Attendance:

Discussion:

• Outstanding problem with deep reads: it has been fixed.

• We should be in a good place for v3.1.3, there aren’t any outstanding issues.

•  

  • Cary: I’ve been cherry picking commits into a branch and submitting as a PR to merge into RB-3.1, because that gives an easy-to-view list of the merged commits. 

  • Larry: In my projects I merge to the branch right after merging to master, but either is fine.

• Why does OpenEXR do chroma subsampling?

•  

  • Rod: Florian had an obsession about 422 and chroma subsampling, he put it in early. He wasn’t happy with how chroma subsampling was done in the early years at ILM.

  • Larry: It’s different from any other image format, except ones that are specifically video.

  • EXRs are linear, but video isn’t.

  • Kimball: It’s an arbitrary number: subsampling set to 17.

  • Rod: B44 compression was also from Florian.

  • Kimball: That’s ok, it’s what ProRes does.

  • Rod: Whether there is subsampling is orthogonal to whether it’s linear or not.

• Joseph: MacPorts

•  

  • Feels like trying to understand the language of a vanished civilization

  • All the C++ stuff builds and installs

  • The python modules don’t install. There’s a problem with the venvs

  • Nick: That’s why there’s been movement towards conda, away from venvs

  • Joseph: I’m working with an old codebase that’s built around macports, but a new codebase that uses conan. I’ll keep struggling with it.

• Cary: I now have an account on a Fedora project ILM machine, can investigate the s390x test failure.

•  

  • Kimball: We ported Manuka to the Power series. There’s something weird, like char is not the same signed-ness as the default. Unsigned by default?

  • Cary: The suggestion is that it’s an endian issue, since the x390x is big-endian.

• Christina: A windows build issue with static linking of zlib