ReadNodeIds is the one GLB parse and now records, per aetheria.id node, whether it carries a mesh. Bind refuses a
thruster-emitter or radiator-mesh anchor whose node has none, naming the ship and anchor (ruling
thrusters-radiators-are-meshes), so Compose and validate refuse such a package before the catalog sees it. AetherDb
'ship-authoring validate' judges the GLB when it exists (next commit).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's scenarios-adopt probe S4 as a committed test: on generated ships,
stations and turrets every weapon, thruster and radiator holds a non-null
Hardpoint containing all its cells, and Tool gear and the hull hold none.
It pins the input the muzzle, arc, thruster-emitter and radiator readers
dereference (finding headless-reader-input).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The one conflict, ShipInstance.cs's thruster emitter lookup, keeps the
branch's thruster.Item.Hardpoint.Transform and master's new
ThrusterHardpoint.WireExhaust(particles, thrusterHardpoint). Master's side
as written would have brought back the origin-cell hardpoint lookup.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The one conflict is EntityInstance's SoundBank block, which this branch had
moved onto item.Hardpoint and ship-mounts deleted: it stays deleted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The emitter carries the invisible material, whose alpha is 0, so an exhaust
prefab left at Unity's default (Use Mesh Colors on) emitted nothing. The
shipped thruster prefabs only worked because theirs is off. ThrusterHardpoint
.WireExhaust now owns the wiring and turns it off; ShipInstance calls it. The
preview smoke wires a default particle system to each emitter on the clone and
requires every particle emitted and none transparent.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PreloadAsync rethrew on the first failing hull, so one bad package faulted
Loading and no mod ship could spawn. Each hull now loads in its own try: a
failure destroys its partial visual, is logged with the hull's name, and the
rest still preload, as Compose skips a bad package. The play smoke installs the
package beside a broken sibling that composes but cannot assemble (its map icon
moved onto a meshless muzzle node) and proves the sibling has no prototype while
the good ship spawns.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
forceRenderingOff is not serialized, so Instantiate and SaveAsPrefabAsset dropped
it and every cloned mod ship (and every FBX-built prefab) drew its collision hull.
The collider renderer now takes the template's invisible material, as the thruster
emitter does. The preview smoke checks the collider on the clone and the prefab
smoke checks it on the configured prefab.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's scenarios batch-4 probe P30 lived only in a session scratchpad. Its
fixtures are now tests: an L gun in an L hardpoint whose origin lies outside
it, an edge gun whose origin lies in a neighbouring radiator hardpoint,
TryFindSpace finding what ItemFits accepts (unrotated and turned), reload into
the same hardpoint, and Tool gear holding none. Three batch-4 minors are
pinned too: IsFilledBy against a cell oracle written in the test, the last-row
placement search, and billing independent of heater equip order.
The file is a part of RunStartTests, so it shares that fixture's scratch-hull
and catalog helpers instead of a second copy of them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The placement rule (HardpointAt) already finds the one hardpoint an item lies
in at equip; TryEquip threw that answer away and eighteen readers recomputed it
from the item's origin cell, which need not lie in the hardpoint (Soul's
scenarios batch-4 P30). EquippedItem.Hardpoint now carries the answer, null for
Tool gear and the hull, and every origin-cell lookup reads it instead:
FireControl.ArcFor, the eleven weapon managers, EntityInstance, ShipInstance
and ActionGameManager's articulation grouping. Entity.Hardpoints[,] stays the
per-cell map for armour and display.
TryFindSpace's hardpoint search tries every origin the fit rule can accept,
including ones up and left of the hardpoint, so a design whose shape lacks its
origin cell is found wherever ItemFits would place it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Kills the Stryker survivor that armed every weapon with any muzzle on the ship
(the && in the per-weapon muzzle check turned into ||): a second weapon mount
with no muzzle is refused even when the first has one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The one conflict is AetheriaStores.CatalogTypes: keep ShipAuthoring from the
ships lane and Scenario from the scenarios lane.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
forceRenderingOff is not serialized, so the prototype's clone drew the
emitter again; the preview smoke's clone check caught it. The emitter's
materials become the template's invisible material, which survives
Instantiate and which EntityInstance keeps as the renderer's own material.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ruling thrusters-radiators-are-meshes: thruster and radiator mounts are meshes;
the thruster's is not drawn and is the surface its exhaust emits from, the
radiator's is drawn. The assembler already required a mesh for both; it now
turns rendering off on the thruster emitter, and the preview smoke checks it.
