Fixed domains not working with <use_normals>

[akenmorris]

This is a new issue with the newer fixed domains code where we not keep the images around for the fixed domains.

The problem is that the MeshBased code later in initialization wants to query the gradient to establish the base normals for the fixed domains. But we no longer have the images at this point.

To fix it and keep the memory savings in the new fixed_domain code, we would need to restructure the initialization so that the AddParticle events take place as the images are added, rather than as a separate stage. This will require a significant refactoring of the code.

[sheryjoe]

And these are the gradients for the not-fixed domains, i.e, the new shape samples? How about the normals of the given fixed domains? Do we read in particles with their normals?

[akenmorris]

These are for the fixed domains. For non-fixed domains, the images will all be there.

I'm aware that some feature files are written, but I don't believe there is any mechanism to read them.