From https://docs.python.org/3/library/multiprocessing.html#multiprocessing.Value

    Note that lock is a keyword-only argument.

But the signature of Value (in Lib/multiprocessing/context.py)

    def Value(self, typecode_or_type, *args, lock=True):

and (in Lib/multiprocessing/sharedctypes.py)

    def Value(typecode_or_type, *args, lock=True, ctx=None):

I'd suggest to remove that part of the documentation in 3.4+. We can also make it a keyword-only argument in Python 3.6. That will make the API more similar with https://docs.python.org/3/library/multiprocessing.html#multiprocessing.Array

I’m not familiar with this module, but I believe “lock” is indeed keyword-only. If you were to try a positional argument, it would be picked up as part of *args:

>>> multiprocessing.Value("I")  # Default to 0, lock=True
<Synchronized wrapper for c_uint(0)>
>>> multiprocessing.Value("I", False)  # False == 0, still lock=True
<Synchronized wrapper for c_uint(0)>
>>> multiprocessing.Value("I", lock=False)
c_uint(0)

Agreed. Having a named positional varargs argument makes subsequently defined arguments keyword only automatically.

Berker: It looks to me like the docs are indeed in sync with the code on Value in that lock really is a keyword-only argument.  It looks like it's been that way since at least 2.7 (I didn't look at earlier).  There are enough other things in the multiprocessing.sharedctypes, maybe it was one of the others that caught your attention instead?

I wished that block/blocking had turned out to be a keyword-only argument in multiprocessing.Lock as part of issue23484, which I immediately thought of when reading this issue.

I didn't test it carefully. It was a mistake on my part :) But I think the function signature could be clearer by changing that part to "*, lock=True".