The authoring doc states the mount rules and the mesh rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ShipAuthoringStore.Validate now alone decides which hardpoints carry anchors
and which role each plays. A thruster, radiator, energy, ballistic or launcher
hardpoint needs an anchor of its mount id with role thruster-emitter,
radiator-mesh or the new weapon-mount; every other hardpoint is internal and
no anchor may take its id. A mount-role anchor must be a hardpoint's mount, a
muzzle must hang off a weapon mount, and each weapon needs a muzzle; these
checks move out of ShipModPlan.Build, which now only maps a validated pair to
slots and throws nothing itself. 'articulation' stays in the vocabulary for
pivots and is never a mount.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The array's only reader was the dead SoundBank block in EntityInstance, which
looked up a hardpoint transform and did nothing with it. ShipModPlan.Equipment,
ShipModShips' assignment, the preview check, the FBX builder's HP.Equipment
case and its smoke assertion all fed that one dead reader. An HP.Equipment node
is now refused by the builder as an unknown hardpoint kind.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's S6. Generation always gives Zenith's 16-cell Reactor hardpoint a reactor
that fills it though smaller ones fit; a generated Djinni's eight thrusters
share one lot; across 5 K either side of a heater's target the bus bills it
exactly when its thermostat is open; and a thermostat gates only its own
behaviour group (a two-group fixture: the closed group's 10 is not billed, the
other group's 7 is).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's S3 and the S2 ruling ("B is the design intent, saving slots in a
hardpoint can be a valid tradeoff for crowded ships if you really need to fit
an extra tool"). Entity.HardpointAt is the one placement rule: at an origin,
the item lies within a hardpoint of its type at that hardpoint's rotation
(HardpointData.TakesAt), its own cells are free, and that hardpoint holds no
other hardpoint item. The origin no longer has to be a hardpoint cell, and
TryFindSpace tries every offset inside a hardpoint, the offsets the fit rule
tries, so an L-shaped hardpoint takes its L design instead of generation
throwing. Leftover hardpoint cells may hold general gear, and a hardpoint still
holds one hardpoint item.
Tests: an L hardpoint takes its L design by hand and in generation (Soul's
P11); a turret's gun goes back beside the tool gear in its leftover cells
(P21); a second gun in a roomy hardpoint is refused; a saved entity reloads to
the same loadout with its items in reverse order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling, 2026-09-30: "removing the 11 self-entries was my intention".
IsAvailable, the one reach rule, now says a faction always reaches its own
manufacturer's gear, and a transient `AetherDb allegiance-self [apply]` drops
the self-entries from the eleven factions that had one (Rossum & Douglas never
did, which is why its stations could not carry its own gear).
Tests: in a main-sector galaxy every faction is offered all it makes and no
allegiance map names its own faction; and, through a fixture faction outside
the galaxy with no self-entry, its own product is on offer to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling, 2026-09-30: "Keep the variety, not every galaxy will have
Zhestokost in it." A transient `AetherDb station-reactors [apply]` gives each
manufacturer that makes reactors a design for Zenith's 16-cell Reactor
hardpoint, scaled by cells from its largest reactor with every other value,
stat term, role, temperature and sound bank its own, and a product by it:
Vulcan Station Reactor (AU, from Vulcan 12 cells x1.333) charge 66.7..266.7, mass 666.7, price 1,550,000
Notorious Station Reactor (Lucent, from Notorious 9 x1.778) charge 711.1..2133.3, mass 1333.3, price 1,800,000
MoveOnPro Station Reactor (NiteLife, from MoveOnPro 4 x4) charge 100..500, mass 1200, price 3,600,000
Core Power Station Reactor (R&D, from Core Power 4 x4) charge 400, mass 200, price 1,000,000
products: Vulcan Station, Notorious Station, MoveOnPro Station, Steadfast Station
The Manhattan-derived one is renamed "Manhattan Station Reactor" (product
"Manhattan Station").
