The IConstraint interface defines an overload where one can pass a actual value by reference.
Nothing in NUnit tests seems to call this as far as I can tell.
There is one overload in DelayedConstraint but that drops the ref prefix and would pass actual by value to the base constraint.
The only place where this might have been intended for is a ref struct that modifies itself, like in:
private struct PretendList
{
public int PollCount { get; private set; }
public int Count
{
get
{
PollCount++;
return 0;
}
}
}Mutable structs are a bad concept unless one can guarantee it is always passed by ref.
The DelayedConstraint calls BaseConstraint.Apply(actual) without ref, it therefore uses copy semantics and the PollCount will always be 0.
Even changing the code to pass ref actual doesn't fix it as the PropertyConstraint called by Has.Count looses the ref.
I wouldn't even know of passing ref to PropertyInfo.GetValue even works.
I suggest to remove the Apply<TActual>(ref TActual) overloads in the next major version as it is technically breaking.
But it is currently not working for any of our constraints as they all behave like Apply<TActual>(TActual) and any mutations are lost.
The
IConstraintinterface defines an overload where one can pass a actual value by reference.Nothing in NUnit tests seems to call this as far as I can tell.
There is one overload in
DelayedConstraintbut that drops therefprefix and would pass actual by value to the base constraint.The only place where this might have been intended for is a
ref structthat modifies itself, like in:Mutable structs are a bad concept unless one can guarantee it is always passed by ref.
The
DelayedConstraintcallsBaseConstraint.Apply(actual)withoutref, it therefore uses copy semantics and thePollCountwill always be0.Even changing the code to pass
ref actualdoesn't fix it as thePropertyConstraintcalled byHas.Countlooses theref.I wouldn't even know of passing
reftoPropertyInfo.GetValueeven works.I suggest to remove the
Apply<TActual>(ref TActual)overloads in the next major version as it is technically breaking.But it is currently not working for any of our constraints as they all behave like
Apply<TActual>(TActual)and any mutations are lost.