in which situations anonymous enum should be used?

In C (but not in C++), enum can be [ab]used to define int constants.

For example, given this declaration:

const int MAX = 1024;

MAX is not a constant expression, it's the name of a read-only object. That means you can't use it in a case label, as the size of an array declared at file scope or with static, or in any other context requiring a constant expression.

But if you write:

enum { MAX = 1024 };

then MAX is a constant expression of type int, usable in any context where you could use the constant 1024.

Of course you could also write:

#define MAX 1024

but there are disadvantages to using the preprocessor: the identifier isn't scoped the way it would be given an ordinary declaration, for example.

The drawback is that such a constant can only be of type int.

C++ has different rules; enumeration constants are of the enumerated type, not int, but you can use declared constant objects as constant expressions (as long as the initializer is a constant expression).

To address the original question, when you use an enum declaration to create constants like this, there's no point in having either a tag or a typedef, since you'll never use the type itself.

Background: This:

enum foo { zero, one, two };
enum foo obj = two;

creates a type enum foo and constants zero, one, and two. In C, the constants are always of type int, which is admittedly odd, and the initialization of obj involves an implicit conversion from int to enum foo.

In C++, the type enum foo can also be referred to as just foo, and the constants are of type enum foo (which is compatible with some integer type, not necessarily int).

Another use case is as an element of a struct or union, typically when it doesn't make sense to use it by itself (because it's there solely to satisfy the ABI of a communication protocol or etc. and there is a more appropriate representation for programmatic use).