ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-626244825
(We've got that code in this diff comment, we can bring it back when we want
it, but that will need more changes elsewhere to serialize more fields
ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-626244699
I was wrong about where that fn was used, so I've removed that complex code
an replace it with this in the test instead:
```python
if serialized_task.resources is
ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-626013593
> Oh, cos it's parsing in a subprocess. H!
>
> I don't think we need to parse _all_ dags in a subprocess, just one is
enough.
Changed it to only parse
ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-625814879
Oh, cos it's parsing in a subprocess. H!
I don't think we need to parse _all_ dags in a subprocess, just one is
enough.
ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-625814458
Curious, the DAG with the problem has `'start_date': datetime.utcnow()`
-- that's "wrong" but I wouldn't have expected this test to fail. I guess we
are loading the dag twice,
ashb commented on pull request #8772:
URL: https://github.com/apache/airflow/pull/8772#issuecomment-625803075
Curious failure:
```
E AssertionError: default_args[start_date] matches
E assert ==
```
I wonder why I didn't see that
