Access violation exception

[anu]

I am trying to update the version of or tools we are using, we are currently at 6.9 and trying to get it to the latest. However, i am receiving a fatal error below 

Fatal error. System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.

   at Google.OrTools.ConstraintSolver.operations_research_constraint_solverPINVOKE.Solver_Solve__SWIG_0(System.Runtime.InteropServices.HandleRef, System.Runtime.InteropServices.HandleRef, System.Runtime.InteropServices.HandleRef)

   at Google.OrTools.ConstraintSolver.operations_research_constraint_solverPINVOKE.Solver_Solve__SWIG_0(System.Runtime.InteropServices.HandleRef, System.Runtime.InteropServices.HandleRef, System.Runtime.InteropServices.HandleRef)

   at Google.OrTools.ConstraintSolver.Solver.Solve(Google.OrTools.ConstraintSolver.DecisionBuilder, Google.OrTools.ConstraintSolver.SearchMonitorVector)

But, if i change the SetSpanUpperBoundForVehicle to a soft constraint to use setcumulvarsoftupperbound, then i do not get the accessviolationexception. Did anyone have a similar issue before? Thank You.

[anu]

Actually, the error happens irrespective of using  setcumulvarsoftupperbound or  SetSpanUpperBoundForVehicle. 

[Mizux Seiha]

Just to be sure using the last version v9.0 in C# (platform will be appreciated) you have some issue when trying to use `SetSpanUpperBoundForVehicle()`

[anu]

Yes, v9.0 in C# (.net core 3.1)

So, I changed the SetSpanUpperBoundForVehicle() to this.solver.GetDimensionOrDie(dimension).CumulVar(this.solver.End(c)).SetMax(limit); and it looks to work without throwing the access violation exception.

But, there is one more path that fails for the same dimension where i set this.solver.GetDimensionOrDie(dimension).SetSpanCostCoefficientForVehicle(coefficient, c);
Also, just as a note this worked with version 6.9 and fails on any version after that.

Thank you for looking into it.

[Mizux Seiha]

I don't remember any .Net sample/test using SetSpanUpperBoundForVehicle() or SetSpanCostCoefficientForVehicle() so a regression since 6.9 and having these methods broken in C# is not surprising...

On my way to do some test and check the SWIG generated file if I can find some clue...

[Mizux Seiha]

Strange using master and hacking constraint_solver/samples/VrpTimeWindows.cs here  https://github.com/google/or-tools/blob/b37d9c786b69128f3505f15beca09e89bf078a89/ortools/constraint_solver/samples/VrpTimeWindows.cs#L160

```cs

        for (int i = 0; i < manager.GetNumberOfVehicles(); ++i) {

            timeDimension.SetSpanUpperBoundForVehicle(22, i);

        }

```

seems to work:

```sh

make rdotnet_VrpTimeWindows

...

Time Elapsed 00:00:00.78

"/usr/bin/dotnet" run --no-build  --project ortools/constraint_solver/samples/VrpTimeWindows.csproj

Objective 71:

Route for Vehicle 0:

0 Time(0,0) -> 9 Time(2,3) -> 14 Time(7,8) -> 16 Time(11,11) -> 0 Time(18,18)

Time of the route: 18min

Route for Vehicle 1:

0 Time(0,0) -> 12 Time(4,4) -> 13 Time(6,6) -> 15 Time(11,11) -> 11 Time(14,14) -> 0 Time(20,20)

Time of the route: 20min

Route for Vehicle 2:

0 Time(2,2) -> 7 Time(4,4) -> 1 Time(8,11) -> 4 Time(10,13) -> 3 Time(16,16) -> 0 Time(24,24)

Time of the route: 24min

Route for Vehicle 3:

0 Time(0,0) -> 5 Time(3,3) -> 8 Time(5,5) -> 6 Time(7,7) -> 2 Time(10,10) -> 10 Time(14,14) -> 0 Time(20,20)

Time of the route: 20min

Total time of all routes: 82min

make[1]: Leaving directory '/home/mizux/work/master'

```

note: On vehicle 2 we can see start is now (2,2) while it was (0,0) before the change...
note2: I've also looked at the ortools/gen/ortools/constraint_solver/RoutingDimension.cs nothing seems strange

So I would say SetSpanUpperBoundForVehicle() is working as expected at least on master branch.

[Mizux Seiha]

Same story for SetSpanCostCoefficientForVehicle() ...

```cs

      for (int i = 0; i < manager.GetNumberOfVehicles(); ++i) {

            timeDimension.SetSpanUpperBoundForVehicle(22, i);

            timeDimension.SetSpanCostCoefficientForVehicle(/*coefficientt=*/1+i, i);

        }

```

```sh

make rdotnet_VrpTimeWindows

...

Time Elapsed 00:00:00.78

"/usr/bin/dotnet" run --no-build  --project ortools/constraint_solver/samples/VrpTimeWindows.csproj

Objective 254:

Route for Vehicle 0:

0 Time(2,2) -> 7 Time(4,4) -> 1 Time(8,8) -> 4 Time(13,13) -> 3 Time(16,16) -> 0 Time(24,24)

Time of the route: 24min

Route for Vehicle 1:

0 Time(1,1) -> 12 Time(5,5) -> 13 Time(7,7) -> 15 Time(12,12) -> 11 Time(15,15) -> 0 Time(21,21)

Time of the route: 21min

Route for Vehicle 2:

0 Time(1,1) -> 9 Time(3,3) -> 5 Time(5,5) -> 6 Time(7,7) -> 2 Time(10,10) -> 10 Time(14,14) -> 0 Time(20,20)

Time of the route: 20min

Route for Vehicle 3:

0 Time(2,2) -> 8 Time(5,5) -> 14 Time(8,8) -> 16 Time(11,11) -> 0 Time(18,18)

Time of the route: 18min

Total time of all routes: 83min

make[1]: Leaving directory '/home/mizux/work/master'

```

note: objective is correctly updated, smaller id vehicle have longer route (since coefficient value is linear to the vehicle id)

last possibility is the nuget package is broken but I can see why...