It is possible to call .c_str() for a C++ const string.
In order to efficiently implement this, it means that it must internally store an extra null character. (Otherwise, copying or modifying data would be required for a .c_str() call)
Therefore a C++ string is always null-terminated, even before calling c_str().
Right? Or wrong?
Yes, it's required from C++11 onwards.
For a std::string s, s[s.size()] must be 0. (Note that the behaviour on attempting to modify that element is undefined.)
Note that s is allowed to contain 0 before that final terminator. In this respect, it differs from a plain C-style string.
Reference: https://en.cppreference.com/w/cpp/string/basic_string
Prior to C++11, the string’s data buffer was NOT required to be null-terminated, or even contiguous. An implementation of c_str() was allowed to copy the data to a separate contiguous null-terminated buffer. However, in practice, no implementation is known to actually have done this, the data buffer was commonly contiguous and null-terminated for efficiency.
Since C++11, the data buffer is required to be contiguous and null-terminated, specifically to keep the implementation of the c_str() and data() methods simple.
In C++11 and later, mystring.c_str() is equivalent to mystring.data() is equivalent to &mystring[0], and mystring[mystring.size()] is guaranteed to be '\0'.