How would you improve this algorithm? (c string reversal)

Question

Working through some programming interview challenges I found online, I had to write an algorithm to reverse a const char * and return a pointer to a new char *. I think I have it, but to make it work properly I had to do some wonky stuff - basically having to account for the null-terminating character myself. Somehow I feel this is wrong, but I'm stumped, and I was wondering if someone could help me out:

char * reverse(const char * str)
{
  int length = strlen(str);
  char * reversed_string = new char[length+1];

  for(int i = 0; i < length; ++i)
  {
    reversed_string[i] = str[(length-1) - i];
  }
  //need to null terminate the string
  reversed_string[length] = '\0';

  return reversed_string;

}

int main(int argc, char * argv[])
{

  char * rev_str = reverse("Testing");

  cout << "Your string reversed is this: " << rev_str << endl;

  delete rev_str;
  rev_str = 0;

  return 0;
}

Answer 1

std::reverse from <algorithm> works for strings and char arrays:

string str = "Hello";
char chx[] = "Hello";

reverse(str.begin(), str.end());
reverse(chx, chx + strlen(chx));

cout << str << endl;
cout << chx << endl;

/EDIT: This, of course, modifies the original string. But STL to the rescue. The following creates a new reversed string. Unfortunately (?), this doesn't work directly on C char arrays without creating an additional (implicit) copy:

string reverse_string(string const& old) {
    return string(old.rbegin(), old.rend());
}

cout << reverse_string("Hello") << endl;