// error reporting system.
  mlir::StatusScopedDiagnosticHandler error_handler(mlir_context);

  // Parse the module.
  *mlir_module = mlir::parseSourceString<mlir::ModuleOp>(serialized_mlir_module,
                                                         mlir_context);
  if (!*mlir_module)
    return error_handler.Combine(
        errors::InvalidArgument("could not parse MLIR module"));

  mlir::StatusScopedDiagnosticHandler error_handler(module.getContext());
  mlir::LogicalResult result = mlir::TF::InferShapeForFunction(
      main_func, arg_shapes_copy, producer_version);

  if (failed(result)) {
    return error_handler.Combine(
        errors::Internal("MLIR Shape refinement failed"));
  }
  return error_handler.ConsumeStatus();
}

    return tensorflow::errors::Aborted(
        "Failed to apply MLIR pass manager CL options.");
  }

  auto error_handler = [&](const Twine& msg) {
    emitError(UnknownLoc::get(pm.getContext())) << msg;
    return failure();
  };
  if (failed(pass_pipeline.addToPipeline(pm, error_handler))) {
    return tensorflow::errors::Aborted("Failed to add passes to pipeline.");
  }

}
}  // namespace

LogicalResult ReshapeOp::verify() {
  ReshapeOp op = *this;
  auto error_handler = [&op](const llvm::Twine &message) -> LogicalResult {
    return op.emitOpError() << message;
  };
  TensorType expected_ty;
  if (failed(GetReshapeOutputType(op.getTensor(), op.getShape(), error_handler,
                                  expected_ty)))
    return failure();

  return success();
}

mlir::LogicalResult ReshapeOp::verify() {
  ReshapeOp op = *this;
  auto error_handler = [&op](const llvm::Twine& message) -> LogicalResult {
    return op.emitOpError() << message;
  };
  TensorType expected_ty;
  if (failed(GetReshapeOutputType(op.getInput(), op.getShape(), error_handler,
                                  expected_ty)))
    return failure();

}

template <typename T>
static bool HasValidTFLiteType(Value value, T& error_handler) {
  // None type is allowed to represent unspecified operands.
  if (mlir::isa<NoneType>(value.getType())) return true;

  auto type = mlir::dyn_cast<TensorType>(value.getType());
  if (!type) {
    if (auto op = value.getDefiningOp()) {
      error_handler.emitError()
          << '\'' << op << "' should produce value of tensor type instead of "

		panic("test")
	}()
	if result != "test" {
		t.Errorf("did not receive custom handler")
	}
}

func TestCustomHandleError(t *testing.T) {
	old := ErrorHandlers
	defer func() { ErrorHandlers = old }()
	var result error
	ErrorHandlers = []ErrorHandler{
		func(_ context.Context, err error, msg string, keysAndValues ...interface{}) {
			result = err
		},
	}
	err := fmt.Errorf("test")
	HandleError(err)

	}
}

// ErrorHandlers is a list of functions which will be invoked when a nonreturnable
// error occurs.
// TODO(lavalamp): for testability, this and the below HandleError function
// should be packaged up into a testable and reusable object.
var ErrorHandlers = []ErrorHandler{
	logError,
	func(_ context.Context, _ error, _ string, _ ...interface{}) {

	return w.watcher.Add(path)
}

func (w *fsnotifyWatcher) Init(eventHandler FSEventHandler, errorHandler FSErrorHandler) error {
	var err error
	w.watcher, err = fsnotify.NewWatcher()
	if err != nil {
		return err
	}

	w.eventHandler = eventHandler
	w.errorHandler = errorHandler
	return nil
}

func (w *fsnotifyWatcher) Run() {
	go func() {
		defer w.watcher.Close()
		for {
			select {

	attempts := 5
	for i := 0; i < attempts; i++ {
		gotStatus := getSyncStatus(t, s)
		var errorHandler func(string, ...any)
		if i == attempts-1 {
			errorHandler = t.Errorf
		} else {
			errorHandler = t.Logf
		}
		for _, ss := range gotStatus {
			if ss.ProxyID == nodeID {
				if ss.ProxyVersion == "" {
					errorHandler("ProxyVersion should always be set for %v", nodeID)
				}