Update config tutorial section to 1.0#468
Conversation
becker33
left a comment
becker33
left a comment
There was a problem hiding this comment.
In the first bit where we cover the sections of the available config, we should remove the compilers section since it's deprecated. Around line 27 in the rst file.
It would be great if we could find time/space for an include example, but I don't know what the changes have done to the timing so I don't know if that's possible. It might be more valuable than some of the stuff in the "high level config" segment but that doesn't mean we have time to make that happen for the HPCIC tutorial.
| all: | ||
| compiler: [clang, gcc, intel, pgi, xl, nag, fj] | ||
| require: | ||
| - one_of: ["%llvm", "%gcc"] |
There was a problem hiding this comment.
Why are we adding this? Wouldn't this make it impossible to use our new toolchain (because it requests some of each)?
I think it would be truer to the original intent of the example to set
all:
prefer:
- %clang_gfortran
There was a problem hiding this comment.
This doesn't seem to work for me when I test it out (doesn't seem to encourage use of the toolchain, or hybridizing the clang/gcc compilers)
I did run through the whole tutorial with one_of, and I agree I'm not clear why it ends up allowing mixing (a bug maybe), but anyhow
- Replacing it with any_of worked for me
- And I think that ought to fit with what we intend
Have you used a preference for a toolchain before?
(trying prefer: [%clang_gfortran] also appears to significantly increase concretization times)
There was a problem hiding this comment.
I retried this with prefer: [%llvm] and it went better, so that's what the tutorial looks like now
2 participants
just the compilers part and intro are updatedeverything up to the "Installation Permissions" section is updated (I don't anticipate any major edits in that section or any after)::) useconfig:section as examplepackages:sectiontargetin the compiler entry anymoreI useUpdate August 4: I'm now usingrequire:prefer:and later on using require forhdf5~mpi, which is smoother overall)hdf5+mpi ^mpich@4.0weren't quite the same, so I truncated that part and emphasized the error message (vs. intermediate step where you concretize and don't seehdf5+mpi)spack spec(pending updates to tutorial image)