This is a bug. Or should it be?

[uselessgoddess]

Here is my code

#include <concepts>

auto foo(auto x)
{
   if constexpr (std::same_as<int, decltype(x)>)
   {
      return 1;
   }
   return 2;
}

int main()
{
	foo(1);
        foo(1.0);
}

Here is the C++ Insights translation

#include <concepts>

auto foo(auto x)
{
   if constexpr (std::same_as<int, decltype(x)>)
   {
       return 1;
   }
   return 2;
}

#ifdef INSIGHTS_USE_TEMPLATE
template<>
int foo<int>(int x)
{
  if constexpr(std::same_as<int, int>) {
    return 1;
  } 
  
  return 2;
}
#endif


#ifdef INSIGHTS_USE_TEMPLATE
template<>
int foo<double>(double x)
{
  if constexpr(std::same_as<int, double>) {
  } 
  
  return 2;
}
#endif


int main()
{
  foo(1);
  foo(1.0);
}

Why

if constexpr(std::same_as<int, int>) {
    return 1;
} 

But

if constexpr(std::same_as<int, double>) {
} 

If C++ Insights doesn't expand constexpr expressions into their values, then why does the second option look so strange?

[andreasfertig]

Hello @uselessgoddess,

thanks for reporting this issue :-) It made me (re-)discover ConstantExpr and its value. A fix is on its way.

Andreas