ShadowReceiverGlsl

Reusable GLSL fragments concatenated into every shadow receiver's fragment shader.

Receiver vertex shader must emit:

• varying vec3 worldPos — camera-relative world position (worldVertex - cameraPoint), computed from uniforms the drawable composed against the camera point in double precision on the CPU. A raw ECEF position quantizes to ~0.5 m in a float32 varying — enough to erase street-scale shadow structure entirely.

• varying float viewDepth — positive distance to the camera plane (gl_Position.w)

Receiver contract - one lighting model engine-wide (see earth.worldwind.render.program.LightingGlsl): shadows attenuate only the direct sun term, never the ambient term, so material Lambert and the cascade term can't double-darken the same fragment.

• Lit receivers call shadowLitFactor(lambert, worldPos, viewDepth, worldNormal) - the whole lighting multiplier with the sun term shadow-attenuated. Emitted when fragmentDeclarations is built with lit = true; requires LightingGlsl.DECLARATIONS earlier in the shader. A zero normal degrades to depth-only sampling.

• Unlit receivers (terrain imagery, photogrammetry, points) call shadowAlbedoFactor(worldPos, viewDepth) - the shadow term is their only sun shading, so it darkens the whole albedo toward ambientShadow.

• shadowSunVisibility overloads expose the raw [0, 1] visibility; the normal-aware overload adds a normal-offset receiver bias scaled by the active cascade's texel world size. All short-circuit to 1.0 when applyShadow is false.

Occlusion resolve is a single hardware depth-compare tap by default (its free 2x2 bilinear PCF softens the edge by one texel), or a bilinear-weighted 3x3 nine-tap kernel when DrawContext.SOFT_SHADOWS is set. Stable under camera motion thanks to the cascade texel snapping.