Skinned model fix

[lilleyse]

While testing a simple skinned model exported from Blender and then going through the Collada-glTF converter, I noticed a bug with the skeleton generation.

When searching for a joint in searchForest, it loops through the children of the runtime node. The problem is the runtime nodes don't have children yet, since createSkins comes before createRuntimeNodes in createResources.

The solution I have loops over the children of the gltf node instead of the runtime node. Another solution is to change the order of calls in createResources, but I was worried this would create unintended problems. Or maybe the runtime node's children should be set in parseNodes.

The reason this wasn't noticed before is our Cesium guy lists all its joints in the skeletons property, so searchForest never had to loop over children.

[lilleyse]

Tested with: simple_anim.txt

[pjcozzi]

Great catch, thanks @lilleyse!

[pjcozzi]

Please update CHANGES.md.

[pjcozzi]

Can you add a test? Put the .dae/.gltf here: https://github.com/AnalyticalGraphicsInc/cesium/tree/master/Specs/Data/Models

[pjcozzi]

@tparisi @julien-moreau @deltakosh here is a simple skinned glTF model for testing: #3224 (comment)

@julien-moreau @deltakosh this is a glTF 1.0 model. If you need to upgrade your importer from glTF 0.8, see https://github.com/KhronosGroup/glTF/wiki/glTF-0.8-to-1.0-Guide

[pjcozzi]

@lilleyse can you also open a pull request into the glTF repo to include this sample model? See https://github.com/KhronosGroup/glTF/tree/master/sampleModels

[pjcozzi]

Please update CHANGES.md.

Ah, nevermind. I see.

[pjcozzi]

@lilleyse will this help this issue: https://groups.google.com/forum/#!topic/cesium-dev/VAyj40HJzgk

[pjcozzi]

The solution I have loops over the children of the gltf node instead of the runtime node.

OK with me. Given how subtle this is, can you please add a comment?

[lilleyse]

Updated.

[lilleyse]

glTF repo pull request: KhronosGroup/glTF#480

[pjcozzi]

Looks great, thanks!