Fix service registration delay causing next service to be skipped

[pierlon]

Summary

Fixes the regression @westonruter reported in #6459 (comment).

When a service's registration is delayed, the service is moved to the end of the array list of services to be registered. When the service is moved to the end, the array's internal pointer moves to the next service in the array, pointing to the next service to be registered. By calling next($services) the array's internal pointer is moved once again to the next service, skipping the to-be service and preventing it from being registered.

See #6484.
Fixes #6459.

Checklist

• My code is tested and passes existing tests.
• My code follows the Engineering Guidelines (updates are often made to the guidelines, check it out periodically).

[github-actions]

Plugin builds for 0fc0374 are ready 🛎️!

• Download development build
• Download production build

[westonruter]

The symptom is fixed:

But tests are failing due to an apparent infinite loop/recursion. Running tests locally I'm getting the same issue as GHA:

.....................S....................................... 2013 / 2439 ( 82%)
................................S.........................Killed

[pierlon]

But tests are failing due to an apparent infinite loop/recursion.

Fixed in 0fc0374.

[westonruter]

@schlessera If you could double check the changes here when you get a chance. In the meantime I'm merging because it fixes the issue.

[schlessera]

That definitely looks off, and I think it only happens to work because we don't have a service yet that has more than one delayed requirement.

If you look at the structure, you can see that the next($services) is located within the foreach ( $missing_requirements as $missing_requirement ) loop, meaning that if you check requirements for a service that has 3 delayed dependencies, it will forward the index into the $services array by 3 entries... that can't be right.

[schlessera]

In case the logic is too complex and finicky with the array index, we can work with a copy of the array that we pop/shift elements out of. That will probably make for cleaner code, even if at a slight performance overhead.

[schlessera]

Ah, nevermind. Due to the return false in the loop, this will never be executed more than once.

[schlessera]

The fix LGTM, but I noticed another problem and opened #6548 for that.

[westonruter]

QA Passed

The ValidationCounts service is initializing as expected: