new TensorSliceReader::VarToDataTypeMap);
  v2_reader_->Seek(kHeaderEntryKey);
  for (v2_reader_->Next(); v2_reader_->Valid(); v2_reader_->Next()) {
    if (filtered_keys.count(string(v2_reader_->key())) > 0) continue;
    CHECK(entry.ParseFromArray(v2_reader_->value().data(),
                               v2_reader_->value().size()))
        << entry.InitializationErrorString();
    string key(v2_reader_->key());

 	    {
 	      cp += 2 * nhere;
 	      continue;
 	    }
 
-	  if (cp[nhere - 1] > usrc[nhere -1])
+	  if (cp[nhere - 1] > usrc[nhere - 1])
 	    {
 	      cp += 2 * nhere;
 	      continue;
 	    }
 
-	  if (cp[2 * nhere - 1] < usrc[nhere -1])
+	  if (cp[2 * nhere - 1] < usrc[nhere - 1])
 	    {
 	      cp += 2 * nhere;
 	      continue;
diff --git a/locale/xlocale.h b/locale/xlocale.h

      continue;
    }
    auto tag = tensorflow::strings::StrCat("Device #", i, " (", name, ")");
    // Copy to device
    TFE_TensorHandle* hdevice =
        TFE_TensorHandleCopyToDevice(hcpu, ctx, name.c_str(), status.get());
    if (TF_GetCode(status.get()) != TF_OK) {
      ADD_FAILURE() << tag << " -- " << TF_Message(status.get());
      continue;
    }

    if (op_id_it == tensor_tape.end()) {
      continue;
    }
    int64_t op_id = op_id_it->second;
    auto op_it = op_tape->find(op_id);
    auto result_op_it = result.op_tape.find(op_id);
    if (op_id == -1 || op_it == op_tape->end() ||
        result_op_it != result.op_tape.end()) {
      continue;
    }
    CHECK(result.op_tape.emplace(op_id, op_it->second).second);

    const Tensor& src = outputs[i];
    if (!src.IsInitialized() || src.NumElements() == 0) {
      c_outputs[i] =
          EmptyTensor(static_cast<TF_DataType>(src.dtype()), src.shape());
      continue;
    }
    c_outputs[i] = TF_TensorFromTensor(src, &status->status);
    if (!status->status.ok()) return;
  }
}

extern "C" {

void TF_Run(TF_DeprecatedSession* s, const TF_Buffer* run_options,

                    /*use_function=*/use_function);
  ASSERT_EQ(errors::OK, s.code()) << s.message();

  for (int i = 0; i < num_inputs; ++i) {
    if (!outputs[i]) continue;

    AbstractTensorHandlePtr numerical_grad;
    {
      AbstractTensorHandle* numerical_grad_raw;
      s = CalcNumericalGrad(ctx, model, inputs,
                            /*input_index=*/i, use_function,

    // make the pointers invalid.
    all_input_tensors.reserve(num_inputs);
    for (int i = 0; i < num_inputs; ++i) {
      if (input_tensors[i] == nullptr) continue;
      all_input_tensors.emplace_back();
      Tensor& input_tensor = all_input_tensors.back();
      status->status = TF_TensorToTensor(input_tensors[i], &input_tensor);
      if (!status->status.ok()) return;

  current_url = None
  for i in range(len(log_lines)):
    line = log_lines[i]
    result_store_line_match = re.search(RESULT_STORE_LINK_RE, line)
    if not result_store_line_match:
      continue

    url = result_store_line_match.group(1)
    url_lines = result_store_links.setdefault(url, {})
    # Each bazel RBE invocation should produce two

# downloads with Invoke-WebRequest not show the progress bar and is MUCH faster).
SHELL ["powershell.exe", "-ExecutionPolicy", "Bypass", "-Command", "$ErrorActionPreference='Stop'; $ProgressPreference='SilentlyContinue'; $VerbosePreference = 'Continue';"]

# Workaround for networking (b/112379377) was closed as won't fix for MTU setting.
# Remaining lines handle making the metadata server on the VM accessible inside docker.

      d="$${d#bazel-out/*/bin/}"

      if [[ $${d} == *local_config_* ]]; then
        continue
      fi

      if [[ $${d} == external* ]]; then
        extname="$${d#*external/}"
        extname="$${extname%%/*}"
        if [[ $${TF_SYSTEM_LIBS:-} == *$${extname}* ]]; then
          continue
        fi

        d="$${d#*external/farmhash_archive/src}"