Entry 272

Kyvia's first fight and its disc

2026-09-28battleworldtest

Filescrates/piney-battle/src/boss/kyvia.rscrates/piney-battle/src/boss/kyvia/core.rscrates/piney-battle/src/boss/kyvia/gomora.rscrates/piney-world/src/evarea_b8.rscrates/piney-world/src/combat/boss.rscrates/piney-world/src/bosscam.rstools/test_battle_kyvia_rs.py

Mutation's event 108 now fights Kyvia on the disc of field 9 and runs to its end. Block 21's entry type=7 code=12 makes ccBossKyvia01(1) with its core and five gomoras (MUT gcmn 0x004d2a80, 0x004f8a30, 0x004ff220), and block 23's absent type=7 code=12 follows once the core dies and the body falls. The disc is EVENTAREAB8 (0x0041d200), which GO(1) makes for fields 9-12. The references are docs/engine/boss-kyvia.md and the EVENTAREAB8 section of docs/engine/evarea.md.

How it was taken apart

Of the 98 functions (74.4 KB) the fight shares with Kyvia's later ones, the body's 18, the core's 46 and the gomoras' 32 were lifted to a straight-line form (a register flow over each function's blocks, the DWARF names of Infection's layouts on Mutation's offsets) and ported from that, checked against the listings wherever a count or a float mattered. The layouts match Infection's DWARF except one byte Mutation adds at the body's +0x295f8: its cinema flag.

The lifted form hid one pattern. Where the source had actCount++ == 40, the compiler loads the counter, stores it raised, and compares the old register; the lifted form printed the store and then the compare, which reads as the new value. Eleven places do this (the body's death at 60, LightAtk at 40, 30, 10 and 30, MegidFlame at 100, the core's wave at 25 and its death at 30, CheckCountAtkTime's NowDmgtime++ < Dmgtime, CheckCountGomora's GsCount++ == GsRimit, a gomora's AtkWait-- == 0). The harness found the first; a scan of the listings for a raised counter compared through its old register found the rest. The second gomora comes out 31 frames after the first, not 30.

Two of the game's own slips are kept. The core's constructor sets each gomora's LifeFlg from &GomoraFlg[i] != 0, an address, so always 1. SetCoreState(8) tests CoreState != 8 || CoreState != 6, which always holds. The blur's fields are words at +0x24 (m_enabled), +0x28 (m_exit), +0x0c (m_scale) and +0x1c (m_abgr), not the bytes and offsets first assumed.

The shape of the port

The core and each gomora are a character and a Boss of their own (AffectFunc::Boss), so the party's hits reach them through the same queue as any boss's, and the field runs the queued affects of all seven before the body's frame. The body's frame takes the parts out of their characters (Tree), runs the body, the core (Slave) and the gomoras, and puts them back. The tables are six new entries of the combat group (the three animation tables, Skill_VARIOUS, Skill_DOWNER, and AllGomoraList_1[0], which is its only list: 3 3 3 2 0, 0 1 1 1 2, 3 3 0 1 2); the data version is 12.

The boss camera needed Mutation's modes 3 and 4 (SetTransfer, 0x00476380): HandAtk waits on ResetFlg, and without the ease the fight stood still in its first attack.

Checks

tools/test_battle_kyvia_rs.py runs the game's constructor and Main natively against battle_probe's kyvia, comparing every member the port keeps for the body, the core and each gomora, the calls, the camera, the blur, the effects, the party and both generators each frame. 410 cases, 561,061 frames, agree, and every act of all three classes is reached. The harness stands in for the skills the gomoras ask for: ccSkillRequestParam sets the caster's skillID and skillStatus as _ccSkillRequest does, because the port's request does and a gomora with a skill running does not move. HitEnable/HitDisable only set the flag: the native list of hits looped once a gomora's hit was enabled twice.

event_108_ends_with_kyvia (piney-game's session tests) puts the save at block 21 on field 9 and runs to eventStatus[0] 2: the disc stops, the core rises, Kite's hits (queued on the core every 20 frames while it is on the lists) bring the body's flinch and HandAtk, the core dies, the body dies and exits. boss::kyvia::tests and the disc's own tests hold the rest.

The survey

PINEY_SURVEY_ONLY=108 PINEY_SURVEY_GOD=1 PINEY_SURVEY_FRAMES=150000 ends story 108: blocks 0xe0083; 446 places; last: The World - area 2 field 47. The run enters area 47's dungeon (block 17) and from frame 13,950 to the end fights a Hackberry King there whose HP keeps coming back (536, then 718 of 1410), so it never reaches field 9: blocks 21-23 never run and Kyvia is never made. Run earlier the same day, before 270's pilot changes were committed, the same survey reached field 9 and waited at block 23 (blocks 0x6e0083). The stop is in the pilot's dungeon walk, not in the fight.