random.Random._randbelow_with_getrandbits loops forever for n=0

[Dutcho]

Bug report

Bug description:

random.Random()._randbelow_with_getrandbits(0)  # loops indefinitely

Same applies to (conditional) synonym ._randbelow().

Note the alternative implementation (below) fails:

random.Random()._randbelow_without_getrandbits(0)  # raises ZeroDivisionError

I suggest adding a check on lower bound (like _randbelow_without_getrandbits has on upper bound).

CPython versions tested on:

3.14

Operating systems tested on:

Windows

[ZeroIntensity]

Is this causing a problem in practice somewhere? If you break a private API, you're usually on your own.

[Dutcho]

True, I was forewarned by the leading _.
I still ran into this when proposing to enhance choice().

Going to the random.py source soon clarified the stated (but unchecked) precondition "Defined for n > 0".
Within random.py, _randbelow gets called in a number of places. These all properly guard that n > 0 (in 3.14).
Doubling the check within _randbelow would be defensive against future errors within random.py and against other careless trespassers like myself.

[ZeroIntensity]

Feel free to open a PR then. Let's just consider this an internal refactor, though.

[picnixz]

I honestly think we should not do defensive programming like that on private APIs. It does not make sense to do double the work IMO.

cc @rhettinger