}

    /**
     * Adds {@code node} if it is not already present.
     *
     * <p><b>Nodes must be unique</b>, just as {@code Map} keys must be. They must also be non-null.
     *
     * @return this {@code Builder} object
     */
    @CanIgnoreReturnValue
    public ImmutableValueGraph.Builder<N, V> addNode(N node) {
      mutableValueGraph.addNode(node);

        .fastFallback(false)
        .callTimeout(1_000, TimeUnit.MILLISECONDS)
        .build()
    val call = client.newCall(Request(url))
    assertFailsWith<IOException> {
      call.execute()
    }.also { expected ->
      assertThat(expected).hasMessage("timeout")
    }
  }

  /**
   * This test reproduces a crash where OkHttp attempted to use a deferred connection when the call

   * is permitted and similarly for decrease/remove(). We assume that a
   * setCount() no-op is permitted if either add() or remove() is permitted,
   * though we also allow it to "succeed" if neither is permitted.
   */

  private void assertSetCount(E element, int count) {
    setCountCheckReturnValue(element, count);

    assertEquals(

we've pushed its bytes onto the socket. Otherwise, if the write fails we are unable to deliver its IOException to the application. We would have told the application layer that the write succeeded, but it didn't!

The application-layer can also do blocking reads. If the application asks to read and there's nothing available, we need to hold that thread until either the bytes arrive, the stream is closed, or a timeout elapses. If we get bytes but there's nobody asking for them, we buffer them....

        shift                   # remove old arg
        set -- "$@" "$arg"      # push replacement arg
    done
fi


# Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
DEFAULT_JVM_OPTS='"-Xmx64m" "-Xms64m"'

# Collect all arguments for the java command:

    return new CompactLinkedHashSet<>(expectedSize);
  }

  private static final int ENDPOINT = -2;

  // TODO(user): predecessors and successors should be collocated (reducing cache misses).
  // Might also explore collocating all of [hash, next, predecessor, successor] fields of an
  // entry in a *single* long[], though that reduces the maximum size of the set by a factor of 2

  /**

          </description>
        </field>
      </fields>
      <codeSegments>
        <codeSegment>
          <version>1.2.0+</version>
          <comment>Generated hashCode() and equals() based on identifier also calls its super, which breaks comparison
          </comment>
          <code>
            <![CDATA[
    /**
     * Computes a hash value based on {@link #getType()} and {@link #getProvides()} values.
     */

## Example Lambda handler

Install the necessary dependencies.
```sh
pip install flask requests
```

 * instances. A discrete domain is one that supports the three basic operations: {@link #next},
 * {@link #previous} and {@link #distance}, according to their specifications. The methods {@link
 * #minValue} and {@link #maxValue} should also be overridden for bounded types.
 *
 * <p>A discrete domain always represents the <i>entire</i> set of values of its type; it cannot
 * represent partial domains such as "prime integers" or "strings of length 5."
 *

     * setException(stackOverflowError) would fail:
     *
     * - If the stack overflowed before set()/setValue() could even store the result in the output
     * Future, then a call setException() would likely also overflow.
     *
     * - If the stack overflowed after set()/setValue() stored its result, then a call to
     * setException() will be a no-op because the Future is already done.
     *