An extent read from the first or the last asteroid now fails the belt test (finding s3.belt-fixture-first-read). Builds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first fixture hid the big sun's weight in the barycentre (it stood at the origin) and tested reach with two roots only, where nearest and farthest coincide.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
WellsSuperpose is removed (its rule, roots superpose everywhere, is the one multi-root-zone-shape decides); HeightIsThePotentialOverOneK and FlowFieldTests probe inside each well's reach. ZoneWorlds gains Belt and AsteroidSize; fade and ring tolerances are relative to the Plummer pull. Not yet run on dotnet at this commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One zone node at the massive roots' centre of mass owns the only global term (SubzoneReach is the one place the subzone shape is decided: isolated-subzones). WellNode gains Extent and Anchor; a belt orbit's extent is its outermost asteroid. GravityField.SeriesFrom = 0.4 is emitted into the shader template and Gravity.cginc. Tests: the spec's WellLawTests; the old seam test is replaced by TheBandPotentialIsExactInSinglePrecision. GravityFieldParityVerify runs the multi-root fixture (second sun, moon inside the planet's Roche limit, far window): PASS, largest drawn tau 1.228, max error 0.01028 u. The dotnet suite has not been run at this commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Full suite at the previous head: 2003 passed, 1 failed (the seam test, whose 3e-6 was tighter than the closed form's
single-precision floor, 1e-5 measured, 1.5e-5 in the research).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Hand-built zones move from OrbitsFromFieldTests into ZoneWorlds so WellLawTests reuses them. Not yet run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>