Error: LibXC libraries not found in ./lib or ./lib64

Okay, that partly worked... MRCC is now looking in the correct directory for the libraries, but for some reason it's not finding them. In the ./build/ directory there are bin, include and lib directories, all created when I installed libxc. ls from the lib directory gives:

libxcf03.la libxcf03.so.15 libxc.la libxc.so.15 pkgconfig
libxcf03.so libxcf03.so.15.1.2 libxc.so libxc.so.15.1.2

However, for some reason, I'm still getting the error that the LibXC directories aren't found in /home/granem94/mrcc/mrcc-26.1.3/build/lib.

Should I try reinstalling LibXC? I'm not sure what difference it would make, but if you think it would help...?

You have dynamic build. Maybe you should comment .a files detection and uncomment .la.

I used that command yesterday to compile LibXC 7.0.0 statically using ifort/icx (and it works, hehe):

cmake -Bbuild -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=icx -DCMAKE_Fortran_COMPILER=ifort -DENABLE_FORTRAN=ON -DCMAKE_INSTALL_PREFIX=$(pwd)/install

Then copied installed library into proper place.
And it produced only .a files and that's why I removed .la detection of libxc in build.mrcc.common.

Probably that check is imperfect.

Swapping the comments in the config file fixed it!

It's compiling with a LOT of warnings (type and rank mismatches), though, and more errors than I think is probably healthy... I even saw a fatal error at one point. Expected the build to stop, but it's still running...

Now you can relax... With GCC, MRCC compilation will take like 12-20 hours. I do not remember exact time.

If there are errors, let's wait until first run of ./build.mrcc. Maybe some of them in integ.f and other files about libxc modules since it might not able to find them properly. For the other files I do not know.

Do you use GCC 15/16? It might raise some problems... With GCC 14.1, I was able to compile MRCC 25.1.3. I did not compile GCC 16.X for testing MRCC 26.X.Y yet.

Okay, I just checked, and the version of gfortran in this easybuild module is 13.3.0...

Could this be part of why there are so many errors/warnings? Might it cause the installation to fail?

One other thing before I embark on this epic compilation: Part of the reason I want to do this is to interface MRCC with CFOUR. On our HPC, CFOUR was compiled with GNU compilers and so cannot be loaded at the same time as the intel compilers. I'm hoping that running MRCC's CCSD(T) calculations through CFOUR will both allow MRCC to use the analytic gradients while also speeding up the calculations themselves (CFOUR doesn't parallelise well...). Is this basically what happens, or does running MRCC through CFOUR (or vice versa) not provide any speed advantages? If it doesn't, I might just install it with the intel compilers again and save a lot of time...

No, old versions (I've tested 10.5, 11.2, 13.4) can compile MRCC well...

Can you show example of errors? I've just recompiled without optimizations without LibXC and see only warnings (+linking error from -lblas -llapack). I suppose you see them from Libxc interfaces.

For better perf, you have to link MRCC with some performance-oriented BLAS/LAPACK, not with default -lblas -llapack.