Generate error (instead of segfault) when trying to copy string tensor

to GPU in EagerTensor constructor.

PiperOrigin-RevId: 168457320

tensorflow/c/eager/c_api.cc

@@ -222,6 +222,15 @@ TFE_TensorHandle* TFE_TensorHandleCopyToDevice(TFE_TensorHandle* h,
     return nullptr;
   }
   tensorflow::Tensor* src = &(h->t);
+  if (!dst_cpu && !tensorflow::DataTypeCanUseMemcpy(src->dtype())) {
+    TF_SetStatus(
+        status, TF_INVALID_ARGUMENT,
+        tensorflow::strings::StrCat("Can't copy Tensor with type ",
+                                    tensorflow::DataTypeString(src->dtype()),
+                                    " to device ", DeviceName(dstd), ".")
+            .c_str());
+    return nullptr;
+  }
   if (src_cpu) {
     tensorflow::Tensor dst(
         dstd->GetAllocator(tensorflow::AllocatorAttributes()), src->dtype(),

tensorflow/python/eager/BUILD

@@ -92,13 +92,15 @@ py_library(
     srcs_version = "PY2AND3",
 )
 
-py_test(
+cuda_py_test(
     name = "tensor_test",
     srcs = ["tensor_test.py"],
-    srcs_version = "PY2AND3",
+    additional_deps = [
-    deps = [
+        ":context",
         ":tensor",
         ":test",
+        "//tensorflow/python:errors",
+        "//tensorflow/python:framework_ops",
         "//third_party/py/numpy",
     ],
 )

tensorflow/python/eager/tensor_test.py

@@ -20,9 +20,12 @@ from __future__ import print_function
 
 import numpy as np
 
+from tensorflow.python.eager import context
 from tensorflow.python.eager import tensor
 from tensorflow.python.eager import test
 from tensorflow.python.framework import dtypes
+from tensorflow.python.framework import errors
+from tensorflow.python.framework import ops
 from tensorflow.python.framework import test_util
 
 
@@ -136,6 +139,15 @@ class TFETensorTest(test_util.TensorFlowTestCase):
     t_np = t.numpy()
     self.assertTrue(np.all(t_np == t_np_orig), "%s vs %s" % (t_np, t_np_orig))
 
+  def testStringTensorOnGPU(self):
+    if not context.context().num_gpus():
+      self.skipTest("No GPUs found")
+    with ops.device("/device:GPU:0"):
+      with self.assertRaisesRegexp(
+          errors.InvalidArgumentError,
+          "Can't copy Tensor with type string to device"):
+        tensor.Tensor("test string")
+
 
 if __name__ == "__main__":
   test.main()