Ghost collision in StaticCompoundShape?

[JHPush]

Hi, I'm using Jolt Physics to develop a small ball rolling game.
MeshShape was used to drill a certain hole with NoCollisions on the HeightFieldShape and cover it. (The boundaries of the two meshes were correctly positioned in vertex units.)

However, when the ball rolls from the Height FieldShape to the MeshShape area, GhostCollision occurs where the ball bounces, and the document confirmed that you can use the mEnhanced InternalEdgeRemoval after tying it with CompoundShape, but in my situation, GhostCollision is still occurring.

For the exact question, I've done a similar reproduction on the Sample app of Jolt Physics,
As shown below, the same ghost collimation occurs in the sample app.
Can't HeightFieldShape and MeshShape be grouped into StaticCompound?

What should I do in this case?
I don't know if this is a bug or if it's intended.

Attached is a reproduction video from the Sample app below.
In the Sample app, we simply placed them in order of HeightField - MeshShape - HeightField - MeshShape, which grouped these four shapes into StaticCompoundMesh.
And we reduced the size of the physical object.

The version currently in use is 5.5.1.

Jolt.Physics.Samples.2026-06-09.16-41-50.mp4

[jrouwe]

Can't HeightFieldShape and MeshShape be grouped into StaticCompound?

They can be grouped.

Looking at the video, I see that I didn't document a requirement that I thought was obvious (but isn't): The algorithm can only match triangles that share the same edge (with the same vertices). In your mesh large triangles connect to small triangles, meaning there are T-edges. The algorithm cannot connect these, so the 'enhanced internal edge removal' doesn't work here. I think it should work if you ensure that the tesselation is the same on both sides.

[JHPush]

Thank you for your quick reply. As you said, I have the same tessellation on both sides, so the Enhanced InternalEdgeRemoval processing seems to work well.