make configtype from_str hypothesis check skip subclasses too

We are already skipping round trip tests for partial functions made in
gen_classes from compound classes (list, dict, ...). This change makes it so
we skip trying it for subclasses of those compound classes too (eg FlagList
and Padding). This is due to the test failing with the input "- A" to
FlagList.

Also switches the Dict example in the comment to be the same order as the
other two.

Some questions I don't know the answers to:

* Why are List and Dict using `json.dumps()` in `to_str()` instead of `yaml_dump()`?
* Is `to_str()` used anywhere apart from tests?
* Should we have a `to_yaml()` or just use `yaml_dump()` in the test to get back
  the original value?

1 files changed, 8 insertions, 5 deletions

diff --git a/tests/unit/config/test_configtypes.py b/tests/unit/config/test_configtypes.py
index e4351d619..99b8a5de0 100644
--- a/tests/unit/config/test_configtypes.py
+++ b/tests/unit/config/test_configtypes.py
@@ -255,11 +255,14 @@ class TestAll:
                 configtypes.PercOrInt,  # ditto
         ]:
             return
-        elif (isinstance(klass, functools.partial) and klass.func in [
-                configtypes.ListOrValue, configtypes.List, configtypes.Dict]):
-            # ListOrValue: "- /" -> "/"
-            # List: "- /" -> ["/"]
-            # Dict: '{":": "A"}' -> ':: A'
+
+        # Dict and List deserialize to yaml but serialize to json, so we can't
+        # do a naive round trip test with them. For example:
+        #   ListOrValue: "- /" -> "/"
+        #   List: "- /" -> ["/"]
+        #   Dict: ':: A' -> '{":": "A"}'
+        compound_types = (configtypes.ListOrValue, configtypes.List, configtypes.Dict)
+        if isinstance(typ, compound_types):
             return
 
         assert converted == s