potiuk opened a new pull request, #70888: URL: https://github.com/apache/airflow/pull/70888
`start_trigger_args` is a class attribute on `DateTimeSensorAsync`, `TimeSensor` and `FileSensor`, and each one assigned *through* it in `__init__`: ```python self.start_trigger_args.trigger_kwargs = dict(moment=self._moment, ...) ``` That mutates the shared class-level object instead of creating an instance attribute, so every task built from the operator ends up advertising the arguments of whichever task was constructed last. Reproducer on `main`: ```python a = DateTimeSensorAsync(task_id="a", target_time=datetime(2030, 1, 1), start_from_trigger=True) b = DateTimeSensorAsync(task_id="b", target_time=datetime(2040, 6, 6), start_from_trigger=True) a.start_trigger_args is b.start_trigger_args # True a.start_trigger_args.trigger_kwargs["moment"] # 2040-06-06 <- b's moment ``` With `start_from_trigger=True` the task starts directly in the triggerer from these arguments, so two `DateTimeSensorAsync` tasks in one Dag both wait for the same moment. The same applies to `TimeSensor`'s moment and to `FileSensor`'s `filepath`, `recursive`, `poke_interval` and `timeout`. Each sensor now replaces `start_trigger_args` with a fresh copy via `dataclasses.replace`, which binds an instance attribute and leaves the class-level template untouched for the next task built from it. Regression tests are added for all three sensors; each fails on `main`. --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes — Claude Code (Opus 5) Generated-by: Claude Code (Opus 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]
