cpython difference in bind(), connect() arguments

[jepler]

CircuitPython version

9.0.0-beta

Code/REPL

>>> s = pool.socket() # or socket.socket
>>> b = bytearray(b'0.0.0.0')
>>> s.connect((b, 443))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: can't convert 'bytearray' object to str implicitly

Behavior

In standard Python, the bytearray is accepted (and the connection fails, because there's no listener):

Python 3.11.2 (main, Mar 13 2023, 12:18:29) [GCC 12.2.0] on linux
>>> b = bytearray(b'0.0.0.0')
>>> s.connect((b, 443))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ConnectionRefusedError: [Errno 111] Connection refused

Similar for bind():

# circuitpython
>>> s.bind((bytearray(), 1234))
TypeError: can't convert 'bytearray' object to str implicitly

vs

# cpython 3.11
>>> s.bind((bytearray(), 1234))
>>> 

Description

No response

Additional information

I ran into this while working on #8954. In that PR I was converting the bind/connect argument to a bytearray (because this let me avoid an allocation) but it didn't work.

[jepler]

While working on this it might be nice to make the empty hostname function the same as '0.0.0.0', again for compatibility with standard python.

[tannewt]

What is the advantage of allowing bytearray ASCII? This seems to go against the idea that strings are for bytes + encoding.

[jepler]

standard Python accepts a bytearray here, even if it's surprising. if we can accept a bytearray without (say) code bloat, I don't see why not to.

[tannewt]

Where does the bytearray come from? It should be a string. We don't do many things that CPython allows. Only the reverse needs to be true. (That CPython is ok with what we do.)

[jepler]

In the draft ssl-anything PR I arrange to just pass the original objects down, instead of unparsing them and rebuilding them. So except for compatiblity with a weird nook of cpython this is moot for me.