static constexpr初始化数组遇到的问题

template<>
struct params {

typedef uint16_t value_type;
typedef int16_t signed_value_type;
typedef uint32_t greater_value_type;

typedef value_type* poly_t;

// The moduli used in each 16 bit block (14 bits long each)
// They are of the form p = 2*14 - i2*kMaxPolyDegree + 1 for increasing i
static constexpr unsigned int kMaxNbModuli = 2;
static constexpr value_type P[kMaxNbModuli] = { 15361U, 13313U};//有问题

// The associated lower word of their Newton quotients
static constexpr value_type Pn[kMaxNbModuli] = { 17458U, 60470U};//有问题

static constexpr unsigned int kModulusBitsize = 14;
static constexpr unsigned int kModulusRepresentationBitsize = 16;

// A primitive 2*kMaxNbModuli (i.e., 2**10) root of unity for each one of the moduli
static constexpr value_type primitive_roots[kMaxNbModuli] = { 4989U, 10076U};//有问题

// Inverses of kMaxPolyDegree (for the other degrees it can be derived easily)
// for the different moduli
static constexpr value_type invkMaxPolyDegree[kMaxNbModuli] = { 15331U, 13287U};//有问题

// Polynomial related data
static constexpr unsigned int kMaxPolyDegree = 512;
};

vs2015编译时遇到问题：
（代码中标注//有问题的语句都是下面形式的error，我把代码改成 static constexpr int P[2] = { 1，1};也是这样的error）
error C4579: 'nfl::params::P': in-class initialization for type 'const nfl::params::value_type [2]' is not yet implemented; static member will remain uninitialized at runtime but use in constant-expressions is supported
error C2131: 表达式的计算结果不是常数
note: 遇到非常量(子)表达式

static 常量成员，除完全用于类内初始化的整型（kMaxNbModuli ），其余必须类内声明初始化，类外定义。

#include <cstdint>

struct params
{
    using value_type = uint16_t;
    using signed_value_type = int16_t;
    using greater_value_type = uint32_t;
    using poly_t = value_type *;

    // The moduli used in each 16 bit block (14 bits long each)
    // They are of the form p = 2*14 - i2*kMaxPolyDegree + 1 for increasing i
    static constexpr unsigned int kMaxNbModuli = 2;
    static constexpr value_type P[kMaxNbModuli] = {15361U, 13313U};

    // The associated lower word of their Newton quotients
    static constexpr value_type Pn[kMaxNbModuli] = {17458U, 60470U};
    ;

    static constexpr unsigned int kModulusBitsize = 14;
    static constexpr unsigned int kModulusRepresentationBitsize = 16;

    // A primitive 2*kMaxNbModuli (i.e., 2**10) root of unity for each one of
    // the moduli
    static constexpr value_type primitive_roots[kMaxNbModuli] = {4989U, 10076U};

    // Inverses of kMaxPolyDegree (for the other degrees it can be derived
    // easily) for the different moduli
    static constexpr value_type invkMaxPolyDegree[kMaxNbModuli] = {15331U,
                                                                   13287U};

    // Polynomial related data
    static constexpr unsigned int kMaxPolyDegree = 512;
};

constexpr params::value_type params::P[kMaxNbModuli];
constexpr params::value_type params::Pn[kMaxNbModuli];
constexpr params::value_type params::primitive_roots[kMaxNbModuli];
constexpr params::value_type params::invkMaxPolyDegree[kMaxNbModuli];

int main()
{
    params a;
    return 0;
}

这个是不是要把定义和声明分开写啊

是不是结构里不允许这样子初始化数组啊

如果我把static constexpr value_type P[kMaxNbModuli] = { 15361U, 13313U};这里改成value_type P[kMaxNbModuli] = { 15361U, 13313U};是可以的，但是另一段代码就会出现另外的问题：
namespace nfl {
constexpr typename params::value_type params::P[];//error C2761: nfl::params::value_type nfl::params::P[2]”: 不允许成员函数重新声明
constexpr typename params::value_type params::Pn[];
constexpr typename params::value_type params::primitive_roots[];
constexpr typename params::value_type params::invkMaxPolyDegree[];
}

你肯定是别的地方有问题，建议你查看一下params 泛化的声明
如果要使用模板，最好系统的学习一下它，毕竟它是c++比较复杂的部分