// The implementation may use a change index or a commit log for the revision.
	// The config client should not make any assumptions about revisions and rely only on
	// exact equality to implement optimistic concurrency of read-write operations.
	//
	// The lifetime of an object of a particular revision depends on the underlying data store.

* Shade/Shadow
* Jetty
* Checkstyle
* JaCoCo
* AntRun (see further down)

Why does this matter? Because many plugins rely on standard Java conventions, migration is just a matter of replicating the configuration of the Maven plugin in Gradle.
As an example, here's a simple Maven Checkstyle plugin configuration:

[source,xml]
----
...

			// Thus, setup the regFileReader again after merging PAX headers.
			if err := tr.handleRegularFile(hdr); err != nil {
				return nil, err
			}

			// Sparse formats rely on being able to read from the logical data
			// section; there must be a preceding call to handleRegularFile.
			if err := tr.handleSparseFile(hdr, rawHdr); err != nil {
				return nil, err
			}

	}

	corefileMigrationRequired, err := isCoreDNSConfigMapMigrationRequired(corefile, currentInstalledCoreDNSVersion)
	if err != nil {
		return err
	}

	// Assume that migration is always possible, rely on migrateCoreDNSCorefile() to fail if not.
	canMigrateCorefile := true

	if corefile == "" || migration.Default("", corefile) {
		// If the Corefile is empty or default, the latest default Corefile will be applied

		const (
			searchBits = 7
			regionBits = 9
			regionMask = 1<<regionBits - 1
		)
		idx := n >> searchBits
		buf := fromM49[m49Index[idx]:m49Index[idx+1]]
		val := uint16(n) << regionBits // we rely on bits shifting out
		i := sort.Search(len(buf), func(i int) bool {
			return buf[i] >= val
		})
		if r := fromM49[int(m49Index[idx])+i]; r&^regionMask == val {
			return Region(r & regionMask), nil
		}
	}

	ns := apps.Namespace
	labels := map[string]string{
		"request_protocol": "http",
		"response_code":    "200",
		// For out of mesh server, client side metrics rely on endpoint resource metadata
		// to fill in workload labels. To limit size of endpoint resource, we only populate
		// workload name and namespace, canonical service name and version in endpoint metadata.

In Gradle and Ivy, a module can have multiple artifacts.
Each artifact can have a different set of dependencies.
====

[[sub:file_dependencies]]
=== File dependencies

Projects sometimes do not rely on a binary repository product e.g. JFrog Artifactory or Sonatype Nexus for hosting and resolving external dependencies.

)

const (
	ntpServerEnv = "MINIO_NTP_SERVER"
)

var ntpServer = env.Get(ntpServerEnv, "")

// UTCNowNTP - is similar in functionality to UTCNow()
// but only used when we do not wish to rely on system
// time.
func UTCNowNTP() (time.Time, error) {
	// ntp server is disabled
	if ntpServer == "" {
		return time.Now().UTC(), nil
	}
	return ntp.Time(ntpServer)
}

    So, this way you will probably have cleaner code.

If you used to rely on this behavior, now you should create the resources for background tasks inside the background task itself, and use internally only data that doesn't depend on the resources of dependencies with `yield`.

  return false;
}

absl::StatusOr<bool> CreateCycleDetectionGraph(const Graph* graph,
                                               GraphCycles* cycles) {
  for (int i = 0; i < graph->num_node_ids(); ++i) {
    // We rely on the node IDs in the cycle detection graph being consecutive
    // integers starting from 0.
    CHECK_EQ(i, cycles->NewNode());
  }

  // Compute the loop structure of the graph.