Your compiler is interpreting the standard correctly. Yes, this is a tricky corner case that many interviewers ask so they appear smarter than they really are.
The route const char[2]
(the formal type of the literal "1"
) to const char*
to bool
is a standard conversion sequence, since it uses exclusively built-in types.
Your compiler must favour that to a user-defined conversion sequence, viz. the std::string
constructor from a const char*
.
The presence of the overload void foo(long long int a)
is a red herring.
You can rather elegantly work around this in C++11 by dropping your overload to bool
, and writing
#include <type_traits>
template <
typename Y,
typename T = std::enable_if_t<std::is_same<Y, bool>{}>
>
void foo(Y)
{
std::cout << "bool" << std::endl;
}
in its place. The compiler will then favour the std::string
for const char[N]
over the template (as that is one of the requirements of overload resolution). Nice!