fix: stop divide by zero warning in LU solvers

[wd15]

Prompted by this.

[wd15]

Test failures seem unrelated to changes

[guyer]

Should this go in #791?

[wd15]

Done

[wd15]

VTK failure on Travis is probably version incompatibility with Numpy.

[guyer]

I agree, the VTK errors are safe to ignore

[guyer]

I don't understand why the polyxtal example is stalling with SciPy on Py3k

[wd15]

I don't understand why the polyxtal example is stalling with SciPy on Py3k

Is this the first time you've seen that?

[guyer]

I can't remember a stall like this for a serial run. I was wondering if the error was somehow negative, so that the <= test would fail, but I don't think that's possible and I don't know why that would make it stall.

[wd15]

I can't remember a stall like this for a serial run. I was wondering if the error was somehow negative, so that the <= test would fail, but I don't think that's possible and I don't know why that would make it stall.

Not to worry. I'll try and debug it at some point. No rush to finish this.

[guyer]

polyxtal is stalling with SciPy on other branches. This doesn't have anything to do with it.

[wd15]

@guyer I merged from master and tests are failing for various reasons seemingly not related to the changes.

[guyer]

@guyer I merged from master and tests are failing for various reasons seemingly not related to the changes.

Weird. Looks like a slew of unrelated failures. Should be a joy to debug.

[guyer]

Looks like a lot of things started failing about 5 months ago when I relaxed the installation constraint on Gmsh from "gmsh<4.0" to just gmsh. (see #820)