Postgres Operator task doesn't sequence properly with TaskFlow API

Apache Airflow version: 2.1.0.dev0

Kubernetes version (if you are using kubernetes) (use kubectl version): NA

Environment: Docker Apache/Airflow master-python3.6

What happened: some_pg_task it out of sequence

What you expected to happen:
Task sequence should be some_py_task >> some_pg_task

How to reproduce it:

from airflow.models import DAG
from airflow.providers.postgres.operators.postgres import PostgresOperator
import os

from airflow.utils.dates import days_ago

default_args = {'start_date': days_ago(1)}

dag_name = os.path.splitext(os.path.basename(__file__))[0]

with DAG(dag_name, default_args=default_args) as dag:

    @dag.task
    def some_py_task() -> str:
        return some_str

    some_str = some_py_task()

    some_pg_task = PostgresOperator(task_id='some_pg_task', 
                     sql='select * from %s',
                     parameters=[some_str],
                     postgres_conn_id='postgres_bob')

Anything else we need to know: The MSSQL Op behaves the same way

[ephraimbuddy]

This is what you need: test_pg_operator >> order_data >> order_summary >> load_task_flow There’s a difference between passing xcoms and creating dependency. We are working on documentation to clarify that. See : https://github.com/apache/airflow/issues/17686#issuecomment-937599461

Closing this for now

[ephraimbuddy]

cc: @alex-astronomer