yuqian90 commented on a change in pull request #14416:
URL: https://github.com/apache/airflow/pull/14416#discussion_r582548614
##########
File path: tests/models/test_taskinstance.py
##########
@@ -1488,6 +1488,30 @@ def test_previous_start_date_success(self, _,
schedule_interval, catchup) -> Non
assert ti_list[3].get_previous_start_date(state=State.SUCCESS) ==
ti_list[1].start_date
assert ti_list[3].get_previous_start_date(state=State.SUCCESS) !=
ti_list[2].start_date
+ def test_get_previous_start_date_none(self):
+ """
+ Test that get_previous_start_date() can handle TaskInstance with no
start_date.
+ """
+ with DAG("test_get_previous_start_date_none", start_date=DEFAULT_DATE,
schedule_interval=None) as dag:
+ task = DummyOperator(task_id="op")
+
+ dag.create_dagrun(
+ execution_date=DEFAULT_DATE,
+ state=State.RUNNING,
Review comment:
Hi @XD-DENG , in fact, any scenario that results in the previous
execution_date's TaskInstance not having a `start_date` causes this crash.
Marking the previous day's TaskInstance success is one way to run into this
situation. There are many other ways, e.g. the DagRun for previous
execution_date's TaskInstance has not run yet because it's still waiting for
the scheduler to run it, or because its dependencies are not yet met). In this
test here, the previous dagrun is created without running op, so it has no
start_date. This triggers the bug described in the PR summary. I also updated
the summary to include more example scenarios.
Regarding passing `start_date=None` to DagRun. I don't fully understand the
use of `DagRun.start_date`. The issue in this PR is caused by the
`TaskInstance` not having a `start_date`, not caused by the `DagRun`
start_date. Maybe it's because i dont' understand how `DagRun.start_date` is
used. Is it used to set the `TaskInstance.start_date` in some way?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
