*
     * The required runtimes for a project should be set if that project has
     * specific runtime requirements. For example, they must be able to run in
     * a worker (a compiler worker, test worker, worker API worker action),
     * or must be able to run in a client (TAPI model, TAPI client, CLI client).
     */
    @get:Nested
    abstract val requiredRuntimes: ModuleTargetRuntimes

}

extensions.create<PluginPublishExtension>("pluginPublish", gradlePlugin)

tasks.validatePlugins {
    enableStricterValidation = true
}

// Remove gradleApi() and gradleTestKit() as we want to compile/run against Gradle modules
// TODO consider splitting `java-gradle-plugin` to provide only what's necessary here
configurations.all {
    withDependencies {
        remove(project.dependencies.gradleApi())

 * iteratorTester.testForEachRemaining();
 * }
 *
 * <p><b>Note</b>: It is necessary to use {@code IteratorTester.KnownOrder} as shown above, rather
 * than {@code KnownOrder} directly, because otherwise the code cannot be compiled.
 *
 * @author Kevin Bourrillion
 * @author Chris Povirk
 */
@GwtCompatible
@NullMarked
public abstract class IteratorTester<E extends @Nullable Object>
    extends AbstractIteratorTester<E, Iterator<E>> {

/**
 * Converts raw javadoc comments into docbook.
 */
public class JavadocConverter {
    private static final Pattern HEADER_PATTERN = Pattern.compile("h(\\d)", Pattern.CASE_INSENSITIVE);
    private static final Pattern ACCESSOR_COMMENT_PATTERN = Pattern.compile("(?:returns|sets)\\s+(the|whether)\\s+", Pattern.CASE_INSENSITIVE);
    private final Document document;
    private final JavadocLinkConverter linkConverter;

    } catch (e: IllegalAccessException) {
      throw AssertionError("failed to get ALPN selected protocol", e)
    }
  }

  /**
   * Handle the methods of ALPN's ClientProvider and ServerProvider without a compile-time
   * dependency on those interfaces.
   */
  private class AlpnProvider(
    /** This peer's supported protocols. */
    private val protocols: List<String>,
  ) : InvocationHandler {

  * For polynull public APIs, the solution has to be decided on a case-by-case basis.

Do not remove `null` checks on public API boundaries, even if the annotations (or rather lack of them) suggest this.
Not all client code is compiled with NullAway.
Do not make a parameter `@Nullable` just to check and throw a `NullPointerException` when it is `null`. 

## Consequences

* Classes are forced to have consistent nullability annotations.

        // For Java compilation incremental analysis
        "it.unimi.dsi.fastutil.ints.IntOpenHashSet",
        "it.unimi.dsi.fastutil.ints.IntSets",
        // For the embedded Kotlin compiler
        "it.unimi.dsi.fastutil.ints.Int2ObjectMap",
        "it.unimi.dsi.fastutil.ints.Int2ObjectOpenHashMap",
        "it.unimi.dsi.fastutil.objects.Object2IntMap",

 * <p>
 * The path matchers created by this service support Maven's traditional include/exclude
 * pattern syntax, which is compatible with the behavior of Maven 3 plugins like
 * maven-compiler-plugin and maven-clean-plugin.
 * <p>
 * Pattern syntax supports:
 * <ul>
 *   <li>Standard glob patterns with {@code *}, {@code ?}, and {@code **} wildcards</li>

        InternalEntry entry = map.getEntry(sample);
        if (entry != null) {
          Object canonical = entry.getKey();
          if (canonical != null) { // only matters if weak/soft keys are used
            // The compiler would know this is safe if not for our use of raw types (see above).
            @SuppressWarnings("unchecked")
            E result = (E) canonical;
            return result;
          }
        }

Pero aquí está el punto clave.

El tema de seguridad e inyección de dependencias se escribe una vez.

Y puedes hacerlo tan complejo como desees. Y aún así, tenerlo escrito solo una vez, en un solo lugar. Con toda la flexibilidad.

Pero puedes tener miles de endpoints (*path operations*) usando el mismo sistema de seguridad.