undefined reference to `tflite::DefaultErrorReporter()'

[ashishmagar600]

System information:
Have I written custom code (as opposed to using a stock example script provided in TensorFlow): Yes
OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 16.04
TensorFlow installed from (source or binary): source
TensorFlow version (use command below): v1.4.0-19-ga52c8d9 1.4.1
Python version: 2.7.12
Bazel version (if compiling from source): 0.8.1
GCC/Compiler version (if compiling from source): g++ 5.4.0
CUDA/cuDNN version: none
GPU model and memory: none
Exact command to reproduce:

g++ -I../tensorflow -Llib -ltensorflow_cc -Llib -ltensorflowlite test.cpp -o exec

Describe the problem
I used this tutorial to build tensorflow link. Also this link

Additionally, I built tensorflow-lite with:

bazel build //tensorflow/lite:libtensorflowlite.so

I am using this code for testing tflite:

#include < cstdio >
#include "tensorflow/lite/interpreter.h"
#include "tensorflow/lite/kernels/register.h"
#include "tensorflow/lite/model.h"
#include "tensorflow/lite/tools/gen_op_registration.h"

using namespace tflite;

int main(int argc, char* argv[]) {
if (argc != 2) {
fprintf(stderr, "minimal \n");
return 1;
}
const char* filename = argv[1];

std::unique_ptrtflite::FlatBufferModel model =
tflite::FlatBufferModel::BuildFromFile(filename);
return 0;
}

Here is the error:

/tmp/ccPKK7am.o: In function main': test.cpp:(.text+0x57): undefined reference to tflite::DefaultErrorReporter()'
test.cpp:(.text+0x6d): undefined reference to tflite::FlatBufferModel::BuildFromFile(char const*, tflite::ErrorReporter*)' /tmp/ccPKK7am.o: In function std::default_deletetflite::FlatBufferModel::operator()(tflite::FlatBufferModel*) const':
test.cpp:(.text.ZNKSt14default_deleteIN6tflite15FlatBufferModelEEclEPS1[ZNKSt14default_deleteIN6tflite15FlatBufferModelEEclEPS1]+0x1e): undefined reference to `tflite::FlatBufferModel::~FlatBufferModel()'
collect2: error: ld returned 1 exit status

[terryheo]

Could you try to separate build and link stage?
For example)
$ g++ -I../tensorflow -ltensorflow_cc -c test.cpp
$ g++ -Llib -ltensorflow_cc -Llib -o exec test.o -ltensorflowlite
Please make sure to provide your object file first than "-ltensorflowlite "

[ashishmagar600]

I am getting this :
./exec: error while loading shared libraries: libtensorflowlite.so: cannot open shared object file: No such file or directory

My lib/
libtensorflow_cc.so libtensorflow_framework.so.2 libtensorflowlite.so-2.params lite
libtensorflow_cc.so.2 libtensorflow_framework.so.2.1.0 libtensorflowlite.so.runfiles
libtensorflow_cc.so.2.1.0 libtensorflow_framework.so.2.1.0-2.params libtensorflowlite.so.runfiles_manifest
libtensorflow_cc.so.2.1.0-2.params libtensorflowlite.so libtensorflow.so.2.1.0-2.params

[terryheo]

You need to update LD_LIBRARY_PATH to make linker can find libtensorflowlite.so

[ashishmagar600]

Thanks a lot...
So I have to:

1. Include my library after object file
2. Add path to LD_LIBRARY_PATH

I wasted so many days on this.
Thank you once again

[rajveershringi]

System information:
Have I written custom code (as opposed to using a stock example script provided in TensorFlow): Tiny modification to existing minimal.cc file
OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 18.04
TensorFlow installed from (source or binary): source
TensorFlow version (use command below): not sure how to get this via c++, I cloned the latest tensorflow repository from here https://github.com/tensorflow/tensorflow
Python version: 3.7.7
Bazel version (if compiling from source): not installed
GCC/Compiler version (if compiling from source): g++ 7.5.0
CUDA/cuDNN version: none
GPU model and memory: none

Exact command to reproduce:
g++ -o minimal minimal.o -Itensorflow/lite/tools/make/gen/linux_x86_64/lib/libtensorflow-lite.a

I have build tensorflowlite for x86 following the instructions here : link
But instead of running ./build_aarch64_lib.sh, I ran ./build_lib.sh for x86
I can see the generated .a files under : tensorflow/lite/tools/make/gen/linux_x86_64/lib/ which are benchmark-lib.a and libtensorflow-lite.a

I can see an executable already made for minimal.cc under tensorflow/lite/tools/make/gen/linux_x86_64/bin, which I am able to run with a .tflite model and get the expected outputs.

