running the `tf.train.Server`.

## Untrusted inputs during training and prediction

TensorFlow supports a wide range of input data formats. For example it can
process images, audio, videos, and text. There are several modules specialized
in taking those formats, modifying them, and/or converting them to intermediate
formats that can be processed by TensorFlow.

These modifications and conversions are handled by a variety of libraries that

  client_job->mutable_tasks()->insert(
      {0, strings::StrCat("localhost:", client_port)});
  server_def.set_job_name("localhost");
  auto serialized = server_def.SerializeAsString();
  TFE_ContextSetServerDef(ctx, 0, serialized.data(), serialized.size(), status);
  EXPECT_EQ(TF_GetCode(status), TF_OK) << TF_Message(status);

  EXPECT_EQ(TF_GetCode(status), TF_OK) << TF_Message(status);

        ->mutable_rewrite_options()
        ->set_min_graph_nodes(-1);
    string serialized_config;
    ASSERT_TRUE(config.SerializeToString(&serialized_config));
    TFE_OpSetAttrString(
        op, "config_proto",
        reinterpret_cast<const void*>(serialized_config.c_str()),
        serialized_config.length());
  }

  if (use_tfrt) {
    // Set some test-only graph compiler options.

  TF_ASSERT_OK(OpRegistry::Global()->LookUpOpDef("Add", &expected_op_def));
  string expected_serialized;
  expected_op_def->SerializeToString(&expected_serialized);
  string actual_string(reinterpret_cast<const char*>(buffer->data),
                       buffer->length);
  EXPECT_EQ(expected_serialized, actual_string);

  TF_GraphGetOpDef(graph, "MyFakeOp", buffer, status);

                  server_def, tensorflow::Env::Default(), &worker_server2)
                  .ok());
  ASSERT_TRUE(worker_server2->Start().ok());

  TFE_ContextSetServerDef(context.get(), 0, serialized.data(),
                          serialized.size(), status.get());
  EXPECT_EQ(TF_OK, TF_GetCode(status.get())) << TF_Message(status.get());

  BasicTestsForTwoDevices(context.get(),

// Get a list of input OpInfo::TensorProperties given node name.
// Return the serialized list `properties`.
TF_CAPI_EXPORT extern void TF_GetInputPropertiesList(
    TF_GraphProperties* graph_properties, const char* name,
    TF_Buffer** properties, int num_values, TF_Status* status);

// Get a list of output OpInfo::TensorProperties given node name.
// Return the serialized list `properties`.
TF_CAPI_EXPORT extern void TF_GetOutputPropertiesList(

                                       int* num_retvals, TF_Status* status);

// Add a function (serialized FunctionDef protocol buffer) to ctx so
// that it can be invoked using TFE_Execute.
TF_CAPI_EXPORT extern void TFE_ContextAddFunctionDef(
    TFE_Context* ctx, const char* serialized_function_def, size_t size,
    TF_Status* status);

// Adds a function (created from TF_GraphToFunction or

TF_CAPI_EXPORT extern void TFE_OpAttrsSerialize(const TFE_OpAttrs* attrs,
                                                TF_Buffer* buf,
                                                TF_Status* status);

// Set an op's attribute from a serialized AttrValue protocol buffer.
//
// Analogous to TF_SetAttrValueProto for building graph operations.
TF_CAPI_EXPORT extern void TFE_OpSetAttrValueProto(const TFE_Op* op,

*   Fixes a null-dereference when specializing tensor type
    ([CVE-2022-23570](https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-23570))
*   Fixes a crash when type cannot be specialized
    ([CVE-2022-23572](https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-23572))
*   Fixes a heap OOB read/write in `SpecializeType`
    ([CVE-2022-23574](https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-23574))

  TFE_Context* ctx = TFE_NewContext(opts, status);
  EXPECT_EQ(TF_GetCode(status), TF_OK) << TF_Message(status);
  TFE_ContextSetServerDefWithTimeout(ctx, 0, serialized_server_def.data(),
                                     serialized_server_def.size(),
                                     init_timeout_in_ms, status,
                                     /*clear_existing_contexts=*/false);