January 9th, 2023

Host: Carol Payne

Secretary: Carol Payne

Attendees:

  • Rémi Achard (TSC) - DNEG
  • Mark Boorer (TSC) - Industrial Light & Magic

  • Mei Chu (TSC) - Sony Pictures Imageworks

  • Sean Cooper (TSC ACES TAC Rep) - ARRI

  • Michael Dolan (TSC) - Epic Games

  • Patrick Hodoul (TSC) - Autodesk

  • John Mertic - Academy Software Foundation / Linux Foundation

  • Carol Payne (TSC Chair) - Netflix

  • Mark Titchener (TSC) - Foundry

  • Carl Rand (TSC) - Weta Digital

  • Doug Walker (TSC Chief Architect) - Autodesk

  • Kevin Wheatley (TSC) - Framestore

  • Zach Lewis - Method Studios

Apologies:

  • Michael Dolan

OCIO TSC Meeting Notes

  • Dev Updates:
    • Released updates for 2.2.1, 2.1.3, 2.0.5
    • ABI compatible patch releases and updates, cmake improvements & documentation updates
  • Third party security policy
    • OpenEXR is ahead of OCIO right now
    • OCIO runs static analysis which flags a lot of things, Patrick made a first stab at some updates but needs more work
    • Second bit is the third party apps that we use/depend on
      • zlib for example - we set a recommended version for it in our cmake script, which will download and install if the user's version is older
      • There is both a minimum version and "recommended" version - what should be default? How should we handle this keeping security issues in mind?
        • Remi likes the recommended version being the default, but studios should be able to more easily override it if they want
        • Doug - yes, thinking about updates to the cmake scripts to allow this functionality
        • Kevin - agree, studios need the flexibility to control their own versioning. 
        • Zach - agree
        • Mark - yes, seems consistent as that's what we already do with imath, etc
        • Doug, yes except for if it's under our current recommendation we can't currently do that
  • CI Build Matrix Review 
    • Proposal:
      • Drop Python 2.7, add 3.10 and/or 3.11
      • Drop 2019 VFX Platform, add 2023 (when available)
      • Add C++ 20
      • Add more Linux compilers
      • Prune some MacOS and Windows builds?
    • Want to create an issue with the proposed changes, and see what feedback we get, even without the 2023 docker containers available yet, other changes still worth doing
    • Remi: We already don't build wheels for 2.7, never have. 
    • Kevin: seems reasonable to drop things now, most apps are already using python 3
    • Main reason to drop 2019 in favor of 2023 is to limit number of jobs running (also 2019 was last platform that supported python 2.7)
    • C++ 20 is forward thinking - C++ 11 is still there. It's more to make sure we're compatible and that things aren't being deprecated, for future version development not necessarily for current version testing
    • Next major release we could think about migrating to C++ 14 or 17, to take advantage of newer features which might break C++ 11 backwards compatibility. 
    • Remi: we don't need to test every C++ version on every OS etc,
    • Kevin: but we need to pay attention to compilers. 
    • Remi - maybe we can do just minimum and maximum for gcc/clang
    • Carol - will set up a time to meet with Michael to brain dump around analysis workflows
    • Doug - what's the earliest versions we should support? 
    • Kevin - ref platform 2020 is still gcc 6.3 so we should keep that. 
    • Kevin: does the various versions of python impact the windows/mac builds? Maybe we could just test latest python and that would remove some windows/mac builds
  • TSC Members addition: Zach and Thomas
    • No objections, all +1's from present TSC members
  • Remi - a lut small enough to be represented by a 1d texture shows up as black, once it's big enough for the 2d texture it works. Not sure if it's a unreal or windows issue
    • Doug - could be a GLSL vs HLSL issue
    • Unit test definitely test 1D luts small enough, definitely working on openGL
    • Remi will chat with Michael to see if he can replicate it
    • Doug: would be great to have HLSL test suite in addition to GLSL
    • Ask Michael how the GPU tests are currently hosted/run
  • No labels