(T2)(new T1()) type casting puzzle

[Xyncgas]

If I cast T1 to T2, which T1 is derived from, now the object is in a narrow-er type, I am aware that the actual type is still T1, you are just pretending it as T2.
My question is :if you store this variable as T2, after a while will this object's data that's not visible right now in a narrow-er type be deleted? or it will remain T1 forever and you can get your data back (if there's any inside T1 thatsn't from T2) by casting back to T1

[gfoidl]

or it will remain T1 forever

That's the case.
You created a T1, so it is a T1.
With e.g. T2 obj = (T2)(new T1()) the variable obj just holds the reference to object of type T1 created on the heap. The machine executing your code doesn't care which type obj is. This is "only" for the language and the compiler to let have you type-safeness. I.e. with T2 obj you are allowed to perform operations on obj that are declared on T2.

As obj is actually a T1 you can safely cast it back, and access further member declared on T1.

[Xyncgas]

It's just a bit counter instinct from c++ point of view, if I casted a pointer of an object to a narrow-er pointer, how does the garbage collect even know the actual size of the object. well I guess it's possible, maybe object is keeping track of it.

[Xyncgas]

dotnet is smart

[svick]

@Xyncgas From a C++ point of view, similar code would still work, if you defined a virtual destructor in the base class. The difference is that in C#, it always works.

[gfoidl]

@Xyncgas also note that from a .NET point of view -- especially from the runtime part -- each object has a header, where additional info is stored like actual type, info for GC, etc.
For more info see e.g. Managed object internals, Part 1. The layout.

[huoyaoyuan]

You are confusing with objects and references.

With C++/CLI syntax, you are not casting T1 to T2. You are casting T1^ to T2^ instead, which are similar to T1* and T2*, but tracked by GC.