2343d5e154
work-items.md (project-management/), design-wi-1-injectable-rng.md (analysis-design/), test-report-wi-1-injectable-rng.md (test/) move into the canonical OpenUP home per the 2026-10-03 decision reversal. Verbatim, no content changes; project-top docs/ husk removed.
2.2 KiB
2.2 KiB
WI-1 Test Report — Injectable RNG (skara-brae-server#1, PSP1)
Date: 2026-10-02 · Phase: TST · Executor: agent (PM session, Kilo/glm-5.3)
Scope under test
Branch wi-1-injectable-rng (uncommitted working tree vs base 64e081a):
sb-core/src/character.rs (create_character signature + new determinism test),
sb-api/src/routes/mod.rs (request_rng helper, create handler, run_adventure
consolidation).
Test cases
| # | Test | Purpose | Result |
|---|---|---|---|
| 1 | create_character_is_deterministic_under_seeded_rng (NEW) |
Two StdRng::seed_from_u64(42) runs yield identical stats (Debug-string) and equal hp/max_hp/energy/nerve — proves RNG injectability end-to-end |
pass |
| 2 | Existing sb-core suite (44 tests: character 20, combat 8, services 17 -1 overlap after adding #1) | No regression from signature change / rng threading | pass |
| 3 | Acceptance audit: grep -rn "rand::rng()" crates/ |
Zero occurrences in the whole repo (incl. tests) — acceptance criterion | pass (only remaining entropy: sb-auth/src/hash.rs:8 getrandom::fill, excluded by design) |
| 4 | cargo clippy --all-targets -- -D warnings |
Zero-warning standard | pass (clean) |
| 5 | cargo build --workspace |
Compile integrity | pass |
| 6 | rustfmt --check (touched lines) |
Coding standard on new/rewritten code | pass (repo-wide pre-existing fmt debt documented, not touched — see CMP log) |
Evidence
cargo test --workspace: 45 passed, 0 failed (sb-core lib target; remaining targets empty)- Determinism test present and green:
test character::tests::create_character_is_deterministic_under_seeded_rng ... ok - Full output captured in session log (PSP time log, TST phase, 2026-10-02 22:17)
Defects found in test phase
None. (All defects D1–D11 were found in Design Review / Code / Code Review phases; see defect log.)
Notes
- The new test replaced a whole-struct
Debugcomparison after D8 (HashMap iteration order instability) — comparison now covers the random-dependent state exactly. - No behavioral change expected from ThreadRng→StdRng: same distributions, one OS-entropy seeding per request (design rev 2, D1 correction).