spack: concretizer does not take a unique provider
Steps to reproduce
spack:
specs:
- sirius ^intel-mkl
"packages:":
all:
compiler: [gcc, intel, pgi, clang, xl, nag, fj, aocc]
providers:
blas: [openblas, amdblis]
fftw-api: [fftw, amdfftw]
lapack: [openblas, amdlibflame]
mkl: [intel-mkl]
mpi: [openmpi, mpich]
scalapack: [netlib-scalapack, amdscalapack]
spack -e . concretize shows openblas@0.3.18, intel-mkl@2020.4.304, fftw@3.3.9
meaning, repeated providers for blas, lapack (openblas / intel-mkl) and fftw-api (fftw, intel-mkl).
I can’t find a Spack version where this does not happen right now.
About this issue
- Original URL
- State: closed
- Created 5 years ago
- Comments: 17 (17 by maintainers)
Commits related to this issue
- ASP-based solver: a package eligible to provide a virtual must provide it fixes #26866 This semantics fits with the way Spack currently treats providers of virtual dependencies. It needs to be revis... — committed to alalazo/spack by alalazo 5 years ago
- ASP-based solver: a package eligible to provide a virtual must provide it fixes #26866 This semantics fits with the way Spack currently treats providers of virtual dependencies. It needs to be revis... — committed to alalazo/spack by alalazo 5 years ago
- ASP-based solver: a package eligible to provide a virtual must provide it fixes #26866 This semantics fits with the way Spack currently treats providers of virtual dependencies. It needs to be revis... — committed to spack/spack by alalazo 5 years ago
Thanks for the quick fix!
@alalazo if it helps, try and remove that depends_on statement for intel-mkl in sirius, and note what happens: intel-mkl becomes a provider of fftw-api as dependency of spfft, but sirius gets openblas and fftw for its virtuals… Does that help with debugging?
The condition for
intel-mklto be a possible provider are met and in fact: