sganggs 9f36db32e8 feedback7 R22-LOADOUT: 干员调配's module card reads the module's talent changes at the operator's 潜能, as 局内数值 and the battle do; golden: unchanged ...

Grok's review of the branch (low): ModuleInfo read talentChanges from the data's module record — the full-potential one —
so at 潜能1 余's 闲云隐市 still read 「生命上限2%（+0.5%）」 while 局内数值 and the battle use 1.5 % (异客 +10 %（+2 %） vs
+8 %). ModuleInfo takes `potential` and composes the elite at it first (shared/potential.js atPotential; the module's
traitOverride does not depend on potential); the detail passes the operator's setting.

- Test: test/ui/loadout-rows.test.js — 余's module card at none / 潜能6 / 潜能1, 不装备's base 特性, the detail's prop.

2026-10-08 22:20:30 +08:00

..

content

feedback6: merge R21P — every operator at full potential (潜能 6; GitHub #252, PR #255 by @SpiritedAwayCN, the owner's decision of 2026-10-07)

2026-10-07 18:48:38 +08:00

e2e

feedback6 R21B: 联防 plays on the round's battlefield again — its terrain, crates, water, devices and runes, as in 0.1.4 (the owner's decision of 2026-10-07, 「官服保留地形，改回去」; GitHub #244). 0.2.0's escaped-level map (8a436249, §25.6.4) is withdrawn: that map is the placeholder grid all 39 wave template levels carry, tile for tile; only the 11 stage levels have terrain. match/unitePhase.js _uniteOpts fields this.stageId in both combat modes, unite.js uniteStageId is removed, the client draws the 联防 on the board's stage again (screens/game.js: 0.1.4's ownStage / baseStage, no map switch, nothing to restore on leaving), the dev mock likewise. Kept from 0.2.0: own-board 突袭 / 乌尔比安 S3 landings (onOwnBoard, raidTile untouched), the bounty payee, the 联防 BGM, the SETTLE result box. data/stages.json keeps the two kind 'unite' records (no field uses them; engine tests use them as a plain two-halves road); tools/build-data.mjs comments only, an offline rebuild reproduces data/ byte for byte. The 联防 battle input is byte-identical to v0.1.4's in 20 cases (5 stages × 1 / 2 helpers × both combat modes); GitHub #41 item 3 (野鬃 pushing walkers back across the gate onto the other half) is back with v0.1.4's numbers, none off the map. Tests: test/match/feedback5-unite-map.test.js rewritten (6, all fail on the parent), test/e2e/coop.e2e.mjs checks the 联防 field's drawn stage; docs: DATA §12, META §4, PLAYING §5, design/geometry-time, I18N, the §25.6.4 pointer; golden: fields unite-1 (escaped_single → act1autochess_m02) and unite-2 (escaped_multi → act1autochess_m07), the stages of 0.1.4, and the co-op matches whose 联防 rounds come out differently on the round's map — coop2-NORMAL-3, coop2-HARD-4, coop3-ABYSS-5, coop4-ABYSS-7, coop2-NORMAL-8-serverrun, coop2-NORMAL-14-standins (now a victory)

2026-10-07 17:40:03 +08:00

fixtures

feedback1 WB: bounty drafts and the 机密商店 from 66 official screenshots (22 matches)

2026-10-03 07:42:20 +08:00

golden

feedback7 R22-LOADOUT: the server takes each player's 潜能 / 练度 per operator and every owned operator fights at them (default 潜能 6, 精英2 Lv.60); golden: roster 49/49, bonds 46/46, fields 22/22, matches 18/18, diy 134/138 moved (+10 % 练度), standins unchanged

