Xmipp and Scipion installation problem on Debian 12 machine

[TomCrey]

Dear all,

We're trying to install Xmipp in order to use Scipion3, but we're having a few problems with the installation (cf terminal output below). We're trying to install it on a machine running Debian 12; do you know if it's possible to install Xmipp on this type of machine?

Thank you for your help,

Tom

---

Unable to install Xmipp. Some error found compiling 'basic_mem_manager.o' program.. You can skip this program by including 'CUDA=False' in the config file. For more information about the error check compileLOG.txt file if exist. You can visit our guide of installation https://github.com/I2PC/xmipp, visit the wiki page with some details https://github.com/I2PC/xmipp/wiki and you can contact us xmipp@cnb.csic.es or Discord/Scipion/xmipp To report the details of the error please share the reportInstallation.tar.gz file

[pconesa]

Let's see if anyone from xmipp team have a clue.

@oierlauzi?

Have you tried to follow what the message says:

"You can skip this program by including 'CUDA=False' in the config file"

If it is not clear, just do:

1.- scipion3 config --show (should pop up a text file with variables and values) 2.- add/edit a line : CUDA=False 3.- install xmipp again: scipion3 install xmippSrc

[oierlauzi]

Hello, Xmipp should work on Debian. I personally use it. As Pablo says, maybe you could try compiling without CUDA at least you are likely to get a working Scipion installation. In any case, please, could you send me your /path/to/scipion/software/em/xmippSrc-?????/compileLOG.txt

[TomCrey]

Hi @oierlauzi, Yes, of course, here it is:

Compilation Date: 2024-04-23 16:31:45.402295

########################### Compiling cuFFTAdvisor-------- mkdir -p build/ nvcc --cudart static -shared -link -Wno-deprecated-gpu-targets -o build/libcuFFTAdvisor.so obj/advisor.o obj/benchmarker.o obj/benchmarkResult.o obj/cudaAsserts.o obj/generalTransform.o obj/inputParser.o obj/main.o obj/sizeOptimizer.o obj/transform.o obj/transformGenerator.o obj/validator.o mkdir -p build/ nvcc -o build/cuFFTAdvisor obj/advisor.o obj/benchmarker.o obj/benchmarkResult.o obj/cudaAsserts.o obj/generalTransform.o obj/inputParser.o obj/main.o obj/sizeOptimizer.o obj/transform.o obj/transformGenerator.o obj/validator.o -lcufft -Wno-deprecated-gpu-targets

########################### Compiling googletest-------- -- Configuring done -- Generating done -- Build files have been written to: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/googletest/build [100%] Built target gtest [ 50%] Built target gtest [100%] Built target gtest_main

########################### Compiling libsvm-------- if [ "Linux" = "Darwin" ]; then \ SHARED_LIB_FLAG="-dynamiclib -Wl,-install_name,libsvm.so"; \ else \ SHARED_LIB_FLAG="-shared -Wl,-soname,libsvm.so"; \ fi; \ g++ ${SHARED_LIB_FLAG} svm.o -o libsvm.so

########################### Compiling libcifpp-------- Not trying to recreate symop_table_data.hpp since CCP4 is not defined -- Testing for known regex bug, since you're using GNU libstdc++ -- You are probably trying to compile using the g++ standard library which contains a crashing std::regex implementation. Will use boost::regex instead -- Configuring done -- Generating done -- Build files have been written to: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/build [ 4%] Building CXX object CMakeFiles/cifpp.dir/src/utilities.cpp.o [ 9%] Linking CXX shared library libcifpp.so [100%] Built target cifpp -- Install configuration: "Release" -- Installing: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/libcifpp.so.5.0.9 -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/libcifpp.so.5 -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/libcifpp.so -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/cmake/cifpp/cifppTargets.cmake -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/cmake/cifpp/cifppTargets-release.cmake -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/include/cif++/exports.hpp -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/share/libcifpp/mmcif_ddl.dic -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/share/libcifpp/mmcif_pdbx.dic -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/share/libcifpp/mmcif_ma.dic -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/cmake/cifpp/cifppConfig.cmake -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/cmake/cifpp/cifppConfigVersion.cmake -- Up-to-date: /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/lib/pkgconfig/libcifpp.pc

########################### Compiling XmippCore-------- scons: Reading SConscript files ... scons: done reading SConscript files. scons: Building targets ... scons: lib/libXmippCore.so' is up to date. scons:lib/xmippCore.so' is up to date. scons: done building targets.

########################### Compiling Xmipp------------ scons: Reading SConscript files ... scons: done reading SConscript files. scons: Building targets ... scons: lib/libXmipp.so' is up to date. scons:lib/_swig_frm.so' is up to date. /usr/local/cuda/bin/nvcc -o libraries/reconstruction_cuda/basic_mem_manager.o -c --x cu -D_FORCE_INLINES -Xcompiler -fPIC -ccbin -std=c++14 --expt-extended-lambda -gencode=arch=compute_60,code=compute_60 -gencode=arch=compute_61,code=compute_61 -gencode=arch=compute_75,code=compute_75 -gencode=arch=compute_86,code=compute_86 -I../ -I/home/tcrey/miniconda3/envs/scipion3/include -Iexternal -Ilibraries -I/storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/xmippCore -I/storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/include libraries/reconstruction_cuda/basic_mem_manager.cpp -std=c++14: No such file or directory Building compilation database /storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/compile_commands.json nvcc fatal : Failed to preprocess host compiler properties. /usr/local/cuda/bin/nvcc -o libraries/reconstruction_cuda/cuda_all.o -c --x cu -D_FORCE_INLINES -Xcompiler -fPIC -ccbin -std=c++14 --expt-extended-lambda -gencode=arch=compute_60,code=compute_60 -gencode=arch=compute_61,code=compute_61 -gencode=arch=compute_75,code=compute_75 -gencode=arch=compute_86,code=compute_86 -I../ -I/home/tcrey/miniconda3/envs/scipion3/include -Iexternal -Ilibraries -I/storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/xmippCore -I/storage/home/tcrey/scipion3/software/em/xmippSrc-v3.23.11.0/src/libcifpp/include libraries/reconstruction_cuda/cuda_all.cpp scons: [libraries/reconstruction_cuda/basic_mem_manager.o] Error 1 -std=c++14: No such file or directory nvcc fatal : Failed to preprocess host compiler properties. scons: [libraries/reconstruction_cuda/cuda_all.o] Error 1 scons: building terminated because of errors.

[oierlauzi]

Hello again. What nvidia driver do you have? and what cuda-toolkit? A quick look of the error (nvcc fatal : Failed to preprocess host compiler properties.) does not shed any light on the issue.

[azazellochg]

Does /usr/local/cuda/bin/nvcc exist?

[TomCrey]

Hello again. What nvidia driver do you have? and what cuda-toolkit? A quick look of the error (nvcc fatal : Failed to preprocess host compiler properties.) does not shed any light on the issue.

Hello @oierlauzi,

Nvidia driver version : 535.154.05 Cuda-toolkit : V11.8.89

Thanks for your help !

[oierlauzi]

Thanks! That looks okay. Please, also attach /path/to/scipion/software/em/xmippSrc-?????/xmipp.conf

[oierlauzi]

I think I have a clue. You may have g++-12 installed, but CUDA 11.8 only supports up to g++-11. 2. If the config file confirms this hypothesis, the solution will be rather simple.

[TomCrey]

We have solved our problem, it was a bad version of CUDA (11.8) which prevented the correct installation of Xmipp; with the CUDA 12.2 version everything seems to work normally for the moment.

Thank you all for your help!

Tom