* can take 1-10 seconds to find colliding objects. Mocking frameworks that
     * can do magic to mock static method calls still can't do so for a system
     * class, so we need the indirection. In production, Hotspot should still
     * recognize that the call is 1-morphic and should still be willing to
     * inline it if necessary.
     */
    int identityHashCode(Object object) {

    assertEquals("{a=[a1, a4, a4], b=[b6]}", transformed.toString());
  }

  @J2ktIncompatible // Synchronized
  public void testSynchronizedMultimapSampleCodeCompilation() {
    // Extra indirection for J2KT, to avoid error: not enough information to infer type variable K
    this.<@Nullable Object, @Nullable Object>genericTestSynchronizedMultimapSampleCodeCompilation();
  }

  @J2ktIncompatible // Synchronized

     * can take 1-10 seconds to find colliding objects. Mocking frameworks that
     * can do magic to mock static method calls still can't do so for a system
     * class, so we need the indirection. In production, Hotspot should still
     * recognize that the call is 1-morphic and should still be willing to
     * inline it if necessary.
     */
    int identityHashCode(Object object) {

- Dynamic Resource Allocation (DRA): in the `pod.spec.recourceClaims` array, the `source` indirection is no longer necessary. Instead of e.g. `source: resourceClaimTemplateName: my-template`, one can write `resourceClaimTemplateName: my-template`. ([#125116](https://github.com/kubernetes/kubernetes/pull/125116), [@pohly](https://github.com/pohly))...

    "1" also makes cuDNN convolution and max-pooling operate deterministically.
    This makes Keras Conv\*D and MaxPool\*D layers operate deterministically in
    both the forward and backward directions when running on a CUDA-enabled GPU.

## Breaking Changes

*   Deletes `Operation.traceback_with_start_lines` for which we know of no
    usages.