# use the wheel's TensorFlow installation instead of the one made available
# through bazel. This must be done in a different root directory, //bazel_pip/...,
# because "import tensorflow" run from the root directory would instead import
# the folder instead of the venv package.
# 
# Pass --config=pip to run the same suite of tests. If you want to run just one
# test for investigation, you'll need --config=pip_venv instead, and then you

  // Returns the operation's device name.
  //
  // The value returned may be different from the one set by SetDeviceName, but
  // it will be compatible with it: the name will be updated by device placement
  // logic to refer to the specific device chosen.
  //
  // Example: If one calls `op->SetDeviceName("/device:GPU")`, the value
  // returned by DeviceName should be "/device:GPU:*" until a particular GPU is

    const char* childpath;
    ASSERT_TRUE(TF_StringStreamNext(children, &childpath));
    ASSERT_EQ(::tensorflow::string(childpath), "somefile.txt");
    // There should only be one file in this directory.
    ASSERT_FALSE(TF_StringStreamNext(children, &childpath));
    ASSERT_EQ(childpath, nullptr);
    TF_StringStreamDone(children);

    TF_FileStatistics stats;

# ==============================================================================
set -euxo pipefail

# Run this from inside the tensorflow github directory.
# Usage: setup_venv_test.sh venv_and_symlink_name "glob pattern for one wheel file"
# Example: setup_venv_test.sh bazel_pip "/tf/pkg/*.whl"
# 
# This will create a venv with that wheel file installed in it, and a symlink
# in ./venv_and_symlink_name/tensorflow to ./tensorflow. We use this for the

  TensorHandlePtr value_one(FloatTensorHandle(3., status.get()));
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  TensorHandlePtr value_two(FloatTensorHandle(-2., status.get()));
  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
  std::array<TFE_TensorHandle*, 2> in_components{value_one.get(),

- TensorFlow installation (pip package or built from source):
- TensorFlow library (version, if pip package or github SHA, if built from source):

### 2. Code

Provide code to help us reproduce your issues using one of the following options:

#### Option A: Reference colab notebooks

# script to run pip-compile for keras, tensorboard, estimator deps.
# if there is a change in requirements.in then all lock files will be updated
# accordingly.

# -e: abort script if one command fails
# -u: error if undefined variable used
# -o pipefail: entire command fails if pipe fails. watch out for yes | ...
# -o history: record shell history
set -euo pipefail -o history

# Check for required arguments

    airtime before a decision is made regarding whether they are to be migrated
    to the core.
*   As every PR requires several CPU/GPU hours of CI testing, we discourage
    submitting PRs to fix one typo, one warning,etc. We recommend fixing the
    same issue at the file level at least (e.g.: fix all typos in a file, fix
    all compiler warnings in a file, etc.)
*   Tests should follow the

  std::unique_ptr<TFE_TensorHandle, decltype(&TFE_DeleteTensorHandle)> one(
      TestScalarTensorHandle(context.get(), 111.f), TFE_DeleteTensorHandle);
  op.reset(TFE_NewOp(context.get(), "AssignVariableOp", status.get()));
  TFE_OpSetAttrType(op.get(), "dtype", TF_FLOAT);
  TFE_OpAddInput(op.get(), var_handle, status.get());
  TFE_OpAddInput(op.get(), one.get(), status.get());
  TFE_OpSetDevice(op.get(), name, status.get());

  template <typename G>
  Cleanup(Cleanup<G>&& src)  // NOLINT
      : released_(src.is_released()), f_(src.release()) {}

  // Assignment to a Cleanup object behaves like destroying it
  // and making a new one in its place, analogous to unique_ptr
  // semantics.
  Cleanup& operator=(Cleanup&& src) {  // NOLINT
    if (!released_) f_();
    released_ = src.released_;
    f_ = src.release();
    return *this;
  }