refactor: replace go-multierror with native std go errors

[kruskall]

Proposed commit message

Go 1.20 added errors.Join allowing for multierrors. Replace go-multierror usage with native go errors and drop the extra dependency on github.com/hashicorp/go-multierror

Checklist

• My code follows the style guidelines of this project
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have made corresponding change to the default configuration files
• I have added tests that prove my fix is effective or that my feature works
• I have added an entry in CHANGELOG.next.asciidoc or CHANGELOG-developer.next.asciidoc.

Disruptive User Impact

Author's Checklist

• [ ]

How to test this PR locally

Related issues

Use cases

Screenshots

Logs

[mergify]

This pull request does not have a backport label.
If this is a bug or security fix, could you label this PR @kruskall? 🙏.
For such, you'll need to label your PR with:

• The upcoming major version of the Elastic Stack
• The upcoming minor version of the Elastic Stack (if you're not pushing a breaking change)

To fixup this pull request, you need to add the backport labels for the needed
branches, such as:

• backport-v8./d.0 is the label to automatically backport to the 8./d branch. /d is the digit

[elasticmachine]

Pinging @elastic/elastic-agent-data-plane (Team:Elastic-Agent-Data-Plane)

[belimawr]

@kruskall, thanks for the PR! That's a great refactoring I'd love to see merged.

However, as it is, I believe it will break the logic on our, there are a couple of places we rely the multierror.MultiError API:

beats/x-pack/libbeat/management/managerV2.go

Lines 818 to 847 in db9406b

	if err := obj.Reload(inputBeatCfgs); err != nil {
	merror := &multierror.MultiError{}
	realErrors := multierror.Errors{}
	// At the moment this logic is tightly bound to the current RunnerList
	// implementation from libbeat/cfgfile/list.go and Input.loadStates from
	// filebeat/input/log/input.go.
	// If they change the way they report errors, this will break.
	// TODO (Tiago): update all layers to use the most recent features from
	// the standard library errors package.
	if errors.As(err, &merror) {
	for _, err := range merror.Errors {
	causeErr := errors.Unwrap(err)
	// A Log input is only marked as finished when all events it
	// produced are acked by the acker so when we see this error,
	// we just retry until the new input can be started.
	// This is the same logic used by the standalone configuration file
	// reloader implemented on libbeat/cfgfile/reload.go
	inputNotFinishedErr := &common.ErrInputNotFinished{}
	if ok := errors.As(causeErr, &inputNotFinishedErr); ok {
	cm.logger.Debugf("file '%s' is not finished, will retry starting the input soon", inputNotFinishedErr.File)
	cm.forceReload = true
	cm.logger.Debug("ForceReload set to TRUE")
	continue
	}
	// This is an error that cannot be ignored, so we report it
	realErrors = append(realErrors, err)
	}
	}

beats/x-pack/libbeat/management/managerV2.go

Lines 676 to 687 in db9406b

	if err := cm.reloadInputs(inputUnits); err != nil {
	merror := &multierror.MultiError{}
	if errors.As(err, &merror) {
	for _, err := range merror.Errors {
	unitErr := cfgfile.UnitError{}
	if errors.As(err, &unitErr) {
	unitErrors[unitErr.UnitID] = append(unitErrors[unitErr.UnitID], unitErr.Err)
	delete(healthyInputs, unitErr.UnitID)
	}
	}
	}
	}

Those bits would need updating as part of this PR.

[kruskall]

Thanks for the feedback! 🙇

I think the code you linked is using github.com/joeshaw/multierror while this PR is removing github.com/hashicorp/go-multierror. The type is different and nothing should change for the errors.As call. Am I missing something ?

[belimawr]

Thanks for the feedback! 🙇

I think the code you linked is using github.com/joeshaw/multierror while this PR is removing github.com/hashicorp/go-multierror. The type is different and nothing should change for the errors.As call. Am I missing something ?

You're right 🤦‍♂️. I forgot we use two multierror libs...

Let me check with the team if anybody can think of any unseen issue with this change.