Globe appearing underneath large polygons

[sanjeetsuhag]

This issue may be a subset of #4871, however, since there are multiple issues affecting large polygons - it may be helpful to break them up into smaller parts.

🟡 178Lon-Rhumb in this Sandcastle

This band that appears is showing the near side of the globe

[sanjeetsuhag]

Here's the bounding volumes for the example above:

View 1	View 2

I think the bounding volumes may be playing a part here - even if they do not cause this particular rendering issue, they do seem problematic. For example, here's the rendering of the bounding volumes at certain scales:

62° Longitude Span	63° Longitude Span

[sanjeetsuhag]

Another thing I noticed is the difference in the appearance of this "band" based on the longitudinal extent. In the examples below, there are two polygons: one of extent 179° and 181°. When the extent is >180°, the polygon wraps around the other side.

179°	181°

[sanjeetsuhag]

I forced the use of planar extents by editing the snippet below:

cesium/Source/Scene/GroundPrimitive.js

Lines 804 to 812 in d6aa763

	let usePlanarExtents = true;
	for (i = 0; i < length; ++i) {
	instance = instances[i];
	geometry = instance.geometry;
	rectangle = getRectangle(frameState, geometry);
	if (ShadowVolumeAppearance.shouldUseSphericalCoordinates(rectangle)) {
	usePlanarExtents = false;
	break;
	}

and achieved this result:

@lilleyse I am unfamiliar with how shadow volumes work in CesiumJS - do you think that's the core problem here?

[sanjeetsuhag]

A couple more findings using the presets in this local Sandcastle:

🟡 178Lon-Rhumb-Edge

🟡 178Lon-Rhumb-Precision

Both of these seem to be influenced by the padding done in ShadowVolumeAppearance:

cesium/Source/Scene/ShadowVolumeAppearance.js

Lines 846 to 850 in 7cabc7a

	// Slightly pad extents to avoid floating point error when fragment culling at edges.
	south -= CesiumMath.EPSILON5;
	west -= CesiumMath.EPSILON5;
	north += CesiumMath.EPSILON5;
	east += CesiumMath.EPSILON5;

[ggetz]

I think the bounding volume issue is documented in #7116.

[ggetz]

And I think the rest of the issue is covered by #4746. Closing this as a duplcaite.