Possible post optimization error when using a locale with a comma as floating point

[PierreMartinon]

Sample case

using CTDirect
using QML
include("ocp.jl")
sol = solve(ocp1)

The problem is solved correctly but an error occurs after the optimization

Total seconds in IPOPT                               = 1,618

EXIT: Optimal Solution Found.
ERROR: LoadError: MethodError: no method matching +(::Float64, ::Expr)

Closest candidates are:
  +(::Any, ::Any, ::Any, ::Any...)
   @ Base operators.jl:587
  +(::AbstractFloat, ::ForwardDiff.Dual{Ty}) where Ty
   @ ForwardDiff ~/.julia/packages/ForwardDiff/PcZ48/src/dual.jl:145
  +(::AbstractFloat, ::ReverseDiff.TrackedReal)
   @ ReverseDiff ~/.julia/packages/ReverseDiff/p1MzG/src/derivatives/scalars.jl:18
  ...

Stacktrace:
  [1] solve!(solver::NLPModelsIpopt.IpoptSolver, nlp::ADNLPModels.ADNLPModel{…}, stats::SolverCore.GenericExecutionStats{…}; callback::Function, kwargs::@Kwargs{…})
    @ NLPModelsIpopt ~/.julia/packages/NLPModelsIpopt/0YgvC/src/NLPModelsIpopt.jl:249
  [2] solve!
    @ ~/.julia/packages/NLPModelsIpopt/0YgvC/src/NLPModelsIpopt.jl:161 [inlined]
  [3] #ipopt#6
    @ ~/.julia/packages/NLPModelsIpopt/0YgvC/src/NLPModelsIpopt.jl:158 [inlined]
  [4] solve(docp::CTDirect.DOCP; init::Nothing, display::Bool, print_level::Int64, mu_strategy::String, kwargs::@Kwargs{})
    @ CTDirect ~/.julia/packages/CTDirect/wWLhZ/src/solve.jl:96

Removing the using other than CTDirect removes the error.
Same with using Mousetrap. Adding the other 'using' directives present in CTDirect, or changing the order of the using directives, does not change the problem.

[PierreMartinon]

Tracking the error in NLPModelsIpopt.jl l.249

ipopt_output = readlines(ipopt_log_file)
Δt = 0.0
...
for line in ipopt_output
	if occursin("Total seconds", line)
		Δt += Meta.parse(split(line, "=")[2])

So the result of Meta.parse would be Expr instead of a Float ?
Does something in CTDirect.jl (or CTBase) mess up with the julia parser ?
Same error with ocp defined in non-abstract mode

[PierreMartinon]

More tests: works fine with NonlinearSolve (although needs prefixing) (prefixing does not fix the error with the other 2)

using CTDirect
using NonlinearSolve
include("ocp.jl")
sol = CTDirect.solve(ocp1)

[PierreMartinon]

FOUND IT: some packages apparently change the way floating point numbers are displayed, namely with a comma instead of a dot. The resulting expression is not properly parsed as a Float in NLPModelsIpopt...

Total seconds in IPOPT                               = 0,213

vs the usual

Total seconds in IPOPT                               = 1.854

Quick fix: set locale (LC_NUMERIC) to US for instance to prevent the use of the comma. Not very satisfying though.

[PierreMartinon]

Todo: signal the possible issue at NLPModelsIpopt