2026-10-08 20:52:30 +08:00

helpers

feedback5 WW: split DESIGN.md into docs/design/ (the current rules) and docs/history/ (the per-release revisions), section numbers kept — a pure move: every ## N. block (heading, text, the --- after it) moves verbatim, each part file only gains a title line and a pointer to the index; docs/DESIGN.md keeps its header and becomes the index (a § → file table). design/overview.md §0 §1 §2 §11 §12, geometry-time.md §3 §4, engine.md §5 §7, match.md §6 §16, network.md §8 §14, client.md §9 §10 §13 §15; history/v2.2-playtest3.md §17, v2.3-playtest4.md §18, v2.4-playtest5.md §19, v2.5-playtest6.md §20, 0.1.1.md §21, 0.1.2.md §22, 0.1.3.md §23, 0.1.4.md §24, 0.2.0.md §25. Round trip: the index header plus the blocks in § order reproduce the pre-split DESIGN.md byte for byte (813 785 bytes, sha256 fe81884c…). Tests: test/helpers/designDocs.js designText() (the split script's splitBlocks / designText) reassembles the text in § order; docs-consistency and feedback1-gaps read it instead of docs/DESIGN.md (every pin unchanged); docs-paths guards the split (each § exactly once in docs/design/ or docs/history/, the index row of every file names its sections, every "DESIGN §N(.M)" cited in the repository is a heading — 779 in 327 files today); package.test keeps docs/design/ and docs/history/ out of the zips; golden unchanged

2026-10-06 19:08:44 +08:00

match

feedback7 R22-LOADOUT: the server takes each player's 潜能 / 练度 per operator and every owned operator fights at them (default 潜能 6, 精英2 Lv.60); golden: roster 49/49, bonds 46/46, fields 22/22, matches 18/18, diy 134/138 moved (+10 % 练度), standins unchanged

2026-10-08 20:52:30 +08:00

render

feedback6: the 联防 down-ring render test reads 德克萨斯's redeploy time and cost from the data — at full potential 60 s and 11 (it hard-coded 70 s / 13; an opt-in RENDER_E2E suite, so the R21P stream's broad run did not reach it)

2026-10-07 20:04:10 +08:00

sim

feedback7 R22-LOADOUT: the battle takes each operator's potential and 练度 from its spec entry; golden: unchanged

2026-10-08 20:38:08 +08:00

ui

feedback7 R22-LOADOUT: 干员调配's module card reads the module's talent changes at the operator's 潜能, as 局内数值 and the battle do; golden: unchanged

2026-10-08 22:20:30 +08:00

asset-network.test.js

Asset downloads can use a gh-proxy mirror when chosen by hand (--asset-source / SP_ASSET_SOURCE): one try per URL, the original source as fallback, the mirror off for the run after 3 failures in a row (#24)

2026-10-06 14:18:30 +08:00

assets-diy.test.js

feedback5 WU: module type icons get one lower-case file per type (dec-X and dec-x were two names for one file)

2026-10-06 17:59:13 +08:00

assets.test.js

feedback6 R21F: enemy animation states, the client part of PR #275 by @xcdoge re-implemented — the resolver gives an enemy every numbered skill clip (anims.skills; the caller's first index stays the primary, operators unchanged, 宴's PREFAB_SPINE_ROLES fix kept), the stun spellings Stun_1 and Dizzy_Begin/Loop/End, a separate run role (Run_Begin/Loop/End); the renderer walks an enemy faster than moveSpeed 1 on its Run cycle (猎狗pro, 深池侦察犬 ×2; [ASSUMED] threshold), swaps an enemy's skill clip on a cast event (UnitView.setSkillSlot; a form in force is kept), and plays the 重生 of 巨大的丑东西 (Revive, then the fleeing 大祭司 Idle_2 / Move_2 / Stun_2) and 自在 (Revive_01); data/assets.json regenerated offline: 18 enemies gain anims.skills, 6 anims.run, 3 a stun clip (巨大的丑东西, 斩胄之剑, 破胄之锤), nothing else; the sim's cast slot (a new event) is not part of it; golden: unchanged