Reach is read through the one rule, LoadoutGenerator.IsAvailable. Test: a
faction allied only to AU, Lightsail and NiteLife is offered none of
Zhestokost's products and AU's, its stations carry nothing it cannot reach and
nothing by Zhestokost, and they are powered.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AFailedBeltTaskSurfacesOnce swapped a belt's asteroids while that belt's task
could still be running, so it failed in 5 of 8 runs including unmutated ones
(Soul S1). It tests belt-task machinery that codex/mining deletes (Cut 1,
22a54ccb), so it goes rather than being reworked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling B, 2026-09-30: "a station with no access to Zhestokost gear
should not have Zhestokost gear." Test: in a main-sector galaxy (not the
prelude, which offers every product), every faction's generated stations carry
a heater and a reactor, and nothing, equipped or stocked, whose manufacturer is
outside the galaxy or the faction's allegiance.
With today's catalog every faction's allegiance names every faction (weights
only bias preference) except Rossum & Douglas, which leaves out itself; the
test killed a reach that ignores allegiance through an R&D station carrying
R&D reactors.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's F5 and F6. The asteroid-belt tasks read the clock, the orbits and each
belt's damage and respawn state. Update advanced the clock before waiting on
last tick's tasks (Soul measured 242k transforms placed for the next tick in
576 of 3000 ticks), MineAsteroid wrote damage and respawn state under a
running task, CreateOrbit and AddOrbit wrote the orbits, and a task that threw
stayed in the list, so Wait rethrew it on every tick after.
SettleBelts waits for the tasks and clears the list even when one failed; Update
settles before it advances the clock, and MineAsteroid, CreateOrbit and
AddOrbit settle before they write.
Tests: over 3000 ticks every asteroid of a belt is placed for the time of the
tick that started its task; mining every asteroid of a belt, chipping and
breaking each four times a tick for 1000 ticks, never races the task; a task
that fails surfaces once and later ticks run clean.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The bus billed every active IPowerConsumer, but a behaviour group stops at its
Thermotoggle: a heater above its target never runs its EnergyDraw, yet its
full draw (1000-2500/s) was charged every tick. On an idle station that kept
the reactor in permanent deficit, overloading at overload efficiency. The bus
now skips a consumer whose group has a closed thermostat ahead of it, the same
rule it already applies to inactive items; Thermotoggle.Open is the one
statement of when a thermostat is open, and Execute returns it.
Tests (Soul's F2; S17 survived before): from a hull at one uniform temperature,
one tick below the heater's target raises its cells and one above does not, so
the heater's own contribution is measured with no reactor heat in it; and the
bus bills the heater's draw below its target and not above.
This does not keep an idle station warm (Soul's F1, reported separately): the
reactor's throttled idle heat alone still drives it past its plateau.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling, 2026-09-30: "Hardpoint fit is loose: nobody's gonna stop you
from putting a small reactor in a large reactor's hardpoint." There is one rule,
HardpointData.Takes: gear of the hardpoint's type fits when its shape, at the
hardpoint's rotation, lies within the hardpoint's cells. TakesAt is the same
rule at one offset, and Entity.ItemFits (equipping, save load, presets) now
asks it instead of repeating the containment test. Generation's hardpoint
picks, HasHome (so station stock and loadouts), product reuse and `AetherDb
hardpoint-fit` all read Takes.
Generation keeps its own placement preference, named IsFilledBy and not a fit
rule: it offers gear that fills the hardpoint when any is on offer and falls
back to anything that fits only when none is, so loadouts change only where a
hardpoint used to stay empty. RestoredHullsTests pick fillers through
IsFilledBy.
Consequence: the Autocannon, Vulcan, scorched void policy, Notorious, pretty
pretty bang bang, Manhattan, plight, DeathCluster, Small Drive and pswarm now
have a home and are offered. Still homeless: CShot RainbowLite Lazer,
ChargeBlast SG, ChargeBlast+-, OK Disperser.
