Commit Graph
5 Commits
Author SHA1 Message Date
Cloudnyco ae22335cfe feedback7: the battle sim uses engine-independent math — Chrome, Firefox, Safari and the server agree bit for bit (PR #373 by @Cloudnyco)
- server/sim/detmath.js: hypot (sqrt of the squares, rescaled outside 2^±500), powi (integer powers in double-double,
  one rounding), sin / cos (fdlibm 5.3) and atan2 (fdlibm / FreeBSD msun) built only from correctly rounded IEEE
  operations; the sim's Math.hypot / sin / cos / atan2 / pow calls go through it (65 files); eslint.config.js forbids
  the approximated Math functions under server/sim and test/sim/detmath.test.js scans for them.
- sin / cos / atan2 are the algorithms V8 runs in Node, so the server's results stay; only hypot differs from V8's.
  Golden: 2 moved — diy-027 (a hypot distance, damage +80) and coop3-ABYSS-5 (a later card pick from round 9, same end).
- Applied unchanged after review (Grok static review: MERGE; the diff read in full: no network, no new dependency).
- Tests: broad node --test 5155 / 5155 with the new file tracked (detmath 7 / 7, package 10 / 10); lint 0 errors,
  typecheck, check:imports.
2026-10-08 18:35:55 +08:00
sganggs 806a047717 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
sganggs 17c96843ee feedback5 CB2: the countdown summons (医疗探机, 诅咒娃娃, 海嗣; 自选 工程蓄水炮, 沙地兽, 打字机, 本能的召唤) take no damage — 活性源石 included —, are never healed, and their bar shows the life left (community report of 2026-10-06 「赫默的无人机、浊心斯卡蒂的海嗣等这种召唤物不会受到外来伤害…血量随时间减少」)
Official: PRTS 斯卡蒂的海嗣 备注 「持有禁疗、无敌」, the token prefabs of 海嗣 and 沙地兽 carry abnormal flags 5 + 7 (无敌,
禁疗), 工程蓄水炮 7 (禁疗); the report describes the official mode for the drone too — [ASSUMED] 无敌 + 禁疗 for the summons
the base prefabs give no flag (医疗探机, 诅咒娃娃, 打字机, 本能的召唤), and the bar as a timer.
Cause: content/tokens.js / the kits only made them untargetable: a 海嗣 on 活性源石 lost 70 of its 100 HP every second, a
hurt drone drew every medic's heals, and the bar was plain HP. Now: COUNTDOWN_SUMMONS / startCountdown (called where each
lifetime is scheduled: tokens.js scheduleLifetime, 浊心斯卡蒂, 温蒂, 莱伊, 鸿雪, 酒神 kits) give 无敌 + 禁疗 and
Unit.countdown; snapshot.js sends hp = maxHp × the share of the life left (the sim HP stays full, so no selector sees a
low-HP summon). Attacked summons (沙之碑, 狼群 …) are unchanged. The 吟游者 trait therefore cannot refill such a bar (item 32).

Golden: snapshot hashes only (the bar) — roster-003, -016, -018, -035, -045, -046, bond-visiShip-low / -high,
boss-boss_1 / 3 / 5 / 7-pair, unite-1; gameplay too — roster-005 (medics no longer heal the drone), roster-037 and
hidden-boss_9-pair (a 海嗣 / drone under enemy and leader damage stays up: fights change).
Tests: test/sim/feedback5-countdown-summons.test.js (7; fails on the parent: no COUNTDOWN_SUMMONS — and there the 海嗣
lost 70 HP to an 活性源石 tick). Docs: SIM.md token pieces, PLAYING.md 作战中.
2026-10-07 02:27:33 +08:00
sganggs ac6454d171 feedback5: merge feedback4 (7ee815e) — W4D ported: a sleeping enemy cannot be blocked and takes no block slot (_checkBlock in server/sim/battle/blocking.js, the applyStatus release in battle/status.js, the plain-buff release re-applied to ai.js updateEnemy's cannot-act branch after WE1's swing cut, 瑕光's comment in kits/ops/chess_char_3_12-blemsh.js, buffs.js; 酒神's 迷狂牢笼 skips a sleeper too); golden: roster-003 / 005 / 013 / 032 / 038 / 039 / 042, bond-maniShip-high, solo-FUNNY-1, coop4-FUNNY-6, coop4-ABYSS-7, coop2-NORMAL-8-serverrun, coop2-NORMAL-70-diy, diy-131 / 132 moved
New on feedback4 since the last merge: W4D (bbe59b9, merged as 7ee815e) and the 0.1.4 version commit (30e7ace).
feedback5 keeps its own package.json / package-lock.json / APP_VERSION / README / DEPLOY version lines (0.2.0 sets
its own version later) and takes feedback4's CHANGELOG 0.1.4 entry, with W4D's line in it.

