Mike Perlov e15419396c [red-knot] Fix Stack overflow in Type::bool (#15843) ... ## Summary This PR adds `Type::call_bound` method for calls that should follow descriptor protocol calling convention. The PR is intentionally shallow in scope and only fixes #15672 Couple of obvious things that weren't done: * Switch to `call_bound` everywhere it should be used * Address the fact, that red_knot resolves `__bool__ = bool` as a Union, which includes `Type::Dynamic` and hence fails to infer that the truthiness is always false for such a class (I've added a todo comment in mdtests) * Doesn't try to invent a new type for descriptors, although I have a gut feeling it may be more convenient in the end, instead of doing method lookup each time like I did in `call_bound` ## Test Plan * extended mdtests with 2 examples from the issue * cargo neatest run		2025-02-04 12:40:07 -08:00
..
custom.md	Don't special-case class instances in binary expression inference (#15161)	2025-01-06 13:50:20 -05:00
integers.md	[red-knot] Autoformat mdtest Python snippets using blacken-docs (#13809)	2024-10-19 15:57:06 +01:00
invert_add_usub.md	[red-knot] Infer unary not operation for instances (#13827)	2024-11-13 23:31:36 +00:00
not.md	[red-knot] Fix Stack overflow in Type::bool (#15843)	2025-02-04 12:40:07 -08:00