Tensorflow uses protobuf v3.6.1.2 tarball instead of protobuf provided by
Tumbleweed, which leads to a build failure on armNN as Caffe is using protobuf
from Tumbleweed (v3.9.1) which confilcts with protobuf v3.6.1.2 from
Tensorflow.

[  368s] [ 52%] Building CXX object
CMakeFiles/armnnCaffeParser.dir/src/armnnCaffeParser/RecordByRecordCaffeParser.cpp.o
[  368s] /usr/bin/c++  -DARMCOMPUTENEON_ENABLED -DARMNN_CAFFE_PARSER
-DARMNN_SERIALIZER
-DARMNN_SERIALIZER_SCHEMA_PATH=\"/home/abuild/rpmbuild/BUILD/armnn-19.05/src/armnnSerializer/ArmnnSchema.fbs\"
-DARMNN_TF_LITE_PARSER
-DARMNN_TF_LITE_SCHEMA_PATH=\"/usr/include/tensorflow/lite/schema/schema.fbs\"
-DARMNN_TF_PARSER -DBOOST_ALL_DYN_LINK -DPROTOBUF_USE_DLLS
-DarmnnCaffeParser_EXPORTS -I/home/abuild/rpmbuild/BUILD/armnn-19.05/include
-I/home/abuild/rpmbuild/BUILD/armnn-19.05/src/armnnUtils -isystem
/usr/lib/python3.7/site-packages/tensorflow/include -isystem /usr/include/half
-isystem /home/abuild/rpmbuild/BUILD/armnn-19.05/third-party  -O2 -Wall
-D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables
-fasynchronous-unwind-tables -fstack-clash-protection -g -pthread -std=c++14
-Wall -Werror -Wold-style-cast -Wno-missing-braces -Wconversion  -O2 -g  -fPIC 
 -o
CMakeFiles/armnnCaffeParser.dir/src/armnnCaffeParser/RecordByRecordCaffeParser.cpp.o
-c
/home/abuild/rpmbuild/BUILD/armnn-19.05/src/armnnCaffeParser/RecordByRecordCaffeParser.cpp
[  368s] In file included from /usr/include/caffe/proto/caffe.pb.h:37,
[  368s]                  from
/home/abuild/rpmbuild/BUILD/armnn-19.05/src/armnnCaffeParser/RecordByRecordCaffeParser.hpp:12,
[  368s]                  from
/home/abuild/rpmbuild/BUILD/armnn-19.05/src/armnnCaffeParser/RecordByRecordCaffeParser.cpp:6:
[  368s] /usr/include/google/protobuf/port_def.inc:74:2: error: #error
PROTOBUF_DEPRECATED was previously defined
[  368s]    74 | #error PROTOBUF_DEPRECATED was previously defined
[  368s]       |  ^~~~~

As a workaround, Tensorflow has been disabled in armNN until this bug is fixed.