Test issue with DFG v.10

[GearsAD]

When I run the tests against masters of DFG and IIF in preparation for v0.10, I see an error:

ArgumentError: start and stop must be finite, got NaN and NaN
Stacktrace:
 [1] _linspace(::Float64, ::Float64, ::Int64) at ./twiceprecision.jl:620
 [2] _range(::Float64, ::Nothing, ::Float64, ::Int64) at ./twiceprecision.jl:616
 [3] #range#43 at ./range.jl:91 [inlined]
 [4] getKDEMax(::BallTreeDensity; N::Int64, addop::NTuple{4,typeof(+)}, diffop::NTuple{4,typeof(-)}) at /home/gearsad/.julia/packages/KernelDensityEstimate/zJ4Va/src/DualTree01.jl:596
 [5] calcVariablePPE(::DistributedFactorGraphs.DFGVariable{DynPoint2}, ::DynPoint2; solveKey::Symbol, method::Type{DistributedFactorGraphs.MeanMaxPPE}) at /home/gearsad/.julia/dev/IncrementalInference/src/FGOSUtils.jl:133
 [6] calcVariablePPE(::DistributedFactorGraphs.DFGVariable{DynPoint2}; method::Type{DistributedFactorGraphs.MeanMaxPPE}, solveKey::Symbol) at /home/gearsad/.julia/dev/IncrementalInference/src/FGOSUtils.jl:151
 [7] setVariablePosteriorEstimates!(::DistributedFactorGraphs.DFGVariable{DynPoint2}, ::Symbol) at /home/gearsad/.julia/dev/IncrementalInference/src/CliqStateMachineUtils.jl:739
 [8] setVariablePosteriorEstimates!(::DistributedFactorGraphs.DFGVariable{DynPoint2}) at /home/gearsad/.julia/dev/IncrementalInference/src/CliqStateMachineUtils.jl:736
 [9] setVariablePosteriorEstimates!(::LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor}, ::Symbol) at /home/gearsad/.julia/dev/IncrementalInference/src/CliqStateMachineUtils.jl:746
 [10] determineCliqIfDownSolve_StateMachine(::CliqStateMachineContainer{BayesTreeNodeData,LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor},LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor},BayesTree}) at /home/gearsad/.julia/dev/IncrementalInference/src/CliqStateMachine.jl:173
 [11] (::StateMachine{CliqStateMachineContainer})(::CliqStateMachineContainer{BayesTreeNodeData,LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor},LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor},BayesTree}; breakafter::Function, verbose::Bool, iterlimit::Int64, recordhistory::Bool) at /home/gearsad/.julia/packages/FunctionalStateMachine/L9Zbf/src/StateMachine.jl:53
 [12] cliqInitSolveUpByStateMachine!(::LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor}, ::BayesTree, ::IncrementalInference.TreeClique; N::Int64, verbose::Bool, oldcliqdata::BayesTreeNodeData, drawtree::Bool, show::Bool, incremental::Bool, limititers::Int64, upsolve::Bool, downsolve::Bool, recordhistory::Bool, delay::Bool, logger::Base.CoreLogging.SimpleLogger) at /home/gearsad/.julia/dev/IncrementalInference/src/CliqStateMachine.jl:1063
 [13] tryCliqStateMachineSolve!(::LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor}, ::BayesTree, ::Int64; verbose::Bool, N::Int64, oldtree::BayesTree, drawtree::Bool, limititers::Int64, downsolve::Bool, incremental::Bool, delaycliqs::Array{Symbol,1}, recordcliqs::Array{Symbol,1}) at /home/gearsad/.julia/dev/IncrementalInference/src/SolveTree.jl:1186
 [14] (::IncrementalInference.var"#354#355"{BayesTree,Bool,Bool,Int64,Int64,Bool,Bool,Array{Symbol,1},Array{Symbol,1},LightDFG{SolverParams,DistributedFactorGraphs.DFGVariable,DistributedFactorGraphs.DFGFactor},BayesTree,Int64})() at ./task.jl:356┌ Warning: printCliqHistorySummary -- No CSM history found.

[dehann]

Hi, that is a numerical solution error on one of the factors. This is a NLsolve.jl or Optim.jl error, when gradient descent ended up at a NaN. Not sure why that is happening in RoME, obviously something is broken. If the test is not a flake then the root cause must be found.