2026-10-07 17:59:11 +08:00

backups.test.js

feedback6 R21P: every operator fights at full potential (潜能 6) — the owner's decision of 2026-10-07 (GitHub #252, PR #255; no host slider): tools/build-data.mjs OPERATOR_POTENTIAL adds every character_table potentialRanks attribute step (部署费用 / 再部署时间 / ATK / HP / DEF / RES / ASPD, all ADDITION, before rounding) to every chess (normal and elite), 补位 and 自选 unit form and picks talent / trait / module-talent candidates up to requiredPotentialRank 5 (「天赋效果增强」); a summon picks its talent candidates at its owner's potential (their requiredPotentialRank mirrors the owner's — PRTS 凯尔希: Mon3tr 3.5 s / 1400) and takes none of the owner's attribute steps; the 原型干员 stand-ins have no potential ranks (unchanged). 刺玫 413 / 17 → 435 / 15 as in the tournament video, 砾 18 → 16 s, 至简 80 → 70 s. data/chess.json (256 records), backups.json (the 71 owned picks, 8 summons), tokens.json (5 summons) and the data translations en / ja / ko / zh-TW rebuilt — only potential-driven fields (JSON-compared); the 干员调配 局内数值 and detail cards read the records. op-wang.js reads her stock cap 「最多拥有8（+1）枚」 through the potential step (the pattern fell back to 7: she sat at the cap from the start). Docs: DATA §2 / §18 / tokens, PLAYING (detail card), research 03, kits README, shared/diy.js / standIn.js, the 71 自选 kit headers and the kit comments that quoted potential-0 numbers. Tests: test/full-potential.test.js (new, fails on the parent); data.test.js re-derivation adds the potential steps (+ respawnTime); every test that hard-coded a potential-0 number carries the full-potential one (from the data and the test's own formula; no check loosened); golden: moves almost everywhere — roster 49/49, bonds 46/46, fields 22/22, matches 18/18, diy 133/138 (each battle parts on a stat: roster-001 / roster-020 at tick 1 on 角峰's HP (生命上限+250), bond-arcaneShip-high at 4.73 s and diy-001 at 3.70 s on an enemy's HP from a higher ATK; diy-110 / 111 also by 望's cap; the matches through those battles); unchanged: standins 10/10 and diy-135…138 (no potential ranks), diy-012 (twelve 夜莺 never hit, never redeployed)

2026-10-07 18:35:39 +08:00

build.test.js

feedback5 R1: docs and comments follow the split — DESIGN §2 layout lists server/http/, the README table names the entry and http/, path comments in shared/media.js, public/js/media.js, buildGuard.js, net.js and three tests point at the new modules; .git-blame-ignore-revs lists the pure-move commit

2026-10-05 16:11:30 +08:00

check-imports.test.js

feedback5 R6: check that sim and match stay off the net and the client

2026-10-05 16:20:55 +08:00

client-static.test.js

feedback5: merge feedback4 (3af5b84) — PRs #186 / #134 / #185 / #24 / #183 ported to the split layout (terrainInfo / TERRAIN_TIPS into a new ui/gameLogic/terrain.js re-exported by the gameLogic.js barrel, closesOnFieldPress into ui/gameLogic/shortcuts.js, the rest in place; 42 strings through t(), #183's 设置 too); golden unchanged

2026-10-06 14:41:40 +08:00

data.test.js

feedback7 R22-LOADOUT: the data carries every potential (潜能 1–6), not only full potential; golden: unchanged

2026-10-08 20:29:06 +08:00

diy.test.js

feedback5 WE2: 自选 rosters refuse the 集成战略 SO-A / SO-B and the 生息演算 RA-A modules like ISW-A (#25, #10)

2026-10-06 15:18:25 +08:00

docs-consistency.test.js

feedback5 CB1: a 复活 (埃芒加德's 命结之秘, M3茧甲) is a knock-out and an immediate free redeploy — 砾's and 瑕光's 被击倒时 / 部署时 layers fire on every revive (community report 「卡西米尔和艾芒加德策略似乎有BUG：像砾和瑕光这种死亡和部署的叠层效果…无法触发」, items 9 / 37)

