}

TEST(CAPI, StatusEnum) {
  EXPECT_EQ(TF_OK, static_cast<TF_Code>(tensorflow::error::OK));
  EXPECT_EQ(TF_CANCELLED, static_cast<TF_Code>(tensorflow::error::CANCELLED));
  EXPECT_EQ(TF_UNKNOWN, static_cast<TF_Code>(tensorflow::error::UNKNOWN));
  EXPECT_EQ(TF_INVALID_ARGUMENT,
            static_cast<TF_Code>(absl::StatusCode::kInvalidArgument));
  EXPECT_EQ(TF_DEADLINE_EXCEEDED,

    ASSERT_EQ(errors::OK, s.code()) << s.message();
    x.reset(x_raw);
  }

  PartialTensorShape shape;
  Status s = x->Shape(&shape);
  ASSERT_EQ(errors::OK, s.code()) << s.message();
  ASSERT_TRUE(shape.unknown_rank());
}

TEST_P(UnifiedAPI, TestPartialShapeTracing) {
  if (!UseFunction()) {
    GTEST_SKIP() << "Tracing only test.";
  }
  if (UseMlir()) {

    case "$TF_WHEEL" in
        # CPU:
        *cpu*manylinux*) LARGEST_OK_SIZE=220 ;;
        # GPU:
        *manylinux*)     LARGEST_OK_SIZE=580 ;;
        # Unknown:
        *)
            echo "The wheel's name is in an unknown format."
            exit 1
            ;;
    esac
    # >&3 forces output in bats even if the test passes. See

                                   " dimensions which is over the limit of ",
                                   TensorShape::MaxDimensions(), ".");
  }
  TensorShapeProto proto;
  if (num_dims < 0) {
    proto.set_unknown_rank(true);
  } else {
    for (int d = 0; d < num_dims; ++d) {
      proto.add_dim()->set_size(dims[d]);
    }
  }

  forward_op_->attrs.Set(attr_name, proto);

                                                                    &device)) {
    methods.deallocator(data);
    status->status =
        tensorflow::errors::InvalidArgument(device_name, " unknown device.");
    return nullptr;
  }
  return tensorflow::wrap(new tensorflow::CAPICustomDeviceTensorHandle(
      context, device, *reinterpret_cast<tensorflow::DataType*>(&dtype), data,
      methods));
}

    return;
  }

  if (num_dims == 0) {
    // Output shape is a scalar.
    return;
  }

  // Rank is greater than 0, so fill in the values, if known, and
  // -1 for unknown values.
  for (int i = 0; i < num_dims; ++i) {
    auto dim = ic->Dim(shape, i);
    int64_t value = -1;
    if (ic->ValueKnown(dim)) {
      value = ic->Value(dim);
    }
    dims[i] = value;

# Run any CI script and env configuration to bisect a failing target in some
# build configuration. You must set the following variables to control this
# script:
#
#   TF_BISECT_GOOD: Last known good commit (e.g. commit from the last passing job)
#   TF_BISECT_BAD: First bad commit (e.g. commit from the first failing job)
#   TF_BISECT_SCRIPT: The build script path relative to the TF root dir, e.g.
#     ci/official/wheel.sh

    TF_RETURN_IF_ERROR(
        operation->SetOpName(absl::StrCat("_input_", inputs_.size()).c_str()));
    TF_RETURN_IF_ERROR(operation->SetAttrType("dtype", dtype));
    if (!shape.unknown_rank()) {
      TF_RETURN_IF_ERROR(operation->SetAttrShape(
          "shape", reinterpret_cast<int64_t*>(shape.dim_sizes().data()),
          shape.dims()));
    }
    int num_outputs = 1;

    *   `LinearOperator.matmul` now returns a new `LinearOperator`.
*   New ops and improved op functionality
    *   Add a Nearest Neighbor Resize op.
    *   Add an `ignore_unknown` argument to `parse_values` which suppresses
        ValueError for unknown hyperparameter types. Such * Add
        `tf.linalg.matvec` convenience function.
    *   `tf.einsum()`raises `ValueError` for unsupported equations like
        `"ii->"`.

#
#    - New environments like new CUDA versions, changes to manylinux,
#      compilers, etc. can cause undefined behavior such as build failures
#      or tests passing incorrectly.
#    - Automatic LLVM updates are known to extend build time even with
#      the cache; this is unavoidable.
export TFCI=py311,linux_x86,public_cache,disk_cache

# Recommended: Configure Docker. (Linux only)
#