Move color default to parse time, clarify deferred sniff and comments

lib/galaxy/model/deferred.py

@@ -270,7 +270,12 @@ class DatasetInstanceMaterializer:
         path: str,
         dataset_instance: HistoryDatasetAssociation | LibraryDatasetDatasetAssociation,
     ) -> None:
-        """Replace ``extension="auto"`` with a sniffed extension after download."""
+        """Resolve ``extension="auto"`` once the deferred source is streamed to ``path``.
+
+        A deferred fetch/upload with no explicit ``ext`` is registered as ``auto`` because
+        the content is not available to sniff at request time. ``path`` is the first point
+        the bytes exist locally, so we sniff here while materializing.
+        """
         if self._datatypes_registry is None:
             return
         if dataset_instance.extension != "auto":

lib/galaxy/tool_util/parameters/convert.py

@@ -492,14 +492,11 @@ def _fill_default_for(tool_state: dict[str, Any], parameter: ToolParameterT) ->
             # see test_tools.py -> expression_null_handling_boolean or test cases for gx_boolean_optional.xml
             tool_state[parameter_name] = parameter.value or False
 
-    if isinstance(parameter, (IntegerParameterModel, FloatParameterModel, HiddenParameterModel)):
+    if isinstance(
+        parameter, (IntegerParameterModel, FloatParameterModel, HiddenParameterModel, ColorParameterModel)
+    ):
         if parameter_name not in tool_state:
             tool_state[parameter_name] = parameter.value
-    elif isinstance(parameter, ColorParameterModel):
-        # Legacy ``value=""`` on an optional color means "no default"; treat as unset (arriba).
-        if parameter_name not in tool_state:
-            default = parameter.value
-            tool_state[parameter_name] = None if default == "" else default
     elif isinstance(parameter, GenomeBuildParameterModel):
         if parameter_name not in tool_state and parameter.optional:
             tool_state[parameter_name] = None
@@ -636,8 +633,7 @@ def _initialize_repeat_state(parameter: RepeatParameterModel, tool_state: dict[s
     if parameter_name not in tool_state:
         tool_state[parameter_name] = []
     repeat_instances = cast(list[dict[str, Any]], tool_state[parameter_name])
-    # Match sync's ``Repeat.get_initial_value()``: seed ``max(min, default)`` empty
-    # instances when the request supplies none (bigscape).
+    # Seed ``max(min, default)`` empty instances when the request supplies none.
     floor = max(parameter.min or 0, parameter.default or 0)
     while len(repeat_instances) < floor:
         repeat_instances.append({})

lib/galaxy/tool_util/parameters/factory.py

@@ -198,11 +198,16 @@ def _from_input_source_galaxy(input_source: InputSource, profile: float) -> Tool
             )
         elif param_type == "color":
             optional = input_source.parse_optional()
+            color_value = get_color_value(input_source)
+            # A color default must be a valid color or None. The legacy ``value=""``
+            # on an optional color means "unset".
+            if optional and color_value == "":
+                color_value = None
             return ColorParameterModel(
                 type="color",
                 name=input_source.parse_name(),
                 optional=optional,
-                value=get_color_value(input_source),
+                value=color_value,
                 **_common_param_kwargs(input_source),
             )
         elif param_type == "rules":

lib/galaxy/tool_util_models/parameters.py

@@ -2397,8 +2397,8 @@ class RepeatParameterModel(BaseGalaxyToolParameterModelDefinition):
     parameters: list["ToolParameterT"]
     min: int | None = None
     max: int | None = None
-    # XML ``<repeat default="N">``; matches sync's ``Repeat.get_initial_value()`` to seed
-    # N empty instances when the request supplies none.
+    # Legacy XML ``<repeat default="N">``: number of empty instances to seed when the
+    # request supplies none.
     default: int = 0
 
     def field_kwargs(self) -> dict[str, Any]:

test/unit/data/test_dataset_materialization.py

@@ -435,8 +435,11 @@ def test_deferred_hdas_with_deferred_metadata():
 
 
 def test_deferred_hda_with_auto_extension_gets_sniffed_from_content():
-    # An HDA created with extension="auto" should get a real extension after materialize
-    # when a datatypes_registry is provided.
+    # Scenario: a deferred fetch/upload with no explicit ``ext`` is stored as
+    # extension="auto" because the bytes aren't available to sniff at request time
+    # (data_fetch.py: ``requested_ext = item.get("ext", "auto")``). When the dataset is
+    # later materialized for a tool run, the real type is sniffed from the downloaded
+    # content. Requires a datatypes_registry.
     fixture_context = setup_fixture_context_with_history()
     store_dict = deferred_hda_model_store_dict()
     store_dict["datasets"][0]["extension"] = "auto"
@@ -460,7 +463,8 @@ def test_deferred_hda_with_auto_extension_gets_sniffed_from_content():
 
 
 def test_deferred_hda_without_registry_keeps_auto_extension():
-    # Without a datatypes_registry the sniff step is a no-op; extension stays "auto".
+    # Guards backward compatibility: existing materializer callers that don't pass a
+    # datatypes_registry keep the prior behavior (no sniffing); extension stays "auto".
     fixture_context = setup_fixture_context_with_history()
     store_dict = deferred_hda_model_store_dict()
     store_dict["datasets"][0]["extension"] = "auto"