auto &body_block = while_op.getBody().front();
    auto &cond_block = while_op.getCond().front();
    bool changed = false;
    for (int op_idx : llvm::seq<int>(0, while_op.getNumOperands())) {
      auto body_arg = body_block.getArgument(op_idx);
      auto cond_arg = cond_block.getArgument(op_idx);
      auto while_operand = while_op.getOperand(op_idx);

  auto* cc_ctx = reinterpret_cast<::tensorflow::OpKernelContext*>(ctx);
  int start = -1, stop = -1;
  auto status = cc_ctx->op_kernel().InputRange(name, &start, &stop);
  args->start = start;
  args->stop = stop;
  tensorflow::Set_TF_Status_from_Status(args->status, status);
}

TF_DataType TF_InputDatatype(TF_OpKernelContext* ctx, int index) {

                 string* init_op_name) {
  const auto& sig_def_map = meta_graph_def.signature_def();
  const auto& init_op_sig_it =
      meta_graph_def.signature_def().find(kSavedModelInitOpSignatureKey);
  if (init_op_sig_it != sig_def_map.end()) {
    const auto& sig_def_outputs = init_op_sig_it->second.outputs();
    const auto& sig_def_outputs_it =
        sig_def_outputs.find(kSavedModelInitOpSignatureKey);

  groups_to_only = 0;
  for (const auto& [group, branch] : from) {
    auto to_iter = to.find(group);
    if (to_iter == to.end()) {
      ++groups_from_only;
    } else {
      auto to_branch = to_iter->second;
      if (to_branch == branch) {
        ++groups_same_branch;
      } else {
        ++groups_different_branch;
      }
    }
  }
  for (const auto& [group, _] : to) {

    return failure();
  }

  auto session_init_func = GetOrCreateSessionInitFunc(module);
  OpBuilder builder(session_init_func.getContext());

  for (auto var_and_tensor : llvm::zip(var_ops, resource_tensors_or.value())) {
    auto& var_op = std::get<0>(var_and_tensor);
    auto& resource_tensor = std::get<1>(var_and_tensor);
    if (resource_tensor.dtype() != tensorflow::DT_RESOURCE) {

  Operation* result_user = *result.getUsers().begin();
  auto assign_var = dyn_cast<TF::AssignVariableOp>(result_user);
  if (!assign_var) return resource;

  auto handle = assign_var.getResource();
  // Skip result if cluster writes to the same variable via multiple results.
  for (Operation* handle_user : handle.getUsers()) {
    if (handle_user == assign_var) continue;
    auto assign_var_user = dyn_cast<TF::AssignVariableOp>(handle_user);

ResourceHandle GetResourceHandle(Operation *op) {
  llvm::StringRef device;
  if (auto attr = op->getAttrOfType<StringAttr>("device")) {
    device = attr.getValue();
  }

  llvm::StringRef container;
  if (auto attr = op->getAttrOfType<StringAttr>("container")) {
    container = attr.getValue();
  }

  llvm::StringRef shared_name;
  if (auto attr = op->getAttrOfType<StringAttr>("shared_name")) {
    shared_name = attr.getValue();

    auto guard = OpBuilder::InsertionGuard(*rewriter);
    auto inputs = branch_func.getFunctionType().getInputs();
    Block *block = rewriter->createBlock(
        &branch_func.getBody(), branch_func.begin(), inputs,
        SmallVector<Location>(inputs.size(), branch_func.getLoc()));

    auto input_shape = block->getArgument(1);
    auto size_diff = block->getArgument(2);

                .run()

        then: "the JDK is auto-provisioned again and its files, even though they are already there don't trigger an error, they just get overwritten"
        markerFile.exists()
    }

    def "issue warning on using auto-provisioned toolchain with no configured repositories"() {

  mlir::MLIRContext context(registry);
  auto module = mlir::OwningOpRef<mlir::ModuleOp>(
      mlir::parseSourceString<mlir::ModuleOp>(kMLIR, &context));
  auto module_op = module.get();
  auto serialized_result_fb = ExportRuntimeMetadata(module_op);
  const auto* result = GetRuntimeMetadata(serialized_result_fb.value().c_str());
  const auto* expected = GetRuntimeMetadata(kExpectedFB.c_str());