Fix for material_name = "" bug

[wxmerkt]

Found this while debugging mesh loading - if the material name is empty, a segfault happened. This also defaults to the default material.

[rhaschke]

The material shouldn't be empty actually, so I assume there is a deeper bug.
Can you provide a URDF that caused this issue?

[wxmerkt]

Hi @rhaschke,
The way I triggered this bug was by switching to a PreFabType entity in createEntity. While I am now pursuing a new fix to #1461 by replacing the meshes, my first attempt is in this branch:

melodic-devel...wxmerkt:wxm-smooth-spheres

To trigger the bug, I loaded a URDF that had a mesh as the visual and a sphere as the collision element. Without this check for empty(), we would trigger a segmentation fault using the above branch with the Fix commit reverted when it tried to setMaterial on the submesh.

[rhaschke]

If I understand you correctly, the issue is only observed in your branch to solve #1461, but not in the current master. Hence, I think your branch doesn't set the materials correctly.
If you follow another solution now (modifying sphere.mesh), I think this PR becomes obsolete.

[wxmerkt]

I am now following another solution, so this won't be triggered. However, with material = "", the current master results in a segfault. Wouldn't be defaulting to the default material instead of a broken pointer resulting in a segmentation fault be a safer way to handle things?

[rhaschke]

The segfault is an indicator that something is wrong with the material handling.
Catching this error will just hide the actual problem.

[wxmerkt]

Okay, I understand your reasoning. I still believe that if we can trigger this by using OGRE native features (PrefabType), this may happen in user plug-ins or with malformed meshes/OGRE meshes. A segmentation fault here is an indicator of a pointer being broken - and the other way to fix it could be to catch that or check it.

[rhaschke]

As you said: A segfault indicates a malformed plugin (which doesn't comply to the rules of rviz)
and my reasoning is that this bug should be fixed there then.

[wxmerkt]

I understand. I tend to catch/avoid/fix segfaults and present a helpful message/error/warning (which this PR did not do - but may consider a fix along those lines in the future). If a plugin can trigger undefined/uncaught behaviour, that's usually bad.