annotationlib: Improve test coverage

[dr-carlos]

Feature or enhancement

Proposal:

The existing test suite for annotationlib is good, but there is still a decent amount of line/branch coverage missing. coverage.py reports 75% coverage, but this is mostly the library is already imported somewhere in coverage.py, so all global statements are marked as 'missing'.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

Linked PRs

• gh-141174: Improve ForwardRef test coverage #141175
• gh-141174: Improve annotationlib.call_annotate_function() test coverage #141176
• gh-141174: Improve annotationlib._Stringifier test coverage #141220
• gh-141174: Improve annotationlib.get_annotations() test coverage #141286
• [3.14] gh-141174: Improve annotationlib.get_annotations() test coverage (GH-141286) #141353
• [3.14] gh-141174: Improve ForwardRef test coverage (GH-141175) #141354
• [3.14] gh-141174: Improve annotationlib.call_annotate_function() test coverage (GH-141176) #141355
• [3.14] gh-141174: Improve annotationlib._Stringifier test coverage (GH-141220) #141358
• gh-141174: Test type_repr() of Template with invalid syntax in Interpolation #155661
• gh-141174: Test get_annotations() on wrapped cyclical functools.partial #155663
• gh-141174: Test re-stringifying a ForwardRef #155689
• [3.15] gh-141174: Test re-stringifying a ForwardRef (GH-155689) #155707
• [3.14] gh-141174: Test re-stringifying a ForwardRef (GH-155689) #155708
• [3.15] gh-141174: Test get_annotations() on wrapped cyclical functools.partial (GH-155663) #155709
• [3.14] gh-141174: Test get_annotations() on wrapped cyclical functools.partial (GH-155663) #155710
• [3.15] gh-141174: Test type_repr() of Template with invalid syntax in Interpolation (GH-155661) #155711
• [3.14] gh-141174: Test type_repr() of Template with invalid syntax in Interpolation (GH-155661) #155712

[dr-carlos]

Currently there is no coverage for the 'else' case in _build_closure():

cpython/Lib/annotationlib.py

Lines 842 to 845 in 14c6222

	if i < len(freevars):
	name = freevars[i]
	else:
	name = "__cell__"

But I'm not sure what conditions would cause __closure__ and __code__.co_freevars to be different lengths.
Git blame says that @JelleZijlstra wrote this code, is there a particular object you'd expect to behave this way?

The only other place in the stdlib where both are used together is inspect.getclosurevars(), which simply zips the closure and freevars together:

cpython/Lib/inspect.py

Lines 1503 to 1509 in 14c6222

	if func.__closure__ is None:
	nonlocal_vars = {}
	else:
	nonlocal_vars = {
	var : cell.cell_contents
	for var, cell in zip(code.co_freevars, func.__closure__)
	}

[JelleZijlstra]

I think that could only happen with a manually constructed code object (or if there's a bug in the compiler). I wrote it that way to make the code robust to this case, but it might have been better to throw an error instead if the lengths mismatch.