sudo modprobe loop
	for i in 1 2 3 4; do
		dd if=/dev/zero of=${WORK_DIR}/disks/img.${i} bs=1M count=2000
		device=$(sudo losetup --find --show ${WORK_DIR}/disks/img.${i})
		sudo mkfs.ext4 -F ${device}
		mkdir -p ${WORK_DIR}/mnt/disk${i}/
		sudo mount ${device} ${WORK_DIR}/mnt/disk${i}/
		sudo chown "$(id -u):$(id -g)" ${device} ${WORK_DIR}/mnt/disk${i}/
	done
	set +e
}

			// - mount.Path doesn't match (means cross-device mount), should error out.
			if mount.Path != path {
				crossMounts = append(crossMounts, mount)
			}
		}
	}
	msg := `Cross-device mounts detected on path (%s) at following locations %s. Export path should not have any sub-mounts, refusing to start.`
	if len(crossMounts) > 0 {
		// if paths didn't match then we do have cross-device mount.
		return fmt.Errorf(msg, path, crossMounts)

==============================================================================*/

#ifndef TENSORFLOW_C_EAGER_CUSTOM_DEVICE_TESTUTIL_H_
#define TENSORFLOW_C_EAGER_CUSTOM_DEVICE_TESTUTIL_H_

// A simple logging device to test custom device registration.
#include <memory>

#include "tensorflow/c/c_api.h"
#include "tensorflow/c/eager/c_api.h"
#include "tensorflow/c/eager/c_api_experimental.h"
#include "tensorflow/c/tf_status.h"

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

  // The device of var_handle0 is local device which is the same before and
  // after cluster update. Remove resource with valid device should succeed.
  TFE_Op* op = TFE_NewOp(ctx, "DestroyResourceOp", status);
  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_OpAddInput(op, var_handle0, status);

[CUDA-enabled GPU cards](https://www.tensorflow.org/install/gpu) *(Ubuntu and
Windows)*:

```
$ pip install tensorflow
```

Other devices (DirectX and MacOS-metal) are supported using
[Device plugins](https://www.tensorflow.org/install/gpu_plugins#available_devices).

A smaller CPU-only package is also available:

```
$ pip install tensorflow-cpu
```

	}
	// Failure case where we detected successfully cross device mounts.
	{
		absPaths := []string{"/path/to/1"}
		if err = checkCrossDevice(absPaths, mountsPath); err == nil {
			t.Fatal("Expected to fail, but found success")
		}

		mp := []mountInfo{
			{"/dev/2", "/path/to/1/2", "type2", []string{"flags"}, "2", "2"},
		}

  virtual ImmediateExecutionContext* GetContext() const = 0;

  // Following two methods are used to support custom device.
  // Return true if the inputs contain custom device tensor handle. It means
  // that the argument need to be handled by a custom device.
  virtual bool HasCustomDeviceInput() const = 0;

  virtual const tensorflow::OpDef* OpDef() const = 0;

// Create a variable handle with name `variable_name` on a device with name
// `device_name`.
TFE_TensorHandle* CreateVarHandle(TFE_Context* ctx,
                                  const tensorflow::string& device_name,
                                  const tensorflow::string& variable_name);

// Create a variable with value `value` and name `variable_name` on a device
// with name `device_name`.

	if len(udids) == 0 {
		fail("no udid found; is a device connected?")
	}

	mps := detectMobileProvisionFiles(udids)
	if len(mps) == 0 {
		fail("did not find mobile provision matching device udids %q", udids)
	}

	fmt.Println("# Available provisioning profiles below.")
	fmt.Println("# NOTE: Any existing app on the device with the app id specified by GOIOS_APP_ID")

  }
  absl::Status s;
  const char* device_name = DeviceName(&s);
  if (!s.ok()) {
    device_name = "<error fetching device name>";
  }
  return absl::StrCat("TensorHandle(", value_string, ", shape=", shape_string,
                      ", dtype=", DataType_Name(DataType()), ", device=\"",
                      device_name, "\")");
}

absl::Status ImmediateExecutionTensorHandle::SummarizeValue(