C++中static_assert放置位置不同的行为差异问题咨询

在C++的CRTP（奇异递归模板模式）实现中，将static_assert放在类的不同位置会导致完全不同的编译结果，以下是具体现象和原因解析：

第一种实现（可正常编译运行） #

#include <iostream>
#include <string>
#include <type_traits>

template <typename, typename T>
struct has_f
{
    static_assert(
        std::integral_constant<T, false>::value,
        "Second template parameter needs to be of function type.");
};
template <typename C, typename Ret, typename... Args>
struct has_f<C, Ret(Args...)>
{
private:
    template <typename T>
    static constexpr auto check(T *)
        -> typename std::is_same<
            decltype(std::declval<T>().f(std::declval<Args>()...)),
            Ret>::type;
    template <typename>
    static constexpr std::false_type check(...);
    typedef decltype(check<C>(0)) type;

public:
    static constexpr bool value = type::value;
};

// T must have a method called f
template <class Derived>
class BaseCRTP
{

public:
    int f(const std::string &s)
    {
        std::cout << "not optimized" << std::endl;
        static_assert(has_f<Derived, int(const std::string &)>::value, "Derived must have a method called f");
        return static_cast<Derived *>(this)->f(s);
    }

    void test()
    {
        f("Fsafas");
    }
};

class Derived : public BaseCRTP<Derived>
{
public:
    int f(const std::string &s)
    {
        std::cout << "Derived::f()" << std::endl;
        return 0;
    }
};

int main()
{
    Derived d;
    d.test();
    return 0;
}

运行结果：代码可正常编译，运行后输出两行内容：

not optimized
Derived::f()

第二种实现（编译失败） #

#include <iostream>
#include <string>
#include <type_traits>

template <typename, typename T>
struct has_f
{
    static_assert(
        std::integral_constant<T, false>::value,
        "Second template parameter needs to be of function type.");
};
template <typename C, typename Ret, typename... Args>
struct has_f<C, Ret(Args...)>
{
private:
    template <typename T>
    static constexpr auto check(T *)
        -> typename std::is_same<
            decltype(std::declval<T>().f(std::declval<Args>()...)),
            Ret>::type;
    template <typename>
    static constexpr std::false_type check(...);
    typedef decltype(check<C>(0)) type;

public:
    static constexpr bool value = type::value;
};

// T must have a method called f
template <class Derived>
class BaseCRTP
{

public:
    int f(const std::string &s)
    {
        std::cout << "not optimized" << std::endl;
        return static_cast<Derived *>(this)->f(s);
    }

    static_assert(has_f<Derived, int(const std::string &)>::value, "Derived must have a method called f");

    void test()
    {
        f("Fsafas");
    }
};

class Derived : public BaseCRTP<Derived>
{
public:
    int f(const std::string &s)
    {
        std::cout << "Derived::f()" << std::endl;
        return 0;
    }
};

int main()
{
    Derived d;
    d.test();
    return 0;
}

编译报错信息：

crtp.cc:76:5: error: static_assert failed due to requirement 'has_f<Derived, int (const std::string &)>::value' "Derived must have a method called f"
    static_assert(has_f<Derived, int(const std::string &)>::value, "Derived must have a method called f");
    ^             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
crtp.cc:84:24: note: in instantiation of template class 'BaseCRTP<Derived>' requested here
class Derived : public BaseCRTP<Derived>

差异原因解析 #

两种写法的核心差异在于模板实例化时机和C++对不完全类型的限制：

1. 第二种写法编译失败的原因：
   当static_assert直接放在BaseCRTP类的作用域中时，定义Derived类（class Derived : public BaseCRTP<Derived>）时，编译器需要立即实例化BaseCRTP<Derived>。此时Derived是不完全类型——编译器只知道它是一个类，但还没处理到它的成员定义（f方法还没被编译器看到）。因此has_f检测时无法找到Derived::f，导致value为false，触发static_assert失败。

2. 第一种写法可以正常编译的原因：
   当static_assert放在成员函数f内部时，成员函数的实例化是延迟触发的——只有当这个函数被实际调用时，编译器才会实例化它。在main函数中调用d.test()进而调用BaseCRTP<Derived>::f时，Derived的完整定义已经被编译器处理完毕（是完全类型），此时has_f可以正确检测到Derived::f的存在，value为true，断言通过。

内容的提问来源于stack exchange，提问作者Junhui Zhu