A class is defined like this:
class Singleton {
public:
static Singleton instance_;
private:
Singleton() {
cout << "constructor\n";
}
};
In a function:
Singleton instance = Singleton::instance_;
Code can be compiled and no error was thrown. If I use it like this:
Singleton &instance = Singleton::instance_;
A link error was thrown. I wonder why the first case can be compiled correctly? And I know the constructor function was not called. What's the object state of instance in the first case? Does the first case make any sense?
Both forms are an error in your code, but in both cases compilers/linkers are allowed to silently ignore the error, they're not required to detect it.
Singleton instance = Singleton::instance_; uses the implicitly generated copy constructor. It copies all 0 non-static data members from Singleton::instance_, and is therefore likely to be optimised away entirely. Therefore, the fact that you're using Singleton::instance_ without a definition is likely to go unnoticed.
Singleton &instance = Singleton::instance_; binds to Singleton::instance_ and requires its address to be known. Therefore, it is more likely to result in an error if Singleton::instance_ is not defined.
The way you can provide a definition of Singleton::instance_ is
Singleton Singleton::instance_;
at file scope.
Collected from the Internet
Please contact [email protected] to delete if infringement.
Comments