Mesh formats and shader adaptation tests

[ragner]

• Mesh formats and shader combination tests
• Shader adaptation tests
• New utils class GVRSceneMaker to create GVRScene from JSON format

We need help with some known issues:

1. Fix for: Shader type is always invalid Samsung/GearVRf#1717 because we have to check material type at our JSON parser.

2. Help to create the right screenshot with the expected results and upload them to github.

3. meshFormatsWithShaderCombinationsTest() :

   • Screenshot of meshes without textures but with color doesn't work.
   • Help to create the expected screenshots for each combination.

4. meshWithoutLightingTest():

   • GVRf looks not working as expected after add the lighting.

5. meshSwitchTexcoordSet():

   • GVRf looks not working as expected after change the textcoords.

GearVRf-DCO-1.0-Signed-off-by: Ragner Magalhaes [email protected]
GearVRf-DCO-1.0-Signed-off-by: Cristiano Bozzetto [email protected]
GearVRf-DCO-1.0-Signed-off-by: Eliamara Silva [email protected]
GearVRf-DCO-1.0-Signed-off-by: Diego Azulay [email protected]

[liaxim]

meshStartWithtTextureTest crashes always for me. Probably a gvrf bug.

Golden masters missing. Could you add some for couple of phones?

[liaxim]

obj.getRenderData().getMaterial().setMainTexture(null); leads to a crash in gvrf. The null must propagate to c++ and the pointer cleared. The Java texture object gets gc-ed (no strong ref to it), takes down the c++ object but we still have a pointer to it in shaderdata.

[liaxim]

meshWithoutLightingTest leads to another crash too. Looks like a problem in the Samsung/GearVRf#1737 pull request I am using. Good tests.

[NolaDonato]

When I run these tests individually they all work. Running them as a group crashes frequently (I am using PR #1743 to run them)

[liaxim]

Planning to merge on Thursday and continue fixing the crashes. There is no fault in the test code from what I can tell. Some of the crashes have been fixed by pending pull requests for the framework.