use new ChainRulesCore correctly

[oxinabox]

No description provided.

[ChrisRackauckas]

Oh god, I need to change all of these?

https://github.com/SciML/DiffEqSensitivity.jl/blob/master/src/concrete_solve.jl#L188

Right now I have both:

https://github.com/SciML/DiffEqBase.jl/blob/master/src/solve.jl#L290-L308

but I'm scared to remove the Zygote.@adjoint because it works. Can I get some help making sure that's all good?

[oxinabox]

Oh god, I need to change all of these?

https://github.com/SciML/DiffEqSensitivity.jl/blob/master/src/concrete_solve.jl#L188

Yeah, those a kinda wrong for ChainRules since they use nothing when they should use NoTangent and Tuple when they should use Tangent{Tuple}.

but I'm scared to remove the Zygote.@adjoint because it works

Yes, right now having the @adjoint will be blocking the the rrule from ever being hit, which is probably why those errors never were detected.
(though Zygote might gracefully accept rrules that violate the convention, Diffractor and Nabla are less likely to)

Can I get some help making sure that's all good?

Yes sure, but after JuliaCon stuff is all submitted.