2026-10-07 01:56:43 +08:00

docs-paths.test.js

feedback5 WW: split DESIGN.md into docs/design/ (the current rules) and docs/history/ (the per-release revisions), section numbers kept — a pure move: every ## N. block (heading, text, the --- after it) moves verbatim, each part file only gains a title line and a pointer to the index; docs/DESIGN.md keeps its header and becomes the index (a § → file table). design/overview.md §0 §1 §2 §11 §12, geometry-time.md §3 §4, engine.md §5 §7, match.md §6 §16, network.md §8 §14, client.md §9 §10 §13 §15; history/v2.2-playtest3.md §17, v2.3-playtest4.md §18, v2.4-playtest5.md §19, v2.5-playtest6.md §20, 0.1.1.md §21, 0.1.2.md §22, 0.1.3.md §23, 0.1.4.md §24, 0.2.0.md §25. Round trip: the index header plus the blocks in § order reproduce the pre-split DESIGN.md byte for byte (813 785 bytes, sha256 fe81884c…). Tests: test/helpers/designDocs.js designText() (the split script's splitBlocks / designText) reassembles the text in § order; docs-consistency and feedback1-gaps read it instead of docs/DESIGN.md (every pin unchanged); docs-paths guards the split (each § exactly once in docs/design/ or docs/history/, the index row of every file names its sections, every "DESIGN §N(.M)" cited in the repository is a heading — 779 in 327 files today); package.test keeps docs/design/ and docs/history/ out of the zips; golden unchanged

2026-10-06 19:08:44 +08:00

doctor.test.js

启动器「发给朋友」不再把代理 TUN 当公网 IP；删掉永不生效的 ^tun\d

2026-10-03 15:41:57 +08:00

feedback1d-models.test.js

灼热 / 炽焰源石虫 keep their tinted alias: notes and a test on why no PRTS model download is used (#74)

2026-10-05 22:34:54 +08:00

feedback5-operator-sfx.test.js

feedback5 CB3: 银灰's normal attack no longer plays his S3 swing (and S2 / S3 impact); 雷蛇, 初雪 and 圣约送葬人 lose a skill mode's impact the same way (community report of 2026-10-06, item 54)

2026-10-07 02:38:07 +08:00

feedback5-utage-anims.test.js

feedback5: 宴 stands during 落地斩·破门 — her one unnumbered skill set (Skill_Start / Loop / End) is S1 分神's sit-down rest, and the by-name resolver gave it to S2 too, so she knelt through the opening seconds of every battle (the owner's playtest of 2026-10-07); tools/assets/spine.mjs PREFAB_SPINE_ROLES now also fixes operator models by character id (front and back): 宴's S2 plays her ordinary clips (via 'attack'), S1 keeps the rest; data/assets.json rebuilt offline (only her entries changed); test/feedback5-utage-anims.test.js; CHANGELOG 界面, §25.22.9; golden unchanged (client art only)

2026-10-07 13:28:12 +08:00

fetch-assets-shrink.test.js

Operator battle voices: the official combat lines play by voice slot during battles, one at a time (#73)

2026-10-06 01:44:46 +08:00

full-potential.test.js

feedback7 R22-LOADOUT: the data carries every potential (潜能 1–6), not only full potential; golden: unchanged

2026-10-08 20:29:06 +08:00

gitignore.test.js

feedback6 R21E: .gitignore lists .npm-cache/ — a local npm cache (npm_config_cache=./.npm-cache, npm ci --cache .npm-cache) is never committed by any contributor (GitHub #258); test/gitignore.test.js asks git itself; golden: unchanged

