`vintagedistribution` encounters an instability for the 2deg TMI

[b-r-hamilton]

The following line runs forever, and then encounters an instability

julia> @time g = vintagedistribution(TMIversion, γ, L, B, 2000, 2010, tmodern = 2015) 
┌ Warning: Instability detected. Aborting
└ @ SciMLBase ~/.julia/packages/SciMLBase/VdcHg/src/integrator_interface.jl:596
ERROR: BoundsError: attempt to access 1-element Vector{Vector{Float64}} at index [2]
Stacktrace:
 [1] getindex
   @ ./essentials.jl:13 [inlined]
 [2] getindex(A::SciMLBase.ODESolution{Float64, 2, Vector{Vector{Float64}}, Nothing, Nothing, Vector{Float64}, Vector{Vector{Vector{Float64}}}, SciMLBase.ODEProblem{Vector{Float64}, Tuple{Float64, Float64}, true, SciMLBase.NullParameters, SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, TMItransient.var"#f#18"{SparseMatrixCSC{Float64, Int64}}, UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}, SciMLBase.StandardODEProblem}, OrdinaryDiffEq.QNDF{5, 12, true, LinearSolve.UMFPACKFactorization, OrdinaryDiffEq.NLNewton{Rational{Int64}, Rational{Int64}, Rational{Int64}, Rational{Int64}}, typeof(OrdinaryDiffEq.DEFAULT_PRECS), Val{:forward}, true, nothing, Nothing, Nothing, NTuple{5, Float64}}, OrdinaryDiffEq.InterpolationData{SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, TMItransient.var"#f#18"{SparseMatrixCSC{Float64, Int64}}, UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing}, Vector{Vector{Float64}}, Vector{Float64}, Vector{Vector{Vector{Float64}}}, OrdinaryDiffEq.QNDFCache{5, StaticArraysCore.SMatrix{5, 5, Float64, 25}, Matrix{Float64}, Vector{Float64}, OrdinaryDiffEq.NLSolver{OrdinaryDiffEq.NLNewton{Rational{Int64}, Rational{Int64}, Rational{Int64}, Rational{Int64}}, true, Vector{Float64}, Float64, Nothing, Float64, OrdinaryDiffEq.NLNewtonCache{Vector{Float64}, Float64, Float64, Vector{Float64}, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, SciMLBase.UJacobianWrapper{SciMLBase.ODEFunction{true, SciMLBase.FullSpecialize, TMItransient.var"#f#18"{SparseMatrixCSC{Float64, Int64}}, UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, Nothing}, Float64, SciMLBase.NullParameters}, SparseDiffTools.ForwardColorJacCache{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 12}}, Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 12}}, Vector{Float64}, Vector{Vector{NTuple{12, Float64}}}, Vector{Int64}, SparseMatrixCSC{Float64, Int64}}, LinearSolve.LinearCache{SparseMatrixCSC{Float64, Int64}, Vector{Float64}, Vector{Float64}, SciMLBase.NullParameters, LinearSolve.UMFPACKFactorization, SparseArrays.UMFPACK.UmfpackLU{Float64, Int64}, LinearSolve.InvPreconditioner{Diagonal{Float64, Vector{Float64}}}, Diagonal{Float64, Vector{Float64}}, Float64, true}}}, Matrix{Float64}, Float64, Float64, StaticArraysCore.SVector{5, Float64}, Vector{Float64}, Vector{Float64}}}, DiffEqBase.DEStats, Nothing}, I::Int64)
   @ SciMLBase ~/.julia/packages/SciMLBase/VdcHg/src/solutions/solution_interface.jl:42
 [3] vintagedistribution(TMIversion::String, γ::Grid, L::SparseMatrixCSC{Float64, Int64}, B::SparseMatrixCSC{Float64, Int64}, t₀::Int64, tf::Int64; tmodern::Int64)
   @ TMItransient ~/Code/TMItransient.jl/src/TMItransient.jl:374
 [4] top-level scope
   @ ./timing.jl:273 [inlined]
 [5] top-level scope
   @ ./REPL[31]:0

[ggebbie]

is it related to Issue #17 ?

The solution was to pin OrdinaryDiffEq to 6.33.3.

I am trying to automate the solution in the Manifest now.

[b-r-hamilton]

Hm this might make sense. I can run this line with no issues in the TMItransient.jl project, which has ODE 6.33.3, but I tried to add TMItransient to a different project and it won't work there.

[ggebbie]

Try updating to latest main branch

[b-r-hamilton]

Revisiting this - I've tried updating to main branch, still throws the error. My current soluiton is to install OrdinaryDiffEq into my package (even though it is not actually needed) and pin it to 6.33.3

[ggebbie]

My attempted solution was to add a compat entry like this: https://github.com/ggebbie/TMItransient.jl/blob/62675cb3ff95a4f165a081ff4f3c2a0354076a50/Project.toml#L22

If that doesn't solve it, here's another idea. Is your environment using DifferentialEquations.jl? I think it may conflict with OrdinaryDiffEq.jl and that only the latter is probably necessary.

[b-r-hamilton]

Hm, I didn't have DiffEq installed, but I did have Sundials installed - removed that and it seems to work fine now. Not 100% that was really the issue, but it's working fine now.