Chaining @wraps with non-modifying then modifying signatures leads to erroneous __wrapped__ flag been kept.

[smarie]

def a(foo):
    return foo + 1

assert a(1) == 2

# create a first wrapper that is signature-preserving

@wraps(a)
def wrapper(foo):
    return a(foo) - 1

assert wrapper(1) == 1

# the __wrapped__ attr is here:
assert wrapper.__wrapped__ is a

# create a second wrapper that is not signature-preserving

@wraps(wrapper, append_args="bar")
def second_wrapper(foo, bar):
    return wrapper(foo) + bar

assert second_wrapper(1, -1) == 0

with pytest.raises(AttributeError):
    second_wrapper.__wrapped__   # <---- does not raise !