2026-10-07 18:03:54 +08:00

golden-diy.test.js

feedback5 WK: golden — one 自选 match (coop2-NORMAL-70-diy): a human seat on AI 托管 slots 推进之王 (S3, SOL-X) and the prototypes Sharp (tiers 5 and 6) and 领主·Sharp, its 调度中心 at level 5 from the first prep; its own shop draws 推进之王 in round 3 (and Sharp in round 7), its AI fields 推进之王 — the operator kit — in rounds 3–7; the digest's diy lists per round the 自选 shop draws and the fielded pieces, and the stock left

2026-10-06 00:16:47 +08:00

golden-standins.test.js

feedback5 WG: golden — a standins family (10 battles: every NORMAL chess record, normal and elite, fielded as its 补位 stand-in on real stages and waves plus ground enemies; all 17 stand-ins, every active backup skill cast) and one co-op match whose AI-托管 human seat does not own 7 chess (coop2-NORMAL-14-standins); the 133 existing digests unchanged (roster / bonds / fields byte-identical, the 16 old matches identical)

2026-10-05 19:42:49 +08:00

golden.test.js

feedback5: merge WE1 — fidelity (sim rules): 秘术师 store at the attack check (深靛; 维伊 / 黑键 plug into it), enemy attacks land at their damage frame so a short stun / freeze / sleep / 浮空 / 缴械 / 恐惧 before it cancels the swing, 武者 heal per damage instance through 禁疗, 庇护 the strongest only, PHY-X 不高于 50 %, 链愈师 jumps in the 3×3, a summon readied at its owner's redeploy, Misery's duration skills, 伺夜 S2 in 战术点形态, centurions hit up to block-count targets; merge fixes: 黑键's kit keeps the up() import WE2's S2 cast needs (WE1 dropped it with its tick store — the first regeneration had stored 9288 swallowed errors in diy-114), the 自选 refusal names 集成战略 / 生息演算, tools/golden.mjs prints a missing value, test/golden.test.js fails on any stored engine error; golden: vs WE1 only the W4B / WE2 scenarios differ (roster-042, bond-egirShip-high, coop2-ABYSS-10-boosted, coop2-NORMAL-70-diy, diy-045 / 078 / 114 / 115 / 116)

2026-10-06 16:52:11 +08:00

i18n-data.test.js

feedback6 R21P: every operator fights at full potential (潜能 6) — the owner's decision of 2026-10-07 (GitHub #252, PR #255; no host slider): tools/build-data.mjs OPERATOR_POTENTIAL adds every character_table potentialRanks attribute step (部署费用 / 再部署时间 / ATK / HP / DEF / RES / ASPD, all ADDITION, before rounding) to every chess (normal and elite), 补位 and 自选 unit form and picks talent / trait / module-talent candidates up to requiredPotentialRank 5 (「天赋效果增强」); a summon picks its talent candidates at its owner's potential (their requiredPotentialRank mirrors the owner's — PRTS 凯尔希: Mon3tr 3.5 s / 1400) and takes none of the owner's attribute steps; the 原型干员 stand-ins have no potential ranks (unchanged). 刺玫 413 / 17 → 435 / 15 as in the tournament video, 砾 18 → 16 s, 至简 80 → 70 s. data/chess.json (256 records), backups.json (the 71 owned picks, 8 summons), tokens.json (5 summons) and the data translations en / ja / ko / zh-TW rebuilt — only potential-driven fields (JSON-compared); the 干员调配 局内数值 and detail cards read the records. op-wang.js reads her stock cap 「最多拥有8（+1）枚」 through the potential step (the pattern fell back to 7: she sat at the cap from the start). Docs: DATA §2 / §18 / tokens, PLAYING (detail card), research 03, kits README, shared/diy.js / standIn.js, the 71 自选 kit headers and the kit comments that quoted potential-0 numbers. Tests: test/full-potential.test.js (new, fails on the parent); data.test.js re-derivation adds the potential steps (+ respawnTime); every test that hard-coded a potential-0 number carries the full-potential one (from the data and the test's own formula; no check loosened); golden: moves almost everywhere — roster 49/49, bonds 46/46, fields 22/22, matches 18/18, diy 133/138 (each battle parts on a stat: roster-001 / roster-020 at tick 1 on 角峰's HP (生命上限+250), bond-arcaneShip-high at 4.73 s and diy-001 at 3.70 s on an enemy's HP from a higher ATK; diy-110 / 111 also by 望's cap; the matches through those battles); unchanged: standins 10/10 and diy-135…138 (no potential ranks), diy-012 (twelve 夜莺 never hit, never redeployed)

