bug(rush-lib,operation-weighting): UNASSIGNED_OPERATION causing memory leak

[aramissennyeydd]

Summary

Followup to #4672

Having a weight of 0 for UNASSIGNED_OPERATION was causing a build up of objects for the instance that was waiting. This quickly causes a GC error, locally I can trigger it in 30 seconds, in our CI it's about 2 mins.

Details

I tried to upgrade our cobuild instance this morning and ran into this issue. Monkeypatching the weight to 1 locally fixes the issue.

This is caused by the weight of 0 being an immediate execution, UNASSIGNED_OPERATIONs must have an actual weight so the process is stuck waiting for some time.

How it was tested

Tested this in our internal rush instance, I also tested this with the cobuild test project by setting the timeout in the build.js script to a long interval (2000 seconds). The instance that is waiting will throw a GC error pretty quickly.

With this change, the waiting process is stable and doesn't cause a GC error.

Impacted documentation

None.

[iclanton]

This is an interesting issue. Is the underlying issue inside Async.forEachAsync?

[aramissennyeydd]

@iclanton I don't think so? I think this is caused by a value of 0 being an automatic execution item, so it just keeps throwing unassigned operations at the queue.

FYI: I'm looking into a separate issue with the weighting that came up when I ran this in CI - not seeing the weighting be respected during execution.