populating a constexpr array using C++17

Question

I'm trying to initialize a constexpr array using C++17 via the following:

template <size_t N>
struct CStr
{
    static constexpr std::array<int, N> getArr()
    {
        std::array<int, N> a;
        for (auto idx = 0; idx < a.size(); ++idx)
        {
            a[idx] = idx * idx;
        }
        return a;
    }
    static constexpr auto arr { getArr()};
};


int main()
{
    for (const auto &el : CStr<10>::arr)
    {
        std::cout << el << std::endl;
    }
}

However this results in the following compile error about getArr not being a constexpr function. Can someone help explain why ?

<source>: In instantiation of 'constexpr const std::array<int, 10> CStr<10>::arr':
<source>:18:27:   required from 'struct CStr<10>'
<source>:24:35:   required from here
<source>:18:39: error: 'static constexpr std::array<int, N> CStr<N>::getArr() [with long unsigned int N = 10]' called in a constant expression
   18 |     static constexpr auto arr { getArr()};
      |                                 ~~~~~~^~
<source>:9:41: note: 'static constexpr std::array<int, N> CStr<N>::getArr() [with long unsigned int N = 10]' is not usable as a 'constexpr' function because:
    9 |     static constexpr std::array<int, N> getArr()
      |                                         ^~~~~~
<source>:12:9: error: 'goto' is not a constant expression
   12 |         for (auto idx = 0; idx < a.size(); ++idx)
      |         ^~~
Compiler returned: 1   |                                         ^~~~~

You have cut off the error message, it still continues - and contains a hilarious remark about goto :D — dyp
– dyp, Commented Dec 31, 2020 at 11:17
The issue disappears when you initialize std::array<int, N> a{};. clang has a much more intelligible error message for this case :) Also, using -std=c++20 makes the error disappear. — dyp
– dyp, Commented Dec 31, 2020 at 11:18
Thanks dyp! Would you mind letting us know the procedure used to locate the relevant working group paper? — user3882729
– user3882729, Commented Dec 31, 2020 at 11:35
I was a little confused about godbolt.org/z/GM1cEG compiling OK with gcc. However it turns out it does not compile with clang.. Likely a compiler bug w/ the former. — user3882729
– user3882729, Commented Dec 31, 2020 at 17:02

dyp · Accepted Answer · 2020-12-31 11:32:35Z

4

This is a restriction in C++ until C++20, see Permitting trivial default initialization in constexpr contexts: You have to initialize all variables inside of a constexpr function. The Standard gives the following example:

C++17

constexpr int uninit() {
  int a;                  // error: variable is uninitialized
  return a;
}

C++20

constexpr int uninit() {
  struct { int a; } s;
  return s.a;                   // error: uninitialized read of s.a
}

Note that the initialization itself is OK in C++20. Reading the indeterminate value (inside the "uninitialized" object) is not OK.

To fix your issue, initialize the array or switch to C++20:

static constexpr std::array<int, N> getArr()
{
    std::array<int, N> a{}; // now initialized!
    for (auto idx = 0; idx < a.size(); ++idx)
    {
        a[idx] = idx * idx;
    }
    return a;
}

answered Dec 31, 2020 at 11:32

dyp

39.4k13 gold badges119 silver badges190 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

populating a constexpr array using C++17

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related