The whole family was excited by the news. Everyone knew grandpa had been an extremely

good bridge player for decades, but when it was announced he would be in the Guinness Book

of World Records as the most successful bridge player ever, whow, that was astonishing!

The International Bridge Association (IBA) has maintained, for several years, a weekly ranking

of the best players in the world. Considering that each appearance in a weekly ranking

constitutes a point for the player, grandpa was nominated the best player ever because he got

the highest number of points.

Having many friends who were also competing against him, grandpa is extremely curious to

know which player(s) took the second place. Since the IBA rankings are now available in the

internet he turned to you for help. He needs a program which, when given a list of weekly

rankings, finds out which player(s) got the second place according to the number of points.

Input

The input contains several test cases. Players are identified by integers from 1 to 10000. The

first line of a test case contains two integers N and M indicating respectively the number of

rankings available (2 <= N <= 500) and the number of players in each ranking (2 <= M <= 500).

Each of the next N lines contains the description of one weekly ranking. Each description is

composed by a sequence of M integers, separated by a blank space, identifying the players who

figured in that weekly ranking. You can assume that:

• in each test case there is exactly one best player and at least one second best player,

• each weekly ranking consists of M distinct player identifiers.

The end of input is indicated by N = M = 0.

The input must be read from standard input.

Output

For each test case in the input your program must produce one line of output, containing the

identification number of the player who is second best in number of appearances in the rankings.

If there is a tie for second best, print the identification numbers of all second best players in

increasing order. Each identification number produced must be followed by a blank space.

The output must be written to standard output.

Sample Input

4 5

20 33 25 32 99

32 86 99 25 10

20 99 10 33 86

19 33 74 99 32

3 6

2 34 67 36 79 93

100 38 21 76 91 85

32 23 85 31 88 1

0 0

Output for the Sample Input

32 33

1 2 21 23 31 32 34 36 38 67 76 79 88 91 93 100