Conversation
|
Are you sure you want to make these changes? I believe the bug in the concretizer that necessitated these changes is still present... |
| # http://www.unidata.ucar.edu/software/netcdf/docs/getting_and_building_netcdf.html | ||
| depends_on('[email protected]:+hl~mpi', when='~mpi') | ||
| depends_on('[email protected]:+hl+mpi', when='+mpi') | ||
| depends_on('[email protected]:+hl') |
There was a problem hiding this comment.
We should mention a non-default value for the variant of hdf5 separately. Otherwise, we will need to mention it in all other constraints that we have.
| conflicts('~mpi', when='@:4.3^hdf5+mpi', | ||
| msg='netcdf-c@:4.3~mpi requires hdf5~mpi') |
There was a problem hiding this comment.
We do not really need this conflict anymore because we put a more strict constraint below as depends_on.
| # However, it is very unlikely that someone needs such combination. | ||
| # Therefore, we put the constraint for all versions to be able to avoid the | ||
| # dependency on MPI by simply running: spack install netcdf-c~mpi. | ||
| depends_on('hdf5~mpi', when='~mpi') |
There was a problem hiding this comment.
As I mentioned before, we need to split the constraints into several depends_on statements. So, the result is the same as before when='~mpi'.
| conflicts('+mpi', when='^hdf5~mpi', | ||
| msg='netcdf-c+mpi requires hdf5+mpi') | ||
| depends_on('hdf5+mpi', when='+mpi') |
There was a problem hiding this comment.
This is what is left from the splitting of the constraints for the case when='+mpi'. Again, we don't' need the conflict anymore because it is basically impossible to have netcdf-c+mpi ^hdf5~mpi due to the depends_on statements.
|
@adamjstewart I am pretty sure because it's the same as before but without redundant conflicts and the problem with
But then I thought that the second problem might be not so serious given the usability improvement that we get, so I just fixed the first one in this PR. |
|
This also makes #11893 and similar workarounds redundant. |
2 participants
This PR fixes
spack spec [email protected]. Plus I have updated my old comments.