Don't release from runahead if paused.

[hjoliver]

These changes close #4278

Requirements check-list

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Applied any dependency changes to both setup.cfg and conda-environment.yml.
• Appropriate tests are included (unit and/or functional).
• Appropriate change log entry included.
• No documentation update required.

[hjoliver]

If starting the #4278 example paused, task pool logging now shows only one task //1/foo:

• * 1/foo waiting(held)(runahead) job:00 flows:1 status=waiting runahead=True queued=False

i.e. we before the preparing stage, so all good.

The initial task pool looks different because subsequent foos have not even been spawned yet, but that's fine.

[hjoliver]

Integration tests fixed and extended.

This could break func tests, if any, that check the task pool at paused start-up. Hopefully there are none, gotta go ... 👋

[hjoliver]

Damn it, one integration test is failing here, but I can't get it to fail locally. That's me done for now though.

[MetRonnie]

I also get test_filter_task_proxies[Task state] passing locally. I think the reason it might fail is because you've changed it so it is no longer paused at start, and the GHA runner must be quicker than our local machines, so the tasks get released and are no longer waiting when the mod_run fixture shuts down?

cylc-flow/tests/integration/utils/flow_tools.py

Lines 78 to 105 in 072acd3

	async def _run_flow(
	run_dir: Path,
	caplog: Optional[pytest.LogCaptureFixture],
	scheduler: Scheduler,
	level: int = logging.INFO
	):
	"""Start a scheduler."""
	contact = (run_dir / scheduler.workflow / WorkflowFiles.Service.DIRNAME /
	WorkflowFiles.Service.CONTACT)
	if caplog:
	caplog.set_level(level, CYLC_LOG)
	task = None
	started = False
	await scheduler.install()
	try:
	task = asyncio.get_event_loop().create_task(scheduler.run())
	started = await _poll_file(contact)
	yield caplog
	finally:
	if started:
	# ask the scheduler to shut down nicely
	async with timeout(5):
	scheduler._set_stop(StopMode.REQUEST_NOW_NOW)
	await task
	if task:
	# leave everything nice and tidy
	task.cancel()

[MetRonnie]

I assume you'll want to re-comment-out this?

Suggested change

	self.pool.log_task_pool(logging.CRITICAL)
	# self.pool.log_task_pool(logging.CRITICAL)

[MetRonnie]

Does the note need updating?

cylc-flow/tests/integration/test_task_pool.py

Lines 28 to 29 in 24147ea

	# NOTE: foo & bar have no parents so at start-up (even with the workflow
	# paused) they are spawned out to the runahead limit.

+# NOTE: foo & bar have no parents so at start-up they are
+spawned out to the runahead limit (only when not paused at start-up).

[oliver-sanders]

Hmm, this would appear to be the reverse of a previous commit:

22e7a20#diff-90711b005fe27b9f894e6cd09f08816d847dcf853518894c6b7e2139784cbcb8R1444

[oliver-sanders]

I think runahead tasks should still be released in "paused" mode, but that the "paused" state should stop them from running.

I.E. as implemented in #4436

[hjoliver]

Good spotting! Admittedly this attempted fix was done in a desperate rush.

[oliver-sanders]

This is a regression of #4436, suggesting #4620 as an alternative.

[MetRonnie]

Closing this, then (if needs be can reopen)