elchao96 opened a new issue, #29903:
URL: https://github.com/apache/airflow/issues/29903

   ### Apache Airflow version
   
   Other Airflow 2 version (please specify below)
   
   ### What happened
   
   When running a DAG that is structured like:
   ```
   @dag{dag_id="my_dag", default_args={"retries":0"}}
   def dag():
       op = MyOperator.partial(task_id="my_task", retries=3).expand(...)
   ```
   
   The following test fails:
   ```
       def test_retries(self) -> None:
           dag_bag = DagBag(dag_folder=DAG_FOLDER, include_examples=False)
           dag = dag_bag.dags["my_dag"]
           for task in dag.tasks:
               if "my_task" in task.task_id:
                   self.assertEqual(3, task.retries) # fails - this is 0
   ```
   
   When printing out `task.partial_kwargs`, and looking at how the default args 
and partial args are merged, it seems like the default args are always taking 
precedence, even though in the `partial` global function, the `retries` do get 
set later on with the task-level parameter value. This doesn't seem to be 
respected though.
   
   ### What you think should happen instead
   
   _No response_
   
   ### How to reproduce
   
   If you run my above unit test for a test DAG, on version 2.4.3, it should 
show up as a test failure.
   
   ### Operating System
   
   OS Ventura
   
   ### Versions of Apache Airflow Providers
   
   _No response_
   
   ### Deployment
   
   Google Cloud Composer
   
   ### Deployment details
   
   _No response_
   
   ### Anything else
   
   _No response_
   
   ### Are you willing to submit PR?
   
   - [ ] Yes I am willing to submit a PR!
   
   ### Code of Conduct
   
   - [X] I agree to follow this project's [Code of 
Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to