Combat and TurretController spelled GetHeight + GridOffset inline; they call target.RideHeight now. The drawn-ship test authors a hull with GridOffset 2.5 on the well's slope. Two new tests aim at a target hull raised 40 u.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DrawAhead moved xz ahead but carried the last step's y, so on a slope the ship's height stepped once per sim tick.
Entity.RideHeight is the one owner of 'an unparented entity sits on the field'; the step and DrawAhead both call it.
A child is drawn where its parent is drawn.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
core-at-roche-and-exaggeration-up: PlanetSettings.CoreRadius and Planet.CoreRadius are
deleted; the well's softened core reads the RocheLimit curve (one curve of mass). The
core < Roche guard cannot fail and is replaced by a Roche-limit-positive-and-finite
check. Settings.asset: CoreRadius removed, HeightExaggeration 10 -> 4.0983607
(10 / (7.32 / 3)). Tests pin the new curves and the peak drawn slope / 2.44x steeper tail.
Not yet built or run at this commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The crosser is unpiloted and launched at 50 u/s; hull drag slows it to about 25 u/s relative at the bow line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The mirror test compared a normalised reading of two spellings and was beaten twice (a statement before the return, gravityHeight's sum). GravityShader writes the whole file (potential, zero-strength guard, reference term, height sum) from GravityField's constant (TailOffset); GravityShaderTests fails on any difference. GravityShaderMirrorTests deleted. GravityFieldParityVerify opens Main Menu.unity: no ZoneRenderer there, its Fog Surface Displacement camera is the menu's own brush presentation (layers 9 and 10) by spec.
Builds not yet verified on this commit; verified at the final merge.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The mutation table showed a join at 0.995R surviving: no sample lay at or past it. At 0.996R the log form
differs from the bowl by 1.6e-5 c, over the test's tolerance.
Not built or run locally; verified in the Yggdrasil job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GravityShaderMirrorTests compares the expression of GravityField.Potential with gravityPotential in Gravity.cginc.
GravityFieldParityVerify now builds four off-axis bodies (one of mass 0), lowers through three frames with bodies moved
between them, and checks the map screen's gravity camera mask. RTS/Planet.prefab loses the Gravity Well object bound to
the Gravity Field material. Builds: yes (verify passes on Starfire; the new test is unrun until the Yggdrasil job).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
TheAuthoredCurvesBuildAtEveryMass loads PlanetSettings from the asset the way RunStartTests does. The unphysical-strength
theory adds -1e-6 and -float.Epsilon and a new test builds zero and negative zero. TheCoreIsTheBowlUpToItsEdge compares
the drop from a lone body's centre with c r^2/2R^2 at 0.5R to 0.99R, tightly enough that a join at 0.95R fails.
Not built or run locally; verified in the Yggdrasil job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Zone.WellReference owns each body's reference zero (GetHeight uses it). Gravity.cginc mirrors GravityField.Potential;
Brushes/Gravity Field draws depth = -h on layer 9 from a StructuredBuffer that GravityLowering.Lower fills from the
zone each frame; GravityLowering.CreateField makes the one quad. GravityFieldParityVerify renders it and compares every
texel with -Zone.GetHeight (max error 0.0059 u), the gradient with GetForce (0.03 degrees) and the wiring.
Deleted: the Gravity Well children of Planet, Gas Giant, Sun and PlanetOutpost prefabs, Wormhole 1's three gravity
children, their overrides in Main Menu and TestScene, five wormhole/wave materials and two brush shaders,
Planet.GravityWellDepth/Radius, BodyData keys 6-7, PlanetSettings.GravityDepth/GravityRadius (and the Settings.asset
keys). Surface Displacement culls layer 9 only. Headless build and tests not yet run at this commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A WellStrength that evaluates negative, NaN or infinite now throws at construction (it built an
antimass well with NaN height). New tests: the core-inside-Roche guard judged at every mass on
curves that cross in the generated range; per-body multipliers move neither strength, core nor
Roche limit; potential and force consistency inside the cores' outer bands (0.5 R to R); a
moving Ship feels the field where it has moved to.
Not built or run locally; verified in the Yggdrasil job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Under the potential-owner field the zone pulls everywhere (about 1.4 u/s^2 even in
its weakest places), so the staged runs bend. Crossing's player sat on the star at the
arena's origin and fell in; its crosser, braked by its pilot, never reached the bow
line from 150 m to port. The crosser now starts 96.5 m to port and 70 m ahead of a
player staged 250 m or more from every body, which keeps the pass at 3.2 s and 40 m
(the test's 3.0-4.2 s and 30-60 m unchanged). Lock On's hostile started on the side
where the zone bends its run by 7 degrees; the lock warning must point within 5 degrees
of where the hostile was staged. It now starts ahead and to port, where the bend is under
one degree and the lock comes at about 14 s. No assertion, limit or gravity is changed.
Not built or run locally; verified in the Yggdrasil job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two hand mutants of GetNormal's K (ignored; times 3) survived because the test compared the ship's
tilt with the same GetNormal it was testing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GravityField is the law of one well (logarithmic tail, softened core).