However I need to write a tweaked version of minimal.cc file for my model interpretation. How do I build this file ? Currently I see linker errors :
minimal.cc:(.text+0x6e): undefined reference to tflite::DefaultErrorReporter() minimal.cc:(.text+0x8a): undefined reference to tflite::FlatBufferModel::BuildFromFile(char const*, tflite::ErrorReporter*) minimal.cc:(.text+0xe5): undefined reference to tflite::ops::builtin::BuiltinOpResolver::BuiltinOpResolver() minimal.cc:(.text+0x110): undefined reference to tflite::impl::InterpreterBuilder::InterpreterBuilder(tflite::FlatBufferModel const&, tflite::OpResolver const&)

I have compiled the file via :
g++ -I. -Itensorflow/lite/tools/make/downloads/googletest/googletest/include/ -Itensorflow/lite/tools/make/downloads/absl/ -Itensorflow/lite/tools/make/downloads/flatbuffers/include/ -c tensorflow/lite/examples/minimal/minimal.cc

Things already done :
echo $LD_LIBRARY_PATH /home/rajveer/Downloads/tensorflow/tensorflow/lite/tools/make/gen/linux_x86_64/lib/

[rajveershringi]

Resolved above via compiling in the tensorflow/ root directory :
$ g++ tensorflow/lite/examples/minimal/minimal.cc -o minimal -std=c++11 -I. -Itensorflow/lite/tools/make/downloads/googletest/googletest/include/ -Itensorflow/lite/tools/make/downloads/absl/ -Itensorflow/lite/tools/make/downloads/flatbuffers/include/ -L/tf-lite/ -ltensorflow-lite -lrt -ldl -pthread

Where linker path /tf-lite is the source directory containing libtensorflow-lite.a

Hope it helps others.

However if I use the similar approach, I am not able to compile label_image.cc example under tensorflow/lite/examples/label_image/label_image.cc and get undefined references for :
label_image.cc:(.text+0x55): undefined reference to "tflite::evaluation::CreateGPUDelegate()"

[pranv12]

Hello,

@rajveershringi @ashishmagar600 @jdduke @terryheo

I'm trying to run a custom code written in C++ using TFLite.
I.m trying to compile and run the code. But it's giving some errors.

I also had undefined reference to `tflite::DefaultErrorReporter()' this error but managed to remove it.

Somehow I've managed to install TensorflowLite from the source.
I've put tensorflow/tensorflow directory in /usr/local/include. Other dependencies (absl,flatbuffers, etc) are also inside same directory. Using this command bazel build //tensorflow/lite:libtensorflowlite.so whcichever files (*.so, *.params, *.runfiles.manifest) are generated and put in /usr/local/lib/ directory.

As said in this comment I tried to execute two commands

g++ -ltensorflow -c main.cc
g++ -Llib -ltensorflow -o ./main.o -ltensorflowlite

And error was:

/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 1603 (>= sh_info of 1)
/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 2710 (>= sh_info of 1)
/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 2711 (>= sh_info of 1)
/usr/bin/ld: /usr/lib/gcc/x86_64-linux-gnu/9/../../../x86_64-linux-gnu/Scrt1.o: in function _start':
(.text+0x24): undefined reference to main'
collect2: error: ld returned 1 exit status

second time I did something like this:
g++ -ltensorflow -Llib main.cc -ltensorflowlite
And now error is :

/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 1603 (>= sh_info of 1)
/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 2710 (>= sh_info of 1)
/usr/bin/ld: /usr/local/lib/libtensorflow.so: .dynsym local symbol at index 2711 (>= sh_info of 1)
/usr/bin/ld: /tmp/ccR2jeyn.o: in function `main':
main.cc:(.text+0x314): undefined reference to `infer_float(tflite::impl::Interpreter*)'
/usr/bin/ld: main.cc:(.text+0x32a): undefined reference to `infer_uint8(tflite::impl::Interpreter*)'
collect2: error: ld returned 1 exit status

Please help to correct me if I've done something wrong with the installation process or coding error.

infer_float and infer_uint8 are two functions declared in a header file. Then two separate .cc files contain the definition of each function. main.cc file has included header file and rest of code.

Here's link to my repo which contains the code.