TSC Meeting Notes 2020-06-18

Owned by Cary Phillips
Nov 04, 2022
2 min read

Attending:

  • Cary Phillips

  • Christina Tempelaar-Lietz

  • Eskil Steenburg

  • Joseph Goldstone

  • Kimball Thurston

  • Larry Gritz

  • Nick Porcino

  • Owen Thompson

  • Peter Hillman

  • Alison Rowan

Discussion:

  • Review OpenEXR logo concepts with Alison Rowan from the LF's Creative Services group.

    • General consensus on preferring the design that illustrates 3 exposure levels.

    • Based loosly Nick's idea sketch.

    • How well will it read when it’s small?

    • Could EXR be in red?

    • Discusion of perspective vs orthographic. Does the perspective look a little too precious? Consensus prefers the orthographic.

  • Discussion of exception handling.

    • Kimball will submit his proposal to a branch in the Imath repo, for further review. He had another idea that avoids any change to the API.

  • Discussion of performance metrics

    • Peter: can we somehow link a test with two different variations of the library and compare metrics between them? Using namespaces?

    • Need further experiments.

  • Should PyImath be a separate repo, or a submodule of Imath?

    • Owen configured PyImath as a submodule of Imath, but Cary suggested we actually want the opposite: Imath makes no mention of PyImath.

    • Cary: the goal is simplicity and reliability of the build process, especially for the default build behavior. Someone wanting the basic c++ library, and possibly not having much technical experience, should not get tripped up because of python or boost configuration on their machines.

    • Kimball: I've always had trouble getting git submodule's to work, they've tended to get out of sync.

    • Larry: OpenVDB is eager for Imath but doesn't want PyImath, so maybe there is merit in Imath w/o python bindings.

    • Peter: The current cmake configuration tries to build the bindings, but if it can’t, it doesn't, so there shouldn't be a problem.

    • Cary: Don't underestimate the amount of work to reimplement the imath FixedArray bindings, it's more complex that it seems. Nick mentioned issue #367, regarding python buffer protocols.

    • Conclusion: move PyImath to a subfolder of Imath. The same repo, but not built by default. The cmake configuation is completely indepedent, and disconnected. "If you want the python bindings, cd to PyImath and run cmake." This satisfies both goals: simple and reliable, but a single repo.

  • Discussion of issue #765, leaks on static initialization. Resolved in 2.5, but the problem was errant c++ version #if's. It's time to remove all that stuff. Just remove all conditionals related to old c++ versions, and put them back in only where they're absolutely necessary. We may need to make a distinction between c++14 and 17 for certain constexpr constructions.

  • Update on the Disney CLA logjam, which came to a head with OSL, due to Pixar's involvement. The governing board's legal group met to discuss the various iterations of the proposed CLA's and may be close to a decision but nothing is final yet.