Zone.GetForce and GetHeight sum it over the bodies; GetNormal derives from
the force for ship tilt; Flow and Ship read GetForce. Well strength, core
radius and Roche limit are PlanetSettings curves of mass, height
exaggeration K one global. Deletes GravityAcceleration, GravityReferenceRate,
GravityStrength, the PowerPulse well, GetNormal's finite differences and
BodyData.GravityDepthExponent. Settings.asset carries the calibrated curves.
Tests: PotentialOwnerTests, the moon's moon placement, retired keys over every
BodyData type, a body-derived giant sweep. Not yet built or run (verify job
follows); this commit may not build.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
UnderALookalikeLieNoGearIsLocated: claimed and true hulls of one type and size with a reactor mount of equal type, position and shape. AMountAimReadsThePositionAxesInOrder and TwoMountsOfOneTypeAreTwoRows use a Wing hull off the diagonal with two Sensors mounts. Assert.Same and DoesNotContain over Entities become ReferenceEquals asserts. The docked-stance test reads IsHostileTo on a hostile-by-default subject instead of _iffOverrides. Test-only; no production change. Not yet built.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The enemy schematic now lists every non-weapon claim as a gear row (cut scanner-catalog), so the old "no row" expectation became "one row, the claimed class, no range". Builds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Format probe showed a Reveal formats as "Reveal" (no graph walk), so no ToString and no test is added (follow-up bearing-markers-test-residuals (2)). Builds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
InspectionScannerCatalogTests and InspectionScannerCommandTests (new); AFinishedScanListsTheGearOnTheSchematic; AScanReadsNoHostility pins pace, emit and a target-side sensor; AScanTakesTheSamePathForAnyPilot uses a Minion in CombatState; an emission heard within Range over the lot's alpha (ping and scan); the docked-subject test is a Theory over true and false; a ReveEal formatting probe. Not yet run: may not build.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GetOrbitPosition places a parent as a side effect of placing its child, so with a child listed first the parent's previous position was already its new one and its Velocity came out zero. The constructor's unread PreviousPosition write is dropped; the first Update takes it from the placed Position. The velocity test now runs moon-first and planet-first packs over varying dt. Builds not verified yet (verify follows).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The five bowl fixtures compare to the zone's own normal, so nothing pinned
the shared well's depth or radius to the bowl (a depth-scaled fixture
mutant survived). TheWellFixtureReproducesTheFormerBowl checks heights
against the bowl's own definition. This commit's test has not run yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A failing Assert.DoesNotContain/Contains over a collection of Entities made xunit format the Entity graph and overflow the stack, aborting the host; the five such assertions now assert a bool. Adds a docked-subject stance test and a pending-paint-on-leaving test. Builds: yes (test-only).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A ship stepped alone drifts under the zone's wells while it idles. Orbits are
now placed at build, so those wells are where the bodies are and the drift is
real; the scenarios (Crossing, FusedRounds) staged geometry the wrong wells
happened to leave undisturbed. Settling is about stats and power, not place.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
OrbitData and BodyData are run-store types; a cache opened with a catalog
only has no home for the well body. The six fixtures that dig a well now
open with a run path. The previous commit does not pass its tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GetHeight sums body wells only. Removes RadialWaves, GasGiant wave fields,
GasGiantData keys 14-17, PlanetSettings Zone*/Wave* fields, the minimap
gravity range, the Boundary brush and Gravity Waves objects from the scene
and both prefabs. The five bowl fixtures become one WellFixture well at the
origin with unchanged expected numbers. Not yet built or tested: this commit
has not been through the Yggdrasil job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Orbit.Position and PreviousPosition are set from GetOrbitPosition at the pack's
time in the constructor, so wells, lights and poses are valid before the first
Update. AddOrbit and CreateOrbit had no caller. Tests not yet run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The hostility stamp (Reveal.Hostile, Contact.Hostile) is deleted: the observer knows its own
stance live. The real fault was the IFF override being dropped when its subject departs while
the book still remembers it; the override now goes when the book drops the subject (ReleaseStance),
so RememberedHostiles reads IsHostileTo. A ship removed from the zone clears its paint inbox and
incoming warnings at the owner, so the Parent == null gate in ActionGameManager is deleted.
Tests pin both stance directions after the act, the departed override hostile, the exclusive
memory window at the equality, and the empty inbox. Unbuilt until the verify job; Unity side not compiled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
RememberedHostiles read the stance the freshest reveal was made under (Reveal.Hostile, surfaced as Contact.Hostile) instead of IsHostileTo now, so a hostile held by the observer's override keeps its marker and the music after it dies or warps, until the bearing lapses. The departure test no longer re-sets the override. New tests: the Detected clause for an in-zone lapsed hostile, position independence for far hostiles on three sides, a within-category lie in ClaimedClasses, and the stale warning of a docked ship. The HUD no longer collects sweeps for a docked ship, and a sweep flash resets the pooled lock indicator's motion.
Untested here: the Unity-side changes (ActionGameManager) have no headless test and no compile has run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
