I re-did the spack environment, and finally it works. Changes include:
the custom compilation of OpenMPI and UCX is not specified explicitly as a dependency of fftw and hdf5, because for some reason spack tends to crash that way. It is picked up automatically however, so there is no problem.
All compilation is done inside of spack environments. This increases encapsulation a bit and also makes the whole setup a bit faster due to better parallelism. (roughly speaking, spack env depfile && make is the same as spack install but can build unrelated packages simultaneously).
next step for me: put it into the CI in order to be notified swiftly if anything breaks again.
I re-did the spack environment, and finally it works. Changes include:
* the custom compilation of `OpenMPI` and `UCX` is not specified explicitly as a dependency of `fftw` and `hdf5`, because for some reason spack tends to crash that way. It is picked up automatically however, so there is no problem.
* All compilation is done inside of spack environments. This increases encapsulation a bit and also makes the whole setup a bit faster due to better parallelism. (roughly speaking, `spack env depfile && make` is the same as `spack install` but can build unrelated packages simultaneously).
next step for me: put it into the CI in order to be notified swiftly if anything breaks again.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
I re-did the spack environment, and finally it works. Changes include:
OpenMPIandUCXis not specified explicitly as a dependency offftwandhdf5, because for some reason spack tends to crash that way. It is picked up automatically however, so there is no problem.spack env depfile && makeis the same asspack installbut can build unrelated packages simultaneously).next step for me: put it into the CI in order to be notified swiftly if anything breaks again.