Isaac Wheeler

It doesn't just involve the axis limits either; something odd is happening to the values themselves. ```julia plot(0:25, 10.0 .^(-25:0), yscale=:log10, ylims=(1e-25, 1)) ``` produces 

Nothing too fancy here, I think: ``` julia> collect(10.0 .^ (-25:0)) 26-element Vector{Float64}: 1.0e-25 1.0e-24 1.0e-23 1.0e-22 1.0e-21 1.0e-20 1.0e-19 1.0e-18 1.0e-17 1.0e-16 1.0e-15 1.0e-14 1.0e-13 1.0e-12 1.0e-11 1.0e-10 1.0e-9...

Looks like the culprit is here, from the error: https://github.com/matthieugomez/LeastSquaresOptim.jl/blob/10465c1ba1d50b72b3a16dfff43aae62a8ee8399/src/optimizer/dogleg.jl#L111-L115 The Git blame indicates that was previously updated for Julia 0.7, perhaps something about LinearAlgebra changed with 1.0?

Any progress here?

Just to add to the above list of PRs for v7: #2602

Has Plots moved far enough towards 2.0 that it would be difficult to revive this PR? Time is limited as always but this seems like a functionality related to eventually...

Turns out that, on master at least, making the callback a `const` resolves this type instability. That is, augmenting the above MWE: ```julia const constcallback = ContinuousCallback((u,_,__)->u[1]-1, terminate!) function eval_prm_ccb(pr)...

In my code base (which has some more intricacies), I'm still seeing something that isn't type inferrable. I'll make another issue later if I manage to identify where the problem...

No, not currently (didn't know that was an option). My problem only has two equations, so I haven't been worried on that front, but I have a parameter struct full...

I can replicate my problem without Unitful; here is an example which resembles my code base, without using Unitful: ```julia using OrdinaryDiffEqRosenbrock using LinearAlgebra: Diagonal using Accessors # using OrdinaryDiffEqNonlinearSolve...