Versions Compared

Old Version 1

changes.mady.by.user Cary Phillips

Saved on

compared with

New Version 2

changes.mady.by.user Cary Phillips

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Attendance:

  •  Cary Phillips
  •  Christina Tempelaar-Leitz
  •  John Mertic
  •  Joseph Goldstone
  •  Kimball Thurston
  •  Larry Gritz
  •  Nick Porcino
  •  Peter Hillman
  •  Rod Bogart
  •  Scott Wilson
  •  Anders Langlands

Agenda:

  • 3.1 release
    • Monday?
  • documentation
    • comments style
    • Theory of Deep
    • Standard Optional Attributes
  • Open Source Days
  • soversion
  • fuzz testing new core
  • Rust

Key Dates for Review:

  • 3.1 release early next week.

Discussion:

  • 3.1 release. Any outstanding issues?
    • Kimball: only if there are other architectures we don’t know about, I’ve only tested on x86.
    • Larry: Seems fine, OIIO is working, tests don’t fail.
    • Larry: does the new core support all compression types? Kimball: DWAA, DWAB not there. Technically, the reading works, it’s the decoding that hasn’t been implemented. Can copy files, but can’t get to pixels. Needs a serious refactor. Not clear yet if that can be a patch into 3.1 or wait for 3.2.
    • Haven’t done all the packing side of deep.
    • Kimball: test harness does a cross-matrix, where the C++ writes files to test reading by the C and vice versa.
    • Kimball: 3.1 is a “tech preview,” lets people start to hook in.
    • Kimball: Later, we will have to discuss how to take on the underlayment of the C++ API.  Might deserve a 4.0 release.
  • Fuzz testing:
    • Just need to hook the new core into the OSS fuzz configuration.
    • Peter: We have all the existing test cases, confirm it runs with those first.
  • Documentation:
    • Peter will look into converting “Theory of Deep Pixels” from tex to rst.
  • Open Source Days:
    • Cary will start a slide deck.
    • Kimball will discuss the new core.
    • Alban volunteered to repeat his presentation about spectral data, slides included. 
    • Kimball: Hoping I can convince Jean-Marie, but the time is like 6am New Zealand time.
  • Rust:
    • Current proposal: hand rust bindings over to openexr project. Find a balance between making sure you have the support you need, while getting access to resources.
    • What license? OpenEXR’s license is BSD3. Rust is Apache2. But you know all the authors, you can track them down, relicense under BSD3.
    • How big is it? 31K lines.
    • Versioning and releases? Rev the rust bindings along with releases?
    • Rust: Have to figure out things like lifetime of objects, etc. 
    • Anders: Personal preference: bake both the openexr version and rust crate version into one number. Will downstream rust users be confused by that? 
    • Simplest thing: we just stick on 0.X, then we’re free to change whatever necessary. When we’re confident, just push to lock step version.
    • Peter: patches could be different. Patch number is C++ but rust has its own version. 
    • Crates at https://crates.io. Some crates we own, some were donated by the rust community. Scott: I’m not on the crate as an owner.
    • Anders: A rust crate is a rust library.
    • Should nominate one of us for the TSC. Both make sense.
    • No rust bindings for Imath. There are also already a number of CG linear algebra crates. We didn’t feel like adding another contender. Depend on “Imath-traits”.