Why does integer overflow cause errors with C++ iostreams?

Question

Ok, so I have some problems with C++ iostreams that feels very odd, but it is probably defined behaviour, considering this happens with both MSVC++ and G++.

Say I have this program:

#include <iostream>
using namespace std;

int main()
{
   int a;
   cin >> a;
   cout << a << endl;
   cin >> a;
   cout << a << endl;

   return 0;
}

If I intentionally overflow by giving the first cin a value that is larger than the max limit of an int, all further calls to cin.operator>>() will immediately return for some reason, and a is set to some value. The value seems to be undefined.

Why, and where is this behavior documented? Is there a way to figure out if such an overflow occured?

Also, this similar program seems to work as I intend. If I overflow the value, it will give a some value, and continue on as if the overflow never happened.

#include <cstdio>
using namespace std;

int main()
{
   int a;
   scanf("%d", &a);
   printf("%d\n", a);
   scanf("%d", &a);
   printf("%d\n", a);
   scanf("%d", &a);
   printf("%d\n", a);

   return 0;
}

Answer 1

iostreams is designed to detect errors and enter an error state. You get the same result from integer overflow as from entering a non-numeric string.

Cast cin (or any stream) to bool or check cin.rdstate() to determine if an error has occurred.

Call cin.clear() and cin.ignore() to flush out the error. It will pick up at the point of the characters that failed.

As for the official documentation, the Standard unfortunately gets a bit inscrutable in the bowels of iostreams. See §27.6.1.2.1, 27.6.1.2.2, and 22.2.2.1.1/11 (no kidding):

— The sequence of chars accumulated in stage 2 would have caused scanf to report an input failure. ios_base::failbit is assigned to err.

The documentation for scanf is just as impenetrable, and I'll take it on faith that overflow is supposed to be an error.