• As in the original translation: "Please don't use public for the reference solution..."
• For the recursion test I would test that the method was actually called with arguments n, n-1, ..., 1 via ArgumentCaptors since you can just call factorial(0) n-1 times and then calculate with a for loop or so. Additionally the recursion should not force to stop at n==1 but could also go to n==0 so exactly n times is too restrictive. Also I would recommend to use a bigger value for this test as n==2 can be rolled and this not really representative value to check for recursion.
• You can go up to n==20 in the random tests without overflowing long range.
• Please do some more random tests, e.g. 10
• If you want to put the effort in you could update to JUnit 5 and use org.junit.jupiter.params.ParameterizedTest and org.junit.jupiter.api.RepeatedTest

Hello @user8436785 ,

Forked and added this translation - https://www.codewars.com/kumite/5d0437ab11bd8c002086eeaa?sel=62b6cf6d0ee74b011f9f9458

Please use Java 11.
Please don't use public for the reference solution...

OK this it's resolved. Even if this kata isn't accepted, I learned how to inject bytecode at Java Runtime, so it's a win :)
In Java, test that a method is called recursively is a pain in the ass, it took me 4 hours.

You're right, I was to hasty to publish this kata, I'll fix it and publish it later. Thanks !

You're right, fixed

None of the "restrictions" are actually present.

Duplicate.

Use assertEquals(expected, actual, delta) to compare floating-point numbers

The tests are not runnable.

The initial solution is wrong. The expected data type is long (at least the author used long in his solution).