Experiment adding libhdf5

[willGraham01]

Opening this PR so we can have an ongoing discussion and a record of it somewhere.

NOT INTENDED FOR REVIEW OR MERGING ANYTIME SOON

Tagging #181

[willGraham01]

Linux build fails, and this is replicable on my local machine.

Can't make much from the error message - looks like (for some reason) HDF5 is expecting a different version to what's installed? H5check_version() has some documentation - it's meant to be a consistency check for things that are already compiled.

[samcunliffe]

Linux build fails, and this is replicable on my local machine.

Can't make much from the error message - looks like (for some reason) HDF5 is expecting a different version to what's installed? H5check_version() has some documentation - it's meant to be a consistency check for things that are already compiled.

That's indeed strange. We had it working in a little sandbox toy experiment, right? On your machine too? So it seems our CMakeLists.txt or FindHDF5.cmake is to blame.

[giordano]

What happened is that CMake correctly found system hdf5 during the configuration stage:

-- Found HDF5: /usr/lib/x86_64-linux-gnu/hdf5/serial/libhdf5.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/usr/lib/x86_64-linux-gnu/libsz.so;/usr/lib/x86_64-linux-gnu/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so (found version "1.10.4") 

but then when compiling the program the linker picks up matlab's own hdf5:

/usr/bin/c++ -Werror -O3 -DNDEBUG CMakeFiles/tdms.dir/src/openandorder.cpp.o CMakeFiles/tdms.dir/src/fields/base.cpp.o CMakeFiles/tdms.dir/src/fields/electric.cpp.o CMakeFiles/tdms.dir/src/fields/magnetic.cpp.o CMakeFiles/tdms.dir/src/fields/split.cpp.o CMakeFiles/tdms.dir/src/fields/td_field_exporter_2d.cpp.o CMakeFiles/tdms.dir/src/argument_parser.cpp.o CMakeFiles/tdms.dir/src/array_init.cpp.o CMakeFiles/tdms.dir/src/arrays.cpp.o CMakeFiles/tdms.dir/src/dimensions.cpp.o CMakeFiles/tdms.dir/src/fdtd_grid_initialiser.cpp.o CMakeFiles/tdms.dir/src/grid_labels.cpp.o CMakeFiles/tdms.dir/src/interface.cpp.o CMakeFiles/tdms.dir/src/interpolate.cpp.o CMakeFiles/tdms.dir/src/iterator.cpp.o CMakeFiles/tdms.dir/src/interpolation_methods.cpp.o CMakeFiles/tdms.dir/src/matrix_collection.cpp.o CMakeFiles/tdms.dir/src/mesh_base.cpp.o CMakeFiles/tdms.dir/src/numerical_derivative.cpp.o CMakeFiles/tdms.dir/src/shapes.cpp.o CMakeFiles/tdms.dir/src/simulation_parameters.cpp.o CMakeFiles/tdms.dir/src/source.cpp.o CMakeFiles/tdms.dir/src/timer.cpp.o CMakeFiles/tdms.dir/src/utils.cpp.o CMakeFiles/tdms.dir/matlabio/matlabio.cpp.o -o tdms   -L/usr/local/MATLAB/R2022b/bin/glnxa64  -Wl,-rpath,/home/runner/work/TDMS/build:/usr/local/MATLAB/R2022b/bin/glnxa64: libtdms_lib.so /usr/lib/x86_64-linux-gnu/libfftw3.so -lmex -lmx -lmat /usr/lib/gcc/x86_64-linux-gnu/9/libgomp.so /usr/lib/x86_64-linux-gnu/libpthread.so _deps/spdlog-build/libspdlog.a -lpthread 
/usr/bin/ld: CMakeFiles/tdms.dir/src/openandorder.cpp.o: undefined reference to symbol 'H5check_version@@MWHDF51.8.12'
/usr/bin/ld: /usr/local/MATLAB/R2022b/bin/glnxa64/libhdf5-1.8.so.8: error adding symbols: DSO missing from command line

There are two problems:

• since you're using the C++ API, you need to require the C++ component:

  find_package(HDF5 REQUIRED COMPONENTS CXX)

• the target tdms isn't told to explicitly link to the HDF5 library found during configuration, so the linker just picks up the first one it finds. Adding

  target_link_libraries(tdms ${HDF5_CXX_LIBRARIES})

  should do the trick

With these two changes I'm able to build tdms correctly on this branch.

[samcunliffe]

• the target tdms isn't told to explicitly link to the HDF5 library found during configuration, so the linker just picks up the first one it finds. Adding

    target_link_libraries(tdms ${HDF5_CXX_LIBRARIES})
  

  should do the trick

Mosè's fix is also needed for the dynlib target (at least on my ARM64 Mac which tries to link the wrong HDF5 libraries).

[samcunliffe]

I am starting to actually develop (for #181) over on 181-hdf5-io.
And, in parallel, get HDF5 working in Windows on #188.

Let's keep this PR open for now.

[samcunliffe]

Also closing this (c.f. #193) since it can just as happily live in the list of closed issues.
I'll need to revisit #193 (or something similar) given #215.