fix: possible nil ptr for invalid records like too long name

[szuecs]

fix: possible nil ptr for invalid records like too long name

see also: https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/DomainNameFormat.html

closes #4293

Checklist

• Unit tests updated

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please ask for approval from szuecs. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[mloiseleur]

Mmmmh 🤔
Wdyt about avoiding to append it when it's nil ?

Something like this on line 399

				ep := endpoint.NewEndpointWithTTL(wildcardUnescape(aws.StringValue(r.Name)), aws.StringValue(r.Type), ttl, targets...)
                if ep != nil {
				  if aws.StringValue(r.Type) == endpoint.RecordTypeCNAME {
					ep = ep.WithProviderSpecific(providerSpecificAlias, "false")
				  }
				  newEndpoints = append(newEndpoints, ep)
                }

And add a nil check also on line 412:

				ep := endpoint.
					NewEndpointWithTTL(wildcardUnescape(aws.StringValue(r.Name)), endpoint.RecordTypeA, ttl, aws.StringValue(r.AliasTarget.DNSName))
                if ep != nil {
				  ep = ep.WithProviderSpecific(providerSpecificEvaluateTargetHealth, fmt.Sprintf("%t", aws.BoolValue(r.AliasTarget.EvaluateTargetHealth))).
					  WithProviderSpecific(providerSpecificAlias, "true")
				  newEndpoints = append(newEndpoints, ep)
                }

[yurrriq]

Should we add a test for a label >63 chars? I've been using this-is-an-exceedingly-long-label-that-external-dns-should-reject locally which is 65.

[yurrriq]

If I understand correctly, records() is a callback for ListResourceRecordSetsPagesWithContext() so isn't it a fair assumption that all endpoints here are non-nil?

[szuecs]

Some slack discussion happened with #4293, so I will close this one in favor of the upcoming changes.