Ledger: Audit: the eft0061 straight fire shot (data record 130)
The template skill for the data-driven shots: every observable part, its state and where the decode or the lead lives. Rows above carry the details; this ledger only points.
Part
- Status: State
- Now: Where / lead
Record fields read (cost, uses, damage, reaction ids, kind, projectile block, effect, sounds, cast profile)
- Status: DECODED
- Now: FORMATS.md “Skill
table”;
data_skill.gdsetup_record
Cooldown, recovery, knockback
- Status: PLACEHOLDER
- Now: row “Cooldown, recovery, knockback”
Cast motion slot and release tick (from the cast profile)
- Status: DECODED
- Now: FORMATS.md “Cast animation profiles”; row “eft0061 skill cast pose” left for the footage check
Hand bone the forming effect follows
- Status: INFERRED (footage)
- Now: FORMATS.md “Skill cast animation”; lead: the bone the exe’s cast event attaches slot 0 to
Effect slot meanings (forming, flight, hit, launch)
- Status: INFERRED (footage)
- Now: FORMATS.md “Skill effects”; lead: the cast/shot code that spawns each slot
Shot spawn point (chest offset)
- Status: INFERRED
- Now:
data_skill.gd_throw,fire_shot.gd: start = feet + 1.2 m, 0.6 m ahead (ours); the launch direction is the decoded frame (vaim 11-10), no longer aimed at target + 0.9 m. decoded (docs 11-10 vaim, M): start = the cast’s fire node position (several nodes: shot index modulo count), else the body centre (feet + centre height, character record 0x12); lead: fire-node id at cast state + 0xc20, node lookup 0x14018a310; FORMATS “Vertical aim of a cast”
Vertical aim per skill kind
- Status: DECODED (M, vaim 11-10 built) / MISSING (kinds not built)
- Now:
data_skill.gdPITCHED_KINDS (straight, ricochet, multishot fly along the pitched frame),blast.gd(our lob: arc solve, pitch ignored). Straight, (shotvar) ricochet and multishot (with a target: its own 3-D line to the target point) follow it; (autocrawl) crawl and moving use the cast frame laid flat (the pitch is ignored; decoded from their inits); (autocrawl) kinesis throws along the cast heading laid flat and turns toward the target (INFERRED, row “Kinesis lift”). (fixes) The target end is the posed torso root node (DataSkill.target_point, row “Target record point”); the caster end stays feet + 0.9 m (decoded: feet + the centre height C + 0x12E98, 9 u). MISSING: the start at the fire node. (lobbed: the parabola / fall / rain landing point without a target, 0x14026e410, is decoded and played: 190 u, range class 1: 300 u, along the heading) Not followed: whether the drawn body leans with the cast pitch
Target record point
- Status: DECODED (doc-body; built fixes) / INFERRED (models without a known joint)
- Now: The point a locked
cast aims and steers at (body record + 0x60,
written each tick by 0x1402513a0): the world
position of the posed torso root node (packed
node id 0x4D1D1710),
DataSkill.target_pointviaPdCharacter.torso_bone(): joint 9 of pc00a and joint 3 of enm12a0 (found by the packed id in the user’s model files; their rest heights 11.261 / 12.249 u match “Body point, sight point and aim point”). Used by every reader of target_point (locked cast pitch, shot homing, multishot / parabola aim, melee hit test, mines; (followups 11-15) the lock grade distance at both ends,lock_on.gdtick(), so the marker ring and the lunge gate grade torso root to torso root) and the deflect / homing of straight shots. (fixes r2) Also the mirror reflection’s new heading (the kind’s init re-runs with the old owner as the target, FORMATS “Reflect”). INFERRED: the ricochet’s last bounce toward its target takes the same point (0x14026ca50’s target point not read). The hand-made test skills (skills/blast.gd,skills/beam.gd, row “Test skills”) keep their own feet + 0.9 m aim. INFERRED: a body without a converted model, or a model whose joint index is not inTORSO_BONES, takes feet + 11.261 u (pc00a’s rest height) - Where:
