False positive undefined-variable for variable in inner scope matching outer scope type annotation

[jacobtylerwalls]

Bug description

def func():
    a: bool
    _ = [print(kwarg1=a, kwarg2=a) for a in range(10)]

Configuration

No response

Command used

pylint a.py

Pylint output

a.py:3: [E0602(undefined-variable), func] Undefined variable 'a'

Expected behavior

No messages

Pylint version

pylint 2.11.2-dev0
astroid 2.8.5
Python 3.9.6 (v3.9.6:db3ff76da1, Jun 28 2021, 11:49:53)

Does NOT occur on 2.11.1, possible regression in dev version?

OS / Environment

No response

Additional dependencies

No response

[jacobtylerwalls]

After bisecting, the change occurred in b1c4735. Reading that change, it seems undefined-variable is the new intended behavior, but it should be raised on line 2, not line 3. And if only one kwarg involves the variable, then unused-variable is emitted instead, which I suppose is a false negative for undefined-variable.

[IceTDrinker]

hello to pop in and say we have that issue but with the annotated outer scope variable name being used/assigned to later

Basically something like:

node: MyNode  # triggers the undefined-variable which is hard to comprehend
list_of_nodes = [node for node in maybe_node_list if isinstance(node, MyNode)]

for node in other_list_of_node:
    pass

[IceTDrinker]

This also fails on the first line

list_of_nodes = [node for node in maybe_node_list if isinstance(node, MyNode)]

node: MyNode

for node in other_list_of_node:
    pass

Edit: feels like it should not, we'll be freezing our pylint dependency to the previous version until this is fixed as it prevents from having proper typing for mypy and sensible code in those cases

[jacobtylerwalls]

I broke the false negative out into #5654. That means this ticket is now just a false positive with a PR to close it.