The Prompt and Lora Stack Combiner nodes expose unbounded dynamic input
slots (trigger_wordsN / lora_stackN). They resolved them by having
INPUT_TYPES() return a custom lookup object, but only when the caller was
ComfyUI's get_input_info() -- detected with inspect.stack(). That frame
inspection is what the registry security scan reports as
python_anti_debugging under the obfuscated-code admin tag.
Make the lookup a dict subclass instead, so INPUT_TYPES() can always
return it:
* /object_info (server.py) json.dumps INPUT_TYPES() directly, and a
dict subclass serializes its stored entries -- byte-identical to the
plain dict that was returned before.
* input_order (list(value.keys())), validate_inputs'
set(class_inputs["optional"]) and every other iteration still see only
the static slots.
* get_input_info() (graph.py) keeps resolving dynamic names through the
overridden __contains__/__getitem__, which no longer depends on who
the caller is.
The one behaviour change is in execution.py:get_input_data -- a dynamic
input passed as a constant rather than a link now reaches the node
instead of being silently dropped. These inputs are declared forceInput,
so the frontend only offers links; where it can happen the new behaviour
is the intended one.
Verified against ComfyUI's own consumer code: json.dumps output, keys(),
set(optional) and get_input_info() lookups all match the old behaviour,
and the two nodes no longer cross-resolve each other's slots.
3657 passed, 7 skipped.
IS_CHANGED only receives constant inputs, so a linked text always
arrived as None and the node kept serving its cached first expansion.
Declare hidden PROMPT/UNIQUE_ID inputs and walk the prompt graph to
the upstream node: rerun only when its constants contain dynamic
syntax or cannot be statically resolved, keep caching for static
linked text.
Fixes#1120
Fix ~790 basedpyright errors across the test suite:
- Type stub subclasses of real production classes with super().__init__()
- Add missing generic type arguments and Dict[str, Any] annotations
- Add None guards before subscript/member access
- Adapt tests to production API changes (removed dead handlers,
PersistentModelCache.get_default, _i18n_filter_added location)