Attendees:

  • Alban Fichet
  • Cary Phillips
  • Christina Tempelaar-Lietz
  • Eskil Steenberg
  • Joseph Goldstone
  • Kimball Thurston
  • Larry Gritz
  • Peter Hillman
  • Nick Porcino
  • Stela Espindola

Discussion items:

  • This week's release problems:
    • A change accidentally broke the install settings so OpenEXR didn't
      install its headers. The CI passed because we don't test the
      installation.
    • We can add a step that validates the install. Just another step in
      the build process. The CMakeLists.txt has to be separate from the
      main openexr build, just pointing to the previously built install
      directory.
    • Larry: Our policy should be to tag the release and let it
      percolate for atleast 48 hours before making the release.
    • Larry: maybe it would be better to integrate patches into the
      Cary: Then I overlooked cherry-picking that patch into the v3.0.2
      branch. I staged it a few days before through the cherry-pick PR,
      then mistakenly assumed it was complete. The v3.0.2 release went
      out without the patch.
    • Larry: maybe it would be better to integrate patches into the
      release branch at the same time they're merged into master, rather
      than right before the release.
  • Half+core changes:
    • Kimball: Almost ready to go. Need to clean up some warnings. I've
      redone the API quite a bit and clean that up quite a bit after
      some feedback from the rust guys in terms of consistency about
      const-ness
    • Get it in soon, in time for a code freeze
      mid-June in preparation for beta release early July.
    • How do we hook it into the oss-fuzz system? Peter: there's a
      directory of tests that currently has a single entry. Just add
      another one.
  • Other changes for 3.1:
    • Cary: I'd like to submit ILM additions to PyImath. Buffer protocol
      and copy-on-write support for FixedArrays. I think it's better to
      have the public repo in sync with ILM's internal version, even if
      there's little use for the features.
  • Rust bindings:
    • Did we decide, do these go into our project or are rust bindings
      consolidated elsewhere?
    • cppmm - "c++ minus-minus" turns C++ into C
      bindings. https://github.com/vfx-rs/cppmm.
    • Haven't gotten a good read from them about whether they want to
      keep all of the rest bindings for all the projects in one repo
      together or in separate repos. Is their eventual goal to push
      those into the individual projects?
  • Alban Fichet: spectral image data
    • Joseph: the paper is awesome, read it.
    • Joseph: the memory requirements are terrifying.
    • Kimball: It's not so bad, we have exr files that have 50 or 60
      light groups in them right now in production and so having having
      10 spectral channels isn't bad.
    • Joseph: but it covers things like how does fluorescence work when
      you have energy at one point in the Spectrum where does it emit?
      It's basically a 2D graph.
  • CII badge
    • A question came up with at the TAC, how much of a lift is it to go
      from passing to silver and then to gold.
    • Have to have a working build system. Ha.
    • By the way: we don't support "uninstall". Because it's not a
      standard CMake feature.
    • A silver/gold requirement: dependency tracking to identify
      vulnerabilities.
    • Not that big an issue for us, OpenEXR only depends on Zlib.
    • Bigger deal for OSL.
    • Dependabot can help: https://github.com/dependabot
    • Let's just start checking off the CII items and see how it
      goes. If it gets onerous, we can reconsider.