data_skill.gdtarget_point,pd_character.gdTORSO_BONES - Lead: the joint index per model from the packed name (the converter could export it, FORMATS implementation notes); the body radius (record + 0xb0, 5 u) is a separate row
Speed, homing rate, lifetime from range
- Status: DECODED
- Now: FORMATS.md “Straight behavior”
Homing percent, homing cone (60°)
- Status: INFERRED
- Now: row “Homing percent of the cast event”; the cone has no source: lead, the straight kind’s update
Hit test (swept segment vs a sphere of object radius + character radius, line of sight)
- Status: DECODED in FORMATS, NOT FOLLOWED
- Now: FORMATS.md “Rules
shared by all kinds”; ours is a 0.25 m Area3D
sphere plus a step ray
(
projectile.gd) and ignores the kind’s object radius; the character centre’s height is not traced Decoded 11-12 (doc-body): the target’s point is its body point (torso root node, posed; record + 0x60) and the character radius is the character table radius, 5 u for the playable cast (record + 0xb0, M: one writer found); the melee line of sight starts at the caster’s feet + centre height (9 u). Ours still uses its own sphere
Shots against the stage
- Status: INFERRED
- Now: row “Shots against the stage collision”
Flight effect, launch effect, trail
- Status: DECODED draw, INFERRED slot pick
- Now: FORMATS.md “Effect trails”; row “Effect trails: UV format, history rate”
Hit effect: slot, place, matrix
- Status: DECODED (bof3)
- Now: FORMATS.md “Shot end”:
slot 2 at the position the hit tick’s move
started, the shot’s own matrix (not turned),
scaled by 2 × blast radius (m) for an exploding
record, not attached to the target; no hit slot
at the lifetime end.
projectile.gdhit_effect_transform; test_shot_end
Hit effect look
- Status: INFERRED
- Now: rows “eft0061 hit burst” and “Effect element joints”
Character hit cue
- Status: INFERRED, narrowed (bof3, docs 11-09)
- Now: A hit on a character
(reason 2, not exploding) also plays common-bank
cue 6 or 7 at the shot
(0x14026d0dc..0x14026d144, volume and pitch
0x1000). Decoded: the pick is the target’s
character table impact sound flag (charaset
+0x34): 0 → 6, 1 → 7; it is 0 for every playable
model, so ours now plays 6
(
Projectile.CHAR_HIT_CUE; was 7) - Where:
projectile.gd - Lead: one more check on the target (character + 0x10 == 0, else 7) is not traced; read the flag from converted data when non-playable models become targets
Flames staying on a hit character (~1 s, xemu 1419)
- Status: MISSING (bof3)
- Now: Not the hit slot (world-fixed, decoded) and not slot 5 of the same effect (0x1402729c0, spawned on some hit paths at a per-character point of 0x14058bd40; eft0061 has no slot 5). Lead: the character’s damage reaction (0x140272bc0 → 0x1402729c0 / 0x14017ba00, the character draw) and a PC capture of a hit on a character. (docs 11-12 effects): to re-check after the joint fix: the hit slot’s petals stand upright 0..1.5 m at the hit place for frames 5–25 (~0.7 s) once joint rotations are applied, which may be what the footage shows (INFERRED; the exclusion above was made with the petals lying flat); slot 5 on a hit character is the other source for effects that have it
Target reaction tiers, damage
- Status: DECODED
- Now: FORMATS.md “Hit reactions”; rows “Hit direction of a reaction”, “Hit-reaction root motion”, “Follow-up reaction (0x74)”
Sounds per event (cast profile sound, voice, flight loop, hit)
- Status: DECODED ids, INFERRED playback
- Now: FORMATS.md “Sound banks”; Audio rows above
Effect light: envelope, pool life and fade, per-character pick and channel cross-fade, cP0 only in battle
- Status: DECODED (bof2)
- Now: row “Effect dynamic light (light02)” (left: pick point, tick order)
Stage lit by the latest effect light
- Status: DECODED, built (bof4; decoded bofcmp 11-13); capsules MISSING
- Now: Under light set 0 the
stage’s point light 0 is the most recently
allocated live effect record (index 0x140527cc8,
set at allocation 0x14010cd10, cleared at battle
setup 0x14019a020; no re-pick when it ends: once
it is gone the stage’s own point light is back).
