Sync meeting on EESSI test suite (2025 07 31) - EESSI/meetings GitHub Wiki

Meeting (2025-07-31)

Attending: Caspar, Lara, Sam, Satish

• Merged PRs

  • None :(

• Open PRs

  • Mare Nostrum 5 config https://github.com/EESSI/test-suite/pull/277
  • OpenBLAS https://github.com/EESSI/test-suite/pull/268
    • Caspar: tested, works with local stack, fails with EESSI due to lack of RPATH-ing (see here)
  • numpy https://github.com/EESSI/test-suite/pull/276
  • TensorFlow https://github.com/EESSI/test-suite/pull/275 => You can use the resource dir from the test with self.current_system.resourcesdir. That, icm with adding the path variable to the tf.datasets call should hopefully help to progress this
  • OpenFOAM smaller test case (1M): https://github.com/EESSI/test-suite/pull/278
    • Not tested yet, but minimal changes compared to 8M test case. "should work" (famous last words :))

• Issues

  • Satish: there's an uncaught if-elif still in the API docs PR. https://github.com/EESSI/docs/blob/67457e1d08ccca72979cf7c03202f8c7f73a2991/docs/generate_eessi_testsuite_api_docs.py#L37
    • Satish will make a fix for this => Still open
      • [WIP] #513

Discussed - GROMACS & Metallwalls test download stuff on the fly - Offline testing of e.g. GROMACS is problematic, because it's a test from hpctestlib, and we can't easily change that to use the resourcesdir. Asked Vasileos on chat, but not sure if there's a response yet.

- Create a follow-up PR for creating a standard function that sets a minimum core count in the EESSI mixin class.
    - Use a standard phrase for the test skip.
    - Tests currently using this kind of functionality are: CP2K, OSU and OpenFOAM.
- Satish will explore Stream
- Improve docstrings so that autogenerated API docs actually become useful (maybe with some AI help?)