cylc
Don't process messages from non-pool tasks.
Close #6686
Bug caused by job polling callback, with a delayed poll result, on tasks that have already left the task pool as completed. (The core scheduler handles this OK, but an erroneous state change was logged and registered with the datastore).
We already have a task proxy flag whose exact purpose is to indicate "not task pool": TaskProxy.transient.
So the easiest way to fix this bug seems to be:
- set the transient flag on a task proxy when it is removed from the pool
- don't process task messages if the target task is marked as transient
Check List
- [x] I have read
CONTRIBUTING.mdand added my name as a Code Contributor. - [x] Contains logically grouped changes (else tidy your branch by rebase).
- [x] Does not contain off-topic changes (use other PRs for other changes).
- [x] Applied any dependency changes to both
setup.cfg(andconda-environment.ymlif present). - [ ] Tests are included (or explain why tests are not needed).
- [ ] Changelog entry included if this is a change that can affect users
- [x] Cylc-Doc pull request opened if required at cylc/cylc-doc/pull/XXXX.
- [x] If this is a bug fix, PR should be raised against the relevant
?.?.xbranch.
(If test failures are meaningful, might have to consider a different approach such as checking task completion before processing messages...).
(OK the initial two-line "fix" is a bit too simplistic - turns out we are calling process_message artificially, with a transient task proxy, to spawn on the outputs of inactive tasks targeted by cylc set).
Is there any relationship to https://github.com/cylc/cylc-flow/issues/6412?
Does this relate to https://github.com/cylc/cylc-flow/issues/4513 ?
That issue predates "transient" task proxies.
For reference, "transient task proxies" explained here: https://github.com/cylc/cylc-flow/issues/6412#issuecomment-2746488195
