Wrong Floyd

Valera conducts experiments with algorithms that search for shortest paths. He has recently studied the Floyd's algorithm, so it's time to work with it.

Valera's already written the code that counts the shortest distance between any pair of vertexes in a non-directed connected graph from n vertexes and m edges, containing no loops and multiple edges. Besides, Valera's decided to mark part of the vertexes. He's marked exactly k vertexes a₁, a₂, ..., a_k.

Valera's code is given below.

ans[i][j] // the shortest distance for a pair of vertexes i, j
a[i]  // vertexes, marked by Valera

for(i = 1; i <= n; i++) {
    for(j = 1; j <= n; j++) {
        if (i == j)
            ans[i][j] = 0;
        else
            ans[i][j] = INF;  //INF is a very large number 
    }
}    

for(i = 1; i <= m; i++) {
    read a pair of vertexes u, v that have a non-directed edge between them;
    ans[u][v] = 1;
    ans[v][u] = 1;
}

for (i = 1; i <= k; i++) {
    v = a[i];
    for(j = 1; j <= n; j++)
        for(r = 1; r <= n; r++)
            ans[j][r] = min(ans[j][r], ans[j][v] + ans[v][r]);
}

Valera has seen that his code is wrong. Help the boy. Given the set of marked vertexes a₁, a₂, ..., a_k, find such non-directed connected graph, consisting of n vertexes and m edges, for which Valera's code counts the wrong shortest distance for at least one pair of vertexes (i, j). Valera is really keen to get a graph without any loops and multiple edges. If no such graph exists, print -1.