*   XLA reduction emitter is deterministic when the environment variable
        `TF_DETERMINISTIC_OPS` is set to "true" or "1". This extends
        deterministic `tf.nn.bias_add` back-prop functionality (and therefore
        also deterministic back-prop of bias-addition in Keras layers) to
        include when XLA JIT compilation is enabled.
    *   Fix problem, when running on a CUDA GPU and when either environment

        String content = "test\ttest\ttest"
        when:
        def actual = UserGuideTransformTask.normalise(content)

        then:
        actual == "test    test    test"
    }

    def usesUnixLineEndings() {
        given:
        String content = "test\r\ntest\r\ntest"
        when:
        def actual = UserGuideTransformTask.normalise(content)

        then:
        actual == "test\ntest\ntest"
    }

        def annotations = []
        jApiType.annotations >> annotations

        when:
        annotations.clear()

        then:
        rule.maybeViolation(jApiType).humanExplanation =~ 'Is not annotated with @Incubating'

        when:
        annotations.add(incubatingAnnotation)

        then:
        rule.maybeViolation(jApiType) == null

        where:
        apiElement    | jApiTypeName

    def canAddMetaData() {
        TestDomainObject value = new TestDomainObject('a')

        when:
        repository.put('class', value)

        then:
        repository.find('class') == value
        repository.get('class') == value
    }

    def findReturnsNullForUnknownClass() {
        expect:

=== Dead Section Links
This section doesn't exist: <<missing_section>>
Also see this one, which is another dead link: <<other_missing_section>>
        """

        when:
        run('checkDeadInternalLinks').buildAndFail()

        then:
        assertFoundDeadSectionLinks(sampleDoc, "missing_section", "other_missing_section")
    }

    def "validates present section links"() {
        given:

   * request, and ensure that (1/QPS) seconds have elapsed since then. For example, for a rate of
   * QPS=5 (5 tokens per second), if we ensure that a request isn't granted earlier than 200ms after
   * the last one, then we achieve the intended rate. If a request comes and the last request was
   * granted only 100ms ago, then we wait for another 100ms. At this rate, serving 15 fresh permits

   * <p>If you are escaping input in arbitrary successive chunks, then it is not generally safe to
   * use this method. If an input string ends with an unmatched high surrogate character, then this
   * method will throw {@link IllegalArgumentException}. You should ensure your input is valid <a
   * href="http://en.wikipedia.org/wiki/UTF-16">UTF-16</a> before calling this method.
   *

fi
set -x

# When cross-compiling with RBE, we need to copy the macOS sysroot to be
# inside the TensorFlow root directory. We then define them as a filegroup
# target inside "tensorflow/tools/toolchains/cross_compile/cc" so that Bazel
# can register it as an input to compile/link actions and send it to the remote
# VMs when needed.

   * <p>If the dataset contains {@link Double#NaN} then the result is {@link Double#NaN}. If it
   * contains {@link Double#NEGATIVE_INFINITY} and not {@link Double#NaN} then the result is {@link
   * Double#NEGATIVE_INFINITY}. If it contains {@link Double#POSITIVE_INFINITY} and finite values
   * only then the result is the lowest finite value. If it contains {@link
   * Double#POSITIVE_INFINITY} only then the result is {@link Double#POSITIVE_INFINITY}.
   *

     * If the given {@code paths} argument contains no element, then this method returns an empty string.
     *
     * <p><b>Examples:</b>
     * If {@code paths} is a list containing two elements, {@code path1} and {@code path2}, then:
     * </p>
     * <ul>
     *   <li>If this type is {@link JavaPathType#MODULES}, then this method returns