solve callback parameter ignored by NLopt solvers

[mottelet]

Callback solve parameter seems to be ignored when using NLopt solvers (Optim.jl solvers seem OK):

using GalacticOptim, NLopt, Optim

rosenbrock(x,p) =  (p[1] - x[1])^2 + p[2] * (x[2] - x[1]^2)^2
function mycallback(x,l)
    display(l)
    return false
end
x0 = zeros(2)
p  = [1.0,100.0]

fun = OptimizationFunction(rosenbrock, GalacticOptim.AutoForwardDiff())
prob = OptimizationProblem(fun,x0,p)
for algo in (Optim.BFGS(), NLopt.LD_LBFGS(), NLopt.LD_TNEWTON)
    display(algo)
    solve(prob,algo,cb=mycallback);
end

BFGS{LineSearches.InitialStatic{Float64}, LineSearches.HagerZhang{Float64, Base.RefValue{Bool}}, Nothing, Nothing, Flat}(LineSearches.InitialStatic{Float64}
 ...
1.0
0.8415971436977181
0.6518877381414275
0.4302799075988325
0.30035641273635594
0.2529701829959191
0.059753770420490335
0.031050735596379386
0.009988655287113285
0.005532996435004709
0.002062241207605042
0.0004268680595462597
3.489760336773075e-5
7.427113352970077e-7
2.4211861160515415e-9
6.905104953391281e-14
7.645684434510211e-21
NLopt.LD_LBFGS: Limited-memory BFGS (L-BFGS) (local, derivative-based)
NLopt.LD_TNEWTON: Truncated Newton (local, derivative-based)

[ChrisRackauckas]

NLopt.jl doesn't have a callback system so we'll need to hook in ourselves at the function call, i.e. https://github.com/SciML/GalacticOptim.jl/blob/v2.2.0/src/solve/nlopt.jl#L96 right there after the loss evaluation put the call to the callback.