TSC Meeting Notes 2021-03-11

Owned by Cary Phillips
Nov 04, 2022
2 min read

Attending:

  • Cary Phillips

  • Dan Rolinek

  • Joseph Goldstone

  • Kimball Thurston

  • Larry Gritz

  • Nick Porcino

  • Peter Hillman

Discussion:

  • Symbol visibility:

    • Foundry guy confirms what Chris from Pixar says.

    • Kimball: couple of things: the dynamic cast solution I came up with really bothers me, but I did it so that the macOS version works. Fundamental issue is with attributes.

    • Typeinfo that gets cooked into the the library is not the same as where you use it. GCC is fine because it uses the type name, not code. I believe we can correctly export those templated types. If that works.

    • Kimball: I believe I can export the type infos. Same solution that the foundry guy was proposing, the export macro. Can get rid of the dynamic_cast stuff. Will work on it this weekend.

    • Should apply to Imath as well. Haven’t gone header-only yet. But there are no virtuals in Imath.

  • Weta and Epic should have now signed CLA 2.1 ARRI is waiting for OCIO to change, then they’ll sign both.

  • Joseph: Standard optional attributes: the definitions in there are less ambiguous than the code today, but the definitions might be invalidated if I get my way with the SYMPE. Camera position is not locked won int terms of whether is focul plan or entry point. Ambiguity is in the comments, not the code, but that’s all there is. Can we add the code now without the comments. Reserves the keyword, so people don’t start using them.

    • Larry: could make a wiki page. People would see it regardless of what.

    • Joseph: I’ll identity the ones that will go through with no fuss. I’ll strip down the PR, then someone else can submit.

  • Nick: CMake install prefix doesn’t get transmitted to Imath.

    • OTIO is replacing OTIO’s math with Imath. Imath config was being installed into the build directory not the install directory.

    • We’re doing a fairly new thing, CMake external dependency features, it could be that external dependencies have subtleties that we need to account for.

  • Fuzz issue is a timeout. Takes the fuzz test more than a minute to decompress.

  • ThreadPool destruction. Larry: It’s been screwing up OIIO, not 100% sure what the fix is. Happens every nth time. It’s on shutdown, it’s not doing anything tricky. Log doesn’t tell you anything, other than never terminating.