False positive not callable for function call in dictionary

[hbraha]

Bug description

ATTRIBUTES = {
        'DOMAIN': ("domain", str),
        'IMAGE': ("image", str),
}

for key, (name, validate) in ATTRIBUTES.items():
    name = validate(1)

Configuration

No response

Command used

pylint test1.py

Pylint output

************* Module test1
test1.py:8:11: E1102: validate is not callable (not-callable)

Expected behavior

no error.

Pylint version

pylint 2.10.0
astroid 2.7.3
Python 3.6.8 (default, Mar 18 2021, 08:58:41)

OS / Environment

Red Hat Enterprise Linux release 8.4

Additional dependencies

No response

[mbyrnepr2]

It seems that when astroid visits the tuple node key, (name, validate), the newnode which is produced has elts which contain only name and validate elements (and the key is missing).

Possibly this is due to recursion on this line where the visit method recurses into the visit_tuple method again and overwrites the newnode.elts which at this point contains the key object:
https://github.com/PyCQA/astroid/blob/main/astroid/rebuilder.py#L2208

elts assigned here: https://github.com/PyCQA/astroid/blob/main/astroid/nodes/node_classes.py#L335

@Pierre-Sassoulas do you think this sounds like the right track?

Aside from this the reported sample code works ok if it is simplified to:

ATTRIBUTES = {
        'DOMAIN': str,
        'IMAGE':  str,
}

for key, validate in ATTRIBUTES.items():
    name = validate(1)

In general, pylint doesn't report any issues with the following:

for i, j, k, l, m, n, o, p in {"a": "b"}.items():
    print(i, j, k, l, m, n, o, p)

[DanielNoord]

@mbyrnepr2 That certainly seems like an issue! Good job on identifying it!

[Pierre-Sassoulas]

the visit method recurses into the visit_tuple method again and overwrites the newnode.elts

This sounds problematic indeed, good job finding this !

[mbyrnepr2]

@Pierre-Sassoulas @DanielNoord thanks for looking.
I take back what I said about the recursion being an issue in my previous comment. On further examination that logic looks ok - I must have confused myself when debugging.

The following changes seem to fix it, but I'll leave as draft for now as I would like to share progress but also double-check when I have more time:
https://github.com/PyCQA/pylint/pull/5593/files
https://github.com/PyCQA/astroid/pull/1311/files

[mbyrnepr2]

I think we can close once #5593 is merged (functional test).