SPH and Octree AMR Datasets Failing in volume rendering, OAProj, Interactive Data Viz

Originally reported by: chummels (Bitbucket: chummels, GitHub: chummels)

---

This similar to Issues #788, #986, #1008, #1183, but I think this description generalizes and distills all of these problems.

It appears that all particle-based (or Octree-based) datasets are segfaulting when one attempts to do anything involving the KDTree with them. The following three scripts all fail with the same result:

#!python

import yt
ds = yt.load("gizmo_cosmology_plus/snap_N128L16_151.hdf5")
L = ds.arr([1,0,0], 'unitary')
yt.OffAxisProjectionPlot(ds, L, 'density').save()

#!python

import yt
ds = yt.load("gizmo_cosmology_plus/snap_N128L16_151.hdf5")
im, sc = yt.volume_render(ds, 'density', fname='rendering.png')

#!python

import yt
ds = yt.load("gizmo_cosmology_plus/snap_N128L16_151.hdf5")
yt.interactive_render(ds)

These all fail with some variation on this output:

#!python

Failed to split grids.
Failed to split grids.
Segmentation fault

The above scripts each use the publicly available dataset gizmo_cosmology_plus, but one gets similar results by using the same simple commands on any of the SPH datasets on http://yt-project.org/data/ e.g. TipsyGalaxy, snapshot_033, GadgetDiskGalaxy.

Notably, however, when you run the same script with Octree-based AMR outputs, you get similar results. Example datasets that do this are Ramses (e.g. output_00080), and ART (e.g. D9p_500).

These scripts were run on an OS X installation with the tip of yt-dev (843a342ee510).

---

• Bitbucket: https://bitbucket.org/yt_analysis/yt/issue/1212