Ours:
PdEffect.latest_lightis set when a record is first written,PdData.apply_stage_point_lightwrites its colour × range, colour, range and position into the stage materials’p0_*(the same terms the stage shader uses for the stage’s own point light) and falls back to the stage’s own set 0 point light. Not fed: the capsules (own shader with no point light term) - Where:
pd_effect.gd_write_light/stage_light,pd_data.gdapply_stage_point_light,pd_stagefx.gd_apply_light - Lead: FORMATS “The straight
fire shot, the rebuild against the game” section
1; test
test_fireshot_cast
Glow pass
- Status: DECODED, small gaps
- Now: rows “Glow (bloom) pass”, “Glow paths not modelled”
Blend: all elements of this effect are one/one additive
- Status: DECODED
- Now: row “Effect blend from the screen texture” does not apply to this effect
particle01 corners/UVs, size push
- Status: DECODED (bof2, bof3)
- Now: FORMATS.md “Effect particles”; row “Effect particle rand seed”
Release burst brightness and size
- Status: INFERRED, narrowed (bof4; was DECODED, bofcmp 11-13)
- Now: Blend decoded
(bof4): both particle draws
(0x1401a1c00 via 0x1401a1fd0, and 0x1401a4110)
set the fixed blend word 0x1101 (one / one, add)
just before drawing, so the element material’s
blend is never used for particles (it gives the
texture only); ours drew the cast-slot haze with
its material blend (src colour / one), which
squared the dim sprite colours into a dim red.
Now one / one: the charge flash reads yellow
(render probe
tests/render_flash.gd, chest box mean at tick 16 (168,117,47) → (191,140,43), tick 20 (120,53,47) → (170,95,51); game footage ≈ (216,192,60)). The 8-bit-vs-linear lead is ruled out (the compatibility renderer adds stored values). Left: after the release ours turns red within ~2 ticks where the game stays yellow ~8 ticks (leads: the release slot 3 cloud on joint 3 vs the game’s chest/arm; the glow pass; aging rate); the sparks are too large (msparticle00 inferred, Xbox point sprite size). Slot 3 plays at the release on the caster, for every cast kind throughPdEffect.play_release - Where:
pd_effect.gdPARTICLE_DRAW,data_skill.gd - Lead: FORMATS “The straight fire shot, the rebuild against the game” sections 3, 4
Floor glow while forming
- Status: DECODED, built (bof4; decoded bofcmp 11-13)
- Now: The cast slot’s light02 lighting the stage through row “Stage lit by the latest effect light”. Same on Xbox and PC
Body turns with the camera during a cast
- Status: INFERRED, narrowed (bof4)
- Now: Fixed: the body keeps
the yaw it has at the cast start for the whole
cast state (
player.gdskips_facewhile_recover_leftruns), the camera still turns with the lock. Left INFERRED: the yaw is taken as it is at the press; the exe’s own turn step into the cast is not traced - Where:
player.gd_physics_process - Lead: row “Cast aim without
a target”; FORMATS same section, 2; test
test_fireshot_cast
Skill card pops at a cast press
- Status: PLACEHOLDER (bofcmp 11-13)
- Now: Ours shows the held slot’s card for the frames of every cast press; neither build of the game does (footage). The gate is not decoded
- Where:
pd_hud.gdheld_slot()/_card() - Lead: FORMATS same section, 6
Aura cost paid at the release
- Status: DECODED (bof4; r2: refusal decoded), set-6 payer INFERRED
- Now: The press only checks
the cost (0x140262d80); the cost is paid at the
release tick (0x1402626d0 computes it,
0x1402621f0 pays). Ours:
Skill.try_castchecks,DataSkill.release_paypays at the end of the windup; skills without a windup pay at the press (Skill.pay). DECODED (r2): when the aura (or level) fell below the cost during the windup, the pay routine refuses, the paid slot is set to 0 and the release returns before the power (ours: no release effect, no power, nothing paid). That branch is skipped when a global equals 0x7777 (0x1404c01d8) or the player index is above a count at 0x1404bdc40 + 0x10 (not traced; ours always takes it). The refused release also returns before the use counter, the used-pile step and the slot routine 0x140263460, so the card is probably not spent; ours still spends the use at the press as before (INFERRED, lead: the slot routine 0x140263460 and when the press empties the button). INFERRED: a set-6 card (recalled attack) pays its own cost at the recalled card’s release (DataSkill._payer) - Where:
skill.gdtry_cast/pay,data_skill.gd_cast/release_pay - Lead: FORMATS “Skill cast
animation”; tests
test_fireshot_cast,test_cast_pay