TestsCaseValidation: "" not allowed for optional selects?

[bernt-matthias]

Describe the bug

For this test parameter the new linter reports the following error (which seems correct) -- but it remains even if I set optional="true" for the outputs parameter (as it probably should be):

.. WARNING (TestsCaseValidation): Test 4: failed to validate test parameters against inputs - tests won't run on a modern Galaxy tool profile version. Validation errors are [6 validation errors for PydanticModelFor[augustus]
outputs.0.literal['protein']
  Input should be 'protein' [type=literal_error, input_value='', input_type=str]
outputs.0.literal['codingseq']
  Input should be 'codingseq' [type=literal_error, input_value='', input_type=str]
outputs.0.literal['introns']
  Input should be 'introns' [type=literal_error, input_value='', input_type=str]
outputs.0.literal['start']
  Input should be 'start' [type=literal_error, input_value='', input_type=str]
outputs.0.literal['stop']
  Input should be 'stop' [type=literal_error, input_value='', input_type=str]
outputs.0.literal['cds']
  Input should be 'cds' [type=literal_error, input_value='', input_type=str]]

Also this one its really tricky to read, because the name of the parameter is not mentioned anywhere.

Ping @jmchilton

Galaxy Version and/or server at which you observed the bug
Galaxy Version: 24.2

[mvdbeek]

"" is not None, so the error makes sense, you're strictly speaking selecting an option that doesn't exist. value_json="null" should be the precise type for a an unset option (or skip it if the test is meant to test the default).
We might want to relax the type coercion in the pydantic model when we're parsing xml tools where all types are pretty much strings, unless we use workarounds like value_json. The parameter is the first one , outputs.0 is the "path notation" for the output.