问题
As I understand, the name of a function itself serves as a pointer to it.
Therefore, when I have a function, I can create a thread by simply passing its address to the thread constructor, like below:
void thread_function {
}
std::thread threadObj1(thread_function);
My confusion is while passing the address of a non-static member function to a thread. For example:
class ClassA
{
public:
void nonstatic_function()
{
}
};
ClassA instance;
std::thread threadObj2(ClassA::nonstatic_function, &instance);
Passing the address of such a function is done in two ways:
ClassA::nonstatic_function
&ClassA::nonstatic_function
Why is there an extra &? If it is indeed needed, then how come the compiler does not complain even without it?
回答1:
As I understand, the name of a function itself serves as a pointer to it.
This behaviour was inherited from C and only applies to free functions or static member functions.
For non-static member functions there is no implicit conversion to pointer-to-member. A pointer-to-member is a fundamentally different thing from a free function pointer because it can't be called without also providing an object instance.
If it is indeed needed, then how come the compiler does not complain even without it?
I guess you are using a compiler with a non-standard extension to allow the & to be omitted in this case. I believe older versions of MSVC allowed it to be omitted; and gcc/Windows defaults to allowing this for MSVC compatibility.
回答2:
std::bind helps you.
ClassA instance;
std::thread threadObj2(std::bind(&ClassA::nonstatic_function, &instance));
来源：https://stackoverflow.com/questions/56553411/why-is-there-an-extra-to-pass-the-address-of-a-non-static-member-function-to-a