const char tensor1[] = {5, 7};
  const int64_t dims1[] = {1, 2};
  const size_t ndims1 = TF_ARRAYSIZE(dims1);

  const char tensor2[] = {2, 4, 6, 8};
  const int64_t dims2[] = {2, 2};
  const size_t ndims2 = TF_ARRAYSIZE(dims2);

  auto desc = init("list(tensor)");
  TF_Tensor* tmp[] = {
      Int8Tensor(dims1, ndims1, tensor1),
      Int8Tensor(dims2, ndims2, tensor2),
  };

  constexpr int64_t dims[] = {100, 100};
  constexpr int num_elements = dims[0] * dims[1];
  float data[num_elements];
  for (int i = 0; i < num_elements; ++i) {
    data[i] = 1.0f;
  }
  TF_Status* status = TF_NewStatus();
  TF_Tensor* t = TFE_AllocateHostTensor(ctx, TF_FLOAT, &dims[0],
                                        sizeof(dims) / sizeof(int64_t), status);

    std::vector<DimensionHandle> dims;
    const TF_ShapeAndType& input_shape = input_shapes->items[i];
    if (input_shape.num_dims == InferenceContext::kUnknownRank) {
      c.SetInput(i, c.UnknownShape());
      continue;
    }
    dims.reserve(input_shape.num_dims);
    for (int j = 0; j < input_shape.num_dims; ++j) {
      dims.push_back(c.MakeDim(input_shape.dims[j]));
    }
    c.SetInput(i, c.MakeShape(dims));

  */

  // Build an abstract input tensor.
  int64_t dims[] = {2, 2};  // Matrices will be 2 x 2
  int num_dims = sizeof(dims) / sizeof(dims[0]);

  float vals[] = {0.0f, 0.0f, 0.0f, 0.0f};
  TFE_Context* eager_ctx = TF_ExecutionContextGetTFEContext(ctx, status.get());
  TFE_TensorHandle* t =
      TestMatrixTensorHandleWithInput(eager_ctx, vals, dims, num_dims);

      return absl::OkStatus();
    }

    std::vector<int64_t> dims(num_dims, kUnknownDim);
    TF_GraphGetTensorShape(graph_, output_,
                           reinterpret_cast<int64_t*>(dims.data()), num_dims,
                           &status);
    TF_RETURN_IF_ERROR(StatusFromTF_Status(&status));
    TF_RETURN_IF_ERROR(tensorflow::TensorShapeUtils::MakeShape(dims, shape));

    return absl::OkStatus();
  }

    int num_dims = output_shapes->items[0].num_dims;
    int64_t* dims = output_shapes->items[0].dims;

    if (!expected_shape.has_value()) {
      EXPECT_EQ(num_dims, -1);
      EXPECT_EQ(dims, nullptr);
      return;
    }

    EXPECT_EQ(num_dims, expected_shape->size());
    for (size_t i = 0; i < num_dims; ++i) {
      EXPECT_EQ(dims[i], (*expected_shape)[i]);
    }

                                                  float data[], int64_t dims[],
                                                  int num_dims);

// Get a Matrix TensorHandle with given float values and dimensions
TFE_TensorHandle* TestTensorHandleWithDimsFloat(TFE_Context* ctx, float data[],
                                                int64_t dims[], int num_dims);

// Get a Matrix TensorHandle with given int values and dimensions

    int dim1 = 1 << 16;
    int dim2 = 1 << 15;
    assertThat(dim1 * dim2).isLessThan(0);
    testConcatOverflow(dim1, dim2);
  }

  @GwtIncompatible // different overflow behavior; could probably be made to work by using ~~
  public void testConcat_overflow_nonNegative() {
    int dim1 = 1 << 16;
    int dim2 = 1 << 16;
    assertThat(dim1 * dim2).isAtLeast(0);
    testConcatOverflow(dim1, dim2);
  }

    int dim1 = 1 << 16;
    int dim2 = 1 << 15;
    assertThat(dim1 * dim2).isLessThan(0);
    testConcatOverflow(dim1, dim2);
  }

  @GwtIncompatible // different overflow behavior; could probably be made to work by using ~~
  public void testConcat_overflow_nonNegative() {
    int dim1 = 1 << 16;
    int dim2 = 1 << 16;
    assertThat(dim1 * dim2).isAtLeast(0);
    testConcatOverflow(dim1, dim2);
  }

  virtual absl::Status SetAttrBool(const char* attr_name, bool value) = 0;
  virtual absl::Status SetAttrType(const char* attr_name, DataType value) = 0;
  virtual absl::Status SetAttrShape(const char* attr_name, const int64_t* dims,
                                    const int num_dims) = 0;
  virtual absl::Status SetAttrShape(const char* attr_name,
                                    const PartialTensorShape shape);