4.1.0: Test failures with Python 3.8

[hroncok]

I see those build failures on Python 3.8.0b2 and Uranium 4.1.0:

_________________________ test_getUsedSettings[data1] __________________________

data = {'code': '"x"', 'variables': ['x']}

    @pytest.mark.parametrize("data", test_getUsedSettings_data)
    def test_getUsedSettings(data):
        function = SettingFunction(data["code"])
        answer = function.getUsedSettingKeys()
>       assert len(answer) == len(data["variables"])
E       assert 0 == 1
E         -0
E         +1

tests/Settings/TestSettingFunction.py:141: AssertionError
_________________________ test_getUsedSettings[data7] __________________________

data = {'code': "sqrt('x')", 'variables': ['sqrt', 'x']}

    @pytest.mark.parametrize("data", test_getUsedSettings_data)
    def test_getUsedSettings(data):
        function = SettingFunction(data["code"])
        answer = function.getUsedSettingKeys()
>       assert len(answer) == len(data["variables"])
E       assert 1 == 2
E         -1
E         +2

The failures were also present in 4.0.0 with older Python 3.8.0 releases.

[Ghostkeeper]

We're targetting CPython 3.5.2 right now.

I wouldn't expect minor versions of Python to break backwards compatibility. They're usually pretty good on this part. Maybe it's because they are still in beta. In any case, we don't intend to support all versions of all our dependencies. That's why we package CPython with our releases.

In the future we're probably going to upgrade to newer versions of Python, if we need a newer version for some new feature (I love the formatted strings for instance). We'll look to support those new versions then.

[hroncok]

As a Fedora packager of Cura, I unfortunately cannot package CPython 3.5.2 with Cura (note that you should probably look at least at Python 3.5.6). So far CPython 3.7 worked pretty good, but 3.8 fails. If it's because it is beta, than maybe it's a bug in CPython itself that needs to be fixed. Or maybe it is not because it is beta.

Would you at least accept a fix if I figure this out?

[hroncok]

This breakage probably caused by this change in Python:

python/cpython#9445

I have a fix ready, would you accept it, or should I carry it downstream forever?

diff --git a/UM/Settings/SettingFunction.py b/UM/Settings/SettingFunction.py
index 9ca0d30..28739c7 100644
--- a/UM/Settings/SettingFunction.py
+++ b/UM/Settings/SettingFunction.py
@@ -177,10 +177,16 @@ class _SettingExpressionVisitor(ast.NodeVisitor):
             self.values.add(node.id)
             self.keys.add(node.id)
 
+    # This one is used before Python 3.8
     def visit_Str(self, node: ast.AST) -> None:
         if node.s not in self._knownNames and node.s not in dir(builtins):  # type: ignore #AST uses getattr stuff, so ignore type of node.s.
             self.keys.add(node.s)  # type: ignore
 
+    # This one is used on Python 3.8+
+    def visit_Constant(self, node: ast.AST) -> None:
+        if isinstance(node.value, str) and node.value not in self._knownNames and node.value not in dir(builtins):  # type: ignore #AST uses getattr stuff, so ignore type of node.value.
+            self.keys.add(node.value)  # type: ignore
+
     _knownNames = {
         "math",
         "max",

[Ghostkeeper]

Sure, we'll accept that bugfix.

[Ghostkeeper]

See commit here: d7ed5eb

[hroncok]

Thanks.