Support assimp 3.1.1

[esper]

Not possible to run with assimp 3.1.1 instead of 3.0. Reported by issue #118

Source used (Pacman for Arch):
http://sourceforge.net/projects/assimp/files/assimp-3.1/

Changes in 3.1.1:
https://github.com/assimp/assimp/blob/master/CHANGES

Arch and Ubuntu 15.04:

chaoticrage: /assimp/assimp-3.1.1/code/Importer.cpp:738: const aiScene* Assimp::Importer::ApplyPostProcessing(unsigned int): Assertion `_ValidateFlags(pFlags)' failed.
Aborted (core dumped)


[TheJosh]

We're probably using a post processing flag which has been removed, or which isn't valid in some cases.

[TheJosh]

Might be this
assimp/assimp@e6c5095

We'll have to check that all of our postprocess flags are valid

[esper]

The assertion is due to that we use both 'aiProcess_OptimizeGraph' and 'aiProcess_PreTransformVertices'. Also the zombies are not visible any more with assimp 3.1.1...

From Importer.cpp

// Validate post process step flags 
bool _ValidateFlags(unsigned int pFlags) 
{
    if (pFlags & aiProcess_GenSmoothNormals && pFlags & aiProcess_GenNormals)   {
        DefaultLogger::get()->error("#aiProcess_GenSmoothNormals and #aiProcess_GenNormals are incompatible");
        return false;
    }
    if (pFlags & aiProcess_OptimizeGraph && pFlags & aiProcess_PreTransformVertices)    {
        DefaultLogger::get()->error("#aiProcess_OptimizeGraph and #aiProcess_PreTransformVertices are incompatible");
        return false;
    }
    return true;
}

[TheJosh]

Note that the issue with zombies not rendering appears to be only related to the Zombie model. Other models are rendering fine - even other models with animations.

[TheJosh]

When I run assimp info on the zombie.dae model using version 3.0 I get lots of info including:

Minimum point      (-1.761364 -0.027560 -3.411929)
Maximum point      (1.181818 18.128241 3.269613)
Center point       (-0.289773 9.050341 -0.071158)

assimp 3.1.1 gives this instead:

Minimum point      (-0.001761 -0.000028 -0.003412)
Maximum point      (0.001182 0.018128 0.003270)
Center point       (-0.000290 0.009050 -0.000071)

It appears to think that the model is much smaller than it actually is. This might be a regression.

[TheJosh]

I might be this assimp/assimp@d411cbfbf4a5f0