- Port (PRTS 异常效果 SLEEPING 沉睡 = 无法行动+无敌+不可阻挡; GitHub #140; the owner's decision of 2026-10-06):
  `_checkBlock` refuses `flags.sleep` (battle/blocking.js) and `applyStatus` releases an enemy that falls asleep
  (battle/status.js). Battle.js stays feedback5's façade, and tier3.js stays deleted (the 瑕光 comment is in her kits/ops
  file). ai.js: WE1 (89f730d) cuts a swing before enemyAttack on a stun / freeze / sleep. The cannot-act branch is still
  there, and a plain-buff `flags.sleep` still makes the enemy `stun` (units.js), so the plain-buff release stays in that
  branch, after the cut, with its comment rewritten for the new order.
- New on feedback5: 酒神's 迷狂牢笼 (kits/ops/op-phatm2.js cageBlock, O22) duplicated the old three-flag 不可阻挡 check.
  It now skips a sleeper too. No golden scenario spawns a 牢笼 over a sleeper: the corpus is identical with that line
  reverted. 予愿安洁莉娜's S3 helper is unchanged, since canTargetEnemy already drops sleepers.
- Readers that follow without code: blockedTargets (blocked-first targeting, the DEFAULT trigger, hitAllBlocked), the 要塞
  branch of acquireTargets (u.blocking), “庞贝”'s clock (blockedBy), the snapshot's UF.BLOCKED and the client's block
  marker (render/units.js). 死芒 S2 折朽 (自选) lets go a linked enemy that her 悲叹的仆役 held.
- Docs: DESIGN §24.9, its §24 intro entry, the §5.5 Blocking sentence (on feedback5's 阻挡半径倍率 text) and §22.8's
  "(since §24.9)", all as on feedback4; SIM.md as on feedback4 (the plain-buff sentence merged with feedback5's
  noNewBlock / undying text).
- Golden: a probe (counting successful sleeps on a blocked enemy, plain-buff sleeps and 牢笼 spawns over sleepers) ran
  over all 283 scenarios. Every scenario that moved sleeps blocked enemies; the counts are in brackets:
  roster-003 (小满 S2 [5]: same fight, blocked 27 → 29, snapshot hashes); roster-005 (elite 小满 S2 [3]: the 140 s
  timeout with 83 kills and 1 leak, where it used to clear at 135.43 s with 84); roster-013 (缇缇 S3 [11]: same kills,
  blocked 22 → 33); roster-032 (缇缇 S1 [10]: 71 kills instead of 68, 8 leaks instead of 11, 4 knock-outs instead of 8,
  萨尔贡 layers 41 → 35); roster-038 (elite 缇缇 S2 [1]: ends at 197.40 s instead of 197.27 s); roster-039 (瑕光 S2
  normal + elite [31 + 12]: same kills and counted leaks, the 窃贼 gets away, blocked 53 → 87); roster-042 (缇缇 S2 [3]:
  ends at 168.77 s instead of 166.93 s); bond-maniShip-high (小满 [1]: blocked 29 → 30, snapshot hashes); solo-FUNNY-1
  (ai_0's 缇缇 [3]: only ai_0's damage dealt); coop4-FUNNY-6 (ai_0's 缇缇 [15 + 56]: from round 9 ai_0's 精准 / 萨尔贡
  layers and boss damage move, ai_1 loses no LP from round 11, team LP 124 → 130); coop4-ABYSS-7 (ai_2's 小满 [18]: ai_2
  one kill fewer, ai_0 one more LP lost in round 9); coop2-NORMAL-8-serverrun (ai_1's 缇缇 [83]: only ai_1's damage dealt;
  WE1 had already turned this match into a loss); coop2-NORMAL-70-diy (the 自选 seat h_0's 缇缇 [171], ai_0's 缇缇 [74]
  and 小满 [19 + 21]: h_0's layers move from round 4, then its AI 托管 buys differently (boards from round 7); ai_0 246
  kills instead of 233, same end); diy-131 (死芒 S2 at tiers 5 and 6 [1 + 1]: snapshot hashes only); diy-132 (elite
  死芒 S2 [2]: ends at 199.50 s instead of 198.13 s, same kills). One scenario sleeps a blocked enemy and did not move:
  coop2-HARD-4 [1]. ai_1's 小满 lets go the enemy it holds; another enemy takes the slot; the sleeper wakes with no room
  and is blocked further on by 2_16_a. The match digest has no snapshot hashes, and its kills, leaks, damage and funds
  are identical. fields and standins did not move. Most outcomes differ from feedback4's list because feedback5's
  baselines differ (WE1's damage frame, etc.).
- Tests: test/sim/feedback4-sleep-unblock.test.js (from feedback4) passes 5/5 on the split layout. Every case fails with
  the three sim hunks reverted, and case 3 (plain buff) fails with only the ai.js release reverted. Targeted sim tests
  pass 184/184 (blocking, block targets, enemy swing, feedback2-enemy-attack-stand, combat / statuses, robustness,
  stun SP, professions, healers, fidelity …). Content tests plus kits_layout and docs-consistency pass 658/658 (kits
  t1t2 / t3 / t5 and alt t2 / t3 / t5 for 小满 / 瑕光 / 缇缇, 酒神, 死芒, garrisons, bosses, items). Stealth / liftoff /
  enemies pass 36/36. npm run golden: all 283 match. GOLDEN_FULL=1 test/golden.test.js passes 9/9 (no stored engine
  error). lint 0 errors, check:imports and typecheck clean.
2026-10-06 17:49:10 +08:00
sganggs a61d82ee4c feedback5 O22: 自选 kits for 淬羽赫默 (with her 夜灯), 遥 and 酒神 (with her 本能的召唤 and 迷狂牢笼)
Three owned-6★ picks get their operator kits (kits/ops/op-slent2.js, op-haruka.js, op-phatm2.js; OPERATOR_KIT_FILES):
every skill at rank 4 / 7, both talents, the trait and every module at stages 1 and 3, from data/backups.json, PRTS and the
client's battle data (charpacks + buff_template_data of the three operators).

