Telcon: 2016 09 01

Participants

1. Todd Gamblin (LLNL)
2. Ben Boeckel (Kitware)
3. Peter Scheibel (LLNL)
4. Thomas Merrick (TAMU)
5. Greg Lee (LLNL)
6. Greg Becker (LLNL)

News

1. SC16 Tutorial preparations still underway at LLNL

• Making slides
• Updating docs
• Tutorial scheduled for Sunday, November 13, 1:30-5pm. Topics:
  • Spack usage basics, core concepts
  • Intro to packaging
  • Deploying with Spack (Module configuration, etc.)

1. Lots of documentation updates

• Docs now hosted on spack.readthedocs.io
• Thanks to Adam stewart for bringing API docs into compliance
  • no more sphinx warnings!
  • docs are much more readable.
• Elizabeth Fischer working on re-org of documentation
  • separating tutorial material from reference manual
• Expected to stay more current this way

1. Fixed some irritating long-standing bugs

• Done:
  1. spack reindex bugs (#1265, #1320)
  2. Made spack reindex more robust (see 69b6815)
  3. Faster DB read performance: spack find now much faster (#1535)
• In progress:
  1. Finer-grained locking, based on earlier work by Massimiliano (#1562)
  2. Uninstall by hash
  3. Hash stability (#1409)
  4. lowercase architecture names, use of distro package to detect distribution (avoid things like linux-None-x86_64)
  • #1329, #1629
• last three above change spec, so be prepared.

1. Lots of other updates

• See Pulse

1. Resume binary caching conversation

• See Update: 2016-08-18 for details on security, how we could host this.
• CERN to talk about their findings (w.r.t. what they're comfortable with)

1. Update from Kitware:

• Ben working on packaging UVCDAT
• would like Qt to get merged for Mac OS X and for Linux soon
  • Qt 4 working for Linux and Mac
  • Qt 5 has issues with new version (5.5)
• Ben to look at using a Spack platform file to fix Qt build.

1. Peter Scheibel:

• PR #1310 is ready to merge
• Build support for Spack to run in Koji is almost ready -- need to discuss merge.
  • This is a precursor to RPM generation support.

Discussion

1. Greg Lee has had a few issues recently

• initializing spack on TOSS2 machines then initializing on TOSS3 doesn't find compilers properly
  • compiler detection doesn't notice that there are no compilers for the current architecture
    • should re-run detection when this happens.
• spack install hanging on BG/Q machines
  • doesn't get to actual compilation