int inconsistently considered a subtype of float

[decorator-factory]

From PEP 484:

when an argument is annotated as having type float, an argument of type int is acceptable

So mypy considers this legal (mypy 1.10.0, with --strict):

def f(x: float) -> None:
    pass

f(42)

But when checking objects with isinstance, it thinks a variable annotated with float cannot be an int:

def f(x: float) -> None:
    if not isinstance(x, float):
        print("💀")  # error: Statement is unreachable  [unreachable]
        reveal_type(x)  # no diagnostics

f(42)

I think this is a bug, since float is always implicitly float | int. I'd expect mypy to not flag this as unreachable and reveal the type of x as int. (that's the behaviour pyright currently has)

Here's another potentially buggy example:

def f(x: float | str) -> None:
    if isinstance(x, float):
        reveal_type(x)  # Revealed type is "builtins.float"
    else:
        reveal_type(x)  # Revealed type is "builtins.str"

If you call this function as f(42), which mypy allows, the second branch will be selected at runtime.

[InSyncWithFoo]

Another way to put it (playground):

from typing import Literal

a: int = 3

if isinstance(a, float):
   reveal_type(a)  # Never touched at runtime, but Mypy says `int`.

b: Literal[3] = 3

if isinstance(b, float):
   reveal_type(b)  # Never touched at runtime, correctly reported as unreachable.

[hauntsaninja]

Closing as a duplicate of #11516