The performance of this is N ** 2 + NM (where N is the size of s2 and M is the size of s1) since:

• it loops through all the characters of s2 (N)
• for each of those caracters it counts the time the character 'x' appears. For that it has to loop through all s2 (N) and s1 (M)

That's far from ideal, which should be N + M

The description mentions from a to b (inclusive). Ins't this answer wrong in that sense ?

irb > (1...3).to_a
=> [1, 2]