Since I bumped black in the tests to a version which supports 3.14, the following test case:

@pytest.mark.parametrize(
    "source_code,exception",
    [
        ("this isn't valid Python", ShedSyntaxWarning),
        # We request a bug report for valid unhandled syntax, i.e. (upstream) bugs
        ("class A:\\\r# type: ignore\n pass\n", ShedSyntaxWarning),
    ],
)
@pytest.mark.parametrize("refactor", [True, False])
def test_error_on_invalid_syntax(source_code, exception, refactor):
    with pytest.raises(exception):
        assert shed(source_code=source_code, refactor=refactor)

now parses fine in parsed = lib2to3_parse(source_code.lstrip(), target_versions) where it previously used to error and hit the The syntax is valid for Python ... case. Not sure what to do - should I remove the test case and mark that branch as nocover? I don't think we can have a reliable covering test for a branch which requires an upstream bug, bug maybe I am misinterpreting

Not sure what to do - should I remove the test case and mark that branch as nocover? I don't think we can have a reliable covering test for a branch which requires an upstream bug, bug maybe I am misinterpreting

seems like a reasonable response to me; my only hesitation is that ci is failing here with a similar-looking string -> tokenization error.