Cannot build Debug build with WITH_GPU=ON due to file truncation

[Sand3r-]

System information
-PaddlePaddle version 29d9fb5
-CPU: including MKLDNN version 0.17
-GPU: including CUDA 8.0/CUDNN 7 version
-OS Platform Ubuntu 16.04
-Python version 2.7.12

• Install method: build from source(without docker)

To Reproduce
Run cmake:

cmake -DCMAKE_BUILD_TYPE=Debug -DWITH_GPU=ON -DWITH_MKLDNN=ON -DWITH_TESTING=ON -DWITH_PROFILER=ON -DWITH_STYLE_CHECK=OFF -DWITH_INFERENCE=ON -DWITH_INFERENCE_API_TEST=ON .. 

and build with make -j
Describe your current behavior
When GPU build is ran with Debug build type, a linking error occurs:

/usr/bin/ar: /home/mgallus/src/Sander/Paddle/build_gpu_deb/paddle/fluid/inference/libpaddle_fluid.a: File truncated
paddle/fluid/inference/CMakeFiles/paddle_fluid.dir/build.make:4981: recipe for target 'paddle/fluid/inference/libpaddle_fluid.a' failed
make[2]: *** [paddle/fluid/inference/libpaddle_fluid.a] Error 1
make[2]: *** Deleting file 'paddle/fluid/inference/libpaddle_fluid.a'
CMakeFiles/Makefile2:58681: recipe for target 'paddle/fluid/inference/CMakeFiles/paddle_fluid.dir/all' failed
make[1]: *** [paddle/fluid/inference/CMakeFiles/paddle_fluid.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
/usr/bin/ar: /home/mgallus/src/Sander/Paddle/build_gpu_deb/paddle/fluid/inference/libpaddle_fluid_origin.a: File truncated
paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/build.make:4909: recipe for target 'paddle/fluid/inference/libpaddle_fluid_origin.a' failed
make[2]: *** [paddle/fluid/inference/libpaddle_fluid_origin.a] Error 1
make[2]: *** Deleting file 'paddle/fluid/inference/libpaddle_fluid_origin.a'
CMakeFiles/Makefile2:59135: recipe for target 'paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/all' failed
make[1]: *** [paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/all] Error 2

[luotao1]

There need some times to fix this compile bug.
And if you only want to fix the test_concat_mkldnn_op (which fails the TC build), you can remove following lines, which will not compile inference code:

Paddle/paddle/fluid/CMakeLists.txt

Line 11 in c83d5b7

add_subdirectory(inference)

[Sand3r-]

@luotao1 Thank you for the tip, I'll definitely try that 👍

[luotao1]

solved by #15109

[kbinias]

System information:

• PaddlePaddle version c49e604
• CPU: SKX 8180, MKLDNN version 0.18
• GPU: including CUDA 8.0, cuDNN v7
• OS Platform Ubuntu 16.04
• Python version 2.7.12

Reproduce:

• cmake .. -DWITH_GPU=ON -DWITH_PROFILER=ON -DWITH_STYLE_CHECK=OFF -DWITH_MKLDNN=ON -DWITH_TESTING=ON
• make -j

Result

[ 74%] Linking CXX static library libpaddle_fluid_origin.a
/usr/bin/ar: /home/kbinias/dev/paddle/build/paddle/fluid/inference/libpaddle_fluid_origin.a: File truncated
paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/build.make:5701: recipe for target 'paddle/fluid/inference/libpaddle_fluid_origin.a' failed
make[2]: *** [paddle/fluid/inference/libpaddle_fluid_origin.a] Error 1
make[2]: *** Deleting file 'paddle/fluid/inference/libpaddle_fluid_origin.a'
CMakeFiles/Makefile2:64213: recipe for target 'paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/all' failed
make[1]: *** [paddle/fluid/inference/CMakeFiles/paddle_fluid_origin.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....

[luotao1]

@kbinias Please use the latest image paddlepaddle/paddle:latest-dev, or you can solve it by #15109.

[lidanqing-intel]

Building with the same command still reports the same error: file truncation.
Removing the inference and training doesn't work.
PaddlePaddle commit: 81b4fad8b94d85a473875ae0efe634fe46697314

Related to #15615, we can not reproduce the same results as yours. Thanks

[luotao1]

@lidanqing-intel please see #14775 (comment)

[lidanqing-intel]

If this would be searched again, consider update binutils to 2.28.0 or higher version.
To check: ar --version