- 淬羽赫默: 护佑者 (arts attacks, heals while a skill runs; BLS-X 100 %, BLS-Y 2 targets); 无声砥柱 庇护 10 % → 24 % by the
  lost HP (shared `protect`); 丰润羽翼 as 生命回复速度 (the atk_to_hp_recovery template PRTS documents so for 絮雨 / 琴柳), ×2
  for 莱茵生命; BLS-X Lv3 max HP +10 %, BLS-Y Lv3 the once-per-deployment barrier; S2's placed 夜灯 (T1 ×2.1 / ×2.4 around it,
  withdrawn at the skill's end or when she leaves); S3's 不死 for the first lethal hit in her range, 2 uses a battle.
- 遥: bubbles with each attack (or instead of one), 30 % 庇护 ([bonus], ×1.4 / ×1.7 in S3), breaking 0.5 s after an enemy's
  damage, the break heal (BLS-Y Lv3 +1 SP to the holder); S1 re-bubbles around, S2 heal → arts splash and endless from the
  2nd use, S3 y-8 and the levitation with its damage.
- 酒神: T1 神经损伤 on each hit + 1.3 splash; 堕梦 ASPD cut in 神经 bursts and 70 / 90 on enemy attacks (a burst interrupts
  the attack); RIT-X ×1.18 vs elite / leader; S1 two hits + 束缚 ×1.3 / ×1.5; S2 (SP_FULL toggle) the placed 本能的召唤
  (lure 4, exit 停顿 + DoT, 25 s redeploys while it runs); S3 y-8, per-second 神经损伤 until a burst, 1.5× burst recovery
  and the 迷狂牢笼 (blocks its tile's enemies at spawn / refresh, 1 per hit, 30 s after the skill).

Engine (additive, no handler before): ai.js hook `enemyAttackStart` (an enemy's normal attack, before 麻痹), blocking.js
flag `noNewBlock`, items/battle.js `holdsUndying` also sees a buff flagged `undying`; docs/SIM.md §3 / §5, kits/README.md
(docked skill-summon pieces: their kit without offOwner).

Golden: only diy moved — diy-106 … diy-109 (the prototype configurations shifted behind the new operators) and the new
diy-110 … diy-114 (the three operators' configurations); roster, bonds, fields, matches, standins unchanged.
Tests: test/content/op_slent2.test.js (11), op_haruka.test.js (9), op_phatm2.test.js (13).
2026-10-06 05:14:29 +08:00