let summary = "Promote resources reads/writes to function inputs/outputs.";
  let description = [{
    This pass promotes resource accesses in function(s) (by default, the main)
    to input arguments and outputs of the function(s).

    Two types of resources are supported:
    (1) A function argument of TF::ResourceType type (this pass).
    (2) A VarHandleOp in the function (tf-promote-var-handles-to-args).

  for (const FunctionDef& function : actual.function()) {
    actual_index[function.signature().name()] = &function;
  }

  for (const FunctionDef& expected_function : expected.function()) {
    auto it = actual_index.find(expected_function.signature().name());
    if (it == actual_index.end()) {
      if (diff) {
        *diff = absl::StrCat("Did not find expected function '",

		}
	}

	return ret
}

// callReflect is the call implementation used by a function
// returned by MakeFunc. In many ways it is the opposite of the
// method Value.call above. The method above converts a call using Values
// into a call of a function with a concrete argument frame, while
// callReflect converts a call of a function with a concrete argument
// frame into a call using Values.

            ),
        ] = None,
        generate_unique_id_function: Annotated[
            Callable[[routing.APIRoute], str],
            Doc(
                """
                Customize the function used to generate unique IDs for the *path
                operations* shown in the generated OpenAPI.

                This is particularly useful when automatically generating clients or
                SDKs for your API.

  SymbolRefAttr func = op->getAttr("f").template cast<SymbolRefAttr>();
  auto function = symbolTable.lookupNearestSymbolFrom<func::FuncOp>(op, func);
  if (!function) {
    return op.emitError("'f' attribute refers to an undefined function: ")
           << func;
  }

  FunctionType function_ty = function.getFunctionType();
  int func_arg_count = function_ty.getNumInputs();
  int arg_count = op.getArgs().size();

      @def_function.function(
          input_signature=[
              tensor_spec.TensorSpec(
                  name='x', shape=shapes[0], dtype=dtypes.float32
              )
          ]
      )
      def matmul_with_kernel(self, x: core.Tensor) -> Mapping[str, core.Tensor]:
        return self._matmul(x, self._kernel)

      @def_function.function(
          input_signature=[

	// It's ok to ignore the error, returning error is not expected from this function.
	// If an error case occurred during the function generation, this error case(skipped one) will also trigger an error
	// while the generated function is executed. And those errors will be handled during the execution of the generated
	// function with a back off policy.
	if err == nil && attachableVolumePlugin != nil {

    checkNotNull(key);
    checkNotNull(function);
    int hash = hash(key);
    return segmentFor(hash).compute(key, hash, function);
  }

  @Override
  public V computeIfAbsent(K key, Function<? super K, ? extends V> function) {
    checkNotNull(key);
    checkNotNull(function);

	USAGE_MATCH_TYPE_AND = 0
	USAGE_MATCH_TYPE_OR  = 1

	/* msgAndCertEncodingType values for CertOpenStore function */
	X509_ASN_ENCODING   = 0x00000001
	PKCS_7_ASN_ENCODING = 0x00010000

	/* storeProvider values for CertOpenStore function */
	CERT_STORE_PROV_MSG               = 1
	CERT_STORE_PROV_MEMORY            = 2
	CERT_STORE_PROV_FILE              = 3

		// wantAddedToNominated is whether a Pod from the test case should be registered as a nominated Pod in the nominator.
		wantAddedToNominated bool
		// prepareFunc is the function called to prepare pods in the queue before the test case calls Update().
		// This function returns three values;
		// - oldPod/newPod: each test will call Update() with these oldPod and newPod.