Telcon: 2015 06 11

Participants

• Todd Gamblin, LLNL
• Jim Galarowicz, Krell
• Mike Collette, LLNL

New Features & Contributions

Lots of updates in develop:

1. Integrated linear algebra contributions from:

• Saravan Pantham: PR #21

• Kevin Brandstatter: PR #20, and

• Benjamin Walters: PR #22

• New linear algebra packages:

  • arpack
  • atlas
  • boxlib
  • cblas
  • cgm
  • hypre
  • lapack
  • netlib-blas
  • parpack
  • petsc

2. New packages from Tom Scogland (PRs #55, #56, #58)

• libuuid
• py-cffi
• py-pycparser
• task
• taskd

3. Better Vim package from Tom Scogland

• See PR #60
• Apparently vim is way fancier than I thought.

4. SPACK-38: Virtual dependencies should be retrievable from spec by v-dep name

   • The following now work differently:

     ```
     spec['mpi']
     spec['blas']
     ```
     

   • This can return a spec for openmpi, mpich, mvapich, etc., EVEN if the spec is already concretized. This means that in a package that depends_on('mpi'), you can do spec['mpi'] to see what it was concretized to. This should simplify MPI and BLAS packages.

     ```
     'mpi' in spec
     'blas' in spec
     ```
     

   • Previously, if the spec had been concretized, these would be False because there was not a dependency in the DAG with either of these names. These will now be True even if the spec has been concretized. So, e.g., this will print "YES"

     ```
     s = Spec('callpath ^mpich')
     if 'mpi' in spec:
         print "YES"
     ```
     

   • Similarly, this will be True:

     ```
     Spec('mpich').satisfies('mpi')
     ```
     

   • Because of the way virtual dependencies are currently implemented, the above required some fiddling around with package.py

     • figuring out what is a vdep requires loading all packages and seeing what doesn't exist.
     • Now can't ever call Spec.__contains__ (and result in endless recursion).
     • For the time being, package.py shouldn't call __contains__

   • This should be fixed by allowing virutal dependnecies to have their own package class.

     • This would allow a quicker check for vdeps, without a call to all_packages.
     • Would also allow us to put methods/variables/etc. on vdeps

5. Added FileFilter class to make repeated calls to filter_file easier.

• example: arpack
• Thinking about adding a generic Makefile editor class
  • Have methods to set/unset variables in a Makefile (most common operation)

Bug Fixes

• Fix for zsh support for blobbing from Tom Scogland:

  • PR #57

• Fixed issue with Dyninst 8.2.1 build

  • boost CMake build parameters were not getting passed correctly
  • Either CMake or dyninst isn't recognizing BOOST_ROOT/BOOSTROOT properly when supplied as a -D option.
    • maybe this needs to be an env var to work properly?
    • consider adding to boost package's setup_dependent_environment if the env vars work.

Issue

• Greg Lee brought up some path length issues with Spack paths in scripts using #!
  • Install paths got a bit longer due to full hashes.
  • What's the best thing to do about this?
    1. Wrappers?
    2. Different install layout?

General questions/discussion

• Mike reprimanded by LC for too many tmp directories

  • investigate whether Spack is removing tmp directories properly

• Mike having /tmp issues

  • Other LLNL user using Spack at LANL
  • Cielo does not separate /tmp directories by user
  • Check whether /tmp path includes username and add if it doesn't.

• Mike notes there are lots of spacks in pathnames

  • recommends mixing it up with spackitty

• Mike building 20 versions every night

  • dev code, prod code, previous production
  • comes back with a weird error about NoneType not callable
    • result of calling which and not finding the thing you're looking for.
    • may not be finding deps due to filesystem sync issues

• would be useful to have variants and dependencies enabled/disabled by default based on whether they're already installed

  • consider Open|SpeedShop case.