TestHandleFromDLPack(status, ctx, {4}, {});
  TestHandleFromDLPack(status, ctx, {4}, {1});
  TestHandleFromDLPack(status, ctx, {4, 3, 2}, {});
  TestHandleFromDLPack(status, ctx, {4, 3, 2}, {6, 2, 1});
  // Test that dims with size=1 can have any stride.
  TestHandleFromDLPack(status, ctx, {1}, {1});
  TestHandleFromDLPack(status, ctx, {1}, {0});
  TestHandleFromDLPack(status, ctx, {4, 1, 2}, {2, 1, 1});

  return true;
}

// Copies `src_graph` into `dst_graph`. Any node in `src_graph` with input
// `src_inputs[i]` will have that input replaced with `dst_inputs[i]`.  `prefix`
// will be prepended to copied node names. `control_deps` are nodes in
// `dst_graph` that the copied `src_graph` nodes will have control dependencies
// on. `return_nodes` are nodes in `src_graph`, and the new corresponding nodes

  // TODO(b/113217601): This is needed for EagerContext::runner_ to use a
  // threadpool, so that we avoid the possibility of running the runner_ in the
  // threadpool of GPU event mgr, as that can trigger more callbacks to be
  // scheduled on that same threadpool, causing a deadlock in cases where the
  // caller of event_mgr->ThenExecute() blocks on the completion of the callback

    for (int i = 0; i < grad_outputs.size(); i++) {
      auto grad_input = grad_outputs[i];
      // TODO(srbs): Should we add a copy contructor to AbstractTensorHandle
      // that takes care of this similar to `Tensor`?
      if (grad_input) {
        grad_input->Ref();
      }
      grad_inputs[i] = grad_input;
    }
    return absl::OkStatus();
  }

                       const char* dst, TF_Status* status) {
  // If target is a directory return TF_FAILED_PRECONDITION.
  // Target might be missing, so don't error in that case.
  struct stat st;
  if (stat(dst, &st) != 0) {
    if (errno != ENOENT) {
      TF_SetStatusFromIOError(status, errno, dst);
      return;
    }
  } else if (S_ISDIR(st.st_mode)) {

}

Status ImmediateExecutionTensorHandle::SummarizeValue(
    std::string& summary) const {
  Status status;
  AbstractTensorPtr resolved(
      // TODO(allenl): Resolve should be const, and the caches that get updated
      // marked mutable.
      const_cast<ImmediateExecutionTensorHandle*>(this)->Resolve(&status));
  if (!status.ok()) {
    return status;
  }
  summary = resolved->SummarizeValue();

  std::unordered_map<int64_t, TapeTensor> sources_that_are_targets;
  for (int i = 0; i < target_tensor_ids.size(); ++i) {
    int64_t target_id = target_tensor_ids[i];
    if (sources_set.find(target_id) != sources_set.end()) {
      auto tensor = targets[i];
      sources_that_are_targets.insert(
          std::make_pair(target_id, TapeTensor(tensor)));
    }
  }

  bool valid = true;
  int64_t expected_stride = 1;
  for (int i = ndim - 1; i >= 0; --i) {
    // Empty tensors are always compact regardless of strides.
    if (shape_arr[i] == 0) return true;
    // Note that dimensions with size=1 can have any stride.
    if (shape_arr[i] != 1 && stride_arr[i] != expected_stride) {
      valid = false;
    }
    expected_stride *= shape_arr[i];
  }
  return valid;
}

==============================================================================*/
// This file extends/implements core graph optimizer base classes in terms of
// the C API defined in grappler.h. A class "CSomething" represents a
// "Something" that can be manipulated via calls in the C interface and a C
// struct called "TP_Something".

#include "tensorflow/c/experimental/grappler/grappler.h"

#include <algorithm>
#include <cstddef>
#include <cstring>

}

TEST(TensorPjRtBufferUtilTest, GetPjRtCBufferFromTensorNoBuffer) {
  auto allocator = std::make_unique<AsyncValueAllocator>();
  tensorflow::Tensor tensor(allocator.get(), DT_FLOAT, {1});

  EXPECT_THAT(
      GetPjRtCBufferFromTensor(&tensor),
      StatusIs(error::INTERNAL, HasSubstr(absl::StrCat(
                                    "Input tensor does not have PjRtBuffer"))));
}