jem

That was a very clear explanation. Thank you!
Jem

It seems that my main problems with supporting iOS are lamps that cast shadows (as Yuri did point out). I would rather lose the shadows than lose the amazing looking materials. If I disable dynamic shadows, my tests work on iOS.

To test this idea, I exported the GLTF with and without shadow casting and modified the application logic to load the appropriate GLTF. Here is my code:

var iOS = /iPad|iPhone|iPod/.test(navigator.userAgent) && !window.MSStream;
if (iOS) {
    var URL = '__URL__ios_render_test_ios.gltf';
} else {
        var URL = '__URL__ios_render_test.gltf';
}

I did not need two different bin files because the meshes are the same. Here is a link to the new version of the scene on the CDN.
https://cdn.soft8soft.com/AROAJSY2GOEHMOFUVPIOE:027349a7ae/applications/ios_render_test_mk2/ios_render_test.html

A solution like this is acceptable. I think that it would be better if I would probe the WebGL MAX_TEXTURE_IMAGE_UNITS support rather than inferring it from the agent string, but it will do for now.
Thank you,
Jem

Hi Yuri,
Thank you very much. This is not an urgent request, but it would be a nice feature in some scenes.
-Jem

I think that this function will be useful for the community.

I will experiment with a custom preloader

Thanks again,
Jem

I was able to get the object cloning to work. I will share my solution to this problem here for anyone who needs this technique.

The goal was to create copies of objects at runtime. I needed this because my scenes are assembled from many objects and the quantities of each type object are not known ahead of time. The quantities can be large.

I wrote a small function that uses the Object3D.clone() method from the API. The function takes two arguments. The srcObjectName is the name of the object to be cloned. The name is assigned in Blender. The targetObjectName is the name of the object to which we will snap the clone. Again, the name is assigned in Blender. The function requests a non-recursive copy. I think that this means clones will share meshes and materials (I may be wrong). Lastly, the function returns the clone. This is important because we need to keep track of the clones. In my example, I place all of the clones in an array so that I may programmatically destroy them later. The clones will not be callable from the puzzles.

    function cloneObject(srcObjectName, targetObjectName) {
        var srcObject = v3dApp.scene.getObjectByName(srcObjectName);
        //Create a shallow copy
        var cloneObject = srcObject.clone(false);
        var targetObject = v3dApp.scene.getObjectByName(targetObjectName);
        //Place the copy
        cloneObject.position.x = targetObject.position.x;
        cloneObject.position.y = targetObject.position.y;
        cloneObject.position.z = targetObject.position.z;
        cloneObject.rotation.x = targetObject.rotation.x;
        cloneObject.rotation.y = targetObject.rotation.y;
        cloneObject.rotation.z = targetObject.rotation.z;
        v3dApp.scene.add(cloneObject);
        return cloneObject;
    }

If you look at the Blender screenshot, you will see that the chessboard is covered with an array of empties. I use these empties as my snapping targets. I found that using the arrows type empty is useful to set the snap orientation.

Here is a link to the example on the Soft8soft CDN,
https://cdn.soft8soft.com/AROAJSY2GOEHMOFUVPIOE:027349a7ae/applications/clone_pawn/clone_pawn.html

Cheers,
Jem