[Bug c++/102780] Checking constraints using large fold expression is slow

["redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102780

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
In the real code, I find that a separate constrained specialization like this:


template<typename T, typename... U>
union variadic_union<T, U...>
{
  T first;
  variadic_union<U...> rest;

  static constexpr int size = variadic_union<U...>::size + 1;
};

template<typename T, typename... U>
requires (!trivially_destructible<T, U...>)
union variadic_union<T, U...>
{
  variadic_union(const variadic_union&) = default;
  variadic_union(variadic_union&&) = default;
  variadic_union& operator=(const variadic_union&) = default;
  variadic_union& operator=(variadic_union&&) = default;

  // Non-trivial dtor is required for this partial specialization
  constexpr ~variadic_union()
  { }

  T first;
  variadic_union<U...> rest;

  static constexpr int size = variadic_union<U...>::size + 1;
};


Is much faster (14s instead of 20s+) than a constrained destructor in the
primary template:

template<typename T, typename... U>
union variadic_union<T, U...>
{
  variadic_union(const variadic_union&) = default;
  variadic_union(variadic_union&&) = default;
  variadic_union& operator=(const variadic_union&) = default;
  variadic_union& operator=(variadic_union&&) = default;

  ~variadic_union() = default;

  // Conditionally non-trivial dtor, if required.
  constexpr ~variadic_union() requires (!trivially_destructible<T, U...>)
  { }

  T first;
  variadic_union<U...> rest;

  static constexpr int size = variadic_union<U...>::size + 1;
};


I haven't been able to reproduce that time difference in the reduced examples
though.