Dataclass structuring fails with misleading error message.

• cattrs version: 23.2.3
• Python version: 3.11.8
• Operating System: OSX

Description

I'm using cattrs to destructure and restructure a complex hierarchy of classes.
It is mostly working.
But I've boiled down one of the failures to this small example.

What I Did

from __future__ import annotations
from dataclasses import dataclass
from cattrs.preconf.json import make_converter

CONV = make_converter()

@dataclass(kw_only=True)
class A:
    a: str = ""

@dataclass(kw_only=True)
class B:
    b: str = ""

@dataclass(kw_only=True)
class C:
    c: int = 3
    a_or_b: AorB | None = None

AorB = A | B

def test_c_reconstruction():
    c = C(a_or_b=B())
    s = CONV.dumps(c)
    #  attr.exceptions.NotAnAttrsClassError: <class 'tests.test_cattrs_fail.A'> is not an attrs-decorated class.
    c2 = CONV.loads(s, C)

Start with the reproducer in the issue, especially make_converter(), CONV.dumps(), and CONV.loads(s, C) in tests.test_cattrs_fail. Confirm the misleading NotAnAttrsClassError with the dataclass union example; done when the failure reports the relevant dataclass or union problem accurately.