Tested locally, commented on https://issues.apache.org/jira/browse/MNG-8520 (tl;dr: it restores previous behaviour, excess of events are gone)

@alextu Can you confirm?

@alextu Can you confirm?

It looks like it's better, using the reproducer I sent. I'll be able to confirm once it's merged as our test suite runs with our own nightly built from Maven master.

And from where you get the nightly build? You are building it yourselves?

And from where you get the nightly build? You are building it yourselves?

yes on our internal CI. If really needed, I can check tomorrow how to tweak that to build a custom branch instead of master and use that in our tests.

I've built this branch:

mvn --version
Apache Maven 4.0.0-rc-3-SNAPSHOT (129afd91931d6cb8688ab6f03d82c43e6ceb0815)

and run our tests, they're still failing for the same reason(s) 🤔: additional requests that are not traced back to plugins.

"additional requests that are not traced back to plugins" -- isnt this the https://issues.apache.org/jira/browse/MNG-8519 ? Or "additional" as "count of them is still more than was in mvn3"?

"additional requests that are not traced back to plugins" -- isnt this the https://issues.apache.org/jira/browse/MNG-8519 ? Or "additional" as "count of them is still more than was in mvn3"?

The count of them is more, for example, in this test, we see 153 requests with this Maven 4 version and 117 with Maven 3.9.9. We count a request when we receive a pair of RepositoryEvent objects with type ARTIFACT_DOWNLOADING (or METADATA_DOWNLOADING) and ARTIFACT_DOWNLOADED (or METADATA_DOWNLOADED).
As part of this test, we also see that some (all?) of those additional requests cannot be traced back to a plugin or a project.

Resolve #9919