当我使用bazel bulid libtensorflow_inference时。所以

Question

系统：Ubuntu 16.04 Bazel：0.4.5 TensorFlow：cpu版本r1.0 gcc版本：（Ubuntu 5.4.0-6Ubuntu1~16.04.4)5.4.0 201 60609

bazel build-c opt//tensorflow/contrib/android:libtensorflow_inference.so\

tensorflow/core/kernels/BUILD:3749:1: C++ compilation of rule '//tensorflow/core/kernels:android_tensorflow_kernels' failed: Process exited with status 1 [sandboxed].
tensorflow/core/kernels/split_v_op.cc:212:28: error: non-constant-expression cannot be narrowed from type 'value_type' (aka 'long long') to 'int' in initializer list [-Wc++11-narrowing]
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
tensorflow/core/kernels/split_v_op.cc:172:12: note: in instantiation of member function 'tensorflow::SplitVOpCPU<int, long long>::Compute' requested here
  explicit SplitVOpCPU(OpKernelConstruction* c) : Base(c) {}
           ^
tensorflow/core/kernels/split_v_op.cc:355:19: note: in instantiation of member function 'tensorflow::SplitVOpCPU<int, long long>::SplitVOpCPU' requested here
TF_CALL_ALL_TYPES(REGISTER_SPLIT_LEN);
                  ^
tensorflow/core/kernels/split_v_op.cc:212:28: note: insert an explicit cast to silence this issue
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
                           static_cast<int>( )
tensorflow/core/kernels/split_v_op.cc:212:28: error: non-constant-expression cannot be narrowed from type 'value_type' (aka 'long long') to 'int' in initializer list [-Wc++11-narrowing]
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
tensorflow/core/kernels/split_v_op.cc:172:12: note: in instantiation of member function 'tensorflow::SplitVOpCPU<float, long long>::Compute' requested here
  explicit SplitVOpCPU(OpKernelConstruction* c) : Base(c) {}
           ^
tensorflow/core/kernels/split_v_op.cc:355:19: note: in instantiation of member function 'tensorflow::SplitVOpCPU<float, long long>::SplitVOpCPU' requested here
TF_CALL_ALL_TYPES(REGISTER_SPLIT_LEN);
                  ^
tensorflow/core/kernels/split_v_op.cc:212:28: note: insert an explicit cast to silence this issue
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
                           static_cast<int>( )
tensorflow/core/kernels/split_v_op.cc:212:28: error: non-constant-expression cannot be narrowed from type 'value_type' (aka 'long long') to 'int' in initializer list [-Wc++11-narrowing]
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
tensorflow/core/kernels/split_v_op.cc:172:12: note: in instantiation of member function 'tensorflow::SplitVOpCPU<tensorflow::bfloat16, long long>::Compute' requested here
  explicit SplitVOpCPU(OpKernelConstruction* c) : Base(c) {}
           ^
tensorflow/core/kernels/split_v_op.cc:356:1: note: in instantiation of member function 'tensorflow::SplitVOpCPU<tensorflow::bfloat16, long long>::SplitVOpCPU' requested here
REGISTER_SPLIT_LEN(bfloat16);
^
tensorflow/core/kernels/split_v_op.cc:353:3: note: expanded from macro 'REGISTER_SPLIT_LEN'
  REGISTER_SPLIT(type, int64);
  ^
tensorflow/core/kernels/split_v_op.cc:349:27: note: expanded from macro 'REGISTER_SPLIT'
                          SplitVOpCPU<type, len_type>);
                          ^
tensorflow/core/kernels/split_v_op.cc:212:28: note: insert an explicit cast to silence this issue
          prefix_dim_size, split_sizes_vec[i], suffix_dim_size};
                           ^~~~~~~~~~~~~~~~~~
                           static_cast<int>( )
3 errors generated.
Use --strategy=CppCompile=standalone to disable sandboxing for the failing actions.
Target //tensorflow/contrib/android:libtensorflow_inference.so failed to build
Use --verbose_failures to see the command lines of failed build steps.

Answer 1

这些错误来自Android NDK R13中从gcc到Clang的切换。为了获得最好的效果，当用Bazel构建时，暂时使用NDK r12b。