refactor: tol should map to a relative tolerance, not an absolute one

Author: SebastianM-C

lib/OptimizationIpopt/src/OptimizationIpopt.jl

@@ -106,8 +106,8 @@ function __map_optimizer_args(cache,
     if !isnothing(maxtime)
         Ipopt.AddIpoptNumOption(prob, "max_cpu_time", maxtime)
     end
-    if !isnothing(abstol)
-        Ipopt.AddIpoptNumOption(prob, "tol", abstol)
+    if !isnothing(reltol)
+        Ipopt.AddIpoptNumOption(prob, "tol", reltol)
     end
     if verbose isa Bool
         Ipopt.AddIpoptIntOption(prob, "print_level", verbose * 5)

lib/OptimizationIpopt/test/additional_tests.jl

@@ -93,7 +93,7 @@ using LinearAlgebra
         prob = OptimizationProblem(optfunc, x0, nothing;
                                  lcons = fill(4.0, n-2),
                                  ucons = fill(6.0, n-2))
-        sol = solve(prob, IpoptOptimizer(); maxiters = 1000, abstol=1e-6)
+        sol = solve(prob, IpoptOptimizer(); maxiters = 1000, reltol=1e-6)
 
         @test SciMLBase.successful_retcode(sol)
         @test length(sol.u) == n

lib/OptimizationIpopt/test/advanced_features.jl

@@ -18,7 +18,7 @@ using SparseArrays
 
         # Test with tight tolerances
         sol = solve(prob, IpoptOptimizer();
-                   abstol = 1e-10,
+                   reltol = 1e-10,
                    acceptable_tol = 1e-8,
                    acceptable_iter = 5)