Question

I was wondering is it safe to return a fixed pointer from one method to another method – does the pinned object still stay fixed? e.g.

struct TestData
{
    public int value;
}

public class Class1
{
    private TestData data; 

    public unsafe TestData* GetDataPtr()
    {
        fixed (TestData* ptr = &data)
        {
            // IS THIS SAFE?
            return ptr; 
        }
    }
}

public class Class2
{
    public unsafe void Test()
    {
        Class1 x = new Class1();

        TestData* ptr = x.GetDataPtr(); // is this still fixed?
        ptr->value = 2; 
    }
}

The reason i’m asking is using something in this style I’ve been getting AccessViolation exceptions. But since I changed it to e.g. set value direct from Class1 i haven’t seen the issue occur.

EDIT:
the reason i thought it may be fixed is if from outside the “TestData* ptr = x.GetDataPtr()” you try to put fixed( ) you get “you cannot take the address of an already fixed expression”. .. i get it now though it’s speaking about my var “ptr” in “Test()” already being fixed..

Answer 1

Returning from the method ends the scope of fixed, hence the pointer is no longer fixed once you return. It is safe to pass fixed pointers up the chain, e.g.

fixed(TestData* ptr = &data) {
    MyFunction1(ptr);
    MyFunction2(ptr);
}

But returning makes the pointer non-fixed again.

Logically, this makes sense: there is no alternative way for the CLR to decide when the pointer should become non-fixed after you have returned it from the call.