Files
ark-stronghold-protocol/test/ui
sganggs 2e83a6b977 feedback5: Japanese, Korean and Traditional Chinese ship — the official game texts of the JP / KR / TW clients (data/i18n/ja.json 96.5 %, ko.json 97.4 %, zh-TW.json 91.1 % of 22 114 texts, built with build-i18n --lang from ArknightsAssets/ArknightsGamedata) and minimal UI packs public/i18n/ja.json, ko.json, zh-TW.json (a _meta manifest, no UI strings yet): ja / ko fall back to the English UI, zh-TW to the Simplified Chinese UI (no fallback; numbers in 萬 / 億); the owner's decision of 2026-10-07; UI translations welcome from the community; golden: unchanged (client text only)
- The packs: _meta = type lang, the code, its own and English name (日本語 / Japanese, 한국어 / Korean, 繁體中文 /
  Traditional Chinese), version 0.2.0, app >=0.2.0, authors, credits (the official client data, © Hypergryph (/ Yostar
  for JP and KR), via ArknightsAssets/ArknightsGamedata), fallback ["en"] for ja / ko and [] for zh-TW; formatted as
  `node tools/i18n.mjs template` writes them, so a translator runs it on the shipped file and keeps the manifest.
- Built offline from the tables W5AD downloaded and checked today (jp / kr 51.10.0, tw 51.2.0; copied into
  .cache/gamedata-<code>): coverage as W5AD measured; `build-i18n --lang <code> --check` up to date for en, ja, ko,
  zh-TW. `node tools/i18n.mjs check --all --strict`: 0 errors (en complete; ja / ko / zh-TW 0 of 1022 msgids, fallback).
  `node tools/packs.mjs list` lists the four packs with their game texts; `tools/package.mjs --dry-run --list` ships
  data/i18n and public/i18n of all four and the generated packs/index.json.
- With five languages the menu is a list (ui/lang.js SEGMENTED_MAX 4, W5AD's rule): test/ui/i18n.e2e.test.js reads and
  picks either form, expects the shipped languages, and checks each new pack in the browser (its official game text,
  the fallback UI, <html lang>) — SP_E2E=1: 3 / 3; the title page of the five devices (devices.e2e "title page") still
  fits: 5 / 5.
- Docs: docs/I18N.md (intro, §2, §5 "Shipped packs" with the call for UI translations), docs/PLAYING.md (语言),
  DESIGN §25.20 and the §25 intro's decisions of 2026-10-07.

Gate: non-browser suites 5024 / 5024 (test/i18n*, test/package, client-static 358 / 358), npm run golden ok,
test/golden.test.js 9 / 9, lint 0 errors, check:imports ok, typecheck ok.
2026-10-07 04:16:59 +08:00
..
feedback5 WE: the 整备区 rules of PRTS 卫戍协议/帮助 §手牌区 (GitHub #82, second part) — a full regular hand refuses every purchase and reward pick, also one whose copy would complete a merge at once ("当常规手牌区全满无空位时,玩家将无法执行使手牌溢出的操作" — "例如招募/购入等通常情况下会增加手牌的操作"; BWIKI 盟约 says the same with no exception; the reporter checked it in the official game), and a free regular slot pulls the 临时整备区 in at once ("常规手牌区出现空位时自动移入"): PlayerState._fillHandFromTemp runs in every recompute, the temp row right to left (the order it fills, [ASSUMED]) into the hand's free slots right to left — after a sale, a deployment from the hand, a merge that consumed hand copies, an item equipped / destroyed / used, a summon stack removed with its owner. invariants.js: temp holds pieces only while the hand is full. The client's refusal path follows (gameLogic.shopBlockReason: 整备区已满 on every card — a one-line change in a file the R5 split moves). Bots: a full hand never tries a purchase, an affordable third copy of a held pair sheds the weakest hand single first (never one of this prep), resolveTemp decides temp right to left. Tests: test/match/feedback5-hand.test.js; merge / hardening / playtest6-elite-to-board / economy / prep-bench / uiFixes updated to the rule. Docs: META §1 Hand, DESIGN §6, PLAYING §3, research 00-INDEX / 01 (the old "a merge-completing purchase is allowed" corrected). Golden: 3 match scenarios moved (coop2-NORMAL-3, solo-HARD-9-boosted, coop2-ABYSS-10-boosted — bot seats that reached a full hand: the merge-completing purchase refused or a single shed first, temp pieces moving in by themselves); every battle digest unchanged
2026-10-05 18:17:08 +08:00