Entry 281

Mutation's attack spells: the ccSkill's request-time aim and the later release

2026-09-28battle render volumes test

Documented in Battle rules.

Filescrates/piney-effect/src/spell.rscrates/piney-effect/src/convergence.rscrates/piney-effect/src/fall.rscrates/piney-effect/src/upheaval.rscrates/piney-effect/src/summons.rscrates/piney-effect/src/summoned.rscrates/piney-effect/src/files.rscrates/piney-effect/examples/spell_probe.rstools/test_effect_spell_rs.pydocs/engine/effects.md

tools/test_effect_spell_rs.py had only run on Infection. On Mutation's disc the game's ccSkill died at frame 0, because the harness built it with Infection's layout. MUT's _ccSkillRequest (gcmn 0x00598000) news 192 bytes, not 176; OUT's and QUA's do too. A vector at +0x70 moves creator and everything after it 0x10 on: creator +0x80, target +0x84, effPtr +0x94, m_effElm +0xb4.

The new field

The request copies the target's position into +0x60 (tPos) and into +0x70 when the target is on the lists (MUT 0x005981d4). The constructor (MUT 0x00598730) leaves +0x70 alone.
ccSkill::Main (MUT 0x00598980) is Infection's up to the offsets. It refreshes +0x60 each frame and never touches +0x70.
So +0x70 is where the target stood when the spell was asked for. The port keeps it as Spell::t_pos_req; Spell::aim(volume) answers tPos on INF and +0x70 from MUT on.

The elements still read +0x60. A diff of all 190 gcmn functions named *Element* (INF against MUT, registers and addresses dropped) shows only the skill's creator and target offsets, except two elements (below).

The systems

Each MUT system differs from INF's in two ways: what INF aims at tPos MUT aims at +0x70, and the caster is released later.

system	MUT gcmn	aimed at +0x70	released (INF)
Fall	0x0059d1e0	the meteor's height, sound 57, `ccSkillDamage2`	at 0 only with no target; else at 120, or at the end or a lost target before 120 (always at 0)
Tornado	0x0059ccf0	smoke, rings, their sound, the hits round the place, the shake's range	after 75, with the end (at 30)
Convergence	0x0059d930	count 20 writes the target's pos to +0x70 (INF: `tPos`); sounds 63, 64, the shake's range, note 66	at 93, or at the end before 93 (at 40)
Upheaval	0x0059e170	the ground generator, the pillars, `ccSkillDamage2`, note 56, the shake's range	at 93, or at the end before 93 (at 35)
Summons	0x0059ea10	sounds 62, 290's lock-on, 289's drills, `ccSkillDamage2`	at 60; 100 from level 3 and for 289, 290 (at the end)

The convergence's burst, smoke and effSkillBreakSE stay at the target's middle, read from the target as on INF.

Main and two elements

A diff of every main eff* function, and of the rest of INF's effect code (main 0x001bc000-0x001dc000), leaves these beside the three already in the tree (the pieces' aim, effSkillChargeObj's new argument, the drill's endFlag):

effSkillChargeObject (MUT main 0x001eb650) sets the controller's posT to the target's middle (INF: its feet). The pieces start round it. Each piece's own posT is still the target's feet, and MUT's pieces fly to that.
effTCDrillMissile (MUT main 0x001edcf0) builds each drill's place from tPos in a local. INF built it in the caller's cPos, whose z gathered tPos.z each time round. Each drill now lives 600 frames (INF: the row's -1).
ccTreeSummonsElement's roots: INF's loop never moves past m_rootsTbl[0] (the addiu $s0, 16 is missing). MUT's (gcmn 0x00511108) places all eight 100 round the tree through ccTransPosW2P and ccTransPosP2W.
ccThunderSummonsElement::Main (MUT gcmn 0x0050e0e8) plays the hit's four sounds once even when the shake is out of the camera's range. INF plays them only in range.

The harness

The ccSkill is volume.SKILL_SIZE and its fields go through volume.skill_at. From MUT on the request sets +0x70 and the skill's state reports it as tPosReq; the probe prints the same.
The elements are read with INF's DWARF. MUT has no .debug section, and its element classes keep INF's layout.
The zeroing hooks (locals the game reads before writing) are per volume, and each decodes the move it replaces. OUT's and QUA's recompiled frames keep the tree upheaval's throw at sp+112, the ice Delete throws at sp+80 and the goblin's rings at sp+256.

Results

python3 tools/test_effect_spell_rs.py bulk 20 (the release probe), all 102 attack spell ids, 20 random casts each (2,040 casts a volume):

system	INF frames	MUT frames
tornado	37,777	37,777
fall	32,457	32,456
convergence	44,484	44,439
upheaval	44,942	44,942
summons	87,626	85,148
all	247,286	244,762

0 mismatches on both. The summons ran in six shards of the same cases. Before the ports, MUT parted at frame 0 (the harness's layout), then at the pieces' posT (convergence), then at 289's drills' life. cargo test --release -p piney-effect passes; clippy is clean.

Outbreak and Quarantine

Run with PINEY_VOLUME=outbreak or quarantine, 3 casts each (before the last two fixes below), the two discs gave the same answers:

tornado: 0 mismatches (5,614 frames).
fall, convergence, summons: parted by one ulp in a matrix or an element's m_dirc. That is the recompiled float code (sqrt.s and the like), not chased.
upheaval: parted at 219's ice Delete, whose frame keeps the throws' leftovers elsewhere. The hooks are now per volume (above).
220: OUT's and QUA's _Level4 raises 42 spikes a wave (OUT gcmn 0x005067e0, QUA 0x003f90f0), not 43. Ported; not rerun on OUT or QUA.

Still unknown

why MUT moved the releases (93, 120, 60 and 100). piney-game's AreaFx::spell requests a spell at its first system call, so its +0x70 is the target's place at count 0, not at _ccSkillRequest; a runtime that asks earlier would need to pass it. OUT's and QUA's one-ulp float differences, and whether the per-volume hooks and the 42-spike wave make upheaval agree there (not rerun).