Nevermind - it is a known bug. @le chevalier posted a link to it in a comment on the solutions page

https://bugs.python.org/issue13305

Thanks for the info lechevalier -- I was really wondering what the problem was.

In Python 3.6, something weird happened where the %Y in strftime specification failed to pad with 4 zeros in the year for two of the random tests on CW, but did pad correctly in external interpreter. Any idea how that could happen?

My Python 3.6 code is timing out, but I'm not sure why.
Is the random test particularly strict on performance?

I've attached the code in a reply in case anyone wants to give me a hint.

Interesting.... so apparently eval is recursing somehow... do you happen to know of any online resources that might shine any more light on this? I searched on 'recursion errors without recursive functions' but didn't see anything about eval on SO.

thanks

OK - so I guess I am learning something new here -
I thought 'maximum recursion depth' errors only occur as consequence of excessive recursive function calls per se.
Since my function does not call itself, there is no recursion - correct?
If so, then I should deduce that something else can trigger this error, such as a large number of non-recursing function calls?
Is this correct? If so, why is it not just giving a stack overflow error?

Thanks in advance

p.s. I have since solved it using a standard approach

Random tests throw false recursion depth errors.

My code has no recursion in it.

It is annoying -- and seems to be an oversight -- that zeros should be kept as valid values while discarding negatives. Anyone who ever worked with geometric means in real life knows that the zeros need to be removed too. This annoyance is compounded by the lack of zeros in the example tests.

Better yet: "Could John be telling the truth?"

"Is John lying?" is a different question than "Is John necessarily lying?", which is what this kata is actually asking. The problem should be rephrased.

I believe this should work even for stupidly large values of a

indecipherable description needs to be re-written