It's not as simple as that. Everything fails for some reason when I change it to 3.6, and I don't know why. I've tried, though – does that work?

Python doesn't seem to have a language version any more.

This really needs to have more test cases. The example test case provided should definitely be in the mandatory section too, and it would be great if you could add some randomised test cases. It shouldn't be too hard to generate three random strings, where the first and third can be any length including 0 and the second has a minimum length of 1, then repeat that a few hundred times.

Yup! You broke the attack I was going to post about just after I submitted it. One of them still gets through, though. (The ctypes one iirc.)

I've got a clever one saved up but it's only useful once you start blocking the importing of certain modules.

You can fix the problem by checking for the vulnerabilities that eval causes. Try running arbitrary code, using = etc. That will patch the real problem, then you can free up all of the features you've blocked for programmers to use again.

In the meantime, I'm going to submit another attack. :-)

I'm frustrated by the implementation, but for now this works.

I've broken the Python one again (though with more code than I would have liked), and I've thought of another, shorter, cleaner one. I'm beginning to find this tricky though. You've patched almost all of the obvious ones.

https://www.codewars.com/kata/reviews/5ad0d4dd6165e69a5f00141d/groups/5b0140003d841f832f0012a2

Another way of bypassing it! https://www.codewars.com/kata/reviews/5ad0d4dd6165e69a5f00141d/groups/5b006cf149db933387002d53

Sorry it took me so long; I had to do something irl.

There are a practically unbounded number of these; you won't be able to stop them all in the way you're trying to do it. This is a fun game of cat and mouse; I've got another one that I'd like to try so could you tell me when you've fixed this one?

Has the problem gone away?

As far as I can tell it saved before I told you about it. :-/

That "incorrect output" thing is not from the third test; it's a global test that checks whether anything was sent to output that shouldn't have been (e.g. by an uncaught exception). Try suppressing stderr. That might fix it.

It's caused by a change in the testing logic that wasn't tested properly. I'll fix it.