2026-10-07 18:35:39 +08:00

i18n-packs.test.js

feedback6: the i18n template test reads the app range from APP_VERSION like tools/i18n.mjs appRange (it hard-coded >=0.2.0; 0.2.1-dev writes >=0.2.1)

2026-10-07 16:51:35 +08:00

i18n-server.test.js

feedback5 W5S: i18n phase 2, match texts the server sends — a 机变 card's and an effect entry's name and description, the draft's name, a reward offer's label and the result title are shown from the localized data when the server's Chinese text is the data record's own (data.lookupRaw + ui/lang.js sentText; names through tName), a text the server reworded (a multi-round 悬赏) stays as sent; Chinese output unchanged by construction (in Chinese the localized record is the raw one); test/i18n-server.test.js covers it

2026-10-06 18:27:13 +08:00

i18n.test.js

feedback5 W5S: i18n phase 2 coverage — test/i18n.test.js scans the whole client, shared and the server messages (no Chinese literal unwrapped, every msgid translated: extract 0 and check --strict as a test; floors 1 000 catalog entries, 1 350 msgid uses), public/i18n/en.json drops the 64 fragments / renamed msgids the sweep replaced (none is a data text), docs/I18N.md: phase-2 coverage, what stays Chinese and why, the server's game texts, the audit's skip rules

2026-10-06 18:31:28 +08:00

lobby-diy.test.js

feedback5 WK: 自选编队 on the server — room.diy { picks } (lenient: illegal picks dropped, only malformed input is BAD_MSG; stored per session / seat; a match takes the picks its seat had at the start, a change during a match is for the next one; welcome lists diyKitted), PlayerState.diy with the player's data view (a slotted DIY slot is its operator for every rule: bonds from its factions, no 特质, the pick's skill and module, the slot's tier / price / merge), a per-player stock per slotted piece (8 at tier 5, 5 at tier 6 [ASSUMED]; never in the shared pool; none when every bond of it is banned), DIY pieces in that player's shop rolls and reward offers once the 调度中心 reaches the slot's level, battleInput diy, m.private.diy / diyBanned, prep scouting and results name the operator, 信标 never sends a 自选 piece to a teammate [ASSUMED]

2026-10-05 23:35:01 +08:00

lobby-kick.test.js

feedback3 W3C: room.kick names the confirmed player; a seat that changed hands while the host confirmed is refused (BAD_TARGET)

2026-10-04 17:38:29 +08:00

lobby-loadout.test.js

feedback7 R22-LOADOUT: the server takes each player's 潜能 / 练度 per operator and every owned operator fights at them (default 潜能 6, 精英2 Lv.60); golden: roster 49/49, bonds 46/46, fields 22/22, matches 18/18, diy 134/138 moved (+10 % 练度), standins unchanged

2026-10-08 20:52:30 +08:00

lobby-ownership.test.js

feedback5 WG: 补位 on the server — room.ownership { notOwned } (lenient check, stored per session / seat, a match takes the list its seat had at the start), PlayerState.standIns: the player's not-owned NORMAL chess fight as their stand-ins (battleInput standIn: true in every field, the stand-in's position / range, no summons), m.private.standIns, stand-in art in prep scouting

