Will Tebbutt

I would be more inclined to suggest using either forwards- or reverse- differencing around discontinuities. i.e you set up your tests so that finite differences avoids the discontinuities. See eg....

> For example, does this in anyway require JuliaDiff/FiniteDifferences.jl#97 Hmm this seems like a separate issue from checking the numerical correctness of the gradients. Rather, this would require some functionality...

Yeah, it feels fine (arguably quite natural) to make `ChainRulesTestUtils` a test-time dep for `ChainRulesCore` -- arguably all we're doing with CRTU is keeping a `TestUtils` module outside of `ChainRulesCore`...

Indeed. We could also test that it's an appropriate differential type for the primal in question (we've nearly got that functionality from `rand_tangent`).

Update: `eigen` was resolved by #16. Just leaves implementing the `cholesky` in a similar manner.

I had a working prototype a while ago i think. Will dig it out and try to finish it off + make a PR in the next few days.

Required a bit of work to get anything off the ground, so make #44 .

Ahhh okay, thanks for letting me know. Will leave this open until that happens then.

Does this mean that the tex and ruby stuff will disappear also?

Oh, they've still not released the proceedings?