Suicide trigger prevents later cycles starting

[wxtim]

Closes #6594

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 if present).
• Tests are included (or explain why tests are not needed).
• Changelog entry included if this is a change that can affect users
• Cylc-Doc pull request opened if required at cylc/cylc-doc/pull/XXXX.
• If this is a bug fix, PR should be raised against the relevant ?.?.x branch.

[oliver-sanders]

This undoes a change @hjoliver made in #4970

5e7ffad#diff-b46dc58446bab522ef576c6d2e32a4276f57d60ff62e930e6f47762d47bf9507R296

I think this change makes sense, but will need to be reviewed in the context of whatever #4970 was trying to resolve.

[hjoliver]

Good spotting ...

[oliver-sanders]

(leaving you two to resolve the differences between these changes)

[wxtim]

@hjoliver what do you think we should do about this?

[hjoliver]

Sorry, I got distracted on the way to looking at that...

[hjoliver]

Actually, by testing, it seems that line is no longer needed at all. Correction, 17-c7backcompat-self-suicide-cycling.t fails without the remaining bit of the line.

[wxtim]

Isn't that the test for your earlier change though?

Why did you make this change

Because it appears to work in this case, but on more thought I don't think that its right, I think that I need to have another look. Marked PR as draft pending further thought.

I think that in the example from the ticket

            install_cold[^] => foo? => bar?

            # The line below stops cycle points 1991 and 1992
            # from starting:
            bar:fail? => !foo?

If bar fails, then foo will only have absolute triggers, but if bar succeeds then foo has a cycling trigger (bar:fail). The outcome of this method depends on somthing which hasn't yet happened!

[hjoliver]

Isn't that the test for your earlier change though?

Yes, it's one of the two functional tests added in the old PR.

If bar fails, then foo will only have absolute triggers, but if bar succeeds then foo has a cycling trigger (bar:fail). The outcome of this method depends on somthing which hasn't yet happened!

It's not dynamically determined like that. The trigger exists in the graph, even if runtime events result it in not being needed.

[wxtim]

The problem that leaves us with is that the user expects a to act like it is parentless, but it's not: It's not unreasonable to suggest that the non existance of an optional outcome to optional suicide trigger dependency (b:fail? => !a?) shouldn't stop a from triggering in subsequent cycles.

[wxtim]

Actually, by testing, it seems that line is no longer needed at all. Correction, 17-c7backcompat-self-suicide-cycling.t fails without the remaining bit of the line.

Does it?