2. **Enemy-pick failure** — data-impossible `DomainError::UnknownEncounterEnemy` maps to `StatusCode::INTERNAL_SERVER_ERROR` (`routes/mod.rs:61`) where the pre-refactor code panicked; same in-memory side effects (nerve already spent).

## Verification limits

- **2B finding #3 evidence limit (docs move):** the verbatim-ness of the OpenUP work-set move from untracked `skara-brae-server/docs/openup/` into `skara-brae-common/docs/openup/` is **not cryptographically verifiable** — the source files were never committed, so no git hash chain exists to diff against. Verification is therefore **content sanity only**: exports regenerate byte-identical from the moved markdown sources (CMP gate 4), the `markdown-to-json.py` invariants hold (12 use cases, 36 realization elements), and the tree is clean. W2–W6 goldens were recorded by the coder per the design protocol (§7: serialized from the pre-refactor `json!` expressions with fixed representative values); **W1 was independently reconstructed by the 2B reviewer** and both agree with the shipped bytes.
- **No manual wire spot-check** was executed (design regression-protocol step 4 is "recommended once", not mandatory); byte-identity of the live HTTP surface rests on the W1–W6 golden tests.

## Platform note

The EA pipeline (`scripts/openup/build-model.ps1`, `export-model.ps1`) is **Windows-gated** (PowerShell + EA COM automation); both scripts are committed but were **not executed on this Linux host** — only the pure-Python intermediate step (`markdown-to-json.py`, stdlib-only) ran. The human gate (design §11.d: EA session, `.eapx` commit) is **pending** and outside this verification's scope.

## Defects found in test phase

None. No defect candidates.

## Verdict

**PASS** — all four CMP gates green (fmt, clippy zero-warning, build, docs drift check byte-identical); all 71 tests green across the workspace; the two recorded wire-invisible deltas verified as documented; docs work set internally consistent. Human EA gate outstanding by design.