thelioo opened a new pull request, #76653:
URL: https://github.com/apache/airflow/pull/76653

   A cross-Dag `xcom_pull(dag_id=..., include_prior_dates=True)` returns `None` 
since Airflow 3.2. The SDK sends the caller's `run_id`, but the prior-date 
cutoff looks that run up in the Dag being read. It finds no such run, the 
cutoff is `NULL`, and every row is dropped.
   
   When `run_id` does not exist in the target Dag and it is the calling task 
instance's own run, the cutoff now comes from the caller's run. That keeps the 
Airflow 3.1 meaning of "on or before my run". A pull that names a run which 
does exist in the target Dag keeps its current behaviour. This needs no 
Execution API schema change.
   
   This takes the same direction as #70199, which was closed as stale before 
review. The difference is that it only uses the caller's date when the target 
run is missing, so an explicit `run_id` of the target Dag is not overridden.
   
   The new test covers the single-value and slice endpoints and fails without 
this change. The XCom suites pass in Breeze (297 passed: `test_xcoms.py`, 
`models/test_xcom.py` and the public `test_xcom.py`). This adds up to two small 
queries, only when `include_prior_dates` is set, and I did not run a benchmark 
for it.
   
   closes: #70173
   
   ---
   
   ##### Was generative AI tooling used to co-author this PR?
   
   - [X] Yes — Claude Code (Opus 5.5)
   
   Generated-by: Claude Code (Opus 5.5) following [the 
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
   


-- 
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