Skip to content

Conference call notes 20230104

Jump to bottom
Kenneth Hoste edited this page Jan 18, 2023 · 7 revisions

(back to Conference calls)

Notes on the 213th EasyBuild conference call, Wednesday 4 January 2023 (09:00 UTC)

Attendees

List of attendees (12):

  • Sebastian Achilles (JSC, Germany)
  • Simon Branford (Univ. of Birmingham, UK)
  • Alain van Hoof (TU Eindhoven, Netherlands)
  • Kenneth Hoste (HPC-UGent)
  • Alexander Grund (TU Dresden, Germany)
  • Kurt Lust (UAntwerpen & LUST)
  • Sebastien Moretti (SIB, Switzerland)
  • Mikael Öhman (Chalmers University of Technology, Sweden)
  • Jurij Pečar (EMBL, Germany)
  • Jörg Saßmannshausen (Imperial College London, UK)
  • Åke Sandgren (Umeå University, Sweden)
  • Lars Viklund (Umeå University, Sweden)

Agenda

  • overview of recent developments + outlook to next EasyBuild release
  • update on porting of EasyBuild docs to MarkDown
  • Q&A

Recent developments

  • release timeline
    • latest release: EasyBuild v4.6.2 (21 Oct 2022)
    • ETA next release (EasyBuild v4.7.0): ASAP
  • recent changes
    • framework
      • bug fixes
        • use -march=x86-64 -mtune=generic instead of -xSSE2 when using Intel oneAPI compilers (PR #4147)
        • partially skip test_det_copy_ec_specs if no GitHub token is available (PR #4149)
      • enhancements
        • add easybuild.tools.LooseVersion (since distutils is deprecated in Python 3.10) (PR #3794)
        • add support for --filter-rpath-sanity-libs to skip RPATH sanity check for designated libraries (PR #4119)
        • add support for postinstallmsgs (PR #4145)
        • make iimkl toolchain aware of intel-compilers (PR #4146)
      • changes
        • trim CI test configurations to avoid hitting GitHub rate limits (PR #4148)
      • docs
        • ...
    • easyblocks
      • bug fixes
        • remove -openmp build option for MRtrix v3.x (PR #2822)
      • enhancements
        • set $TEMPDIRPATH for testsuite in the BerkeleyGW easyblock, to avoid polluting /tmp (PR #2836)
      • updates
        • ...
      • changes
        • tweak docstring in some generic easyblocks so it renders nicely in auto-generated documentation (PR #2849)
      • new software
        • ...
    • easyconfigs
  • work in progress
    • framework
      • reported bugs / bug fixes
        • Python packages are being installed to local/lib/python*/dist-packages due to tweaks to sysconfig.py in OS (issue #4128)
        • improve error when checksum dict has no entry for a file (PR #4150)
      • enhancements
        • add support for using --output-format=md (MarkDown) (PR #4117)
          • BLOCKER for EasyBuild v4.7.0
        • allow to directly import EasyConfig constants from the module (PR #4144)
        • add support for --summarize-prs (WIP) (PR #4151)
        • add support for --list-software --output-format=json (PR #4152)
          • first step towards including EasyBuild in repology
      • changes
        • ...
    • easyblocks
      • bug reports/fixes
        • ...
      • enhancements
        • allow use of test_cmd without runtest for ConfigureMake (PR #2837)
        • enhance CMakeMake easyblock to run ctest command if runtest is True (PR #2838)
      • updates
        • update scipy easyblock for scipy >= 1.9 to correctly configure BLAS/LAPACK library to use via Meson (WIP PR #2848)
      • new software
        • ...
      • changes
        • ...
    • easyconfigs

Update on porting of EasyBuild docs to MarkDown

Q&A

  • simint (PR #16886)
    • problematic tests: long-running, several seemingly fail with precision errors
    • but test command always has zero exit code...
  • NVHPC toolchain should be ready to go (PR #16724)
    • still need to add nvofbf toolchain definition to framework;
  • recent versions of Intel MPI and Slurm and PMIx
    • Intel MPI's hydra is reporting that it's ignoring PMI stuff when$I_MPI_PMI_LIBRARY is set, when using srun;
    • seems like unsetting $I_MPI_PMI_LIBRARY works fine, although it disagrees with the documentation (cfr. https://slurm.schedmd.com/mpi_guide.html#intel_mpi);
  • should we consider switching to iomkl as common toolchain, instead of intel
    • Sebastian: at JSC, recent software is now installed on top of Parastation MPI, and users can switch to Intel MPI (without rebuilding);
Clone this wiki locally