2026-10-05 19:13:46 +08:00

lobby.test.js

feedback3 WE: spectator seats in the browser — 观战 in the lobby, the 观战席 strip in the room (host removes, 入座), watch-only match screen; docs (community report #26)

2026-10-04 17:39:30 +08:00

local-extract.test.js

feedback5 WX: the 39 summon models no web dump carries come from the local client (extract.py TOKEN_SPINES) and are drawn instead of the avatar diamond when data/local-assets.json lists them

2026-10-06 20:16:20 +08:00

local-token-models.test.js

feedback5 WX: the 39 summon models no web dump carries come from the local client (extract.py TOKEN_SPINES) and are drawn instead of the avatar diamond when data/local-assets.json lists them

2026-10-06 20:16:20 +08:00

media-url.test.js

按 review 意见：/media/ 的路径与扩展名列表提到 shared/ 共用

2026-10-03 22:33:20 +08:00

open-browser.test.js

打开浏览器改用 explorer.exe（shell 关联），找不到时才用 rundll32 (#29)

2026-10-03 23:05:06 +08:00

package.test.js

fb5-patch-pkg: every release zip carries MANIFEST.json (size + sha256 of each shipped non-art file, written after npm ci; the full, lite and update zips of a version list the same), and node tools/package.mjs --update --from <base>[,<base>…] builds Stronghold-Protocol-v<version>-update.zip — the full stage built as before, compared file by file with every base (an earlier full zip or its extracted folder; tools/package-update.mjs reads zips itself, zip64 included), shipping each file that is new or differs from at least one base plus MANIFEST.json and UPDATE.json (from, count, bytes, files, removed with each base's bytes); bases must be older, full (not lite or update) and distinct; the update stage gets the refusal and personal-info checks (sha256 values left out of the name scan); test/update-package.test.js (8), test/package.test.js expects MANIFEST.json in the zip

2026-10-07 11:23:55 +08:00

potential.test.js

feedback7 R22-LOADOUT: the data carries every potential (潜能 1–6), not only full potential; golden: unchanged

2026-10-08 20:29:06 +08:00

static-local-art.test.js

feedback5 WX: the 39 summon models no web dump carries come from the local client (extract.py TOKEN_SPINES) and are drawn instead of the avatar diamond when data/local-assets.json lists them

2026-10-06 20:16:20 +08:00

update-boot.test.js

fb5-patch-pkg: the update package's boot step and install check (server/update.js) — before the server starts (runMain: npm start, launch.mjs, the Windows service's run-server.cmd, NSSM, systemd; launch.mjs also before setup) an extracted UPDATE.json is checked: malformed → logged and skipped; the install verified against MANIFEST.json, and a missing or different file the server or browser runs stops the start with the bilingual "download the full package" message (nothing deleted, UPDATE.json kept); else each removed file still holding a base's bytes is deleted (never a changed file, a player's pack, a symlink, a per-machine name or a path outside the install) and UPDATE.json becomes .update-applied.json; doctor gets a MANIFEST.json row; test/update-boot.test.js (10)

2026-10-07 11:23:42 +08:00

update-package.test.js

修复：Windows 上打包和读取 zip 时中文文件名出错（master 的 Windows CI 失败） (#311)

2026-10-08 17:45:45 +08:00

version.test.js

feedback5: the public dev branch is marked as a development build (the owner's decision of 2026-10-06): version 0.2.0-dev (title footer, boot banner, /healthz), a README warning under the title (unstable, not for public servers, how to report, shared browser storage, no release zips), a visible 开发版 tag on the title screen (zh / en), a boot-banner warning for hosts, and tools/package.mjs refuses a dev version without --allow-dev; test/version.test.js checks all of it

2026-10-06 19:05:40 +08:00

windows-bundle.test.js

Windows 便携包：只收 git 跟踪的文件、生产依赖、钉死的 Node 版本与许可证 (#30)

2026-10-04 14:20:11 +08:00