Tests: generation offers exactly the sold hardpoint gear some hull fits; a 7x7
design no hull fits is offered only once a hull that takes it exists; a 5x5
Reactor hardpoint nothing fills still gets a reactor; equipping and the rule
agree for every hull, hardpoint and design; the homeless list is stated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul's F4: LRMM72, SRMM72, Autocannon, pswarm and plight had DamageCurve ==
null, and once they had products NPCs were generated with them and
CombatState.SampleDps threw at InstantWeapon.cs:64 / ChargedWeapon.cs:81.
A transient `AetherDb damage-curves [apply]` copies each one's curve from its
nearest comparable weapon, same class first, then closest range and damage:
Autocannon <- DeathCluster (auto-fire Ballistic, 4 cells, range reaching 1000)
LRMM72 <- GT 3K (the only other unguided launcher)
SRMM72 <- scorched void policy (guided; 750 range and 128 damage inside its 500-1250, 128-512)
pswarm <- scorched void policy (guided; identical damage, range, cooldown)
plight <- ChargeBlast+- (charged Energy; nearest range)
Every curve in the catalog is the same one, keys (0, 1, 0, 0) and
(1, 0, -1.7656243, -1.7656243): full damage at minimum range falling to none
at maximum. Test: every weapon a product sells has a damage curve.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling, 2026-09-30, on the Autocannon: "Leave the autocannon unsold,
we'll need to author a bunch more hulls before all the gear variety in the game
has a home." The fit rule gets one owner, HardpointData.Takes (the gear's type,
laid in at the hardpoint's rotation, filling every cell); LoadoutGenerator's
hardpoint picks, its reuse of a previous product and `AetherDb hardpoint-fit`
all read it. RandomProducts, which both station stock and NPC loadouts draw
through, offers only gear with a home: Tool gear, hulls, or hardpoint gear some
hull in the catalog takes. The Autocannon keeps its product and price.
Test: the Autocannon is never drawn in 2000 draws of its hardpoint type; after
a scratch hull whose hardpoint takes it is authored, it is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Update cleared `_updatedOrbits` and only then waited on the previous tick's
asteroid-belt tasks, which read the cache through GetOrbitPosition. A task
still running could see the set mid-clear ("concurrent update ... corrupted its
state") or recompute an orbit into last tick's position. The wait now comes
first.
Test (Soul's orbit counter, committed): 3000 back-to-back 0.02 s ticks of a
quiet arena, and no orbit stands still against its parent. Clearing first gave
500+ stalls or the corruption exception. Measured against the parent because a
moon can genuinely pause in zone space: one such reading (tick 1321, a moon
whose motion cancels its planet's) is real motion, not the race.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operator ruling, 2026-09-30: "Yep, we need a station reactor." Zenith's 16-cell
Reactor hardpoint had no design to fill it, so generated stations had no power.
A transient `AetherDb station-reactor [apply]` authored it from Manhattan, the
largest and most industrial reactor, scaled by cells (16/9):
Station Reactor (GearData, Reactor hardpoint, full 4x4)
mass 500 -> 888.9, price 750,000 -> 1,350,000 (to the nearest 50,000),
charge 500..1000 -> 888.9..1777.8; every other value, the efficiency,
overload and throttling stats and all stat terms, the roles (core,
regulator), temperatures, durability and sound bank are Manhattan's.
Product "Station Reactor" by Zhestokost (Manhattan's maker), default role
quality.
Station generation fills the slot through its ordinary hardpoint pass. Tests:
every generated station carries one heater and a reactor; an idle station (a
quiet arena, nothing else running) holds its heater's cells above freezing over
the second five of ten minutes (unpowered they fell to ~247 K); a station hull
with no room for a heater fails generation by name instead of skipping it
(Soul's S11 survived before).
The station's border cells still sit near 265 K with the heater and reactor
running (the reactor core reaches ~433 K); how warm the whole hull should run
is left as a tuning question.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>